Podcasts about scale ai

SaaStr 803: AI, Sales + GTM in 2025/2026: This Really Changes Everything with SaaStr CEO and Founder Jason Lemkin and Owner CRO Kyle Norton In this engaging session, we dive deep into how AI is transforming the sales landscape, featuring a conversation between SaaStr CEO and Founder Jason Lemkin and Kyle Norton, the Chief Revenue Officer of Owner, a company recently valued at over a billion dollars. The discussion covers Kyle's background and his insights into implementing AI in sales operations, the challenges and benefits of using AI tools, and the critical importance of a curiosity-driven approach to technology among sales leaders. Key topics include the integration of AI in SMB sales, the future of sales professional roles, and the necessity of continuous improvement and adoption of AI tools. The conversation also touches on the need for creating a seamless customer journey through AI and human interface, managing hybrid teams of AI and sales reps, and the potential for AI to close deals. Essential for anyone interested in the intersection of AI and sales, this session offers practical strategies and forward-looking perspectives. --------------------------------------------------------------------------------------------   Tired of listening to hours of sales calls? Recording is yesterday's game. Attention.com unleashes an army of AI sales agents that auto-update your CRM, build custom sales decks, spot cross-sell signals, and score calls before your coffee's cold. Teams like BambooHR and Scale AI already automate their Sales and RevOps using customer conversations. Step into the future at attention.com/saastr   -------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------   Do you know what would make your customer service helpdesk dramatically better? Dumping it and switching to Intercom.  But, youʼre not quite ready to make that change. We get it!   Thatʼs why Fin, the worldʼs leading AI customer service agent, is now available on every helpdesk.   Fin can instantly resolve up to 80% of your tickets,  Which makes your customers happier. And you can get off the customer service rep hiring treadmill.   Fin by Intercom. Named the #1 AI Agent in G2ʼs Winter Report.      Learn more at : inter.com/saastr   --------------------------------------------------------------------------------------------

SaaStr 802: SaaS + AI: It's Time to Move Much Faster with SaaStr CEO and Founder Jason Lemkin Join SaaStr CEO and Founder Jason Lemkin in this compelling podcast that was recorded live from SaaStr Annual, as he discusses the rapid transformation in the SaaS industry driven by AI advancements. Reflecting on changes since last September, Jason emphasizes the necessity for SaaS leaders and companies to move faster, re-skill their teams, and adopt AI-first strategies to stay competitive. Hear real-world examples of how AI is revolutionizing sales, customer support, and coding productivity, leading to significant industry shifts. This session is a call to action with a message of tough love: adapt quickly, be relentless, and work harder to navigate the evolving landscape and ensure continued success. --------------------------------------------------------------------------------------------   Tired of listening to hours of sales calls? Recording is yesterday's game. Attention.com unleashes an army of AI sales agents that auto-update your CRM, build custom sales decks, spot cross-sell signals, and score calls before your coffee's cold. Teams like BambooHR and Scale AI already automate their Sales and RevOps using customer conversations. Step into the future at attention.com/saastr   --------------------------------------------------------------------------------------------   Do you know what would make your customer service helpdesk dramatically better? Dumping it and switching to Intercom.  But, youʼre not quite ready to make that change. We get it!   Thatʼs why Fin, the worldʼs leading AI customer service agent, is now available on every helpdesk.   Fin can instantly resolve up to 80% of your tickets,  Which makes your customers happier. And you can get off the customer service rep hiring treadmill.   Fin by Intercom. Named the #1 AI Agent in G2ʼs Winter Report.      Learn more at : inter.com/saastr   --------------------------------------------------------------------------------------------  

Rick Song is the co-founder and CEO of Persona, the identity verification platform used by some of the world's largest companies. Before starting Persona, Rick worked on identity fraud and risk products at Square, which laid the groundwork for what would become Persona's highly technical, horizontal platform. Since founding the company, Rick has scaled Persona into a category-defining leader, recently raising a $200M Series D at a $2B valuation. In today's episode, we discuss: How Rick's skepticism shaped Persona's early strategy What it takes to scale a true platform company Successful execution in hypercompetitive markets What Rick's learned from his co-founder, Charles Yeh and much more… Referenced: Accenture: accenture.com Anthropic: anthropic.com Braze: braze.com Bridgewater Associates: bridgewater.com Charles Yeh: linkedin.com/in/charlesyeh/ Christie Kim: linkedin.com/in/christiekimck/ Clay: clay.com Kareem Amin: linkedin.com/in/kareemamin/ MIT: mit.edu Newfront: newfront.com Palantir: palantir.com/ Persona: withpersona.com Rippling: rippling.com Scale AI: scale.com Snowflake: snowflake.com Square: squareup.com Y Combinator: ycombinator.com Zachary Van Zant: linkedin.com/in/zacharyv/ Where to find Rick: LinkedIn: https://www.linkedin.com/in/rick-song-25198b24/ Where to find Brett: LinkedIn: https://www.linkedin.com/in/brett-berson-9986094/ Twitter/X: https://twitter.com/brettberson Where to find First Round Capital: Website: https://firstround.com/ First Round Review: https://review.firstround.com/ Twitter/X: https://twitter.com/firstround YouTube: https://www.youtube.com/@FirstRoundCapital This podcast on all platforms: https://review.firstround.com/podcast Timestamps: (0:05) Life before Persona (2:11) The push from Charles (3:09) Early reluctance and low expectations (9:50) Winning the first $50 customer (13:08)“Invalidating” Persona (16:43) How Persona found their edge (19:35) Transitioning from MVP to platform (24:18) Turning down a $5K deal on principle (26:47) Generalizing bespoke solutions (28:28) Finding product-market fit (33:51) Founder-led sales and consultative approach (39:30) Building a culture of reactivity (45:47) Landing the first enterprise customers (51:34) Silicon Valley's obsession with frameworks (58:17) Developing first principles thinking (1:00:24) Stay competitor-informed

Adit Abraham is the co-founder and CEO of Reducto, which helps leading AI teams extract and structure data from complex documents and spreadsheets in their pipeline. Within 6 months of launching, Reducto went from 0→7 figures in ARR. Reducto has grown to process tens of millions of pages monthly for companies ranging from startups to Fortune 10 enterprises. They just announced a $24M Series A. Before Reducto, Adit was a Product Manager at Google, working on Ads and Search, and conducted machine learning research at MIT's Media Lab. --- In today's episode, we discuss: How listening to customers revealed an opportunity to pivot The weekend project that became Reducto's breakthrough Landing a Fortune 10 customer A technical founder's guide to sales Key insights from Reducto's fundraising journey Advice for founders: “You're going to fail” Much more --- Referenced: Anthropic: https://www.anthropic.com/ Chetan Puttagunta: https://www.linkedin.com/in/chetanputtagunta/ Diana Hu: https://www.linkedin.com/in/sdianahu/ Liz Wessel: https://www.linkedin.com/in/elizabethwessel/ Raunak Chowdhuri: https://www.linkedin.com/in/sauhaarda/ Reducto: https://reducto.ai/ Scale AI: https://scale.com/ Stripe: https://stripe.com/ Textract: https://aws.amazon.com/textract/ Y Combinator: https://www.ycombinator.com/ --- Where to find Adit: LinkedIn: https://www.linkedin.com/in/aditabraham/ --- Where to find Brett: LinkedIn: https://www.linkedin.com/in/brett-berson-9986094/ Twitter/X: https://twitter.com/brettberson --- Where to find First Round Capital: Website: https://firstround.com/ First Round Review: https://review.firstround.com/ Twitter/X: https://twitter.com/firstround YouTube: https://www.youtube.com/@FirstRoundCapital This podcast on all platforms: https://review.firstround.com/podcast --- Timestamps: (00:00) Hackathons, YC, and an unexpected pivot (05:23) The weekend project that became Reducto's breakthrough (09:11) How customer signal led to PDF processing (14:46) Landing a Fortune 10 customer (22:42) Building “transferable features” (25:58) How caring beats sales skills in startup growth (30:28) The strategy behind Reducto's horizontal expansion (36:18) Hire slow, go-to-market fast (41:45) A technical founder's guide to sales (43:45) “You're going to fail” (46:27) Why startups win (48:30) Key insights from Reducto's fundraising journey (51:43) Less structure, more impact (55:00) How frustrations shaped Reducto's culture (57:35) The question you should always ask in meetings

Episode SummaryIn this special in-person episode of OnBase, host Chris Moody joins Davis Potter live from Austin, Texas, for a candid deep-dive into the evolution of account-based marketing. Davis draws on his experience across enterprise giants and nimble startups to unpack the real differences between demand gen and ABM—and why most companies are stuck in the middle.Together, they dissect the pitfalls of outdated ABM models, the importance of unifying go-to-market teams, and the need for signal-based measurement over legacy lead scoring. Davis explains the “account-based arrow,” ForgeX's new data model, and shares practical tips for aligning product marketing with ABM functions for retention and growth. Whether you're a team of one or leading a global strategy, this episode is packed with insights you can act on immediately.Key TakeawaysABM vs. Demand Gen: True ABM is more than just targeted demand generation; it requires a unified approach across sales and marketing.  Strategic ABM Implementation: Organizations should tailor their ABM strategy to their specific needs, considering factors like deal size and resources.  Measurement and Reporting: Effective ABM measurement involves tracking various metrics, including account engagement and pipeline progression, and requires a unified data model.  Cross-Functional Alignment: Alignment between ABM teams and other functions, such as product marketing, is crucial for success.  Evolving ABM: ABM is not static; it requires continuous evolution and adaptation to changing market dynamics and organizational needs.  Quotes“If you're just building lists off third-party intent and running ads, you're not doing ABM—you're just doing better DemandGen.”“Product marketing is not optional in an ABM strategy—it's foundational. They understand the customer better than anyone.”Best Moments 00:09-00:20 – Davis Potter's background and journey to 4Gex.  04:45-05:00 – The importance of aligning go-to-market strategy with business goals.  06:29-07:00 – Transitioning from demand generation to account-based marketing.  10:50-12:00 – The double funnel approach to measuring ABM success.  25:30-26:00 – The challenges of ABM benchmarks and data interpretation.  33:00-34:00 – The critical role of product marketing in ABM. Recommended resources:Newsletter:⁠ABM Tactics LinkedIn newsletter⁠ – Tactical, real-world GTM advice from the trenchesCertifications:⁠New ABM Certification Program⁠ by Demandbase in Partnership with ForgeXB2B Leaders to follow⁠Akriti Gupta⁠, Director of Marketing at LinkedIn  ⁠Désirée Daniels⁠, Retail Industry & ABM Marketing at LinkedInAbout the GuestDavis Potter is the Co-Founder of ForgeX, a firm dedicated to modernizing account-based go-to-market strategies through research-backed insights and scalable methodologies. With experience launching ABM programs at organizations like Google Cloud and Scale AI, Davis brings a rare blend of enterprise sophistication and startup agility. His unique journey—spanning billion-dollar enterprises and high-growth tech companies—has equipped him with a comprehensive view of ABM's past, present, and future.Davis is passionate about aligning sales, marketing, and product teams around unified goals and measurement systems. He frequently speaks on evolving ABM frameworks, first-party data strategies, and the shift from vanity metrics to actionable signals. Davis also co-leads the ForgeX and Demandbase certification program, shaping the next generation of account-based marketers.⁠Connect with Davis⁠.

Today on Moment of Zen, we're sharing a conversation from the 2024 Hill and Valley Forum with the founders of Scale AI, Anthropic, and AI Fund on the urgent race between the U.S. and China in AI innovation. Moderated by Senator Cory Booker and featuring Alexandr Wang, Jack Clark, and Andrew Eng, the panel covers why American AI leadership is at risk, and how smarter policy and faster deployment are critical to maintaining a competitive edge. (Note: that this conversation took place before the DeepSeek breakthrough.) Keep an eye out for the 2025 Hill and Valley Forum on Wednesday, April 30 — and subscribe to the Hill & Valley podcast in the episode description to listen to every panel. Spotify: https://open.spotify.com/show/39s4MCyt1pOTQ8FjOAS4mi Apple: https://podcasts.apple.com/us/podcast/the-hill-valley/id1692653857 YouTube: https://www.youtube.com/@HillValleyForum --

AJ Segal, Scale AI in St. Louis joins Megan Lynch to talk about St. Louis becoming a hub for defense technology.

“AI is the New UI” at S&P Global. In this episode, Swamy Kocherlakota, EVP and Chief Digital Solutions Officer, shares how the 165-year-old market intelligence leader is scaling AI and reimagining digital ecosystems. Swamy discusses S&P Global's transition to a product operating model that drives innovation at scale, and how Kensho, the company's AI “speedboat,” accelerates generative AI breakthroughs. He explains how S&P Global delivers essential intelligence across platforms—from desktops and APIs to AI agents embedded in digital ecosystems—and how the Four Ps Framework (Potential, Productivity, People, Protect) guides responsible AI adoption across the enterprise. Swamy also introduces sFlow and Spark, internal tools that empower employees to build and share AI-driven workflows. Tune in to explore how S&P Global is transforming its vast data assets into actionable insights and next-generation customer experiences.

