Podcasts about Multimodal

#youtube #chatgpt #supergrok #podcastWelcome back to So Lead Saturday. I'm Chibi version of Vaishali Lambe, and today we are talking about Multimodal A I Explained Simply.The core idea for today is simple: A I is moving from text-only systems to models that understand images, video, audio, and actions.This topic matters because A I is no longer something that sits only inside research labs or innovation teams. It is now entering everyday workflows, business decisions, product roadmaps, leadership conversations, and career planning. But with that growth comes a lot of confusion. We hear big words, exciting demos, and bold predictions. Yet in the real world, success with A I usually comes down to something much more practical: clarity, workflow design, trust, and measurable value.Let's start with what people often misunderstand.When a new A I trend becomes popular, many people immediately focus on the tool. They ask: Which model should I use? Which platform is best? Which app is trending right now? Those questions are useful, but they are not the starting point. The better starting point is: what problem are we solving, who is affected, what decision needs to improve, and what workflow needs to become easier?That difference matters.A tool-first approach creates scattered experiments. A workflow-first approach creates business value.Until we meet, happy leading, and let's lead together. Stay safe. Bye for now.

In this episode, host Alyssa Watson, DVM, welcomes back Ellen M. Lindell, VMD, DACVB, to discuss her recent Clinician's Brief article, “Beyond Fluoxetine: A Multimodal Approach to Anxiety, Aggression, & Fear in Cats.” Dr. Lindell uses real-life cases to explore why things like house soiling and anxious behaviors occur in our cats. She shares advice on how to ask the right questions to plan environmental and behavioral modifications and when medication should get involved. Resources: https://www.cliniciansbrief.com/article/feline-anxiety-aggression-fluoxetine-quiz https://www.zoetisus.com/petcare/care-is-your-calling/ Contact: podcast@instinct.vet Where To Find Us: Website: CliniciansBrief.com/Podcasts YouTube: Youtube.com/@clinicians_brief Facebook: Facebook.com/CliniciansBrief LinkedIn: LinkedIn.com/showcase/CliniciansBrief/ Instagram: @Clinicians.Brief X: @CliniciansBrief The Team: Alyssa Watson, DVM - Host Alexis Ussery - Producer & Multimedia Specialist

We also talk about Nintendo's replaceable battery in Europe, some hope for energy storage, and an explanation of Apple's smart glasses strategy as well.Starring Tom Merritt and Jenn CutterShow notes found here. Hosted on Acast. See acast.com/privacy for more information.

Send me a derm question or story through text or voicemail!Skin barrier is having a moment... and for good reason. In this episode, I break down why restoring the skin barrier is a non-negotiable piece of the multimodal approach to managing atopic dermatitis in dogs and cats, and more importantly, how to actually do it when your clients can't keep up with a bathing schedule.Watch The Episode: https://www.youtube.com/@thedermvet3932Follow The Derm Vet Podcast: https://www.instagram.com/thedermvetpod/Follow Me: https://www.instagram.com/thedermvet/Timestamps00:00 Intro00:45 Itch Inquiry: Recurrent Yeast Otitis01:37 Antifungal Resistance in Malassezia03:54 Underlying Allergies and Immunotherapy06:29 Paronychia07:19 Toothpick vs. Tape Methods09:30 Treatment for Yeast15:35 Summary/Outro 

People can be interested without being ready to buy, they can agree to a proof of concept without having a clear path to production, they can praise the product without becoming the kind of customer who helps the company grow. That distinction was at the center of Collin Stewart's conversation with Ankur Patel, founder and CEO of Multimodal, on the Predictable Revenue Podcast. Multimodal builds AI for document-heavy, decision-heavy workflows in financial services, and Ankur's story is useful because it shows how easy it is to mistake activity for traction. Highlights include: Identifying the Niche (02:01), Customer Development and Validation (04:36), Pricing Strategy and First Customer (10:23), Evolving Market Strategies (17:46), Recognizing Product-Market Fit (22:50), and more... Stay updated with our podcast and the latest insights on Outbound Sales and Go-to-Market Strategies!

This is a recap of the top 10 posts on Hacker News on June 03, 2026. This podcast was generated by wondercraft.ai (00:30): Gemma 4 12B: A unified, encoder-free multimodal modelOriginal post: https://news.ycombinator.com/item?id=48385906&utm_source=wondercraft_ai(01:55): Meta workers can opt out of being tracked at work up to 30 minOriginal post: https://news.ycombinator.com/item?id=48383220&utm_source=wondercraft_ai(03:21): Pwnd Blaster: Hacking your PC using your speaker without ever touching itOriginal post: https://news.ycombinator.com/item?id=48382310&utm_source=wondercraft_ai(04:46): Elixir v1.20: Now a gradually typed languageOriginal post: https://news.ycombinator.com/item?id=48388324&utm_source=wondercraft_ai(06:12): I was recently diagnosed with anti-NMDA receptor encephalitisOriginal post: https://news.ycombinator.com/item?id=48384355&utm_source=wondercraft_ai(07:38): DaVinci Resolve 21Original post: https://news.ycombinator.com/item?id=48384482&utm_source=wondercraft_ai(09:03): Uber's $1,500/month AI limit is a useful signal for AI tool pricingOriginal post: https://news.ycombinator.com/item?id=48383056&utm_source=wondercraft_ai(10:29): 32GB of DDR5 now costs $375 – AI shortage continues to squeeze PC buildingOriginal post: https://news.ycombinator.com/item?id=48383241&utm_source=wondercraft_ai(11:54): U.S. to dismantle system tracking Atlantic currents that are at risk of collapseOriginal post: https://news.ycombinator.com/item?id=48392232&utm_source=wondercraft_ai(13:20): MacBook Neo is so popular that Apple doubled productionOriginal post: https://news.ycombinator.com/item?id=48386238&utm_source=wondercraft_aiThis is a third-party project, independent from HN and YC. Text and audio generated using AI, by wondercraft.ai. Create your own studio quality podcast with text as the only input in seconds at app.wondercraft.ai. Issues or feedback? We'd love to hear from you: team@wondercraft.ai

No matter your role, experience or industry, we all (mostly) waste hours a week doing the same thing: manually creating slides. 

Modern work can be frustrating and chaotic—if you don't have the right tools. From context engineering to multimodal search, go behind the scenes and hear how Dropbox engineers are building AI that actually understands you, so you can focus on the work that matters most. If you're new to Working Smarter, we've travelled from the F1 track to the bottom of a lake, and heard real stories from chefs, doctors, lawyers, and founders about how AI is helping them do more of what they love about their jobs. But in our third season, we're talking to the people behind the tools—the engineers and product leaders building helpful, time-saving AI features into the Dropbox experience you already know and trust. You'll hear all about their work on agents, inference, security, and, of course, how the people building AI use AI themselves. ~ ~ ~  Working Smarter is brought to you by Dropbox. Find, organize, and share your work—all in one place—with context-aware AI from Dropbox. You can listen to more episodes of Working Smarter on Apple Podcasts, Spotify, YouTube, Amazon Music, or wherever you get your podcasts. To read more stories and past interviews, visit workingsmarter.ai This show would not be possible without the talented team at Cosmic Standard: producer Ben Montoya, sound engineer Aja Simpson, technical director Jacob Winik, and executive producer Eliza Smith. Special thanks to our illustrator Fanny Luor, marketing consultant Meggan Ellingboe, and editorial support from Catie Keck.  Our theme song was composed by Doug Stuart.  Working Smarter is hosted by Matthew Braga. Thanks for listening!

June is here so guess what? It's officially Hot AI Summer.

Guests: Dr. Kelly Williams, Marion L Brittain Postdoctoral Fellow at Georgia Tech; Dr. Meryem Yilmaz Soylu, research scientist at the Center for 21st Century Universities (C21U); Alison Valk, the Emerging Technologies Librarian at the Georgia Tech Library.  First broadcast May 29 2026. Playlist "That's as you like it."

Google dropped like 197 new AI features this week.

AI suddenly feels like it has crossed a threshold, and Yann Dubois, co-lead of the Post-training Frontiers team at OpenAI, joins Matt Turck to explain why. Yann's team has led the post-training behind the company's reasoning models, including the recent GPT-5.5 release. In this conversation, we go inside the shift from raw model capability to useful, reliable systems: what changed with GPT-5.5, why reinforcement learning is moving beyond math and coding competitions into messy real-world work, how reasoning models like GPT-5.5 actually work, the difference between GPT-5.5 Thinking and GPT-5.5 Pro, why post-training has become one of the most important frontiers in AI, and why evals, model-as-judge, hallucinations, agentic workflows, GDPval, and continual learning are now central to the next phase of frontier models. Yann also shares why continual learning remains one of AI's biggest unsolved problems three years after ChatGPT, and where startups still have massive room to build as frontier models race ahead.(00:00) - Cold open(00:34) - Intro(01:30) - Why recent AI progress feels like a step function(04:13) - Model reliability & the rollercoaster of shipping 5.5(07:33) - How OpenAI structures vertical and horizontal teams(09:49) - Improving model efficiency and test-time compute(12:32) - Yann Dubois' journey from Switzerland to OpenAI(15:37) - Reasoning in 2026: Real-world utility vs verifiable rewards(18:34) - GPT-5.5 Thinking vs Pro: Scaling test-time compute(20:09) - How reasoning models become more efficient(23:23) - Pre-training scaling and overcoming the data wall(27:03) - Multimodal data, synthetic data, and embodied AI(31:05) - Demystifying mid-training and post-training(37:21) - Does RL create new capabilities in AI?(38:53) - The challenges and frontier of scaling RL(43:09) - Is building AI models a craft or a strict science?(48:21) - How AI models generalize across different domains(54:18) - How reinforcement learning cures AI hallucinations(56:04) - Negative generalization and conflicting instructions(58:05) - Can RL scale to law, medicine, and the broader economy?(1:00:19) - The evaluation bottleneck and Model as a Judge(1:04:21) - Continuous AI progress & continual learning(1:08:49) - Will foundation models eat the agent harness?(1:11:23) - Why startups should focus on the last mile of AI

In this session, we discuss how filmmaking encompasses a range of creative skills and also requires a collection of mentors for different aspects of the process.

This research examines the development and scaling laws of Native Multimodal Models (NMMs), which are AI systems trained from scratch to process both images and text simultaneously. The sources compare early-fusion architectures, which integrate raw multimodal signals from the start, against traditional late-fusion models that rely on separate pre-trained encoders. Findings indicate that early-fusion models are more efficient to train, easier to deploy, and perform as well as or better than late-fusion counterparts at lower compute budgets. Furthermore, the study highlights that incorporating a Mixture of Experts (MoE) significantly boosts performance by allowing the model to learn modality-specific weights. This specialized approach enables sparse models to handle heterogeneous data more effectively than dense architectures while maintaining the same inference cost. Ultimately, the reports suggest that NMMs follow predictable scaling properties similar to large language models, providing a blueprint for the next phase of edge AI development.

En este episodio de "Marcas que venden", conversamos con Amalia Carrasco, Directora de Comunicación de BlaBlaCar para España y Portugal. Amalia nos cuenta su transición desde el mundo de las agencias de comunicación hasta liderar la estrategia de una de las marcas más reconocidas en el sector de la movilidad. Exploramos cómo BlaBlaCar ha logrado pasar de ser una web para compartir coche a convertirse en un marketplace multimodal que integra autobuses y trenes. Analizamos el peso de la confianza como motor del negocio, cómo gestionan un rebranding que no pierda la esencia social de la marca y por qué el "boca-oreja" sigue siendo su canal de captación más potente. Si te interesa el marketing basado en el propósito, la gestión de reputación y las anécdotas reales de una comunidad que incluso termina en bodas, ¡este episodio es para ti!. Linkedin de Amalia: ⁠https://www.linkedin.com/in/amaliacarrascolozano/⁠ Linkedin de Capi Herrero: ⁠https://www.linkedin.com/in/juanmiguelherrero/⁠ Linkedin de Chema Martínez: ⁠https://www.linkedin.com/in/chema-martinez-pastor/⁠ Una producción de Formato Podcast: ⁠https://formatopodcast.com/⁠ Con la colaboración de Espacio Eventize: ⁠https://espacioeventize.com/

En este episodio de "Marcas que venden", conversamos con Amalia Carrasco, Directora de Comunicación de BlaBlaCar para España y Portugal. Amalia nos cuenta su transición desde el mundo de las agencias de comunicación hasta liderar la estrategia de una de las marcas más reconocidas en el sector de la movilidad. Exploramos cómo BlaBlaCar ha logrado pasar de ser una web para compartir coche a convertirse en un marketplace multimodal que integra autobuses y trenes. Analizamos el peso de la confianza como motor del negocio, cómo gestionan un rebranding que no pierda la esencia social de la marca y por qué el "boca-oreja" sigue siendo su canal de captación más potente. Si te interesa el marketing basado en el propósito, la gestión de reputación y las anécdotas reales de una comunidad que incluso termina en bodas, ¡este episodio es para ti!. Linkedin de Amalia: ⁠https://www.linkedin.com/in/amaliacarrascolozano/⁠ Linkedin de Capi Herrero: ⁠https://www.linkedin.com/in/juanmiguelherrero/⁠ Linkedin de Chema Martínez: ⁠https://www.linkedin.com/in/chema-martinez-pastor/⁠ Una producción de Formato Podcast: ⁠https://formatopodcast.com/⁠ Con la colaboración de Espacio Eventize: ⁠https://espacioeventize.com/

