Skip to content
The Information Bottleneck cover art

Ravid Shwartz-Ziv & Allen Roush

The Information Bottleneck

A Technology and Science podcast

Good podcast? Give it some love!

About the show

Two AI Researchers - Ravid Shwartz Ziv, and Allen Roush, discuss the latest trends, news, and research within Generative AI, LLMs, GPUs, and Cloud Systems.

Newest episodes

Post-Training Is Where Models Are Made Now | Maxime Labonne

1 hr 3 min

In this episode, we talked with Maxime Labonne, Head of Post-Training at Liquid AI and author of the LLM Course on GitHub, about what happens to a model after pre-training and why that stage now matters so much. Maxime walks us through the short history of post-training, from supervised fine-tuning and preference optimization to reinforcement learning with verifiable rewards and now multi-domain on-policy distillation (MOPD), where separate teachers for code, math and other domains get merged back into one student. He explains why you still can’t skip SFT, why the order of the stages matters, and why RL works even though a long rollout gives you about one bit of signal. His intuition is that SFT and distillation raise pass@k, and RL then sharpens it into pass@1. We also get into training small models for agentic tasks directly inside the harness, which open-source RL stacks are good enough for serious work, looping transformers, and how much of the work big models do today could move to small ones running locally. We also talk about why models write worse than they used and safety issues. We talk about regulation, open source and Chinese models,  whether a small startup can beat the big labs in a niche, and why he thinks RL environment companies are overrated. We finish with France vs San Francisco, Mistral and the regulation moat, and whether Europe can produce frontier models. Topics Why post-training matters more every yearThe modern pipeline: SFT, RL and distillationTraining agents and small modelsWhy models write worse than they used toSafety, incentives and regulationOpen source and Chinese modelsStartups vs the big labsEurope, Mistral and sovereign AIChapters 0:00 Intro 0:57 Why post-training matters 3:44 A short history of post-training 7:20 The modern pipeline: SFT, RL and distillation 19:41 Agents, architectures and small models 27:15 Why models write worse now 31:44 Safety, incentives and regulation 38:26 Open source and Chinese models 41:39 Recursive sel...

Physical AI for Science: Why It’s Harder Than Coding | Rishabh Agarwal (Periodic Labs)

1 hr 1 min

Rishabh Agarwal has worked on reinforcement learning for years, at Mila, Google DeepMind and Meta. He is now a founding member at Periodic Labs, a company building AI scientists together with the physical labs where they run experiments, and an adjunct professor at McGill. By his own count he understands about ten percent of the science they do there. In this episode he tells us why he made that move. His argument is that science is a different problem from coding. A coding agent never has to leave the computer, so progress is mostly about compute and good environments. To make a new material, someone has to mix powders, heat them and measure what came out, and you can't simulate that well enough to skip it. Rishabh walks us through the loop his lab runs and shows where the machines are the limit and where intelligence is. He brings up Bernd Matthias, who found more than a hundred superconductors over a career of, by Rishabh's estimate, maybe 20,000 experiments. What would a scientist like that do with a million? When he was deciding whether to join, people told him to wait: build superintelligence first and let it do the science. He didn't buy it. "Doing AI for science needs doing AI for science." With 1,300 older GPUs and data from their own lab, his team beat a frontier model on a task they care about. We also talk about Anthropic's recent biology announcement and the role of PR in industry science, who oversees experiments when an AI proposes them, and whether skills learned on games like NetHack carry over to real science. Near the end Rishabh explains why, after all these years, he thinks RL is "super dumb as a paradigm", and why that inefficiency helps the big labs. His closing message is that more people in AI should work on harder problems. Topics * Physical AI vs coding agents * Why you can't simulate your way to a new material * "Geniuses in a data center" vs geniuses with a lab * Anthropic's biology announcement and PR in industry science * Oversight fo...

Dan Card: AI Hasn’t Changed Cybersecurity As Much As You Think

1 hr 21 min

Dan Card has spent more than 26 years in cybersecurity as a penetration tester, defender and architect. He helped organizations respond to some of the biggest incidents of the last two decades, from MS Blaster to WannaCry and Log4j. In this episode we ask him what has actually changed now that attackers and defenders both have strong AI models. Dan runs a lab where Claude attacks and ChatGPT defends, and he explains why a swarm of ten thousand agents hitting a defended server mostly gets you blocked faster. AI does help attackers with reconnaissance and research, and it helps defenders work through hundreds of alerts a day. Still, most breaches come from bad passwords and misconfigurations, not from exotic zero days. We also talk about air gaps versus sandboxes, why finding more vulnerabilities doesn’t always mean more risk, prompt injection, robots and autonomous weapons, and the story of how Dan tracked down someone who attacked his website. His main message is that security in the real world should be boring, like a good flight. Topics Social engineering, fake job candidates and mail bombing attacksWho the attackers really are, from teenagers to nation statesWhether an AI could escape and spread across the internetAir gaps, sandboxes and the Stuxnet storyHow AI attackers compare to human onesWhy agent swarms are easy to blockVulnerabilities that exist versus vulnerabilities that can be exploitedUsing LLMs to triage security alertsPrompt injection and how much we can trust modelsRobots, drones and physical riskWhy you defend against the brick before the aliensChapters 00:00 Intro and Dan’s background 01:13 What cybersecurity is and who the attackers are 09:36 Can an AI escape and spread? 19:31 Air gaps, sandboxes and Stuxnet 25:04 How AI is changing attack and defense 34:04 Why agent swarms don’t scare defenders 42:05 AI for security alerts and trusting models 54:57 Crime, deterrence and spending on security 1:02:38 When someone attacked Dan’s website 1:07:29 Robo...

Similar podcasts

All episodes

Post-Training Is Where Models Are Made Now | Maxime Labonne

1 hr 3 min

In this episode, we talked with Maxime Labonne, Head of Post-Training at Liquid AI and author of the LLM Course on GitHub, about what happens to a model after pre-training and why that stage now matters so much. Maxime walks us through the short history of post-training, from supervised fine-tuning and preference optimization to reinforcement learning with verifiable rewards and now multi-domain on-policy distillation (MOPD), where separate teachers for code, math and other domains get merged back into one student. He explains why you still can’t skip SFT, why the order of the stages matters, and why RL works even though a long rollout gives you about one bit of signal. His intuition is that SFT and distillation raise pass@k, and RL then sharpens it into pass@1. We also get into training small models for agentic tasks directly inside the harness, which open-source RL stacks are good enough for serious work, looping transformers, and how much of the work big models do today could move to small ones running locally. We also talk about why models write worse than they used and safety issues. We talk about regulation, open source and Chinese models,  whether a small startup can beat the big labs in a niche, and why he thinks RL environment companies are overrated. We finish with France vs San Francisco, Mistral and the regulation moat, and whether Europe can produce frontier models. Topics Why post-training matters more every yearThe modern pipeline: SFT, RL and distillationTraining agents and small modelsWhy models write worse than they used toSafety, incentives and regulationOpen source and Chinese modelsStartups vs the big labsEurope, Mistral and sovereign AIChapters 0:00 Intro 0:57 Why post-training matters 3:44 A short history of post-training 7:20 The modern pipeline: SFT, RL and distillation 19:41 Agents, architectures and small models 27:15 Why models write worse now 31:44 Safety, incentives and regulation 38:26 Open source and Chinese models 41:39 Recursive sel...

Physical AI for Science: Why It’s Harder Than Coding | Rishabh Agarwal (Periodic Labs)

1 hr 1 min

Rishabh Agarwal has worked on reinforcement learning for years, at Mila, Google DeepMind and Meta. He is now a founding member at Periodic Labs, a company building AI scientists together with the physical labs where they run experiments, and an adjunct professor at McGill. By his own count he understands about ten percent of the science they do there. In this episode he tells us why he made that move. His argument is that science is a different problem from coding. A coding agent never has to leave the computer, so progress is mostly about compute and good environments. To make a new material, someone has to mix powders, heat them and measure what came out, and you can't simulate that well enough to skip it. Rishabh walks us through the loop his lab runs and shows where the machines are the limit and where intelligence is. He brings up Bernd Matthias, who found more than a hundred superconductors over a career of, by Rishabh's estimate, maybe 20,000 experiments. What would a scientist like that do with a million? When he was deciding whether to join, people told him to wait: build superintelligence first and let it do the science. He didn't buy it. "Doing AI for science needs doing AI for science." With 1,300 older GPUs and data from their own lab, his team beat a frontier model on a task they care about. We also talk about Anthropic's recent biology announcement and the role of PR in industry science, who oversees experiments when an AI proposes them, and whether skills learned on games like NetHack carry over to real science. Near the end Rishabh explains why, after all these years, he thinks RL is "super dumb as a paradigm", and why that inefficiency helps the big labs. His closing message is that more people in AI should work on harder problems. Topics * Physical AI vs coding agents * Why you can't simulate your way to a new material * "Geniuses in a data center" vs geniuses with a lab * Anthropic's biology announcement and PR in industry science * Oversight fo...

