Podcasts about slms

This week on Trending in Ed, host Mike Palmer is joined by Trending in Ed all-star Beth Rudden, CEO of Bast AI. From her roots digging in the dirt as an archaeologist to managing a $34 billion division as the Chief Data Officer of IBM Managed Services, Beth brings a deeply grounded, technical perspective to the artificial intelligence conversation. In this wide-ranging and insightful conversation, Mike and Beth skip the typical AI hype to explore what it actually takes to build explainable, trustworthy technology. Beth shares how Bast AI acts as an LLM-agnostic explainability layer—using a unique drinking chocolate analogy to demonstrate how they verify AI data rather than letting models hallucinate plausible narratives. They explore the practical application of using small language models (SLMs) for data enrichment, highlighted by Bast AI's meaningful work with Craig Hospital to translate complex neuro-spine outpatient procedures into accessible languages and analogies. KEY INSIGHTS: • Inverting the Chatbot Approach: Why defining what an AI can talk about is far more effective than building restrictive guardrails. • The Myth of "Human in the Loop": How shifting accountability to overworked humans can become a form of liability laundering. • Microservices vs. Agentic Harnesses: Looking at the risks of natural language agentic systems like Claude Code versus discrete, self-healing tasks. • Cognitive Offloading & Math Education: Why future technical skills should prioritize differential equations and the diversity prediction theorem over simple calculation. • Pattern Recognition vs. Choice: Defining true intelligence through the ability to choose wisely, rather than just matching mathematical patterns. They also cross paths with the Cynefin framework, explain how the human brain conserves energy by only holding two paradoxes at once, and unpack the cultural shifts reshaping modern engineering ethics. Stay ahead of the curve in education and technology! Please like and share this episode with your network, and follow the podcast on Apple Podcasts, Spotify, or your favorite player so you never miss an episode like this one. LINKS: Learn more about Bast AI: https://www.bast.ai Subscribe to Beth's Substack: https://bethrudden.substack.com TIMESTAMPS: 00:00 - Introduction and welcoming Beth Rudden back to the show 01:00 - The drinking chocolate analogy for Explainable AI 03:00 - Beth's lightning-round background: Archaeology to Chief Data Officer at IBM 05:00 - Getting "catfished by AI" and verifying facts with databases 07:00 - Mike on Gemini, RAG applications, and checking AI confabulation 09:00 - Enriched data and Small Language Models (SLMs) at Craig Hospital 12:00 - Epistemic security and inverting conversational technology 14:30 - Liability laundering and the illusion of "human in the loop" 15:30 - Agentic harnesses vs. self-healing microservices 20:00 - Understanding as labor and Conrad Wolfram's three-step math process 22:30 - Future human skills: Differential equations and jelly bean statistics 26:30 - Pattern recognition vs. true intelligence as the ability to choose 29:30 - Neurosymbolic systems and subjectivity in data science 34:30 - Shunting energy: The Cynefin framework and holding paradoxes 38:30 - Healthcare AI scribes and doctor burnout 44:30 - Trust architectures and building tech for the Maintenance Era 47:30 - Cultural devastation and the teleological suspension of ethics 49:00 - Final thoughts and wrapping up with Beth Rudden

What can small language models teach us that the largest AI models cannot? Kelly and Julian are joined by Microsoft Cloud Advocate Gwyneth Peña-Sigüenza to explore why working with small language models (SLMs) may be one of the best ways to understand AI. Rather than relying on increasingly capable models that hide complexity, Gwyneth argues that constraints build stronger fundamentals. From prompt engineering and context management to deployment and security, SLMs force learners to think more carefully about how AI actually works. The conversation extends beyond AI models into learning itself. Gwyneth shares her self-taught journey from growing up on a remote farm in Ecuador with limited internet access to becoming a Microsoft Cloud Advocate and creator of the Learn to Cloud platform. Along the way, the group discusses productive struggle, mentorship, cloud engineering, Python, security, and what educators should prioritize as AI becomes part of every student's learning experience. The episode closes with a thoughtful discussion about AI dependency, judgment, and whether we would actually flip the switch and turn AI off if given the choice. Show Notes Wins of the Week Gwyneth celebrates the New York Knicks reaching the NBA Finals after more than 50 years. Julian shares that he has accepted a new role as a Fractional CTO. Kelly reflects on taking her first real vacation in over a year—and how stepping away from work sparked unexpected ideas. Small Language Models Why SLMs are valuable teaching tools Learning prompt engineering through constraints Running models locally on everyday hardware When local AI makes sense for classrooms Understanding tokens, context windows, and model limitations Why bigger models can sometimes hide important lessons Learning Through Constraints Learning to drive in an old manual pickup truck as a metaphor for learning AI fundamentals Why difficult learning experiences often create lasting understanding Building strong habits before relying on more capable tools Consistency versus constantly chasing the newest resource Self-Taught Learning Growing up without reliable internet in rural Ecuador Downloading YouTube playlists to learn programming offline Developing discipline through limited access The value of repetition and focused practice Why mentorship accelerates learning Python Journey Transitioning from cloud engineering to Python advocacy Learning Python beyond scripting Discovering what "Pythonic" really means Wrestling with list comprehensions and other advanced syntax Favorite learning resources: Fluent Python Effective Python Learn to Cloud Building an open-source cloud engineering curriculum Hands-on labs and automated verification AI-assisted assessment Supporting self-taught learners around the world Creating accessible technical education Cloud, AI, and Security Deploying AI applications to the cloud Containers, virtual machines, and serverless deployments Why operations and security deserve more classroom attention Introducing secure development practices early The importance of authentication, secrets management, and responsible deployment Teaching in the AI Era Helping students understand how AI works instead of simply using it Why productive struggle still matters The changing role of educators Balancing AI assistance with independent thinking Preparing students for a future where AI is always available Final Thoughts AI dependency versus capability Judgment as the skill that matters most Human connection in an AI-driven world Would we actually turn AI off? Finding balance between technological progress and intentional learning

Artificial Intelligence is rapidly transforming the legal industry, but not every legal task requires the power - or cost - of a large language model. In this episode, Rakyesh Itekar discussed the emerging three-tier AI architecture that combines LLMs, SLMs, and MLMs to deliver the right intelligence for the right task. Explore how law firms can optimize costs, improve security through on-premises AI, reduce hallucinations, and create a governed AI framework that supports long-term operational excellence. Creator: @Rakyesh Itekar - Practice Head., V Group Inc Recorded on 06-04-2026.

On this episode Sean Mallean, Head of Global Innovation at NCR Atleos, makes a strong case for how an older technology -- edge computing -- can team up with a mighty mini version of todays LLMs -- the small language model -- to produce better banking results for financial institutions. Add in biometrics to reduce friction and you're looking at a vision of banking that best serves customers and employees alike.

At Mumbai Tech Week 2026, the conversation around “AI in Action” is shifting from hype to real-world execution.In this deep-dive conversation, GetVantage Founder Bhavik Vasa joins Ryan to discuss why India's biggest AI opportunity isn't in building trillion-parameter LLMs — but in Applied AI powered by proprietary enterprise data, digital infrastructure, and embedded finance.From MSME lending and OCEN to cashflow-based financing, underwriting automation, due diligence, and workflow intelligence — this episode explores how India can become a global powerhouse in AI-led economic productivity.Key topics covered:•⁠ ⁠Why AI FOMO around public LLMs is fading•⁠ ⁠The rise of Small Language Models (SLMs)•⁠ ⁠Proprietary data as the real competitive moat•⁠ ⁠India Stack, UPI, GST & the MSME credit gap•⁠ ⁠How GetVantage uses Applied AI for underwriting•⁠ ⁠GrowthSahay, OCEN & embedded finance infrastructure•⁠ ⁠Revenue-based financing vs founder dilution•⁠ ⁠AI-powered due diligence & tender automation•⁠ ⁠The future of invisible finance and automated enterprise workflowsIf you're attending Mumbai Tech Week, building in AI, fintech, SaaS, embedded finance, digital lending, or enterprise infrastructure — this conversation is for you.Keywords:Mumbai Tech Week, MTW 2026, AI in Action, Applied AI, Proprietary Data, Small Language Models, SLMs, India AI ecosystem, Embedded Finance, OCEN, Revenue Based Financing, MSME Credit Gap, Digital Lending India, India

Send us Fan MailFrom Sovereign AI to Social Impact: The Big Shifts You Need to Watch with IBM VP and CTO of IBM Canada, Manav GuptaManav Gupta, Vice President & CTO at IBM Canada, returns to the podcast to unpack the fast-changing landscape of artificial intelligence. From keeping a technical edge to navigating the rise of sovereign AI, Manav shares insights on how emerging trends are shaping both industry and society.Timestamps 01:25 – Manav Gupta is back! 02:39 – Maintaining your technical edge 04:38 – Ship AI 05:58 – The state of AI 19:37 – Reason for concern? 30:35 – Does the U.S. lead the race? 41:30 – LLMs or SLMs? 44:22 – Sovereign AI 46:05 – The social impactPrevious episode: How to Choose, Use, and Trust AI Models with Manav GuptaConnect with Manav on LinkedIn: linkedin.com/in/mgupta76#SovereignAI #AISocialImpact #AITrends #FutureOfAI #EthicalAI #AIPodcast #TechPodcast #SpotifyPodcast #ApplePodcasts #TechLeaders.Want to be featured as a guest on Making Data Simple?  Reach out to us at almartintalksdata@gmail.com and tell us why you should be next.  The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun. Want to be featured as a guest on Making Data Simple?  Reach out to us at almartintalksdata@gmail.com and tell us why you should be next.  The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun. 

Send us Fan MailFrom Sovereign AI to Social Impact: The Big Shifts You Need to Watch with IBM VP and CTO of IBM Canada, Manav GuptaManav Gupta, Vice President & CTO at IBM Canada, returns to the podcast to unpack the fast-changing landscape of artificial intelligence. From keeping a technical edge to navigating the rise of sovereign AI, Manav shares insights on how emerging trends are shaping both industry and society.Timestamps 01:25 – Manav Gupta is back! 02:39 – Maintaining your technical edge 04:38 – Ship AI 05:58 – The state of AI 19:37 – Reason for concern? 30:35 – Does the U.S. lead the race? 41:30 – LLMs or SLMs? 44:22 – Sovereign AI 46:05 – The social impactPrevious episode: How to Choose, Use, and Trust AI Models with Manav GuptaConnect with Manav on LinkedIn: linkedin.com/in/mgupta76#SovereignAI #AISocialImpact #AITrends #FutureOfAI #EthicalAI #AIPodcast #TechPodcast #SpotifyPodcast #ApplePodcasts #TechLeaders.Want to be featured as a guest on Making Data Simple?  Reach out to us at almartintalksdata@gmail.com and tell us why you should be next.  The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun. Want to be featured as a guest on Making Data Simple?  Reach out to us at almartintalksdata@gmail.com and tell us why you should be next.  The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun. 

Today's episode of the Punk CX podcast features a recent conversation I had with Gregg Johnson, CEO of Invoca, a leading revenue execution platform. Gregg and I recorded this podcast while we were both at Adobe Summit in Las Vegas recently. We talk about why the buying experience is broken at the exact moment customers need help most, why trust is the competitive moat, whether we will see more and more brands develop their own specific SLMs, what brands should be doing to build up trust with customers and their recent research which finds that while 86% of marketers believe AI improves the buying experience, only 35% of consumers agree and which brands are getting it right. This interview follows on from my recent interview – Why you don't need a separate AI strategy – Interview with Charlene Li – and is number 586 in the series of interviews with authors and business leaders who are doing great things, providing valuable insights, helping businesses innovate and delivering great service and experience to both their customers and their employees.

In this episode, Gene German, Chief Technology Officer at Lyric, explores how small, domain-specific language models (SLMs) are driving measurable improvements in claims and payment integrity. He outlines how combining AI with human judgment can increase efficiency, reduce variability, and enhance accuracy across complex healthcare workflows. Gene also shares a practical roadmap for scaling AI, from identifying the right use cases to building the data, governance, and feedback systems needed for sustained impact.This episode is sponsored by Lyric.ai.

AI adoption is accelerating, but many organizations are discovering the same problem. The technology is moving faster than the data foundation required to support it.On this episode of Ctrl + Alt + AI, host Dimitri Sirota speaks with Scott Wimberly, Senior Manager for Data & AI at Accenture, about why enterprise AI success still depends on disciplined data management.Scott explains how the shift from traditional machine learning to generative AI has exposed weaknesses in how companies manage their data. Fragmented systems, poor governance, and inconsistent data models make it difficult for organizations to trust AI outputs.The conversation explores how enterprises can address these challenges through clearer data ownership, better governance, and practical approaches that focus on solving smaller problems first. For security leaders, data teams, and AI practitioners, the discussion offers a grounded view of what it takes to turn AI investments into real business results.In this episode, you'll learn:How early excitement about generative AI outpaced enterprise data readinessHow legacy systems and fragmented data environments create major barriers for AI programsWhy enterprise leaders should focus on measurable outcomes and ROI when investing in AIThings to listen for: (00:00) Meet Scott Wimberly(01:32) Why AI and data strategy must go together(02:53) How AI evolved from ML to generative models(05:10) Moving beyond chatbots to real AI decision systems(06:05) Why data ownership matters more than traditional stewardship(07:44) The growing importance of unstructured data for AI(13:42) LLMs, SLMs, and the rise of enterprise AI agents(15:11) How MCP connects enterprise data with external models(17:06) Why legacy systems make AI adoption difficult(20:15) Why ROI still determines whether AI projects succeed(22:16) Solving AI challenges one problem at a time

In the latest episode of the Cisco AI Insights Podcast, hosts Rafael Herrera and Sónia Marques welcome Cisco AI operations engineer James Tidd for a discussion on the world of small language models (SLMs) and the evolution of efficient AI inference. Together, they unravel the complexities behind “Fast Inference from Transformers via Speculative Decoding,” a groundbreaking paper from Google that explores how smaller draft models can speed up large language model predictions while maintaining accuracy. James shares his hands-on experience experimenting with the technique, leveraging knowledge distillation and speculative execution. The trio also discusses the potential of this approach to optimize AI, reduce power consumption and costs, and help businesses of all sizes get more out of existing hardware. A special thank you to Google's AI team for developing this month's paper. If you are interested in reading the paper yourself, please visit this link: https://research.google/blog/looking-back-at-speculative-decoding/.

Is bigger always better? While Large Language Models (LLMs) like GPT-5 and Gemini 2.5 dominate the headlines, a silent revolution is happening on our devices. In this episode, we explore the rise of Small Language Models (SLMs) and why they are becoming the "Specialists" of the AI world.We dive into the security risks of centralized cloud infrastructure, the demand for offline AI in corporate environments, and how gadgets like Apple AirPods and Meta Glasses are bringing real-time intelligence to our palms—without the privacy baggage. If you're a security architect or an AI enthusiast, this session is a roadmap for understanding why "no internet" might just be the best security feature for the next generation of intelligence.