Is AI really the end of creativity, or the biggest emancipation of creative energy we've ever seen? How can authors thrive in a time of super abundance, when anyone can make anything? What happens when publishers become technology providers, and agents start shopping for books on our behalf? With Nadim Sadek. In the intro, my AI-Assisted Artisan Author webinars. This show is supported by my Patrons. Join my Community and get articles, discounts, and extra audio and video tutorials on writing craft, author business, and AI tools, at Patreon.com/thecreativepenn Nadim Sadek is a serial entrepreneur and the founder and CEO of Shimmr AI, an AI-powered book marketing company, as well as the bestselling author of children's books and non-fiction books, including Quiver, don't Quake: How Creativity Can Embrace AI. You can listen above or on your favorite podcast app or read the notes and links below. Here are the highlights and the full transcript is below. Show Notes Using AI as a research partner, editor, and constructive critic when writing a book The ratio of dreaming to execution Why publishers still draw red lines at AI-written words, and why that may change Inside Shimmr's three-engine advertising system: Strategizer, Generator, and Deployer Multimodal interactivity, agentic purchasing, and the idea of the Panthropic You can find Nadim on LinkedIn or at NadimSadek.com. Transcript of Interview with Nadim Sadek Jo: Nadim Sadek is a serial entrepreneur and the founder and CEO of Shimmr AI, an AI-powered book marketing company, as well as the bestselling author of children's books and non-fiction books, including Quiver, don't Quake: How Creativity Can Embrace AI. So welcome to the show, Nadim. Nadim: It is lovely to be here. I feel very privileged to be invited onto this. Thank you. Jo: Oh, I'm excited to talk to you today, and we're really talking about AI. I wanted to start with the fact that you do seem to have a sort of relentless optimism. How do you remain so optimistic about AI when the publishing industry that we both work in seems so overwhelmingly negative? Lift our eyes to the horizon—what is the bigger picture? Nadim: Oh my goodness. That is a big one. I think my optimism is quite confined actually in the area of publishing. If you were to ask me to speak about AI more broadly—which you're not, but I'm going to give you a little bit of it—I've got lots of concerns. That includes the advent of autonomous weapons and economic singularity, where the wealth from AI as an industry is going into just a few hands, and energy usage, and cultural homogenisation, I suppose, and the potential for brain rot. There's a whole pile of stuff which is really not very good about AI, and all the normal things about fraud and theft and so on. However, if you recognise that and then you say what's going on in publishing, then the obvious thing that you first have to deal with is what did happen with copyright. Is it appropriate to say that things have been stolen and taken without permission and so on? It is. It's going through the American courts at one pace. I saw that Penguin Random House have started a case against OpenAI in Germany, where there will be a much faster legal conclusion—a judge's conclusion, I think. This will begin to put parameters on how copyrighted materials can be used, and possibly also some retrospective judgment about what has happened to this point and what can be done about it. So it's good that you've asked questions so early in our conversation, because I think —  It's important to contextualise my optimism. It is whilst noting with regret the behaviour of the AI industry—the models themselves—in not dealing with copyright in the most generous or appropriate fashion. I think we should also recognise that copyright probably wasn't designed for machine learning in the way that it is. Probably the industry wasn't terribly well prepared to note, negotiate with, and navigate the very fast-moving technological culture of AI companies. So I think lots of mistakes have been made on both sides. When you put all that to one side, what's left for me is an amazing emancipation of creative energy and also a huge efficiency being brought to the publishing industry. We can talk about both those things further, but for me that is what's going on. The efficiency of bookmaking and publishing generally—the whole workflow of getting a book out of somebody's head and into a reader's hands—I think is immensely streamlined and improved by AI. Actually, if you talk about it carefully, which I'm sure we will do, the ability of creators to share and let others experience their creative endeavours becomes so much better, so much fuller, so much richer. So that's why I'm excited about it. Jo: Well, let's get into those two things then. You mentioned the emancipation of creative energy, and you've worked with various AI tools as part of your creative and business processes. You've said that AI can be a creative companion. So specifically when it comes to Quiver, don't Quake, for example— How are you using the various tools in such an emancipated way? Nadim: Well, just to put a bit of a broader context on it, we're an AI-native company at Shimmr, and separately I wear a hat as an author. You mentioned the AI books and the children's books. I'm also writing a book about the psychology of motorcycling. So it's a very odd authorial footprint, but it means that I kind of tramp around the place and learn different things. What I've noticed, even within Shimmr, is that the whole team has been using AI tools very differently. Lots of people are very bright in the company. They're all brighter than me, and I salute them and love them. But they've all used AI to become more creative in their own ways. For example, our Chief Commercial Officer is very numerate and logical, and not loquacious. She prefers to say things straight and simply. She has become an unbelievably creative financial modeller and analyst because she uses AI in lots of different ways. So she has flourished and grown so much, and is creative in a way that she never could be before—not only around numeracy and financial matters, but in thinking through new concepts for sales and marketing and for our commercial development. I've just noticed all around me this going on. When it comes to me, I prefer to express myself through writing. I talk a bit as well, as you can tell, but my favourite means of communication is just writing. When I was writing Quiver, don't Quake, I would use AI in a number of different fashions. One would be for research. One of the chapters is about the psychology of creativity. I'm a psychologist, so I tend to come at things from a psychological perspective. What is the psychology of creativity? Well, here comes a million-word answer from an AI—this person said this, this person said that. Then I kind of focused my research in particular areas and assembled them by drawing from the outputs of several AIs about what has been said about AI, what the science says about it, what sociology says about it, what particular creatives that we're all aware of say about it, whether they're in the advertising industry or musicians or artists or whatever. So that was a very rich way of researching things. I would often put a chapter in—this is a slightly different use—a manuscript that I'd written and say, “Read this as if you're somebody just coming across my book, and tell me where the reader might struggle between one paragraph and another, or where there's a logical fallout, or where the concept isn't really very fully excavated and developed.” It would occasionally prompt me to say, “You could probably do with a line that brings the reader from this point to that point.” And usually I listened to that and then wrote something new. In another use case, I eventually gave it the whole book and said, “I think I've done an okay job here and I quite like the flow and I'm sort of satisfied enough, but before I send it to the publisher and say, ‘there you go,' what do you think? Are there any ways in which this book could become a better and more interesting read?” It came back fairly promptly and said, “Well, what you haven't really done is considered what all the naysayers would say. You've done your dark moments of militarism and all that stuff, but what about some of the other stuff closer to publishing or creativity?” So off I went on a new round of research, and did some myself and used the AI for other bits. The funny thing, really the ironic thing here, is that the book is much better, and most people salute the book for the eighth to ninth chapter that talks about the constructive critics. I assemble them all and articulate all their arguments and say how hideous AI is and how terrible it is for the world and all of us. And then I try to repudiate some of them, not in a defensive way, but just to say, actually, yes, that's one perspective and here's another one. That chapter, ironically, about how AI is terrible was prompted by AI. It said, “You should really have a go at me.” And so I did. So that was another use case. Then finally—perhaps I'll say this—I have a friend who is, I think, the Editor-in-Chief of Penguin in India. I got to know her at a book fair or something. We started chatting, and I told her about my kids' books. I said, “I could really do with an editor on these ten books that are due to be published.” She very generously, amiably, and very constructively gave me feedback on each individual book and then on the whole set. I was really happy with it. I said to her, “That was a delight.” She said, “You'd be much better off working with Editrix.” I said, “What's Editrix?” She said, “Well, it's an AI platform I've created where you can go and self-edit.” I said, “You must be kidding. I'd much prefer chatting to you and our interactions.” She said, “Yes, well, go and try it.” So I got an account for the Editrix AI. Off I went, gave it my books, and lo and behold, it came up with some incredibly sophisticated and subtle observations on the books that neither Meru nor I had seen. For example, there's a story where a boy who lives in a house on a hill meets another boy on a bridge, and they end up in a silly confrontation. They're young and foolish, and it sort of transpires that the other boy lived in a local village. Now, I suppose in retrospect, it's pretty obvious that this could be seen to be colonialist, imperialist, and a sense of entitlement from the boy at the top of the hill crossing the bridge first and so on. Hadn't crossed my mind. The AI said, “I can tell from the rest of your writing that you don't really have a sort of racist or imperialist or superior attitude to things, but in this story, there could be a misapprehension that you do.” I thought, wow, what a great warning. So I changed it. There are almost endless ways—and I can tell you others, because I'm writing a book about clouds at the moment—in which AI can help you as an author. I've just shared some of those with you. Jo: Yes, well, I love that. I also use it for research. I definitely use the “give me feedback as a reader avatar, as a reader of this type of genre” or whatever. Nadim: Yes. Jo: I use different tools as well, so I agree with you. All of that is, I think, what a lot of people are doing. You also said you did a lot of the writing and rewriting, so the human was very much there. This was not an AI-generated work in any way. It was using an AI as a sort of collaborator—a creative companion, to use your words—which I think is great. One of the things that AI-positive people like us are finding is that there's so much negativity around the traditional publishers, around other authors, around supposedly negative backlash from readers. I think there's a lot of very noisy people who are probably making this sound worse than it is. Since you are so embedded in traditional publishing in so many ways, how are publishing people thinking about this? Do you think it's just different in terms of the creative side versus say the marketing side? What is happening there, and what do you recommend for authors? Nadim: What I'm observing is that there is increasingly confident adoption of AI for corporate efficiency, which is a polite way of saying where one can see profitability being improved. Could you streamline legal contracting? Yes. Can you manage royalty payments better? Yes. Are there better sustainability prospects with managing a warehouse and distribution and so on with AI? Yes. Could you improve your marketing by looking at competitive titles and trends, and optimising your metadata and your SEO and now your GEO, all using AI? Yes, yes, yes, yes, yes. All of these things can be assisted. Can you manage much more of your backlist, where you don't have the human or financial capital to manage all of those titles in a truly respectful and invested way? Yes, yes, yes. So wherever there's corporate efficiency, I see publishers being increasingly bold about saying they have integrated AI into their workstreams. What's much more tentative and hesitant is where there's discussion of authors—and I do hesitate to use the right words here—being assisted by, employing, working with AI. I kind of shorthand it as creative emancipation. It really means very many different things. Let me give you the example that I referred to briefly a second ago of Cloud Land, which is probably my first real novel. I'm very lucky. I sit working every day at a desk that's got three windows, and I look at the sky, and every day it's different, and I'm fascinated by it. I've been flying around the world since I was very young—my father worked for the World Health Organization, we moved between many countries—so I've also seen clouds from the sky a lot. I've noticed that in different parts of the world there are different cloud formations. It came to me one day that it would be very interesting if the clouds were somehow sentient, and that there is a cloud society, and that Cloud Land lived above human land and absorbed and observed us. Actually, the more I started thinking about it, the more I thought, well, we kind of evaporate. We give off vapour all the time and it rises up to clouds and maybe we're sending DNA signals to it, and it condensates and sends rain and storms and winds and lightning and thunder and all. There's a huge amount of interaction between Cloud Land and human land if you think about it. So I went into an AI. I said, “Hey, I've been thinking about this, blah, blah, blah. Any observations on what I've been saying so far?” I think one of the first things it said to me was, “You are actually playing with quantum physics.” I had no idea what quantum physics were really. I thought, well, this is interesting. I went and researched quantum physics, and actually there is some of that in it. If you count Cloud Land as a creative notion— The original idea, the creativity, came wholly from me, and then the development of it has been assisted by working with AI. I as a creator have spent much more time originating ideas about a story than would historically have been true. I probably would have gone to a library, tried to find the right geography textbook, read up about clouds, discovered what the nomenclature is, thought about whether I could put characters to cumulonimbus versus stratus something or other, and kind of worked my way gradually through it. There is something that I refer to in Quiver, don't Quake, which is what I call the ratio of dreaming to execution. I think previously, without AI, creators would probably spend 80% of their time researching and trying to get information and assembling things and editing documents and spell-checking and doing a whole pile of different tasks None of which I actually dismiss, because I think sometimes those difficult and “menial” tasks give you time to let ideas percolate and flourish and grow. It's just part of the process. But whereas before, I think we probably spent 20% of our time originating and 80% of our time assembling, I think it's inverted now. You can probably do 80% of the time you want creating and 20% of the time fiddling about getting your act together. So I feel that that's a huge emancipation of individual creativity. There's also—and we can talk about this if you wish—I think a much broader sociological phenomenon going on, which is really about every person in the world, all 8 billion of us, being creatives. That's the way I see the world. I think that only a minority of that 8 billion have the gift of craft that we recognise—of writing or drawing or making music or being an architect or a biomedical scientist or something that's creative and assembling things. And AI gives you courage and helps you to identify what you wish to make. I really don't mean creating the artefacts. I don't mean painting or making a song or writing a book. I just mean helping one to express and articulate oneself so that one's creative idea is shareable and experienceable by others. Jo: Well, it's interesting. I mean, everything that we've discussed, you're really saying that the main line is the actual writing of the words, because none of us can articulate how ideas come. Especially with Claude, we might have a creative spark, but I'm sure you've found the same: if I go to Claude, which is my favourite, with my creative spark, by the time we've discussed it, possibly over days, I've lost track of who said what. The idea definitely started with me, because the AI at the moment doesn't have its own creative spark in terms of its own drive to write a book, for example. So it starts with me, but then it goes back and forth, back and forth—sparks new ideas, something it wrote makes me think about something else. I think the difficulty with how publishing seems to be doing this at the moment is that it is just the written words on the page that is their red line around “have you used AI to generate a book?” But even that, I just think, surely that will change. For example, in the publishing industry, ghost writing—or writing dead authors, like Wilbur Smith—I was going to say Wilbur Smith is a good one. I mean, we've seen them, just different dead authors essentially writing in the voice of those people. So I just see that there are many possible places where publishers might want this kind of tool. I don't know— Do you see any openness to the actual words themselves? Nadim: I think you're right to identify that that is the place that it gets stickiest. What you kind of do in your private time—imagining and dreaming things up and interacting—it's a facsimile for talking to your friends or another author or something. It's just an AI companion. So I think that that is, you're right, less scrutinised. It is when one examines the words on the page. It's funny—it's almost as if it's a measure of how hard did you work to do this? Or did you just splatter it down on the page by pressing a button somewhere? It's almost as if, as creatives, we have to evidence that we have suffered, you know? I think there's a different form of suffering when you write with AI. It's true that if you command AI in some way to write for you, the default writing will be pretty anodyne, pretty bland, pretty mundane. It is deliberately so. AI is created and it is tuned to be inoffensive, to please most people, to be accessible to most readers and consumers of it. So it's another thing that I encourage people to do: don't approach AI with a kind of Google mindset where you just do a question and answer—”what time is it in New York now?” “Well, it's five hours behind” or whatever. Instead you say, “Hey, listen, I'm thinking about clouds, but I want a bit of spittle going up and down between the two, and I'd quite like a crazy cloud that harasses us.” Well, now I'm putting in some of my idiosyncrasy and my eccentricity and my personal perspective. The more you do that, the more that even if you did press a button and say, “Command, I want you to write this book,” that will no longer be a bland and mundane bit of output. It'll be very tuned by your interactions, and it'll exhibit some of your nature. So I think there probably are factories—there's always factories. They're probably—and actually I know this—writing a lot of romance, writing a lot of porn, things which are fairly well parametered. You know what happens in both of those genres more or less, so it's pretty easy for a machine to emulate what an author might write there and go and do it. But if you get into something like, “a sand dune was my cousin”—like, okay, well that's a bit different. What do you mean? And there it becomes a much more interesting bit of writing. So I think we're going to see a spectrum. To come back to your question about where publishers draw red lines, I think it's where they just see straight away mundane output that doesn't feel like it had a lot of craft or ingenuity or hard work to it. But I believe that as we go on, that's going to become harder and harder to establish. As we become more sophisticated users of AI, and AI's capabilities to understand us and to work with us become better, then I don't think it'll be such a big question where the words came from. What we'll feast on with each other is our creative ideas and how they're expressed, but not how they were produced. Jo: I mean, I always say to people, I'm not a word generator. That's not what makes me or my books worthy. It is what I do with it. It's the stories I tell, or it's the personal things behind it. So generating millions and millions of words, whether you generate them by typing or handwriting or AI or whatever, it isn't the word generation that is the point. It's all of the things that make that finished thing what it is. So anyway, let's come back to the other thing, because you mentioned that publishers seem very happy around corporate efficiency, anything that drives profitability. You also mentioned that Shimmr is an AI-native company. Now, I, and many people listening—we are a one-person company. So I run my own company. It's a publishing company. I do all my publishing, I do all my marketing, I do all my business as just me. So I also use AI for a lot of this stuff. I wondered— How do you see publishers changing to become more AI-native? How can we as individual author-publishers do that too? Because it feels like a massive mindset shift, not just plug in Opus 4.7 here. Nadim: I have been found saying at various publishing events—and it is deliberately a little bit provocative—that I believe that publishers have always been technology providers to creatives. It's not only what they do, but it is a part that they don't seem to embrace very hard. Even if you just go back to Gutenberg—I mean, here's a printing press, it's a bit of technology. “I'll make your book, I'll make your words into books.” It started there, and it's always been. That applies to distribution and e-commerce and audiobook manufacture and all sorts of other things along the way. So I encourage publishers to accept the notion that what they should do to attract authors in the future is partly—only partly—develop their own house AIs. It can be as ethically trained as that house wishes to deal with the copyright furore. It can be tuned to do editing in a particular way. It can have a specific way of copy editing. It can have a collaborative notion. It can have an assistant that helps you understand genres and hotspots and competitive titles. It can help you to think about, as Americans might say, what's hot and what's not in the world at the moment. So you might be more attuned to what the market demands, if that affects you at all. Some writers don't care, and that's fine. It can certainly help with all the marketing then. How can you produce social media content that's appropriate to your book, and all the rest of it. So I think there's a way in which publishers could massively enable authors. I talk to tons and tons of authors clearly about Shimmr, and what they all resent, I would say, is finding their time stolen by trying to flog their work rather than make it. Jo: Yes. Nadim: So the marketing process is just theft of creative time for most authors, and they hate doing it, and they're often not very good at it, because it's a completely different skillset from creating great stories or writing non-fiction books about particular subjects. So I believe that authors should be embracing the notion that publishers will create their own house AIs. And goodness me, we might even decide which publisher we prefer to go to on the strength of their AI position. Wouldn't that be interesting? But that is what I see the future being. Jo: Yes. I mean, definitely there's some quite significant authors—Dean Koontz, probably one of the biggest—who went to Amazon because of their technical ability around publishing and marketing. He was like, “Yes, I want this because of this.” Not that he'd be in bookshops or whatever—of course Dean Koontz is—but yes, so I think you're right there. For individuals also, as you know, we can use AI to help us market. I upload my books to Claude when they're finished, and I've just been marketing today. I'll say, “create 10 Midjourney images based on this book and give me all the marketing copy.” So I think we can use it now to help us be more efficient. On the other side of that, I think the bigger thing that's starting to happen is marketing is now much easier in one way. Nadim: Yes. Mm-hmm. Jo: So it's getting fuller, or even more. Nadim: Yes. Jo: So how do we deal with this? Because Shimmr is an AI marketing company. How are you thinking about the predominance of very, very good AI marketing now? Nadim: Yes, and it gets better all the time. It's a great question. Obviously, strategically, as an enterprise, we've really had to think about this one. If I go back one step, I always believe that innovation succeeds when it starts in a narrow space. So when Shimmr launched, we put ourselves forward and were quickly embraced, I have to say, as automated advertising that sells books. Nothing particularly more complicated than that. “Okay, you do ads, you automate it for me, and it'll help flog my books. Yes, that's it.” We had a rush. We've worked with about 250 publishers. As you might anticipate, it started with smaller ones, then got bigger. We now work with the biggest as well. That notion of automated advertising selling books was successful. Actually, that was about three years ago—a bit shorter than three years ago. What's happened in that time is that we have now collected a ton of data, and meanwhile the AI models have become more sophisticated and competent. Maybe I should just pause briefly and say what Shimmr actually does. We've got three main engines that are all chained together, to use pretty old language. The first one is what we call the Strategizer. It reads the book, it understands what we call its book DNA. So it's the structural elements of what the narrative is, who the protagonists are, and all the rest of it. It's also a psychological study of it—what's going on, what are the emotions or the values, what are the interests, how they intersect, where are the tensions, all those sorts of things. The Strategizer decides, “Well, reading everything between the covers of this book and understanding the author's intent, this is the best way to put this book forward because here are its strong points.” It hands that off to the second machine, which we call the Generator, which says, “Thanks for the creative brief. I'll make you the ads now.” It does videos and music and captions and all the rest of it. Then it presents its newly baked campaign to the third machine, which is the Deployer, that says, “Okay, well, I know where to find the audiences for this. If that's the DNA of the book and this is the campaign that manifests it, then I know where to find these people.” It goes and autonomously deploys it in various media channels to specific audiences who might be interested in that content. So that's what we started doing, and that generated a huge amount of data. Where we've got to recently—really in the last six months—is understanding that, as you've just said, most people can generate their own stuff. So in some ways they can look just like a mini Shimmr. The thing that differentiates the content is always the strategy. What we have learned to do now—and it's because of an agentic framework—is we've moved beyond what's between the covers of the book to look at life. We look at culture, what's going on, what are the trends, what's in and what's out. Even if you take a particular trend—let's say, fascism—what's the language associated with it that's being treated positively and respectfully, and what's the stuff that leads to it being dismissed straight away? All those sorts of nuances around everything. But equally, as well as going deep with a set of agents on what fascism might be in today's culture, we also go wide and say, “Well, how does that sit next to loyalty or hedonism or ambition or something else?” So we get this very, very circumspect analysis of the market. Then, indeed, if you do write a book about—I'm really going off-piste here, but you know, the hedonism of fascism, like, God, that would be a weird book—you discover that actually you're not really competing with another book, but you are competing with that specific podcast and this movie that came out, and another movement that's born in Italy but it's moving across Europe now or something. So we were able to produce strategies which now lead to a much broader offer, one which is much more sophisticated and much more likely to drive success in a book or in a creative enterprise. It informs product listings, metadata, author communications, PR, SEO, GEO, and of course the thing that we started with, advertising. So things that you see made by Shimmr should be much more resonant and much more attuned to the world, and commercially much more likely to drive success, than simply saying, “Here's a book, make ten Midjourney images out of it.” Jo: Mm-hmm. Nadim: It's really about the quality of the briefing and the quality of the assets that you're able to produce by having a much more sophisticated Strategizer. So we've gone back into the intellectual property and the human analysis, in a way, of the world. To understand where a specific piece of creative work sits in culture and society has become a much bigger proposition. Jo: Right. So you did mention podcasts there. So as in, you might present to a publisher “these are the podcasts that they should pitch” for example? Nadim: There's that, of course, but it's also, don't think that this book is competing with these three titles which your team put together. It's more that, if people want to listen to hedonistic fascism, they can listen to that podcast before they read this book. Jo: Okay, that's interesting. Interesting times. So we don't have much time left, but I think one of the biggest questions that people have—even if they're AI-positive, as I am and many people listening are—it's not that we're worried about AI replacing us, because we know we're individuals and all that, but we are slightly concerned about the volume of books in the market. And not just books, but TV shows and YouTube and TikTok. It's very hard to stand out. You do say in the book: “When anyone can make, maybe creativity lies not in the making, but in making others care.” How can I move up the value chain? So for many of us who make an income this way, what are your recommendations? Nadim: Great question. And actually I think it's really central. My latest catchphrase is that in a time of super abundance, we need super discoverability. So it's exactly as you just said—tons of work, tons of movies, tons of podcasts, and tons of everything. If you believe in what I've been saying, which is that we're emancipating the creative spark of 8 billion people, there's going to be even more. So I believe that the solution is what I call multimodal interactivity. That doesn't mean multimedia—it means multimodal. Multimodal means you can engage with an experience in different modalities—the same idea. So my conviction is that if you write a book or make a painting or have a piece of music that you've come up with—or anything really, creatively—and you wish it to both survive the first six weeks of its birth and then thrive in a more perpetual way in society and culture, then people have to be able to experience and engage with your idea in multiple modalities. I would always write a book, because that's what I do. Others produce a podcast or write a piece of music—whatever the same sort of things. Any one of us needs to make sure that that reappears and is experienceable and interactable with in different modalities. So my book should have some Instagram reels. There might be YouTube shorts, there might be a podcast, there might be a piece of music associated with it, it could be a movie. It could be a game, it could be an app. You really have to think about allowing your creative idea—more than your creative artefact—to live in culture. Sure, you want to make an income from the artefact that you are good at producing. As many of your listeners, and I, would be writers of books, we want that to persist as a revenue stream, and it should do. I would simply argue that making sure that whatever you've produced in your book is manifest, and people can interact with it in other modalities, is the surest way to get it seen and discovered. Jo: Yes, it's interesting. I've actually started looking at making my non-fiction books into skills. Nadim: Yes. Jo: And also making markdown MD files—books as markdown files for agents to buy. Nadim: Very good. You are way ahead of the curve. Jo: Well, I sell on Shopify, as do many listeners, and Shopify, as I'm sure you know, is now enabled for agentic purchasing. We are in ChatGPT. So it's really interesting to think, well, if the agents go shopping for people now and in the future, what you want is to be able to find it. Also, I haven't actually put an explicit licence, but people email me and say, “Can I upload your books into an LLM?” And I'm like, “If you buy a copy from me, then yes, you can.” Nadim: Yes. Jo: So I think it's changing. And as you say, I do think that people are more and more going to want to say “buy the PDF and put it in NotebookLM” or use it as a skill. Nadim: That's right. Jo: That kind of thing. Nadim: Yes, and then they go on a walk with their dog and they listen to the podcast about your book, which they've created on NotebookLM. It's exactly that. I think my worst fear for publishers is that they lose so much of the value chain—distribution, creative collaboration, all sorts of things along the way—that the worst position they could end up in is simply as book manufacturers, which would be just one small manifestation of a creative idea. Jo: Well, I'm excited about the future. I hope you are too. I think you are. What are you particularly excited about in terms of the changes coming? Nadim: Well, if I can be my most extravagant now, my greatest excitement about AI and the changes that are coming are that it'll produce what I describe as the Panthropic. The Panthropic is a way of seeing AI not as a companion or some anthropomorphic being, but instead the repository of everything that humans have ever thought or felt or created or shared, accessible to us all in an anonymised way. It's just a repository of interactable information. My excitement about it is that the liberation that that gives to information—which becomes knowledge, which of course we all know leads to some power—should result in truly new thinking, new philosophy, new spiritualism, possibly new questions about what it is to be a human being and what life on Earth is all about. New economics, new employment, new education. I think one can too easily underestimate the massive liberation of intellectual consideration and creativity that's about to surf across the globe, and I'm so excited by it. Jo: Mm-hmm. Yes, me too. Very interesting times ahead. So where can people find you and your books and everything you do online? Nadim: I think the easiest thing is just to go to LinkedIn and find me there as Nadim Sadek. You can also go to my personal website, which is NadimSadek.com, and that'll take you wherever you want on different journeys and different parts of my career. It'll also give you links to books. Of course, they're available in all formats—audio, paperback, ebook—and in many different languages, all through Amazon and other platforms, and Spotify and Audible and all the usual things. Jo: All the usual things. Well, thanks so much for your time, Nadim. That was great. Nadim: It's a pleasure. Thank you so much for having me.The post AI, Creativity, And The Future of Publishing with Nadim Sadek first appeared on The Creative Penn.

This podcast sponsored by Virbac. In honor of Mobility Awareness Month this May, dvm360 is shining a spotlight on the movement health of veterinary patients. Joining host Adam Christman, DVM, MBA, is Kara Amstutz, DVM, DACVSMR (Canine), CVA, CVPP, CCRT, to discuss how proactive mobility care profoundly impacts the quality and duration of life for aging pets.Together, they explore the necessity of a comprehensive, multimodal approach, integrating pain management, targeted nutrition, and physical exercise, while highlighting the essential role joint supplements play in long-term support.

In this must-listen episode, Dennis sits down with Dr. Jon Andrews—former 5th and 20th Group Special Forces medic turned Duke-trained anesthesiologist (pediatric & cardiac fellowships)—to tackle one of the biggest headaches in austere medicine: you have a tiny box of opioids and ketamine, a long mission, and a patient who needs to stay alive AND comfortable.They break down exactly how to stretch every milligram using real OR strategies adapted for prolonged field care: patient-specific planning, smart titration, multimodal synergy, regional blocks, ketamine myths, and when (and how) to layer non-narcotics without crashing your patient or your supply.Why this episode matters: Acute pain becomes chronic pain. Chronic pain leads to opioid dependence, PTSD, and worse outcomes. In the field, your choices today shape your patient's tomorrow—and whether you still have meds left when the next casualty shows up.Key TakeawaysStart low, titrate smart. Cut your first dose in half on sick or unstable patients. You can always give more—never the other way around.Multimodal is mission-critical. Hit pain from every angle (blocks + ketamine + acetaminophen + judicious NSAIDs) to dramatically reduce opioid requirements and prevent chronic pain pathways.Ketamine IS an analgesic. It's not just dissociation—it's an NMDA antagonist that blunts central sensitization and has proven opioid-sparing effects.Schedule your non-opioids. Acetaminophen (1 g IV/PO/PR q6h) and longer-acting adjuncts form your baseline; use fentanyl or morphine only for breakthrough.Blocks beat everything—if you can do them. Pre-emptive regional anesthesia (when feasible) is the single highest-yield move before surgical stimulus hits.Monitor like your life depends on it. Heart rate, blood pressure, and respiratory rate are your best pain score when the patient can't talk.Plan for worst-case evacuation. Bring more than you think you'll need and dose for the opioid-naïve or opioid-tolerant reality in front of you.Why treating hypertension in the OR (or field) almost always starts with fixing pain firstThe “start low, see response, add more” mantra every austere provider needsWhy Tylenol often performs as well as morphine in blinded ED studies (and why your patients still doubt it)Real talk on ultrasound-guided blocks in 2011 vs. today—and why proficiency still mattersThe dangerous synergy of opioids + benzos + ketamine on respiratory driveWhy you must get comfortable decreasing doses, not just ramping them upChapters01:55 – The austere reality: limited narcotics and why your favorite med won't last forever03:37 – OR planning vs. field reality: opioid-naïve vs. chronic users05:57 – Multimodal analgesia explained (blocks, ketamine, Tylenol, NSAIDs, dexmedetomidine)08:28 – Patient & mission factors that should drive your loadout12:23 – Golden rule: start low, titrate to effect, monitor vitals15:05 – Sick-patient hack: cut your mental dose in half16:01 – Is ketamine actually an analgesic? (NMDA, opioid-sparing, PTSD data)19:12 – Extending your supply: bolus vs. infusion, redosing strategy24:27 – First-line multimodal choices in the field27:43 – Juggling multiple agents: timing, scheduling, and longer-acting blocks30:15 – Regional anesthesia timing—pre-emptive is king (post-injury limitations)32:48 – Ultrasound & blocks in the current PFC world35:08 – Safety considerations for adjuncts (liver, kidneys, bleeding, alcohol)36:59 – Bang-for-buck data on Tylenol vs. morphine38:55 – Practical integration: layering Tylenol/ketamine with fentanyl titration41:54 – Getting comfortable titrating down (and why pain scores can lie)42:53 – Final wisdom: use everything you're comfortable with.For more content go to ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠www.prolongedfieldcare.org⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Consider supporting us: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠patreon.com/ProlongedFieldCareCollective⁠⁠⁠⁠⁠⁠ or ⁠⁠⁠⁠⁠⁠www.lobocoffeeco.com/product-page/prolonged-field-care

The countdown started yesterday in my kitchen, as my daughter flipped the calendar forward for something and realized she had less than thirty days of school left. She loves her teacher and looks forward to school, so she felt sad. It launched her into a story about how her class is trying to convince her teacher to move to the next grade with them. If you, too, are starting to plan ahead and think end-of-year thoughts, today I want to share a way to help students review and reflect on the year in one multimodal activity. I've had requests in The Lighthouse for ways to help students reflect on their own learning - to tell their own learning story. Research backs the importance of metacognitive reflection for students - in other words, it's helpful for them to think not only about what they've learned, but also how they've grown and developed as learners, and where they might want to go next. Before we dive in, feel free to grab the free curriculum that goes along with this episode. Everything pictured below and discussed throughout the episode is already set up to make this activity as easy to implement for you as possible! And yes, the handouts are editable so you can tweak them to suit your own twist on the activity. Grab the free curriculum for this activity: https://sparkcreativity.kartra.com/page/endofyearhexagons  Go Further:  Explore alllll the Episodes of The Spark Creativity Teacher Podcast. Grab the free Better Discussions toolkit Join our community, Creative High School English, on Facebook. Come hang out on Instagram.  Enjoying the podcast? Please consider sharing it with a friend, snagging a screenshot to share on the 'gram, or tapping those ⭐⭐⭐⭐⭐ to help others discover the show. Thank you! 

A proud California girl, Louisa Frahm was born and raised in San Diego. She received her undergraduate degree in journalism from the University of Colorado at Boulder in 2012. Throughout the past decade, she's built a booming career in the news SEO world, conducting search efforts at E! Online, Yahoo!, TMZ, People Magazine, Entertainment Weekly, the Los Angeles Times, and ESPN. She also served as a Trends Curator on the Google Trends team. To bolster her communications skill set, she acquired a master's degree in communication management from the University of Southern California in 2021. Leadership development and mentoring are two of her biggest professional passions. When Louisa isn't busy with work, she loves enjoying any and all things pop culture with her family and friends. Her Funko Pop collection is over 100 figurines strong. Ask her about Prince, Freddie Mercury, and David Bowie.

In this episode of the PRS Global Open Keynotes Podcast, Dr. Paulo Michels discusses his technique for minimally invasive full body remodelling. His technique involves pre-surgery diet, intraop liposuction, rib recontouring, ultrasound guided fat grafting and video assisted minimally invasive lipoabdominoplasty. This episode discusses the following PRS Global Open article: "Full-body Remodeling with Minimally Invasive Techniques" by Paulo Michels, Ricardo Araujo and Rafaela T.B. Michels. Read it for free on PRSGlobalOpen.com: https://journals.lww.com/prsgo/fulltext/2026/01000/full_body_remodeling_with_minimally_invasive.60.aspx Dr. Paulo Michels is a plastic surgeon in Abu Dhabi. Your host, Dr. Damian Marucci, is a board-certified plastic surgeon and Associate Professor of Plastic Surgery at the University of Sydney in Australia. #PRSGlobalOpen; #KeynotesPodcast; #PlasticSurgery; Plastic and Reconstructive Surgery- Global Open The views expressed by hosts and guests are their own and do not necessarily reflect the official policies or positions of ASPS.

Send us Fan MailPaper Discussed in this Episode: Artificial intelligence in clinical oncology: Multimodal integration and translational development. Ruichong Lin, Zhenhui Zhao, Zhonghai Liu, Jin Kang, Kang Zhang, Xiaoying Huang, Yunfang Yu. Cancer Letters 2026; Volume 649, 218493.Episode Summary: In this journal club deep dive, we explore how cutting-edge AI is fundamentally rewriting the rules of cancer diagnostics. We examine a comprehensive 2026 review on clinical oncology that highlights the shift from narrow, single-modality algorithms to highly sophisticated multimodal AI. We discuss how machines are learning to cross-reference patient charts, genomic data, and medical imaging simultaneously to achieve unprecedented feats—like accurately predicting tumor mutations without ever performing a physical biopsy. Plus, we explore the controversial but necessary world of "computational hallucinations" or synthetic data, which is currently being used to solve diagnostic blind spots.In This Episode, We Cover:• The Fragmentation Bottleneck: Why keeping radiology, pathology, genomics, and clinical history in isolated silos limits our ability to treat the whole patient, and why single-modality AI suffers from severe diagnostic "tunnel vision".• Cross-Modal Attention & Non-Invasive Biopsies: How models like LUCID essentially mimic the deductive reasoning of a multidisciplinary tumor board. By utilizing cross-modal attention mechanisms, LUCID dynamically shifts focus between CT scans, routine labs, and text-based clinical charts to predict EGFR gene mutations in lung cancer entirely non-invasively.• Graph Neural Networks (GNNs) & Tumor Social Networks: A look at the NePSTA framework, which uses GNNs and spatial transcriptomics to treat the tumor microenvironment like a mathematical topology. By mapping the "social network" of cells, it can rapidly molecularly subtype notoriously ambiguous central nervous system (CNS) tumors in minutes.• Computational Hallucinations: Introducing MINIM, a generative AI foundation model that creates statistically valid, photorealistic synthetic medical images (like optical CT or chest X-rays) for rare diseases based on textual descriptions. We discuss how intentionally generating these synthesized images solves the critical "data scarcity" problem and directly improves real-world diagnostic accuracy.• The Reality Check - Distribution Shifts: The dangerous logistical reason why an AI model boasting near-perfect accuracy at a massive urban academic center might fail completely in a rural clinic due to differing scanner calibrations and population demographics. We emphasize why the field must transition away from retrospective "vanity metrics" and toward clinically trustworthy prospective validation.• The Virtual Cell Paradigm: A staggering look into the near future where AI constructs completely accurate, computationally interactive digital twins of a patient's cancer. This framework allows doctors to test different drug regimens and simulate cellular responses mathematically in silico before ever administering medicine to the actual patient.Key Takeaway: Multimodal AI proves that cancer diagnostics must go beyond isolated data points. By dynamically synthesizing highly fragmented clinical information and utilizing synthetic imaging to overcome rare disease data scarcity, AI is pushing oncology into an era of robust, individualized molecular phenotyping. Ultimately, these innovations are replacing risky, invasive testing with precSupport the showGet the "Digital Pathology 101" FREE E-book and join us!

Katharine Jarmul, Privacy in ML/AI Expert & Author of Practical Data Privacy, joins Hugo to unpack why most AI privacy advice is theater: and what technical privacy actually looks like when you're shipping LLMs, agents, and multimodal systems into the real world.In this episode, we dig into how to build defensible systems in an era of AI agents and multimodal models: why system prompts (and your entire agent harness!) should be considered public by default, and why “privacy observability” is as critical as data observability for anyone building with LLMs today. Multimodal is what changes the threat model: identifiers hide in images, audio, and metadata, not just text, and the old anonymization playbook doesn't cover it.Vanishing Gradients is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.We Discuss:* No Convenience Tax, you don't have to trade privacy for utility: high-utility AI products can be privacy-preserving through technical controls like privacy routing and input sanitization;* Public Prompts and Harnesses: assume any instruction or secret in a system prompt or agent harness will be exfiltrated; don't put sensitive info there in the first place;* Privacy Observability, tag and track data flows so information is used only for its original intended purpose: catch design flaws before they become legal problems;* Technical Privacy, implement mathematical and statistical constraints directly into ML systems and data flows so privacy is measurable and enforceable, not aspirational;* Tiered Guardrails, a three-layer approach: deterministic filters for hard rules, algorithmic models for nuanced classification, and internal alignment training for behavioral baselines;* Federated Learning Is Not Privacy, model updates in FL leak sensitive data on their own: you must layer differential privacy or encrypted computation on top, or you're reverse-engineerable;* Anonymization Spectrum, navigate the “grayscale” of privacy in multimodal AI, balancing data utility and individual risk as identifiers hide in non-obvious places;* Privacy Champions, embed privacy accountability directly into development by training and incentivizing engineers inside product teams;* Red Teaming as Ritual, your goal is to attack yourself: practice thinking like an attacker, and turn privacy testing into an organization-wide creative ritual rather than a siloed security task.You can also find the full episode on Spotify, Apple Podcasts, and YouTube.You can also interact directly with the transcript here in NotebookLM: If you do so, let us know anything you find in the comments!