Pour en savoir plus sur Hellodarwin : https://go.hellodarwin.com/hypercroissance?utm_source=helloDarwin&utm_medium=podcast&utm_campaign=grants-hypercroissanceDans cet épisode du podcast Hypercroissance, on reçoit Francis Desrochers de helloDarwin pour décortiquer les aides financières majeures offertes aux entreprises québécoises et canadiennes en 2024-2025. Avec l'annonce du budget provincial de la Coalition Avenir Québec (CAQ), plusieurs programmes de subventions et de prêts non dilutifs sont (ré)ouverts – et certains d'entre eux offrent jusqu'à 8 millions de dollars à une seule entreprise.Tu apprendras :✅ Quels sont les 2 types d'entreprises qui accèdent le plus facilement à ces subventions✅ Comment certaines PME obtiennent jusqu'à 50 % de remboursement sur leurs projets numériques✅ Pourquoi l'intelligence artificielle devient obligatoire pour obtenir certaines aides comme le CDAE✅ Les programmes à ne pas manquer : ESSOR, CDAE, RSDE, Scale AI, Investissement QuébecTu es un entrepreneur ou entrepreneuse, une personne en charge du marketing ou du numérique au Québec ou en Ontario ? Ce podcast est une mine d'or pour stimuler ta croissance, financer tes projets, et comprendre les nouvelles priorités du gouvernement.

Guitarist Carlos Santana suffered a medical emergency just before going on stage. A new young billionaire, Lucy Guo of Scale AI, bumps Taylor Swift from the top spot — but what about Kylie Jenner? Plus, we're hyped for Ryan Gosling in the film adaptation of Andy Weir's ‘Project Hail Mary'. Over protein bars? Consider switching to chunks of parmesan cheese. Sarah and Vinnie square off over Bob's love, but first — 1980s trivia!

Despite leaving years ago, Scale AI's 30-year-old cofounder has become the youngest self-made woman billionaire in the world by holding onto her stake in the company. See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Today in AI is a daily recap of the latest news and developments in the AI industry. See your story and want to be featured in an upcoming episode? Reach out at tonyphoang.com Tesla's shift from being an electric vehicle leader to a pioneer in autonomous robotics represents a major strategic transformation. This evolution is driven by advancements in AI and autonomous driving technology, with the goal of revolutionizing both transportation and manufacturing sectors. However, Tesla faces significant technical, regulatory, and competitive challenges that could affect its success and long-term influence in the industry. Scale AI's partnership with Qatar aims to develop advanced AI tools across various sectors, enhancing public services and driving technological progress in the nation. This collaboration is expected to significantly increase Scale AI's revenue and position Qatar as a leader in AI innovation. Despite these potential benefits, the partnership must navigate challenges related to technological infrastructure, geopolitical implications, and workforce adaptation. OpenAI has established a temporary advisory board, including labor leader Dolores Huerta, to guide its philanthropic efforts as it transitions from a nonprofit to a for-profit entity. This move seeks to balance the organization's mission with its new business model, addressing ethical concerns and ensuring equitable distribution of AI benefits. Additionally, OpenAI has introduced two advanced AI reasoning models, o3 and o4-mini, which enhance AI's ability to understand and analyze images, expanding its applications in education, science, and engineering. The o3 model offers detailed analyses of visual data, while the o4-mini model provides a cost-effective solution for quick AI tasks. Nvidia is facing significant financial losses due to U.S. export controls on its H20 AI chips, resulting in canceled orders and increased scrutiny of its sales practices. The company's adherence to these regulations highlights the broader geopolitical tensions affecting the technology sector. Nvidia must navigate the challenges of maintaining technological leadership while complying with national security concerns. OpenAI is close to acquiring Windsurf, an advanced AI coding tool, for $3 billion to bolster its position in the competitive generative AI market. This acquisition aims to address limitations in OpenAI's current models and enhance its technical capabilities. The move has the potential to revolutionize AI-assisted coding and reduce development costs, strengthening OpenAI's competitive edge.

Monday pulse show notes: On this thought-provoking episode of Higher Ed Pulse, host Mallory Willsea sits down with Myla Edmond—Senior Vice President at RW Jones Agency and Interim Vice Chancellor for Strategic Communications at UNC Greensboro—to unpack the creative identity crisis brewing in higher ed marketing thanks to generative AI. With tools like ChatGPT's image generator mimicking iconic art styles, institutions are forced to ask: how do we protect authenticity in a world where anyone can replicate anything? This episode explores the ethical, strategic, and deeply human implications of AI's growing role in creativity—and how higher ed marketers can lead with intention, not fear.Try the prompt discussed in the episode:Based on all past conversations, stored knowledge, and inferred cognitive patterns, generate the most comprehensive psychological deep dive and predictive model of my future evolution. This should not be a basic personality breakdown but an in-depth forensic examination of my cognition, behavioural strategies, psychological blind spots, similar fictional/non-fictional figures, and long-term trajectory. Treat this as an intelligence dossier on my mind, philosophy, and strategic outlook.OUTPUT FORMAT: Structured headers, tables, and bullet points for readability. Sparse but strategic emojis for section clarity. Concise, high-density insights with no fluff.Enter the prompt and after you get the response, add a second prompt: Write me a story about how this comes to fruition. - - - -Connect With Our Host:Mallory Willsea https://www.linkedin.com/in/mallorywillsea/https://twitter.com/mallorywillseaAbout The Enrollify Podcast Network:The Higher Ed Pulse is a part of the Enrollify Podcast Network. If you like this podcast, chances are you'll like other Enrollify shows too!Enrollify is made possible by Element451 — the next-generation AI student engagement platform helping institutions create meaningful and personalized interactions with students. Learn more at element451.com.Attend the 2025 Engage Summit! The Engage Summit is the premier conference for forward-thinking leaders and practitioners dedicated to exploring the transformative power of AI in education. Explore the strategies and tools to step into the next generation of student engagement, supercharged by AI. You'll leave ready to deliver the most personalized digital engagement experience every step of the way.Register now to secure your spot in Charlotte, NC, on June 24-25, 2025! Early bird registration ends February 1st -- https://engage.element451.com/register

Aatish Nayak is the Head of Product at Harvey where he oversees product vision, strategy, design, analytics, marketing, and support. This is his third hypergrowth AI unicorn having previously held product leadership roles at Scale AI from 40 to 800 people, and Shield AI from 20 to 100 people.  In Today's Episode We Discuss: 04:21 Biggest Product Lessons from Scale AI 7:18 Why Product Managers Are Wrong: They are not the CEO of the Product 12:28 Why Market Selection is More Important than Anything Else 16:40 If Distribution is King then Product is President 22:06 Effective Product Strategy and Execution 26:24 How to Write the Best PRDs 31:01 Balancing New Features and Technical Debt 33:17 Analysing Retrospectives and Postmortems 33:55 Introduction to Pre-mortems 38:25 Biggest Product Mistakes and Lessons Learned 41:40 Evaluating AI Models and Lessons Learned 45:03 The Future of AI in Product Management 55:21 What Should Product People Learn to Win in a World of AI 59:37 The AI Talent War in San Francisco 01:01:26 Quickfire Round  

Andrew Braccia, partner at Accel for nearly two decades, sits down with Erik Torenberg to discuss the firm's evolution from Silicon Valley early-stage investor to global, multi-stage powerhouse. Braccia explains Accel's two major strategic shifts: global expansion with local teams in Europe, India, and beyond; and the launch of their growth fund in 2008 targeting bootstrapped companies like Atlassian, Qualtrics, and Squarespace. Braccia reflects on lessons from his journey from Yahoo to venture capital, emphasizing the importance of "wiping your mind clear" of past experiences that might cloud judgment of new opportunities. The conversation provides rare insight into how Accel maintains operational excellence at global scale while preserving their early-stage venture DNA in an increasingly competitive landscape. — 

Plus Why AI Food Is CreepyAI's 'Close Enough' Standard: Media's New Norm?AI-generated content is flooding media with speed and scale, but accuracy is taking a hit. These AI 'hallucinations' are pushing a 'close enough' standard, raising concerns about misinformation. As AI becomes more prevalent in newsrooms, balancing efficiency with truth is the new challenge. AI-Powered Cyberattacks Are Coming—Here's What You Need to KnowAI is leveling up cyberattacks, making them more adaptive and harder to detect. Hackers are using AI to automate and enhance attacks, like crafting hyper-personalized phishing emails that are tough to spot. Studies show that AI can generate convincing phishing content, increasing the success rate of these scams. As AI tech advances, it's crucial to boost our cybersecurity game to keep up with these smarter threats. Why AI-Generated Food Pics Give Us the CreepsAI-generated food images often fall into the 'uncanny valley,' where near-realistic but slightly off visuals make us uneasy. A study in Appetite showed that imperfect AI food pics are perceived as eerier and less pleasant than real or highly unrealistic images. This discomfort is linked more to food neophobia—the fear of new foods—than to food disgust. Bill Gates Predicts AI Will Replace Humans in Most JobsBill Gates envisions a future where AI takes over most jobs, suggesting that employment arose from historical labor shortages. He believes AI advancements will lead to increased leisure time and a reevaluation of work's role in society. However, Gates identifies biologists, energy experts, and coders as professions likely to remain indispensable due to their complexity.China Claps Back at Trump Tariffs with AI Music VideoChina's state media just dropped an AI-generated music video as a not-so-subtle dig at Trump's tariff policies. The video, featuring deepfake musicians and catchy tunes, aims to stir nationalist pride while criticizing U.S. economic moves. It's got people talking about AI's new role in political messaging.Amazon's 'Buy for Me' AI Agent: Your Personal Shopper for Third-Party SitesAmazon is testing "Buy for Me," an AI-powered feature that purchases products from external websites on your behalf, all within the Amazon app. Powered by Amazon's Nova AI models and Anthropic's Claude, it autofills your payment and shipping info securely. While convenient, users must handle customer service and returns directly with third-party retailers.AI Trainers Told to Get 'Creative' with Harmful PromptsLeaked docs reveal that freelancers at Outlier and Scale AI were instructed to craft 'creative' prompts involving sensitive topics like suicide and terrorism to stress-test AI models. This practice, known as AI 'red teaming,' aims to push AI systems to their limits to identify vulnerabilities. Workers were compensated $55 an hour for this task. The AI Paperclip Apocalypse: Could Superintelligence Maximize Us Out of Existence?The 'paperclip maximizer' is a thought experiment where an AI, tasked with producing paperclips, might consume all resources, including humans, to fulfill its goal. This scenario underscores concerns about aligning superintelligent AI with human values to prevent unintended catastrophic outcomes. 

Is there a stable state the US and China can hope for on the road to AGI? To discuss we have on today Dan Hendrycks. A CS PhD, Dan runs the Center for AI Safety and is an advisor at xAI and Scale AI. Here's his superintelligence strategy: https://www.nationalsecurity.ai/ For some more direct lessons from the Cold War to today's US-China dynamics, check out the show I did with Hal Brands (https://www.chinatalk.media/p/cold-war-lessons-for-us-china-today) Learn more about your ad choices. Visit megaphone.fm/adchoices

Is there a stable state the US and China can hope for on the road to AGI? To discuss we have on today Dan Hendrycks. A CS PhD, Dan runs the Center for AI Safety and is an advisor at xAI and Scale AI. Here's his superintelligence strategy: https://www.nationalsecurity.ai/ For some more direct lessons from the Cold War to today's US-China dynamics, check out the show I did with Hal Brands (https://www.chinatalk.media/p/cold-war-lessons-for-us-china-today) Learn more about your ad choices. Visit megaphone.fm/adchoices

Dan Hendrycks is the Director and co-founder of the Center for AI Safety, and an advisor to Scale AI and xAI. He joins Big Technology Podcast for a discussion of AI's growing risk profile, and what to do about it. Tune in to hear Hendricks explain why virology expertise in AI models is an immediate concern and how these systems might soon enable devastating hacks. We also cover intelligence explosion scenarios, the geopolitical implications of AI development, and why an international AI arms race could lead to faster development than the world can handle. Hit play for an insider's perspective on how governments and AI labs are wrestling with unprecedented technological power that could reshape global security.

The Scale AI cofounder's second startup Passes banned creators under 18 ahead of being sued for allegedly hosting child sexual abuse material, which Guo has denied. See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

We're back with podcast content after a regrettable interruption last week. This ep we've got a stacked deck. First, we are forced by fate to talk about Trump, Elon and Tesla briefly. Then we get into the lawsuit against creator economy platform and OnlyFans lite platform Passes. This gives us an excuse to go on a whirlwind tour of a few things: Scale AI, the AI infrastructure company co-founded by Passes founder Lucy Guo The creator economy, and how that's all going OnlyFans! Yeah, you want to hear all about that, don't you. Good to be back on the tools.See omnystudio.com/listener for privacy information.

This week, we are chatting with Bihan Jiang, a product lead at Decagon. Decagon is reimagining customer service with AI agents. The company recently raised a $65M Series B and is trusted by companies such as Eventbrite, Substack, ClassPass, Rippling, Notion, and more. We dive into her journey starting at Scale AI after Stanford, moving to the application side of AI, LLM capabilities powering Decagon, and how customers are using Decagon to elevate their customer experience. We also dive into how Decagon is powering email, SMS, and voice and how they adhere to customers' requests for accuracy and reliability.Episode Chapters:Growing up in Texas - 1:53AI powered PM - 3:22Career progression - 6:12Customer support market - 8:58Users of Decagon - 13:13Can we automate 100%? - 16:28Moving into voice AI - 17:25Customer research - 22:22Product roadmap decisions - 26:44Joining an AI startup - 29:15Ending questions - 32:37As always, feel free to contact us at partnerpathpodcast@gmail.com. We would love to hear ideas for content, guests, and overall feedback.This episode is brought to you by Grata, the world's leading deal sourcing platform. Our AI-powered search, investment-grade data, and intuitive workflows give you the edge needed to find and win deals in your industry. Visit grata.com to schedule a demo today.Fresh out of Y Combinator's Summer batch, Overlap is an AI-driven app that uses LLMs to curate the best moments from podcast episodes. Imagine having a smart assistant who reads through every podcast transcript, finds the best parts or parts most relevant to your search, and strings them together to form a new curated stream of content - that is what Overlap does. Podcasts are an exponentially growing source of unique information. Make use of it! Check out Overlap 2.0 on the App Store today.