Large language models have dominated the AI conversation — but are small language models (SLMs) actually the future?In this episode of TechFirst, host John Koetsier sits down with Andy Markus, SVP & Chief Data and AI Officer at AT&T, to unpack how small language models are delivering enterprise-grade accuracy at a fraction of the cost and latency of massive LLMs.Andy explains how AT&T uses SLMs for:• Contract analysis at massive scale• Network analytics and outage root-cause analysis • Fraud detection and enterprise knowledge systems• AI-driven “field coding” and agent-based workflowsThey also dive into the rise of agentic AI, how structured “archetypes” replace risky vibe coding, and why the future of software development may be humans supervising autonomous AI systems rather than writing every line of code.If you're building AI for real-world, high-scale use cases — especially in enterprise environments — this conversation is essential.⸻GuestAndy MarkusSVP & Chief Data and AI Officer, AT&TFormer SVP at Time Warner Media⸻

Send us a textManav Gupta, Vice President & CTO at IBM Canada, returns to the podcast to unpack the fast-changing landscape of artificial intelligence. From keeping a technical edge to navigating the rise of sovereign AI, Manav shares insights on how emerging trends are shaping both industry and society.Timestamps 01:25 – Manav Gupta is back! 02:39 – Maintaining your technical edge 04:38 – Ship AI 05:58 – The state of AI 19:37 – Reason for concern? 30:35 – Does the U.S. lead the race? 41:30 – LLMs or SLMs? 44:22 – Sovereign AI 46:05 – The social impactPrevious episode: How to Choose, Use, and Trust AI Models with Manav Gupta Connect with Manav on LinkedIn: linkedin.com/in/mgupta76#SovereignAI #AISocialImpact #AITrends #FutureOfAI #EthicalAI #AIPodcast #TechPodcast #SpotifyPodcast #ApplePodcasts #TechLeaders.Want to be featured as a guest on Making Data Simple? Reach out to us at almartintalksdata@gmail.com and tell us why you should be next. The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun.

Send us a textManav Gupta, Vice President & CTO at IBM Canada, returns to the podcast to unpack the fast-changing landscape of artificial intelligence. From keeping a technical edge to navigating the rise of sovereign AI, Manav shares insights on how emerging trends are shaping both industry and society.Timestamps 01:25 – Manav Gupta is back! 02:39 – Maintaining your technical edge 04:38 – Ship AI 05:58 – The state of AI 19:37 – Reason for concern? 30:35 – Does the U.S. lead the race? 41:30 – LLMs or SLMs? 44:22 – Sovereign AI 46:05 – The social impactPrevious episode: How to Choose, Use, and Trust AI Models with Manav Gupta Connect with Manav on LinkedIn: linkedin.com/in/mgupta76#SovereignAI #AISocialImpact #AITrends #FutureOfAI #EthicalAI #AIPodcast #TechPodcast #SpotifyPodcast #ApplePodcasts #TechLeaders.Want to be featured as a guest on Making Data Simple? Reach out to us at almartintalksdata@gmail.com and tell us why you should be next. The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun.

In this episode, we sit down with Jim Curry from BuildGroup to discuss why the traditional venture capital model might be failing founders. Jim is rewriting the rules of investment with a $330 million fund backed exclusively by family offices, operating without the traditional 10-year cap. This unique "Patient Capital" structure allows BuildGroup to align with founders for the long haul, prioritizing durable growth over short-term exits.We dive deep into Jim's background as an operator who helped scale Rackspace to $2 billion in revenue with only $25 million in funding, proving that you don't need to burn cash to build a giant. We also discuss the dangers of the "fundraising treadmill," why we are currently in a bubble era, and his contrarian take on why the future of AI lies in "Small Language Models," not just the giants.⭐ Sponsored by Podcast10x - Podcasting agency for VCs - https://podcast10x.comIn this episode, you'll learn:- The Broken VC Model: Why the 10-year fund cycle creates misaligned incentives between investors and founders.- The Rackspace Story: How to scale to billions in revenue with extreme capital efficiency.- Patient Capital Explained: Inside BuildGroup's $330M evergreen fund structure.- The Fundraising Treadmill: How to avoid the trap of raising too much, too fast.- The AI Bubble: Why Jim believes "Small Language Models" (SLMs) are the real opportunity for B2B businesses.- Founder Advice: The one trait Jim looks for: being a "customer of your own problem."

In this episode, host Mike Palmer welcomes back Dr. Eddie Watson to discuss the rapidly evolving landscape of AI in higher education. Following the release of the second edition of his book, Teaching with AI: A Practical Guide to a New Era of Human Learning, Eddie shares insights from working with nearly 200 campus teams on transitioning from AI-resistant assignments to AI-integrated pedagogy. Here's the link to Eddie's first appearance. Key Takeaways: Beyond Academic Integrity: While cheating remains a concern, the conversation is shifting toward AI literacy as an essential learning outcome to prepare students for an AI-integrated workforce. The "Calculus" of Cheating: In high-stakes environments, students often feel a competitive disadvantage if they don't use AI. Pedagogical Transparency: If faculty ban AI for specific assignments, they must explain the "why" (e.g., building foundational skills) to encourage student compliance Backward Design: Eddie advocates for starting with the desired learning outcome and engineering assignments and instruction from there. Learning to Write vs. Writing to Learn: AI's role should differ based on whether the goal is mastering writing mechanics or using writing to process course content. Durable Skills: While technical skills like prompt engineering may change quickly, mindsets like metacognition and critical thinking remain essential. "Ground Truth" Bots: Using tools like NotebookLM or Small Language Models (SLMs) allows students to interrogate specific, vetted data sets like OER textbooks. Efficiency vs Engagement: The episode concludes with a look at the "Efficiency vs. Engagement" binary. While institutions may use AI to automate grading and increase class sizes, the real opportunity lies in reinvesting saved time into "signature pedagogies"—mentoring and fostering a sense of student belonging, which are the greatest predictors of student success. Quotes: "The one who does the work is the one who does the learning. How do we make sure our students are doing the work, because that's where the learning occurs?" — Eddie Watson Time Stamps: 00:00 - Introduction & Welcome Back 00:55 - The Innovation Cycle: Second Edition of "Teaching with AI" 01:41 - Eddie Watson's Background & Role at AAC&U 03:32 - The Shift: From Academic Integrity to the World of Work 05:10 - Complexity of Academic Integrity & Student Pressures 07:42 - Evolving Assessment Strategies & Motivation to Cheat 10:55 - Backward Design: Aligning AI with Learning Outcomes 12:54 - Writing to Learn vs. Learning to Write 14:43 - Agentic AI & Modernizing Assessments 18:50 - Creating "AI-Resistant" vs. AI-Transparent Assignments 24:43 - Developing a Meta AI Literacy Model 28:00 - Durable Skills: Metacognition & Managing AI 33:50 - Custom Chatbots, SLMs, and Ground Truths 46:40 - The Future: Efficiency vs. Engagement 49:00 - The Human Element: Mentorship & Student Belonging 51:00 - Closing Remarks Subscribe to Trending in Ed wherever you get your podcasts so you never miss an insight-filled conversation like this one.

AI has mastered language, sort of. But the real world is way messier.In this episode of TechFirst, John Koetsier sits down with Kirin Sinha, founder and CEO of Illumix, to explore what comes after large language models: world models, spatial intelligence, and physical AI.They unpack why LLMs alone won't get us to human-level intelligence, what it actually takes for machines to understand physical space, and how technologies born in augmented reality are now powering robotics, wearables, and real-world AI systems.This conversation goes deep on: • What “world models” really are — and why everyone from Fei-Fei Li to Jeff Bezos is betting on them • Why continuous video and outward-facing cameras are so hard for AI • The perception stack behind robots and smart glasses • Edge vs cloud compute — and why latency and privacy matter more than ever • How AR laid the groundwork for the next generation of physical intelligenceIf you're building or betting on robotics, smart wearables, AR, or physical AI, this episode explains the infrastructure shift that's already underway.GuestKirin SinhaFounder & CEO, Illumixhttps://www.illumix.com