Send us Fan MailHow close is pathology AI to making decisions that matter in real workflows, real trials, and real patient care?In this episode of DigiPath Digest, I review five recent papers that approach that question from very different angles. We look at multimodal survival prediction in cervical cancer, pathology-driven response assessment in neoadjuvant immunotherapy for head and neck squamous cell carcinoma, AI-assisted Ki-67 scoring in pulmonary neuroendocrine neoplasms, automation and AI in hematologic diagnostics, and AI-based qFibrosis readouts from the Phase 3 MAESTRO-NASH trial.What I liked about this set of papers is that they do not all tell the same story. Some show clear progress. Some show where AI already works well as an adjunct. Others make it very clear that validation, governance, reproducibility, and workflow design still matter just as much as model performance.Key topics and timestamps00:00 Introduction, Easter edition, and community updates 00:51 USCAP recap, signed book giveaway, and free Digital Pathology 101 PDF 02:04 Partnerships, lab automation preview, and what's coming in this episode 03:25 Multimodal deep learning for cervical cancer survival prediction 13:00 Why pathology may be a better response endpoint than radiology in neoadjuvant HNSCC immunotherapy 23:09 Ki-67 scoring in pulmonary neuroendocrine neoplasms: pathologists vs two AI systems 33:46 AI, digital morphology, and automation in hematologic diagnostics 43:29 qFibrosis, digital biomarkers, and the MAESTRO-NASH Phase 3 trial 51:57 Closing thoughts, community updates, and Easter promotion Resources Deep Learning Can Predict the Overall Survival of Cervical Cancer Based on Histopathological Image, Gene Mutation and Clinical Information https://pubmed.ncbi.nlm.nih.gov/41902378/ Modern Pathology-Driven Strategies in Neoadjuvant Immunotherapy for Head and Neck Squamous Cell Carcinoma: From Residual Tumor Quantification to Spatial and AI-Based Biomarkers https://pubmed.ncbi.nlm.nih.gov/41899621/ Ki-67 Proliferation Index in Pulmonary Neuroendocrine Neoplasms: Interobserver Agreement Among Pathologists and Comparison of Two Artificial Intelligence-Based Image Analysis Systems https://pubmed.ncbi.nlm.nih.gov/41898274/ Molecular Pathology, Artificial Intelligence, and New Technologies in Hematologic Diagnostics: Translational Opportunities and Practical Considerations https://pubmed.ncbi.nlm.nih.gov/41897649/ Quantitative regression of qFibrosis with resmetirom: Exploratory histologic endpoints from the MAESTRO-NASH phase III clinical trial https://pubmed.ncbi.nlm.nih.gov/41895606/Support the showGet the "Digital Pathology 101" FREE E-book and join us!

What does it take to build a multimodal transportation network that actually works for rural communities, growing regions, and everyone in between? In Kansas, that question is critical as the Kansas City Chiefs plan their stadium move and the 2026 FIFA World Cup comes to town.   Listen in as we sit down with Matt Messina, Chief of Multimodal Transportation at the Kansas Department of Transportation (KDOT), to explore how Kansas is prioritizing transit solutions for upcoming projects and how community input shapes decisions. It's an insightful journey into the challenges and opportunities of public transit, pedestrian infrastructure, and the future of mobility.   We'd love to hear from you. Share your comments, questions and ideas for future topics and guests to podcast@sae.org. Don't forget to take a moment to follow SAE Tomorrow Today—a podcast where we discuss emerging technology and trends in mobility with the leaders, innovators and strategists making it all happen—and give us a review on your preferred podcasting platform.   Follow SAE on LinkedIn, Instagram, Facebook, X, and YouTube. Follow host Grayson Brulte on LinkedIn, X, and Instagram.