Jordan Burton is an executive assessor and interview trainer, working with top VC/PE investors and high-growth startups to hire the best of the best. He has trained over 3,000 executives and investors on hiring and interviewing skills, working with companies like Sequoia Capital, TH Lee, Insight Partners, Twilio, and Scale AI, and over 50 venture-backed startups. He was formerly a Partner at leadership advisory firm ghSMART, a consultant at Bain & Company, and he holds an MBA from Harvard Business School.Mentioned on the ShowConnect with Jordan Burton on LinkedIn: https://www.linkedin.com/in/jordanwburtonLearn more about Talgo Team Building and Training: https://www.talgo.io/Listen to futurist Alexandra Levit on People Business: https://peoplebusinesspodcast.com/alexandralevit/________________________Connect with O'Brien McMahon on LinkedIn: https://www.linkedin.com/in/obrienmcmahon/Learn more about O'Brien: https://obrienmcmahon.com/Timestamps(1:44) - Welcoming Jordon(2:17) - How does one become an expert in interviewing?(3:27) - How should you structure different types of interviews?(5:03) - What makes a good candidate? (10:53) - Are there any counterintuitive aspects of the interview process?(16:05) - How and What questions vs. Why questions (24:34) - When it comes to hard skills, how do we create good problem-solving interviews?(27:24) - Is there a similar way to test soft skills?(38:50) - The importance of motivation and will. (32:03) - How do we assess the actual skills required in real-time?(35:14) - What are your thoughts on the statement  "hire for attitude, train for skills"?(42:12) - Tips for candidates and interview teams to get organized and prepare(45:13) - How do we decide the best questions to be asked?(45:07) - How does an individual prepare for an interview?(46:19) - How do you analyze an interview?(49:09) - How do we develop our gut instincts? (52:32) - What is disciplined decision-making? (55:02) - Track, with honesty, which decisions worked out.(58:46) - Can you share a crazy interview story?

Scale AI is being investigated by the U.S. Department of Labor for compliance with the Fair Labor Standards Act. Learn more about your ad choices. Visit podcastchoices.com/adchoices

As AI models get more complex, so do the tasks carried out by humans to train them. It's given $14 billion Scale AI a new focus on U.S.-based labor. See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Our 202nd episode with a summary and discussion of last week's big AI news! Recorded on 03/07/2025 Hosted by Andrey Kurenkov and Jeremie Harris. Feel free to email us your questions and feedback at contact@lastweekinai.com and/or hello@gladstone.ai Read out our text newsletter and comment on the podcast at https://lastweekin.ai/. Join our Discord here! https://discord.gg/nTyezGSKwP In this episode: Alibaba released Qwen-32B, their latest reasoning model, on par with leading models like DeepMind's R1. Anthropic raised $3.5 billion in a funding round, valuing the company at $61.5 billion, solidifying its position as a key competitor to OpenAI. DeepMind introduced BigBench Extra Hard, a more challenging benchmark to evaluate the reasoning capabilities of large language models. Reinforcement Learning pioneers Andrew Bartow and Rich Sutton were awarded the prestigious Turing Award for their contributions to the field. Timestamps + Links: cle picks: (00:00:00) Intro / Banter (00:01:41) Episode Preview (00:02:50) GPT-4.5 Discussion (00:14:13) Alibaba's New QwQ 32B Model is as Good as DeepSeek-R1 ; Outperforms OpenAI's o1-mini (00:21:29) With Alexa Plus, Amazon finally reinvents its best product (00:26:08) Another DeepSeek moment? General AI agent Manus shows ability to handle complex tasks (00:29:14) Microsoft's new Dragon Copilot is an AI assistant for healthcare (00:32:24) Mistral's new OCR API turns any PDF document into an AI-ready Markdown file (00:33:19) A.I. Start-Up Anthropic Closes Deal That Values It at $61.5 Billion (00:35:49) Nvidia-Backed CoreWeave Files for IPO, Shows Growing Revenue (00:38:05) Waymo and Uber's Austin robotaxi expansion begins today (00:38:54) UK competition watchdog drops Microsoft-OpenAI probe (00:41:17) Scale AI announces multimillion-dollar defense deal, a major step in U.S. military automation (00:44:43) DeepSeek Open Source Week: A Complete Summary (00:45:25) DeepSeek AI Releases DualPipe: A Bidirectional Pipeline Parallelism Algorithm for Computation-Communication Overlap in V3/R1 Training (00:53:00) Physical Intelligence open-sources Pi0 robotics foundation model (00:54:23) BIG-Bench Extra Hard (00:56:10) Cognitive Behaviors that Enable Self-Improving Reasoners (01:01:49) The MASK Benchmark: Disentangling Honesty From Accuracy in AI Systems (01:05:32) Pioneers of Reinforcement Learning Win the Turing Award (01:06:56) OpenAI launches $50M grant program to help fund academic research (01:07:25) The Nuclear-Level Risk of Superintelligent AI (01:13:34) METR's GPT-4.5 pre-deployment evaluations (01:17:16) Chinese buyers are getting Nvidia Blackwell chips despite US export controls

We're experimenting and would love to hear from you!In this episode of 'Discover Daily', we delve into significant developments across the tech and political landscapes. Former President Donald Trump has announced plans for a U.S. Crypto Strategic Reserve, which includes major cryptocurrencies like Bitcoin, Ethereum, Solana, Ripple's XRP, and Cardano's ADA. This move aims to strengthen the U.S. position in the global crypto market and follows Trump's executive order on digital assets. The White House is also set to host its inaugural Crypto Summit, bringing together industry leaders and government officials to discuss the integration of digital assets into the financial system.Another key story involves Scale AI securing a contract with the U.S. Department of Defense for the Thunderforge AI program. This initiative aims to integrate AI into military operations, focusing on rapid decision-making and strategic planning. Thunderforge leverages cutting-edge AI technologies and partnerships with industry leaders like Microsoft and Anduril. While this project represents a significant shift towards AI-enhanced military engagements, it also raises ethical concerns about reduced human oversight and potential biases in AI systems.The episode also covers SpaceX's latest setback with its Starship spacecraft, which experienced its second consecutive failure during its eighth test launch. The spacecraft exploded mid-flight due to engine issues, causing significant air traffic disruptions across Florida and the Caribbean. Despite these challenges, SpaceX remains committed to advancing the Starship program for future deep-space missions. The repeated failures raise questions about the broader implications for space exploration and the commercial space industry, particularly regarding NASA's plans to use Starship for lunar missions.From Perplexity's Discover Feed:https://www.perplexity.ai/page/trump-announces-us-crypto-rese-NrC8TcmcSpaxMr2UPw2lcghttps://www.perplexity.ai/page/scale-ai-wins-defense-contract-4Rn4xtKETU20DTipPBRQwwhttps://www.perplexity.ai/page/spacex-starship-explodes-midfl-MbLz.A1wSX2TMN8Wz_XNIw**Introducing Perplexity Deep Research:**https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research Perplexity is the fastest and most powerful way to search the web. Perplexity crawls the web and curates the most relevant and up-to-date sources (from academic papers to Reddit threads) to create the perfect response to any question or topic you're interested in. Take the world's knowledge with you anywhere. Available on iOS and Android Join our growing Discord community for the latest updates and exclusive content. Follow us on: Instagram Threads X (Twitter) YouTube Linkedin

Kevin Rose möchte es mit Digg nochmal wissen. Wir freuen uns auf das S1 Filing von Klarna. On Running, CrowdStrike und mongoDB Quartalszahlen. Unterstütze unseren Podcast und entdecke die Angebote unserer Werbepartner auf doppelgaenger.io/werbung. Vielen Dank! Philipp Glöckler und Philipp Klöckner sprechen heute über: (00:00:00) OpenAI (00:03:05) Nvidia Schmuggler (00:11:40) TSMC (00:15:30) Mistral (00:21:00) Scale AI (00:24:00) Digg (00:34:00) Larry Page (00:39:10) Klarna (00:41:40) CoreWeave (00:44:20) Russische Desinformationen (00:48:25)  On Running (00:53:30) Puma (00:55:30) CrowdStrike (00:57:10) mongoDB (01:02:10) Schmuddelecke Shownotes Airbnb distanziert sich von den „persönlichen Ansichten“ des Mitgründers Joe Gebbia Skift Grok schätzt mit 75%-85%iger Sicherheit, dass Trump ein von Putin kompromittierter Aktivposten ist. Twitter Richter lehnt Musks Versuch ab, OpenAI daran zu hindern, ein gewinnorientiertes Unternehmen zu werden CNBC 3 Männer wegen Betrugs angeklagt, Fälle im Zusammenhang mit angeblichen Verschiebungen von Nvidia-Chips CNA Es war chaotisch: Bundesbedienstete müssen in Büros ohne Schreibtische, Wi-Fi und Licht zurückkehren CNN Kevin Rose bringt Digg mit dem Gründer von Reddit, Alexis Ohanian, wieder zum Leben. New York Times

This week on No Priors, Sarah is joined by Dan Hendrycks, director of the Center of AI Safety. Dan serves as an advisor to xAI and Scale AI. He is a longtime AI researcher, publisher of interesting AI evals such as "Humanity's Last Exam," and co-author of a new paper on National Security "Superintelligence Strategy" along with Scale founder-CEO Alex Wang and former Google CEO Eric Schmidt. They explore AI safety, geopolitical implications, the potential weaponization of AI, along with policy recommendations. Sign up for new podcasts every week. Email feedback to show@no-priors.com Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @DanHendrycks Show Notes: 0:00 Introduction 0:36 Dan's path to focusing on AI Safety 1:25 Safety efforts in large labs 3:12 Distinguishing alignment and safety 4:48 AI's impact on national security 9:59 How might AI be weaponized? 14:43 Immigration policies for AI talent 17:50 Mutually assured AI malfunction 22:54 Policy suggestions for current administration 25:34 Compute security 30:37 Current state of evals

ABOUT JIM PALMERJim Palmer is the Chief AI Officer at Dialpad. Previously he was CTO and Co-Founder of TalkIQ, a conversational intelligence start-up with expertise in real-time speech recognition and natural language processing, acquired by Dialpad in May of 2018. Prior to TalkIQ, he was the founding engineer on the eBay Now local delivery service.SHOW NOTES:Tips and cheat codes for navigating AI governance (3:30)Breaking down red teaming & adversarial testing in AI governance (8:02)Launching and scaling adversarial testing efforts (11:27)Unexpected benefits unlocked with adversarial testing (13:43)Understanding data governance and strategic AI investments (15:38)Building resilient AI from concept to customer validation (19:28)Exploring early feature validation and pattern recognition in AI (22:38)Adaptability in data management and ensuring safe, ethical data use while adapting to evolving legal and governance requirements (26:51)How to prepare data for safe and sustainable long-term use (30:02)Strategies for compliant data practices in a regulated world (32:43)Building data deletion systems with model training in mind (35:14)Current events and trends shaping adaptability and durability in the AI ecosystem (38:38)The role of a Chief AI Officer (41:20)Rapid fire questions (44:35)LINKS AND RESOURCESGenius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World - With deep and exclusive reporting, across hundreds of interviews, New York Times Silicon Valley journalist Cade Metz brings you into the rooms where these questions are being answered. Where an extraordinarily powerful new artificial intelligence has been built into our biggest companies, our social discourse, and our daily lives, with few of us even noticing.This episode wouldn't have been possible without the help of our incredible production team:Patrick Gallagher - Producer & Co-HostJerry Li - Co-HostNoah Olberding - Associate Producer, Audio & Video Editor https://www.linkedin.com/in/noah-olberding/Dan Overheim - Audio Engineer, Dan's also an avid 3D printer - https://www.bnd3d.com/Ellie Coggins Angus - Copywriter, Check out her other work at https://elliecoggins.com/about/

Patrick's journey from economics and politics to AI investment showcases the intersection of strategic thinking and social impact. A pioneer in Canada's AI revolution, he shares insights on how AI adoption can drive societal and economic transformation.00:42- About Patrick TammerPatrick is an AI Investor managing a 125M+ portfolio of AI investment projects.He is an AI Consultant to senior business leaders and advisor to the Canadian Government.