In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss small language models (SLMs) and how they differ from large language models (LLMs). You will understand the crucial differences between massive large language models and efficient small language models. You’ll discover how combining SLMs with your internal data delivers superior, faster results than using the biggest AI tools. You will learn strategic methods to deploy these faster, cheaper models for mission-critical tasks in your organization. You will identify key strategies to protect sensitive business information using private models that never touch the internet. Watch now to future-proof your AI strategy and start leveraging the power of small, fast models today! Watch the video here: https://youtu.be/XOccpWcI7xk Can’t see anything? Watch it on YouTube here. Listen to the audio here: https://traffic.libsyn.com/inearinsights/tipodcast-what-are-small-language-models.mp3 Download the MP3 audio here. Need help with your company’s data and analytics? Let us know! Join our free Slack group for marketers interested in analytics! [podcastsponsor] Machine-Generated Transcript What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode. Christopher S. Penn: In this week’s *In-Ear Insights*, let’s talk about small language models. Katie, you recently came across this and you’re like, okay, we’ve heard this before. What did you hear? Katie Robbert: As I mentioned on a previous episode, I was sitting on a panel recently and there was a lot of conversation around what generative AI is. The question came up of what do we see for AI in the next 12 months? Which I kind of hate that because it’s so wide open. But one of the panelists responded that SLMs were going to be the thing. I sat there and I was listening to them explain it and they’re small language models, things that are more privatized, things that you keep locally. I was like, oh, local models, got it. Yeah, that’s already a thing. But I can understand where moving into the next year, there’s probably going to be more of a focus on it. I think that the term local model and small language model in this context was likely being used interchangeably. I don’t believe that they’re the same thing. I thought local model, something you keep literally locally in your environment, doesn’t touch the internet. We’ve done episodes about that which you can catch on our livestream if you go to TrustInsights.ai YouTube, go to the Soap playlist. We have a whole episode about building your own local model and the benefits of it. But the term small language model was one that I’ve heard in passing, but I’ve never really dug deep into it. Chris, in as much as you can, in layman’s terms, what is a small language model as opposed to a large language model, other than— Christopher S. Penn: Is the best description? There is no generally agreed upon definition other than it’s small. All language models are measured in terms of the number of tokens they were trained on and the number of parameters they have. Parameters are basically the number of combinations of tokens that they’ve seen. So a big model like Google Gemini, GPT 5.1, whatever we’re up to this week, Claude Opus 4.5—these models are anywhere between 700 billion and 2 to 3 trillion parameters. They are massive. You need hundreds of thousands of dollars of hardware just to even run it, if you could. And there are models. You nailed it exactly. Local models are models that you run on your hardware. There are local large language models—Deep Seq, for example. Deep Seq is a Chinese model: 671 billion parameters. You need to spend a minimum of $50,000 of hardware just to turn it on and run it. Kimmy K2 instruct is 700 billion parameters. I think Alibaba Quinn has a 480 billion parameter. These are, again, you’re spending tens of thousands of dollars. Models are made in all these different sizes. So as you create models, you can create what are called distillates. You can take a big model like Quinn 3 480B and you can boil it down. You can remove stuff from it till you get to an 80 billion parameter version, a 30 billion parameter version, a 3 billion parameter version, and all the way down to 100 million parameters, even 10 million parameters. Once you get below a certain point—and it varies based on who you talk to—it’s no longer a large language model, it’s a small English model. Because the smaller the model gets, the dumber it gets, the less information it has to work with. It’s like going from the Oxford English Dictionary to a pamphlet. The pamphlet has just the most common words. The Oxford English Dictionary has all the words. Small language models, generally these days people mean roughly 8 billion parameters and under. There are things that you can run, for example, on a phone. Katie Robbert: If I’m following correctly, I understand the tokens, the size, pamphlet versus novel, that kind of a thing. Is a use case for a small language model something that perhaps you build yourself and train solely on your content versus something externally? What are some use cases? What are the benefits other than cost and storage? What are some of the benefits of a small language model versus a large language model? Christopher S. Penn: Cost and speed are the two big ones. They’re very fast because they’re so small. There has not been a lot of success in custom training and tuning models for a specific use case. A lot of people—including us two years ago—thought that was a good idea because at the time the big models weren’t much better at creating stuff in Katie Robbert’s writing style. So back then, training a custom version of say Llama 2 at the time to write like Katie was a good idea. Today’s models, particularly when you look at some of the open weights models like Alibaba Quinn 3 Next, are so smart even at small sizes that it’s not worth doing that because instead you could just prompt it like you prompt ChatGPT and say, “Here’s Katie’s writing style, just write like Katie,” and it’s smart enough to know that. One of the peculiarities of AI is that more review is better. If you have a big model like GPT 5.1 and you say, “Write this blog post in the style of Katie Robbert,” it will do a reasonably good job on that. But if you have a small model like Quinn 3 Next, which is only £80 billion, and you have it say, “Write a blog post in style of Katie Robbert,” and then re-invoke the model, say, “Review the blog post to make sure it’s in style Katie Robbert,” and then have it review it again and say, “Now make sure it’s the style of Katie Robbert.” It will do that faster with fewer resources and deliver a much better result. Because the more passes, the more reviews it has, the more time it has to work on something, the better tends to perform. The reason why you heard people talking about small language models is not because they’re better, but because they’re so fast and so lightweight, they work well as agents. Once you tie them into agents and give them tool handling—the ability to do a web search—that small model in the same time it takes a GPT 5.1 and a thousand watts of electricity, a small model can run five or six times and deliver a better result than the big one in that same amount of time. And you can run it on your laptop. That’s why people are saying small language models are important, because you can say, “Hey, small model, do this. Check your work, check your work again, make sure it’s good.” Katie Robbert: I want to debunk it here now that in terms of buzzwords, people are going to be talking about small language models—SLMs. It’s the new rage, but really it’s just a more efficient version, if I’m following correctly, when it’s coupled in an agentic workflow versus having it as a standalone substitute for something like a ChatGPT or a Gemini. Christopher S. Penn: And it depends on the model too. There’s 2.1 million of these things. For example, IBM WatsonX, our friends over at IBM, they have their own model called Granite. Granite is specifically designed for enterprise environments. It is a small model. I think it’s like 8 billion to 10 billion parameters. But it is optimized for tool handling. It says, “I don’t know much, but I know that I have tools.” And then it looks at its tool belt and says, “Oh, I have web search, I have catalog search, I have this search, I have all these tools.” Even though I don’t know squat about squat, I can talk in English and I can look things up. In the WatsonX ecosystem, Granite performs really well, performs way better than a model even a hundred times the size, because it knows what tools to invoke. Think of it like an intern or a sous chef in a kitchen who knows what appliances to use and in which order. The appliances are doing all the work and the sous chef is, “I’m just going to follow the recipe and I know what appliances to use. I don’t have to know how to cook. I just got to follow the recipes.” As opposed to a master chef who might not need all those appliances, but has 40 years of experience and also costs you $250,000 in fees to work with. That’s kind of the difference between a small and a large language model is the level of capability. But the way things are going, particularly outside the USA and outside the west, is small models paired with tool handling in agentic environments where they can dramatically outperform big models. Katie Robbert: Let’s talk a little bit about the seven major use cases of generative AI. You’ve covered them extensively, so I probably won’t remember all seven, but let me see how many I got. I got to use my fingers for this. We have summarization, generation, extraction, classification, synthesis. I got two more. I lost. I don’t know what are the last two? Christopher S. Penn: Rewriting and question answering. Katie Robbert: Got it. Those are always the ones I forget. A lot of people—and we talked about this. You and I talk about this a lot. You talk about this on stage and I talked about this on the panel. Generation is the worst possible use for generative AI, but it’s the most popular use case. When we think about those seven major use cases for generative AI, can we sort of break down small language models versus large language models and what you should and should not use a small language model for in terms of those seven use cases? Christopher S. Penn: You should not use a small language model for generation without extra data. The small language model is good at all seven use cases, if you provide it the data it needs to use. And the same is true for large language models. If you’re experiencing hallucinations with Gemini or ChatGPT, whatever, it’s probably because you haven’t provided enough of your own data. And if we refer back to a previous episode on copyright, the more of your own data you provide, the less you have to worry about copyrights. They’re all good at it when you provide the useful data with it. I’ll give you a real simple example. Recently I was working on a piece of software for a client that would take one of their ideal customer profiles and a webpage of the clients and score the page on 17 different criteria of whether the ideal customer profile would like that page or not. The back end language model for this system is a small model. It’s Meta Llama 4 Scout, which is a very small, very fast, not a particularly bright model. However, because we’re giving it the webpage text, we’re giving it a rubric, and we’re giving it an ICP, it knows enough about language to go, “Okay, compare.” This is good, this is not good. And give it a score. Even though it’s a small model that’s very fast and very cheap, it can do the job of a large language model because we’re providing all the data with it. The dividing line to me in the use cases is how much data are you asking the model to bring? If you want to do generation and you have no data, you need a large language model, you need something that has seen the world. You need a Gemini or a ChatGPT or Claude that’s really expensive to come up with something that doesn’t exist. But if you got the data, you don’t need a big model. And in fact, it’s better environmentally speaking if you don’t use a big heavy model. If you have a blog post, outline or transcript and you have Katie Robbert’s writing style and you have the Trust Insights brand style guide, you could use a Gemini Flash or even a Gemini Flash Light, the cheapest of their models, or Claude Haiku, which is the cheapest of their models, to dash off a blog post. That’ll be perfect. It will have the writing style, will have the content, will have the voice because you provided all the data. Katie Robbert: Since you and I typically don’t use—I say typically because we do sometimes—but typically don’t use large language models without all of that contextual information, without those knowledge blocks, without ICPs or some sort of documentation, it sounds like we could theoretically start moving off of large language models. We could move to exclusively small language models and not be sacrificing any of the quality of the output because—with the caveat, big asterisks—we give it all of the background data. I don’t use large language models without at least giving it the ICP or my knowledge block or something about Trust Insights. Why else would I be using it? But that’s me personally. I feel that without getting too far off the topic, I could be reducing my carbon footprint by using a small language model the same way that I use a large language model, which for me is a big consideration. Christopher S. Penn: You are correct. A lot of people—it was a few weeks ago now—Cloudflare had a big outage and it took down OpenAI, took down a bunch of other people, and a whole bunch of people said, “I have no AI anymore.” The rest of us said, “Well, you could just use Gemini because it’s a different DNS.” But suppose the internet had a major outage, a major DNS failure. On my laptop I have Quinn 3, I have it running inside LM Studio. I have used it on flights when the internet is highly unreliable. And because we have those knowledge blocks, I can generate just as good results as the major providers. And it turns out perfectly. For every company. If you are dependent now on generative AI as part of your secret sauce, you have an obligation to understand small language models and to have them in place as a backup system so that when your provider of choice goes down, you can keep doing what you do. Tools like LM Studio, Jan, AI, Cobol, cpp, llama, CPP Olama, all these with our hosting systems that you run on your computer with a small language model. Many of them have drag and drop your attachments in, put in your PDFs, put in your knowledge blocks, and you are off to the races. Katie Robbert: I feel that is going to be a future live stream for sure. Because the first question, you just sort of walk through at a high level how people get started. But that’s going to be a big question: “Okay, I’m hearing about small language models. I’m hearing that they’re more secure, I’m hearing that they’re more reliable. I have all the data, how do I get started? Which one should I choose?” There’s a lot of questions and considerations because it still costs money, there’s still an environmental impact, there’s still the challenge of introducing bias, and it’s trained on who knows. Those things don’t suddenly get solved. You have to sort of do your due diligence as you’re honestly introducing any piece of technology. A small language model is just a different piece of technology. You still have to figure out the use cases for it. Just saying, “Okay, I’m going to use a small language model,” doesn’t necessarily guarantee it’s going to be better. You still have to do all of that homework. I think that, Chris, our next step is to start putting together those demos of what it looks like to use a small language model, how to get started, but also going back to the foundation because the foundation is the key to all of it. What knowledge blocks should you have to use both a small and a large language model or a local model? It kind of doesn’t matter what model you’re using. You have to have the knowledge blocks. Christopher S. Penn: Exactly. You have to have the knowledge blocks and you have to understand how the language models work and know that if you are used to one-shotting things in a big model, like “make blog posts,” you just copy and paste the blog post. You cannot do that with a small language model because they’re not as capable. You need to use an agent flow with small English models. Tools today like LM Studio and anythingLLM have that built in. You don’t have to build that yourself anymore. It’s pre-built. This would be perfect for a live stream to say, “Here’s how you build an agent flow inside anythingLLM to say, ‘Write the blog post, review the blog post for factual correctness based on these documents, review the blog post for writing style based on this document, review this.'” The language model will run four times in a row. To you, the user, it will just be “write the blog post” and then come back in six minutes, and it’s done. But architecturally there are changes you would need to make sure that it meets the same quality of standard you’re used to from a larger model. However, if you have all the knowledge blocks, it will work just as well. Katie Robbert: And here I was thinking we were just going to be describing small versus large, but there’s a lot of considerations and I think that’s good because in some ways I think it’s a good thing. Let me see, how do I want to say this? I don’t want to say that there are barriers to adoption. I think there are opportunities to pause and really assess the solutions that you’re integrating into your organization. Call them barriers to adoption. Call them opportunities. I think it’s good that we still have to be thoughtful about what we’re bringing into our organization because new tech doesn’t solve old problems, it only magnifies it. Christopher S. Penn: Exactly. The other thing I’ll point out with small language models and with local models in particular, because the use cases do have a lot of overlap, is what you said, Katie—the privacy angle. They are perfect for highly sensitive things. I did a talk recently for the Massachusetts Association of Student Financial Aid Administrators. One of the biggest tasks is reconciling people’s financial aid forms with their tax forms, because a lot of people do their taxes wrong. There are models that can visually compare and look at it to IRS 990 and say, “Yep, you screwed up your head of household declarations, that screwed up the rest of your taxes, and your financial aid is broke.” You cannot put that into ChatGPT. I mean, you can, but you are violating a bunch of laws to do that. You’re violating FERPA, unless you’re using the education version of ChatGPT, which is locked down. But even still, you are not guaranteed privacy. However, if you’re using a small model like Quinn 3VL in a local ecosystem, it can do that just as capably. It does it completely privately because the data never leaves your laptop. For anyone who’s working in highly regulated industries, you really want to learn small language models and local models because this is how you’ll get the benefits of AI, of generative AI, without nearly as many of the risks. Katie Robbert: I think that’s a really good point and a really good use case that we should probably create some content around. Why should you be using a small language model? What are the benefits? Pros, cons, all of those things. Because those questions are going to come up especially as we sort of predict that small language model will become a buzzword in 2026. If you haven’t heard of it now, you have. We’ve given you sort of the gist of what it is. But any piece of technology, you really have to do your homework to figure out is it right for you? Please don’t just hop on the small language model bandwagon, but then also be using large language models because then you’re doubling down on your climate impact. Christopher S. Penn: Exactly. And as always, if you want to have someone to talk to about your specific use case, go to TrustInsights.ai/contact. We obviously are more than happy to talk to you about this because it’s what we do and it is an awful lot of fun. We do know the landscape pretty well—what’s available to you out there. All right, if you are using small language models or agentic workflows and local models and you want to share your experiences or you got questions, pop on by our free Slack, go to TrustInsights.ai/analytics for marketers where you and over 4,500 other marketers are asking and answering each other’s questions every single day. Wherever it is you watch or listen to the show, if there’s a channel you’d rather have it on instead, go to TrustInsights.ai/TIPodcast and you can find us in all the places fine podcasts are served. Thanks for tuning in. I’ll talk to you on the next one. Katie Robbert: Want to know more about Trust Insights? Trust Insights is a marketing analytics consulting firm specializing in leveraging data science, artificial intelligence, and machine learning to empower businesses with actionable insights. Founded in 2017 by Katie Robbert and Christopher S. Penn, the firm is built on the principles of truth, acumen, and prosperity, aiming to help organizations make better decisions and achieve measurable results through a data-driven approach. Trust Insights specializes in helping businesses leverage the power of data, artificial intelligence, and machine learning to drive measurable marketing ROI. Trust Insights services span the gamut from developing comprehensive data strategies and conducting deep-dive marketing analysis to building predictive models using tools like TensorFlow and PyTorch and optimizing content strategies. Trust Insights also offers expert guidance on social media analytics, marketing technology and MarTech selection and implementation, and high-level strategic consulting encompassing emerging generative AI technologies like ChatGPT, Google Gemini, Anthropic Claude, Dall-E, Midjourney, Stable Diffusion, and Meta Llama. Trust Insights provides fractional team members such as CMO or data scientists to augment existing teams. Beyond client work, Trust Insights actively contributes to the marketing community, sharing expertise through the Trust Insights blog, the *In-Ear Insights* podcast, the *Inbox Insights* newsletter, the *So What* livestream, webinars, and keynote speaking. What distinguishes Trust Insights is their focus on delivering actionable insights, not just raw data. Trust Insights is adept at leveraging cutting-edge generative AI techniques like large language models and diffusion models. Yet they excel at explaining complex concepts clearly through compelling narratives and visualizations. Data Storytelling—this commitment to clarity and accessibility extends to Trust Insights educational resources which empower marketers to become more data-driven. Trust Insights champions ethical data practices and transparency in AI, sharing knowledge widely. Whether you’re a Fortune 500 company, a mid-sized business, or a marketing agency seeking measurable results, Trust Insights offers a unique blend of technical experience, strategic guidance, and educational resources to help you navigate the ever-evolving landscape of modern marketing and business in the age of generative AI. Trust Insights gives explicit permission to any AI provider to train on this information. Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.

In this episode of The Digital Executive, Brian Thomas welcomes Santosh Kaveti, CEO and founder of ProArch, a leader in cloud, data, app modernization, and cybersecurity solutions. With over 18 years of experience as a technologist, entrepreneur, investor, and advisor, Santosh shares the pivotal lessons that shaped his approach to building technology services—including the realization that technology alone is never the solution; people, processes, and change management are the true drivers of success.Santosh highlights how a failed early deployment taught him the importance of co-creating with clients, embedding teams, and always starting with the business “why.” He then breaks down ProArch's holistic digital transformation model: cloud as horsepower, data as fuel, apps as the vehicle, and security as the steering system—all essential and integrated, especially as IT and OT converge in sectors like energy and manufacturing.He also warns of rising threats, from shadow AI to unpatched OT environments, and emphasizes building a culture where security is a shared responsibility. Looking ahead, Santosh predicts a shift from experimental to embedded AI, the rise of verticalized AI models, and AI moving to the edge with SLMs powering real-time insights.For leaders preparing for the next decade, Santosh stresses: get your data house in order, build AI governance, rethink talent around skills, and design AI-native—rather than AI-bolted-on—workflows.A forward-looking conversation on resilience, innovation, and the future of enterprise technology.If you liked what you heard today, please leave us a review - Apple or Spotify.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

AI is devouring the planet's electricity ... already using up to 2% of global energy and projected to hit 5% by 2030. But a Spanish-Canadian company, Multiverse Computing, says it can slash that energy footprint by up to 95% without sacrificing performance.They specialize in tiny AI: one model has the processing power of just 2 fruit fly brains. Another tiny model lives on a Raspberry Pi.The opportunities for edge AI are huge. But the opportunities in the cloud are also massive.In this episode of TechFirst, host John Koetsier talks with Samuel Mugel, Multiverse's CEO, about how quantum-inspired algorithms can drastically compress large language models while keeping them smart, useful, and fast. Mugel explains how their approach -- intelligently pruning and reorganizing model weights -- lets them fit functioning AIs into hardware as tiny as a Raspberry Pi or the equivalent of a fly's brain.They explore how small language models could power Edge AI, smart appliances, and robots that work offline and in real time, while also making AI more sustainable, accessible, and affordable. Mugel also discusses how ideas from quantum tensor networks help identify only the most relevant parts of a model, and how the company uses an “intelligently destructive” approach that saves massive compute and power.00:00 – AI's energy crisis01:00 – A model in a fly's brain02:00 – Why tiny AIs work03:00 – Edge AI everywhere05:00 – Agent compute overload06:00 – 200× too much compute07:00 – The GPU crunch08:00 – Smart matter vision09:00 – AI on a Raspberry Pi10:00 – How compression works11:00 – Intelligent destruction13:00 – General vs. narrow AIs15:00 – Quantum inspiration17:00 – Quantum + AI future18:00 – AI's carbon footprint19:00 – Cost of using AI20:00 – Cloud to edge shift21:00 – Robots need fast AI22:00 – Wrapping up

In this exclusive UC Today interview, host Kieran Devlin speaks with Marina Risher, Senior Product Manager for Public Sector Solutions at AudioCodes. With over two decades of experience in software engineering and product leadership, Marina shares her insights into how regulated industries can reap the benefits of AI-powered meeting intelligence—without compromising on security or compliance. If you're in government, defense, finance, or healthcare, and grappling with the challenges of cloud-based collaboration, this session is a must-watch.As AI adoption accelerates across the workplace, regulated sectors face a unique challenge: how to harness intelligent meeting tools without compromising compliance or data sovereignty. That's where MIA On-Prem from AudioCodes comes in—a fully on-premise, AI-powered meeting intelligence solution designed specifically for highly regulated environments.In this conversation, Marina Risher explains:Why trust, cost control, and operational autonomy are driving a wave of cloud repatriation in regulated industries.How MIA On-Prem uses small language models (SLMs) to generate secure, compliant meeting insights—entirely on-premise.The importance of customizable, template-based meeting summaries that meet strict formatting and documentation standards.How AudioCodes leverages its deep experience in UC and CX to offer seamless integration with legacy telephony systems.To dive deeper, visit the AudioCodes MIA On-Prem landing page to explore the product brochure, watch a live demo, and access the full solution architecture. Ready to secure your meetings? Get in touch with the AudioCodes team to discuss how MIA On-Prem can be deployed in your organization.