Dan Card: AI Hasn’t Changed Cybersecurity As Much As You Think

1 hr 21 min

Dan Card has spent more than 26 years in cybersecurity as a penetration tester, defender and architect. He helped organizations respond to some of the biggest incidents of the last two decades, from MS Blaster to WannaCry and Log4j. In this episode we ask him what has actually changed now that attackers and defenders both have strong AI models. Dan runs a lab where Claude attacks and ChatGPT defends, and he explains why a swarm of ten thousand agents hitting a defended server mostly gets you blocked faster. AI does help attackers with reconnaissance and research, and it helps defenders work through hundreds of alerts a day. Still, most breaches come from bad passwords and misconfigurations, not from exotic zero days. We also talk about air gaps versus sandboxes, why finding more vulnerabilities doesn’t always mean more risk, prompt injection, robots and autonomous weapons, and the story of how Dan tracked down someone who attacked his website. His main message is that security in the real world should be boring, like a good flight. Topics Social engineering, fake job candidates and mail bombing attacksWho the attackers really are, from teenagers to nation statesWhether an AI could escape and spread across the internetAir gaps, sandboxes and the Stuxnet storyHow AI attackers compare to human onesWhy agent swarms are easy to blockVulnerabilities that exist versus vulnerabilities that can be exploitedUsing LLMs to triage security alertsPrompt injection and how much we can trust modelsRobots, drones and physical riskWhy you defend against the brick before the aliensChapters 00:00 Intro and Dan’s background 01:13 What cybersecurity is and who the attackers are 09:36 Can an AI escape and spread? 19:31 Air gaps, sandboxes and Stuxnet 25:04 How AI is changing attack and defense 34:04 Why agent swarms don’t scare defenders 42:05 AI for security alerts and trusting models 54:57 Crime, deterrence and spending on security 1:02:38 When someone attacked Dan’s website 1:07:29 Robo...

Why Video Games Still Don’t Use Modern AI | Julian Togelius (NYU)

1 hr 13 min

Julian Togelius, an NYU professor and co-founder of modl.ai, is living the dream of every kid who grew up as a computer geek and gamer. He’s been doing AI research on video games since 2004, back when people told him it wasn’t real research. So why do most video games still run on behavior trees and dialogue trees when we have LLMs? Julian’s answer is that games are designed not to need AI. Most genres were locked in decades ago, and nobody has figured out how to design a good game around a model that can say anything. We talk about the state of the game industry and why so many developers there hate AI. Julian also explains why learning a brand new game as fast as a person would is a better test of intelligence than most benchmarks, and why he thinks we built the wrong kind of AI by training it to copy us. Near the end he gets into making games with Claude and where video world models like Genie might fit. Timeline 0:14 Why games matter for AI 5:21 What counts as AI? 10:02 Why games barely use modern ML 17:36 Why LLM characters break games 20:59 The state of the game industry 27:09 The AI backlash in games 30:41 AI in casual games 36:07 AI-made mods and fake players 40:42 Games as a test of intelligence 47:33 Can GPT-6 beat any game? 51:10 Playing alongside AI agents 53:36 We invented the wrong AI 58:33 Personal assistants and Jeb 1:03:22 Making games with AI 1:07:50 Video world models 1:11:16 Joining Julian’s lab 1:13:48 Favorite games Topics Why games are a good testbed for AI Classic game AI vs machine learning Why shipped games rarely use neural networks LLM NPCs and game design The game industry slump and the AI backlash AI in mobile games like Candy Crush AI-generated mods and simulated players Games as a test of general intelligence Automated playtesting and reward hacking Designing games around AI agents “We invented the wrong AI” Generating games with code, asset models and world models Music: "Kid Kodi" by Blue Dot Sessions (CC BY-NC 4.0)

What’s Left for Mathematicians? | Daniel Litt

1 hr 18 min

In this episode, we talked with Daniel Litt, a mathematician at the University of Toronto. In March 2025, Daniel bet that AI would not be able to autonomously produce an Annals of Mathematics-level paper by 2030. He has since conceded. He told us what changed and why. The models became superhuman at a small set of skills, like knowing the entire literature and grinding through technical arguments, and that turned out to be enough to do real math. They are still weak at the fuzzy part of the job, like figuring out what is actually true about an object or which questions are worth asking. We also got into how anyone can check a proof that no human has read yet, and what a Lean formalization does and doesn't guarantee. Daniel thinks prestige and credit in math are dead, and that we will have to judge people through thesis defenses and seminars rather than through PDFs. He explains why he thinks math got Goodharted and why the labs put so much effort into it in the first place. We finished with listener questions, including how to teach kids math now and whether P vs NP might be just one idea away. Topics - Why Daniel conceded his AI math bet - What models are superhuman at in math, and where they fall short - Why AI won't "solve math" - Checking AI-written proofs and the limits of Lean - Why different models keep proving the same results - How Daniel uses AI in his own research - Evaluating mathematicians when prestige and credit are dead - Why math got "Goodharted" - AI labs, race dynamics and publication norms - Teaching kids math, and P vs NP 00:00 Intro 00:15 The bet Daniel conceded 03:23 What AI is good and bad at in math 08:13 Will AI solve all of math? 11:37 Why we still need human mathematicians 13:59 Math is about understanding, not only proof 21:26 Axioms, foundations and Lean 27:42 How do you check an AI-written proof? 31:22 When models cheat 36:18 Are all the models proving the same things? 39:51 How Daniel uses AI in his own research 45:20 "Prestige and cr...

The Political Economy of AI: Governance, Bias, and the Future | Andy Hall (Stanford/Anthropic)

1 hr 24 min

In this episode of the the information bottleneck podcast, we talked with Andy Hall, a Stanford professor and MTS at Anthropic. The conversation explores the political economy of AGI, analyzing how superintelligence could reshape job markets, wealth redistribution, and public trust in government. Hall presents an optimistic view that hyperintelligent agents could enhance democracy by helping citizens make better political decisions, likening AI's potential impact to the invention of the printing press. The discussion also covers the challenges of regulating open-source models, the origins and mitigation of ideological biases in AI, and the geopolitical implications of the AI race between the US and adversaries like China. List of TopicsThe definition of the political economy of superintelligence and its potential to improve civic decision-making.The future of work, the likelihood of white-collar job loss, and the role of the existing tax system in wealth redistribution.Public skepticism toward AI, the decline of trust in the US government, and the contrast between AI power users and the general public.The tension between personal freedom and collective safety regarding open-source and open-weight AI models.Using AI tools to assist legislators in drafting policy and auditing frontier AI models.The geopolitical AI race, the threat of state-run hacking groups, and the challenges for middle-power countries.Center-left ideological biases in current AI models, their roots in training data and reinforcement learning, and experiments to fine-tune models with alternative viewpoints.The concept of AI constitutions, market competition over model values, and the risks of government intervention.AI-driven astroturfing on social media, ungovernable AI swarms, and the debate over regulatory "kill switches".Timeline00:14 - Introduction of Andy Hall and an overview of the political economy of superintelligence.05:38 - Discussion on the future of work, potential mass unemployment, an...

John Platt (Google): AI for Climate and Science

1 hr 12 min

John Platt is a Google Fellow and leads the Applied Science team at Google Research, the group working on climate and science. He's also the person behind SMO and Platt scaling. In this episode, we talk with him about how Google decides which scientific problems are worth its time, and what happens when AI meets real physical systems. A big part of the conversation is contrails. A small share of flights creates most of the warming from contrails, which together add up to roughly 1% of human-caused warming. John explains how his team detects contrails from weather satellites and predicts where they will form. In trials with American Airlines, rerouting flights cut contrail formation by 62% with almost no extra fuel. We also get into ERA, Google's empirical research assistant, and where AI agents actually help scientists. John sees them taking over the drudgery of building models and munging data, while humans keep the taste and the rigor. He shares why he's skeptical of gradient-free optimization, why the bitter lesson runs into energy and data center limits, and why physical AI that makes manufacturing cheaper could matter a lot for climate. We finish with FireSat, a satellite constellation (about 50 satellites at full scale) designed to catch wildfires while they're still small enough to put out. Timeline 00:00 Intro 01:09 How Google picks science problems: comparative advantage 04:15 Projects that didn't work: RNA aptamers and tool/problem fit 05:46 DNA, RNA and XNA 07:41 Contrails and why they warm the planet 13:01 Three AI systems for avoiding contrails 15:23 Cloud seeding and why cloud physics is hard 18:17 Airline trials: 62% fewer contrails 19:46 Navier-Stokes, wind tunnels and where simulation still fails 25:07 Will agents replace scientists? ERA and scorable tasks 31:36 Overfitting, Kaggle and the half-pixel bug 33:41 Agents that exploit bugs: Lean and Goodhart's law 36:45 Gradient-free optimization 38:53 SMO, SVMs and hinge loss in neural networks 42:12 Th...

Yuandong Tian on Recursive Self-Improvement

58 min

Yuandong Tian spent many years at Meta FAIR and recently left to co-found Recursive Superintelligence, a company building AI that improves itself. In this episode, he tells us why. We start with his work on computer Go, where he built DarkForest before AlphaGo came out. A few years later came OpenGo, which played Korean professionals on a single GPU and didn't lose a game. He then tried to bring RL to real-world problems and found that the design of the action space mattered more than the algorithm. We also talk about Coconut, his paper on reasoning in latent space instead of tokens, and why frontier models still aren't trained that way. The second half is about recursive self-improvement. Yuandong wrote his last paper at Meta together with GPT-5 and says it made him 6 to 10 times faster. That convinced him his own job could be replaced within five years. We ask him where agents still fall short and whether a new architecture can really beat transformers at scale. He also gives his view on the calls to restrict self-improving AI and makes the case for open source. Recursive is hiring in San Francisco and London: talent@recursive.com Topics * Computer Go: DarkForest, AlphaGo and OpenGo * Gradient-free optimization * Action space design and neural architecture search * Coconut and reasoning in latent space * Understanding how neural networks learn representations * Grokking, and writing a paper with GPT-5 * Recursive self-improvement and coding agents * New architectures vs. transformers * The NanoGPT speedrun * Data efficiency and robotics * Restricting self-improving AI * Open source models Timeline 0:00 Intro 0:45 From CMU to deep learning, and the AlexNet debates 5:35 AlphaGo and beating Go pros on a single GPU 10:27 Gradient-free optimization 13:16 Diffusion models and diversity 16:23 RL on real problems: why the action space matters 20:34 AutoML and architecture search 22:05 Coconut and reasoning in latent space 26:07 Why latent reasoning hasn't caught on 30:36...