We've been on a bit of a mini World Models series over the last quarter: from introducing the topic with Yi Tay, to exploring Marble with World Labs' Fei-Fei Li and Justin Johnson, to previewing World Models learned from massive gaming datasets with General Intuition's Pim de Witte (who has now written down their approach to World Models with Not Boring), to discussing the Cosmos World Model with with Andrew White of Edison Scientific on our new Science pod, to writing up our own theses on Adversarial World Models. Meanwhile Nvidia, Waymo and Tesla have published their own approaches, Google has released Genie 3, and Yann LeCun has raised $1B for AMI and published LeWorldModel.Today's guests have a radically different approach to World Modeling to every player we just mentioned — while Genie 3 is impressive, its many flaws demonstrate the issues with their approach - terrain clipping, noninteractivity (single player, no physics/no objects other than the player move), and maximum of 60 second immersion. Moonlake AI (inspired by the Dreamworks logo) is the diametric opposite - immediately multiplayer, incredibly interactive, indefinite lifetime, capable of MANY different kinds of world models by simulating environments, predicting outcomes, and planning over long horizons. This is enabled by bootstrapping from game engines and training custom agents: In Towards Efficient World Models, Chris Manning and Ian Goodfellow join Fan-Yun in explaining why their approach to efficiency with structure and casuality instead of just blind scaling is sorely needed:SOTA models still show physical or spatial understanding glitches, such as solid objects floating in mid-air or moving “inside” other solid objects.If the goal is to plan for the next action, how often is a high-resolution pixel view necessary for modeling the world? Our bet is that there is a disproportionately large share of economically valuable tasks where such detail is not required. After all, humans with a wide variety of sensory limitations have little difficulty doing almost everything in the world. Furthermore, for a large number of purposes, describing a scene or a situation in a few words of language (“the car's tires squealed as it cornered sharply”) is sufficient for understanding and planning.Experiments also show that humans only partially process visual input in a top-down, task-directed way, often making use of abstracted object-level modeling. In almost all cases, partial representations combined with semantic understanding are sufficient.…If the goal is to facilitate the understanding of causality in multimodal environments, then the world model—whether it is used in the virtual world or the physical world—must prioritize properties such as spatial and physical state consistency maintained over long time periods, and an ability to evolve the world that accurately reflects the consequences of actions. That's what Moonlake is building.Game engines are the right starting point abstraction to efficiently extract causal relationships, and building the interfaces and community (including their new $30,000 Creator Cup) to kickstart the flywheel of actions-to-observations.We were fortunate enough to attend their sessions at GDC 2026 (the Mecca of Game Devs), and were impressed by the huge variety and flexibility of the worlds people were building with Moonlake's tools already! Live videos on the pod.Full Video Pod on YouTube!Timestamps00:00 Benchmarking Gets Hard00:47 Meet Moonlake Founders01:26 Why Build World Models03:12 Structure Not Just Scale05:37 Defining Action Conditioned Worlds07:32 Abstraction Versus Bitter Lesson14:39 Language Versus JEPA Debate20:27 Reasoning Traces And Rendering Layer37:00 Gameplay Over Graphics38:02 Fiction Rules And World Tweaks39:15 Code Engines Beat Learned Priors41:10 Diffusion Scaling Limits43:23 Symbolic Versus Diffusion Boundary46:14 Platform Vision Beyond Games50:24 Spatial Audio And Multimodal Latents54:23 NLP Roots Hiring And Moon Lake NameTranscript[00:00:00] Cold Open[00:00:00] Chris Manning: Think this whole space is extremely difficult as things are emerging now. And I mean, it's not only for world models, I think it's for everything including text-based models, right? ‘cause in the early days it seemed very easy to have good benchmarks ‘cause we could do things like question answering benchmarks.[00:00:20] But these days so much of what people are wanting to do is nothing like that, right? You're wanting to get some recommendations about which backpack would be best for you for your trip in Europe next month. It's not so easy to come up with a benchmark, and it's the same problem with these world models.[00:00:41] Meet the Founders[00:00:41] swyx: Okay. We're back in the studio with Moon Lake's, two leads. I, I guess there's other founders as well, but, sun and Chris Manning. Welcome to the studio.[00:00:54] Fan-yun Sun: Thanks. Thanks, Chris. Thanks for having us.[00:00:56] swyx: You've got, you guys have, come burst onto the scene with a really refreshing [00:01:00] new take of mold models.[00:01:01] I would just want to, I guess ask how you, the two of you came together. Chris, you're a legend in NLP and just AI in, in, in general. You're, you're his grad student, I guess[00:01:10] Fan-yun Sun: Actually my co-founder.[00:01:11] swyx: Oh, yeah.[00:01:12] Fan-yun Sun: I should give a lot of credit to my co-founder, Sharon. Yeah. She was, she was actually working with Professor Fe Androgyn and then she ended up working with, Ron and Chris Manning here.[00:01:22] And then, so I got connected through to Chris initially, actually through my co-founder,[00:01:26] What is Moon Lake?[00:01:26] swyx: what is Moon Lake? What, what is, actually, I'm also very curious about the name, but like why going into world models?[00:01:33] Fan-yun Sun: So I was working a lot. With actually Nvidia research during my PhD years on essentially generating interactive worlds to train reinforcement learning agents or embody EA agents.[00:01:44] And then there's two observations. One in academia and one in industry. An industry like folks at Nvidia are actually paying a lot of dollars to purchase these types of interactive worlds, whether it's for the sake of evaluation or training the robots, or policies or models. And [00:02:00] then, in academia, same thing is happening.[00:02:02] And more specifically, when I was actually working with Nvidia on the synthetic data foundation model training project, we were actually generating a lot of these synthetic data and showing that, hey, you can actually, these synthetic data are actually as useful as real world data when it comes to multimodal pre-training.[00:02:16] But then, like I said, there's a lot of dollars being paid out to like external vendors or, or like. Other folks to manually curate these types of data. It was very clear to us that, okay, on our way to, let's call it embody general intelligence models need to learn the consequences behind their actions, which means that they need interactive data and the demand for those types of data are growing exponentially.[00:02:38] But everybody's sort of thinking about it from a pure, say, video generation perspective or something else. But we feel like the true actually opportunity is actually building reasoning models that can do these things, like how humans do these things today. So that's a little bit on the genesis of Moon Lake, and I think the reason I got into world models was partly.[00:02:59] A philosophical [00:03:00] take of the on the world where I like, believe the simulation theory and stuff like that. But on the other, on the other hand, it's really just like, oh, like there's an opportunity there that I feel like nobody's doing it the way I think should be done.[00:03:10] Structure, Not Scale: The Vision[00:03:10] Chris Manning: I can say a little bit about that.[00:03:12] Yeah. So of the overall goal is the pursuit of artificial intelligence and most of my career has been doing that in the language space and that's been just extremely productive. As we all know, the story of the last few years, I don't have to tell about how much we've achieved with large language models, but, uh.[00:03:31] Although they have been extremely effective for ramping language and general intelligence, it's clearly not the whole world. There's this multimodal world of vision, sound, taste that you'd like to be dealing with more than just, language. And then the question is how to do it. And despite, a huge investment in the computer vision space, right, as the research field computer [00:04:00] vision has been for decades, far, far larger than the language space, actually.[00:04:05] I think it's fair. Say that, vision, understanding sort of stalled out, right? You got to object recognition and then progress just wasn't being made right? If you look at any of these, vision language models, it's the language that's doing 90% of the work and the vision barely works. And so there's really an interesting research question as to why that is and at heart, the ideas behind Moon Lake are an attempt to answer that, believing that there can be a really rich connection between a more symbolic layer of abstracted understanding of visual domains, which aren't in the mainstream vision models, which are still trying to operate on the surface level of pixels.[00:04:50] swyx: I think one of your blog posts, you put it as structure, not scale. Is that, a general thesis?[00:04:57] Chris Manning: Yeah. Well, scale is good too.[00:04:58] swyx: Yeah. Scale is good. Too[00:04:59] lot,[00:04:59] Chris Manning: [00:05:00] lots of data is good as well and scale, but nevertheless, you want the structure Yeah. To be able to much more efficiently learn.[00:05:07] swyx: Yeah. The other thing I really liked also is you put out an example of what your kind of reasoning traces look like.[00:05:12] Right. Which you would distill is the word that comes to mind. I don't even think that's a good, good description, but it would involve, for example, geometry, physics, affordances, symbolic logic, perceptual mappings, and what, what have you. But like that, that is the kind of example that involves, let's call it spatial reasoning, role model reasoning as as compared to normal LM reasoning.[00:05:35] Yeah.[00:05:36] Defining World Models vs Video Generation[00:05:36] Vibhu: But also like taking it a step back. So how do you guys define world models? A lot of people see okay, you can do diffusion, you can do video generation. But, you guys put out quite a few blog posts. You put out a essay recently, we can even pull it up about efficient world models. You have a pretty like structural definition here, but for the general audience that don't super follow the space, right.[00:05:55] What's, what's the difference in what we see from like a video generation model to [00:06:00] a world gen A simulator? How do you kind of paint that last[00:06:02] Chris Manning: year? Yeah, so I think this is actually a little bit subtle because, people look at these amazing generative AI video models, SAWA VO three, one of these things, and they think Genie, they think, oh, this is amazing.[00:06:17] This is we've solved understanding the world because you can produce these generative AI videos, but. The reality is that although the visuals do look fantastic, those visuals actually are accompanied by an understanding of the 3D world, understanding how objects can move, what the consequences of different actions are, and that's what's really needed for spatial intelligence.[00:06:49] So I mean, a term we sometimes use is that you need action condition, world models. That you only actually have a world model if you can predict, [00:07:00] given some action is taken, what is going to change in the world because of it. And in particular, that becomes hard over longer time scales. So if you're simply, trying to.[00:07:12] Predict the next video frame. That's not so difficult. But what you actually want to do is understand the consequences, likely consequences of actions minutes into the future. And to do that, you actually much more of an abstracted semantic model of the world.[00:07:32] The Bitter Lesson & Data Abstraction[00:07:32] swyx: Yeah, the question comes where you want to have more structure than is available in just predicting the next token.[00:07:41] And typically, well, let's, let's call it the experience of the last five years has been that is just washed away by scale, right? So what is the right middle ground here that, you don't ignore the bitter lesson, but also you. Can be more efficient than what we're doing today.[00:07:57] Chris Manning: One possibility [00:08:00] is, look, if we just collect masses and masses and masses and masses of video data, this problem will be solved.[00:08:11] Under certain assumptions that could be true, but there are sort of multiple avenues in which it could not be true. The first is what's really essential is understanding the, the consequences of actions producing an action conditioned world model. And if you are simply, collecting observational video data, which is the easy stuff to collect, when you're sort of mining online videos, you don't actually.[00:08:41] Know the actions that are being taken to see how the video is changing. And so if you are never collecting directly actions and you are having to try and infer them from what happened in the observed video, that's not impossible. But it's very [00:09:00] hard and it's not really established that you can get that to work at any scale yet.[00:09:05] And so there's a lot of premium on collecting action condition video data, which is part of why there's been a lot of interest in using simulation so that you can be collecting data where you do know the actions, which isn't quite limited supply, but there's also in the limit of as much data as you could possibly have.[00:09:28] Maybe the problem is eventually solvable, but. Even though we collect huge amounts of text data is always at a great level of abstraction, right? Language is a human designed, abstracted representation where there's meaning in each token and it's representing and abstraction of the world, right?[00:09:51] As soon as you are describing someone as a professor, and as soon as you are saying that they're condescending, right? These are very [00:10:00] abstracted descriptions of the world. It's not at what you're observing as pixel level, and to get to that kind of degree of abstraction, starting from pixels is orders and magnitude of extra data and processing.[00:10:14] And so, although, we absolutely want to exploit, get as much data as possible, use the bitter lesson. Nevertheless, if there are ways in which you can work with five orders of magnitude less data than people working purely from pixels, you're gonna be able to make a lot more progress, a lot more quickly.[00:10:34] And that's the bet here. And so you could just say that's only wanting to be able to, do it more efficiently, do it more quickly, do it more cheaply. But I think it's actually more than that, I think. One should be making the analogy to how human beings work at one level. You know? Yes, we have these high [00:11:00] resolution eyes and we can look and see a scene like a video, but all of the evidence from neuroscience and psychology is that most of what comes into people's eyes is never processed.[00:11:13] Right. That you are doing fairly fine ated processing of exactly what you're focusing on. But as soon as it's away from that of yeah, there's another guy over there that you've sort of only processing top down this very abstracted semantic description of the world around you. And so, that's what human beings are doing.[00:11:33] They're working with semantic abstractions and so. I think it is just the right representation. ‘cause we also have other goals we want to be able to do, real time worlds. So that means there's a limit to how much processing you can do and we want to do long-term planning and consistency. And again, that favors abstraction.[00:11:55] I mean, I guess there was actually a recent. Blog posts that [00:12:00] came out from our Friends of physical intelligence and, they were sort of heading in the same direction they were saying Oh, to the pay[00:12:06] swyx: pay model.[00:12:07] Chris Manning: Yeah. Yeah. To maintain a long term memory of what's happening in the world. So we can, do longer term we actually storing text of what is, been happening in the world.[00:12:19] Right. It is not such a successful strategy of trying to keep it all at a pixel level.[00:12:24] Vibhu: And yeah, I mean, you can see it in video models like that Temporal consistency. We're at a scale of train on, all the video data we have. We have it for maybe 30 seconds, a few minutes. That's not the same as a game state played for half an hour.[00:12:37] Right. I thought you guys break it down pretty well. You have a, you have a blog post about. Building multimodal worlds with an agent. I dunno if you guys wanna talk about this. This is one of the things I read, I[00:12:48] swyx: thought, yeah, it's the thing I talked about with the reasoning chain. Yeah.[00:12:51] Vibhu: So there's like different phases to this.[00:12:53] It seems like it's more of an agent, a scaffold, very different approach than just, type in a prompt and you, you don't have the same consistency. [00:13:00] It also, like, for people that are listening, I, I would highly recommend reading it. It breaks down the problem in a different light, right?[00:13:06] So like, what do you need to consider when you're talking about video, like world game models, right? How would, what do you need to consider? What are the factors? What are the elements? What's the state? So I don't know if you guys have stuff to talk about for this one.[00:13:19] Fan-yun Sun: Yeah. Actually, I wanted to add on a little bit Yeah.[00:13:22] On our previous point, which is just like, change topics so quickly. I, I do feel like sometimes people confuse like, oh, like we're taking an an, an method with abstraction. That means they don't believe in bitter lesson. Like that's just false, right? Like we are believed is a bitter lesson. But then I feel like the question that we always discuss is like, what is the right abstraction level today?[00:13:42] The analogy I like to make is like, let's just say we can encode and decode. Represent all of images, videos, audio and bytes. Then the most bitter lesson approached is to train a next byte prediction model as opposed to the next token prediction model where it's just like, okay, it's natively multimodal, can just, but it's like, yeah, like [00:14:00] to, to Chris's point, it's like the scale and computing you need to achieve that.[00:14:03] So that's why we always come back to like, okay, what is the most efficient way to do it? And reasoning models to the point of this blog post is a showcase of like, Hey, we're actually just like reasoning about the world and reasoning about. The aspects of the world that CAGR that matter for me to learn what I want to learn from this role model.[00:14:21] swyx: Yeah, it's like you're improving the en encoder of whatever you're, trying to model. And like a better representation would just represent the important things in less space. Yeah. Which would just be more efficient.[00:14:33] Fan-yun Sun: Yeah.[00:14:34] swyx: So yeah, I, I, I fully agree that it is not, antagonistic to, bitter lesson.[00:14:38] I do wanna wanna mention one more thing. Is there any philosophical differences with the JPA stuff that, Yun is working on? I gotta go there. You, you, you, you're, you're imagining like some latent abstraction. I'm like, okay, fine. Let's, let's talk about it, right? Like it's an elephant in the room.[00:14:52] Chris Manning: Yeah.[00:14:53] JEPA & Philosophical Differences with LeCun[00:14:53] Chris Manning: There are philosophical differences. Jan Lacoon is a dear friend of mine, but. [00:15:00] He has never appreciated the power of language in particular, or symbolic representations in general. Yarn is a very visual thinker. He always wants to claim that he thinks visually and there are no words, symbols, or math in his head.[00:15:21] Maybe that's true of yarn. It's certainly not the way I think. Um. But at any rate, the world according to yarn is the basic stuff of the, the world and of intelligence is visual and language is just. This low bit rate communication mechanism between humans and it doesn't have much other utility and it's far inferior to the high bit rate video, that comes into your eyes.[00:15:53] And I think he's fundamentally missing a number of important things [00:16:00] there. Think of this evolutionary argument looking at animals, right? That the closest analogies, the things with chimps, right? So chimpanzees, have fairly similar brains to human beings. They have great vision systems, they have great memory systems.[00:16:18] They've got, better memory than we do of short term memories. They can plan, they can build primitive tools that, humans. Massively ahead in what we understand about the world, what we can plan, what we can build. And essentially what took off for us was that humans managed to develop language and that gave a symbolic knowledge, representation, and reasoning level, which just, okay if this sort of vaulting of what could be done with the intelligence in brains.[00:16:59] So the [00:17:00] philosopher Dan de refers to language as a cognitive tool and argues that, humans unique among the creatures in the world have managed to build their own cognitive tools and language is the famous first example. But other things like, mathematics and programming languages are also cognitive tools.[00:17:21] They give you an ability to. Think in abstractions, in extended causal reasoning chains. And that allows you to do much more. And we use that for spatial representation and intelligence and planning and gameplay as well. So we believe, and this is, underlying the specific technologies that Moon Lake is making, that symbolic representations are powerful.[00:17:50] And you want to use that in your understanding of the visual world when you want a causal understanding, when you want to maintain long-term [00:18:00] consistency and prediction. And as I understand it, that's just not in ya Koon's worldview. So I think that's the fundamental philosophical difference. Then there's the specific model.[00:18:11] He's been advancing jpa, that's a reasonable. Research bed is a direction as to, to head for building out a model of the visual world. To my mind, it's sort of one reasonable research bed. It's not really established. It's the best one that everyone should be following,[00:18:32] swyx: at least developed at scale, at Meta.[00:18:34] But it's not just vision, right? Like, I mean, JPA is a, just joint admitting prediction can be applied to anything really. And people have done it. The argument is that there is a latent representation or that is probably more. Suited to the task, then why not let machines do it for us instead of predefining it at all?[00:18:50] And isn't something like a JPA shaped thing the right answer? And if not, why not?[00:18:55] Chris Manning: So I think there's a part of jpa that's right, which is [00:19:00] you do want to have a joint. Embedding that gives you a consistent model of the world. And Jan's argument is you can never get that from auto aggressive language models ‘cause they're sort of left to right churning out one token at a time.[00:19:22] I guess this is where we're the research arguments of the field, I'm not actually convinced that's right. ‘cause although the token production is this auto aggressive, process that's heading, left to right, I guess don't have to be left to right. But anyway, in sequence of tokens we could have right to left Arabic.[00:19:40] But although that's true, all of the weights of the model that are internal to the transformer, they are a joint model of the model's understanding of the world. And so I think you can think of the weights of the model as a form of. Joint representation, [00:20:00] and therefore it is plausible to think that could be the basis of a world model, which avoids, ya's objections.[00:20:10] swyx: I think I follow, and obviously that would touch on what Moon Lake eventually ends up doing as well. Right. Like, which it's hard to tell because you put out the end results, but we don't know the inputs that go into it. So it's, it's, that's something that we have to figure out over time.[00:20:25] Vibhu: Yeah. I mean, I guess this kind of breaks down some of the outputs. Do you wanna walk us through it?[00:20:31] Reasoning Traces & Interactive Worlds[00:20:31] Fan-yun Sun: Yeah. So this, this really just walks us through the reasoning traces of like, okay. So that just say, if we wanna build a world in this context, it's really just a game demo that, that shows the, the variety of interactions that this world model can build.[00:20:45] And yeah, it's really just a reasoning traces of like, okay it prompted to create a bowling game. Like how did it achieve what you saw? That level of causality, interaction and consistency, right? So yeah, this is almost just like a, an example of [00:21:00] like a reasoning traces. Very[00:21:01] swyx: detailed.[00:21:01] Fan-yun Sun: Yeah.[00:21:01] Vibhu: Very, very detailed.[00:21:02] You gotta you don't even realize it, right? Like when a video is generated, what happens when a ball strikes a pin, right? So first, like you, there's audio in that, like audio triggers happens, score increments, the world changes. Like pins have to start dropping. There's a timer that goes on. It's just like very similar to how now we're used to reasoning for language models.[00:21:20] There's a whole state of what happens. So geometry, physics, all this stuff. And then yeah, there's kind of that single prompt. So asset, ation all this stuff. It's like a, it's a nice view to see what's going on.[00:21:32] swyx: I think Sun is also too polite to point out that, both like Google's genie, demos as well as world Labs is marble, do not have interactive worlds.[00:21:41] Fan-yun Sun: That's the benefit of having a reasoning model, right? Like, because you can, you can say, oh, like maybe in this particular context, I want to learn how to bowl. And then you can say, okay, then what is it important when it comes to learning how to bowl? Okay, maybe it's like I need to understand the, the basic of like, physics and I want to throw it over [00:22:00] them.[00:22:00] I wanna know that when I, when it resets it's a new game. So I know that yeah, basically, you know to pick up the ball, you know that ball's gonna cause the pins to fall down. You know that what's important to this particular bowling game is to score and you know that the score corresponds to the number of pins that fell down.[00:22:19] So it's just like, if it's a model that sort of knows what it. Looks like, knows what a bowling game looks like, but doesn't actually allows you to practice over and over again and to understand that, oh, like what it takes to actually get a high score. Then it sort of doesn't actually allow you to learn what you set out to learn within the world model.[00:22:38] And I think this is really just one example of showing like the advantages of the approach that we're taking over most the, let's call it the zeitgeist, is today, when people talk about clinical role models,[00:22:51] Chris Manning: right? So it sort of seems like the question to ask when there's a world model is.[00:22:58] Can I not [00:23:00] only just wander around the world and look at the beautiful graphics, can I interact with the objects in the world and see the right consequences of actions?[00:23:11] Vibhu: And you also understand what the consequences would be if you do something right. So it's not just like, okay, there's one thing if I pick it up, something will happen.[00:23:19] But, there's 50 options and I know I can expect, I can infer what would happen if I do any of them. Right. So very different when you can actually see it play around with it.[00:23:28] swyx: There,[00:23:28] Beyond Unity: Cognitive Tools for World Building[00:23:31] swyx: there's two cheeky elements of that. I mean, the, the, the I guess, less ambitious one is, let's really establish for listeners, why is this fundamentally different than writing Unity code, right?[00:23:40] Like just creating a model to translate a prompt into Unity code[00:23:44] Fan-yun Sun: so there is an underlying physics engine. Yeah. In that sense, there's some overlapping things to Unity, but the way we think about it is like physics engine. Tools or code are cognitive tools like borrowing Chris's term, right? Like tools [00:24:00] that the model can employ as means to an end.[00:24:04] So today maybe you say, okay, in this particular context we care about physics, we care about the long-term causality consequences. Then yes, we deploy it, employ physics engine, and then maybe tomorrow we say, okay, we're we're training that. Just say drones where we only care about really fluid dynamics and the visual aspect of the world.[00:24:25] Then, then yeah, maybe we don't actually, the model actually doesn't have to use a physics engine. Or maybe it employs other types of representation or physics engine to achieve the task. So yes, writing code for Unity is sort of similar to a tool that our A model can employ, but our goal is for a model to take a representation conditioned reasoning.[00:24:46] Approach or process.[00:24:47] swyx: Yeah,[00:24:47] Fan-yun Sun: internally.[00:24:48] swyx: Yeah. Using these things as just like general two calls. Right. Which I think is very interesting. The other more ambitious one is, some kind of recursive element where it becomes multiplayer, right? Like here, there's a single player element, you're not [00:25:00] modeling any other people involved.[00:25:01] And that is a whole other thing.[00:25:04] Fan-yun Sun: But in fact, we can really do multiplayers. Oh yeah, okay. I haven't seen any double situations. So just actually just like prompt our, our model to say, Hey, like configure to multiplayer. Then it'll do like this. You'll be able to configure multiplayer[00:25:16] swyx: great[00:25:17] Fan-yun Sun: persistency database for you.[00:25:18] Easy. Yeah.[00:25:19] Vibhu: So what, what are like some of the current limitations in where we're at? So there's one approach of like, okay, scale up video predictors. Obviously there's data issues. With approaches like this, is it data constraints? What are like the next steps? Is it real time? Like, so there's one side of, write an agent to write Unity code, but okay, I want to be streaming a game real time.[00:25:38] I want to have characters being also like agent, but where, where do we kinda see this scaling up? Right?[00:25:44] Fan-yun Sun: Yeah, there's definitely a data constraint. Like the more data, the, the better. This reasoning model can almost basically act as humans to like operate a variety of tools and softwares to build whatever's necessary.[00:25:57] And then there's a sort [00:26:00] of fidelity constraint, which we're actually solving with another model, which we can talk about later. But it's like, it's not as easy to get to photorealism with the approach that we're taking. But we think there are better solutions to that, which is we can dive into later.[00:26:14] Later.[00:26:15] Vibhu: The one one thing you note here is it's a diffusion model, right? So there's, there's a few approaches, diffusion caution, splatting, yeah, so Ry diffusion model, you guys wanna[00:26:25] Fan-yun Sun: Yeah.[00:26:25] Vibhu: Introduce,[00:26:26] Fan-yun Sun: yeah, totally.[00:26:26] Rie: Neural Rendering & Skins for Worlds[00:26:26] Fan-yun Sun: So within our world modeling framework, we think there are two models that we train, right?[00:26:31] Like, there's the multimodal reasoning model that we just talked about that essentially handles. Mainly the, the causality, the persistency and logic determinism of the world. And then RY is our bet on saying, okay, like while all those model, can take care of all these things that we just talked about, it's limitations compared to existing, say, video models, is that it doesn't have as high of a pixel [00:27:00] ality right off the gate, right?[00:27:02] And EE is to say, Hey, we can actually take whatever persistent representation that we generate with our multimodal reasoning model and learn to restyle it into photo photorealistic styles or arbitrary styles you want. So this model is almost to say, Hey, I'm going to respect the persistency and interactivity of the world that you created, but my only job is to make sure that its pixel distribution is close to what we want.[00:27:29] Vibhu: Yeah.[00:27:30] swyx: Great example right there. You kept the KL divergence.[00:27:33] Fan-yun Sun: Oh. Where,[00:27:34] swyx: no, no. I mean this, this is a, a classic like, how you don't stray too far from the source material as you, you kept the kl, which is Oh yeah. Kind of cool. Yeah.[00:27:43] Fan-yun Sun: Yeah.[00:27:44] swyx: I mean, and the[00:27:44] Chris Manning: difference is, and I mean sun was pointing at this, where sort of saying it's in one way a more difficult path, but a better path that, typically the diffusion models are producing the whole scene and it looks lovely, [00:28:00] but there isn't spatial understanding behind it, which is allowing for the real time graphics gameplay, the spatial intelligence, understanding the consequences of worlds where this is, taking a path where it is assuming an abstracted semantic model of the world's state.[00:28:20] And then the diffusion model is then being used on top of that to produce the high quality graphics.[00:28:27] swyx: Is there an intended practical, or business use for this, or is it like a, like a demonstration of capabilities?[00:28:34] Fan-yun Sun: We actually believe that this is gonna be the next paradigm of rendering. So it's gonna replace how ra raizer, it's gonna replace DLSS today because it not only has these pixel prior that's learned from the world such that you can literally play any game in photo realistic styles, which is a lot of people's desire when they do GTA, right?[00:28:51] Like,[00:28:51] Vibhu: all the mods, all the people adding perfect lighting and all this.[00:28:54] swyx: So[00:28:54] Fan-yun Sun: skins[00:28:55] swyx: for worlds, let's call it[00:28:56] Fan-yun Sun: skins, let's call it skin for worlds. I,[00:28:58] Vibhu: it's also like, you can call it skin, you can call it [00:29:00] customization. You can play it how you want, right?[00:29:01] Fan-yun Sun: Yeah, exactly. And I think another thing that we really pointed out specific specifically in this blog is the programmability of it, right?[00:29:09] So what this means is that this render historically render is always a derivative of the game state, right? You're saying, oh, here's the game state, I'm rendering out a frame. But here I'm saying actually this render can be part of the gameplay loop. I can say something along the lines of, if upon getting 10.[00:29:26] Apples, I'm gonna, my weapon of choice, my bullet's gonna turn into apples. And that's, that's possible because we can say, we can basically dynamically have certain game state trigger the, the preconditions to the render such that the rendering is now part of the game loop too. One thing is to just say, okay, it's, it's, it's the appearance.[00:29:47] But the second thing is also to say there's these novel interactions that are possible because this render now has actually priors of the world.[00:29:57] swyx: It is up to the artist to figure out what to do with it.[00:29:59] Fan-yun Sun: It [00:30:00] is up to the creators. Yes.[00:30:01] swyx: Yeah.[00:30:01] Fan-yun Sun: And I also think that's actually another big argument that we're making and the reason that we're picking, taking the bet we're baking is that a lot of the times, whether it's for embody AI gaming, like you want a layer where human can inject their intentions.[00:30:15] So, for example, let's just say in the context of gaming, it's obviously like my creative intent, but maybe in the context of embodied ai, it's like, oh, like I take this foundational policy and I want to actually fine tune it to deploy in my house. So you want to almost say, inject, have a layer where human can say, oh, here's the distribution of things I want to create to achieve my goal.[00:30:35] And I think 3D graphics as it as it is today, is basic, the layer for people to say, Hey, what do I care about in this world? And it allows, basically human intent to be expressed in these worlds much more explicitly and distributionally as opposed to just saying, Hey, I'm gonna generate like, arbitrary.[00:30:54] And it's like just prompts,[00:30:55] swyx: it's one of those things where like, I think you, you're going to build up a series of models, right? [00:31:00] This is just one of, this is probably like the highest utility or heaviest, frequency one, I don't dunno what to call this. Where like you Yeah. You can immediately drop this in on any game and you don't need anything else that.[00:31:10] That you guys do. But, I, I could see, I could see that I think the, the human intent is something that people are not even used to because we're so used to static worlds or, worlds that just don't react, or, I don't know. It's, it, you're kind of blowing my mind right now with like, I'm, I wonder if you've talked to people at GDC Hmm.[00:31:27] And what are they gonna do with it?[00:31:30] Fan-yun Sun: Yeah. Now the stance that we take on this front is like, we're not gonna be more creative than our users to ship[00:31:35] swyx: it out.[00:31:35] Fan-yun Sun: Yeah. But we wanna make sure that we're building things in a way that really allows them to express their intent.[00:31:41] swyx: The thing that you said about, here's the distribution that I want.[00:31:45] I think text may be too low of a bandwidth to. To really demonstrate, because I, I, there, I'm, I'm probably just gonna want to drop in a bunch of, reference assets and then you can figure it out from[00:31:58] Vibhu: there. But you probably wanna do a, a mixture of [00:32:00] both, right? Like you throw in a few images. I wanted this style.[00:32:02] Yeah. I want it to look like this. So it, it's, it's a mixture, right?[00:32:05] Chris Manning: I, I think it's a mixture. I mean, yeah, I mean there's clearly a visual component of this, and it's not that, everything can be text. ‘cause of course you want to give a visual look, but there's also a massive amount of giving the overall picture of the look of the world and the behavior of things that you can express in a few words of text.[00:32:32] And it be very time consuming and difficult to do via visual means. So I think, yeah, you want a combination of both.[00:32:40] Evaluating World Models[00:32:40] Vibhu: So one question I kind of have is, how do we go about evaluating world models? So like, there's many axes, right? One is like, okay. I have preferences. How well do we adhere to prompts? One is the simulation.[00:32:50] One is like do things, is there core logic that's broken? So coming from we know how to evaluate diffusion, there's fidelity, there's [00:33:00] stuff like that. But what are some of the challenges that most people probably aren't thinking about?[00:33:04] Fan-yun Sun: Yeah, I think this is like a great question and probably one of the hardest questions in role models because like, I think it always comes back to what are you building this role model for?[00:33:13] And depending on your end goal and purpose, the evaluation should defer. So in the context of games, then the most direct way of measuring is how much behind are people actually spending in this world that you create? And if your goal is to say, for example, in the context that we just talked about, like, hey, deploying, deploying action in body, a agent, then your, your end.[00:33:33] Metric is then, okay, after training in these worlds that you generate how robust it is to when you actually deploy to the target environment. But then, it's, it's hard to measure these end metrics. So today people have like these proxy metrics that I call that basically try to measure what we really care about, which is the end metrics, but then frankly it's different for every use case.[00:33:57] Yeah,[00:33:57] Vibhu: which seems like quite a challenge, right? Like in [00:34:00] in language models or video models. Image models, your benchmarks are proxies, right? People aren't actually asking instruction, following tool use questions. They're proxies of how well it will do downstream. But for this, so like, should teams, should companies have their own individual benchmarks outside of games?[00:34:16] If you think of stuff like, okay, video production, movies, stuff like that, that also want to use world models. Should, should they sort of internalize like. Their own proxy. Is this something you guys do? Where, where does that connect[00:34:28] Chris Manning: go? Yeah, I think this whole space is extremely difficult as things are emerging now.[00:34:35] And I mean, it's not only for world models, I think it's for everything including text-based models, right? ‘cause in the early days it seemed very easy to have good benchmarks ‘cause we could do things like question answering benchmarks and could you answer the question based on these documents and the various other kinds of, do pieces of logical reasoning or math.[00:34:58] But again, these are sort of. [00:35:00] And there were sort of visual equivalents of things like object recognition, right? For these small component tasks. These days so much of what people are wanting to do also with language models is nothing like that, right? You're wanting to, have an interaction with the language model and get some recommendations about which backpack would be best for you for your trip in Europe next month.[00:35:25] And it's not the same kind of thing, right? And it's not so easy to come up with a benchmark as to does this large language model give you an effective interaction for guiding you in a good way for shopping, right? So, and it's the same problem with these world models. So if we take the game design case, well success is that a game designer can.[00:35:57] Produce what they are [00:36:00] imagining in a reasonable amount of time. And that's really the kind of macro task. That's a very hard thing to turn into a benchmark and I think a lot of this is actually going to turn into people walking, walking with their feet. Right? I mean, I guess that's what's happening, at the large language model level, right?[00:36:23] When people are choosing to use, GPT five or Gemini or clawed, individuals are trying out these different models and deciding, oh, I like the kind of answers that GT five gives me, or no, I feel like I get more accurate detail from Claude, right?[00:36:43] Vibhu: It's a lot of[00:36:43] Chris Manning: vitech, a lot of people just using it.[00:36:45] It's vibe checking. I realize that, but it's actually whether. People feel it's giving them utility in what they want. Right.[00:36:52] Vibhu: And the the interesting thing there is like a lot of people prefer the visual, right? This looks pretty, which is not the objective of what this is [00:37:00] for, right? It's if a, if a game designer is working on something, they care about the game engine, right?[00:37:04] The state, it's, it can look whatever. You can fix that up later. Or you can have a really good game state and you can quickly edit it to 20. 20 different versions, like Keep State,[00:37:14] Chris Manning: right?[00:37:14] Vibhu: So[00:37:14] Chris Manning: that's a really important distinction, for and for speaking to Moon Lake strength, right? So, yeah, great visuals are lovely to look at for a few seconds, but gains are really all about the concept, the game play.[00:37:33] And a lot of the time that doesn't actually even require great visuals. I mean, there are just lots of very successful games which have relatively primitive visuals, and there are other games where people have spent millions producing photo realistic, visuals, and the game sucks, right? So, keeping those two axes apart is really important in thinking about what's important in a [00:38:00] world model for different uses.[00:38:02] swyx: This conversation is reminding me of some game review and fiction discussions I've, had in my sort of non-AI related life. Some, for some people might know Brandon Sanderson, who's a very famous, fiction author, had, is is a big game reviewer. And he, he's a big fan of video games where you change one thing about a normal what you might assume about, about the world.[00:38:22] For example, Baba is you, I don't know if you might have come across that, where like the rules change as you play the game. And also like where, you can do things like reverse time selectively or like change gravity selectively. And I think this is also reminds, reminds me of other kinds of world models that are created by authors.[00:38:38] Where Ted Chang is, is my typical example where he'll take the world that, you know today, but change one thing about it and, but then create a consistent world based on that. Which is long-winded answer of me to, of. For me to say is it's it easy to create alternative roles that don't exist, but you change one thing and then let's, let's run a whole bunch of people through it to see if it works.[00:38:58] Chris Manning: My first dance will [00:39:00] be, that seems a lot easier and more conceivable to do using Techn technology like Moon Lakes than with some of the other world models out there, where the sun can actually make it happen. I'll let him give a second answer.[00:39:15] swyx: If I guess for you, you're constrained by the game engine tool, right?[00:39:18] Like at the end of the day, that's the, that's the thought, partner that you have. If I ask for something where like, if it never is allowed to reverse time or if gravity only ever works one way, then well that's it. But sometimes gravity might change,[00:39:33] Fan-yun Sun: but it's a lot easier to change with code as opposed to a model that is learned primarily on data of.[00:39:42] Real world and virtual worlds that are, I guess, like for example, junior, like there's actually trained on a lot of real world data and a lot of virtual gaming data, and it's hard to say maybe it's easier to say, okay, I wanna change the visuals in like the time period of, of the world. Like, you can't change gravity, for [00:40:00] example.[00:40:00] Vibhu: I feel like you can to light bounds, right? Everything comes down to like, code is a better way to execute it, but the models aren't that diverse and creative, right? You can say, okay, make gravity slower. It can do that, but it's limited to your representation of how you text it out, right? Like they're, they're only gonna do a few iterations, whereas programmatically, if there's a game engine under the hood, you can kind of go wild, right?[00:40:22] So one of the, I dunno, one of the limitations of most models is that they're very overtrained to one style. Right. And extracting diversity is pretty difficult. At least that's something we've seen.[00:40:35] Fan-yun Sun: I mean, are there examples you have in mind where you Existing models? Yeah. Like it would be easier to do that's not using code.[00:40:43] Certain types of creative intent or like transition state transitions,[00:40:47] swyx: Clipping, other models, other wo models are very good at clipping through things. Clipping my, my, my legs clipping through a rock because it's, it's just, it's just bad. [00:41:00] Like, you would have to struggle very hard with your stuff to actually make that happen.[00:41:04] Which I think is maybe a topic that you actually prepared on, Gian Splatting versus, the other stuff.[00:41:09] Vibhu: Yeah. Yeah. It's just for those not super familiar, right? There's a, there's gian splatting, there is diffusion. Like what works, what scales up. I feel like in February when Soro one came out the blog post was literally titled like,[00:41:21] swyx: you bring it up.[00:41:22] You never know.[00:41:23] Vibhu: World, world, video generation models are world simulators. It's super bitter lesson pilled. Yeah, emer, a lot of it is emergence, right? So, not to go through their blog post, basically their whole thing was as you scale up all this consistency, all this stuff just kind of solves, it's a very simple premise, right?[00:41:41] They just scaled up, diffusion, and from there, this is, this is Feb 2024, how much can we, it's already been two years, which is basically five years. How much more in AI time do we need to just scale up or, or do we hit a data cap? But I think we already talked about this a lot, right? Like this is back to the beginning discussion of what's [00:42:00] appropriate for the time.[00:42:01] And that seems like your approach, right?[00:42:03] Fan-yun Sun: Yeah. The point I'm trying to make is that they're very many, many different types of world simulators and like having a world simulator that can produce pixel coherency is very, very useful for games and, marketing and all these things, but it's not as useful as people think when it comes to causal reasoning.[00:42:25] When it comes to embodied ai. Yeah, like it this title is true. We're not saying that it's, it's like, not a great world simulator, but actually in the blog that we, we, we, we wrote, the bet is more so that there are gonna be disproportionately large share of value of real world tasks or, and virtual tasks where high resolution pixel fidelity is not needed.[00:42:47] Yes. Video models have their values.[00:42:50] swyx: Yeah. This is at the absolute limit of my physics understanding, but one example that comes to mind is basically having to solve like ba the equivalent of a three [00:43:00] body problem in a deterministic Well, where the video models, which is approximated good enough. Yeah.[00:43:08] Right. Like there's, there's some point at which your approach kind of runs into like the you now have to simulate the world. Please, thank you very much. And like you're trying to do that, but only to the extent that the game engine lets you and like game engines cannot do some things.[00:43:23] Fan-yun Sun: Yeah, no, I mean, I think the interesting or more technical question here actually is where do you draw the boundary between.[00:43:32] What's handled with, let's say, diffusion prior and what, when? What's handled with symbolic priors?[00:43:38] swyx: Yes.[00:43:38] Fan-yun Sun: Okay.[00:43:38] swyx: Okay.[00:43:39] Fan-yun Sun: Right. Let's go there. Because this, this boundary can actually be fluid. Like I think like maybe what you're trying to get at is like, okay, people are saying pixel prior, everything. But what we're saying is, okay, there's a boundary that we draw where this is where we think provides the most economical value for the domains and things that we care about today.[00:43:59] [00:44:00] And I actually do think, and it's something that we do internally all the time, which is like, okay, given new equations that we learn or new elements of the world and that we, we learn, or maybe some other knowledge that we acquire in the process of developing the models. Should we still be maintaining this line exactly as it is today?[00:44:22] Or should we move it a little bit left or a little bit right? Right. Like sometimes that we realize that, oh, like maybe customers or, or folks like want certain things that are better handled with preop pryor as opposed to, symbolic prior than,[00:44:34] swyx: yeah. Your, your skin thing is a, is a example moving it, right.[00:44:37] Yeah.[00:44:37] Or left. Yeah,[00:44:37] Fan-yun Sun: exactly.[00:44:38] swyx: I dunno what the, the left right is.[00:44:39] Fan-yun Sun: Yeah, yeah, yeah. No the, the model.[00:44:42] swyx: Yes.[00:44:42] Fan-yun Sun: Actually we have a few iterations of them. They're actually at slightly different[00:44:45] swyx: I know boundaries. You should, you should do that. That's a cool dimension to show.[00:44:49] Fan-yun Sun: Yeah.[00:44:50] swyx: Is quantum mechanics the diffusion prior of our world?[00:44:55] Right. It's like that's the boundary of classical mechanics versus quantum. Right? Like, that's it. At one [00:45:00] point God plays dice and the other point doesn't.[00:45:02] Fan-yun Sun: I dunno if Chris, you wanna say it, but I think, I think generally I feel like physics is better with symbol P priors.[00:45:08] Chris Manning: Even quantum physics.[00:45:09] Fan-yun Sun: Even quantum physics.[00:45:11] swyx: Yeah. This is starts against to, MLST territory is, is what I call it, where, he, he likes to get philosophical. We, we we're quite friendly.[00:45:18] Vibhu: I mean, we need to get, we need to get singularity. I heard some of that.[00:45:23] swyx: No, no, I think that is actually really helpful and man, I just want you to productize this like, as a product guy, I'm just like, oh, also[00:45:32] Vibhu: a gamer, I[00:45:33] swyx: wanna, it's like a researcher, like, it's cool.[00:45:35] Like this is a, the theoretical, like you have a very good, I don't know, like the way of thinking about these things, but I just wanna see you like, express it. I do think like your fundamentally things when, when you leave open new tools, like, okay, use, use human intent to incorporate it into how you render.[00:45:52] Artists are gonna have to take like two to three years to figure out what to do with this. And you just don't know.[00:45:57] Chris Manning: Right. But I think, this is, [00:46:00] gives a much more approachable and controllable world for the society, which is the beauty, the beauty of, NLP, that that will enable it to be adopted and used.[00:46:10] And we are very hopeful about that. Yeah,[00:46:13] Fan-yun Sun: yeah. Yeah. I mean, we are, we are very focused actually on commercialization in the sense that like we do, we do really believe in the data flywheel app approach. Yeah. Where, we put this in the hands of the creators and the users and then they will teach us when, what capability our model should improve.[00:46:27] And that's why we are, we are actually, like products and beta[00:46:31] swyx: Yeah. Focusing on gaming. What, what's like the adjacent thing to gaming[00:46:34] Fan-yun Sun: embody adjacent, basically. So maybe we can, we can I'll maybe start with where we see the platform in three years. Yeah. Which is like, okay. The users would tell us what they want to achieve.[00:46:45] The end goal could be, Hey, I just, I wanna make something to teach my kids the value of humility. Or it could be, Hey, I wanna fine tune my, drones to be really good at rescue situations. I could be vacuum robots. I want to like train [00:47:00] my manipulation or like vacuum robot to be very robust to my office, right?[00:47:04] But it's like, whatever it is, scenario robust to[00:47:06] swyx: my office[00:47:07] Fan-yun Sun: or like navigate very robustly in my office. But then it's like, whatever end goal that you want, our role model will say, okay, given what you want to achieve, let me generate a distribution of environments such that I can train and evaluate whatever it is you want.[00:47:24] Yeah. Right. Maybe for the purpose of games, it's just the end simulation and that's the end product for certain policies. It's like I can train it within these environments and then help you see where your policy is failing or not. Yeah. And then, so I think,[00:47:37] swyx: so in that case, much more of a training tool.[00:47:40] Than in other training[00:47:41] Vibhu: evaluation? Both. Right?[00:47:43] swyx: Sure. Same. Same thing.[00:47:43] Fan-yun Sun: Yeah, same thing. I think it's just this role model that allows people to train any policy that can act in any multimodal environments.[00:47:51] swyx: Would it be harder to reward hack? Is there an angle here where it is harder to reward hack? Like it's just, I'll just put it generally because I think that's a, that's obviously a key [00:48:00] problem that a lot of people face when in training agents in these environments, and I don't know, can you solve it?[00:48:07] Chris Manning: I think not necessarily. To the extent that there's a mis specified reward that. It seems like it could be hacked in a more symbolic world or in a more pixel based world. I dunno if Sun's got any thoughts, but I don't think that's really being solved.[00:48:26] swyx: The other thing that comes to mind is just you could just build a better sawa as a video generator model, right?[00:48:31] Because then you, you would move the diffusion, side a bit more further to the right. I think if I got the directionality correct. And that's it.[00:48:40] Vibhu: It's better on domains, right? Like on consistency over now, or for sure it exists versus something doesn't, right.[00:48:46] Chris Manning: So[00:48:46] swyx: yeah. Yeah. Is[00:48:49] Vibhu: is a question more like, like[00:48:51] swyx: I'm just riffing on like, how do you, what can you build, you know?[00:48:54] Oh, with the stuff that you have. I do think that the minor, the academic does go immediately to training [00:49:00] and in eval evaluation, but like art tends to take unusual directions. Like you might end up,[00:49:06] Chris Manning: okay. Yeah. But the question is, can you use this piece of software to develop compelling gameplay and. I don't think you can take SOAR and produce compelling gameplay, right?[00:49:19] If you want to have a world that you can wander around in a bit, you are good. But what are your abilities to have gameplay mechanics implemented the way you'd like them to be and to have things stay, with the long-term history of your gameplay that influences future actions. I think there's just nothing there for that.[00:49:39] swyx: Yeah, I do tend to agree. I, I'm just trying to sort of test the boundaries. I would also make the observation that as AAA games industry has developed the line between what is a movie and what is a game has blurred. And you, you, you do end up basically producing a two hour movie as part of your game.[00:49:57] Fan-yun Sun: No, honestly, there, there's so many actually [00:50:00] applications in adjacent markets that our world model can go into. Yeah. But yeah, it, it's sort of fun to riff, riff on. Although on the execution side, we we, we need to stay focused with like, okay, what are the capabilities we want to unlock over time?[00:50:11] And there's a roadmap for that. But yeah, if we're just riffing on sort of like the possibilities, I feel like, whether it's endless Yeah, it's like classic[00:50:18] swyx: and the embedding for a possibility and endless in my mind, it's very close. Yeah. I do wanna, focus on one, like weird choice. I, I don't know if it's weird.[00:50:28] Maybe I'm, I got something here. Audio, right? You could have just said no audio And audio in my mind has a lot of recursion, whereas in video you can just do recasting and that's much computationally much simpler. Audio just seems way harder. I don't know if you wanna just comment on just the special 3D audio.[00:50:46] Problem. Did you really have to do it? I guess you do to be immersive, but like a lot of people do treat it as like, well, you just stick a, a tt S model on top of[00:50:57] Vibhu: Well, there's a lot more to game audio than [00:51:00] just speech. Right. It's not just[00:51:01] swyx: tts. Yeah. Tts. S Fxt, GM Spatial in my mind Echoes[00:51:06] Chris Manning: Yeah.[00:51:06] swyx: And reflections.[00:51:07] And I, I don't even know what's, what else? I don't know what, what other problems in this space.[00:51:13] Fan-yun Sun: Yeah, I think this point like the, it's sort of a more, more pointing to the benefits of using an game engine as a tool that's available to the model, right? Because like part of the spatial audio is from the code that is underlying the simulation.[00:51:32] And while we do give our model access to other types of audio models as. Tools.[00:51:39] swyx: None of them would be spatial, I think.[00:51:41] Fan-yun Sun: But that's exactly sort of more 0.2. We're giving our model an abstraction or a suite of tools such that it's able to achieve that. And you can argue that sort of spatial is like a, like a emergence out of the, the tools that we and abstraction that we provide to the agents.[00:51:59] And I think that's the beauty of [00:52:00] this, this, this approach is like there's a lot of things kind of like how human's built technology and they're like Lego blocks that build on top of each other. And it's the same thing here. There's gonna be things that sort of just sort of emerges from being able to put these things together in like combinatorially interesting ways,[00:52:14] Chris Manning: right?[00:52:15] So this integrated audio model exploits the understanding and semantics of the Moon Lake world, right? And whereas in general for the Gen AI video models. There's no actual integration across to audio at all, right? That someone might stick some music or stick a soundscape or whatever else on top of their video.[00:52:44] So it's not a silent video, but they're in no way connected into a consistent world model. And there's nothing that's okay. An action is happening in the video. Therefore there should be a sound that's [00:53:00] coming from this part of the visual field.[00:53:03] swyx: Yeah.[00:53:03] Vibhu: Is that different than Sora too? Does it not have audio?[00:53:06] Not to say it's not like[00:53:08] swyx: amazing[00:53:08] Vibhu: isn't a spatial[00:53:09] swyx: audio.[00:53:09] Vibhu: It doesn't,[00:53:10] swyx: no. I've played around it with it enough. It just sounds like someone put an 11 laps voice on top of it and just tried to do the lip sync.[00:53:18] Vibhu: Oh, yeah. I've seen, okay. Generate a dog at the beach and reactions to big wave and move[00:53:23] swyx: around.[00:53:23] It's definitely like, so have the dog, have the dog move away from camera and see if the, the song goes down. It doesn't. ‘Cause they don't have facial audio.[00:53:32] Fan-yun Sun: We do want to basically like we, our moral model, like the one we're training is basically towards the goal of having a combined latent representation across all these different modalities.[00:53:42] Right? Such that it can like reason across these different modalities. So for example, if I close my eyes and like you play a video, you play a sound of like a car skidding away from me. I almost can like, visually extrapolate that trajectory in my mind. And I think that type of capability, we want our model to be able to reason, right?[00:53:59] And that's the reason that [00:54:00] we're sort of taking this multimodal reasoning approach. It's like we want this combine late in space that can[00:54:05] swyx: Yeah. Oh, you said late in space. We like that. Here we have to play the, the bell Every time that someone says late in space, no, you gotta train daredevil one. Where you, you, you, it's only audio, but you have to work out.[00:54:15] Where everything is.[00:54:19] Cool. I I think that that was, that was about it for our Moon Lake coverage. I do think that we have like a couple of, Chris Madden questions on, on IR and, just any, any other sort of attention topics or n NLP topics.[00:54:31] Vibhu: Okay.[00:54:31] swyx: Go ahead.[00:54:32] Chris Manning's Journey: From NLP to World Models[00:54:32] Vibhu: Well, no, I mean, yeah, it's just fun. We talked a bit about how you guys met, but you basically, you, you were like the godfather of NLP per se, right?[00:54:39] You spent the whole career from early embeddings, early early attention. You did 2015 attention for machine translation, everything. You, you had information retrieval, so RAG before rag, we just wanna shout that out and admire a lot of that. Right? So what prompted the switch over to world models?[00:54:56] How, how'd all that come about?[00:54:58] Chris Manning: To some answer it [00:55:00] is, the enthusiasms and creativity of students, but there's a bit of a history there, right? So, yeah. So clearly most of my career has been doing stuff with language and how I got into research was thinking, ah, this is just so amazing how humans can produce speech and understand each other in real time.[00:55:21] And somehow they managed to learn languages from their kids. How could this possibly happen? And so, yeah, starting off I was very focused on language, but as it sort of got into the 2000 and tens, I started, going, I'd been working on question answering, and then I started to get, interest in visual question answering.[00:55:42] And that was an area where it was very noticeable. That the visual understanding was bad. Right. These were the days when like, it sort of seemed like there's almost no visual [00:56:00] understanding. You were just getting answers that came from priors. So, if you asked how many people are sitting at the table, it'd always answer two regardless of how many, how many people you could see in the picture.[00:56:11] And so it seemed like, oh, these models actually aren't able to get semantic information outta