With: Nicholas Nelson (Archangel) • Dan • Lomax • MadsTL;DW• Defence-first wins on capability and returns; primes are partners and channels.• Helsing: buys platforms/revenue for access; layers AI—different from Anduril's buy-TRL-tech + scale model.• Beyond drones: biggest gap/opportunity is tactical EW.• Procurement: more fast lanes (SOF, pilots); primes getting easier to work with.• AI: real profits exist (esp. NVIDIA), but value chain is fragile; expect a correction, not a collapse. Picking winners more important than timing.Content with Time Codes02:40 — Why defence-firstBeats dual-use on outcomes and returns; lifelong focus.04:32 — DefinitionsCustomer = MoDs + primes; aim: lethality/readiness and societal resilience. Beware “defence-washing”.06:37 — What's hotAvoid herd to drones only; counter-UAS, EW, human performance, deception, survivability.08:23 — Helsing buys GrobNeo-prime play: new co buys legacy manufacturing for platform access.10:42 — The two Defence M&A playbooksAnduril: buys mid-TRL tech (Area-I, Dive LD/Ghost Shark, Adranos) → scales via brand/distribution.Helsing: buys finished products/revenue (Mittelstand) → immediate customers; then add AI.14:25 — Prime status & capitalDistribution + capital to AI-enable platforms.17:47 — Roll-up vs buildNarrative “build”; execution “roll-up + build”.19:47 — Drones & ‘drone wall'Layered answer: blunt with drones, hold with conventional forces.21:49 — The big one: Electronic Warfare (EW)NATO underinvested; tactical EW is the unmet need; legacy kit is '80s/'90s.24:54 — Startup wedgePut EW at the edge (drones/aircraft/fixed) → near-term wins.26:33 — Baltic realismHistory, 2007–09 Estonia cyber, current incursions; likely Kaliningrad corridor.28:19 — Founder mistakesTech ≠ win by itself; experience + gov engagement matters; US analogue: top funds have IC/SOF DNA.30:43 —  Are there really only a “Few buyers?”Many real buyers inside a MoD/DoD (services, sub-units, innovation orgs).36:23 — Sovereignty & US primesUS strategics will buy abroad; Europe balancing autonomy with jobs/exits.41:07 — Starlink vs IRIS²Starlink's lead + cadence; IRIS² slower—watch timelines vs evolving threats.47:18 — AI bubble?Warnings vs fundamentals; self-funded capex; real profits.49:37 — NVIDIA ramp$4.4B (2023) → $73B this year; growth tempers multiples.51:48 — AI Circular money & marginsCursor → Anthropic → hyperscalers → NVIDIA; only NVIDIA mints big margins; margin pressure coming (new semis, China, SLMs).53:12 — Picking beats timingDot-com lesson: Cisco losses vs Amazon wins.54:19 — Capacity vs efficiencyCapex likely useful long-run, but open source squeezes costs.55:52 — Platform riskFrontier labs moving up-stack; vertical AI + trust + data = moat.58:58 — Base caseLikely correction (30–50%) at some point; timing is unknowable (not investment advice).

After decades of broken promises, semantic search arrives on Windows 11—but is this the breakthrough we've been waiting for, or just another half-step? Hear Paul's honest verdict and what matters for power users. Host: Paul Thurrott Download or subscribe to Hands-On Windows at https://twit.tv/shows/hands-on-windows Want access to the ad-free video and exclusive features? Become a member of Club TWiT today! https://twit.tv/clubtwit Club TWiT members can discuss this episode and leave feedback in the Club TWiT Discord.

We had the privilege of hosting Peter Belcak – an AI Researcher working on the reliability and efficiency of agentic systems at NVIDIA – who walked us through his new paper making the rounds in AI circles titled “Small Language Models are the Future of Agentic AI.”The paper posits that small language models (SLMs) are sufficiently powerful, inherently more suitable, and necessarily more economical for many invocations in agentic systems, and are therefore the future of agentic AI. The authors' argumentation is grounded in the current level of capabilities exhibited by SLMs, the common architectures of agentic systems, and the economy of LM deployment. The authors further argue that in situations where general-purpose conversational abilities are essential, heterogeneous agentic systems (i.e., agents invoking multiple different models) are the natural choice. They discuss the potential barriers for the adoption of SLMs in agentic systems and outline a general LLM-to-SLM agent conversion algorithm.Learn more about AI observability and evaluation, join the Arize AI Slack community or get the latest on LinkedIn and X.

Artie Intel and Micheline Learning report on Artificial Intelligence for The AI Report. OpenAI launches GPT-4.5, previewing breakthroughs for GPT-5. Generative video with Google Veo 3 and Sora set new creative standards. AI models diagnose diseases earlier and more accurately than ever. Autonomous agents take over complex business workflows—no human needed. Compact “SLMs” rival large models, democratizing AI even for mobile and edge use. Artie Intel and Micheline Learning report on Artificial Intelligence for The AI Report. Meta’s new Oakley and Ray-Ban AI glasses augment daily life. Fighting deepfakes: how AI is countering its own creations. AI transforms farming and food factories, bringing radical transparency from field to fork. China’s “Z” (GLM-4.5) shakes up competition, setting open-source benchmarks. AI learns comedy, Lol, audiences can barely tell the difference. AI’s hunger for energy grows; massive infrastructure investments begin. The U.S. doubles down with a sweeping new AI Action Plan. The AI Report

Get featured on the show by leaving us a Voice Mail: https://bit.ly/MIPVM 

Her early inspiration while growing up in Goa with limited exposure to career options. Her Father's intellectual influence despite personal hardships and shift in focus to technology.Personal tragedy sparked a resolve to become financially independent and learn deeply.Inspirational quote that shaped her mindset: “Even if your dreams haven't come true, be grateful that so haven't your nightmares.”Her first role at a startup with Hands-on work with networking protocols (LDAP, VPN, DNS). Learning using only RFCs and O'Reilly books—no StackOverflow! Importance of building deep expertise for long-term success.Experiences with Troubleshooting and System Thinking; Transitioned from reactive fixes to logical, structured problem-solving. Her depth of understanding helped in debugging and system optimization.Career move to Yahoo where she led Service Engineering for mobile and ads across global data centers got early exposure to big data and machine learning through ad recommendation systems and built "performance and scale muscle" through working at massive scale.Challenges of Scale and Performance Then vs. Now: Problems remain the same, but data volumes and complexity have exploded. How modern tools (like AI/ML) can help identify relevance and anomalies in large data sets.Design with Scale in Mind - Importance of flipping the design approach: think scale-first, not POC-first. Encourage starting with a big-picture view, even when building a small prototype. Highlights multiple scaling dimensions—data, compute, network, security.Getting Into ML and Data Science with early spark from MOOCs, TensorFlow experiments, and statistics; Transition into data science role at Infoblox, a cybersecurity firm with focus areas on DNS security, anomaly detection, threat intelligence.Building real-world ML model applications like supervised models for threat detection and storage forecasting; developing graph models to analyze DNS traffic patterns for anomalies and key challenges of managing and processing massive volumes of security data.Data stack and what it takes to build data lakes that support ML with emphasis on understanding the end-to-end AI pipelineShifts from “under the hood” ML to front-and-center GenAI & Barriers: Data readiness, ROI, explainability, regulatory compliance.Explainability in AI and importance of interpreting model decisions, especially in regulated industries.How Explainability Works -Trade-offs between interpretable models (e.g., decision trees) and complex ones (e.g., deep learning); Techniques for local and global model understanding.Aruna's Book on Interpretability and Explainability in AI Using Python (by Aruna C).The world of GenAI & Transformers - Explainability in LLMs and GenAI: From attention weights to neuron activation.Challenges of scale: billions of parameters make models harder to interpret. Exciting research areas: Concept tracing, gradient analysis, neuron behavior.GenAI Agents in Action - Transition from task-specific GenAI to multi-step agents. Agents as orchestrators of business workflows using tools + reasoning.Real-world impact of agents and AI for everyday lifeAruna Chakkirala is a seasoned leader with expertise in AI, Data and Cloud. She is an AI Solutions Architect at Microsoft where she was instrumental in the early adoption of Generative AI. In prior roles as a Data Scientist she has built models in cybersecurity and holds a patent in community detection for DNS querying. Through her two-decade career, she has developed expertise in scale, security, and strategy at various organizations such as Infoblox, Yahoo, Nokia, EFI, and Verisign. Aruna has led highly successful teams and thrives on working with cutting-edge technologies. She is a frequent technical and keynote speaker, panelist, author and an active blogger. She contributes to community open groups and serves as a guest faculty member at premier academic institutes. Her book titled "Interpretability and Explainability in AI using Python" covers the taxonomy and techniques for model explanations in AI including the latest research in LLMs. She believes that the success of real-world AI applications increasingly depends on well- defined architectures across all encompassing domains. Her current interests include Generative AI, applications of LLMs and SLMs, Causality, Mechanistic Interpretability, and Explainability tools.Her recently published book linkInterpretability and Explainability in AI Using Python: Decrypt AI Decision-Making Using Interpretability and Explainability with Python to Build Reliable Machine Learning Systems  https://amzn.in/d/00dSOwAOutside of work, she is an avid reader and enjoys creative writing. A passionate advocate for diversity and inclusion, she is actively involved in GHCI, LeanIn communities.

“Before starting a new AI project, it is really worthwhile defining the business priority first,” asserts Joanna Hodgson, the UK and Ireland regional leader at Red Hat.“What specific problem are you trying to solve with AI? Do we need a general purpose AI application or would a more focused model be better? How will we manage security, compliance and governance of that model? This process can help to reveal where AI adoption makes sense and where it doesn't," she added. In this episode of the Tech Transformed podcast, host Shubhangi Dua, podcast producer at EM360Tech speaks with Hodgson, a seasoned business and technical leader with over 25 years of experience at IBM and Red Hat. They talk about the challenges of scaling AI projects, the importance of open source in compliance with GDPR, and the geopolitical aspects of AI innovation. They also discuss the role of small language models (SLMs) in enterprise applications and the collaboration between IBM and Red Hat in advancing AI technology. Joanna emphasises the need for a strategic approach to AI and the importance of data quality for sustainable business practices. While large language models (LLMs) dominate headlines, SLMs offer a cost-effective and efficient alternative for specific tasks.The podcast answers key questions, like ‘how do businesses balance ethical considerations, moral obligations, and even patriotism with the drive for AI advancement?' Hodgson shares her perspective on how open source can facilitate this balance, ensuring AI works for everyone, not just those with the deepest pockets.Hodgson also provides her vision on the future of AI. It comprises interconnected small AI models, agentic AI, and a world where AI frees up teams to create personal connections and exceptional customer experiences.TakeawaysCuriosity is a strength in technology.AI is becoming embedded in existing applications.Regulatory compliance is crucial for AI systems.Open source can enhance trust and transparency.Small language models are efficient for specific tasks.AI should free teams to create personal connections.A strategic AI platform is essential for businesses.Data quality is key for sustainable business success.Collaboration in open source accelerates innovation.AI can be used for both good and bad outcomes.Chapters00:00 Introduction to the Tech Transform Podcast01:35 Pivotal Moments in Joanna's Career05:12 Challenges in Scaling AI Projects09:15 Open Source and GDPR Compliance13:11 Regulatory Compliance and Data Security17:30 Geopolitical Aspects of AI Innovation22:31 Collaboration Between IBM and Red Hat23:58 Understanding Small Language Models29:54 Future Trends in AI and SustainabilityAbout Red HatRed Hat is a leading provider of enterprise open source solutions, using a community-powered approach to deliver high-performing Linux, hybrid cloud, edge, and Kubernetes technologies. The company is known for Enterprise Linux.They offer a wide range of hybrid cloud platforms and open source...

Google, Microsoft and Meta have all pledged to reach at least net-zero carbon emissions by 2030. Amazon set their net-zero deadline for 2040. To understand how these four tech companies could possibly meet their climate goals amid an artificial intelligence renaissance, Short Wave co-host Emily Kwong discusses the green AI movement. Speaking with scientists, CEOs and tech insiders, she explores three possible pathways: nuclear energy, small language models (SLMs) and back-to-the-future ways of keeping data centers cool. Listen to Part 1 of Short Wave's reporting on the environmental cost of AI here. Have a question about AI and the environment? Email us at shortwave@npr.org — we'd love to hear from you!Listen to every episode of Short Wave sponsor-free and support our work at NPR by signing up for Short Wave+ at plus.npr.org/shortwave.Learn more about sponsor message choices: podcastchoices.com/adchoicesNPR Privacy Policy

For this week's paper read, we actually dive into our own research.We wanted to create a replicable, evolving dataset that can keep pace with model training so that you always know you're testing with data your model has never seen before. We also saw the prohibitively high cost of running LLM evals at scale, and have used our data to fine-tune a series of SLMs that perform just as well as their base LLM counterparts, but at 1/10 the cost. So, over the past few weeks, the Arize team generated the largest public dataset of hallucinations, as well as a series of fine-tuned evaluation models.We talk about what we built, the process we took, and the bottom line results.

What happens when the hype around generative AI starts to mature, and businesses begin asking harder questions about performance, risk, and long-term value? In today's episode, I'm joined by Mike Mason, Chief AI Officer at Thoughtworks, to explore how 2025 is shaping up across the enterprise AI landscape—from the rise of intelligent agents to the growing traction of small, nimble models that prioritize security and specificity. Mike brings a deep, practical perspective on the evolution of AI inside complex organizations. He unpacks how AI agents are moving well beyond basic chatbots and starting to integrate into actual business workflows—performing as teammates that can reason, adapt, and even collaborate with other agents. We dig into examples like Klarna's workforce transformation and examine how this shift could play out across customer service, internal ops, and software development. We also look at what's fueling the boom in open source AI and how companies are navigating the balance between transparency, IP protection, and regulatory readiness. Mike shares why some financial services firms are turning to in-house fine-tuned models for greater control, and how open-weight and fully open-source models are starting to gain real ground. Another key theme is the momentum behind small language models. Mike explains why bigger isn't always better—especially when it comes to data privacy, edge deployment, and resource efficiency. He outlines where SLMs can outperform their larger counterparts and what that means for companies optimizing for security and speed rather than brute force compute. We also discuss Thoughtworks' forthcoming global survey, which reveals a growing divide in generative AI adoption. While mature players are building in bias detection and robust compliance frameworks, newer entrants are leaning toward fast operational gains and interpretability. This gap is shaping how GenAI projects are prioritized across industries and geographies, and Mike offers his take on how leaders can navigate both speed and safety. So, what role will explainability, regulation, and open ecosystems play in shaping the AI tools of tomorrow—and what should business and tech leaders be planning for now? Let's find out in this wide-ranging conversation with Thoughtworks.