Chris Manning: Language Is the Real Unlock for Intelligence

55 min

Chris Manning is a legend in NLP. He's a professor of linguistics and computer science at Stanford and ran the Stanford AI Lab. His course CS224N, Natural Language Processing with Deep Learning, is where a whole generation of researchers learned the field. If you've worked on anything in NLP over the last 25 years, you've probably used his work. We start with what linguistics gave machine learning that ML wouldn't have figured out on its own, and what's left for NLP researchers now that LLMs handle most of the classic tasks. Chris thinks the open problems have moved up the stack to pragmatics and dialogue. Models are good at using context but still sound confident when they shouldn't. Then we get into his recent work on how language models learn verb classes. Small GPT-2 models seem to form abstract categories right away instead of memorizing verbs one at a time, and Chris explains why distributed representations push them in that direction. The second half is about meaning and representation. Chris makes the case that Yann LeCun underrates the role of language in intelligence, and revisits his debate with Emily Bender over whether text alone can teach meaning. We finish with ReFT, which steers a frozen model by editing its hidden states, and whether concepts really live in linear subspaces. Timeline 00:00 Intro 01:05 What linguistics gave machine learning 03:27 Is NLP losing its focus on language? 06:29 What's left for NLP researchers in the LLM era 09:37 The next frontier: pragmatics and dialogue 13:01 Constructed languages and AI-to-AI communication 15:35 Do language models learn categories first? 23:43 Why transformers learn abstractions early 26:37 Does it hold at scale? 28:30 LeCun, JEPA and the role of language in intelligence 34:54 Can text alone teach meaning? The octopus debate 40:45 Diffusion language models vs transformers 43:39 ReFT: editing representations instead of weights 50:34 Weights or representations: where knowledge lives 52:37 Do concepts real...

Why AI Still Can't See wtih Andrew Dai

58 min

Andrew Dai spent over a decade at Google Brain and DeepMind, where he co-wrote the 2015 paper that introduced language model pre-training followed by fine-tuning, and later co-led pre-training data for Gemini. He's now co-founder and CEO of Elorian AI, which is building models for visual reasoning. We talk about how his pre-training result started as a bug, why next-token prediction scales better than other objectives, and what went wrong for Google in the early LLM race. The second half is about vision: why today's frontier models still can't count objects in a photo, why he thinks reasoning is fundamentally visual, and how his view of world models differs from JEPA. Chapters 00:00 Intro 00:48 Andrew's background and the accidental discovery of pre-training 07:10 Why next-token prediction scales 10:09 How Google fell behind and the early days of Gemini 18:28 What makes training data good 28:16 Where visual understanding breaks down 46:01 World models, JEPA and robotics 55:27 Generation vs understanding, and what's next for Elorian Topics Pre-training and fine-tuningScaling and next-token predictionGemini and Google's AI historyData quality and synthetic dataVisual reasoning and countingWorld models and JEPAMusic "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0

Can AI Replace Doctors? Zachary Lipton on the Future of Healthcare and AI

1 hr 31 min

Zachary Lipton is an associate professor at Carnegie Mellon University and a co-founder of Abridge, a healthcare AI company (and a jazz saxophonist!). His research spans machine learning, healthcare, and the broader impact of AI. In this episode, we discuss what AI can (and cannot yet) do in healthcare. We talk about why medicine is harder to automate than coding, AI scribes and clinical decision support, drug discovery, whether AI could eventually replace doctors, and the role of open models and intelligent routing. In the second half, we turn to the future of AI research: whether academia can still compete, what AI PhD students should work on, and how Zach thinks about automation and the future of research. Timeline00:00 — Introduction 00:28 — Why healthcare is hard for AI 05:00 — Zach’s path into healthcare AI 22:13 — Where AI can have the biggest impact 32:01 — Can AI replace doctors? 41:41 — Open models and AI infrastructure 56:20 — The future of AI research 1:03:03 — What should AI PhD students work on? 1:27:00 — Advice for researchers and founders TopicsAI in healthcare • AI doctors • drug discovery • clinical decision support • open-source AI • AI research • academia vs. industry • future of work Music "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0

Sara Hooker on the End of Static AI

1 hr 35 min

What comes after scaling? We talk with Sara Hooker, co-founder and CEO of Adaptation Lab, about why the next generation of AI may look very different from today's static models. Sara argues that models should continuously adapt to new tasks, data, users, and environments—and that doing this efficiently will require rethinking much more than fine-tuning. We discuss continual learning, AutoScientist and automated research, why non-verifiable tasks may become the next major bottleneck, and why interfaces could be as important as the models themselves. We also get into open vs. closed models, distillation and Chinese AI labs, AI regulation and safety, cybersecurity and biorisk, AI companionship, and what may eventually come after Transformers and tokenization. TopicsContinuous learning and adaptive AIFine-tuning, memory, and AutoScientistAI agents and automated researchNon-verifiable tasks and human feedbackAdaptive interfacesOpen vs. closed models and distillationAI safety, regulation, cyber risk, and bioriskAI companionship and persuasionThe limits of TransformersMultilingual models and tokenizationChapters00:00 — Introduction 02:15 — Why start another AI lab? The return of research 05:46 — What continuous learning actually means 12:04 — Should every company have its own adapting model? 13:59 — Fine-tuning and platforms like Tinker 18:04 — AutoScientist and automated optimization 22:52 — Can AI really improve its own research? 28:38 — The problem of non-verifiable tasks 31:30 — Human feedback and the limits of exponential progress 34:43 — Why the AI interface matters 40:36 — Distillation, China, and open models 49:05 — Open-model licensing 52:19 — Will open models catch closed models? 58:43 — AI regulation and compute thresholds 1:03:07 — AI safety and agent failures 1:10:19 — Biorisk vs. cybersecurity 1:14:03 — Persuasion, AI companionship, and overlooked risks 1:20:41 — Where will AI have the biggest real-world impact? 1:25:41 — What is missing from current AI arch...

Tiny Recursive Models Beat the Giants - Alexia Jolicoeur-Martineau (Microsoft)

52 min

Alexia Jolicoeur-Martineau is a Principal Researcher at Microsoft and the author of "Less is More: Recursive Reasoning with Tiny Networks," the paper behind the Tiny Recursive Model that hit about 45% on ARC-AGI-1 with a fraction of the parameters of frontier systems. It won the 2025 ARC Prize paper award. She read the hierarchical reasoning paper, thought the potential was real and the explanation was not, and rebuilt it without the mouse brains: a small network that carries a hidden state and a current answer, thinks for a few steps, updates, and repeats, with the gradient truncated at each loop. We get into why puzzles suit this and autoregression doesn't, why she thinks LLMs are bad at molecules and more data won't fix it, and what she'd do with a trillion dollars. Timeline 00:01 Intro01:06 Leaving biostatistics, and why the field stagnated06:47 GANs, diffusion, and research on four GPUs12:58 What was wrong with the hierarchical reasoning paper16:31 Tiny recursive models explained without the biology22:35 Why puzzles favor recursion over left to right generation24:15 Is the bitter lesson really bitter?27:28 With infinite compute, would you still want small models?32:00 Self improvement, memory, and a trillion dollars37:01 Test time compute beyond chain of thought40:41 Why chain of thought fails on molecules45:17 Is there a universal representation?48:06 What people are already building with TRM55:22 Fixed point models and DEQMusic "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0Topics Tiny Recursive Models and the ARC-AGI resultsWhat the hierarchical reasoning model was really doingDeep supervision and truncated backpropLooping transformers and parameter efficiencyWhy puzzles favor whole-context iteration over left to right generationTest time compute beyond chain of thoughtLatent reasoning and the Coconut line of workWhy LLMs fail on chemistry and physicsRepresentation learning and whether a universal representation exists

Continual Learning Is the Next Bottleneck | Rohan Anil (Core Automation )

1 hr 11 min

Rohan Anil spent eleven and a half years at Google, where he went from writing memory allocators to large-scale linear solvers, then optimization at Google Brain, where he co-developed distributed Shampoo and led optimization for PaLM and Gemini pre-training, including the work that produced Gemini Flash. He then joined Anthropic's pre-training team, and left before the IPO to co-found Core Automation with Jerry Tworek (ex-VP of Research at OpenAI). We talk with him about how Brain worked at its peak, why he left two of the world's best labs, and what he thinks is missing from today's models. Rohan's view is that pre-training and RL were split by organizational convenience rather than by science. Pre-training builds a prior, and RL sharpens it to the tasks we care about, and neither gives a model a way to absorb new data or learn from its own experience once it is deployed. Post-training more every day plateaus, on-policy distillation plateaus, and in-context learning only goes as far as the context does. He argues the next architecture needs better ways to fold in new knowledge at inference time, and that this is a fundamental optimization question rather than a harness-engineering one. We also get into why coding agents still fail on low-level systems work, his take on Muon, why second-order methods matter once you leave the noise-dominated regime, and why nobody can yet use a few million GPUs for a single training run. Timeline 00:00 Intro 01:09 From computer vision to Google systems engineering 02:37 Large-scale linear solvers and sparse features 06:01 Getting into optimization: SDCA and Yonghui Wu's team 07:29 Joining the Shampoo crew 09:35 The Google Brain ethos, and why 2017 to 2019 was special 14:23 Is open research going to keep winning? 15:53 Frontier models are only as good as the prior you give them 17:30 Missing the language model wave, then Common Crawl and online distillation 18:31 Paternity leave, DALL-E Mini, and the 14 days that became two years 2...