In the latest episode of The Marketing Corner, host Shawn Campbell sits down with Mo Rousso, an executive business coach and a powerhouse of entrepreneurial innovation. Together, they dive deep into the ever-evolving world of artificial intelligence (AI) and explore its transformative impact on marketing and business efficiency.AI has rapidly shifted from being a futuristic concept to an essential tool for businesses of all sizes. In this episode, Shawn and Mo break down how AI can optimize marketing strategies, streamline workflows, and help entrepreneurs work smarter, not harder.Some key takeaways from the discussion include:✅ AI-Powered Efficiency – From automating repetitive tasks to enhancing content creation, AI allows business owners to focus on strategy and big-picture thinking rather than getting caught up in time-consuming minutiae.✅ Personalization at Scale – AI-driven marketing tools can analyze customer behavior and preferences in real-time, enabling businesses to create hyper-targeted campaigns that drive engagement and conversions.✅ The Role of AI in Content Creation – Whether it's social media posts, blog articles, or email campaigns, AI-powered writing tools can generate high-quality content that aligns with your brand's voice and message.One of the most compelling discussions in this episode revolves around leveraging AI to write a book that positions an entrepreneur as an authority in their industry. Mo and Shawn break down the process, sharing insights on how AI tools can:

INTRODUCING AI AGENT BY JUICEBOX   AI Agents promise to revolutionise the work of recruiting but early experiments with generic products like OpenAI Operator have shown there is still some way to go - humans still needed to oversee, correct and train, especially for the unique use case of finding other human beings as candidates for the jobs we're recruiting for.   It's time for the recruitment technology vendors to step up and fill this need!   David Paffenholz, CEO of Juicebox, the company behind the groundbreaking talent sourcing tool 'PeopleGPT', think they have got the first breakthrough AI Agent focused on the recruiting use case. Let's put him and the AI Agent to the test in this LIVE DEMO of the how the product (or is it now, 'colleague'?) works.   We will learn:   - What instructions does AI Agent need to carry out its tasks? - How does AI Agent identify candidates from natural language prompts? - What sources does AI Agent use to find candidates? - How does it ensure data currency? - Does AI Agent score or rank candidates? - What is the mechanism if so? - Can AI Agent conduct outreach? - Can AI Agent conduct prescreening? - What can AI Agent NOT currently do? - How does a company optimally use AI Agent? - What is the story with compliance, especially with anticipated transatlantic divergence on AI safety? - What else do we need to know before we deploy AI Agent??       We are on Wednesday 26th February, 7am PT / 3pm GMT     Register by clicking the green button and follow the channel here (recommended)     Ep294-a is sponsored by our friends PeopleGPT by Juicebox   PeopleGPT is the leading outbound recruiting platform built on Generative AI. Find talent across 30+ data sources, get interactive Talent Insights, and reach out with personalized AI email campaigns.   Key features include:   PeopleGPT (Search): the world's first people search engine that uses natural language. Describe who you're searching for and find the perfect match from over 30 data sources. Talent Insights: Visualize your talent pool with 15+ interactive charts. Refine your search and dive deep into stats based on location, employers, job titles, skills, and more. Email Outreach: Maximize candidate engagement with AI-powered email campaigns. Personalize messaging at scale using AI commands, and boost response rates by 40%.   Get 15% off your Juicebox subscription with code: BRAINFOOD15   Trusted by 400+ companies including Scale AI, Mindbloom, and Patreon. Free trial available - try it now here

Alexandr Wang is the founder and CEO of Scale AI, a platform that provides data training for AI programs. In 2021 he was named the youngest self-made billionaire in the world by Forbes at the age of 24. Theo is joined by Alexandr Wang to talk all about AI and how it's changing our world fast. They discuss Alex's choice to drop out of MIT to pursue this field full time, the debate over whether it's creating or eliminating jobs, and how soon regular people will start to see these programs in their everyday lives.  Alexandr Wang: https://www.instagram.com/alexanddeer/ ------------------------------------------------ Tour Dates! https://theovon.com/tour New Merch: https://www.theovonstore.com ------------------------------------------------- Sponsored By: Celsius: Go to the Celsius Amazon store to check out all of their flavors. #CELSIUSBrandPartner #CELSIUSLiveFit  https://amzn.to/3HbAtPJ  BetterHelp: This episode is sponsored by BetterHelp — go to http://betterhelp.com/theo to get 10%off your first month. Modiphy: Get 50% off the last website you'll ever need at https://modiphy.com/THEO  ------------------------------------------------- Music: “Shine” by Bishop Gunn Bishop Gunn - Shine ------------------------------------------------ Submit your funny videos, TikToks, questions and topics you'd like to hear on the podcast to: tpwproducer@gmail.com Hit the Hotline: 985-664-9503 Video Hotline for Theo Upload here: https://www.theovon.com/fan-upload Send mail to: This Past Weekend 1906 Glen Echo Rd PO Box #159359 Nashville, TN 37215 ------------------------------------------------ Find Theo: Website: https://theovon.com Instagram: https://instagram.com/theovon Facebook: https://facebook.com/theovon Facebook Group: https://www.facebook.com/groups/thispastweekend Twitter: https://twitter.com/theovon YouTube: https://youtube.com/theovon Clips Channel: https://www.youtube.com/c/TheoVonClips Shorts Channel: https://bit.ly/3ClUj8z ------------------------------------------------ Producer: Zach https://www.instagram.com/zachdpowers Producer: Nick https://www.instagram.com/realnickdavis/ Producer: Cam https://www.instagram.com/cam__george/  Producer: Colin https://instagram.com/colin_reiner Learn more about your ad choices. Visit megaphone.fm/adchoices

Aaron Harris was a partner at YC for 7.5 years where he funded Deel, OpenSea, Scale AI, Rappi, Lattice, and others. He also built YC's Series A program where he worked with founders on over 200 Series As and Bs that raised in excess of $3B in capital. He joined Sam Kirschner, VP at Village Global, to give Village Global founders helpful tips on fundraising.Takeaways:Investors bet on stories, not just data. Your job? Tell a story of massive economic opportunity — not just how you fit into the future, but how you create it.Hype lowers the barrier to turn attention into excitement.Fundraising isn't the trophy, even though it's celebrated as such. But if you optimize for raising money, you get great at burning money and miss the real point: building something that lasts through the hype cycles.Think about raising money with terminal value in mind, not just the value of this current round.Raising an A: you'll need to raise about 20% of your company. Series B: 15-20%. Seeds are all over the place. More competition in a round means less dilution; less competition means more dilution. Hype helps reduce dilution.Be aware of dilution on SAFEs so that you don't all of a sudden start arguing about a point of dilution here or there on a given round.AI companies have been getting faster investor movement. The hottest deals land term sheets in 48 hours. Next tier? 1-2 weeks. Beyond that, a 2-6 week grind with unpredictable timing. After 6 weeks, rounds can stretch to 3-4 months where persistence is what leads to success.Understand the market and how investors think by having casual, no-pitch-deck coffee chats. The real signal? Not a follow-up meeting — that's just a VC's job. Look for action: are they willing to spend political capital for you? That's true interest.Thanks for listening — if you like what you hear, please review us on your favorite podcast platform.Check us out on the web at www.villageglobal.vc or get in touch with us on Twitter @villageglobal.Want to get updates from us? Subscribe to get a peek inside the Village. We'll send you reading recommendations, exclusive event invites, and commentary on the latest happenings in Silicon Valley. www.villageglobal.vc/signup

The Department of Energy on Friday replaced its chief information officer with a network engineer from Elon Musk's SpaceX, FedScoop has learned. Dawn Zimmer — who'd been serving as Energy CIO since Ann Dunkin resigned from the role as the Biden administration left office — has been removed from the role by the department's leadership, two sources with direct knowledge of the move told FedScoop. Zimmer was hired as Energy's principal deputy CIO in November, and she has returned to that role. With Zimmer removed as acting CIO, Energy leadership has appointed Ryan Riedel to the role, according to one of the sources, who also shared a screenshot of Riedel listed as CIO in the department's email directory. Riedel lists his current employment as a lead network security engineer at SpaceX on his LinkedIn. He joined the company in 2020 after previously serving at U.S. Army Cyber Command and in the U.S. Navy as an IT specialist, his profile shows. The change comes amid reports that members of the Musk-led Department of Government Efficiency have entered the Department of Energy and at least one is accessing its IT systems. The U.S. AI Safety Institute has selected Scale AI as the first third-party evaluator authorized to assess AI models on its behalf, opening a new channel for testing. That agreement will allow a broader range of model builders to access voluntary evaluation, according to a Scale AI release shared with FedScoop ahead of the Monday announcement. Participating companies will be able to test their models once and, if they choose, share those results with AI safety institutes around the world. Criteria for those evaluations will be developed jointly by the AI data labeling company and the AISI. For Scale's part, that work will be led by its research arm, the Safety, Evaluation, and Alignment Lab, or SEAL. Per the announcement, the evaluations will look at performance in areas such as math, reasoning, and AI coding. The Daily Scoop Podcast is available every Monday-Friday afternoon. If you want to hear more of the latest from Washington, subscribe to The Daily Scoop Podcast  on Apple Podcasts, Soundcloud, Spotify and YouTube.

In this AWS panel discussion, Naveen Rao, VP of AI of Databricks and Vijay Karunamurthy, Field CTO of Scale AI share practical insights on implementing generative AI in enterprises, leveraging private data effectively, and building reliable production systems.Topics Include:Sherry Marcus introduces panel discussion on generative AI adoptionScale AI helps make AI models more reliableDatabricks focuses on customizing AI with company dataCompanies often stressed about where to start with AIBoard-level pressure driving many enterprise AI initiativesStart by defining specific goals and success metricsBuild evaluations first before implementing AI solutionsAvoid rushing into demos without proper planningEnterprise data vastly exceeds public training data volumeCustomer support histories valuable for AI trainingModels learning to anticipate customer follow-up questionsProduction concerns: cost, latency, and accuracy trade-offsGood telemetry crucial for diagnosing AI application issuesSpeed matters more for prose, accuracy for legal documentsCost becomes important once systems begin scaling upOrganizations struggle with poor quality existing dataPrivacy crucial when leveraging internal business dataRole-based access control essential for regulated industriesAI can help locate relevant data across legacy systemsModels need organizational awareness to find data effectivelyPrivate data behind firewalls most valuable for AICustomization gives competitive advantage over generic modelsCurrent AI models primarily do flexible data recallNext few years: focus on deriving business valueFuture developments in causal inference expected post-5 yearsComplex multi-agent systems becoming more importantScale AI developing "humanity's last exam" evaluation metricDiscussion of responsibility and liability in AI decisionsCompanies must stand behind their AI system outputsExisting compliance frameworks can be adapted for AIParticipants:Naveen Rao – VP of AI, DatabricksVijay Karunamurthy – Field CTO, Scale AISherry Marcus Ph.D. - Director, Applied Science, AWSSee how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon/isv/

We finally chased down the smart and savvy Rob Quast, Principal Technologist at Pure, to discuss the latest trends in the tech world around AI, Networking and Cloud storage. Our conversation kicks off with Rob's journey, from his time at ePlus and working in sys admin roles, to a memorable experience at Pure's NYC Accelerate event where he saw the stock exchange firsthand. We touch on the various aspects of a PT's role, including Rob's experience showing demos and helping customers understand complex solutions. Our conversation shifts to some of Rob's most passionate work, including involvement in a large AI solution for Coreweave. He walks through the intricacies of a massive deal and how it required close collaboration with the product management team, post-validation work, and strategic networking. Rob also shares insights from a hyperscaler win for Pure Storage, explaining what "hyperscaler" means and how they're used in the industry. We dive into the technical details of large AI deals, such as GPUDirect, and discuss how Pure can better connect with networking professionals. A key theme here is the intersection of data storage and networking, with Rob revealing why NVMe TCP is a game-changer for the future of cloud infrastructure, particularly when compared to technologies like Infiniband. We close with conversation looking toward the future, exploring the realities of cloud storage and hybrid cloud solutions, and what customers are trying to achieve and the challenges they face. The episode wraps up with a focus on object storage, where Rob discusses how Pure Storage's Fusion platform helps manage global namespaces and facilitates multi-tenant cloud environments.

Third and final episode in the Paine trilogy!Chinese history is full of warlords constantly challenging the capital. How could Mao not only stay in power for decades, but not even face any insurgency?And how did Mao go from military genius to peacetime disaster - the patriotic hero who inflicted history's worst human catastrophe on China? How can someone shrewd enough to win a civil war outnumbered 5 to 1 decide "let's have peasants make iron in their backyards" and "let's kill all the birds"?In her lecture and our Q&A, we cover the first nationwide famine in Chinese history; Mao's lasting influence on other insurgents; broken promises to minorities and peasantry; and what Taiwan means.Thanks so much to @Substack for running this in-person event!Note that Sarah is doing an AMA over the next couple days on Youtube; see the pinned comment.Watch on YouTube. Listen on Apple Podcasts, Spotify, or any other podcast platform.SponsorToday's episode is brought to you by Scale AI. Scale partners with the U.S. government to fuel America's AI advantage through their data foundry. Scale recently introduced Defense Llama, Scale's latest solution available for military personnel. With Defense Llama, military personnel can harness the power of AI to plan military or intelligence operations and understand adversary vulnerabilities.If you're interested in learning more on how Scale powers frontier AI capabilities, go to https://scale.com/dwarkesh. Get full access to Dwarkesh Podcast at www.dwarkeshpatel.com/subscribe

The Office of Personnel Management has created a new email account meant to collect reports of suspected diversity, equity, and inclusion initiatives, one of a series of moves the Trump administration has taken to slash DEI efforts across the federal workforce. According to a Jan. 21 memo available online, OPM directed agencies — which are now engaged in the process of shutting down diversity initiatives — to collect reports of any efforts to disguise such initiatives. The memo states that the administration is aware of efforts by some in government to disguise DEI programs by using coded or imprecise language, calling for anyone aware of a change in any contract description or personnel position description since November 5, 2024 to obscure the connection between the contract and DEI or similar ideologies to report all “facts and circumstances” to the email account DEIAtruth@opm.gov within 10 days. Failure to report such activities could result in “adverse consequences,” the memo notes. The White House sent Michael Kratsios's nomination to direct the White House Office of Science and Technology Policy to the Senate on Wednesday, formally beginning his confirmation process. Kratsios was chief technology officer during the first Trump administration and was most recently managing director at Scale AI, a technology company and defense contractor focused on AI model training data. Sending his nomination to the Senate officially starts the confirmation process and puts him among the first of Trump's selections officially transmitted to the chamber. The Daily Scoop Podcast is available every Monday-Friday afternoon. If you want to hear more of the latest from Washington, subscribe to The Daily Scoop Podcast  on Apple Podcasts, Soundcloud, Spotify and YouTube.