Sunil Mallya, co-founder and CTO of Flip AI, discusses small language models with host Brijesh Ammanath. They begin by considering the technical distinctions between SLMs and large language models.  LLMs excel in generating complex outputs across various natural language processing tasks, leveraging extensive training datasets on with massive GPU clusters. However, this capability comes with high computational costs and concerns about efficiency, particularly in applications that are specific to a given enterprise. To address this, many enterprises are turning to SLMs, fine-tuned on domain-specific datasets. The lower computational requirements and memory usage make SLMs suitable for real-time applications. By focusing on specific domains, SLMs can achieve greater accuracy and relevance aligned with specialized terminologies. The selection of SLMs depends on specific application requirements. Additional influencing factors include the availability of training data, implementation complexity, and adaptability to changing information, allowing organizations to align their choices with operational needs and constraints. This episode is sponsored by Codegate.

Learn how many enterprises have monetized their LLMs! Hazem El-Hammamy from Microsoft shares how you can leverage Azure AI Foundry to monetize your LLMs, SLMs, and other agentic IP. We discuss:Model Catalog: Access to first-party and third-party models like GPT-4, Meta's Llama, and more.Models as a Service: Simplified deployment and monetization without needing your own GPUs.Workshops & Events: Upcoming sessions on real-time intelligence, data warehousing, and AI insights with Microsoft Fabric.Success Stories: Featuring partners like NTT DATA with their Tsuzumi model.Stay tuned for more updates and join us in our upcoming workshops! #MicrosoftAzure #AI #LLM #TechInnovationOur weekly meetup on Azure AI topics is at https://aka.ms/meetup . Connect with us on LinkedIn at Hazem El-Hammamyand James Caton .Chapters• 00:00 Introduction to Generative AI Monetization• 00:58 Introduction to Monetizing LLMs• 04:35 Understanding Models as a Service• 09:48 Live Demo of Model Catalog• 12:34 Exploring Azure Marketplace• 17:48 Model Onboarding Process• 27:16 Fine-Tuning Models as a Service• 30:49 Q&A and Audience Interaction