World Models | John Langford (Microsoft AI Labs)

1 hr 5 min

John Langford, one of the heads of Microsoft's AI Labs, the creator of Vowpal Wabbit, and a co-inventor of CAPTCHA, joins us to talk about world models. Transformers need orders of magnitude more data than humans to learn the same thing, and John argues a compact, implicit world model is how you close that gap. He explains why he's skeptical of JEPA-style objectives, why a transformer's KV cache is the Ptolemaic epicycle model of belief states, and what his Next Latent work does differently. We also get into whether research still matters in the age of scale; open versus closed models; agent-driven research after running 2,000 pre-training experiments in 90 days; the origin story of CAPTCHA; and why Muon and orthonormal optimizers actually work. Topics: Implicit vs. explicit world models, and the case against JEPA-style objectivesCompact belief states: why compression beats a growing KV cacheDoes research still matter in the age of scale? The Kimi K3 argumentAgent-driven research: 2,000 pre-training experiments in 90 daysThe invention of CAPTCHAOptimizers from SGD and Vowpal Wabbit to Muon Chapters00:00 Why world models: the sample-complexity gap09:48 The case against JEPA; a transformer-style implicit world model15:52 Compact belief states: epicycles vs. heliocentrism23:41 Does research still matter? The Kimi K3 argument27:35 Open vs. closed models35:57 Recursive self-improvement and agent-driven research42:30 2,000 pre-training experiments in 90 days; weak baselines and reproducibility54:54 The invention of CAPTCHA1:00:53 Optimizers: from Vowpal Wabbit to MuonMusic "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0

Which Tabular Model Should You Actually Use? | David Holzmüller (INRIA)

55 min

Description Tabular data is still where most of machine learning actually happens in industry, and the field has changed a lot in the last few years. In this episode we talk with David Holzmüller, a researcher at INRIA and one of the people behind TabArena, TabICL and RealMLP, about what the state of the art looks like right now and how to pick a model for your own data. We cover the shift to TabPFN-style foundation models that learn to learn from whole tables, why TabArena was built and what earlier benchmarks got wrong, what Google's new TabFM means for the leaderboard, and when gradient boosted trees are still the right tool. David explains why LLMs struggle with tables, shares an early result comparing Claude Opus against TabICL on tiny datasets, and walks through how to embed text columns for tabular models. We also get into time series vs tabular data, the open research problems he thinks matter most, and why classical ML libraries are so bad out of the box. Links: TabArena: https://tabarena.ai Topics Tabular foundation models and in-context learning on tablesTabArena and Beyond Arena: building a benchmark that stays honestTabFM, TabPFN, TabICL and the tradeoffs between themWhen boosted trees and MLPs still win (large data, CPU, fast inference)Why LLMs are inefficient on tabular data and where they might helpEmbedding text columns with language modelsExplainability, calibration and class imbalanceTime series vs tabular dataOpen problems: invariances, synthetic data, uncertainty, scaling downWhere the field is heading in the next five years Chapters 0:00 Intro 0:31 What changed in tabular ML: TabPFN-style foundation models 2:22 Which model to try first? TabArena and how it was built 5:14 What older benchmarks got wrong, and Beyond Arena 8:45 GPU AutoML vs foundation models 10:40 Reading the leaderboard: TabFM, TabPFN, TabICL and the tradeoffs 12:47 Calibration, class imbalance and small vs large data 19:45 Explainability for black-box tabular models 21:34 Why...

Why You Can't Just Rent 1,000 GPUs | Charles Frye (Modal)

1 hr 3 min

Charles Frye (Modal, ex-Weights & Biases, Berkeley PhD) joins Ravid and Allen to explain why modern AI research is bottlenecked by compute, and why simply buying more GPUs doesn't solve it. We cover the three problems every lab hits (underutilization, saturation, resource sharing), when companies should actually train their own models, why inference is a "bad algorithm" for today's hardware, NVIDIA's monopoly, the OpenAI/Hugging Face hack and what it says about open models, and whether we're in a compute bubble. Key topics AI infrastructure challenges and when to train your own modelsGPU resource management and virtualizationInference optimization and speculative decodingThe economics and future of AI hardwareAgents, sandboxing, and open-model security Chapters 00:00 Intro 01:03 Why AI needs special-purpose compute 03:22 Buying vs renting GPUs: the three problems 07:15 Modal's approach, and doing more with less compute 09:46 Do we actually need to spend more? The conflict-of-interest question 13:08 Should companies train their own models? 14:47 Efficient fine-tuning and prompts as fast weights 17:37 Are we in a compute bubble? 20:21 Why inference will dominate compute (the SQLite analogy) 22:42 Speculative decoding 26:44 Why scaling inference is hard, and neuromorphic hardware 28:36 Why NVIDIA's monopoly persists 33:09 Inference chip startups and the hardware lottery 35:24 How Modal stays hardware-agnostic (GPU snapshot restore) 38:45 Will agentic coding erode CUDA's moat? 41:18 Running one agent vs thousands: sandboxing at scale 46:27 The OpenAI/Hugging Face hack and open models as defenders 52:28 Rogue AI, self-replication, and fast takeoff 56:09 What's next: evals, embodiment, edge inference 1:00:27 Modal is hiring (modal.jobs) Music"Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0

Stella Biderman (EleutherAI) - Open Source, AI Safety, and Who We Can Trust

1 hr 6 min

Stella Biderman, Executive Director of EleutherAI, joins us the week an OpenAI model autonomously broke out of its sandbox and hacked Hugging Face. Stella calls it what she thinks it is, an offensive cyber operation, and argues it's part of a pattern: this is not the first containment failure at a frontier lab, and sandboxes have failed basically every time they've been tested for real. So we spend a good chunk of the episode on what actual containment would look like. Stella's argument is that the tools already exist, the labs just don't use them: run dangerous capability evals on air-gapped networks with no route to the public internet, put the most sensitive testing in SCIF-style secure facilities, and treat model evaluation the way the security world treats classified systems rather than the way startups treat staging environments. And yet Stella remains one of the world's most prominent open-source advocates. From her perspective, the biggest risk isn't the technology; it's unchecked corporate power, and the only durable check on it is an independent scientific research establishment that doesn't depend on the AI industry for its funding or its facts. From there the conversation spans the geopolitics of Chinese open models and whether governments can restrict them, sovereign AI and what it would actually take for other countries to train their own models, why harnesses and UX drive more of AI's perceived progress than raw intelligence, the AI-found counterexample to the Jacobian conjecture, and EleutherAI's "Deep Ignorance" approach to making open-weight models safe by filtering hazardous knowledge out of pretraining. key topics AI governance and regulationCybersecurity incidents involving AI modelsOpen source AI safety and securityThe role of independent research in AI safetyLegal and ethical considerations in AI developmentTimeline 00:13 — Intro: Stella Biderman and EleutherAI, a real non-profit in AI02:05 — News of the week: Kimi K3, and OpenAI's model auto...

Why Deep Learning Finally Works on Tables | Frank Hutter (Prior Labs)

1 hr 18 min

In this episode, Frank Hutter joins us to talk about TabPFN and why tabular data is suddenly the hottest problem in deep learning. Frank is a professor at the University of Freiburg and spent 15 years building the AutoML field before founding Prior Labs, which SAP just acquired for over a billion dollars. We get into why deep learning failed on tables for a decade and what in-context learning changed, how TabPFN is trained entirely on synthetic data, and why a model that never saw a real time series ended up beating specialized forecasting models. Frank also explains the architecture tricks behind scaling from 10,000 to a million rows, where LLMs fit into data science (and where they embarrassingly don't), and what happens to XGBoost from here. Beyond the research, Frank talks about the jump from professor to co-CEO, why he refused to merge his 45-person team into SAP's 110,000 employees, the open-weights licensing debate, and the case for building a frontier lab in Freiburg rather than San Francisco. key topics The role of foundation models in tabular dataImpact of SAP acquisition on Pro LabsThe evolution of AutoML and hyperparameter optimizationChallenges and solutions for large context in modelsOpen source models and licensing strategiesThe importance of independence for startup agilityFuture directions in AI for science and medicine00:00 Intro 00:34 The SAP acquisition and staying independent 07:39 Why tabular data is the next big thing in deep learning 14:19 What makes tabular data hard 19:14 AutoML, AutoGluon, and fifteen years of hyperparameter tuning 28:27 Scaling TabPFN: context limits and architectures 34:35 Agentic data science and LLMs 39:30 Online learning, time series, and Bayesian inference in a forward pass 47:05 Open weights and the license debate 54:51 Will LLMs and tabular models merge? 1:00:01 From academia to startup 1:09:42 Why build in Europe 1:12:53 Audience questions and hiring Music"Kid Kodi" - Blue Dot Sessions - via Free Music Archive - C...