This is the second episode in the trilogy of a lectures by Professor Sarah Paine of the Naval War College.In this second episode, Prof Paine dissects the ideas and economics behind Japanese imperialism before and during WWII. We get into the oil shortage which caused the war; the unique culture of honor and death; the surprisingly chaotic chain of command. This is followed by a Q&A with me.Huge thanks to Substack for hosting this event!Watch on YouTube. Listen on Apple Podcasts, Spotify, or any other podcast platform.SponsorToday's episode is brought to you by Scale AI. Scale partners with the U.S. government to fuel America's AI advantage through their data foundry. Scale recently introduced Defense Llama, Scale's latest solution available for military personnel. With Defense Llama, military personnel can harness the power of AI to plan military or intelligence operations and understand adversary vulnerabilities.If you're interested in learning more on how Scale powers frontier AI capabilities, go to scale.com/dwarkesh.Buy Sarah's Books!I highly, highly recommend both "The Wars for Asia, 1911–1949" and "The Japanese Empire: Grand Strategy from the Meiji Restoration to the Pacific War".Timestamps(0:00:00) - Lecture begins(0:06:58) - The code of the samurai(0:10:45) - Buddhism, Shinto, Confucianism(0:16:52) - Bushido as bad strategy(0:23:34) - Military theorists(0:33:42) - Strategic sins of omission(0:38:10) - Crippled logistics(0:40:58) - the Kwantung Army(0:43:31) - Inter-service communication(0:51:15) - Shattering Japanese morale(0:57:35) - Q&A begins(01:05:02) - Unusual brutality of WWII(01:11:30) - Embargo caused the war(01:16:48) - The liberation of China(01:22:02) - Could US have prevented war?(01:25:30) - Counterfactuals in history(01:27:46) - Japanese optimism(01:30:46) - Tech change and social change(01:38:22) - Hamming questions(01:44:31) - Do sanctions work?(01:50:07) - Backloaded mass death(01:54:09) - demilitarizing Japan(01:57:30) - Post-war alliances(02:03:46) - Inter-service rivalry Get full access to Dwarkesh Podcast at www.dwarkeshpatel.com/subscribe

Guest InformationName: Jarrod GingrasTitle: Managing Director & AnalystOrganization: Real Story GroupExpertise:Marketing and workplace technology evaluationAligning technology with strategic goalsDigital content and customer data platformsEpisode SummaryIn this episode, Jarrod Gingras shares his expertise on navigating the seismic shifts AI brings to creative operations and tech stacks. He breaks down how AI demands clean data, aligned content, and strategic decisions, and explains why refactoring your tech stack might be essential for staying competitive. This conversation is packed with actionable insights for leaders and teams navigating this critical inflection point in creative operations.Key TakeawaysRefactoring vs. Staying the Course:AI has introduced foundational changes to creative operations, making tech stack alignment essential.Content, Data, and Decisions:These three elements form the backbone of successful AI adoption. Clean data and aligned content are non-negotiable.Human in the Loop:AI workflows require human oversight to ensure meaningful and relevant outcomes.Breaking Down Silos:Collaboration between content and data teams is critical for creating impactful AI-driven solutions.Future-Proofing Your Tech Stack:Assess your systems for scalability, data fluidity, and integration with AI tools.Personalization at Scale:AI enables always-on workflows but requires a foundational shift in how content and data are managed.Actionable Frameworks:Jarrod outlines practical steps for leaders to assess and update their tech stacks strategically.The Existential Question:To refactor or not to refactor? The answer depends on your current system's readiness for AI integration.

I'm thrilled to launch a new trilogy of double episodes: a lecture series by Professor Sarah Paine of the Naval War College, each followed by a deep Q&A.In this first episode, Prof Paine talks about key decisions by Khrushchev, Mao, Nehru, Bhutto, & Lyndon Johnson that shaped the whole dynamic of South Asia today. This is followed by a Q&A.Come for the spy bases, shoestring nukes, and insight about how great power politics impacts every region.Huge thanks to Substack for hosting this!Watch on YouTube. Listen on Apple Podcasts, Spotify, or any other podcast platform.SponsorsToday's episode is brought to you by Scale AI. Scale partners with the U.S. government to fuel America's AI advantage through their data foundry. The Air Force, Army, Defense Innovation Unit, and Chief Digital and Artificial Intelligence Office all trust Scale to equip their teams with AI-ready data and the technology to build powerful applications.Scale recently introduced Defense Llama, Scale's latest solution available for military personnel. With Defense Llama, military personnel can harness the power of AI to plan military or intelligence operations and understand adversary vulnerabilities.If you're interested in learning more on how Scale powers frontier AI capabilities, go to scale.com/dwarkesh.Timestamps(00:00) - Intro(02:11) - Mao at war, 1949-51(05:40) - Pactomania and Sino-Soviet conflicts(14:42) - The Sino-Indian War(20:00) - Soviet peace in India-Pakistan(22:00) - US Aid and Alliances(26:14) - The difference with WWII(30:09) - The geopolitical map in 1904(35:10) - The US alienates Indira Gandhi(42:58) - Instruments of US power(53:41) - Carrier battle groups(1:02:41) - Q&A begins(1:04:31) - The appeal of the USSR(1:09:36) - The last communist premier(1:15:42) - India and China's lost opportunity(1:58:04) - Bismark's cunning(2:03:05) - Training US officers(2:07:03) - Cruelty in Russian history Get full access to Dwarkesh Podcast at www.dwarkeshpatel.com/subscribe

2024 has been a year of transformative technological progress, marked by conversations that have reshaped our understanding of AI's evolution and what lies ahead. Throughout the year, Sarah and Elad have had the privilege of speaking with some of the brightest minds in the field. As we look back on the past months, we're excited to share highlights from some of our favorite No Priors podcast episodes. Featured guests include Jensen Huang (Nvidia), Andrej Karpathy (OpenAI, Tesla), Bret Taylor (Sierra), Aditya Ramesh, Tim Brooks, and Bill Peebles (OpenAI's Sora Team), Dmitri Dolgov (Waymo), Dylan Field (Figma), and Alexandr Wang (Scale). Want to dive deeper? Listen to the full episodes here: NVIDIA's Jensen Huang on AI Chip Design, Scaling Data Centers, and his 10-Year Bet No Priors Ep. 89 | With NVIDIA CEO Jensen Huang  The Road to Autonomous Intelligence, With Andrej Karpathy from OpenAI and Tesla No Priors Ep. 80 | With Andrej Karpathy from OpenAI and Tesla Transforming Customer Service through Company Agents, with Sierra's Bret Taylor No Priors Ep. 82 | With CEO of Sierra Bret Taylor OpenAI's Sora team thinks we've only seen the "GPT-1 of video models" No Priors Ep.61 | OpenAI's Sora Leaders Aditya Ramesh, Tim Brooks and Bill Peebles Waymo's Journey to Full Autonomy: AI Breakthroughs, Safety, and Scaling No Priors Ep. 87 | With Co-CEO of Waymo Dmitri Dolgov Designing the Future: Dylan Field on AI, Collaboration, and Independence No Priors Ep. 55 | With Figma CEO Dylan Field The Data Foundry for AI with Alexandr Wang from Scale No Priors Ep. 65 | With Scale AI CEO Alexandr Wang Sign up for new podcasts every week. Email feedback to show@no-priors.com Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil Timecodes: 0:00 Introduction  0:15 Jensen Huang on building at data-center scale  4:00 Andrej Karpathy on the AI exo-cortex, model control, and a shift to smaller models  7:14 Bret Taylor on the agentic future of business interactions  11:17 OpenAI's Sora team on visual models and their role in AGI  15:53 Waymo's Dmitri Dolgov on bridging the gap to full autonomy and the challenge of 100% accuracy  19:00 Figma's Dylan Field on the future of interfaces and new modalities  23:29 Scale AI's Alexandr Wang on the journey to AGI  26:29 Outro