Send us Fan MailDeep Learning Can Predict the Overall Survival of Cervical Cancer Based on Histopathological Image, Gene Mutation and Clinical Information. Shen J, Miao Z, Wang L, et al. IET Systems Biology 2026.Episode Summary: In this deep dive, we explore a groundbreaking 2026 study that uses multimodal deep learning to act as a "master diagnostician" for cervical cancer. We examine what happens when an AI is fed a combination of standard clinical data, cutting-edge genetic sequencing, and century-old H&E tissue slides. The results force us to rethink how cancer operates: what happens when the genetic "blueprint" of a tumor lies to us, and the real biological truth is hiding in the seemingly chaotic pink and purple pixels of the connective tissue?In This Episode, We Cover:The Murky Diagnostics of Oncology: Understanding why predicting an individual patient's overall survival (OS) in cervical cancer is profoundly difficult. Getting this prediction wrong means risking either lethal undertreatment (distant metastasis) or subjecting stable patients to devastating overtreatment toxicities.The Three Modalities (The Suspect, The DNA, and The Security Footage):Clinical Data: The "suspect's description," utilizing standard patient metrics like age and tumor stage.Molecular Data: The genetic "blueprint" and somatic gene mutations. The AI isolated major red flags like RGR, DBN1, and CALCR mutations, which drive metastasis and signal poor prognosis.Histopathological Images (H&E): The "security footage" showing the physical tissue battlefield via whole slide images.The Model Showdown: Researchers trained a deep learning model (ResNet18) and fused these modalities using Multimodal Compact Bilinear (MCB) fusion. The AI was tasked with classifying patients into short-term (under 3 years) or long-term (over 3 years) survival, and it was rigorously validated on a completely independent dataset (PUMCH) to ensure generalizability.Round 1 - The Genetic Curveball: Despite being the cell's source code, genetic mutation data was the absolute worst predictor of survival, achieving an AUC of just 0.559. Adding it to the AI actually caused the "curse of dimensionality," making the model worse by overwhelming it with mathematical noise.Round 2 - The AI's "Aha!" Moment: The tissue phenotype dictates what actually happens. Fusing simple clinical data (age) with H&E images achieved a highly accurate 0.783 AUC. Even more shockingly, for aggressive short-term survival cases, the AI didn't focus heavily on the tumor itself. It looked at the stroma (connective tissue), deducing on its own that the host's inflammatory battleground dictates the lethality of the disease.The Future of the Lab: How automated quality control (HistoQC) and mathematical techniques (Macenko color normalization) strip away lab technician error and chemical dye variations. We also look ahead to how hyperspectral imaging might soon reveal the foundational chemical signatures of living cells.Key Takeaway: Throwing more data at an algorithm isn't always better. By successfully extracting profound biological truths from routine, inexpensive H&E slides, the AI proved that we don't necessarily need $1,000 genomic sequencing panels to accurately predict prognosis. The physical manifestation of the tumor microenvironment tells us exactly who is winning the battle, paving the way for accessible precision medicineSupport the showGet the "Digital Pathology 101" FREE E-book and join us!

In this episode of the PFC Podcast, Dennis and a panel of experts discuss the intricacies of facial trauma management, focusing on the use of anesthesia and nerve blocks. They explore various techniques for achieving effective pain control in trauma situations, emphasizing the importance of understanding facial anatomy and the application of dental blocks beyond traditional uses. The conversation also highlights the significance of multimodal pain management strategies and the role of cross-training in enhancing trauma care skills.TakeawaysFacial blocks can be used for more than just dental procedures.Understanding the anatomy of facial nerves is crucial for effective anesthesia.The infraorbital block is essential for mid-face trauma management.Lidocaine with epinephrine can provide longer-lasting anesthesia in vascular areas.The mandibular nerve requires precise techniques for effective anesthesia.Ring blocks are effective for ear and nose trauma.X-Brow is a long-acting anesthetic that can reduce narcotic use post-surgery.Cross-training with dental professionals can enhance trauma care skills.Multimodal pain management is key in treating facial injuries.Effective pain control can significantly improve patient outcomes in trauma situations.Chapters00:00 Introduction to Facial Trauma and Anesthesia03:37 Understanding Facial Blocks and Their Applications10:31 Anatomy of Facial Nerves and Block Techniques24:32 Mandibular Nerve Considerations and Techniques40:34 Special Considerations for Facial Injuries54:49 Multimodal Pain Management in Facial TraumaFor more content, go to ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠www.prolongedfieldcare.org⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Consider supporting us: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠patreon.com/ProlongedFieldCareCollective⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ or ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠www.lobocoffeeco.com/product-page/prolonged-field-care

In the electric utility industry, lineworkers often retire with years of experience and take that knowledge with them. Eversource discovered a way to bridge the knowledge gap by overhauling its electric and gas training using multimodal digital tools, e-books and mobile apps. This episode of the Line Life Podcast features a narrated version of the article "Eversource Advances Workforce Training" from the November 2025 field-focused Electric Utility Operations section of T&D World magazine.  This audio story outlines the utility's plan to deploy roughly 85 custom training applications by 2026–27 across core programs like overhead and underground electric, substations, gas maintenance, corrosion and field communications. The episode highlights how tablet-delivered, interactive content preserves institutional knowledge, supports different learning styles and improves safety through virtual practice. It also provides scalable, just-in-time training to meet growing workforce demands and close the experience gap. For more information, read the article on T&D World's website. 

Karan Singhal, Head of Health AI at OpenAI, explains how ChatGPT Health is achieving attending-physician-level performance and already serving hundreds of millions of users. He details how OpenAI works with over 250 doctors, built the 49,000-criteria HealthBench evaluation, and ran one of the first randomized trials of AI copilots in clinical care. The conversation explores privacy and safety safeguards, medical multimodality, N-of-1 treatment plans, and how AI could become a standard part of global medical practice. Use the Granola Recipe Nathan relies on to identify blind spots across conversations, AI research, and decisions: https://bit.ly/granolablindspot LINKS: modeling human wellness Sponsors: Claude: Claude is the AI collaborator that understands your entire workflow, from drafting and research to coding and complex problem-solving. Start tackling bigger problems with Claude and unlock Claude Pro's full capabilities at https://claude.ai/tcr Serval: Serval uses AI-powered automations to cut IT help desk tickets by more than 50%, freeing your team from repetitive tasks like password resets and onboarding. Book your free pilot and guarantee 50% help desk automation by week 4 at https://serval.com/cognitive Framer: Framer is an enterprise-grade website builder that lets business teams design, launch, and optimize their.com with AI-powered wireframing, real-time collaboration, and built-in analytics. Start building for free and get 30% off a Framer Pro annual plan at https://framer.com/cognitive Tasklet: Tasklet is an AI agent that automates your work 24/7; just describe what you want in plain English and it gets the job done. Try it for free and use code COGREV for 50% off your first month at https://tasklet.ai CHAPTERS: (00:00) About the Episode (06:11) Cancer story and mission (11:46) Designing safe health AI (Part 1) (17:49) Sponsors: Claude | Serval (21:09) Designing safe health AI (Part 2) (26:48) Uncertainty, HealthBench and robustness (Part 1) (30:23) Sponsors: Framer | Tasklet (32:50) Uncertainty, HealthBench and robustness (Part 2) (38:11) Chain-of-thought and evaluation (46:49) Real-world performance and frontiers (55:35) Multimodal data and science (01:05:36) Personalization, privacy and monitoring (01:15:47) Models, data and incentives (01:29:31) Doctor adoption and workflows (01:38:13) Scalable oversight and alignment (01:51:06) Move 37 and future (02:00:50) Episode Outro (02:03:06) Outro PRODUCED BY: https://aipodcast.ing

The integration of Artificial Intelligence (AI) into post-injury rehabilitation is transforming recovery paradigms by enabling personalized, adaptive, and efficient rehabilitation pathways tailored to individual patient needs. This podcast reviews the current advances in AI applications that facilitate assessment, monitoring, and optimization of rehabilitation programs following injuries. Through machine learning algorithms, wearable sensors, and predictive analytics, AI enhances the precision of therapy plans, tracks patient progress in real-time, and predicts recovery trajectories. The discussion includes the benefits of AI-driven rehabilitation, including improved functional outcomes, reduced recovery times, and increased patient engagement. It also addresses challenges such as data privacy, algorithmic bias, and integration with clinical workflows.  1. Transforming recovery paradigms Traditional post‑injury rehab relies on periodic in‑person assessments, therapist intuition, and standardized protocols that only partially account for individual variability. AI is shifting this model toward: Continuous, data‑driven care: Instead of snapshots in clinic, rehab can be informed by near real‑time streams of kinematic, physiological, and behavioral data from wearables, smart devices, and robot interfaces. Dynamic adaptation: Therapy intensity, task difficulty, and exercise selection can be automatically adjusted based on ongoing performance, fatigue, and recovery trends, rather than fixed schedules. Precision rehabilitation: Algorithms can identify which patients are likely to respond to specific interventions (e.g., constraint‑induced movement therapy vs robotics) and tailor plans accordingly. This moves rehabilitation from a "one‑size‑fits‑many" paradigm toward precision, context‑aware therapy, analogous to precision oncology but focused on function and participation. 2. Assessment, monitoring, and optimization AI for assessment Sensor‑based movement analysis: Machine learning models process accelerometer, IMU, EMG, and pressure data to quantify gait symmetry, joint kinematics, balance, and fine motor control with higher resolution than visual observation alone. Automated scoring: AI can approximate or support standardized scales (e.g., Fugl‑Meyer, Berg Balance Scale) by mapping sensor features or video-derived pose estimates to clinical scores, reducing inter‑rater variability and saving clinician time. Continuous monitoring Home and community tracking: Wearable and ambient sensors enable monitoring of daily steps, walking speed, arm use, posture, and adherence to exercises outside the clinic, feeding rich longitudinal datasets into AI models. Real‑time alerts: Algorithms can detect abnormal patterns—such as increased fall risk, reduced limb use, or signs of over‑exertion—and flag the clinician or adjust digital therapy content automatically. Optimization and decision support Predictive models: Using historical data, AI can forecast functional gains, plateau points, or risk of complications (e.g., falls, readmission), supporting individualized goal‑setting and resource allocation. Reinforcement learning and "digital twins": Emerging work in neurorehabilitation treats rehab as a sequential decision problem, using model‑based reinforcement learning and patient "digital twins" to recommend optimal timing, dosing, and progression of interventions over weeks to months.​ 3. Technologies: ML, wearables, analytics Machine learning algorithms: Supervised ML classifies movement quality (normal vs compensatory), detects exercise type from sensor streams, and estimates clinical scores. Unsupervised learning clusters patients into phenotypes (e.g., gait patterns after stroke), revealing subgroups that respond differently to certain therapies. Reinforcement learning and contextual bandits explore which therapy adjustments yield the best long‑term functional outcomes for a given individual.​ Wearable sensors and robotics: Inertial sensors, EMG, pressure insoles, and exoskeleton sensors capture high‑frequency movement and muscle activity data during training. Robotic devices (upper‑limb exoskeletons, gait trainers) coupled with AI can modulate assistance, resistance, or task difficulty in real time based on performance and predicted fatigue. Predictive and prescriptive analytics: Predictive analytics estimate trajectories (e.g., time to independent walking, expected upper‑limb function) to inform shared decisions with patients and families. Prescriptive analytics recommend therapy intensity, modality mix, and scheduling to maximize functional gains under resource constraints. 4. Benefits: outcomes, efficiency, engagement Improved functional outcomes: Studies report better motor recovery, gait quality, and ADL performance when AI‑assisted training is used—especially when robotics and intelligent feedback are involved. Reduced recovery time and resource use: More precise dosing and earlier identification of non‑responders can reduce ineffective sessions, shorten time to key milestones, and support safe earlier discharge with robust remote follow‑up. Increased adherence and engagement: AI‑driven digital rehab platforms use gamification, adaptive difficulty, and personalized feedback to keep patients engaged in home programs, improving adherence compared to static paper instructions. Support for clinicians: Instead of replacing therapists, AI can offload repetitive measurement tasks, highlight concerning trends, and offer data‑driven suggestions, allowing clinicians to focus on relational, motivational, and complex decision‑making aspects of care. 5. Challenges and ethical considerations Data privacy and security: Rehab AI often relies on continuous collection of sensitive motion, physiological, and sometimes audio/video data, raising questions about consent, storage, secondary use, and breach risk. Approaches like federated learning and on‑device processing are being explored to reduce centralization of identifiable data while still enabling model training. Algorithmic bias and fairness: If training data under‑represent older adults, women, certain racial/ethnic groups, or people with severe disability, AI models may misestimate performance or risk for those groups, potentially widening disparities in rehab access and outcomes. Ongoing auditing, diverse datasets, and participatory design with patients and clinicians are needed to ensure equitable performance. Integration with clinical workflows: Many AI tools are developed in research settings and are not yet seamlessly integrated into EHRs, scheduling systems, or therapist documentation workflows. Poorly integrated tools risk adding documentation burden or "alert fatigue," reducing adoption. Successful implementations co‑design interfaces with frontline therapists and physicians. Regulation, liability, and trust: It remains unclear in many jurisdictions how to regulate adaptive rehab algorithms (as medical devices, clinical decision support, or wellness tools) and who is liable when AI‑informed plans cause harm.​ Transparent, explainable models and clear communication to patients about the role of AI are critical for maintaining trust. 6. Case studies and emerging trends Remote and hybrid digital rehabilitation: AI‑driven platforms providing home‑based stroke, orthopedic, or Parkinson's rehab with clinician dashboards are improving adherence and extending care beyond brick‑and‑mortar clinics. Collaborative AI for precision neurorehabilitation: Frameworks combining patient‑clinician goal setting, digital twins, and reinforcement learning exemplify "collaborative AI" that augments rather than replaces therapists.​ Multimodal personalization: Integration of movement data, EMG, heart rate, sleep, and self‑reported pain/fatigue is enabling more nuanced adaptation to daily fluctuations in capacity. Conversational AI for education and coaching: Early work is assessing tools like ChatGPT as low‑risk supports for exercise education and motivation, though they are not yet precise enough to replace professional plan design AI is moving rehab toward patient‑centered, continuously adapting, and data‑rich care, but realizing this promise depends on addressing privacy, bias, workflow, and regulatory challenges in partnership with clinicians and patients.

Dormir no es simplemente descansar, sino un proceso activo en el que el cerebro reorganiza información, regula emociones y consolida la memoria. Cuando este proceso se altera de manera crónica, aumenta el riesgo de desarrollar trastornos como ansiedad, depresión y dificultades en la regulación emocional, por lo que su abordaje preventivo resulta fundamental en salud mental. Abordar la conducta del dormir desde las neurociencias y mediante una atención multimodal implica reconocer que el sueño es un pilar fundamental del bienestar psicológico en este podcast de El Expresso de las 10 la Dra. Teresita Villaseñor Jefa del Servicio de Neuropsicología en el Antiguo Hospital Civil de Guadalajara Fray Antonio alcalde, nos habla de ello, parte de nuestra cobertura de CIAM 2026.

Send a textIs AI in pathology actually improving diagnosis — or just adding complexity?In DigiPath Digest #37, we reviewed four recent publications covering AI-based biomarker quantification in glioblastoma, real-world digital workflow integration in prostate cancer, multimodal AI combining histopathology and genomics, and patient perspectives on AI in cancer diagnostics.This episode connects technical performance with something equally important: trust.Episode Highlights[00:02] Community & updates Digital Pathology 101 free PDF, upcoming patient-focused book, and global attendance.[04:07] AI-based image analysis in glioblastoma AI showed strong consistency with pathologists when quantifying Ki-67, P53, and PHH3. Significant biological correlations (Ki-67 ↔ PHH3, PHH3 ↔ P53) were detected by AI — not by manual assessment. Takeaway: computational quantification improves precision.[09:28] Real-world digital workflow + AI in prostate cancer (France) AI-pathologist concordance: • 93.2% (high probability cancer detection) • 99.0% (low probability slides) Gleason concordance: 76.6% 10% failure rate due to pre-analytical artifacts. Takeaway: infrastructure and sample quality still matter.[15:58] Multimodal AI (MARBIX framework) Combines whole slide images + immunogenomic data in a shared latent space using binary “monograms.” Performance in lung cancer: 85–89% vs 69–76% unimodal models. Takeaway: integrated data improves case retrieval and similarity reasoning.[22:13] AI-powered paper summary subscription introduced Structured summaries for busy professionals who want more than abstracts.[26:17] Patient roundtable on AI in pathology (Belgium) Patients expect: • Better accuracy • Faster turnaround • Stronger collaborationTrust is high when: • Algorithms use diverse datasets • Pathologists retain final responsibilityClinical validity mattered more than full algorithm transparency. Privacy concerns focused more on insurer misuse than cloud transfer.Key TakeawaysAI improves biomarker precision in glioblastoma.Digital pathology implementation works — but pre-analytics can limit AI performance.Multimodal AI represents the next meaningful step in precision diagnostics.Patients are not afraid of AI — they want validation, oversight, and governance.Human–AI collaboration remains central.If you're working in digital pathology, computational pathology, or precision oncology, this episode connects evidence, implementation, and patient perspective.Support the showGet the "Digital Pathology 101" FREE E-book and join us!