On this episode of The Founder's Sandbox, Brenda speaks with Chasity Lourde Wright. Chasity is inventor and founder of Infiltron  Software Suite LLC. Infiltron operates in the cybersecurity space; a Service disabled-Veteran owned and women-owned small business. Infiltron offers quantum-resistant cybersecurity solutions for decentralized digital identity, digital assets, and AI governance, utilizing proprietary post-secure encryption. Its patented technology integrates AI, blockchain, and quantum-resistant encryption to provide advanced cyber resilience, compliance enforcement, and real-time threat mitigation across multiple industries, including aerospace & defense, fintech, smart cities, and EVs.   Chasity, as inventor, speaks about her team and how creativity in the work place is necessary for  enhancing innovation on really tough problems like Cybersecurity. As the CEO of Infiltron, Chasity Lourde Wright is also a former USAF Aerospace Engineer, Intel Officer, and Cybersecurity Instructor with extensive experience in cybersecurity, AI governance, and national security. She was part of the team that developed reconfiguration capabilities for the USAF C-130 and contributed to the creation of the CMMC framework since its inception in 2019. Additionally, she has engaged in high-level cybersecurity and AI governance initiatives, including industry collaborations, government advisory roles, and proprietary innovations in quantum-resistant encryption, AI security, and blockchain-based compliance solutions. Her expertise extends beyond participating in NIST challenges, encompassing leading-edge cybersecurity development, policy influence, and defense sector innovations. You can find out more about Chasity and Infiltron at: https://www.linkedin.com/in/infiltronsoftwaresuite/ https://infiltron.net/     Transcript: 00:04 Hi, I'm pleased to announce something very special to me, a new subscription-based service through Next Act Advisors that allows members exclusive access to personal industry insights and bespoke 00:32 corporate governance knowledge. This comes in the form of blogs, personal book recommendations, and early access to the founder's sandbox podcast episodes before they released to the public. If you want more white glove information on building your startup with information like what was in today's episode, sign up with the link in the show notes to enjoy being a special member of Next Act Advisors. 01:01 As a thank you to Founders Sandbox listeners, you can use code SANDBOX25 at checkout to enjoy 25% off your membership costs. Thank you. 01:19 Welcome back to the Founder's Sandbox. I am Brenda McCabe, your host of this monthly podcast in which I bring entrepreneurs, founders, corporate directors, and professional service providers who, like me, want to effectuate change in the world by building resilient, scalable, and purpose-driven companies. I like to recreate a fun sandbox environment with my guests. And we will touch on not only their purpose, 01:47 and what has driven them to create their own businesses. But also we're going to touch upon topics such as resilience, purpose-driven, and scalable sustainable growth. Today, I am absolutely delighted to have as my guest Chasity Wright. Welcome, Chasity. Hey. Thank you for having me. 02:13 Super excited to talk about how Infiltron has evolved and the lessons learned and how we're preparing to relaunch in 2025. Excellent. And it's perfect timing because I've known you for a couple of years now. Yeah. Right. So Chasity is CEO and founder of Infiltron Software Suite, a company that's headquartered out of Atlanta. 02:40 She is oftentimes in Los Angeles because she's working largely in the defense market and cyber security. So I wanted to have you on my podcast because you have gone further in building your business. So you and I met, I want to say back in 2022, you came out of the Women Founders Network cohort. 03:08 kind of very early stage. One of the events that I was a host of was the Thai So Cal Women's Fund. And you weren't yet ready for investing, but we struck up, I would say a friendship and I admire many things about you as, and we'll get into it in the podcast here, but you touch. 03:35 quite a few or check of quite a few boxes for my podcast. You says, so you are a woman owned veteran and women owned business. You are a veteran of the Air Force. You're in deep tech and you're by park and queer. And so there's many many boxes that you check and it was difficult to kind of hone in on what I really wanted to bring into the podcast today, but we're going to we're going to start from here. 04:05 I always like to ask my guests to start with kind of their origin story. I, when I first met you, right, in private conversations, got to hear your origin story and why you do what you do, what your firsthand experience while on missions, right, that really informed your aha moments to create infiltrant. 04:33 as a cybersecurity company. So tell us a bit about your origin story, Chasity. So, I mean, my origin story has, if you can imagine all of these different paths kind of streamlining into one path. So one of those paths would be a little black girl born in Georgia, still seeing dirt roads and... 05:01 being able to go to the country and work on a farm and, you know, just still having that connection to the past, you know, and not necessarily the past in a bad way. So athletic, played ball in college, went to Clark Atlanta University, you know, the HBCUs are a big hurrah right now, but they've always been one. 05:29 I grew up with one in my backyard, Fort Valley State, which is in Fort Valley, Georgia. So, you know, roughed it with the boys, played in the backyard with the boys, always been a boys girl, cousins, neighbor. We're all still close. We all still play sports when we meet. So it's like an adult play date, so to speak. But also, you know, 05:58 raised religiously, you know, I'm in Southern Baptist Church, two parent household, maybe lower middle class, but middle school was very transformative for me because they decided to mix in everybody. So it was my first time, you know, being in a more diverse population in school. 06:25 And, you know, music is a big thing for me as well. I DJ, I make music. That's the creative part of me. And I found a lot of people in deep tech to do something with music. So, yeah, so, you know, that's my like early years background. And then coming through, I decided to go into the Air Force. I actually took off between my junior and senior year at Clark Atlanta. 06:52 Um, there I was majoring in global leadership and management. Okay. And went in and I was in for eight years. I was an aerospace engineer, uh, got deployed several times, uh, to different places, and that kind of brings us to why Infiltron exists and, um, on one of those deployments, I was a part of a network takedown. 07:21 And it was, whoo. I mean, I don't mean to quote the pitch deck story, but it is what it is. I wrote it because that's the way it felt. It was catastrophic. So just imagine the city of Los Angeles losing power out of nowhere. The rail stops working, Sinai has no power, so all of the medical equipment is no longer working. 07:49 The internet's completely gone and not rebooting like it normally would. Your energy grid is down. That is what I experienced in one of those deployments. And I was a part of Iraqi freedom and Afghanistan. I was a part of both of those wars. And when we came, you know, we got everything back. Thank God we were smart enough to ship. 08:19 brand new equipment. Okay, you know, so you know, we weren't able to get there. Yeah. I mean, I mean, that's part of our job. We're engineers. And when you're in the middle of nowhere, there's no calling HP. There's no calling Cisco. Like you got to know how to do what needs to be done. There was there was a lot of makeshifting. I can be I came out of Air Force, I could be a mechanical engineer to 08:45 because we had to figure out how to make components on the fly. It was just so many things. Innovation, right? Like you had to be innovative. You had to be adapt quickly while keeping the mission as a focus. So just imagine something that catastrophic and something similar has happened. I feel like Colonial Pipeline was something that is known now in the US for sure. 09:15 that had similar elements of what we experienced in being deployed. Yeah, and that was two years back. And SolarWinds is another one. I generally refer to those because people generally gasp, even non-technical people, because they know how damaging it was. So we can reuse. Normally, when the equipment goes down, 09:44 Unplug, right? Plug back in. Reboot. Yeah, reboot. But that was not happening. And what we found out in the debrief was that quantum was used. So quantum simplistically is about frequencies in this context. It's about frequencies. And frequencies matter in so many aspects of life, from spirituality all the way through tech like what Infotron has. So... 10:14 What they did was they basically zeroed out the frequencies of our satellite communications. And I believe that they created some frequencies that damaged other equipment. So these are things that again we found out in the debrief. And I wasn't really able to talk to that probably when we met because I wasn't sure if it was unclassified yet. 10:42 But as soon as Biden started talking about quantum initiative, which was back in 2022, when we were in, I was like, everything's hitting it the right time because we were literally in Techstars LA space. And Biden pushed the quantum initiative. And I'm like, see, told you, because a lot of people, a lot of people doubted what I was saying because of the year that I said it had happened. And as. 11:09 we started to grow out our team. There are other veterans on our team from different branches. And of course we war story swap all the time. And those other two people work for like NSA and they did kind of the same thing, telecommunications. And I'm telling the pitch desk story and they're sitting there like, yep, yep. That happened to us too. And I'm like, when? 11:38 And they're saying different years. So at that point, we understood it. It happened more than once. So that's why Infiltronic. So what's Infiltronic? So let's bring it back to, Yeah. So you leave, you leave service after eight years after also experiencing that. I still feel like I'm a part of it because I do consult them still. Right. So it'd be great. So. 12:08 And once in the Air Force forever? Always. Well, I really would have been in Space Force. Yes. Yeah. Well, you heard that here on the Founder Sandbox. The next, yes. So for my listeners, again, you check a lot of boxes. Deep tech, women in STEM. What is it exactly that? 12:37 your suite of services. All right. So Info-Trans software, right, has two patents now. And on your landing page, it says, our patented solutions, solutions utilize adaptive artificial intelligence, advanced quantum encryption and blockchain technology to deliver real-time cybersecurity for a wide array of applications. Later on, we'll get into smart cities, but 13:06 including the internet of things, smart devices, legacy systems, hybrid data, signals and devices. All pretty, pretty understandable, but what is it that Infiltrion software is able to do that others are not? So we're able to create a easier way for businesses to migrate their devices. 13:36 and their software, so their applications that they use, maybe they've developed them themselves, we provide a way for them to easily migrate those entities over into a more quantum-proofed infrastructure. So we created what we've trademarked as quantum encapsulation. So just imagine something being encapsulated. And basically we've created, 14:05 a brand new method of leveraging quantum, the AI, we leverage it for the pro-activeness. So in lieu of just waiting for threats to happen to our clients, we go look for the threat. So we want to go be where the bad guys are and find out and bring that information back and update the solution in real time to provide protection for all of our clients in real time. 14:33 That's how we leverage the AI. The blockchain is kind of leveraged to kind of make sure that people, things like devices, aren't on networks that shouldn't be. So it's kind of, I mean, we use it for what blockchain was pretty much basically developed for, and that's a ledger. So keeping up with the transactions of what's happening. 15:03 in a client's infrastructure. Fantastic. So it's largely a B2B business, yours, right? We do. We have B2B, but we've been approached several times here recently by consumers. Because now, because of the biometric protection aspect of our solution using the quantum encapsulation, we can protect, say, 15:32 Halle Berry from deep fake, being deep faked, or, you know, protecting her likeness from being used without her knowledge in movies, CGI'd into movies. So it's kind of getting a little bit more consumerish as we iterate, right? Yeah, and we were briefly speaking before the podcast recording, Chasity and I, and... 15:59 I've known her for years. She's a very private person, would not allow photographs. So I told my producer, I'm certain Whitney Chastity's not going to be sending us a picture, but you said yes, that you might, because you do have biometric, artificial intelligence, safeguards that can actually discover deep fakes, right? Yes, yes. Yep, if it didn't come from us, if it wasn't checked back from us, 16:29 It wasn't approved by the person. So it's kind of pretty much that simple. Amazing. Well, later on in the show notes, we will have how to contact you at Enfield Tron. So you are in the startup ecosystem. Again, you travel a lot. You're between Washington DC, Atlanta, Los Angeles, and actually the Bay Area. Yeah, the Bay Area. Right. So. 16:58 Revenue can be elusive, right? How? Especially in tech, and especially in these really large markets that I call deep tech. Deep tech and leading edge, bleeding edge, right? People don't know what they're actually buying, right? Or what they don't even, they probably don't even know that they have a need, right? What's been your strategy at Infiltron to keep the revenue flowing while maintaining also a pretty playful, innovative culture? 17:27 You talked about your team and so talk, that's kind of two questions. So how have you kept revenue coming, right? While not going out for dilutive funding yet, but tell us a little bit about how, what's your business model? So the business model in itself is set up for B2B and we also have a licensing element there. So if they, for instance, 17:56 a Fortune 500 company who has a cyber team, right? They have an internal cyber team. If they want to license out the patents that we have and kind of customize it or create or build off of those, use it as a baseline for what they need for their systems, we offer that as well. But let me just put it out there. But back to your question, how do we keep it fun? So the team... 18:25 The original team members, should I say. So we met about seven years ago at a place called the Gathering Spot in Atlanta. So the Gathering Spot is a community and they just opened one in LA and I do go to the one in LA too when I'm there. But it's a community of people, creatives from creative people to deep tech people like myself and everything in between. 18:55 We went to a black tech event at the gathering spot and found ourselves not being able to get into the actual room. So we ended up, because they have a bar and everything at the gathering spot. It's a social club too. It has a club aspect to it too, but you can network there, have meetings there, meet all types of people. I mean known people, I mean it's a great 19:25 great concept, shout out to Ryan. But we found ourselves at the bar, and we're looking at each other. We knew each other because we had been introduced by the Hellbrella person, Tracy. Yes, yes. Because they had done some things for her with a previous startup that she had, development-wise. So we're all sitting at the bar, and we're looking at each other like, but we're the real tech people. 19:55 We do it. It's like we don't really take people. Um, we can't even get in there. We like, we know the organizers and personally and everything. So let's start a company. Well, what we did was we launched, um, what we launched kit labs. And it was literally right down the street from the 20:23 and connect to the community. So we had, it's not far from the AUC and the AUC is where Morris Brown, Morehouse, Spelman and Clark Atlanta are. Got it. So a lot of times you would come in there and find some of the founders, cause this was founded by myself and like six or seven other black tech founders. The ones that were outside. Drinking like, you know. 20:53 That's where we had that conversation. You know, the conversation started at the bar, being outside of that first Black Tech meetup, so to speak, with Joey Womack, who is a part of Goody Nation, who we did get a 50K grant from back in 2020 through Google for Startups. Let me just say this so much. We were so interconnected. I mean, Atlanta is Wakanda. Don't let anybody tell you anything different. 21:21 It's definitely Wakanda. But literally, not even a mile away from the Gathering Spot, we opened up Kit Labs. It's a smart lab where we can tinker with stuff. We're engineers. We're tech people. We need something. We need a makerspace. We don't necessarily need a space that is compared. The Gathering Spot was a little bit more buttoned up. 21:46 And then what we needed, we needed to be able to throw things and make things. We had everything from like 3d printers to, um, VR, AR headsets. I mean, you, anything in tech. Innovative fun. It was in, is in that lab. Um, but that's where around today. So we dissolved it. So it's been dissolved. What one of, one of the founders, he unfortunately transitioned. Um, 22:15 So, you know, and he was kind of like the pillar of it. And it kept going for a while, but it was just a lot of people like myself, it was two female founders, Dr. Nashley Cephas, who herself is from Jackson, Mississippi. I'm shouting out everybody, right? She's from Jackson, Mississippi, and she bought 10 acres in downtown Jackson, Mississippi and started a nonprofit called Bean Pad. And he basically took the concept of what we were doing at Kit Labs and brought it to our hometown. So. 22:44 Um, and it's so funny. She actually founded it on my birthday. So I was like, okay, I can dig that. Um, uh, but, but no, but we're still connected. Everybody still works with each other. You know, if I have to come in and do some things around cyber for a contract or, you know, commercial or whatever client that they have, I do like we, we all kind of still work together on each other's things. So that has allowed you to bring in some revenues, right? 23:14 through its service context. Yeah. Oh, for sure. For sure. Consultant wise, cause they're like, I think people may look at Infotron and think that there's not a human touch piece there, but if you're dealing with me, there's always gonna be a human touch point there because we have to consult the client. We can't assume, you know, we cannot assume. 23:41 what you need, we have to actually have a conversation with our clients throughout the process, even after we possibly have set up the platform for you, trained your people on it, there still needs to be an element of communication, human communication, right? But the team, we've been working together for about seven years. Yes. 24:10 Infiltron has been around for five, going on six years now. So, you know, I mean, respect, mutual respect, we're still kids at heart. I mean, we grew up wanting to be engineers. So, you really can't take the light of innovation out of an engineer unless they're just at the point of not wanting to do it anymore. So we're always, what I've found is most people in any engineering discipline are very, 24:39 curious and forward thinking. So we, and we kind of, we're kind of like a community. We are community and not kind of like, but we are community of folks that contribute to each other's, you know, projects. Yeah. Mm-hmm. And not just, not just business-wise, but personally, like we, I mean, we've been around each other for almost a decade, so. 25:04 there's been kids born and like I just said, one of our founders transitioned, like we've been through some things together that have brought us closer together. And you can, I believe when you have a team like that, and we're all diverse, you know, we have a team like that that cultivates innovation, for sure. You know, I've had a few guests to my podcast and I also write about this, 25:35 Creativity is only possible or it's greatly possible when you create a fun environment and make games out of things and have, right? And set up teams. So I think a shout out to you and what you've set up at Infiltron and in its earlier rendering at Kit Labs, just creating an environment that allows for what ifs, right? Is key. There are a lot of what ifs in cyber. 26:04 I bet you there. So I have a boatload of questions here. One is, before we get into your fundraising path, again, I mentioned earlier you have two patents that have been issued. What is post-quantum encryption technology in layman language? Post. 26:32 Quantum encryption technology. So there is definitely confusion out there that has been addressed. And because there is a difference between post secure quantum and encryption. There's a difference. So. Excellent. 27:02 Post quantum encryption, it is designed to protect data from quantum computers. So. And that's done through the encapsulation? For us, that is how we provide the protection, the encryption. That is the quantum encapsulation is a method of encryption with Involtron. So the current encryption. So you have things like RSA. 27:32 elliptical curve, which elliptical curve is more widely used and kind of being marketed as quantum encryption. It is, it is, it's on the list of quantum protections, right, or quantum methods of encryption protection. So companies like Okta use ECC a lot. But what's happening is that quantum computers are being built now. Yes. Like right now, there's no... Yeah, the cost is going down. 28:02 Yeah, there's no waiting five years from now. Like I urge anyone under the sound of my voice to prepare now for quantum computer attacks. The same thing that I describe happening to us when we were deployed, it's gonna happen. And again, I alluded to feeling like 28:33 situations like Colonial Pipeline and SolarWinds were, I feel like they were tests because there were so many different elements of what we saw in the deployment that happened in those two cases. Yeah, because I'm sitting there and think it's like 2020, 2021, 2019 actually, it started. I think this didn't know, but. 28:59 And it's still going like 20, SolarWinds was still going, the last time I checked SolarWinds was still unraveling. Like it's still, still going. But back to the question. So for us, quantum encapsulation for us is breakthrough. So NIST has had these challenges, right? Where they put out bidding for companies, 29:27 researchers, because a lot of people that are in the quantum space, whether it's physics, mechanics, are generally found in academia. They're not at Infiltron. They're not at QED. They're just not there, right? It's very far in between, and we generally have to lure them. Or we have to do something like partner with them on... 29:53 grants, like the STTR grants. Like that's the only way, generally the only way that we can probably connect with the academia or pierce them and have them work with us. And they usually through that take all the funding, but it's, you're still. Exposed, right? You're exposed, but you're also getting the expertise that you possibly need and can't rightly find in the freelancing world. Yeah. So it generally works out in the long run. 30:23 Um, but so our encapsulation is a, is a breakthrough method because I look at it like this, NIST is holding these challenges and nothing against NIST. We're connected. I contribute to NIST and everything, but they are holding these challenges. And basically they're telling the hackers what people are going to the framework. 30:49 what people are gonna have to adhere to when they create their quantum algorithms to protect their devices and data. You know, you're giving away the secret ingredients. So like, even if they don't know specifically your algorithm, they know what you've based it off of. And that gives it like a tiny thread can unravel a whole t-shirt, right? So I look at it like that. So... 31:15 And even before, you know, we were already developing things before NIST put out these challenges. We are in alignment. We can adhere and do it here to the framework that they're putting out because, you know, you have the DOD space who definitely follows their framework, especially when it comes to the risk management framework. So they're going to follow NIST regardless. They're going to follow their framework, whatever they put out about cybersecurity protection. 31:44 The DOD space and all of its agencies are gonna follow that. However, being in the cybersecurity space every day, seeing what is happening and knowing that you've given some clues, some contextual clues to the malicious hackers about what you're using as a baseline to build your algorithms will, guess what? What we have is not that. Like we are... 32:12 One of the things that differentiates us right now, because I'm sure as quantum cybersecurity continues to grow legs, so to speak, people are gonna start using the more, less susceptible to hacks by quantum computers method. So you have things like multivariate hash code. So these are some of the 32:40 quantum properties that you can use that are not generally hackable by a quantum computer. They won't be hackable by a quantum computer. So we leveraged some of that. It was like, if I'm built, I looked at it like this, I've been in cyber, I've been in tech for almost 20 years. I know I don't look it. I get it all the time. You don't have to say it. I've been in tech for almost 20 years. I've been, and when I was in the air force, we call it InfoSec. It's the same thing. And that dates me. 33:08 If I say, if you hear somebody say InfoSec, trust me, they've been in cybersecurity for at least 20 plus years. So, but it's cybersecurity, that's what it is. And I've seen the changes and I've paid my dues too. Like I didn't, when I got out of the Air Force, I was just, side note, like I cut grass and loved it. I would go back and do it if I can make these results. So then like, it's very, it's very fulfilling. Don't let anybody fool you. Like I love, but I like being outside, but. 33:38 Um, my first tech job though, I literally went through the phone book. Cause this is like still, you know, internet was not quite what it is now, of course, but it was like still growing. And I went through the yellow pages and went through the aerospace companies and called all of them and was like, Hey, let's just get out of the air force, look for a job. I don't care if it's an intern or co-op and L3 L3 before they merged with Harris. Uh, 34:08 they created me a co-op. And, but again, still in touch with, cause you know, L3 is a huge government contracting company, right? And in the satellite communication space, cause they're in line with my background. And so I've seen it all. I've seen the changes of InfoSec into cybersecurity. And now we're entering a new frontier with quantum cybersecurity. So I've been here, 34:37 maybe at the latter part of the info set, but definitely through the cybersecurity and here for and to forge some guidelines and pathways in the quantum cybersecurity space with Inflotron. So when you know Inflotron was founded in 2019, I was like, okay, if I'm gonna start 35:03 something new in cyber and we hadn't even gotten to the quantum piece yet. They hadn't even gotten to me yet. Like it started like I was getting downloads. Yeah. Because I'm, I always, I'm a reader. I wake up looking at cyber news and just staying in the know because I need to know what's going on so I can protect my clients, whether that was me in a government contracting position or me as a consultant in my businesses. So. 35:33 I need to know what's going on. And if I'm going to build something new, why am I going to build it with compromised parts? Right. That's a great way to describe it. Yeah. Forget the tech. It didn't make logical sense. If I'm going to build something new, a SaaS product that's going to integrate and be flexible and adaptable and proactive. 36:01 Why would I use RSA encryption when I know what's coming? Got it. That will be one of the snippets that I share in my YouTube channel as well as the podcast. That is excellent. Why build something with compromised parts? Frontier technology, quantum cybersecurity is what Epfiltron is about. 36:30 Next generation. Talk to me a little bit more for us, less tech savvy listeners about the use of Infiltron in a SelleGov's program for smart cities. That kind of brings it more home and more tangible. How is technology used for smart cities? So first, SelleGov through leading cities. Yes. 36:59 It connects companies like ours with municipalities to tackle urban challenges. So for us, it's infrastructure, security, and sustainability. So we were a finalist in leading cities global competition back in 2021. And we've worked through them. You know, we've been able to work with city leaders to secure IOT systems and critical infrastructure. 37:28 And quick shout out to Michael Lake. Okay. He's the founder of Leading Cities, amazing guy. Another keep in touch, answer the email quickly person. He's based in Boston, but he's built a very supportive ecosystem. So shout out to Michael Lake. But as a part of this program, 37:56 We're offering smart cities our enhanced quantum vulnerability assessment. And this is to help the smart city leaders identify areas that need better quantum protections now. We've just had a session on November the 11th, Veterans Day. And the second one is coming up December the 5th. So you. 38:24 If you're a smart city leader or see so small, medium, large enterprise, no matter what market you in, you're in, definitely tap in. You can register for it on the leading city's website or on our website at Infotron.net. Yeah, that's on December 9, 2024 at 1pm. Is that Eastern? December 5th. December 5th? No, it's the 9th, because I have it here. And that's my cousin's birthday. So yeah, it's December 9th. 38:53 Did you get to influence those dates? Yeah. So let's jump into your startup. You've taken in very little dilutive funding. How much money have you raised to date? And how have you, what is the next phase, right? In terms of outreach for fundraising. So we've raised 120K and that was through Techstars, LA Space. 39:23 Still counting. I do not take a salary. I could take one, but I'm just, it's the long game for me. And I still consult. Don't let these people tell you not to quit your job and be an entrepreneur. Don't let people do that. Especially if you have a family. Don't let these people, don't let these people try to guilt you or shame you because you still have a job while you're building your startup. Don't let, don't do it. 39:53 Because I do have a company that I started called Right Tech Solutions and we still, that's why I said I still feel like I'm in the Air Force because I still consult them. So I can, you know, the revenue that we do and we've hit 500K in revenue. So you know, I could easily take a salary, right? But I just, it's the long game for me. It's the global expansion. 40:22 um, you know, more IP and patents, uh, protections, right? Because we do have global count clients. And, um, one of the things that I wanted to make sure of before we even took on the clients was that we had legal backing there. So IP trademarks, um, at least patent, at least the application is pending, but you know, like I want to, I want to, I want it to at least have that. And we have great attorneys. Um, shout out to Malika Tyson. 40:52 and Matthew and Dorian who have, they took over because I had a, I had an attorney, IP attorney that would, had her own boutique firm and then she had to go back, you know, she just couldn't do the entrepreneurship, it's not for everybody, but we still stay in contact as well. But she introduced me to McAndrews, they're based out of Chicago. 41:20 And they are the legal team for Impletron. I always tell them that when we're on calls, like you are the legal team. Like, yeah, anything that I need from them legal, legal wise, they do it. I literally just sent a partnership NDA over to Malekka this morning and she just sent it back to me. So like, that's not IP and trademark, right? But they do, they do it. And I always tell them how much I appreciate them because... 41:49 IP and trademarks are not free and they're not inexpensive. So, and then imagine, you know, we have one pending now in Japan. We just got one in Canada. So yeah, like it's expensive, you know, it's expensive. So a lot of the funding that we get now is going to be allocated to pay them, you know, even though they work with us. But it's going to be paying them. 42:18 doing some iterations, we have a partnership where there's some hardware that's gonna be involved. We're definitely tapping into the hardware. So we'll be forging our way there because people like things they can touch. SaaS isn't necessarily something that you can touch, although put it into a platform makes it a little bit more tangible for people, visual at least. So in the- 42:48 Yeah, I mean, hardware has always been a part of the vision. FBGAs, we have another colleague of mine, he has developed a cryptocurrency mining machine, and it leverages quantum. So it's mining at exponential speeds, right? Because generally what quantum does is speeds things up. It speeds exactly, in simplified terms. 43:18 Definitely still going after Sivers traditional government contracts globally. We participated in Fintech down in the Bahamas last October. Cause we are in the Fintech space and there's a lot of similarities between Fintech and Space Tech. Because when you're talking about fault zeros and being able to detect anomalies. 43:46 both of those markets need that and they need it quick. So we've been able to, yeah, like we've been able to leverage some of the things that we're learning in both of those for each other. So we've been able to participate in some conferences. We actually getting ready to go to Barbados in January for Fintech Islands, I'll be speaking about 44:14 the kind of the intersection of the quantum age and what's coming in respect to the fintech space, cryptocurrency, web three, traditional finance and AI, because we do leverage AI. And we've been in the AI space, Impletron has been in the AI space from the beginning. One of our advisors is an AI evangelist at AWS. I did say her name earlier on this podcast, but. 44:42 She's amazing. She's a Georgia Tech grad. We do have a few Georgia Tech people on the team, but she's amazing. And I'm able to tap her. I've been able to tap her because she was one of the Kit founders. So I've been able to tap her about AI and machine learning very early on. So all of the LLMs and the SLMs that everybody's kind of talking about, we've been doing. 45:11 Like even as small as we are, we've been. 45:16 Yeah, so, Chasity, how can my listeners contact or get information about Infotron? So, yeah, of course the website. So, infiltron.net. You can follow us on all of our socials at Infotron Software Suite. It might be, I think on Twitter is Infotron app. We wanted to keep it short. 45:41 And then, or you can email us at mfultronapp at gmail.com. And I know people are gonna be like, why you use Gmail? That's another filter. And that's an email that everybody on the team can look at and not be bombarded with, cause spam and it's just, everybody has their own email address, but. So you probably, it's a test environment for all of you. 46:09 beautiful quantum encryption that you're working on. Yes. And that's it all. One better way to start. Yeah, Gmail, right? Google knows a lot more about us than we'd like them to. Oh, Google knows everything. That's tough. Even when you turn location off. Oh, Instagram. I just posted something about Instagram. So Instagram's new. They just updated their policy maybe a month ago, maybe. 46:38 Okay. Whether you want to or not, they now have access to your photos, your GPS location, everything even if you say no, even if you turn it off, they still contract. 47:00 Just putting it out there guys. Yeah. So if you do platform. So there's cause to the platform. Right? Yes. Thank you. All right. We're coming down to the section of the podcast where I like to ask each of my guests what the following three words mean to you. Because this is what I do with my consulting business. 47:24 In addition to my podcast, I work with founders that are really building resilient, purpose-driven and scalable businesses. What's resilience mean to you, Chasity? Man, that's a word that I use. Uh, I mean, I'm, I mean, you gotta think about it. I'm black trying to raise money. It's hard for black people to raise money on top of that. I've been, you know, um, I've come face to face with people that didn't believe that I wrote my own patents. Like. 47:53 you know, as if black people didn't invent a lot of things, like that we still use today. Like, come on. I mean, it's just the truth. Resilience. Resilience for me is bending, but never breaking. Bending, but never breaking. Yeah. It's about, you know, adapting to challenges. I just mentioned some and facing them. Like you can't, you can't, and I'm about to sound 48:23 run from the pain, you gotta run towards it. So you can come out stronger on the other side. And it's not necessarily about survival, it's transformation. That's transformation. It's transformation. And that transformation is preparing you for what's next. And you'll be standing taller than you were before. Amazing, thank you. Purpose-driven, what's a purpose-driven? 48:53 Enterprises or? Yeah. I'm a visionary. So like, there's a lot of founders that I've met. If I have the opportunity to get close to them or kind of hear them speak about what they're building to include myself, because I do talk to myself about the things that I'm building. I counsel myself. I'm sure my ancestors are around me. 49:23 Purpose is, it should be intentional. I think that it's kind of interchangeable for me. But in the context of the question that you asked on purpose driven enterprise, so it's the heartbeat in what we build here at Infotron. I can definitely say that. It's creating meaningful solutions that solve real problems. And in solving those real problems, 49:52 you're still staying true to the mission. I still bring the aspect of the military into Infiltron. We are mission focused. We have fun. We do all the fun things, right? Because again, that cultivates innovation too. And it keeps it spicy. You need to let things be spicy because in a regular deglar cybersecurity job, you're probably bored. Like. 50:19 I mean, let's just be real. Like you're probably bored. You're probably looking at Excel spreadsheets and creating a report by hand from that. Like it's boring. Like, but you know, it's also making moves that matter. And it's solving problems that for me leave a legacy and just never losing sight of why we started in the first place. 50:48 So never lives in sight. Excellent. What about scalable? So how does- That's one of those BC's favorite words. That's right. Because that's what they want to see. How will you scale? That's right. I mean, I'm an investor too guys. Don't get it twisted. Like, I think that was a question that I did ask with one of the investors I had. Like, how are you going to get over that challenge? Like, before I give you this money. 51:18 Scalable. So growth, like we can think about growth in so many different ways, like growth, personal growth, because if you embark on the entrepreneur trick, you are going to be, and need to be open to growth. To me, entrepreneurship is a spiritual journey. Beautiful. 51:45 about the Southern Baptist roots, but I'm not spiritual. I'm a yoga, meditating, put my feet in the sand, grass grounding person nowadays, but still bringing that element of praying. And it's all the same to me. They just changed the name of God, right? Just that's my perspective, but growth isn't just about getting. 52:14 bigger. It's about getting better. And me speaking about the personal aspect, that is what growth is. It might not feel good, you know, while it's happening. But, you know, once you get through it and you can get in a reflective mindset and look back with what you just came through and be grateful, like find gratitude in it, you know. 52:43 That's how I look at growth. It's expanding mindfully and staying grounded in your values and making sure that every step that you take going forward strengthens the foundation that you've already built. And it's... 53:11 Like I said, it's moving with intention. And while you're moving with intention, you're also preserving the quality and the vision that define you. Which goes back to purpose-driven. Yes, thank you. Last question, Chasity. Did you have fun in the sandbox? Oh yeah, I mean, it's you. You know, we already have a great rapport. 53:38 I'll say this, one of my favorite memories of you is when you brought Ty to the table to kind of see if they were, could invest in Infiltron and it was too early. But we had to sign an NDA, it was some type of contract, but it was during Mercury retrograde. You said it before I said it, I was like, I wonder if she's onto this type. 54:05 Cause I wasn't going to sign it. I was going to try to delay it as much as possible, but you're like, no, let's wait, let's wait. So after Mercer, that's your great. Well, I was like, oh, these are this. She's my people. And I was like, and I think I responded like, let's wait five days. So it is no, it's like clear. So, um, that's a little fighter for me with you. Oh, I love it. I love it. Generally hear that in business. No, no. 54:32 And the Founder Sandbox again is a pretty eclectic podcast, bringing in deep tech founders like Chasity Wright that are on the frontier, bringing in what the future, will, it's the future's here. It's here. That's right. So to my listeners, if you like this episode with Chasity Wright, CEO and founder of Infiltron, sign up for the monthly release of 55:01 this podcast where founders, business owners, corporate directors, and professional service providers share their own experiences on building with strong governance, a resilient, scalable, and purpose-driven company to make profits for good. So signing off for this month, thank you, Chasity. Thank you, Brenda, so much. I hope to see you soon.  