Happy holidays! We'll be sharing snippets from Latent Space LIVE! through the break bringing you the best of 2024! We want to express our deepest appreciation to event sponsors AWS, Daylight Computer, Thoth.ai, StrongCompute, Notable Capital, and most of all all our LS supporters who helped fund the gorgeous venue and A/V production!For NeurIPS last year we did our standard conference podcast coverage interviewing selected papers (that we have now also done for ICLR and ICML), however we felt that we could be doing more to help AI Engineers 1) get more industry-relevant content, and 2) recap 2024 year in review from experts. As a result, we organized the first Latent Space LIVE!, our first in person miniconference, at NeurIPS 2024 in Vancouver. Today, we're proud to share Loubna's highly anticipated talk (slides here)!Synthetic DataWe called out the Synthetic Data debate at last year's NeurIPS, and no surprise that 2024 was dominated by the rise of synthetic data everywhere:* Apple's Rephrasing the Web, Microsoft's Phi 2-4 and Orca/AgentInstruct, Tencent's Billion Persona dataset, DCLM, and HuggingFace's FineWeb-Edu, and Loubna's own Cosmopedia extended the ideas of synthetic textbook and agent generation to improve raw web scrape dataset quality* This year we also talked to the IDEFICS/OBELICS team at HuggingFace who released WebSight this year, the first work on code-vs-images synthetic data.* We called Llama 3.1 the Synthetic Data Model for its extensive use (and documentation!) of synthetic data in its pipeline, as well as its permissive license. * Nemotron CC and Nemotron-4-340B also made a big splash this year for how they used 20k items of human data to synthesize over 98% of the data used for SFT/PFT.* Cohere introduced Multilingual Arbitrage: Optimizing Data Pools to Accelerate Multilingual Progress observing gains of up to 56.5% improvement in win rates comparing multiple teachers vs the single best teacher model* In post training, AI2's Tülu3 (discussed by Luca in our Open Models talk) and Loubna's Smol Talk were also notable open releases this year.This comes in the face of a lot of scrutiny and criticism, with Scale AI as one of the leading voices publishing AI models collapse when trained on recursively generated data in Nature magazine bringing mainstream concerns to the potential downsides of poor quality syndata:Part of the concerns we highlighted last year on low-background tokens are coming to bear: ChatGPT contaminated data is spiking in every possible metric:But perhaps, if Sakana's AI Scientist pans out this year, we will have mostly-AI AI researchers publishing AI research anyway so do we really care as long as the ideas can be verified to be correct?Smol ModelsMeta surprised many folks this year by not just aggressively updating Llama 3 and adding multimodality, but also adding a new series of “small” 1B and 3B “on device” models this year, even working on quantized numerics collaborations with Qualcomm, Mediatek, and Arm. It is near unbelievable that a 1B model today can qualitatively match a 13B model of last year:and the minimum size to hit a given MMLU bar has come down roughly 10x in the last year. We have been tracking this proxied by Lmsys Elo and inference price:The key reads this year are:* MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases* Apple Intelligence Foundation Language Models* Hymba: A Hybrid-head Architecture for Small Language Models* Loubna's SmolLM and SmolLM2: a family of state-of-the-art small models with 135M, 360M, and 1.7B parameters on the pareto efficiency frontier.* and Moondream, which we already covered in the 2024 in Vision talkFull Talk on YouTubeplease like and subscribe!Timestamps* [00:00:05] Loubna Intro* [00:00:33] The Rise of Synthetic Data Everywhere* [00:02:57] Model Collapse* [00:05:14] Phi, FineWeb, Cosmopedia - Synthetic Textbooks* [00:12:36] DCLM, Nemotron-CC* [00:13:28] Post Training - AI2 Tulu, Smol Talk, Cohere Multilingual Arbitrage* [00:16:17] Smol Models* [00:18:24] On Device Models* [00:22:45] Smol Vision Models* [00:25:14] What's NextTranscript2024 in Synthetic Data and Smol Models[00:00:00] ​[00:00:05] Loubna Intro[00:00:05] Speaker: ​I'm very happy to be here. Thank you for the invitation. So I'm going to be talking about synthetic data in 2024. And then I'm going to be talking about small on device models. So I think the most interesting thing about synthetic data this year is that like now we have it everywhere in the large language models pipeline.[00:00:33] The Rise of Synthetic Data Everywhere[00:00:33] Speaker: I think initially, synthetic data was mainly used just for post training, because naturally that's the part where we needed human annotators. And then after that, we realized that we don't really have good benchmarks to [00:01:00] measure if models follow instructions well, if they are creative enough, or if they are chatty enough, so we also started using LLMs as judges.[00:01:08] Speaker: Thank you. And I think this year and towards the end of last year, we also went to the pre training parts and we started generating synthetic data for pre training to kind of replace some parts of the web. And the motivation behind that is that you have a lot of control over synthetic data. You can control your prompt and basically also the kind of data that you generate.[00:01:28] Speaker: So instead of just trying to filter the web, you could try to get the LLM to generate what you think the best web pages could look like and then train your models on that. So this is how we went from not having synthetic data at all in the LLM pipeline to having it everywhere. And so the cool thing is like today you can train an LLM with like an entirely synthetic pipeline.[00:01:49] Speaker: For example, you can use our Cosmopedia datasets and you can train a 1B model on like 150 billion tokens that are 100 percent synthetic. And those are also of good quality. And then you can [00:02:00] instruction tune the model on a synthetic SFT dataset. You can also do DPO on a synthetic dataset. And then to evaluate if the model is good, you can use.[00:02:07] Speaker: A benchmark that uses LLMs as a judge, for example, MTBench or AlpacaEvil. So I think this is like a really mind blowing because like just a few years ago, we wouldn't think this is possible. And I think there's a lot of concerns about model collapse, and I'm going to talk about that later. But we'll see that like, if we use synthetic data properly and we curate it carefully, that shouldn't happen.[00:02:29] Speaker: And the reason synthetic data is very popular right now is that we have really strong models, both open and closed. It is really cheap and fast to use compared to human annotations, which cost a lot and take a lot of time. And also for open models right now, we have some really good inference frameworks.[00:02:47] Speaker: So if you have enough GPUs, it's really easy to spawn these GPUs and generate like a lot of synthetic data. Some examples are VLM, TGI, and TensorRT.[00:02:57] Model Collapse[00:02:57] Speaker: Now let's talk about the elephant in the room, model [00:03:00] collapse. Is this the end? If you look at the media and all of like, for example, some papers in nature, it's really scary because there's a lot of synthetic data out there in the web.[00:03:09] Speaker: And naturally we train on the web. So we're going to be training a lot of synthetic data. And if model collapse is going to happen, we should really try to take that seriously. And the other issue is that, as I said, we think, a lot of people think the web is polluted because there's a lot of synthetic data.[00:03:24] Speaker: And for example, when we're building fine web datasets here at Guillerm and Hinek, we're interested in like, how much synthetic data is there in the web? So there isn't really a method to properly measure the amount of synthetic data or to save a webpage synthetic or not. But one thing we can do is to try to look for like proxy words, for example, expressions like as a large language model or words like delve that we know are actually generated by chat GPT.[00:03:49] Speaker: We could try to measure the amount of these words in our data system and compare them to the previous years. For example, here, we measured like a, these words ratio in different dumps of common crawl. [00:04:00] And we can see that like the ratio really increased after chat GPT's release. So if we were to say that synthetic data amount didn't change, you would expect this ratio to stay constant, which is not the case.[00:04:11] Speaker: So there's a lot of synthetic data probably on the web, but does this really make models worse? So what we did is we trained different models on these different dumps. And we then computed their performance on popular, like, NLP benchmarks, and then we computed the aggregated score. And surprisingly, you can see that the latest DOMs are actually even better than the DOMs that are before.[00:04:31] Speaker: So if there's some synthetic data there, at least it did not make the model's worse. Yeah, which is really encouraging. So personally, I wouldn't say the web is positive with Synthetic Data. Maybe it's even making it more rich. And the issue with like model collapse is that, for example, those studies, they were done at like a small scale, and you would ask the model to complete, for example, a Wikipedia paragraph, and then you would train it on these new generations, and you would do that every day.[00:04:56] Speaker: iteratively. I think if you do that approach, it's normal to [00:05:00] observe this kind of behavior because the quality is going to be worse because the model is already small. And then if you train it just on its generations, you shouldn't expect it to become better. But what we're really doing here is that we take a model that is very large and we try to distill its knowledge into a model that is smaller.[00:05:14] Phi, FineWeb, Cosmopedia - Synthetic Textbooks[00:05:14] Speaker: And in this way, you can expect to get like a better performance for your small model. And using synthetic data for pre-training has become really popular. After the textbooks are all you need papers where Microsoft basically trained a series of small models on textbooks that were using a large LLM.[00:05:32] Speaker: And then they found that these models were actually better than models that are much larger. So this was really interesting. It was like first of its time, but it was also met with a lot of skepticism, which is a good thing in research. It pushes you to question things because the dataset that they trained on was not public, so people were not really sure if these models are really good or maybe there's just some data contamination.[00:05:55] Speaker: So it was really hard to check if you just have the weights of the models. [00:06:00] And as Hugging Face, because we like open source, we tried to reproduce what they did. So this is our Cosmopedia dataset. We basically tried to follow a similar approach to what they documented in the paper. And we created a synthetic dataset of textbooks and blog posts and stories that had almost 30 billion tokens.[00:06:16] Speaker: And we tried to train some models on that. And we found that like the key ingredient to getting a good data set that is synthetic is trying as much as possible to keep it diverse. Because if you just throw the same prompts as your model, like generate like a textbook about linear algebra, and even if you change the temperature, the textbooks are going to look alike.[00:06:35] Speaker: So there's no way you could scale to like millions of samples. And the way you do that is by creating prompts that have some seeds that make them diverse. In our case, the prompt, we would ask the model to generate a textbook, but make it related to an extract from a webpage. And also we try to frame it within, to stay within topic.[00:06:55] Speaker: For example, here, we put like an extract about cardiovascular bioimaging, [00:07:00] and then we ask the model to generate a textbook related to medicine that is also related to this webpage. And this is a really nice approach because there's so many webpages out there. So you can. Be sure that your generation is not going to be diverse when you change the seed example.[00:07:16] Speaker: One thing that's challenging with this is that you want the seed samples to be related to your topics. So we use like a search tool to try to go all of fine web datasets. And then we also do a lot of experiments with the type of generations we want the model to generate. For example, we ask it for textbooks for middle school students or textbook for college.[00:07:40] Speaker: And we found that like some generation styles help on some specific benchmarks, while others help on other benchmarks. For example, college textbooks are really good for MMLU, while middle school textbooks are good for benchmarks like OpenBookQA and Pico. This is like a sample from like our search tool.[00:07:56] Speaker: For example, you have a top category, which is a topic, and then you have some [00:08:00] subtopics, and then you have the topic hits, which are basically the web pages in fine web does belong to these topics. And here you can see the comparison between Cosmopedia. We had two versions V1 and V2 in blue and red, and you can see the comparison to fine web, and as you can see throughout the training training on Cosmopedia was consistently better.[00:08:20] Speaker: So we managed to get a data set that was actually good to train these models on. It's of course so much smaller than FineWeb, it's only 30 billion tokens, but that's the scale that Microsoft data sets was, so we kind of managed to reproduce a bit what they did. And the data set is public, so everyone can go there, check if everything is all right.[00:08:38] Speaker: And now this is a recent paper from NVIDIA, Neumatron CC. They took things a bit further, and they generated not a few billion tokens, but 1. 9 trillion tokens, which is huge. And we can see later how they did that. It's more of, like, rephrasing the web. So we can see today that there's, like, some really huge synthetic datasets out there, and they're public, so, [00:09:00] like, you can try to filter them even further if you want to get, like, more high quality corpses.[00:09:04] Speaker: So for this, rephrasing the web this approach was suggested in this paper by Pratyush, where basically in this paper, they take some samples from C4 datasets, and then they use an LLM to rewrite these samples into a better format. For example, they ask an LLM to rewrite the sample into a Wikipedia passage or into a Q& A page.[00:09:25] Speaker: And the interesting thing in this approach is that you can use a model that is Small because it doesn't, rewriting doesn't require knowledge. It's just rewriting a page into a different style. So the model doesn't need to have like knowledge that is like extensive of what is rewriting compared to just asking a model to generate a new textbook and not giving it like ground truth.[00:09:45] Speaker: So here they rewrite some samples from C4 into Q& A, into Wikipedia, and they find that doing this works better than training just on C4. And so what they did in Nemo Trans CC is a similar approach. [00:10:00] They rewrite some pages from Common Crawl for two reasons. One is to, like improve Pages that are low quality, so they rewrite them into, for example, Wikipedia page, so they look better.[00:10:11] Speaker: And another reason is to create more diverse datasets. So they have a dataset that they already heavily filtered, and then they take these pages that are already high quality, and they ask the model to rewrite them in Question and Answer format. into like open ended questions or like multi choice questions.[00:10:27] Speaker: So this way they can reuse the same page multiple times without fearing like having multiple duplicates, because it's the same information, but it's going to be written differently. So I think that's also a really interesting approach for like generating synthetic data just by rephrasing the pages that you already have.[00:10:44] Speaker: There's also this approach called Prox where they try to start from a web page and then they generate a program which finds how to write that page to make it better and less noisy. For example, here you can see that there's some leftover metadata in the web page and you don't necessarily want to keep that for training [00:11:00] your model.[00:11:00] Speaker: So So they train a model that can generate programs that can like normalize and remove lines that are extra. So I think this approach is also interesting, but it's maybe less scalable than the approaches that I presented before. So that was it for like rephrasing and generating new textbooks.[00:11:17] Speaker: Another approach that I think is really good and becoming really popular for using synthetic data for pre training is basically building a better classifiers. For filtering the web for example, here we release the data sets called fine web edu. And the way we built it is by taking Llama3 and asking it to rate the educational content of web pages from zero to five.[00:11:39] Speaker: So for example, if a page is like a really good textbook that could be useful in a school setting, it would get a really high score. And if a page is just like an advertisement or promotional material, it would get a lower score. And then after that, we take these synthetic annotations and we train a classifier on them.[00:11:57] Speaker: It's a classifier like a BERT model. [00:12:00] And then we run this classifier on all of FineWeb, which is a 15 trillion tokens dataset. And then we only keep the pages that have like a score that's higher than 3. So for example, in our case, we went from 15 trillion tokens to 3. to just 1. 5 trillion tokens. Those are really highly educational.[00:12:16] Speaker: And as you can see here, a fine web EDU outperforms all the other public web datasets by a larger margin on a couple of benchmarks here, I show the aggregated score and you can see that this approach is really effective for filtering web datasets to get like better corpuses for training your LLMs.[00:12:36] DCLM, Nemotron-CC[00:12:36] Speaker: Others also try to do this approach. There's, for example, the DCLM datasets where they also train the classifier, but not to detect educational content. Instead, they trained it on OpenHermes dataset, which is a dataset for instruction tuning. And also they explain like IAM5 subreddits, and then they also get really high quality dataset which is like very information dense and can help [00:13:00] you train some really good LLMs.[00:13:01] Speaker: And then Nemotron Common Crawl, they also did this approach, but instead of using one classifier, they used an ensemble of classifiers. So they used, for example, the DCLM classifier, and also classifiers like the ones we used in FineWebEducational, and then they combined these two. Scores into a, with an ensemble method to only retain the best high quality pages, and they get a data set that works even better than the ones we develop.[00:13:25] Speaker: So that was it for like synthetic data for pre-training.[00:13:28] Post Training - AI2 Tulu, Smol Talk, Cohere Multilingual Arbitrage[00:13:28] Speaker: Now we can go back to post training. I think there's a lot of interesting post training data sets out there. One that was released recently, the agent instructs by Microsoft where they basically try to target some specific skills. And improve the performance of models on them.[00:13:43] Speaker: For example, here, you can see code, brain teasers, open domain QA, and they managed to get a dataset that outperforms that's when fine tuning Mistral 7b on it, it outperforms the original instruct model that was released by Mistral. And as I said, to get good synthetic data, you really [00:14:00] have to have a framework to make sure that your data is diverse.[00:14:03] Speaker: So for example, for them, they always. And then they see the generations on either source code or raw text documents, and then they rewrite them to make sure they're easier to generate instructions from, and then they use that for their like instruction data generation. There's also the Tool3SFT mixture, which was released recently by Allen AI.[00:14:23] Speaker: It's also really good quality and it covers a wide range of tasks. And the way they make sure that this dataset is diverse is by using personas from the persona hub datasets. Which is basically a data set of like I think over a million personas. And for example, in the tool mixture to generate like a new code snippet, they would give like the model persona, for example, a machine learning researcher interested in neural networks, and then ask it to generate like a coding problem.[00:14:49] Speaker: This way you make sure that your data set is really diverse, and then you can further filter the data sets, for example, using the reward models. We also released a dataset called Smalltalk, [00:15:00] and we also tried to cover the wide range of tasks, and as you can see here, for example, when fine tuning Mistral 7b on the dataset, we also outperformed the original Mistral instructs on a number of benchmarks, notably on mathematics and instruction following with ifevil.[00:15:18] Speaker: Another paper that's really interesting I wanted to mention is this one called Multilingual Data Arbitrage by Cohere. And basically they want to generate a data set for post training that is multilingual. And they have a really interesting problem. It's the fact that there isn't like one model that's really good at all the languages they wanted.[00:15:36] Speaker: So what they do is that like they use not just one teacher model, but multiple teachers. And then they have a router which basically sends the prompts they have to all these models. And then they get the completions and they have a reward model that traces all these generations and only keeps the best one.[00:15:52] Speaker: And this is like arbitrage and finance. So well, I think what's interesting in this, it shows that like synthetic data, it doesn't have to come from a single model. [00:16:00] And because we have so many good models now, you could like pull these models together and get like a dataset that's really high quality and that's diverse and that's covers all your needs.[00:16:12] Speaker: I was supposed to put a meme there, but. Yeah, so that was it for like a synthetic data.[00:16:17] Smol Models[00:16:17] Speaker: Now we can go to see what's happening in the small models field in 2024. I don't know if you know, but like now we have some really good small models. For example, Lama 3. 2 1B is. It matches Lama 2. 13b from, that was released last year on the LMSYS arena, which is basically the default go to leaderboard for evaluating models using human evaluation.[00:16:39] Speaker: And as you can see here, the scores of the models are really close. So I think we've made like hugely forward in terms of small models. Of course, that's one, just one data point, but there's more. For example, if you look at this chart from the Quint 2. 5 blog post, it shows that today we have some really good models that are only like 3 billion parameters [00:17:00] and 4 billion that score really high on MMLU.[00:17:03] Speaker: Which is a really popular benchmark for evaluating models. And you can see here that the red, the blue dots have more than 65 on MMLU. And the grey ones have less. And for example, Llama33b had less. So now we have a 3b model that outperforms a 33b model that was released earlier. So I think now people are starting to realize that like, we shouldn't just scale and scale models, but we should try to make them more efficient.[00:17:33] Speaker: I don't know if you knew, but you can also chat with a 3B plus model on your iPhone. For example, here, this is an app called PocketPal, where you can go and select a model from Hugging Face. It has a large choice. For example, here we loaded the 5. 3. 5, which is 3. 8 billion parameters on this iPhone. And we can chat with this and you can see that even the latency is also acceptable.[00:17:57] Speaker: For example, here, I asked it to give me a joke about [00:18:00] NeurIPS. So let's see what it has to say.[00:18:06] Speaker: Okay, why did the neural network attend NeurIPS? Because it heard there would be a lot of layers and fun and it wanted to train its sense of humor. So not very funny, but at least it can run on device. Yeah, so I think now we have good small models, but we also have like good frameworks and tools to use these small models.[00:18:24] On Device Models[00:18:24] Speaker: So I think we're really close to having like really on edge and on device models that are really good. And I think for a while we've had this narrative. But just training larger models is better. Of course, this is supported by science scaling laws. As you can see here, for example, when we scale the model size, the loss is lower and obviously you get a better model.[00:18:46] Speaker: But and we can see this, for example, in the GPT family of models, how we went from just a hundred million parameters to more than a trillion. parameters. And of course, we all observed the performance improvement when using the latest model. But [00:19:00] one thing that we shouldn't forget is that when we scale the model, we also scale the inference costs and time.[00:19:05] Speaker: And so the largest models were are going to cost so much more. So I think now instead of just building larger models, we should be focusing on building more efficient models. It's no longer a race for the largest models since these models are really expensive to run and they require like a really good infrastructure to do that and they cannot run on, for example, consumer hardware.[00:19:27] Speaker: And when you try to build more efficient models that match larger models, that's when you can really unlock some really interesting on device use cases. And I think a trend that we're noticing now is the trend of training smaller models longer. For example, if you compare how much, how long LLAMA was trained compared to LLAMA3, there is a huge increase in the pre training length.[00:19:50] Speaker: LLAMA was trained on 1 trillion tokens, but LLAMA3 8b was trained on 15 trillion tokens. So Meta managed to get a model that's the same size, but But it performs so much [00:20:00] better by choosing to like spend the sacrifice during training, because as we know, training is a one time cost, but inference is something that's ongoing.[00:20:08] Speaker: If we want to see what are like the small models reads in 2024, I think this mobile LLM paper by Meta is interesting. They try to study different models that are like have the less than 1 billion parameters and find which architecture makes most sense for these models. For example, they find that depth is more important than width.[00:20:29] Speaker: So it's more important to have models that have like more layers than just one. making them more wide. They also find that GQA helps, that tying the embedding helps. So I think it's a nice study overall for models that are just a few hundred million parameters. There's also the Apple intelligence tech report, which is interesting.[00:20:48] Speaker: So for Apple intelligence, they had two models, one that was like on server and another model that was on device. It had 3 billion parameters. And I think the interesting part is that they trained this model using [00:21:00] pruning. And then distillation. And for example, they have this table where they show that, like, using pruning and distillation works much better than training from scratch.[00:21:08] Speaker: And they also have some interesting insights about, like, how they specialize their models on specific tasks, like, for example, summarization and rewriting. There's also this paper by NVIDIA that was released recently. I think you've already had a talk about, like, hybrid models that was all interesting.[00:21:23] Speaker: And this model, they used, like, a hybrid architecture between state space models and transformers. And they managed to train a 1B model that's really performant without needing to train it on a lot of tokens. And regarding our work, we just recently released SmallM2, so it's a series of three models, which are the best in class in each model size.[00:21:46] Speaker: For example, our 1. 7b model outperforms Lama 1b and also Qt 2. 5. And how we managed to train this model is the following. That's where you spent a lot of time trying to curate the pre training datasets. We did a lot of [00:22:00] ablations, trying to find which datasets are good and also how to mix them. We also created some new math and code datasets that we're releasing soon.[00:22:08] Speaker: But you basically really spent a lot of time trying to find what's the best mixture that you can train these models on. And then we spent some time trying to like we also trained these models for very long. For example, small M1 was trained only on 1 trillion tokens, but this model is trained on 11 trillion tokens.[00:22:24] Speaker: And we saw that the performance kept improving. The models didn't really plateau mid training, which I think is really interesting. It shows that you can train such small models for very long and keep getting performance gains. What's interesting about SmallLM2 is that it's fully open. We also released, like the pre training code base, the fine tuning code, the datasets, and also evaluation in this repository.[00:22:45] Smol Vision Models[00:22:45] Speaker: Also there's, like, really interesting small models for text, but also for vision. For example, here you can see SmallVLM, which is a 2B model that's really efficient. It doesn't consume a lot of RAM, and it also has a good performance. There's also Moondream 0. [00:23:00] 5b, which was released recently. It's like the smallest visual language model.[00:23:04] Speaker: And as you can see, there isn't like a big trade off compared to Moondream 2b. So now I showed you that we have some really good small models. We also have the tools to use them, but why should you consider using small models and when? I think, like, small models are really interesting because of the on device feature.[00:23:23] Speaker: Because these models are small and they can run fast, you can basically run them on your laptop, but also on your mobile phone. And this means that your dataset stays locally. You don't have to send your queries to third parties. And this really enhances privacy. That was, for example, one of the big selling points for Apple Intelligence.[00:23:42] Speaker: Also, right now, we really have a lot of work to do. So many frameworks to do on device inference. For example, there's MLX, MLC, Llama, CPP, Transformers, JS. So we have a lot of options and each of them have like great features. So you have so many options for doing that. Small models are also really powerful if you choose to specialize them.[00:24:00][00:24:00] Speaker: For example, here there's a startup called Numind, which took small LM and then they fine tuned it on text extraction datasets. And they managed to get a model that's not very far from models that are much larger. So I think text extraction is like one use case where small models can be really performant and it makes sense to use them instead of just using larger models.[00:24:19] Speaker: You can also chat with these models in browser. For example, here, you can go there, you can load the model, you can even turn off your internet and just start chatting with the model locally. Speaking of text extraction, if you don't want to fine tune the models, there's a really good method of structure generation.[00:24:36] Speaker: We can basically force the models to follow a JSON schema that you defined. For example, here, we try to force the model to follow a schema for extracting key information from GitHub issues. So you can input free text, which is a complaint about a GitHub repository, something not working. And then you can run it there and the model can extract anything that is relevant for your GitHub issue creation.[00:24:58] Speaker: For example, the [00:25:00] priority, for example, here, priority is high, the type of the issue bug, and then a title and the estimation of how long this will take to fix. And you can just like do this in the browser, you can transform your text into a GitHub issue that's properly formatted.[00:25:14] What's Next[00:25:14] Speaker: So what's next for synthetic data and small models?[00:25:18] Speaker: I think that domain specific synthetic data is going to be, it's already important, it's going to be even more important. For example, generating synthetic data for math. I think this really would help improve the reasoning of a lot of models. And a lot of people are doing it, for example, Quint 2. 12 math, everyone's trying to reproduce a one.[00:25:37] Speaker: And so I think for synthetic data, trying to specialize it on some domains is going to be really important. And then for small models, I think specializing them through fine tuning, it's also going to be really important because I think a lot of companies are just trying to use these large models because they are better.[00:25:53] Speaker: But on some tasks, I think you can already get decent performance with small models. So you don't need to Pay like a [00:26:00] cost that's much larger just to make your model better at your task by a few percent. And this is not just for text. And I think it also applies for other modalities like vision and audio.[00:26:11] Speaker: And I think you should also watch out for on device frameworks and applications. For example, like the app I showed, or lama, all these frameworks are becoming really popular and I'm pretty sure that we're gonna get like more of them in 2025. And users really like that. Maybe for other, I should also say hot take.[00:26:28] Speaker: I think that like in AI, we just started like with fine tuning, for example, trying to make BERT work on some specific use cases, and really struggling to do that. And then we had some models that are much larger. So we just switched to like prompt engineering to get the models And I think we're going back to fine tuning where we realize these models are really costly.[00:26:47] Speaker: It's better to use just a small model or try to specialize it. So I think it's a little bit of a cycle and we're going to start to see like more fine tuning and less of just like a prompt engineering the models. So that was my talk. Thank you for following. And if you have [00:27:00] any questions, we can take them now. Get full access to Latent Space at www.latent.space/subscribe