CEO & President Kyle and Graphic Designer & Brand strategist Kelsey explore how prompting has evolved from using AI like a “smarter Google” to structured strategies that deliver sharper, less generic results.They break down the CRIT framework (Context, Role, Interview, Task), share why detailed context reduces hallucinations, and explain how prompt libraries and model memory speed up repeatable work. The conversation also dives into context engineering with tools like Microsoft 365 Copilot and Google Workspace Gemini to make AI outputs more relevant and secure.Plus: common prompting mistakes, model comparisons, multimodal inputs, and how to onboard teams without losing brand consistency.Listen now to level up how you work with AI.00:00 Prompting Then vs Now: From “Smarter Google” to Strategic Skill 00:39 Why AI Sounds Vanilla: Averages, Models & AI Slop 01:33 Prompt Engineering & the CRIT Framework 02:35 Interview-Style Prompts: Fewer Hallucinations, Better Results 04:10 Garbage In, Garbage Out: Treat AI Like a New Hire 05:04 Let AI Help Write Prompts + Tools & Libraries 07:08 Why One-Liners Fall Flat (Contractor Analogy) 07:55 From Prompts to Systems: Templates & Model Memory 11:21 Context Engineering: Files, Memory & Workplace Data (Copilot/Gemini) 13:27 Over-Prompting: Context Limits & When to Reset 16:26 Set Outcomes, Don't Micromanage 18:22 Smarter Models: Gemini & Claude Need Less Steering 19:06 Claude Opus vs ChatGPT: Speed vs Detail 20:27 Multi-Model Workflow: Use Each for Its Strength 21:20 Why New Models Feel Smarter 22:11 Ask AI to Improve Your Prompts 24:42 Planning Mode: Structured Builds & AI Interviews 26:13 Training Teams: Frameworks, SOPs & Safe Experimentation 31:47 Multimodal & Voice Prompting (Gemini's Edge) 33:15 Wrap-Up & What's Next

What happens when AI safety filters fail to catch harmful content hidden inside images? Alessandro Pignati, AI Security Researcher at NeuralTrust, joins Sean Martin to reveal a newly discovered vulnerability that affects some of the most widely used image-generation models on the market today. The technique, called semantic chaining, is an image-based jailbreak attack discovered by the NeuralTrust research team, and it raises important questions about how enterprises secure their multimodal AI deployments.How does semantic chaining work? Pignati explains that the attack uses a single prompt composed of several parts. It begins with a benign scenario, such as a historical or educational context. A second instruction asks the model to make an innocent modification, like changing the color of a background. The final, critical step introduces a malicious directive, instructing the model to embed harmful content directly into the generated image. Because image-generation models apply fewer safety filters than their text-based counterparts, the harmful instructions are rendered inside the image without triggering the usual safeguards.The NeuralTrust research team tested semantic chaining against prominent models including Gemini Nano Pro, Grok 4, and Seedream 4.5 by ByteDance, finding the attack effective across all of them. For enterprises, the implications extend well beyond consumer use cases. Pignati notes that if an AI agent or chatbot has access to a knowledge base containing sensitive information or personal data, a carefully structured semantic chaining prompt can force the model to generate that data directly into an image, bypassing text-based safety mechanisms entirely.Organizations looking to learn more about semantic chaining and the broader landscape of AI agent security can visit the NeuralTrust blog, where the research team publishes detailed breakdowns of their findings. NeuralTrust also offers a newsletter with regular updates on agent security research and newly discovered vulnerabilities.This is a Brand Highlight. A Brand Highlight is a ~5 minute introductory conversation designed to put a spotlight on the guest and their company. Learn more: https://www.studioc60.com/creation#highlightGUESTAlessandro Pignati, AI Security Researcher, NeuralTrustOn LinkedIn: https://www.linkedin.com/in/alessandro-pignati/RESOURCESLearn more about NeuralTrust: https://neuraltrust.ai/Are you interested in telling your story?▶︎ Full Length Brand Story: https://www.studioc60.com/content-creation#full▶︎ Brand Spotlight Story: https://www.studioc60.com/content-creation#spotlight▶︎ Brand Highlight Story: https://www.studioc60.com/content-creation#highlightKEYWORDSAlessandro Pignati, NeuralTrust, Sean Martin, brand story, brand marketing, marketing podcast, brand highlight, semantic chaining, image jailbreak, AI security, agentic AI, multimodal AI, LLM safety, AI red teaming, prompt injection, AI agent security, image-based attacks, enterprise AI security Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