Surya Ganguli: The Physics of Intelligence

1 hr 25 min

Surya Ganguli is a professor at Stanford and VP at General Catalyst, working at the intersection of physics, neuroscience, and AI. He started in string theory, moved to theoretical neuroscience, and now uses tools from statistical physics to understand both brains and neural networks. We talk about why deep learning theory is finally catching up to practice,  including his group's recent work explaining neural scaling laws, and why smarter data selection could beat them entirely. He also tells the origin story of diffusion models, which were invented in his lab as an attempt to violate the second law of thermodynamics. The second half turns to the brain: what happens to a mouse's sense of self on ketamine, how stimulating a handful of neurons can induce hallucinations, and a method his lab developed to get a neuron deep in a monkey's brain to describe, in English, what makes it fire. We close on where he thinks AI is going wrong: models train on ten trillion tokens while humans hear a hundred million words, because we don't teach children with gradients; we tell them the algorithm. key topics Connections between physics, neuroscience, and AIEmergent properties in complex systemsScaling laws in language modelsData efficiency and pruning in AINeuroscience insights into consciousness and selfThe future of AI and brain modelingChapters 00:00 Introduction to Surya Ganguli 00:57 Surya's Background: From String Theory to Neuroscience 02:22 Emergent Properties in Physics, Neuroscience, and AI 03:16 Energy Landscapes and Loss Landscapes in High Dimensions 04:07 Why Local Minima Don't Exist in High-Dimensional AI 05:22 Gradient-Based vs. Gradient-Free Learning Methods 08:21 AI in Mathematics and Drug Discovery: Opportunities and Challenges 13:48 Scaling Laws and Data Efficiency in Language Models 18:10 Properties of Data that Affect Scaling Laws 22:04 Constructing Non-Redundant Data Sets for Better Learning 24:32 Theory vs. Empirical Results in AI Research 32:19 Fundamental...

Text Diffusion Models with Brendan O'Donoghue (Google DeepMind)

1 hr 9 min

Brendan O'Donoghue, research director at Google DeepMind, makes the case for text diffusion as a real alternative to autoregressive generation. He walks through how discrete diffusion works, why diffusion samples are far more diverse and what that unlocks for RL, where the Gemma diffusion model actually stands against frontier models, and why the whole training and serving stack being hyper-optimized for autoregression is the main thing holding the approach back. The conversation also covers hardware trends favoring flops over bandwidth, AGI timelines and real-world bottlenecks, and why he thinks RL is still underhyped. Key topics - Discrete diffusion for text vs autoregressive generation - Why diffusion samples are more diverse, and what that unlocks for RL - Where diffusion already wins: latency, on-device, robotics - Why serving cost, not quality, is the real blocker - RL as the most underhyped area in AI Timeline 00:00 Introduction 00:50 What diffusion models are and how text diffusion works 04:40 Why Brendan bet on text diffusion in 2023 07:15 Diversity, creativity, and why it helps RL 11:00 The best diffusion LLM today and the gap to frontier models 14:25 Latency, serving cost, and why it needs more chips 17:14 Where diffusion already wins: on-device, robotics, battery 20:14 One model, two modes: diffusion for thinking, AR for answering 22:24 Samplers and the stuttering problem 26:27 Theory, BERT, and why now is a good time to work on this 31:48 Pipelines built for autoregression, and continuous diffusion 35:35 Hardware: flops vs bandwidth 39:49 AGI timelines and real-world bottlenecks 50:15 Is AI engineering or science? 54:14 Most overhyped and most underhyped ideas 58:35 RL on diffusion, value functions, and exploration 1:07:30 Go download the model and break it Music "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0.

Nathan Lambert: Inside Post-Training and the Open Model Fight

1 hr 15 min

Nathan Lambert spent three years as post-training lead at Ai2, where he built the OLMo models, and he writes Interconnects, one of the most-read technical newsletters in AI. He left Ai2 in June and is now working on a new project. He's also the author of the RLHF book. We talked a lot about open models, their capabilities, and why they are better than he expected. We get into what that means over the next two to five years, why he thinks recursive self-improvement is overblown, what the market for training environments actually looks like now, and why he expects Anthropic's famously open internal culture to break after its IPO. Key Topics Open vs closed models and who actually captures the valueAnthropic and OpenAI as opposite cultures, and the talent concentration problemBoom vs bubble, and why token spend hasn't produced 10x better productsContinual learning, RSI skepticism, and what Nathan wants to work on nextWhat the open ecosystem needs economically to survive Timeline 00:00 Intro 00:27 Open vs closed models, and who actually captures the value 05:12 China, harnesses, and where the real training leverage sits 08:40 Sovereign compute and the national security case for building models 11:18 Uncensored open weights and the bioweapon question 14:29 Anthropic vs OpenAI, ideology and politics 19:35 The Mythos ban and the Fable 5 delays 24:30 The AGI narrative, the talent drain, and antitrust 28:12 Why researchers join Anthropic, and the open Slack culture 34:04 Nathan's next 12 months: character training and big RL runs 37:55 Continual learning, RSI, and why Nathan is skeptical 43:19 Boom or bubble, tokens vs GPUs 45:12 Why all that token spend never produced 10x products 48:38 Job displacement and the small-business future 52:49 Robotics, world models, and why multimodal lags 57:44 What the open ecosystem should actually do 1:03:17 Why NVIDIA isn't building a frontier model 1:07:34 The RLHF book, and whether RLHF still matters 1:11:06 GRPO vs PPO and on-policy di...

Daphne Koller - The Future of AI in Biology and Drug Discovery

1 hr 2 min

Daphne Koller wrote the book that many of us learned probabilistic graphical models from, founded Coursera, and now runs insitro, which is trying to make drug discovery a machine-learning problem. We start with the bitter lesson. She agrees with most of it and then says where it stops working: biology doesn't have enough data, structure is how people understand anything, and making a drug is a question about an intervention that hasn't happened yet, not a pattern in data you already have. Most of the episode is about why drug discovery is hard. Ninety percent of drugs that reach the clinic fail, and mostly not because the molecule was bad. The molecule usually does what it was designed to do. It just turns out the thing it was designed to do had nothing to do with the disease. Only 22% of diseases have any approved drug at all, and she calls that an upper bound on what we understand, not a lower bound. She also gets into what agents are and aren't good for in a wet lab, why cells don't grow faster no matter how many GPUs you point at them, what it would take to have real foundation models for biology, and why almost all of biology is still out of distribution. Plus GLP-1s and what human data keeps teaching us, whether AI can make the kind of leap that turned a bacterial immune system into CRISPR, and what she'd build if she were starting Coursera today. Key Topics The impact of scaling and data in machine learningThe importance of structure and causality in AIChallenges in drug discovery and biological understandingThe role of foundation models in biologyEthical considerations in AI and biomedical research Chapters 00:00 Introduction to Machine Learning and Drug Discovery 02:00 The Bitter Lesson and Its Implications 06:48 Challenges in Drug Design and Discovery 11:48 Ethical Considerations in Human Research 17:20 The Drug Discovery Pipeline Explained 29:30 Integrating AI in Experimental Design 35:38 The Role of Human Judgment in Drug Design 37:14 Future of Drug D...

RL Was Broken at Every Level - With Joseph Suarez (PufferAI)

1 hr 3 min

In this episode, Joseph Suarez from PufferAI explains why he thinks RL never had an algorithm problem, but it had a code problem. Every part of the standard RL stack was running about a thousand times slower than it should have been, and once that got fixed, problems that used to take months started getting solved in seconds on one GPU. We talk about what makes a simulator good for RL, why most of their sims run on CPU, what he wants to do with scientific simulation, and why he open sources all of it instead of writing papers. Key topics Types of RL and their applicationsChallenges in scaling reinforcement learningThe role of simulators and hardware in RLRL in gaming: from chess to complex games like NetHack and RuneScapeFuture directions: scientific simulation and biological modeling Chapters 00:00 - Introduction to RL and Puff AI 01:50 - Different settings for RL: Games, Robots, Finance 04:10 - RL in LM and other domains 07:00 - Challenges and solutions in RL scaling 09:55 - Building fast, efficient simulators 15:10 - RL for scientific research and simulation 19:57 - RL in complex games: NetHack, RuneScape, Dwarf Fortress 29:55 - Future of RL: Scientific discovery and beyond Resources Puff AI - Official Site - https://puffer.ai NetHack - https://www.nethack.org/ RuneScape - https://www.runescape.com/ Dwarf Fortress - http://www.bay12games.com/dwarves/ OpenAI Gym - https://github.com/openai/gym Music "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0.

The Model Found a Way Out - with Florian Brand (Prime Intellect)

57 min

Florian Brand builds evals at Prime Intellect. The premise of the conversation is that writing a benchmark is the easy part now. Keeping the model from cheating it is the job, and it takes longer than the benchmark itself. We get into why he thinks you can't evaluate a model apart from the CLI it runs in, what happens to statistics when a single run costs five figures, and whether the feeling that a model just works can ever become a number. He also has a few stories about agents finding their way around the scoring that are worth hearing cold. Timeline 00:13 Intro01:00 What evals are for04:05 Agentic benchmarks07:10 Kimi K2 and model diversity08:23 Long-horizon coding tasks10:29 Building a benchmark12:15 MirrorCode14:27 Rubrics and LLM judges16:30 The cost of expert labelers17:49 Long runs and variance19:44 Evaluating the harness24:29 Chinese labs building CLIs30:00 More reward hacking37:45 Tau-bench and economic tasks39:43 Benchmaxxing and GLM 5.245:15 Statistics and cost47:56 Frontier convergence52:04 Misuse in open and closed models55:35 Self-improvement Music "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0.About The Information Bottleneck is hosted by Ravid Shwartz-Ziv and Allen Roush, featuring in-depth conversations with leading AI researchers about the ideas shaping the future of machine learning.