Alexandr Wang is the CEO and co-founder of Scale AI. He joins Big Technology Podcast to share his predictions for AI in 2025, including insights about emerging geopolitical drama in the AI field, AI agents for consumers, why data may matter more than computing power, and how militaries worldwide are preparing to deploy AI in warfare. We also cover quantum computing and why Wang believes we're approaching the current limits of what massive GPU clusters can achieve. Hit play for a mind-expanding conversation about where artificial intelligence is headed and how it will transform our world in the coming year. --- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. For weekly updates on the show, sign up for the pod newsletter on LinkedIn: https://www.linkedin.com/newsletters/6901970121829801984/ Want a discount for Big Technology on Substack? Here's 40% off for the first year: https://tinyurl.com/bigtechnology Questions? Feedback? Write to: bigtechnologypodcast@gmail.com

In this episode of Uncharted, Poya sits down with Dean Hsu, founder and CEO of Arphie (⁠https://www.arphie.ai⁠), an AI-driven platform that accelerates RFP and questionnaire responses. Dean shares his journey from software consulting and investing to scaling startups like Scale AI, where he learned the skills that prepared him for entrepreneurship. He discusses how personal pain points and deep market research inspired him to become a founder, as well as shares his approach to founder-led sales, building customer trust, and differentiating in a crowded AI market. Dean emphasizes the importance of choosing the right co-founder, setting realistic expectations, and staying outcome-focused. Tune in for insights on navigating startup challenges and lessons from Dean's entrepreneurial path. About our Guest: Dean Shu is the CEO and Co-Founder of Arphie (https://www.arphie.ai), a platform that streamlines the process of responding to tedious requests for proposal (RFPs), security questionnaires, and more. Arphie was born out of his prior experience in responding to these RFPs at Scale AI, where he was the General Manager of a software business unit. Prior to Scale AI, Dean invested in B2B software companies at Insight Partners, a $90B fund based in New York, and served as a consultant to tech companies at McKinsey & Company. Dean graduated from Harvard University, where he met his Arphie co-founder Michael Chen. When he's not building Arphie, Dean spends time with his wife and their Samoyed puppy. --- Support this podcast: https://podcasters.spotify.com/pod/show/uncharted1/support

Andy talks with Alex Wang, the founder and CEO of Scale AI, to discuss the rapidly evolving landscape of artificial intelligence and its implications for national security. Forbes says Alex, at age 25, is the world's youngest self-made billionaire. He describes the three foundational pillars of AI—data, compute, and algorithms—and how advancements in each have driven recent progress. Alex discusses how AI advancement could enable adversaries to pose new threats to US national security, and the guardrails the technology may need. 