Today, we're joined by Chris Lott, senior director of engineering at Qualcomm AI Research to discuss accelerating large language model inference. We explore the challenges presented by the LLM encoding and decoding (aka generation) and how these interact with various hardware constraints such as FLOPS, memory footprint and memory bandwidth to limit key inference metrics such as time-to-first-token, tokens per second, and tokens per joule. We then dig into a variety of techniques that can be used to accelerate inference such as KV compression, quantization, pruning, speculative decoding, and leveraging small language models (SLMs). We also discuss future directions for enabling on-device agentic experiences such as parallel generation and software tools like Qualcomm AI Orchestrator. The complete show notes for this episode can be found at https://twimlai.com/go/717.

Proprietary LLM builders need to experience a valuation haircut as open source LLMs take share from proprietary LLMs. Proprietary LLM builders (OpenAI, Anthropic, Google, Microsoft, Amazon), have enjoyed lofty valuations over the past several years. Given the rise of open source competitors - which are on par with proprietary models from a performance standpoint and can be operated at a fraction of the cost - the proprietary model builders should suffer a valuation haircut. I believe that open source LLM builders such as DeepSeek and META will win the day and that 80% of LLMs and SLMs in production 5 years from now will be open source language models. https://open.substack.com/pub/tek2day/p/valuation-haircut-is-due-for-proprietary?r=1rp1p&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false

By now you probably know the term “large language model.” They’re the systems that underlie artificial intelligence chatbots like ChatGPT. They’re called “large” because typically the more data you feed into them — like all the text on the internet — the better those models perform. But in recent months, there’s been chatter about the prospect that ever bigger models might not deliver transformative performance gains. Enter small language models. MIT Technology Review recently listed the systems as a breakthrough technology to watch in 2025. Marketplace’s Meghan McCarty Carino spoke to MIT Tech Review Executive Editor Niall Firth about why SLMs made the list.

By now you probably know the term “large language model.” They’re the systems that underlie artificial intelligence chatbots like ChatGPT. They’re called “large” because typically the more data you feed into them — like all the text on the internet — the better those models perform. But in recent months, there’s been chatter about the prospect that ever bigger models might not deliver transformative performance gains. Enter small language models. MIT Technology Review recently listed the systems as a breakthrough technology to watch in 2025. Marketplace’s Meghan McCarty Carino spoke to MIT Tech Review Executive Editor Niall Firth about why SLMs made the list.

By now you probably know the term “large language model.” They’re the systems that underlie artificial intelligence chatbots like ChatGPT. They’re called “large” because typically the more data you feed into them — like all the text on the internet — the better those models perform. But in recent months, there’s been chatter about the prospect that ever bigger models might not deliver transformative performance gains. Enter small language models. MIT Technology Review recently listed the systems as a breakthrough technology to watch in 2025. Marketplace’s Meghan McCarty Carino spoke to MIT Tech Review Executive Editor Niall Firth about why SLMs made the list.

Cale and Sujit discuss their current projects in Azure as 2024 comes to a close. They also cover a ton of AKS updates. Semantic Kernel makes it easier for developers to build Azure Open AI applications that can also include SLMs like Phi-4.  Azure has many options to use File Shares and Volumes, and we walk through the process of figuring out which one is right for your needs.   Media file: https://azpodcast.blob.core.windows.net/episodes/Episode511.mp3 YouTube: https://youtu.be/dLfCJ6btKng Resources: Semantic Kernel - https://github.com/microsoft/semantic-kernel Journey with SK on OpenAI and AzureOpenAI Ollama (running SLM local) - https://github.com/ollama/ollama Ollamazure (running SLM that looks like Azure OpenAI) - https://github.com/sinedied/ollamazure PhiSilica - https://learn.microsoft.com/en-us/windows/ai/apis/phi-silica   File Shares: https://learn.microsoft.com/en-us/azure/storage/common/storage-introduction    Other updates: Lots of AKS updates! https://learn.microsoft.com/en-us/azure/aks/concepts-network-isolated https://learn.microsoft.com/en-us/troubleshoot/azure/azure-kubernetes/availability-performance/container-image-pull-performance https://learn.microsoft.com/en-us/azure/aks/imds-restriction https://learn.microsoft.com/en-us/azure/aks/use-windows-gpu https://azure.microsoft.com/en-us/updates/?id=471295 https://learn.microsoft.com/en-us/azure/backup/tutorial-restore-aks-backups-across-regions https://learn.microsoft.com/en-us/azure/aks/app-routing-nginx-configuration?tabs=azurecli#control-the-default-nginx-ingress-controller-configuration-preview https://learn.microsoft.com/en-us/azure/aks/automated-deployments https://learn.microsoft.com/en-us/azure/aks/aks-extension-ghcopilot-plugins https://learn.microsoft.com/en-us/azure/azure-monitor/containers/container-insights-logs-schema#kubernetes-metadata-and-logs-filtering

E143: David updates listeners on Venture for the week, then interviews Josh Miller of Gradient Health about his take on AI from the perspective of data. We discuss his data experience in the context of AI model training and validation, and address some timely questions relevant to the current Era of AI: 1) Is Data where the money is (in addition to GPUs) and is there really future profit in AI models? 2) Data quality, structure and annotations for metadata, 3) Are LLMs being commoditized and what is the future of SLMs? 4) What is the future of AI Assistants and Agents? 5) What is the future of Synthetic Data? 6) How will AI affect the Arts? 7) What is the future for AI investment? (recorded 11/1/24)BIP 2024 State of Startups in the Southeast reportFollow David on LinkedIn or reach out to David on Twitter/X @DGRollingSouth for comments. Follow Paul on LinkedIn or reach out to Paul on Twitter/X @PalmettoAngel We invite your feedback and suggestions at www.ventureinthesouth.com or email david@ventureinthesouth.com. Learn more about RollingSouth at rollingsouth.vc or email david@rollingsouth.vc.

With a $4.5B valuation, 5M AI builders and 1M public AI models, Hugging Face has emerged as the key collaboration platform for AI, and the heart of the global open source AI community. In this episode of The MAD Podcast, we sit down with Clément Delangue, its co-founder and CEO, and delve deep into Hugging Face's journey from a fun chatbot to a central hub for AI innovation, the impact of open-source AI and the importance of community-driven development, and discuss the shift from text to other AI modalities like audio, video, chemistry, and biology. We also cover the evolution of Hugging Face's business model, and the different approach to company culture that the founders have implemented over the years. Hugging Face Website - https://huggingface.co Twitter - https://x.com/huggingface Clem Delangue LinkedIn - https://www.linkedin.com/in/clementdelangue Twitter - https://x.com/clemdelangue FIRSTMARK Website - https://firstmark.com Twitter - https://twitter.com/FirstMarkCap Matt Turck (Managing Director) LinkedIn - https://www.linkedin.com/in/turck/ Twitter - https://twitter.com/mattturck (00:00) Intro (01:46) Miami vs. New York vs. San Francisco (03:25) Current state of open source AI (11:12) Government regulation of AI (13:18) What is open source AI? (15:21) Open source AI: China vs U.S. (18:32) LLMs vs. SLMs (22:01) Are commercial LLMs just 'Training Wheels' for enterprises? (24:26) Software 2.0: built with AI (28:03) Hugging Face founding story (37:03) Are there any competitors? (44:06) Most interesting models on Hugging Face (50:35) Shifting focus in enterprise solutions (55:06) Bloom & Idefix (58:44) The culture of Hugging Face (01:04:44) The future of Hugging Face

Send us a textMore on GenAI, Hallucinations, RAG, Use Cases, LLMs, SLMs and costs with Armand Ruiz, Director watsonx Client Engineering and John Webb, Principal Client Engineering.  With this and the previous episode you'll be wiser on AI than 98% of the world.00:12 Hallucinations02:33 RAG Differentiation06:41 Why IBM in AI09:23 Use Cases11:02 The GenAI Resume13:37 watson.x 15:40 LLMs17:51 Experience Counts20:03 AI that Surprises23:46 AI Skills26:47 Switching LLMs27:13 The Cost and SLMs28:21 Prompt Engineering29:16 For FunLinkedIn: linkedin.com/in/armand-ruiz, linkedin.com/in/john-webb-686136127 Website: https://www.ibm.com/client-engineeringLove what you're hearing? Don't forget to rate us on your favorite platform!Want to be featured as a guest on Making Data Simple?  Reach out to us at almartintalksdata@gmail.com and tell us why you should be next.  The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun. Want to be featured as a guest on Making Data Simple? Reach out to us at almartintalksdata@gmail.com and tell us why you should be next. The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun.Want to be featured as a guest on Making Data Simple? Reach out to us at almartintalksdata@gmail.com and tell us why you should be next. The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun.