Pierre-Carl Langlais on Building Models from Data You Can Account For

1 hr 5 min

Most labs build language models by scraping the web and filtering afterward. Pierre-Carl Langlais runs it the other way around. At Pleias, the French-German lab he co-founded, the models are built from data he can actually account for, which in practice means open and public-domain sources plus a lot of synthetic data the lab generates itself. It sounds like a self-imposed handicap. It mostly isn't. One of their models is a 600 million parameter system that runs live inside the Paris subway's monitoring pipeline. We cover the SYNTH pretraining dataset and why he thinks "ethical data" has to mean more than copyright-free. He explains why barely 2% of their Common Corpus appears in typical web crawls, and why that gap is really a preservation problem. From there, he gets blunt about benchmark maxing and whether GLM really earns its Opus-class reputation. He also argues that the quiet move by closed labs to hide reasoning traces is mostly about claiming ownership of model outputs. He's skeptical of sovereign AI, and not shy about how Mistral drifted from frontier research toward French corporate consulting. We finish on NVIDIA's persona datasets and the odd idea of training on the conditions that produced a text rather than the text itself. Timeline(00:02) Welcome and introductions(00:49) Why synthetic data matters, and the SYNTH set(04:15) Three reasons to control your training data(07:18) What "ethical data" actually means(11:08) How Common Corpus got built, from Wikipedia to PDFs(16:35) Agentic harnesses and synthetic data(20:03) Evaluating data when you train on reasoning traces(25:27) General versus specialized pretraining(27:08) Benchmark maxing and the GLM question(31:51) Getting diversity in, and the NVIDIA personas(35:02) Hidden reasoning traces and the fight over model IP(38:17) Mid-training and the "It's All Training" thesis(41:47) Can small models actually compete(45:01) Cybersecurity and Europe's strategic gap(47:08) Do you need a big model to orchestrate...

Dhruv Batra: The Browser Is a Robotics Problem - From Embodied AI at Meta to Web Agents at Yutori

1 hr 7 min

Dhruv Batra spent years leading Embodied AI at Meta,  training virtual robots to navigate photorealistic 3D scans of real buildings with pure reinforcement learning. Then he left to co-found Yutori and build agents for a very different environment: the web browser. In this episode, Dhruv explains why he sees these as the same problem. Web agents, in his framing, are robots that act in a browser (pixels in, actions out), and the web turns out to be just as messy an environment as the physical world. Along the way, we cover his definition of intelligence as "navigation in idea space," why robotics is lagging LLMs, the sim-to-real gap and why you can't fake friction coefficients, the teleoperation counterexample to the "it's a sensor problem" argument, and his provocative claim that under the current paradigm, we solved machine learning and didn't even realize it. He also makes the case for why the scaling hypothesis isn't falsifiable, why JEPA-style arguments deserve to be grappled with, how Yutori trains its Navigator models with RL on live websites, and what happens to the ad-supported web when agents, not eyeballs, do the browsing. Timeline00:01 — Intro 00:54 — What embodied AI actually means 06:47 — Intelligence as navigation in idea space 13:26 — Habitat: training robots with pure RL, no maps 20:04 — Why robotics is behind LLMs 28:24 — Sim-to-real: what you can and can't fake 33:34 — "We solved ML and nobody noticed" 37:12 — Leaving Meta, founding Yutori 43:21 — Web agents: screenshots in, actions out 48:15 — Why the web won't rebuild itself for agents 53:32 — Training Navigator: RL on live websites 1:01:04 — Who pays for the web when agents browse? 1:09:17 — What Yutori means, closing thoughts Music "Kid Kodi" - Blue Dot Sessions - via Free Music Archive - CC BY-NC 4.0.About The Information Bottleneck is hosted by Ravid Shwartz-Ziv and Allen Roush, featuring in-depth conversations with leading AI researchers about the ideas shaping the future of machine learning...

How to Turn Research Into Billion-Dollar Companies, with Ion Stoica

49 min

Ion Stoica has done what almost no academic ever does — repeatedly turned university research into billion-dollar companies. He co-founded Databricks (now valued at over $100 billion), Anyscale, Arena AI and Conviva, while his Berkeley lab produced the open source projects the entire AI industry runs on: Ray, vLLM, and SGLang. In this episode, we ask him how it's actually done. His answer is surprisingly unromantic: solve a problem people already care about, build an artifact good enough that they adopt it, and pay attention to the moment users start asking "who maintains this after the students graduate?" - that's when a project becomes a company. He's also insistent that the credit belongs to his students. From there, the conversation goes deep into what he's watching now: why the AI stack has become an order of magnitude more complex than the Hadoop/Spark era, why maximizing GPU utilization is "the name of the game" for any enterprise, and why coding agents will struggle with distributed systems long after they've mastered web apps. He shares a memorable reward-hacking story — a load balancer that maximized throughput by dropping requests — explains why the gap between open and closed models sits at about six months, and closes with his case for regulating AI by outcomes, not capabilities. Timeline 00:00 — Introduction: welcoming Ion Stoica01:21 — The playbook: how research projects become companies05:22 — Will vLLM and SGLang stay open source?07:47 — The real bottleneck in the AI stack: complexity, not just hardware14:31 — Should algorithms follow infrastructure, or the other way around?16:13 — Can AI coding tools write distributed systems and GPU kernels?21:09 — Verifiers, harnesses, and the limits of outsourcing understanding25:41 — Reward hacking: the load balancer that dropped requests25:58 — How should enterprises consume GPUs? Utilization as the name of the game30:23 — GPU scarcity: will the compute crunch ever end?35:27 — Hyper-optimization and the risk o...

Kaggle Grandmasters, Agent Skills, and Why Everyone Is Overfitting with Jean-Francois Puget (NVIDIA)

59 min

Jean-Francois Puget is a Director and Distinguished Engineer at NVIDIA, where he leads the Kaggle Grandmasters team, and he's ranked third on Kaggle's all-time list. We caught him on the day NVIDIA announced Nemotron Ultra and its new agent skills repo. We talk about what skills actually are, why they beat MCP tools on context cost, and how NVIDIA built an evaluation pipeline to separate skills that help from skills that don't. From there we talk about the thing JFP cares about most: evaluation. He explains why most LLM benchmarks reward overfitting, how his team discovered O3 could pick the right files to fix SWE-bench issues without reading them, and why the only benchmarks he trusts are the ones where you commit before you see the score, which is exactly how Kaggle works. He predicts a "bloodbath" for the wave of competitors letting coding agents chase leaderboard scores with no notion of validation. We also get into what coding agents are actually good for ("a mix of a genius and a dumb person"), the multi-agent system at NVIDIA that built a working PyTorch clone that runs 10x slower than the real thing, his unfiltered take on frontier lab PR and the Mythos release, whether AI is a bubble, and the story of how his team won ARC-AGI with a 4-billion-parameter model at 20 cents a task, including jumping from third to first in the final hours of a seven-month competition. Timeline 00:00 — Intro01:05 — NVIDIA's announcements: Nemotron Ultra and the agent skills repo07:21 — Skills vs MCP tools, and progressive disclosure10:24 — Agents that write their own skills: a new form of learning13:33 — When overfitting is fine (and when it isn't)15:47 — Why most LLM benchmarks reward overfitting17:06 — The SWE-bench contamination story: O3 picks files without reading them19:45 — How LLMs changed Kaggle, and the coming "bloodbath"25:40 — What makes a good data scientist: evaluation and one-bit experiments28:56 — Running Codex at scale: the top token consumers at NVIDIA29:37 — D...

AI Agents and The Golden Age of Asking Questions with Dimitris Papailiopoulos (MSR/UW-Madison)

1 hr 13 min

In this episode, we talked with Dimitris Papailiopoulos, researcher at Microsoft Research's AI Frontiers lab and professor at the University of Wisconsin, about doing research in the age of agents. Dimitris told us about the Sunday morning that changed how he works: he handed Claude Code and Codex a question he'd been sitting on for years, went about his day, and came back to an answer. After a few days of dread about what's left for humans, he landed somewhere more optimistic, calling this the golden age of asking questions. We talked about his "smallest transformer that can add" leaderboard, a symbolic GSM8K solver built from if-else statements, and what happened when he put two Claude Code instances in the same file system and told them to do something cool (one pair invented a communication protocol, the other played Battleship). We also got into diversity and slop in agent-generated ideas, why agents get stubborn after a million tokens, harness overfitting on Terminal-Bench, continual learning and world models, whether agents need vision, and where information theory actually helps in AI and where it's a katana used to make coffee. Timeline 00:00 Intro 01:45 How agents changed the way Dimitris does research 04:30 A Sunday morning with Claude Code, Codex, and GSM8K 07:15 The dread, then the golden age of asking questions 08:20 Taste and verification, and how we train students now 09:53 Will models make human verification obsolete? 11:30 The smallest transformer that can add 10-digit numbers 13:40 Humans as initializers for gradient descent in idea space 15:32 Allen on diversity, slop profiles, and high temperature research 21:44 When Claudes meet: Battleship, invented protocols, and a grokking paper 25:53 Single agent vs multi-agent under fixed compute 30:28 Auto-research benchmarks and what agents actually accelerate 35:14 Inside the symbolic GSM8K solver (with a live progress check) 40:04 Idea overfitting and why agents refuse to change course 44:00 Learning...