From rewriting Google's search stack in the early 2000s to reviving sparse trillion-parameter models and co-designing TPUs with frontier ML research, Jeff Dean has quietly shaped nearly every layer of the modern AI stack. As Chief AI Scientist at Google and a driving force behind Gemini, Jeff has lived through multiple scaling revolutions from CPUs and sharded indices to multimodal models that reason across text, video, and code.Jeff joins us to unpack what it really means to “own the Pareto frontier,” why distillation is the engine behind every Flash model breakthrough, how energy (in picojoules) not FLOPs is becoming the true bottleneck, what it was like leading the charge to unify all of Google's AI teams, and why the next leap won't come from bigger context windows alone, but from systems that give the illusion of attending to trillions of tokens.We discuss:* Jeff's early neural net thesis in 1990: parallel training before it was cool, why he believed scaling would win decades early, and the “bigger model, more data, better results” mantra that held for 15 years* The evolution of Google Search: sharding, moving the entire index into memory in 2001, softening query semantics pre-LLMs, and why retrieval pipelines already resemble modern LLM systems* Pareto frontier strategy: why you need both frontier “Pro” models and low-latency “Flash” models, and how distillation lets smaller models surpass prior generations* Distillation deep dive: ensembles → compression → logits as soft supervision, and why you need the biggest model to make the smallest one good* Latency as a first-class objective: why 10–50x lower latency changes UX entirely, and how future reasoning workloads will demand 10,000 tokens/sec* Energy-based thinking: picojoules per bit, why moving data costs 1000x more than a multiply, batching through the lens of energy, and speculative decoding as amortization* TPU co-design: predicting ML workloads 2–6 years out, speculative hardware features, precision reduction, sparsity, and the constant feedback loop between model architecture and silicon* Sparse models and “outrageously large” networks: trillions of parameters with 1–5% activation, and why sparsity was always the right abstraction* Unified vs. specialized models: abandoning symbolic systems, why general multimodal models tend to dominate vertical silos, and when vertical fine-tuning still makes sense* Long context and the illusion of scale: beyond needle-in-a-haystack benchmarks toward systems that narrow trillions of tokens to 117 relevant documents* Personalized AI: attending to your emails, photos, and documents (with permission), and why retrieval + reasoning will unlock deeply personal assistants* Coding agents: 50 AI interns, crisp specifications as a new core skill, and how ultra-low latency will reshape human–agent collaboration* Why ideas still matter: transformers, sparsity, RL, hardware, systems — scaling wasn't blind; the pieces had to multiply togetherShow Notes:* Gemma 3 Paper* Gemma 3* Gemini 2.5 Report* Jeff Dean's “Software Engineering Advice fromBuilding Large-Scale Distributed Systems” Presentation (with Back of the Envelope Calculations)* Latency Numbers Every Programmer Should Know by Jeff Dean* The Jeff Dean Facts* Jeff Dean Google Bio* Jeff Dean on “Important AI Trends” @Stanford AI Club* Jeff Dean & Noam Shazeer — 25 years at Google (Dwarkesh)—Jeff Dean* LinkedIn: https://www.linkedin.com/in/jeff-dean-8b212555* X: https://x.com/jeffdeanGoogle* https://google.com* https://deepmind.googleFull Video EpisodeTimestamps00:00:04 — Introduction: Alessio & Swyx welcome Jeff Dean, chief AI scientist at Google, to the Latent Space podcast00:00:30 — Owning the Pareto Frontier & balancing frontier vs low-latency models00:01:31 — Frontier models vs Flash models + role of distillation00:03:52 — History of distillation and its original motivation00:05:09 — Distillation's role in modern model scaling00:07:02 — Model hierarchy (Flash, Pro, Ultra) and distillation sources00:07:46 — Flash model economics & wide deployment00:08:10 — Latency importance for complex tasks00:09:19 — Saturation of some tasks and future frontier tasks00:11:26 — On benchmarks, public vs internal00:12:53 — Example long-context benchmarks & limitations00:15:01 — Long-context goals: attending to trillions of tokens00:16:26 — Realistic use cases beyond pure language00:18:04 — Multimodal reasoning and non-text modalities00:19:05 — Importance of vision & motion modalities00:20:11 — Video understanding example (extracting structured info)00:20:47 — Search ranking analogy for LLM retrieval00:23:08 — LLM representations vs keyword search00:24:06 — Early Google search evolution & in-memory index00:26:47 — Design principles for scalable systems00:28:55 — Real-time index updates & recrawl strategies00:30:06 — Classic “Latency numbers every programmer should know”00:32:09 — Cost of memory vs compute and energy emphasis00:34:33 — TPUs & hardware trade-offs for serving models00:35:57 — TPU design decisions & co-design with ML00:38:06 — Adapting model architecture to hardware00:39:50 — Alternatives: energy-based models, speculative decoding00:42:21 — Open research directions: complex workflows, RL00:44:56 — Non-verifiable RL domains & model evaluation00:46:13 — Transition away from symbolic systems toward unified LLMs00:47:59 — Unified models vs specialized ones00:50:38 — Knowledge vs reasoning & retrieval + reasoning00:52:24 — Vertical model specialization & modules00:55:21 — Token count considerations for vertical domains00:56:09 — Low resource languages & contextual learning00:59:22 — Origins: Dean's early neural network work01:10:07 — AI for coding & human–model interaction styles01:15:52 — Importance of crisp specification for coding agents01:19:23 — Prediction: personalized models & state retrieval01:22:36 — Token-per-second targets (10k+) and reasoning throughput01:23:20 — Episode conclusion and thanksTranscriptAlessio Fanelli [00:00:04]: Hey everyone, welcome to the Latent Space podcast. This is Alessio, founder of Kernel Labs, and I'm joined by Swyx, editor of Latent Space. Shawn Wang [00:00:11]: Hello, hello. We're here in the studio with Jeff Dean, chief AI scientist at Google. Welcome. Thanks for having me. It's a bit surreal to have you in the studio. I've watched so many of your talks, and obviously your career has been super legendary. So, I mean, congrats. I think the first thing must be said, congrats on owning the Pareto Frontier.Jeff Dean [00:00:30]: Thank you, thank you. Pareto Frontiers are good. It's good to be out there.Shawn Wang [00:00:34]: Yeah, I mean, I think it's a combination of both. You have to own the Pareto Frontier. You have to have like frontier capability, but also efficiency, and then offer that range of models that people like to use. And, you know, some part of this was started because of your hardware work. Some part of that is your model work, and I'm sure there's lots of secret sauce that you guys have worked on cumulatively. But, like, it's really impressive to see it all come together in, like, this slittily advanced.Jeff Dean [00:01:04]: Yeah, yeah. I mean, I think, as you say, it's not just one thing. It's like a whole bunch of things up and down the stack. And, you know, all of those really combine to help make UNOS able to make highly capable large models, as well as, you know, software techniques to get those large model capabilities into much smaller, lighter weight models that are, you know, much more cost effective and lower latency, but still, you know, quite capable for their size. Yeah.Alessio Fanelli [00:01:31]: How much pressure do you have on, like, having the lower bound of the Pareto Frontier, too? I think, like, the new labs are always trying to push the top performance frontier because they need to raise more money and all of that. And you guys have billions of users. And I think initially when you worked on the CPU, you were thinking about, you know, if everybody that used Google, we use the voice model for, like, three minutes a day, they were like, you need to double your CPU number. Like, what's that discussion today at Google? Like, how do you prioritize frontier versus, like, we have to do this? How do we actually need to deploy it if we build it?Jeff Dean [00:02:03]: Yeah, I mean, I think we always want to have models that are at the frontier or pushing the frontier because I think that's where you see what capabilities now exist that didn't exist at the sort of slightly less capable last year's version or last six months ago version. At the same time, you know, we know those are going to be really useful for a bunch of use cases, but they're going to be a bit slower and a bit more expensive than people might like for a bunch of other broader models. So I think what we want to do is always have kind of a highly capable sort of affordable model that enables a whole bunch of, you know, lower latency use cases. People can use them for agentic coding much more readily and then have the high-end, you know, frontier model that is really useful for, you know, deep reasoning, you know, solving really complicated math problems, those kinds of things. And it's not that. One or the other is useful. They're both useful. So I think we'd like to do both. And also, you know, through distillation, which is a key technique for making the smaller models more capable, you know, you have to have the frontier model in order to then distill it into your smaller model. So it's not like an either or choice. You sort of need that in order to actually get a highly capable, more modest size model. Yeah.Alessio Fanelli [00:03:24]: I mean, you and Jeffrey came up with the solution in 2014.Jeff Dean [00:03:28]: Don't forget, L'Oreal Vinyls as well. Yeah, yeah.Alessio Fanelli [00:03:30]: A long time ago. But like, I'm curious how you think about the cycle of these ideas, even like, you know, sparse models and, you know, how do you reevaluate them? How do you think about in the next generation of model, what is worth revisiting? Like, yeah, they're just kind of like, you know, you worked on so many ideas that end up being influential, but like in the moment, they might not feel that way necessarily. Yeah.Jeff Dean [00:03:52]: I mean, I think distillation was originally motivated because we were seeing that we had a very large image data set at the time, you know, 300 million images that we could train on. And we were seeing that if you create specialists for different subsets of those image categories, you know, this one's going to be really good at sort of mammals, and this one's going to be really good at sort of indoor room scenes or whatever, and you can cluster those categories and train on an enriched stream of data after you do pre-training on a much broader set of images. You get much better performance. If you then treat that whole set of maybe 50 models you've trained as a large ensemble, but that's not a very practical thing to serve, right? So distillation really came about from the idea of, okay, what if we want to actually serve that and train all these independent sort of expert models and then squish it into something that actually fits in a form factor that you can actually serve? And that's, you know, not that different from what we're doing today. You know, often today we're instead of having an ensemble of 50 models. We're having a much larger scale model that we then distill into a much smaller scale model.Shawn Wang [00:05:09]: Yeah. A part of me also wonders if distillation also has a story with the RL revolution. So let me maybe try to articulate what I mean by that, which is you can, RL basically spikes models in a certain part of the distribution. And then you have to sort of, well, you can spike models, but usually sometimes... It might be lossy in other areas and it's kind of like an uneven technique, but you can probably distill it back and you can, I think that the sort of general dream is to be able to advance capabilities without regressing on anything else. And I think like that, that whole capability merging without loss, I feel like it's like, you know, some part of that should be a distillation process, but I can't quite articulate it. I haven't seen much papers about it.Jeff Dean [00:06:01]: Yeah, I mean, I tend to think of one of the key advantages of distillation is that you can have a much smaller model and you can have a very large, you know, training data set and you can get utility out of making many passes over that data set because you're now getting the logits from the much larger model in order to sort of coax the right behavior out of the smaller model that you wouldn't otherwise get with just the hard labels. And so, you know, I think that's what we've observed. Is you can get, you know, very close to your largest model performance with distillation approaches. And that seems to be, you know, a nice sweet spot for a lot of people because it enables us to kind of, for multiple Gemini generations now, we've been able to make the sort of flash version of the next generation as good or even substantially better than the previous generations pro. And I think we're going to keep trying to do that because that seems like a good trend to follow.Shawn Wang [00:07:02]: So, Dara asked, so it was the original map was Flash Pro and Ultra. Are you just sitting on Ultra and distilling from that? Is that like the mother load?Jeff Dean [00:07:12]: I mean, we have a lot of different kinds of models. Some are internal ones that are not necessarily meant to be released or served. Some are, you know, our pro scale model and we can distill from that as well into our Flash scale model. So I think, you know, it's an important set of capabilities to have and also inference time scaling. It can also be a useful thing to improve the capabilities of the model.Shawn Wang [00:07:35]: And yeah, yeah, cool. Yeah. And obviously, I think the economy of Flash is what led to the total dominance. I think the latest number is like 50 trillion tokens. I don't know. I mean, obviously, it's changing every day.Jeff Dean [00:07:46]: Yeah, yeah. But, you know, by market share, hopefully up.Shawn Wang [00:07:50]: No, I mean, there's no I mean, there's just the economics wise, like because Flash is so economical, like you can use it for everything. Like it's in Gmail now. It's in YouTube. Like it's yeah. It's in everything.Jeff Dean [00:08:02]: We're using it more in our search products of various AI mode reviews.Shawn Wang [00:08:05]: Oh, my God. Flash past the AI mode. Oh, my God. Yeah, that's yeah, I didn't even think about that.Jeff Dean [00:08:10]: I mean, I think one of the things that is quite nice about the Flash model is not only is it more affordable, it's also a lower latency. And I think latency is actually a pretty important characteristic for these models because we're going to want models to do much more complicated things that are going to involve, you know, generating many more tokens from when you ask the model to do so. So, you know, if you're going to ask the model to do something until it actually finishes what you ask it to do, because you're going to ask now, not just write me a for loop, but like write me a whole software package to do X or Y or Z. And so having low latency systems that can do that seems really important. And Flash is one direction, one way of doing that. You know, obviously our hardware platforms enable a bunch of interesting aspects of our, you know, serving stack as well, like TPUs, the interconnect between. Chips on the TPUs is actually quite, quite high performance and quite amenable to, for example, long context kind of attention operations, you know, having sparse models with lots of experts. These kinds of things really, really matter a lot in terms of how do you make them servable at scale.Alessio Fanelli [00:09:19]: Yeah. Does it feel like there's some breaking point for like the proto Flash distillation, kind of like one generation delayed? I almost think about almost like the capability as a. In certain tasks, like the pro model today is a saturated, some sort of task. So next generation, that same task will be saturated at the Flash price point. And I think for most of the things that people use models for at some point, the Flash model in two generation will be able to do basically everything. And how do you make it economical to like keep pushing the pro frontier when a lot of the population will be okay with the Flash model? I'm curious how you think about that.Jeff Dean [00:09:59]: I mean, I think that's true. If your distribution of what people are asking people, the models to do is stationary, right? But I think what often happens is as the models become more capable, people ask them to do more, right? So, I mean, I think this happens in my own usage. Like I used to try our models a year ago for some sort of coding task, and it was okay at some simpler things, but wouldn't do work very well for more complicated things. And since then, we've improved dramatically on the more complicated coding tasks. And now I'll ask it to do much more complicated things. And I think that's true, not just of coding, but of, you know, now, you know, can you analyze all the, you know, renewable energy deployments in the world and give me a report on solar panel deployment or whatever. That's a very complicated, you know, more complicated task than people would have asked a year ago. And so you are going to want more capable models to push the frontier in the absence of what people ask the models to do. And that also then gives us. Insight into, okay, where does the, where do things break down? How can we improve the model in these, these particular areas, uh, in order to sort of, um, make the next generation even better.Alessio Fanelli [00:11:11]: Yeah. Are there any benchmarks or like test sets they use internally? Because it's almost like the same benchmarks get reported every time. And it's like, all right, it's like 99 instead of 97. Like, how do you have to keep pushing the team internally to it? Or like, this is what we're building towards. Yeah.Jeff Dean [00:11:26]: I mean, I think. Benchmarks, particularly external ones that are publicly available. Have their utility, but they often kind of have a lifespan of utility where they're introduced and maybe they're quite hard for current models. You know, I, I like to think of the best kinds of benchmarks are ones where the initial scores are like 10 to 20 or 30%, maybe, but not higher. And then you can sort of work on improving that capability for, uh, whatever it is, the benchmark is trying to assess and get it up to like 80, 90%, whatever. I, I think once it hits kind of 95% or something, you get very diminishing returns from really focusing on that benchmark, cuz it's sort of, it's either the case that you've now achieved that capability, or there's also the issue of leakage in public data or very related kind of data being, being in your training data. Um, so we have a bunch of held out internal benchmarks that we really look at where we know that wasn't represented in the training data at all. There are capabilities that we want the model to have. Um, yeah. Yeah. Um, that it doesn't have now, and then we can work on, you know, assessing, you know, how do we make the model better at these kinds of things? Is it, we need different kind of data to train on that's more specialized for this particular kind of task. Do we need, um, you know, a bunch of, uh, you know, architectural improvements or some sort of, uh, model capability improvements, you know, what would help make that better?Shawn Wang [00:12:53]: Is there, is there such an example that you, uh, a benchmark inspired in architectural improvement? Like, uh, I'm just kind of. Jumping on that because you just.Jeff Dean [00:13:02]: Uh, I mean, I think some of the long context capability of the, of the Gemini models that came, I guess, first in 1.5 really were about looking at, okay, we want to have, um, you know,Shawn Wang [00:13:15]: immediately everyone jumped to like completely green charts of like, everyone had, I was like, how did everyone crack this at the same time? Right. Yeah. Yeah.Jeff Dean [00:13:23]: I mean, I think, um, and once you're set, I mean, as you say that needed single needle and a half. Hey, stack benchmark is really saturated for at least context links up to 1, 2 and K or something. Don't actually have, you know, much larger than 1, 2 and 8 K these days or two or something. We're trying to push the frontier of 1 million or 2 million context, which is good because I think there are a lot of use cases where. Yeah. You know, putting a thousand pages of text or putting, you know, multiple hour long videos and the context and then actually being able to make use of that as useful. Try to, to explore the über graduation are fairly large. But the single needle in a haystack benchmark is sort of saturated. So you really want more complicated, sort of multi-needle or more realistic, take all this content and produce this kind of answer from a long context that sort of better assesses what it is people really want to do with long context. Which is not just, you know, can you tell me the product number for this particular thing?Shawn Wang [00:14:31]: Yeah, it's retrieval. It's retrieval within machine learning. It's interesting because I think the more meta level I'm trying to operate at here is you have a benchmark. You're like, okay, I see the architectural thing I need to do in order to go fix that. But should you do it? Because sometimes that's an inductive bias, basically. It's what Jason Wei, who used to work at Google, would say. Exactly the kind of thing. Yeah, you're going to win. Short term. Longer term, I don't know if that's going to scale. You might have to undo that.Jeff Dean [00:15:01]: I mean, I like to sort of not focus on exactly what solution we're going to derive, but what capability would you want? And I think we're very convinced that, you know, long context is useful, but it's way too short today. Right? Like, I think what you would really want is, can I attend to the internet while I answer my question? Right? But that's not going to happen. I think that's going to be solved by purely scaling the existing solutions, which are quadratic. So a million tokens kind of pushes what you can do. You're not going to do that to a trillion tokens, let alone, you know, a billion tokens, let alone a trillion. But I think if you could give the illusion that you can attend to trillions of tokens, that would be amazing. You'd find all kinds of uses for that. You would have attend to the internet. You could attend to the pixels of YouTube and the sort of deeper representations that we can find. You could attend to the form for a single video, but across many videos, you know, on a personal Gemini level, you could attend to all of your personal state with your permission. So like your emails, your photos, your docs, your plane tickets you have. I think that would be really, really useful. And the question is, how do you get algorithmic improvements and system level improvements that get you to something where you actually can attend to trillions of tokens? Right. In a meaningful way. Yeah.Shawn Wang [00:16:26]: But by the way, I think I did some math and it's like, if you spoke all day, every day for eight hours a day, you only generate a maximum of like a hundred K tokens, which like very comfortably fits.Jeff Dean [00:16:38]: Right. But if you then say, okay, I want to be able to understand everything people are putting on videos.Shawn Wang [00:16:46]: Well, also, I think that the classic example is you start going beyond language into like proteins and whatever else is extremely information dense. Yeah. Yeah.Jeff Dean [00:16:55]: I mean, I think one of the things about Gemini's multimodal aspects is we've always wanted it to be multimodal from the start. And so, you know, that sometimes to people means text and images and video sort of human-like and audio, audio, human-like modalities. But I think it's also really useful to have Gemini know about non-human modalities. Yeah. Like LIDAR sensor data from. Yes. Say, Waymo vehicles or. Like robots or, you know, various kinds of health modalities, x-rays and MRIs and imaging and genomics information. And I think there's probably hundreds of modalities of data where you'd like the model to be able to at least be exposed to the fact that this is an interesting modality and has certain meaning in the world. Where even if you haven't trained on all the LIDAR data or MRI data, you could have, because maybe that's not, you know, it doesn't make sense in terms of trade-offs of. You know, what you include in your main pre-training data mix, at least including a little bit of it is actually quite useful. Yeah. Because it sort of tempts the model that this is a thing.Shawn Wang [00:18:04]: Yeah. Do you believe, I mean, since we're on this topic and something I just get to ask you all the questions I always wanted to ask, which is fantastic. Like, are there some king modalities, like modalities that supersede all the other modalities? So a simple example was Vision can, on a pixel level, encode text. And DeepSeq had this DeepSeq CR paper that did that. Vision. And Vision has also been shown to maybe incorporate audio because you can do audio spectrograms and that's, that's also like a Vision capable thing. Like, so, so maybe Vision is just the king modality and like. Yeah.Jeff Dean [00:18:36]: I mean, Vision and Motion are quite important things, right? Motion. Well, like video as opposed to static images, because I mean, there's a reason evolution has evolved eyes like 23 independent ways, because it's such a useful capability for sensing the world around you, which is really what we want these models to be. So I think the only thing that we can be able to do is interpret the things we're seeing or the things we're paying attention to and then help us in using that information to do things. Yeah.Shawn Wang [00:19:05]: I think motion, you know, I still want to shout out, I think Gemini, still the only native video understanding model that's out there. So I use it for YouTube all the time. Nice.Jeff Dean [00:19:15]: Yeah. Yeah. I mean, it's actually, I think people kind of are not necessarily aware of what the Gemini models can actually do. Yeah. Like I have an example I've used in one of my talks. It had like, it was like a YouTube highlight video of 18 memorable sports moments across the last 20 years or something. So it has like Michael Jordan hitting some jump shot at the end of the finals and, you know, some soccer goals and things like that. And you can literally just give it the video and say, can you please make me a table of what all these different events are? What when the date is when they happened? And a short description. And so you get like now an 18 row table of that information extracted from the video, which is, you know, not something most people think of as like a turn video into sequel like table.Alessio Fanelli [00:20:11]: Has there been any discussion inside of Google of like, you mentioned tending to the whole internet, right? Google, it's almost built because a human cannot tend to the whole internet and you need some sort of ranking to find what you need. Yep. That ranking is like much different for an LLM because you can expect a person to look at maybe the first five, six links in a Google search versus for an LLM. Should you expect to have 20 links that are highly relevant? Like how do you internally figure out, you know, how do we build the AI mode that is like maybe like much broader search and span versus like the more human one? Yeah.Jeff Dean [00:20:47]: I mean, I think even pre-language model based work, you know, our ranking systems would be built to start. I mean, I think even pre-language model based work, you know, our ranking systems would be built to start. With a giant number of web pages in our index, many of them are not relevant. So you identify a subset of them that are relevant with very lightweight kinds of methods. You know, you're down to like 30,000 documents or something. And then you gradually refine that to apply more and more sophisticated algorithms and more and more sophisticated sort of signals of various kinds in order to get down to ultimately what you show, which is, you know, the final 10 results or, you know, 10 results plus. Other kinds of information. And I think an LLM based system is not going to be that dissimilar, right? You're going to attend to trillions of tokens, but you're going to want to identify, you know, what are the 30,000 ish documents that are with the, you know, maybe 30 million interesting tokens. And then how do you go from that into what are the 117 documents I really should be paying attention to in order to carry out the tasks that the user has asked? And I think, you know, you can imagine systems where you have, you know, a lot of highly parallel processing to identify those initial 30,000 candidates, maybe with very lightweight kinds of models. Then you have some system that sort of helps you narrow down from 30,000 to the 117 with maybe a little bit more sophisticated model or set of models. And then maybe the final model is the thing that looks. So the 117 things that might be your most capable model. So I think it has to, it's going to be some system like that, that is really enables you to give the illusion of attending to trillions of tokens. Sort of the way Google search gives you, you know, not the illusion, but you are searching the internet, but you're finding, you know, a very small subset of things that are, that are relevant.Shawn Wang [00:22:47]: Yeah. I often tell a lot of people that are not steeped in like Google search history that, well, you know, like Bert was. Like he was like basically immediately inside of Google search and that improves results a lot, right? Like I don't, I don't have any numbers off the top of my head, but like, I'm sure you guys, that's obviously the most important numbers to Google. Yeah.Jeff Dean [00:23:08]: I mean, I think going to an LLM based representation of text and words and so on enables you to get out of the explicit hard notion of, of particular words having to be on the page, but really getting at the notion of this topic of this page or this page. Paragraph is highly relevant to this query. Yeah.Shawn Wang [00:23:28]: I don't think people understand how much LLMs have taken over all these very high traffic system, very high traffic. Yeah. Like it's Google, it's YouTube. YouTube has this like semantics ID thing where it's just like every token or every item in the vocab is a YouTube video or something that predicts the video using a code book, which is absurd to me for YouTube size.Jeff Dean [00:23:50]: And then most recently GROK also for, for XAI, which is like, yeah. I mean, I'll call out even before LLMs were used extensively in search, we put a lot of emphasis on softening the notion of what the user actually entered into the query.Shawn Wang [00:24:06]: So do you have like a history of like, what's the progression? Oh yeah.Jeff Dean [00:24:09]: I mean, I actually gave a talk in, uh, I guess, uh, web search and data mining conference in 2009, uh, where we never actually published any papers about the origins of Google search, uh, sort of, but we went through sort of four or five or six. generations, four or five or six generations of, uh, redesigning of the search and retrieval system, uh, from about 1999 through 2004 or five. And that talk is really about that evolution. And one of the things that really happened in 2001 was we were sort of working to scale the system in multiple dimensions. So one is we wanted to make our index bigger, so we could retrieve from a larger index, which always helps your quality in general. Uh, because if you don't have the page in your index, you're going to not do well. Um, and then we also needed to scale our capacity because we were, our traffic was growing quite extensively. Um, and so we had, you know, a sharded system where you have more and more shards as the index grows, you have like 30 shards. And then if you want to double the index size, you make 60 shards so that you can bound the latency by which you respond for any particular user query. Um, and then as traffic grows, you add, you add more and more replicas of each of those. And so we eventually did the math that realized that in a data center where we had say 60 shards and, um, you know, 20 copies of each shard, we now had 1200 machines, uh, with disks. And we did the math and we're like, Hey, one copy of that index would actually fit in memory across 1200 machines. So in 2001, we introduced, uh, we put our entire index in memory and what that enabled from a quality perspective was amazing. Um, and so we had more and more replicas of each of those. Before you had to be really careful about, you know, how many different terms you looked at for a query, because every one of them would involve a disk seek on every one of the 60 shards. And so you, as you make your index bigger, that becomes even more inefficient. But once you have the whole index in memory, it's totally fine to have 50 terms you throw into the query from the user's original three or four word query, because now you can add synonyms like restaurant and restaurants and cafe and, uh, you know, things like that. Uh, bistro and all these things. And you can suddenly start, uh, sort of really, uh, getting at the meaning of the word as opposed to the exact semantic form the user typed in. And that was, you know, 2001, very much pre LLM, but really it was about softening the, the strict definition of what the user typed in order to get at the meaning.Alessio Fanelli [00:26:47]: What are like principles that you use to like design the systems, especially when you have, I mean, in 2001, the internet is like. Doubling, tripling every year in size is not like, uh, you know, and I think today you kind of see that with LLMs too, where like every year the jumps in size and like capabilities are just so big. Are there just any, you know, principles that you use to like, think about this? Yeah.Jeff Dean [00:27:08]: I mean, I think, uh, you know, first, whenever you're designing a system, you want to understand what are the sort of design parameters that are going to be most important in designing that, you know? So, you know, how many queries per second do you need to handle? How big is the internet? How big is the index you need to handle? How much data do you need to keep for every document in the index? How are you going to look at it when you retrieve things? Um, what happens if traffic were to double or triple, you know, will that system work well? And I think a good design principle is you're going to want to design a system so that the most important characteristics could scale by like factors of five or 10, but probably not beyond that because often what happens is if you design a system for X. And something suddenly becomes a hundred X, that would enable a very different point in the design space that would not make sense at X. But all of a sudden at a hundred X makes total sense. So like going from a disk space index to a in memory index makes a lot of sense once you have enough traffic, because now you have enough replicas of the sort of state on disk that those machines now actually can hold, uh, you know, a full copy of the, uh, index and memory. Yeah. And that all of a sudden enabled. A completely different design that wouldn't have been practical before. Yeah. Um, so I'm, I'm a big fan of thinking through designs in your head, just kind of playing with the design space a little before you actually do a lot of writing of code. But, you know, as you said, in the early days of Google, we were growing the index, uh, quite extensively. We were growing the update rate of the index. So the update rate actually is the parameter that changed the most. Surprising. So it used to be once a month.Shawn Wang [00:28:55]: Yeah.Jeff Dean [00:28:56]: And then we went to a system that could update any particular page in like sub one minute. Okay.Shawn Wang [00:29:02]: Yeah. Because this is a competitive advantage, right?Jeff Dean [00:29:04]: Because all of a sudden news related queries, you know, if you're, if you've got last month's news index, it's not actually that useful for.Shawn Wang [00:29:11]: News is a special beast. Was there any, like you could have split it onto a separate system.Jeff Dean [00:29:15]: Well, we did. We launched a Google news product, but you also want news related queries that people type into the main index to also be sort of updated.Shawn Wang [00:29:23]: So, yeah, it's interesting. And then you have to like classify whether the page is, you have to decide which pages should be updated and what frequency. Oh yeah.Jeff Dean [00:29:30]: There's a whole like, uh, system behind the scenes that's trying to decide update rates and importance of the pages. So even if the update rate seems low, you might still want to recrawl important pages quite often because, uh, the likelihood they change might be low, but the value of having updated is high.Shawn Wang [00:29:50]: Yeah, yeah, yeah, yeah. Uh, well, you know, yeah. This, uh, you know, mention of latency and, and saving things to this reminds me of one of your classics, which I have to bring up, which is latency numbers. Every programmer should know, uh, was there a, was it just a, just a general story behind that? Did you like just write it down?Jeff Dean [00:30:06]: I mean, this has like sort of eight or 10 different kinds of metrics that are like, how long does a cache mistake? How long does branch mispredict take? How long does a reference domain memory take? How long does it take to send, you know, a packet from the U S to the Netherlands or something? Um,Shawn Wang [00:30:21]: why Netherlands, by the way, or is it, is that because of Chrome?Jeff Dean [00:30:25]: Uh, we had a data center in the Netherlands, um, so, I mean, I think this gets to the point of being able to do the back of the envelope calculations. So these are sort of the raw ingredients of those, and you can use them to say, okay, well, if I need to design a system to do image search and thumb nailing or something of the result page, you know, how, what I do that I could pre-compute the image thumbnails. I could like. Try to thumbnail them on the fly from the larger images. What would that do? How much dis bandwidth than I need? How many des seeks would I do? Um, and you can sort of actually do thought experiments in, you know, 30 seconds or a minute with the sort of, uh, basic, uh, basic numbers at your fingertips. Uh, and then as you sort of build software using higher level libraries, you kind of want to develop the same intuitions for how long does it take to, you know, look up something in this particular kind of.Shawn Wang [00:31:21]: I'll see you next time.Shawn Wang [00:31:51]: Which is a simple byte conversion. That's nothing interesting. I wonder if you have any, if you were to update your...Jeff Dean [00:31:58]: I mean, I think it's really good to think about calculations you're doing in a model, either for training or inference.Jeff Dean [00:32:09]: Often a good way to view that is how much state will you need to bring in from memory, either like on-chip SRAM or HBM from the accelerator. Attached memory or DRAM or over the network. And then how expensive is that data motion relative to the cost of, say, an actual multiply in the matrix multiply unit? And that cost is actually really, really low, right? Because it's order, depending on your precision, I think it's like sub one picodule.Shawn Wang [00:32:50]: Oh, okay. You measure it by energy. Yeah. Yeah.Jeff Dean [00:32:52]: Yeah. I mean, it's all going to be about energy and how do you make the most energy efficient system. And then moving data from the SRAM on the other side of the chip, not even off the off chip, but on the other side of the same chip can be, you know, a thousand picodules. Oh, yeah. And so all of a sudden, this is why your accelerators require batching. Because if you move, like, say, the parameter of a model from SRAM on the, on the chip into the multiplier unit, that's going to cost you a thousand picodules. So you better make use of that, that thing that you moved many, many times with. So that's where the batch dimension comes in. Because all of a sudden, you know, if you have a batch of 256 or something, that's not so bad. But if you have a batch of one, that's really not good.Shawn Wang [00:33:40]: Yeah. Yeah. Right.Jeff Dean [00:33:41]: Because then you paid a thousand picodules in order to do your one picodule multiply.Shawn Wang [00:33:46]: I have never heard an energy-based analysis of batching.Jeff Dean [00:33:50]: Yeah. I mean, that's why people batch. Yeah. Ideally, you'd like to use batch size one because the latency would be great.Shawn Wang [00:33:56]: The best latency.Jeff Dean [00:33:56]: But the energy cost and the compute cost inefficiency that you get is quite large. So, yeah.Shawn Wang [00:34:04]: Is there a similar trick like, like, like you did with, you know, putting everything in memory? Like, you know, I think obviously NVIDIA has caused a lot of waves with betting very hard on SRAM with Grok. I wonder if, like, that's something that you already saw with, with the TPUs, right? Like that, that you had to. Uh, to serve at your scale, uh, you probably sort of saw that coming. Like what, what, what hardware, uh, innovations or insights were formed because of what you're seeing there?Jeff Dean [00:34:33]: Yeah. I mean, I think, you know, TPUs have this nice, uh, sort of regular structure of 2D or 3D meshes with a bunch of chips connected. Yeah. And each one of those has HBM attached. Um, I think for serving some kinds of models, uh, you know, you, you pay a lot higher cost. Uh, and time latency, um, bringing things in from HBM than you do bringing them in from, uh, SRAM on the chip. So if you have a small enough model, you can actually do model parallelism, spread it out over lots of chips and you actually get quite good throughput improvements and latency improvements from doing that. And so you're now sort of striping your smallish scale model over say 16 or 64 chips. Uh, but as if you do that and it all fits in. In SRAM, uh, that can be a big win. So yeah, that's not a surprise, but it is a good technique.Alessio Fanelli [00:35:27]: Yeah. What about the TPU design? Like how much do you decide where the improvements have to go? So like, this is like a good example of like, is there a way to bring the thousand picojoules down to 50? Like, is it worth designing a new chip to do that? The extreme is like when people say, oh, you should burn the model on the ASIC and that's kind of like the most extreme thing. How much of it? Is it worth doing an hardware when things change so quickly? Like what was the internal discussion? Yeah.Jeff Dean [00:35:57]: I mean, we, we have a lot of interaction between say the TPU chip design architecture team and the sort of higher level modeling, uh, experts, because you really want to take advantage of being able to co-design what should future TPUs look like based on where we think the sort of ML research puck is going, uh, in some sense, because, uh, you know, as a hardware designer for ML and in particular, you're trying to design a chip starting today and that design might take two years before it even lands in a data center. And then it has to sort of be a reasonable lifetime of the chip to take you three, four or five years. So you're trying to predict two to six years out where, what ML computations will people want to run two to six years out in a very fast changing field. And so having people with interest. Interesting ML research ideas of things we think will start to work in that timeframe or will be more important in that timeframe, uh, really enables us to then get, you know, interesting hardware features put into, you know, TPU N plus two, where TPU N is what we have today.Shawn Wang [00:37:10]: Oh, the cycle time is plus two.Jeff Dean [00:37:12]: Roughly. Wow. Because, uh, I mean, sometimes you can squeeze some changes into N plus one, but, you know, bigger changes are going to require the chip. Yeah. Design be earlier in its lifetime design process. Um, so whenever we can do that, it's generally good. And sometimes you can put in speculative features that maybe won't cost you much chip area, but if it works out, it would make something, you know, 10 times as fast. And if it doesn't work out, well, you burned a little bit of tiny amount of your chip area on that thing, but it's not that big a deal. Uh, sometimes it's a very big change and we want to be pretty sure this is going to work out. So we'll do like lots of carefulness. Uh, ML experimentation to show us, uh, this is actually the, the way we want to go. Yeah.Alessio Fanelli [00:37:58]: Is there a reverse of like, we already committed to this chip design so we can not take the model architecture that way because it doesn't quite fit?Jeff Dean [00:38:06]: Yeah. I mean, you, you definitely have things where you're going to adapt what the model architecture looks like so that they're efficient on the chips that you're going to have for both training and inference of that, of that, uh, generation of model. So I think it kind of goes both ways. Um, you know, sometimes you can take advantage of, you know, lower precision things that are coming in a future generation. So you can, might train it at that lower precision, even if the current generation doesn't quite do that. Mm.Shawn Wang [00:38:40]: Yeah. How low can we go in precision?Jeff Dean [00:38:43]: Because people are saying like ternary is like, uh, yeah, I mean, I'm a big fan of very low precision because I think that gets, that saves you a tremendous amount of time. Right. Because it's picojoules per bit that you're transferring and reducing the number of bits is a really good way to, to reduce that. Um, you know, I think people have gotten a lot of luck, uh, mileage out of having very low bit precision things, but then having scaling factors that apply to a whole bunch of, uh, those, those weights. Scaling. How does it, how does it, okay.Shawn Wang [00:39:15]: Interesting. You, so low, low precision, but scaled up weights. Yeah. Huh. Yeah. Never considered that. Yeah. Interesting. Uh, w w while we're on this topic, you know, I think there's a lot of, um, uh, this, the concept of precision at all is weird when we're sampling, you know, uh, we just, at the end of this, we're going to have all these like chips that I'll do like very good math. And then we're just going to throw a random number generator at the start. So, I mean, there's a movement towards, uh, energy based, uh, models and processors. I'm just curious if you've, obviously you've thought about it, but like, what's your commentary?Jeff Dean [00:39:50]: Yeah. I mean, I think. There's a bunch of interesting trends though. Energy based models is one, you know, diffusion based models, which don't sort of sequentially decode tokens is another, um, you know, speculative decoding is a way that you can get sort of an equivalent, very small.Shawn Wang [00:40:06]: Draft.Jeff Dean [00:40:07]: Batch factor, uh, for like you predict eight tokens out and that enables you to sort of increase the effective batch size of what you're doing by a factor of eight, even, and then you maybe accept five or six of those tokens. So you get. A five, a five X improvement in the amortization of moving weights, uh, into the multipliers to do the prediction for the, the tokens. So these are all really good techniques and I think it's really good to look at them from the lens of, uh, energy, real energy, not energy based models, um, and, and also latency and throughput, right? If you look at things from that lens, that sort of guides you to. Two solutions that are gonna be, uh, you know, better from, uh, you know, being able to serve larger models or, you know, equivalent size models more cheaply and with lower latency.Shawn Wang [00:41:03]: Yeah. Well, I think, I think I, um, it's appealing intellectually, uh, haven't seen it like really hit the mainstream, but, um, I do think that, uh, there's some poetry in the sense that, uh, you know, we don't have to do, uh, a lot of shenanigans if like we fundamentally. Design it into the hardware. Yeah, yeah.Jeff Dean [00:41:23]: I mean, I think there's still a, there's also sort of the more exotic things like analog based, uh, uh, computing substrates as opposed to digital ones. Uh, I'm, you know, I think those are super interesting cause they can be potentially low power. Uh, but I think you often end up wanting to interface that with digital systems and you end up losing a lot of the power advantages in the digital to analog and analog to digital conversions. You end up doing, uh, at the sort of boundaries. And periphery of that system. Um, I still think there's a tremendous distance we can go from where we are today in terms of energy efficiency with sort of, uh, much better and specialized hardware for the models we care about.Shawn Wang [00:42:05]: Yeah.Alessio Fanelli [00:42:06]: Um, any other interesting research ideas that you've seen, or like maybe things that you cannot pursue a Google that you would be interested in seeing researchers take a step at, I guess you have a lot of researchers. Yeah, I guess you have enough, but our, our research.Jeff Dean [00:42:21]: Our research portfolio is pretty broad. I would say, um, I mean, I think, uh, in terms of research directions, there's a whole bunch of, uh, you know, open problems and how do you make these models reliable and able to do much longer, kind of, uh, more complex tasks that have lots of subtasks. How do you orchestrate, you know, maybe one model that's using other models as tools in order to sort of build, uh, things that can accomplish, uh, you know, much more. Yeah. Significant pieces of work, uh, collectively, then you would ask a single model to do. Um, so that's super interesting. How do you get more verifiable, uh, you know, how do you get RL to work for non-verifiable domains? I think it's a pretty interesting open problem because I think that would broaden out the capabilities of the models, the improvements that you're seeing in both math and coding. Uh, if we could apply those to other less verifiable domains, because we've come up with RL techniques that actually enable us to do that. Uh, effectively, that would, that would really make the models improve quite a lot. I think.Alessio Fanelli [00:43:26]: I'm curious, like when we had Noam Brown on the podcast, he said, um, they already proved you can do it with deep research. Um, you kind of have it with AI mode in a way it's not verifiable. I'm curious if there's any thread that you think is interesting there. Like what is it? Both are like information retrieval of JSON. So I wonder if it's like the retrieval is like the verifiable part. That you can score or what are like, yeah, yeah. How, how would you model that, that problem?Jeff Dean [00:43:55]: Yeah. I mean, I think there are ways of having other models that can evaluate the results of what a first model did, maybe even retrieving. Can you have another model that says, is this things, are these things you retrieved relevant? Or can you rate these 2000 things you retrieved to assess which ones are the 50 most relevant or something? Um, I think those kinds of techniques are actually quite effective. Sometimes I can even be the same model, just prompted differently to be a, you know, a critic as opposed to a, uh, actual retrieval system. Yeah.Shawn Wang [00:44:28]: Um, I do think like there, there is that, that weird cliff where like, it feels like we've done the easy stuff and then now it's, but it always feels like that every year. It's like, oh, like we know, we know, and the next part is super hard and nobody's figured it out. And, uh, exactly with this RLVR thing where like everyone's talking about, well, okay, how do we. the next stage of the non-verifiable stuff. And everyone's like, I don't know, you know, Ellen judge.Jeff Dean [00:44:56]: I mean, I feel like the nice thing about this field is there's lots and lots of smart people thinking about creative solutions to some of the problems that we all see. Uh, because I think everyone sort of sees that the models, you know, are great at some things and they fall down around the edges of those things and, and are not as capable as we'd like in those areas. And then coming up with good techniques and trying those. And seeing which ones actually make a difference is sort of what the whole research aspect of this field is, is pushing forward. And I think that's why it's super interesting. You know, if you think about two years ago, we were struggling with GSM, eight K problems, right? Like, you know, Fred has two rabbits. He gets three more rabbits. How many rabbits does he have? That's a pretty far cry from the kinds of mathematics that the models can, and now you're doing IMO and Erdos problems in pure language. Yeah. Yeah. Pure language. So that is a really, really amazing jump in capabilities in, you know, in a year and a half or something. And I think, um, for other areas, it'd be great if we could make that kind of leap. Uh, and you know, we don't exactly see how to do it for some, some areas, but we do see it for some other areas and we're going to work hard on making that better. Yeah.Shawn Wang [00:46:13]: Yeah.Alessio Fanelli [00:46:14]: Like YouTube thumbnail generation. That would be very helpful. We need that. That would be AGI. We need that.Shawn Wang [00:46:20]: That would be. As far as content creators go.Jeff Dean [00:46:22]: I guess I'm not a YouTube creator, so I don't care that much about that problem, but I guess, uh, many people do.Shawn Wang [00:46:27]: It does. Yeah. It doesn't, it doesn't matter. People do judge books by their covers as it turns out. Um, uh, just to draw a bit on the IMO goal. Um, I'm still not over the fact that a year ago we had alpha proof and alpha geometry and all those things. And then this year we were like, screw that we'll just chuck it into Gemini. Yeah. What's your reflection? Like, I think this, this question about. Like the merger of like symbolic systems and like, and, and LMS, uh, was a very much core belief. And then somewhere along the line, people would just said, Nope, we'll just all do it in the LLM.Jeff Dean [00:47:02]: Yeah. I mean, I think it makes a lot of sense to me because, you know, humans manipulate symbols, but we probably don't have like a symbolic representation in our heads. Right. We have some distributed representation that is neural net, like in some way of lots of different neurons. And activation patterns firing when we see certain things and that enables us to reason and plan and, you know, do chains of thought and, you know, roll them back now that, that approach for solving the problem doesn't seem like it's going to work. I'm going to try this one. And, you know, in a lot of ways we're emulating what we intuitively think, uh, is happening inside real brains in neural net based models. So it never made sense to me to have like completely separate. Uh, discrete, uh, symbolic things, and then a completely different way of, of, uh, you know, thinking about those things.Shawn Wang [00:47:59]: Interesting. Yeah. Uh, I mean, it's maybe seems obvious to you, but it wasn't obvious to me a year ago. Yeah.Jeff Dean [00:48:06]: I mean, I do think like that IMO with, you know, translating to lean and using lean and then the next year and also a specialized geometry model. And then this year switching to a single unified model. That is roughly the production model with a little bit more inference budget, uh, is actually, you know, quite good because it shows you that the capabilities of that general model have improved dramatically and, and now you don't need the specialized model. This is actually sort of very similar to the 2013 to 16 era of machine learning, right? Like it used to be, people would train separate models for lots of different, each different problem, right? I have, I want to recognize street signs and something. So I train a street sign. Recognition recognition model, or I want to, you know, decode speech recognition. I have a speech model, right? I think now the era of unified models that do everything is really upon us. And the question is how well do those models generalize to new things they've never been asked to do and they're getting better and better.Shawn Wang [00:49:10]: And you don't need domain experts. Like one of my, uh, so I interviewed ETA who was on, who was on that team. Uh, and he was like, yeah, I, I don't know how they work. I don't know where the IMO competition was held. I don't know the rules of it. I just trained the models, the training models. Yeah. Yeah. And it's kind of interesting that like people with these, this like universal skill set of just like machine learning, you just give them data and give them enough compute and they can kind of tackle any task, which is the bitter lesson, I guess. I don't know. Yeah.Jeff Dean [00:49:39]: I mean, I think, uh, general models, uh, will win out over specialized ones in most cases.Shawn Wang [00:49:45]: Uh, so I want to push there a bit. I think there's one hole here, which is like, uh. There's this concept of like, uh, maybe capacity of a model, like abstractly a model can only contain the number of bits that it has. And, uh, and so it, you know, God knows like Gemini pro is like one to 10 trillion parameters. We don't know, but, uh, the Gemma models, for example, right? Like a lot of people want like the open source local models that are like that, that, that, and, and, uh, they have some knowledge, which is not necessary, right? Like they can't know everything like, like you have the. The luxury of you have the big model and big model should be able to capable of everything. But like when, when you're distilling and you're going down to the small models, you know, you're actually memorizing things that are not useful. Yeah. And so like, how do we, I guess, do we want to extract that? Can we, can we divorce knowledge from reasoning, you know?Jeff Dean [00:50:38]: Yeah. I mean, I think you do want the model to be most effective at reasoning if it can retrieve things, right? Because having the model devote precious parameter space. To remembering obscure facts that could be looked up is actually not the best use of that parameter space, right? Like you might prefer something that is more generally useful in more settings than this obscure fact that it has. Um, so I think that's always attention at the same time. You also don't want your model to be kind of completely detached from, you know, knowing stuff about the world, right? Like it's probably useful to know how long the golden gate be. Bridges just as a general sense of like how long are bridges, right? And, uh, it should have that kind of knowledge. It maybe doesn't need to know how long some teeny little bridge in some other more obscure part of the world is, but, uh, it does help it to have a fair bit of world knowledge and the bigger your model is, the more you can have. Uh, but I do think combining retrieval with sort of reasoning and making the model really good at doing multiple stages of retrieval. Yeah.Shawn Wang [00:51:49]: And reasoning through the intermediate retrieval results is going to be a, a pretty effective way of making the model seem much more capable, because if you think about, say, a personal Gemini, yeah, right?Jeff Dean [00:52:01]: Like we're not going to train Gemini on my email. Probably we'd rather have a single model that, uh, we can then use and use being able to retrieve from my email as a tool and have the model reason about it and retrieve from my photos or whatever, uh, and then make use of that and have multiple. Um, you know, uh, stages of interaction. that makes sense.Alessio Fanelli [00:52:24]: Do you think the vertical models are like, uh, interesting pursuit? Like when people are like, oh, we're building the best healthcare LLM, we're building the best law LLM, are those kind of like short-term stopgaps or?Jeff Dean [00:52:37]: No, I mean, I think, I think vertical models are interesting. Like you want them to start from a pretty good base model, but then you can sort of, uh, sort of viewing them, view them as enriching the data. Data distribution for that particular vertical domain for healthcare, say, um, we're probably not going to train or for say robotics. We're probably not going to train Gemini on all possible robotics data. We, you could train it on because we want it to have a balanced set of capabilities. Um, so we'll expose it to some robotics data, but if you're trying to build a really, really good robotics model, you're going to want to start with that and then train it on more robotics data. And then maybe that would. It's multilingual translation capability, but improve its robotics capabilities. And we're always making these kind of, uh, you know, trade-offs in the data mix that we train the base Gemini models on. You know, we'd love to include data from 200 more languages and as much data as we have for those languages, but that's going to displace some other capabilities of the model. It won't be as good at, um, you know, Pearl programming, you know, it'll still be good at Python programming. Cause we'll include it. Enough. Of that, but there's other long tail computer languages or coding capabilities that it may suffer on or multi, uh, multimodal reasoning capabilities may suffer. Cause we didn't get to expose it to as much data there, but it's really good at multilingual things. So I, I think some combination of specialized models, maybe more modular models. So it'd be nice to have the capability to have those 200 languages, plus this awesome robotics model, plus this awesome healthcare, uh, module that all can be knitted together to work in concert and called upon in different circumstances. Right? Like if I have a health related thing, then it should enable using this health module in conjunction with the main base model to be even better at those kinds of things. Yeah.Shawn Wang [00:54:36]: Installable knowledge. Yeah.Jeff Dean [00:54:37]: Right.Shawn Wang [00:54:38]: Just download as a, as a package.Jeff Dean [00:54:39]: And some of that installable stuff can come from retrieval, but some of it probably should come from preloaded training on, you know, uh, a hundred billion tokens or a trillion tokens of health data. Yeah.Shawn Wang [00:54:51]: And for listeners, I think, uh, I will highlight the Gemma three end paper where they, there was a little bit of that, I think. Yeah.Alessio Fanelli [00:54:56]: Yeah. I guess the question is like, how many billions of tokens do you need to outpace the frontier model improvements? You know, it's like, if I have to make this model better healthcare and the main. Gemini model is still improving. Do I need 50 billion tokens? Can I do it with a hundred, if I need a trillion healthcare tokens, it's like, they're probably not out there that you don't have, you know, I think that's really like the.Jeff Dean [00:55:21]: Well, I mean, I think healthcare is a particularly challenging domain, so there's a lot of healthcare data that, you know, we don't have access to appropriately, but there's a lot of, you know, uh, healthcare organizations that want to train models on their own data. That is not public healthcare data, uh, not public health. But public healthcare data. Um, so I think there are opportunities there to say, partner with a large healthcare organization and train models for their use that are going to be, you know, more bespoke, but probably, uh, might be better than a general model trained on say, public data. Yeah.Shawn Wang [00:55:58]: Yeah. I, I believe, uh, by the way, also this is like somewhat related to the language conversation. Uh, I think one of your, your favorite examples was you can put a low resource language in the context and it just learns. Yeah.Jeff Dean [00:56:09]: Oh, yeah, I think the example we used was Calamon, which is truly low resource because it's only spoken by, I think 120 people in the world and there's no written text.Shawn Wang [00:56:20]: So, yeah. So you can just do it that way. Just put it in the context. Yeah. Yeah. But I think your whole data set in the context, right.Jeff Dean [00:56:27]: If you, if you take a language like, uh, you know, Somali or something, there is a fair bit of Somali text in the world that, uh, or Ethiopian Amharic or something, um, you know, we probably. Yeah. Are not putting all the data from those languages into the Gemini based training. We put some of it, but if you put more of it, you'll improve the capabilities of those models.Shawn Wang [00:56:49]: Yeah.Jeff Dean [00:56:49]:

Let's talk about an incredibly adaptable project in which students experiment with creative ideas across modes. It's easy to plug into a variety of units and times of year, and ready to tap at a moment's notice. It remixes easily for Valentine's Day on the horizon, but it could also work well at Halloween, or as part of a creative writing unit, or when you're reading any verse novel or graphic novel. This project starts with fiction, moves into verse, and lands in a multimodal combination of verse and imagery. I call it a multimodal flash verse project, informed along the way by the brilliant mode collaborations of Jason Reynolds. Let's dig into it. Links Mentioned: Jason Reynolds' Interview with the Kennedy Center: https://www.youtube.com/watch?v=cuXNsJvNaFs  Book Trailer for Ain't Burned all the Bright: https://www.youtube.com/watch?v=EjqvOyAh36Y  Reynolds on his collab with Novgorodoff: https://www.youtube.com/watch?v=0ErpAXd7Swg  There was a Party for Langston Read-Aloud: https://www.youtube.com/watch?v=m4MYO4WmR9s  Go Further:  Explore alllll the Episodes of The Spark Creativity Teacher Podcast. Get my popular free hexagonal thinking digital toolkit Join our community, Creative High School English, on Facebook. Come hang out on Instagram.  Enjoying the podcast? Please consider sharing it with a friend, snagging a screenshot to share on the 'gram, or tapping those ⭐⭐⭐⭐⭐ to help others discover the show. Thank you!   