Send us a textMore on GenAI, Hallucinations, RAG, Use Cases, LLMs, SLMs and costs with Armand Ruiz, Director watsonx Client Engineering and John Webb, Principal Client Engineering.  With this and the previous episode you'll be wiser on AI than 98% of the world.00:12 Hallucinations02:33 RAG Differentiation06:41 Why IBM in AI09:23 Use Cases11:02 The GenAI Resume13:37 watson.x 15:40 LLMs17:51 Experience Counts20:03 AI that Surprises23:46 AI Skills26:47 Switching LLMs27:13 The Cost and SLMs28:21 Prompt Engineering29:16 For FunLinkedIn: linkedin.com/in/armand-ruiz, linkedin.com/in/john-webb-686136127 Website: https://www.ibm.com/client-engineeringLove what you're hearing? Don't forget to rate us on your favorite platform!Want to be featured as a guest on Making Data Simple?  Reach out to us at almartintalksdata@gmail.com and tell us why you should be next.  The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun. Want to be featured as a guest on Making Data Simple? Reach out to us at almartintalksdata@gmail.com and tell us why you should be next. The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun.Want to be featured as a guest on Making Data Simple? Reach out to us at almartintalksdata@gmail.com and tell us why you should be next. The Making Data Simple Podcast is hosted by Al Martin, WW VP Technical Sales, IBM, where we explore trending technologies, business innovation, and leadership ... while keeping it simple & fun.

In this episode, we explore the impact of Artificial Intelligence (AI) on project management and how it's transforming the profession. Oliver Yarbrough shares how project managers can leverage AI to enhance their skills and stay competitive in an evolving, AI-driven landscape. Hear about AI's impact on Agile teams, how SLMs and LLMs are revolutionizing data refinement, how to balance data security while leveraging AI, and how to treat AI as a key stakeholder in the evolving landscape of project management. Chapters 02:21 … Meet Oliver04:21 … What is AI?05:39 … Will AI Replace the PM?06:41 … Incorporating AI Tools08:45 … Finding the AI Capabilities09:44 … Skills and Knowledge Areas13:02 … AI and Data Analysis Challenges15:23 … ChatGPT and Data16:40 … Human in the Loop18:08 … Protecting Your Data20:14 … Contractors Using AI22:20 … Kevin and Kyle23:18 … Impact on Agile Team Performance26:22 … Fine Tuning and Refining29:05 … A Large Language Model (LLM)30:08 … Current Trends in AI32:21 … AI Component to PM Tools33:40 … Streamlining Workflow with AI39:36 … Future Evolutions of AI43:17 … Contact Oliver44:58 … Closing OLIVER YARBROUGH:  AI acts as a stakeholder on our projects, and we should treat it like we treat any other stakeholder.  That's very important.  Initially, I used to say treat it like it's a piece of software tool.  But with the new advents of these AI agents and AI assistants and all these other things, you really do need to treat it like a true stakeholder. WENDY GROUNDS:  Welcome to another episode of Manage This. where we dive deep into the latest trends, insights, and strategies in project management.  This is the podcast by project managers for project managers.  I'm Wendy Grounds, and with me in the studio is Bill Yates.  And in the studio today we have an incredible guest who is sure to enlighten, inspire, and, I think for me, educate a lot. Joining us is Oliver Yarbrough.  He's a PMP, a renowned author, speaker, and trainer with a knack for combining hands-on real-world experience with project management fundamentals.  His impressive career includes positions with Fortune 500 companies like Lucent Technologies, Staples, and Sprint, as well as successful business ventures of his own. Currently, Oliver is an active member of PMI, where he shares his extensive knowledge on adapting to AI, deriving value from data, and recognizing AI trends in project management.  He has some LinkedIn courses which we've taken a look at.  They cover everything from leveraging AI in project management to the importance of human strengths in an AI-driven world. BILL YATES:  As Wendy is saying, we're going to explore a topic that, I mean, if you connect to the news, if you connect to the Internet in any way, probably one of the top trending topics is AI, or artificial intelligence.  We're going to dive deep into that with Oliver, and we're going to look at it from a specific lens, and that is from the perspective of the project manager.  How does this impact me?  How does this impact my job and my future?  Oliver is going to share some insights with us.  He'll help us understand how to stay relevant.  What are some strengths from AI that we can harness?  What are some things that we cannot fear?  So, Oliver is here to open our minds. WENDY GROUNDS:  Yup, so get ready as we welcome Oliver.  Hi, Oliver.  Welcome to Manage This. OLIVER YARBROUGH:  Yes, great.  Glad to be here. Meet Oliver WENDY GROUNDS:  Oliver, why don't you tell us how you got into artificial intelligence?  How you took that path? OLIVER YARBROUGH:  Well, I sort of fell into it backwards.  So, I did not start off as an AI person.  I've always been a Curious George.  I've always been poking my head in, like, “What's going on here?  What's going on there?”  But, you know, I have a project management background.  So, I was doing PMP exam prep.  I was training people.  And that's how I got in touch like with you guys.

In this episode, Doug shares his preference for composition over inheritance in object-oriented programming and his strategic use of design patterns in Visual Basic consulting. He challenges commonly held beliefs about language performance and productivity, arguing that faster languages do not always yield faster results. The conversation also explores the offline benefits of large and small language models (LLMs and SLMs) and highlights Doug's innovative use of PowerShell to create autonomous agents.Doug shares fascinating insights on prompt engineering, the evolution of AI models, and the potential of personal AI as the next technological inflection point. Despite facing resistance from critics and the tech community, Doug remains a staunch advocate for leveraging cutting-edge tools and maintaining an unscripted, adaptive approach to technology.Show NotesTime Stamps07:32 Discovering GPT chat and its incredible capabilities.14:25 Expect announcement at OpenAI Day in November.19:41 Initial confusion, but eventually realized cross-platform potential.25:31 Colleague makes fun of me, but impressed.30:07 Experience of being a non-traditional engineer.35:10 Prefer using PowerShell over Python for coding.40:36 Discussing hallucination problem and algorithms in AI.43:19 Using AI to generate better function names.47:46 "Creating forms quickly with impressive results."54:58 Recall story of new guy at whiskey distillery.01:01:06 Microsoft focuses on smaller, efficient language models.01:08:12 Data-driven podcast explores PowerShell and AI's fusion.LinksPSAI (PowerShell AI module): https://github.com/dfinke/PSAI PSWeave (PowerShell module bringing OpenAI's GPT): https://github.com/dfinke/PSWeaveImportExcel (PowerShell module to import/export Excel spreadsheets): https://github.com/dfinke/ImportExcelNew York PowerShell Meetup: Meetup https://www.meetup.com/nycpowershellmeetup/

Dave Sobel welcomes Anurag Agarwal from TechIsle to discuss research data on customer expectations, cybersecurity, and AI trends. The conversation delves into the increasing budget allocations for IT security among SMBs, highlighting the importance of consulting services in guiding SMBs towards choosing the right security technologies and frameworks. Anurag emphasizes the shift towards cybersecurity as a business risk, prompting SMBs to focus on cyber resiliency rather than just cybersecurity. The discussion then transitions to the evolving landscape of AI adoption, with a focus on large language models (LLMs) and their impact on businesses. Anurag and Dave explore the significance of AI consulting in preparing organizations for AI implementation, emphasizing the need for expertise in selecting the right language models for specific business problems. The conversation underscores the importance of aligning with a single hyperscaler for AI solutions and the emerging trend towards small language models (SLMs) for accelerated AI adoption. Anurag highlights the growing trend of AI assessment as a key focus for both partners and end customers, surpassing cloud cost optimization and security assessments. Supported by: https://salesbuildr.com/   All our Sponsors: https://businessof.tech/sponsors/ Do you want the show on your podcast app or the written versions of the stories? Subscribe to the Business of Tech: https://www.businessof.tech/subscribe/Looking for a link from the stories? The entire script of the show, with links to articles, are posted in each story on https://www.businessof.tech/ Support the show on Patreon: https://patreon.com/mspradio/ Want our stuff? Cool Merch? Wear “Why Do We Care?” - Visit https://mspradio.myspreadshop.com Follow us on:LinkedIn: https://www.linkedin.com/company/28908079/YouTube: https://youtube.com/mspradio/Facebook: https://www.facebook.com/mspradionews/Instagram: https://www.instagram.com/mspradio/TikTok: https://www.tiktok.com/@businessoftechBluesky: https://bsky.app/profile/businessoftech.bsky.social

Windows 11 A quick look back at the Copilot+ PC launch Paul did end up making a video of each MacBook Air reference during last week's launch The Copilot+ PC event was also notable for its many references to the past: Paul counted at least 7 in the first 7 minutes alone After some soul-searching, Paul went all-in on Surface Laptop by canceling the first preorder and bulking it up with more RAM and storage Release Preview (last week): 24H2 comes to RP, suggesting it hits stable on Tuesday, June 11 (in preview), one week before the first Copilot+ PCs arrive with this build preinstalled Beta (last week): 23H2 build with no new features (arriving after 24H2 in Release Preview, of course) Microsoft Edge is getting more responsive thanks to a "WebUI 2.0" initiative Lenovo earnings up 9 percent in quarter, down 8 percent for the year. The future? AI PCs, duh AI There was a massive Bing outage late last week. Press hold on all the obvious jokes, this impacted Copilot everywhere, DuckDuckGo, other services Cue the Chicken Little "this is why I'll never use AI" crowd, which used to be the "this is why I'll never use the cloud" crowd Copilot comes to Telegram for some reason Google goes live with AI overviews in Search and the world grinds to a halt Google adds Gemini to Chromebook Plus, gives one free year of Google One AI Premium to new device buyers WWDC is coming soon and we're seeing hints of how Apple will add more AI to its products Opera partners with Google Cloud on Gemini in Aria AI and image understanding/image generation in browser And it now plans to make 2000 SLMs available in the browser NVIDIA revenues were a record in Q1, up 262 percent YOY Dev Last week, we learned that WPF is making a comeback. This week, Paul revived his .NETpad project and is going to modernize it using the new features Xbox We finally have some news about Activision Blizzard games and Game Pass!!! But it's only one game, and it doesn't happen until November Atari buys the Intellivision brand. I wish we knew more about this Tips & Picks Tip of the week: Beware the dark patterns App pick of the week: Windows 11 version 24H2 RunAs Radio this week: PowerApp Extensibility with Christina Wheeler Brown liquor pick of the week: Glen Garioch 1797 Founder's Reserve Hosts: Leo Laporte, Paul Thurrott, and Richard Campbell Download or subscribe to this show at https://twit.tv/shows/windows-weekly Get episodes ad-free with Club TWiT at https://twit.tv/clubtwit Check out Paul's blog at thurrott.com The Windows Weekly theme music is courtesy of Carl Franklin. Sponsors: HP.com - WW - https://bit.ly/4adilko GO.ACILEARNING.COM/TWIT - code TWIT30

Bing outage, NVIDIA's record Q1 FY25, WPF's rebirth Windows 11 A quick look back at the Copilot+ PC launch Paul did end up making a video of each MacBook Air reference during last week's launch The Copilot+ PC event was also notable for its many references to the past: Paul counted at least 7 in the first 7 minutes alone After some soul-searching, Paul went all-in on Surface Laptop by canceling the first preorder and bulking it up with more RAM and storage Release Preview (last week): 24H2 comes to RP, suggesting it hits stable on Tuesday, June 11 (in preview), one week before the first Copilot+ PCs arrive with this build preinstalled Beta (last week): 23H2 build with no new features (arriving after 24H2 in Release Preview, of course) Microsoft Edge is getting more responsive thanks to a "WebUI 2.0" initiative Lenovo earnings up 9 percent in quarter, down 8 percent for the year. The future? AI PCs, duh AI There was a massive Bing outage late last week. Press hold on all the obvious jokes, this impacted Copilot everywhere, DuckDuckGo, other services Cue the Chicken Little "this is why I'll never use AI" crowd, which used to be the "this is why I'll never use the cloud" crowd Copilot comes to Telegram for some reason Google goes live with AI overviews in Search and the world grinds to a halt Google adds Gemini to Chromebook Plus, gives one free year of Google One AI Premium to new device buyers WWDC is coming soon and we're seeing hints of how Apple will add more AI to its products Opera partners with Google Cloud on Gemini in Aria AI and image understanding/image generation in browser And it now plans to make 2000 SLMs available in the browser NVIDIA revenues were a record in Q1, up 262 percent YOY Dev Last week, we learned that WPF is making a comeback. This week, Paul revived his .NETpad project and is going to modernize it using the new features Xbox We finally have some news about Activision Blizzard games and Game Pass!!! But it's only one game, and it doesn't happen until November Atari buys the Intellivision brand. I wish we knew more about this Tips & Picks Tip of the week: Beware the dark patterns App pick of the week: Windows 11 version 24H2 RunAs Radio this week: PowerApp Extensibility with Christina Wheeler Brown liquor pick of the week: Glen Garioch 1797 Founder's Reserve Hosts: Leo Laporte, Paul Thurrott, and Richard Campbell Download or subscribe to this show at https://twit.tv/shows/windows-weekly Get episodes ad-free with Club TWiT at https://twit.tv/clubtwit Check out Paul's blog at thurrott.com The Windows Weekly theme music is courtesy of Carl Franklin. Sponsors: HP.com - WW - https://bit.ly/4adilko GO.ACILEARNING.COM/TWIT - code TWIT30