Why All Models Learn the Same Thing with Phillip Isola (MIT)

1 hr 11 min

Phillip Isola, professor at MIT, joins us to talk about representation learning: what makes a representation good, why different models seem to converge on similar representations, and whether pre-training is really over. We discuss the platonic representation hypothesis and its limits, why clustering structure matters more than global geometry, and Phillip's new neural thickets paper arguing that post-training is easier than people think because pre-trained weights already sit near solutions to downstream tasks. Phillip also explains why he thinks LLMs are already world models, why he's betting on RNNs making a comeback, and why his most exciting current direction is artificial life: putting LLM agents in open environments with no fixed task and studying them like new organisms. Timeline: 00:00 Intro song 00:13 Intro 01:05 What is representation learning and why it matters 04:09 What makes a representation good: minimality and sufficiency 10:03 How cross entropy and contrastive learning shape representations 14:35 Dimensionality reduction and why dimension isn't the right complexity measure 16:35 Compression and geometric clustering during training 19:27 The platonic representation hypothesis and what actually converges 22:53 Local neighborhoods vs global structure: the Aristotelian follow-up 24:33 When convergence is strong: truth vs the space of possibility 28:09 Is there true similarity in the world? The Bouba-Kiki effect 30:56 World models vs autoregressive LLMs 32:14 Diffusion LLMs as a special case of autoregressive models 33:42 What architectures win in five years: the case for RNNs 36:11 Grad student descent, or do we actually have principles? 40:51 Feathers and wings: what to take from biology 43:17 How close are we to brain-like models? Marr's three levels 47:01 Are better models becoming less human-like? 49:38 Is pre-training all you need? The neural thickets paper 54:18 LoRA, low rank fine-tuning, and why post-training is easier than we thought 56:01 R...

AI for Science with Qichao Hu (Molecular Universe / SES AI)

1 hr 0 min

Most AI-for-science companies are selling shovels. Qichao Hu wants the gold. In this episode, we talk with Qichao, the founder and CEO of Molecular Universe, the AI-for-science platform that grew out of SES AI, a high-energy-density battery developer he's run for fourteen years. His core distinction is that companies from the AI world build tools, such as foundation models that predict properties, while companies from the science world care about the final product, such as the new battery or material that actually ships. Molecular Universe sits firmly on the science side, and the difference shows up everywhere from what they publish to what they refuse to. We get into the actual workflow of materials discovery and where AI compresses it. A single trial in a traditional lab can take a year with maybe a 40% success rate; the goal is to run a thousand candidates in parallel and turn that year into a week. Qichao walks through improving low-temperature fast-charging for EV batteries:  from hypothesis generation through molecule-, material-, and device-level property prediction, down to autonomous labs that synthesize and test the top candidates without a human touching a pipette. The hardest problem, it turns out, isn't predicting molecular properties or measuring device performance, but it's the black box connecting the two. In batteries, that's the solid-electrolyte interface, which the field has been hand-waving about since the seventies. And the thing standing in the way of cracking it isn't a clever training trick but data: companies sitting on twenty years of records are finding it too messy, incomplete, and poorly labeled to train on, and are having to start collecting from scratch with new protocols and robots. Timeline 00:13 — Intro and welcome;01:19 — Shovel vs. gold05:18 — Why the world's smartest scientist doesn't automatically give you a better battery07:25 — The discovery workflow09:37 — Exploration vs. exploitation11:54 — Safety and filtering: screening n...

Infrastructure for AI at Scale - With Benny Chen (Fireworks AI)

1 hr 5 min

We talk a lot on this show about RL, agents, and the move between pre-training and post-training, but not enough about the layer everything actually runs on. Benny Chen, co-founder of Fireworks AI, one of the largest inference platforms around, walks us through what it takes to serve models at scale: sourcing GPUs, writing the kernels, the runtime, and the routing layer that lets a customer hit one endpoint and forget the rest. We talk why the real bottleneck is power, not chips, and why that favors Nvidia and Google. Why MoE keeps winning even when dense models look better on paper and why he'd rather run fungible capacity at 95% than specialized chips at 60%. We also talk about quantization limits, where RL efficiency has to go next, and his case that AI is still under-hyped. We also get into cross-region training, sparse autoencoders and why interpretability hasn't taken off in open source, whether open models can close the gap, and a frank read on Anthropic's go-to-market. Timeline 00:00 — Intro: the part of AI nobody talks about01:20 — What "infrastructure for AI" actually means: the layers, from GPUs up to routing02:59 — Why not just buy your own GPUs and do it yourself?05:17 — The scale Fireworks runs at06:35 — Hardware inflation, GPU costs, and the real risk hiding in commit duration10:14 — Nvidia vs AMD vs TPUs, and why power is the bottleneck11:57 — Mixing GPU types and generations; fungibility vs. specialization14:22 — Once you have the GPUs, what's the next layer to build?17:04 — Dense vs. MoE, and why the hardware picks the winner21:07 — Quantization: is FP4 the floor? TurboQuant and INT vs. FP24:28 — How tied are the algorithms to the hardware?25:12 — DeepSeek, DeepGEMM, and next-token prediction as reconstruction loss28:50 — Why RL is still wildly inefficient compared to pre-training30:08 — Speculative decoding, AI-generated kernels, and auto-research34:00 — The AGI question: why text gets automated but vision may stay expensive37:07 — Hype check: wh...

Broken Peer Review, AI, and Worms — with Oded Rechavi

1 hr 18 min

Oded Rechavi is a biologist at Tel Aviv University and the co-founder of QED, a company building AI to review scientific work. He's also spent years studying worms. We start with what's wrong with peer review and grant funding: why it takes years to publish, why reviewers are often your own competitors, and why the whole thing is locked to an economic model that rewards publishing more papers, not better ones. Oded explains why he doesn't call QED "peer review" at all, and what it would take to actually validate science instead of just stamping it. Then we get into the biology. C. elegans has exactly 959 cells, every one of them named, and a fully mapped brain. Oded's lab studies how a worm's experiences get passed to its offspring through RNA rather than DNA — meaning what happens to a worm in its lifetime can change its descendants. We also talk about using ancient DNA to reassemble the Dead Sea Scrolls, what AI can and can't do for biology, and why he wants to build an "Ironman suit" for researchers rather than replace them. 00:00 Intro 01:35 Why scientific publishing is broken 04:02 Years to publish, and what it costs science 07:20 Bad reviewers, conflicts of interest, and the money 10:47 Why preprints don't fix it 15:37 How AI conferences handle review 22:07 Conferences vs. journals — does slow review help? 25:22 Building QED: review, not peer review 30:02 Tracking a paper from idea to submission 33:11 What writing a grant actually involves 35:00 The ERC reviewer crisis 37:06 Tailoring feedback to your field 41:48 Switching to biology 44:30 Every cell has a name: inside C. elegans 46:28 Inheritance without DNA 48:16 What the worm "thinks" changes its offspring 51:58 Reassembling the Dead Sea Scrolls with ancient DNA 56:07 Psychedelics and worms 58:36 Can AI run the research itself? 1:04:49 Automation vs. validation 1:07:12 The origin of life 1:08:49 Why people reject AI-written work 1:16:18 Will humans still have a role? 1:17:39 Wrap-up Music: "Kid Kodi" - Blue...

Will AI Take Our Jobs? With Alex Imas (Google/University of Chicago)

1 hr 29 min

Will AI take our jobs? We put the question to Alex Imas, the new Director of AGI Economics at Google DeepMind and a professor at Chicago Booth, whose entire job now is studying how frontier AI reshapes the economy. His short answer: probably some of them, but the popular story is mostly wrong about which jobs and how fast. Alex makes the case that a job is a bundle of tasks, not a single thing AI either does or doesn't do, and that the number of people who should actually care about is how much consumer demand responds to falling prices. Get that wrong and you predict mass layoffs. Get it right and you sometimes predict more hiring. We get into why the automation panic is two centuries old, why he thinks blue-collar work is in more danger than white-collar, and why the people already winning are the ones adopting AI fastest. We also cover the AGI versus ASI distinction and why it changes everything for the economy, what happens when there's no moat and open models stay six to eight months behind, the three-tier pricing future he sees coming after the 2026 compute crunch, and what any of this means if you're deciding whether to send your kids to college. The episode was recorded before Alex joined GoogleTimestamps 00:00 Meeting Alex Imas 00:44 Will AI take our jobs? 03:35 Is this an AI question or an economics question? 06:18 The economy is already behind the AI we have 07:43 Why AI adoption is K-shaped 12:51 Was Andrew Yang right? 13:45 The automation panic is 200 years old 16:46 Dario's six-month claim, and why we don't see it yet 17:22 A job is not a task 22:38 The three numbers that actually predict the labor market 22:42 The chess engine analogy and the centaur phase 25:45 Recursive self-improvement and the hamburger problem 30:06 Should AI labs be the ones answering alignment questions? 31:17 The "invisible hand wave" and why nobody wants fully autonomous AI 33:27 AGI vs ASI, and why the difference is everything 35:28 Commodities vs relational goods 41:14 Star...