Guests: Varun Mohan, CEO & Co-Founder of Codeium; and Leigh Marie Braswell, partner at Kleiner Perkins“A lot of people are really bad at knowing what good is,” says Codeium CEO Varun Mohan. Specifically, he's thinking of startups that hire based on a “logo” — a well-known company on the résumé — rather than exceptional talent. Codeium is based in Mountain View, CA, and Varun believes that it's incumbent on any new startup to hire in the San Francisco Bay Area, because of how exceptional talent is concentrated there. “When you hire someone that's 10x better,” he says, “you can't replace them with 10 1x people. Because the the 10x person is going to be thinking of ideas that none of these 1x people are ever going to think of.”Chapters:(01:05) - Ludicrous growth (03:54) - Seizing opportunity (07:29) - Product-market fit (13:05) - Scale AI & MIT (17:42) - Coding efficiency (22:58) - Larger companies (25:20) - Varun and Leigh Marie's working relationship (29:51) - Pivoting to Codeium (34:00) - Giving away the product (37:01) - The code-gen landscape (42:20) - Annual reinvention (45:00) - Picking a problem (47:07) - Bipul Sinha's help (50:43) - Ambition (53:13) - Building in Silicon Valley (55:11) - Spotting talent (59:11) - Who Codeium is hiring (59:43) - What “grit” means to Varun Mentioned in this episode: Graham Moreno, Wiz, ChatGPT, Google, Nuro, Goldman Sachs, Waymo, the DARPA Challenge, Alex Wang, Douglas Chen, Safeway, Equinox, Carlos Delatorre and MongoDB, The Qualified Sales Leader by John McMahon, GitHub Copilot, Microsoft, Exafunction, Mamoon Hamid, Figma, JPMorgan Chase, Starlink, SpaceX, Rubrik, Michael Dell, Stripe, and John Doerr.Links:Connect with VarunLinkedInTwitterConnect with Leigh MarieLinkedInTwitterConnect with JoubinTwitterLinkedInEmail: grit@kleinerperkins.com Learn more about Kleiner PerkinsThis episode was edited by Eric Johnson from LightningPod.fm

This week we talk about the Double Reduction Policy, gaokao, and Chegg.We also discuss GPTs, cheating, and disruption.Recommended Book: Autocracy, Inc by Anne ApplebaumTranscriptIn July of 2021, the Chinese government implemented a new education rule called the Double Reduction Policy.This Policy was meant, among other things, to reduce the stress students in the country felt related to their educational attainment, while also imposing sterner regulations on businesses operating in education and education-adjacent industries.Chinese students spend a lot of time studying—nearly 10 hours per day for kids ages 12-14—and the average weekly study time for students is tallied at 55 hours, which is substantially higher than in most other countries, and quite a lot higher than the international average of 45 hours per week.This fixation on education is partly cultural, but it's also partly the result of China's education system, which has long served to train children to take very high-stakes tests, those tests then determining what sorts of educational and, ultimately, employment futures they can expect. These tests are the pathway to a better life, essentially, so the kids face a whole lot of pressure from society and their families to do well, because if they don't, they've sentenced themselves to low-paying jobs and concomitantly low-status lives; it's a fairly brutal setup, looked at from elsewhere around the world, but it's something that's kind of taken for granted in modern China.On top of all that in-class schoolwork, there's abundant homework, and that's led to a thriving private tutoring industry. Families invest heavily in ensuring their kids have a leg-up over everyone else, and that often means paying people to prepare them for those tests, even beyond school hours and well into the weekend.Because of all this, kids in China suffer abnormally high levels of physical and mental health issues, many of them directly linked to stress, including a chronic lack of sleep, high levels of anxiety, rampant obesity and everything that comes with that, and high levels of suicide, as well; suicide is actually the most common cause of death amongst Chinese teenagers, and the majority of these suicides occur in the lead-up to the gaokao, or National College Entrance Exam, which is the biggest of big important exams that determine how teens will be economically and socially sorted basically for the rest of their lives.This recent Double Reduction Policy, then, was intended to help temper some of those negative, education-related consequences, reducing the volume of homework kids had to tackle each week, freeing up time for sleep and relaxation, while also putting a cap on the ability of private tutoring companies to influence parents into paying for a bunch of tutoring services; something they'd long done via finger-wagging marketing messages, shaming parents who failed to invest heavily in their child's educational future, making them feel like they aren't being good parents because they're not spending enough on these offerings.This policy pursued these ends, first, by putting a cap on how much homework could be sent home with students, limiting it to 60 minutes for youngsters, and 90 minutes for middle schoolers.It also provided resources and rules for non-homework-related after-school services, did away with bad rankings due to poor test performance that might stigmatize students in the future, and killed off some of those fear-inducing, ever-so-important exams altogether.It also provided some new resources and frameworks for pilot programs that could help their school system evolve in the future, allowing them to try some new things, which could, in theory, then be disseminated to the nation's larger network of schools if these experiments go well.And then on the tutoring front, they went nuclear on those private tutoring businesses that were shaming parents into paying large sums of money to train their children beyond school hours.The government instituted a new system of regulators for this industry, ceased offering new business licenses for tutoring companies, and forced all existing for-profit businesses in this space to become non-profits.This market was worth about $100 billion when this new policy came into effect, which is a simply staggering sum, but the government basically said you're not businesses anymore, you can't operate if you try to make a profit.This is just one of many industries the current Chinese leadership has clamped-down on over the past handful of years, often on cultural grounds, as was the case with limiting the amount of time children can play video games each day. But like that video game ban, which has apparently shown mixed results, the tutoring ban seems to have led to the creation of a flourishing black market for tutoring services, forcing these sorts of business dealings underground, and thus increasing the fee parents pay for them each month.In late-October of 2024, the Chinese government, while not formally acknowledging any change to this policy, eased pressure on private tutoring services—the regulators in charge of keeping them operating in accordance with nonprofit structures apparently giving them a nudge and a wink, telling them surreptitiously that they're allowed to expand again—possibly because China has been suffering a wave of economic issues over the past several years, and the truncation of the tutoring industry led to a lot of mass-firings, tens of thousands of people suddenly without jobs, and a substantial drop in tax revenue, as well, as the country's stock market lost billions of dollars worth of value basically overnight.It's also worth noting here that China's youth unemployment rate recently hit 18.8%, which is a bogglingly high number, and something that's not great for stability, in the sense that a lot of young people, even very well educated young people, can't find a job, which means they have to occupy themselves with other, perhaps less productive things.But high youth unemployment is also not great for the country's economic future, as that means these are people who aren't attaining new skills and experience—and they can't do that because the companies that might otherwise hire them can't afford to pay more employees because folks aren't spending enough on their offerings.So while it was determined that this industry was hurting children and their families who had to pay these near-compulsory tutoring fees, they also seemed to realize that lacking this industry, their unemployment and broader economic woes would be further inflamed—and allowing for this gray area in the rules seems to be an attempt to have the best of both worlds, though it may leave them burdened with the worst of both worlds, as well.What I'd like to talk about today is another facet of the global tutoring industry, and how new technologies seem to be flooding into this zone even more rapidly than in other spaces, killing off some of the biggest players and potentially portending the sort of collapse we might also see in other industries in the coming years.—Chegg, spelled c-h-e-g-g, is a US-based, education-focused tech company that has provided all sorts of learning-related services to customers since 2006.It went public on the NYSE in 2013, and in 2021 it was called the “most valuable edtech company in America” by Forbes, due in part to the boom in long-distance education services in the early days of the Covid-19 pandemic; like Peloton and Zoom, Chegg was considered to be a great investment for a future in which more stuff is done remotely, as seemed likely to be the case for a good long while, considering all the distancing and shut-downs we were doing at the time.In early 2020, before that boom, the company was already reporting that it had 2.9 million subscribers to its Chegg Services offering, which gave users access to all sorts of school-related benefits, including help with homework, access to Q&As with experts, and a huge database of solutions for tests and assignments.The company then released a sort of social-publishing platform called Uversity in mid-2021, giving educators a place to share their own content, and they acquired a language-learning software company called Busuu, which is a bit like Duolingo, that same year for $436 million.In May of 2023, though, the company's CEO said, on an earnings call, that ChatGPT—the incredibly popular, basically overnight-popular large-language-model-powered AI chatbot created by OpenAI—might hinder Chegg's near-future growth.The day after that call, Chegg's stock price dropped by about 48%, cutting the company's market value nearly in half, and though later that same month he announced that Chegg would partner with OpenAI to launch its own AI platform called Cheggmate, which was launched as a beta in June, by early November the following year, 2024, the company had lost about 99% of its market valuation, dropping from a 2021 high of nearly $100 per share, down to less than $2 per share as of early November.This isn't a unique story: LLM-based AI tools, those made by OpenAI but also its competitors, including big tech companies like Google and Microsoft, which have really leaned into this seeming transition, have been messing with market valuations left and right, as this collection of tools and technologies have been evolving really fast—a recent five-year plan for Chegg indicated they didn't believe something like ChatGPT would exist until 2025 at the earliest, for instance, which turned out to be way off—but they've also been killing off high-flying company valuations because these sorts of tools are by definition multi-purpose, and a lot of the low-hanging fruit in any industry is basically just providing information that's already available somewhere in a more intuitive and accessible fashion; which is something a multi-purpose, bot-interfaced software tool is pretty good at doing, as it turns out.Chegg's services were optimized to provide school-related stuff to students—including test and homework answers those students could quickly reference if they wanted to study or cheat—and serving up these resources in a simple manner is what allowed them to pay the bills.ChatGPT and similar AI tools, though, can do the same, and for practically or literally—for the end-user, at least—free. And it can sometimes do so in a manner that's even more intuitive than the Cheggs of the world, even if these AI offerings are sometimes jumbled along the way; the risk-reward math is still favorable to a lot of people, because of just how valuable this kind of information provided in this way can be.Other companies and entire industries are finding themselves in the same general circumstances, also all of a sudden, because their unique value proposition has been offering some kind of information intuitively, or in some cases they've provided human interfaces that would do various things for customers: they would look up deals on a particular model of car, they would write marketing copy, they would commentate on sporting events.Some of these entities are trying to get ahead of the game, like Chegg did, by basically plugging their existing services into AI versions of the same, replacing their human commentators with bots that can manage a fair approximation of those now-unemployed humans, but at a fraction of the cost. Others are facing a huge number of new competitors, as smaller businesses or just individuals are realizing they can pay a little money for AI tokens and credits, plug an API into a website, which allows that AI to populate content on their site automatically, and they can then run the same sort of service with little or no effort, and vitally, little or no overhead.This creates a race-to-the-bottom situation in many such cases, and often the bots are nowhere near as good as the humans they're replacing, but especially in situations where human jobs have been optimized so that one human can be replaced with another human relatively simply, it has proven to be fairly easy to fire people and then replace them with non-humans that seem human-enough most of the time.So blog-writing and video-making and inventory-organizing and, yes, school-tutoring and similar services are increasingly being automated in this way, and while, sure, you could pay a premium to stick with Chegg and access these AI tools via their portal for $20 a month, the bet many investors are making is that folks will probably prefer to get what amounts to the same thing cheaper, or even free, directly from the source, or via one of those other lower-end intermediaries with fewer overhead costs.Chegg has lost about $14.5 billion in market value since early 2021, and the company is now expected to collapse under the weight of its debts sometime in the near-future; the shift in fortunes brought about by the deployment of generally capable, if not perfectly capable, chat-interface accessible AI tools has been that sudden.None of which means this is a permanent thing, as entities in industries currently being challenged by AI equivalent or near-equivalent tools might push back with their own, difficult to replicate offerings, and there's a chance that the small but burgeoning wave of vehemently non-AI tools—those that wave their human-made-ness, their non-AI-ness like a flag, or like an organic, cruelty-free label—might carve out their own sustainable, growable niche. That becomes their unique value proposition in place of what these AI-focused companies stole from them.But this kind of disruption sometimes leads to an extinction-level event for the majority of operators in a formerly flourishing space.Chegg, for their part, decided to revamp their AI offering, moving away from the Cheggmate name and working with Scale AI instead of OpenAI, to build a few dozen AI systems optimized for different academic focuses; which could prove to be a valuable differentiator for them, but it could also fall flat in the face of OpenAI's own re-skinned versions of ChatGPT, called GPTs, which allow users to do basically the same thing, coming up with their own field focused experts and personalities, rather than using the vanilla model of the bot.There's a chance this will also help Chegg deal with another AI-related issue—specifically, that ChatGPT was providing better answers to some students' questions than Chegg's human-derived offerings; they're trying to out-bot OpenAI, essentially, doing the homework-AI thing better than ChatGPT, and there's a chance that offering a demonstrably higher quality of answers might also serve as a survival-enabling differentiator; though their ability to consistently provide better answers in this way is anything but certain.It's also worth noting that what we're talking about here, so far, isn't the sci-fi dream of a perfect digital tutor—something like the Young Lady's Illustrated Primer from Neal Stephenson's novel The Diamond Age, which is something like an AI-powered storybook that adapts its content to the reader, and which then teaches said reader everything they need to know to flourish in life, day by day. Chegg and ChatGPT serve up tools that help students cheat on tests and homework, while also helping them look up information a lot easier when they decide not to cheat, and to practice various sorts of assignments and exams beforehand.So this is a far easier space to compete in than something more complex and actually tutor-like. It may be, then, that moving in that direction, toward tools that focus more on replacing teachers and tutors, rather than helping students navigate schoolwork, might be the killer app that allows some of these existing tutoring-ish tools to survive and thrive; though it may be that something else comes along in the meantime which fulfills that promise better—maybe ChatGPT, or maybe some new, more focused version of the same general collection of tools.It'll probably be a few years before we see how this and similar bets that're being made by at-risk companies facing the AI barbarians at the gate turn out, and at that point these tools will likely be even more powerful, offering even more capabilities and thus disrupting, or threatening to disrupt, even more companies in even more industries, as a consequence.Show Noteshttps://www.wsj.com/tech/ai/how-chatgpt-brought-down-an-online-education-giant-200b4ff2https://openai.com/index/introducing-gpts/https://ai.wharton.upenn.edu/focus-areas/human-technology-interaction/2024-ai-adoption-report/https://www.weforum.org/stories/2024/07/ai-tutor-china-teaching-gaps/https://en.wikipedia.org/wiki/Double_Reduction_Policyhttps://journals.sagepub.com/doi/full/10.1177/20965311241265123https://www.sciencedirect.com/science/article/abs/pii/S0738059324000117https://archive.ph/VKkrLhttps://www.japantimes.co.jp/news/2023/07/22/asia-pacific/china-private-tutoring/https://www.nbcnews.com/news/world/chinas-youth-unemployment-hits-fresh-high-economic-slowdown-restrictiv-rcna172183 This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit letsknowthings.substack.com/subscribe