What if your sales motion created real partnerships instead of fragile price wins? That's the thread we pull with Hans, CEO of Odyssey Logistics, as he maps a journey from Danish directness and early Maersk rotations to leading a global multimodal platform through a roll-up-to-one-brand transformation. The conversation is practical, candid, and loaded with moves you can copy tomorrow—whether you're running a desk or running a P&L.We start with the foundation: a value proposition built on facts, not slogans. Hans explains how probing, silence, and quarterly KPI reviews expose true customer pain, unlock share of wallet, and make relationships stick at multiple levels, including the C-suite. He shares why he spends heavy time in the field, what onsite town halls surface that email never will, and how a consistent cadence—global Q&A, divisional sessions, defined values—turns culture from posters into behavior.Then we dig into Odyssey's shift from 16+ legacy brands to One Odyssey. Hans breaks down the integration playbook: centralizing shared services, standardizing procurement, and rebranding fast without crushing entrepreneurial spirit. He's frank about PE carve-outs, IT risk, and why overcommunication beats overpromising during ownership changes. On growth, we get specific: three levers—share of wallet, new logos, and cross-sell—powered by a cross-trained sales force and subject matter experts. Multimodal strategy is the differentiator, with intermodal often beating truckload on cost and CO2 when planned well.Technology underpins the whole plan. A data lake fuels route optimization, predictive analytics, and automated bidding, while better systems lift both customer outcomes and employee satisfaction. Odyssey's rebranded brokerage in Atlanta becomes the easy entry point—truckload and LTL open the door to deeper multimodal solutions. Hans closes with career advice that never expires: choose training over titles, learn every job, stay humble, and remember the team is smarter than any one of us.If this resonates, follow the show, share it with a colleague who sells on price, and leave a quick review so more people can find conversations that move logistics forward.Follow The Freight Pod and host Andrew Silver on LinkedIn.Thanks to our sponsors:Stuut Technologies: Your AI coworker that collects your cash automatically.https://www.stuut.ai/Cloneops.ai: Not just AI. Industry-born AI.https://www.cloneops.ai/Rapido Solutions Group: Nearshore solutions for logistics companies.https://www.gorapido.com/GenLogs: Freight Intelligence on every carrier, shipper, and asset via a nationwide sensor networkhttps://www.genlogs.io/

With developments in generative AI progressing at such a furious pace, how can investors cut through the noise to identify the companies that will really matter? Baillie Gifford's Kyle McEnery shares his approach to meeting the entrepreneurs building the future – including his encounters with AppLovin, Anthropic, NVIDIA, Roblox and Reddit. Background:Kyle McEnery is an investment manager in our Long Term Global Growth Team (LTGG) and previously led Baillie Gifford's Artificial Intelligence Research Project. In this conversation, he tells host Leo Kelion why AI's ever-increasing capabilities make this one of the most exciting times to be a growth investor, and how leadership and culture act as signals in the noise to help identify companies with the greatest long-term growth potential. In addition to discussing which of the firms enabling and using today's language-based ‘frontier' AI models are leading the pack, he explains how efforts to understand and simulate real-world physics could unlock further progress. Portfolio companies discussed include:Anthropic – developer of the Claude AI models, which excel at coding, among other tasks.NVIDIA – the semiconductors firm whose accelerator chips are powering many of the advances in generative AI.Roblox – the video games platform whose Cube 3D technology allows creators to build objects and environments out of text-based descriptions.AppLovin – the ad-tech company whose AI-first strategy keeps the business lean and nimble.Reddit – the online discussion forum, whose authentic human conversations are gaining in value as a counterpoint to AI-generated output. Resources:AI and the future of everything: a long-term perspectiveAnthropic: why we are backing the AI frontrunnerLong Term Global Growth Strategy (institutional investors only)LTGG philosophy and process (institutional investors only)Private companies: from Anthropic to ZetwerkShort Briefings on Long Term Thinking hub Companies mentioned include:Alphabet/GoogleAmazonAnthropicAppLovinHorizon RoboticsNVIDIARedditRobloxTesla  Timecodes:00:00  Introduction – Dartmouth College's artificial intelligence workshop01:50   From quantum to AI via asset management02:50  Creating and then culling a machine-learning initiative08:05  ChatGPT's wake-up call10:35   Exceptional companies at the dawn of generative AI12:10   Anthropic's appeal to business customers14:55   A winner-takes-all opportunity?17:05   Dario Amodei and the scaling laws19:10   NVIDIA's foundational role in neural networks22:55  Making video game items in Roblox with AI25:00  AppLovin – a company built for the next era26:55  Reddit's valuable conversational communities29:35  World models, spatial AI and the physical world32:35  Staying open-minded and humble33:35  Book choice  Glossary of terms (in order of mention): Generative AI: AI systems that create new content such as text, images or code rather than just analysing data.Machine learning: AI techniques where systems learn patterns from data rather than being explicitly programmed.End-to-end, systematic (investment strategy): Fully automated, with decisions made by predefined rules rather than human judgement.Agentic AI: AI systems that can plan and carry out tasks autonomously rather than just responding to prompts.R&D: Research and development.GPT: OpenAI's models, which power its ChatGPT chatbot.Natural language processing: AI that enables computers to understand and generate human language.Token: A chunk of text, such as a word or part of a word, used by language models.Foundation models: Large AI models that can handle a wide variety of tasks.Know your customer (KYC): Financial checks used by banks to verify customers' identities and risks.Scaling laws: The idea that AI performance improves predictably as models, data and computing power increase.Compute: The processing power required to train and run AI models.Jevons' paradox: The counterintuitive idea that efficiency gains can increase, rather than reduce, overall usage.CUDA: NVIDIA's software platform for programming its chips for high-performance computing.Jensen: Jensen Huang, NVIDIA's co-founder and chief executive.Metaverse: Shared virtual worlds where people interact, create and play online.Large language models (LLMs): AI systems trained on vast amounts of text to understand and generate language.Multimodal models: AI systems that can process multiple types of data, such as text, images and video.World models: AI systems that learn how the physical world works in order to predict and simulate it.Embodied AI: AI that learns through physical interaction with the real world, such as robots or vehicles.Imitation learning: Training AI by having it copy actions demonstrated by humans.

In this Omni Talk Retail episode, recorded live from NRF 2026 in the Vusion podcast studio, Mark Propes from Vusion and Art Miller from Qualcomm reveal how their partnership is enabling "detect and connect" capabilities that transform physical retail into personalized experiences, and why retailers still testing need to operationalize now before the gap becomes permanent. From edge computing that processes 4K video locally instead of streaming to the cloud, to closed-loop attribution tracking customer intent in real-time physical space, Mark and Art break down the multimodal signal taxonomy (RFID, Wi-Fi, Bluetooth, vision) powering connected stores. They share insights on why scanning barcodes continuously creates data-poor environments, how agentic AI creates new doorways into physical stores, and the precision needed for sub-30 minute delivery promises. If you've wondered what detect and connect actually means beyond buzzwords, this conversation delivers the technical foundation and business applications.

We are reupping this episode after LMArena announced their fresh Series A (https://www.theinformation.com/articles/ai-evaluation-startup-lmarena-valued-1-7-billion-new-funding-round?rc=luxwz4), raising $150m at a $1.7B valuation, with $30M annualized consumption revenue (aka $2.5m MRR) after their September evals product launch.—-From building LMArena in a Berkeley basement to raising $100M and becoming the de facto leaderboard for frontier AI, Anastasios Angelopoulos returns to Latent Space to recap 2025 in one of the most influential platforms in AI—trusted by millions of users, every major lab, and the entire industry to answer one question: which model is actually best for real-world use cases? We caught up with Anastasios live at NeurIPS 2025 to dig into the origin story (spoiler: it started as an academic project incubated by Anjney Midha at a16z, who formed an entity and gave grants before they even committed to starting a company), why they decided to spin out instead of staying academic or nonprofit (the only way to scale was to build a company), how they're spending that $100M (inference costs, React migration off Gradio, and hiring world-class talent across ML, product, and go-to-market), the leaderboard delusion controversy and why their response demolished the paper's claims (factual errors, misrepresentation of open vs. closed source sampling, and ignoring the transparency of preview testing that the community loves), why platform integrity comes first (the public leaderboard is a charity, not a pay-to-play system—models can't pay to get on, can't pay to get off, and scores reflect millions of real votes), how they're expanding into occupational verticals (medicine, legal, finance, creative marketing) and multimodal arenas (video coming soon), why consumer retention is earned every single day (sign-in and persistent history were the unlock, but users are fickle and can leave at any moment), and his vision for Arena as the central evaluation platform that provides the North Star for the industry—constantly fresh, immune to overfitting, and grounded in millions of real-world conversations from real users.We discuss:* The $100M raise: use of funds is primarily inference costs (funding free usage for tens of millions of monthly conversations), React migration off Gradio (custom loading icons, better developer hiring, more flexibility), and hiring world-class talent* The scale: 250M+ conversations on the platform, tens of millions per month, 25% of users do software for a living, and half of users are now logged in* The leaderboard illusion controversy: Cohere researchers claimed undisclosed private testing created inequities, but Arena's response demolished the paper's factual errors (misrepresented open vs. closed source sampling, ignored transparency of preview testing that the community loves)* Why preview testing is loved by the community: secret codenames (Gemini Nano Banana, named after PM Naina's nickname), early access to unreleased models, and the thrill of being first to vote on frontier capabilities* The Nano Banana moment: changed Google's market share overnight, billions of dollars in stock movement, and validated that multimodal models (image generation, video) are economically critical for marketing, design, and AI-for-science* New categories: occupational and expert arenas (medicine, legal, finance, creative marketing), Code Arena, and video arena coming soonFull Video EpisodeTimestamps00:00:00 Introduction: Anastasios from Arena and the LM Arena Journey00:01:36 The Anjney Midha Incubation: From Berkeley Basement to Startup00:02:47 The Decision to Start a Company: Scaling Beyond Academia00:03:38 The $100M Raise: Use of Funds and Platform Economics00:05:10 Arena's User Base: 5M+ Users and Diverse Demographics00:06:02 The Competitive Landscape: Artificial Analysis, AI.xyz, and Arena's Differentiation00:08:12 Educational Value and Learning from the Community00:08:41 Technical Migration: From Gradio to React and Platform Evolution00:10:18 Leaderboard Delusion Paper: Addressing Critiques and Maintaining Integrity00:12:29 Nano Banana Moment: How Preview Models Create Market Impact00:13:41 Multimodal AI and Image Generation: From Skepticism to Economic Value00:15:37 Core Principles: Platform Integrity and the Public Leaderboard as Charity00:18:29 Future Roadmap: Expert Categories, Multimodal, Video, and Occupational Verticals00:19:10 API Strategy and Focus: Doing One Thing Well00:19:51 Community Management and Retention: Sign-In, History, and Daily Value00:22:21 Partnerships and Agent Evaluation: From Devon to Full-Featured Harnesses00:21:49 Hiring and Building a High-Performance Team Get full access to Latent.Space at www.latent.space/subscribe

Meta just made a multi-billion acquisition for AI agents.

From creating SWE-bench in a Princeton basement to shipping CodeClash, SWE-bench Multimodal, and SWE-bench Multilingual, John Yang has spent the last year and a half watching his benchmark become the de facto standard for evaluating AI coding agents—trusted by Cognition (Devin), OpenAI, Anthropic, and every major lab racing to solve software engineering at scale. We caught up with John live at NeurIPS 2025 to dig into the state of code evals heading into 2026: why SWE-bench went from ignored (October 2023) to the industry standard after Devin's launch (and how Walden emailed him two weeks before the big reveal), how the benchmark evolved from Django-heavy to nine languages across 40 repos (JavaScript, Rust, Java, C, Ruby), why unit tests as verification are limiting and long-running agent tournaments might be the future (CodeClash: agents maintain codebases, compete in arenas, and iterate over multiple rounds), the proliferation of SWE-bench variants (SWE-bench Pro, SWE-bench Live, SWE-Efficiency, AlgoTune, SciCode) and how benchmark authors are now justifying their splits with curation techniques instead of just "more repos," why Tau-bench's "impossible tasks" controversy is actually a feature not a bug (intentionally including impossible tasks flags cheating), the tension between long autonomy (5-hour runs) vs. interactivity (Cognition's emphasis on fast back-and-forth), how Terminal-bench unlocked creativity by letting PhD students and non-coders design environments beyond GitHub issues and PRs, the academic data problem (companies like Cognition and Cursor have rich user interaction data, academics need user simulators or compelling products like LMArena to get similar signal), and his vision for CodeClash as a testbed for human-AI collaboration—freeze model capability, vary the collaboration setup (solo agent, multi-agent, human+agent), and measure how interaction patterns change as models climb the ladder from code completion to full codebase reasoning. We discuss: John's path: Princeton → SWE-bench (October 2023) → Stanford PhD with Diyi Yang and the Iris Group, focusing on code evals, human-AI collaboration, and long-running agent benchmarks The SWE-bench origin story: released October 2023, mostly ignored until Cognition's Devin launch kicked off the arms race (Walden emailed John two weeks before: "we have a good number") SWE-bench Verified: the curated, high-quality split that became the standard for serious evals SWE-bench Multimodal and Multilingual: nine languages (JavaScript, Rust, Java, C, Ruby) across 40 repos, moving beyond the Django-heavy original distribution The SWE-bench Pro controversy: independent authors used the "SWE-bench" name without John's blessing, but he's okay with it ("congrats to them, it's a great benchmark") CodeClash: John's new benchmark for long-horizon development—agents maintain their own codebases, edit and improve them each round, then compete in arenas (programming games like Halite, economic tasks like GDP optimization) SWE-Efficiency (Jeffrey Maugh, John's high school classmate): optimize code for speed without changing behavior (parallelization, SIMD operations) AlgoTune, SciCode, Terminal-bench, Tau-bench, SecBench, SRE-bench: the Cambrian explosion of code evals, each diving into different domains (security, SRE, science, user simulation) The Tau-bench "impossible tasks" debate: some tasks are underspecified or impossible, but John thinks that's actually a feature (flags cheating if you score above 75%) Cognition's research focus: codebase understanding (retrieval++), helping humans understand their own codebases, and automatic context engineering for LLMs (research sub-agents) The vision: CodeClash as a testbed for human-AI collaboration—vary the setup (solo agent, multi-agent, human+agent), freeze model capability, and measure how interaction changes as models improve — John Yang SWE-bench: https://www.swebench.com X: https://x.com/jyangballin Chapters 00:00:00 Introduction: John Yang on SWE-bench and Code Evaluations 00:00:31 SWE-bench Origins and Devon's Impact on the Coding Agent Arms Race 00:01:09 SWE-bench Ecosystem: Verified, Pro, Multimodal, and Multilingual Variants 00:02:17 Moving Beyond Django: Diversifying Code Evaluation Repositories 00:03:08 Code Clash: Long-Horizon Development Through Programming Tournaments 00:04:41 From Halite to Economic Value: Designing Competitive Coding Arenas 00:06:04 Ofir's Lab: SWE-ficiency, AlgoTune, and SciCode for Scientific Computing 00:07:52 The Benchmark Landscape: TAU-bench, Terminal-bench, and User Simulation 00:09:20 The Impossible Task Debate: Refusals, Ambiguity, and Benchmark Integrity 00:12:32 The Future of Code Evals: Long Autonomy vs Human-AI Collaboration 00:14:37 Call to Action: User Interaction Data and Codebase Understanding Research

From creating SWE-bench in a Princeton basement to shipping CodeClash, SWE-bench Multimodal, and SWE-bench Multilingual, John Yang has spent the last year and a half watching his benchmark become the de facto standard for evaluating AI coding agents—trusted by Cognition (Devin), OpenAI, Anthropic, and every major lab racing to solve software engineering at scale. We caught up with John live at NeurIPS 2025 to dig into the state of code evals heading into 2026: why SWE-bench went from ignored (October 2023) to the industry standard after Devin's launch (and how Walden emailed him two weeks before the big reveal), how the benchmark evolved from Django-heavy to nine languages across 40 repos (JavaScript, Rust, Java, C, Ruby), why unit tests as verification are limiting and long-running agent tournaments might be the future (CodeClash: agents maintain codebases, compete in arenas, and iterate over multiple rounds), the proliferation of SWE-bench variants (SWE-bench Pro, SWE-bench Live, SWE-Efficiency, AlgoTune, SciCode) and how benchmark authors are now justifying their splits with curation techniques instead of just “more repos,” why Tau-bench's “impossible tasks” controversy is actually a feature not a bug (intentionally including impossible tasks flags cheating), the tension between long autonomy (5-hour runs) vs. interactivity (Cognition's emphasis on fast back-and-forth), how Terminal-bench unlocked creativity by letting PhD students and non-coders design environments beyond GitHub issues and PRs, the academic data problem (companies like Cognition and Cursor have rich user interaction data, academics need user simulators or compelling products like LMArena to get similar signal), and his vision for CodeClash as a testbed for human-AI collaboration—freeze model capability, vary the collaboration setup (solo agent, multi-agent, human+agent), and measure how interaction patterns change as models climb the ladder from code completion to full codebase reasoning.We discuss:* John's path: Princeton → SWE-bench (October 2023) → Stanford PhD with Diyi Yang and the Iris Group, focusing on code evals, human-AI collaboration, and long-running agent benchmarks* The SWE-bench origin story: released October 2023, mostly ignored until Cognition's Devin launch kicked off the arms race (Walden emailed John two weeks before: “we have a good number”)* SWE-bench Verified: the curated, high-quality split that became the standard for serious evals* SWE-bench Multimodal and Multilingual: nine languages (JavaScript, Rust, Java, C, Ruby) across 40 repos, moving beyond the Django-heavy original distribution* The SWE-bench Pro controversy: independent authors used the “SWE-bench” name without John's blessing, but he's okay with it (”congrats to them, it's a great benchmark”)* CodeClash: John's new benchmark for long-horizon development—agents maintain their own codebases, edit and improve them each round, then compete in arenas (programming games like Halite, economic tasks like GDP optimization)* SWE-Efficiency (Jeffrey Maugh, John's high school classmate): optimize code for speed without changing behavior (parallelization, SIMD operations)* AlgoTune, SciCode, Terminal-bench, Tau-bench, SecBench, SRE-bench: the Cambrian explosion of code evals, each diving into different domains (security, SRE, science, user simulation)* The Tau-bench “impossible tasks” debate: some tasks are underspecified or impossible, but John thinks that's actually a feature (flags cheating if you score above 75%)* Cognition's research focus: codebase understanding (retrieval++), helping humans understand their own codebases, and automatic context engineering for LLMs (research sub-agents)* The vision: CodeClash as a testbed for human-AI collaboration—vary the setup (solo agent, multi-agent, human+agent), freeze model capability, and measure how interaction changes as models improve—John Yang* SWE-bench: https://www.swebench.com* X: https://x.com/jyangballinFull Video EpisodeTimestamps00:00:00 Introduction: John Yang on SWE-bench and Code Evaluations00:00:31 SWE-bench Origins and Devon's Impact on the Coding Agent Arms Race00:01:09 SWE-bench Ecosystem: Verified, Pro, Multimodal, and Multilingual Variants00:02:17 Moving Beyond Django: Diversifying Code Evaluation Repositories00:03:08 Code Clash: Long-Horizon Development Through Programming Tournaments00:04:41 From Halite to Economic Value: Designing Competitive Coding Arenas00:06:04 Ofir's Lab: SWE-ficiency, AlgoTune, and SciCode for Scientific Computing00:07:52 The Benchmark Landscape: TAU-bench, Terminal-bench, and User Simulation00:09:20 The Impossible Task Debate: Refusals, Ambiguity, and Benchmark Integrity00:12:32 The Future of Code Evals: Long Autonomy vs Human-AI Collaboration00:14:37 Call to Action: User Interaction Data and Codebase Understanding Research Get full access to Latent.Space at www.latent.space/subscribe

From building LMArena in a Berkeley basement to raising $100M and becoming the de facto leaderboard for frontier AI, Anastasios Angelopoulos returns to Latent Space to recap 2025 in one of the most influential platforms in AI—trusted by millions of users, every major lab, and the entire industry to answer one question: which model is actually best for real-world use cases? We caught up with Anastasios live at NeurIPS 2025 to dig into the origin story (spoiler: it started as an academic project incubated by Anjney Midha at a16z, who formed an entity and gave grants before they even committed to starting a company), why they decided to spin out instead of staying academic or nonprofit (the only way to scale was to build a company), how they're spending that $100M (inference costs, React migration off Gradio, and hiring world-class talent across ML, product, and go-to-market), the leaderboard delusion controversy and why their response demolished the paper's claims (factual errors, misrepresentation of open vs. closed source sampling, and ignoring the transparency of preview testing that the community loves), why platform integrity comes first (the public leaderboard is a charity, not a pay-to-play system—models can't pay to get on, can't pay to get off, and scores reflect millions of real votes), how they're expanding into occupational verticals (medicine, legal, finance, creative marketing) and multimodal arenas (video coming soon), why consumer retention is earned every single day (sign-in and persistent history were the unlock, but users are fickle and can leave at any moment), the Gemini Nano Banana moment that changed Google's market share overnight (and why multimodal models are becoming economically critical for marketing, design, and AI-for-science), how they're thinking about agents and harnesses (Code Arena evaluates models, but maybe it should evaluate full agents like Devin), and his vision for Arena as the central evaluation platform that provides the North Star for the industry—constantly fresh, immune to overfitting, and grounded in millions of real-world conversations from real users. We discuss: The $100M raise: use of funds is primarily inference costs (funding free usage for tens of millions of monthly conversations), React migration off Gradio (custom loading icons, better developer hiring, more flexibility), and hiring world-class talent The scale: 250M+ conversations on the platform, tens of millions per month, 25% of users do software for a living, and half of users are now logged in The leaderboard illusion controversy: Cohere researchers claimed undisclosed private testing created inequities, but Arena's response demolished the paper's factual errors (misrepresented open vs. closed source sampling, ignored transparency of preview testing that the community loves) Why preview testing is loved by the community: secret codenames (Gemini Nano Banana, named after PM Naina's nickname), early access to unreleased models, and the thrill of being first to vote on frontier capabilities The Nano Banana moment: changed Google's market share overnight, billions of dollars in stock movement, and validated that multimodal models (image generation, video) are economically critical for marketing, design, and AI-for-science New categories: occupational and expert arenas (medicine, legal, finance, creative marketing), Code Arena, and video arena coming soon Consumer retention: sign-in and persistent history were the unlock, but users are fickle and earned every single day—"every user is earned, they can leave at any moment" — Anastasios Angelopoulos Arena: https://lmarena.ai X: https://x.com/arena Chapters 00:00:00 Introduction: Anastasios from Arena and the LM Arena Journey 00:01:36 The Anjney Midha Incubation: From Berkeley Basement to Startup 00:02:47 The Decision to Start a Company: Scaling Beyond Academia 00:03:38 The $100M Raise: Use of Funds and Platform Economics 00:05:10 Arena's User Base: 5M+ Users and Diverse Demographics 00:06:02 The Competitive Landscape: Artificial Analysis, AI.xyz, and Arena's Differentiation 00:08:12 Educational Value and Learning from the Community 00:08:41 Technical Migration: From Gradio to React and Platform Evolution 00:10:18 Leaderboard Delusion Paper: Addressing Critiques and Maintaining Integrity 00:12:29 Nano Banana Moment: How Preview Models Create Market Impact 00:13:41 Multimodal AI and Image Generation: From Skepticism to Economic Value 00:15:37 Core Principles: Platform Integrity and the Public Leaderboard as Charity 00:18:29 Future Roadmap: Expert Categories, Multimodal, Video, and Occupational Verticals 00:19:10 API Strategy and Focus: Doing One Thing Well 00:19:51 Community Management and Retention: Sign-In, History, and Daily Value 00:22:21 Partnerships and Agent Evaluation: From Devon to Full-Featured Harnesses 00:21:49 Hiring and Building a High-Performance Team

Wildest week in AI since December 2024.

You ever see a new AI model drop and be like.... it's so good OMG how do I use it?