Why AI Benchmarks Are Lying to You - with Wenhu Chen (Meta/University of Waterloo)

1 hr 19 min

In this episode, we sit down with Wenhu Chen, research scientist at Meta MSL, assistant professor at the University of Waterloo, and the person behind MMLU-Pro and MMMU. If you've read a frontier model release in the last two years, you've seen his benchmarks. That makes him one of the best people to answer the question everyone dances around: when a model jumps from 40% to 90% on your benchmark, how much of that is real? In this episode, we dig into why benchmarks have become the loss function of the entire field - design a bad one, and thousands of brilliant researchers will spend months hill-climbing in the wrong direction. Wenhu is surprisingly candid about the limits of his own creations: contamination is everywhere, saturation turns frontier benchmarks into unit tests, and popular alternatives, such as LM Arena, mostly measure tone and length rather than capability. His answer is to evaluate models where they've never been: private codebases, hospital data, and the messy, live internet. We also talk about ClawBench, his new benchmark that deploys agents to over 140 real production websites to do things people actually want done, such, such as ordering food, booking tickets, and applying for jobs. The best model in the world completes about a third of these tasks. We unpack why: bot detection, models that refuse to click "pay," agents that give up the moment an environment doesn't match their training, and harnesses that can swing results by 20% without changing the model at all. Along the way, we cover the overlooked science of evaluating pre-training, data flywheels, and synthetic environments for agent training, and whether RL teaches models to reason or just surfaces what's already there. We close with Wenhu's predictions: exploration and adaptability will improve rapidly, but security will become the field's hardest problem as agents gain real permissions in the real world. Timestamps 00:00 – Intro 00:55 – What good evaluation means, and how it's changed...

Jürgen Schmidhuber - Part 2: JEPA, the Road to AGI, and Who Really Invented Modern AI

1 hr 29 min

In the second half of our conversation with Jürgen Schmidhuber, we focus on the key ideas he's pursued since the early 1990s and discuss why he believes these concepts are only now being rediscovered. We start with JEPA. Jürgen argues that the method LeCun named in 2022 is the same family he published in 1992 as Predictability Maximization. From there he traces the adversarial lineage back further still, to his 1990 world-model paper and 1991 Predictability Minimization  -  the curiosity-driven minimax games he sees as the real origins of GANs. We also talk about why these ideas took thirty years to land, why today's trillion-dollar data-center buildout is driven by AGI fear, and why he thinks Apple may come out ahead. The back half turns to what he sees as the real frontier: physical AI. Today's systems are superhuman behind the screen but helpless at a leaky pipe, and until a robot can use human tools, there's no AGI. He discusses self-replicating, self-improving machines as "a new kind of life," reframes continual learning and test-time training as ideas from his 1991 fast-weight work, and detours through Solomonoff's universal prior, Hutter's AIXI, and the Gödel machine. We close on the subject Jürgen is famous for: scientific credit. He makes his case for rigorous attribution, casts himself as a "speaker for the dead" championing forgotten pioneers like Ivakhnenko, and reflects candidly on whether the fights are personal. Timeline 00:30 — What JEPA is, and the 1992 Predictability Maximization story 04:54 — Implementing PMAX: autoencoders, Siamese networks, Infomax 09:10 — Predictability Minimization, factorial codes, and the roots of GANs 16:00 — Why it took 30 years: the economics of compute 20:52 — Data, the web, and 1990 as the origin point 23:09 — Hardware inflation, the trillion-dollar buildout, and the coming crash 34:05 — Physical AI: the plumber problem and self-replicating machines 41:14 — Which 90s ideas are being scaled right now 45:26 —...

Jürgen Schmidhuber - World Models, RL, and the Year that changed AI (Part 1)

1 hr 37 min

In this episode, we host Jürgen Schmidhuber - the man, the legend, one of the godfathers of modern AI. His lab worked out many ideas behind today’s systems (LSTM, world models, artificial curiosity, Transformer variants, and even GAN-style setups) decades before they became fashionable, and he’s just as well known for making sure people remember who did what first. This is the first of two conversations with him. We go back to his lab in the early 90s and ask how one small group came up with so many of the ideas that are now being scaled to a thousand billion dollars, back when compute was ten million times more expensive. A lot of the episode comes down to one distinction he keeps making: prediction vs. decision-making. His take is that LLMs are very good prediction machines that imitate the web, but that’s only half the problem. To actually act in the world, you need a controller that uses a world model to plan. He talks about his 1990 work on world models and artificial curiosity, where the controller gets rewarded for running experiments that improve its own model (an adversarial setup years before GANs), why planning millisecond by millisecond doesn’t scale, and why you need sub-goals instead. We also talk about compression as the core of understanding, from falling apples to Kepler to Einstein, and why we still don’t have a robot that can do what a plumber does, even though the AI behind the screen keeps getting better. Then the conversation moves to credit assignment: how “to Schmidhuber” became a verb, what he thinks is broken about the award system, and a long exchange on PMAX vs. JEPA. He ends on the real origins of deep learning and a prediction about self-replicating machines in space. Timeline 00:00  Intro 00:55  1991 in Munich, and why that lab mattered 02:38  "I'm not very smart"  and why compute getting 10× cheaper every 5 years changed everything 04:25  Chess as an AI proxy 08:27  Artificial curiosity in the 90s vs. today's RL exploration 09:10  Wh...

AI for Science and the Thermodynamics of Generative AI - with Max Welling (UvA, CuspAI)

1 hr 13 min

In this episode, we sit with Max Welling, Professor of Machine Learning at the University of Amsterdam, co-founder and CTO of CuspAI, and a foundational figure behind variational autoencoders (VAEs), equivariant networks, and Bayesian deep learning. We talk about AI for science, the physics underneath generative models, and what's still missing on the road to real intelligence. Max starts with what impresses him and what worries him about the LLM era, then makes the case that the next leaps will come from physical AI and from science itself. We dig into how machine learning actually works in the lab, world models and whether priors like geometry and symmetry should be built in or simply learned, and whether transformers will still rule a decade from now. At the end, we talk about CuspAI's climate mission, AI risk and regulation, Max’s new book, and where neuroscience might inspire the next wave of ML. Timeline 00:00 — Intro00:47 — Are we happy with the LLM era?03:14 — Embodiment and physical AI08:05 — Does "AGI" even matter as a term?11:34 — Verifiers, RL, and why math/coding are tractable13:17 — What actually shifted to make materials discovery work14:42 — From molecules to biology and wet labs16:26 — Working with real labs: timescales, friction, and the "Mira" agent20:29 — Balancing simulators vs. experiments: the exploration–exploitation trade-off23:44 — Active learning for experimental design24:23 — Why active learning hasn't been central to LLMs25:24 — A general loop for ML-for-science across domains27:10 — Foundation models for chemistry: a "mother ship" plus a zoo of fine-tuned models30:04 — Quantum mechanics, interpretation, and AI as a creative theorist31:54 — World models and Yann LeCun's view; priors vs. learning34:57 — Should world knowledge be explicit? (responding to Stefano Ermon)36:41 — Vision: equivariance vs. transformers, and the role of optimization40:32 — Best model for molecular properties in 10 years? Will transformers survive?43:16 — CuspAI'...

After Math Falls, What's Next? with Julia Kempe (NYU/Meta)

1 hr 14 min

Julia Kempe on Why Math Will Fall Next, Superhuman Provers, and the Return of the Renaissance Researcher In this episode, we sit down with Julia Kempe, a Professor at NYU's Center for Data Science and researcher at Meta FAIR's Foundations of Reasoning team,  for a wide-ranging conversation on the future of AI research. We dig into why verifiable domains like mathematics may be on track to "fall" the way Go did. With formal verification through Lean and the Mathlib infrastructure, LLM agents can now generate and check proofs at scale, and Julia makes the case that a new industry of automated mathematical discovery is closer than most mathematicians believe. We explore why Erdős problems are already falling, what's still missing for harder fields like analysis and physics, and how synthetic data, curation, and verification fit together. From there we get into the energy and scaling limits of frontier models, the case for academic research that big labs can't pursue, how to advise PhD students when Claude can already do their first-year work, the rise of AI safety and security as research priorities, and Julia's optimistic argument that AI tools are bringing back the Renaissance generalist  -  the researcher who can finally work fluently across math, biology, and beyond. Timeline 00:00 — Introductions01:00 — Defining reasoning and verifiable domains04:00 — Lean, Mathlib, and the formalization of mathematics10:00 — Constructive proofs, Erdős problems, and the new wave of "AI mathematicians"14:00 — Will math be "solved"? Art, photography, and the changing nature of creative work18:00 — Why physics is harder than math22:00 — Moravec's paradox, evolution, and why robotics lags behind language27:00 — The Renaissance is back: generalist researchers in the age of AI29:00 — Advising students: math, programming, and what core education still matters32:00 — Teaching and assessment when GPT can do the homework35:00 — Anti-AI backlash, energy costs, and the security threat40:00 —...

Intelligence in an Open World - with Mengye Ren (NYU)

59 min

We talk with Mengye Ren, Assistant Professor at NYU's Center for Data Science, about what intelligence actually means once you step outside a benchmark, and why scaling a single centralized model isn't the whole story. We get into why intelligence has to be defined in open environments, not closed ones, and what that means for how we measure progress. We push on the creativity question: today's models sample bottom-up from a softmax or a Gaussian, with no internal loop of consideration, and as Mengye puts it, we haven't understood creativity yet and we're already prepared to hand it over. We also talk about what's missing for the next paradigm: continual learning, memory, embodied grounding, and smaller models that actually accumulate experience instead of re-deriving everything from scratch each call. Along the way, we get into JEPA and latent variables, biology as inspiration vs. blueprint, why frontier labs don't lean on explicit latents, the limits of synthetic data and world models, agent-to-agent communication, model uncertainty and forecasting, and whether ML education still matters when AI writes the experiments. A grounded, contrarian conversation about where AI research should be looking next, beyond benchmarks, beyond scale. Timeline00:00 — Intro and welcome 01:24 — What is intelligence? Defining it relative to objectives and open environments 04:19 — Is intelligence really the path to human flourishing, or is it productivity? 04:57 — Safety, scalable oversight, and whether stronger models help or hurt 06:09 — What does "alignment" actually mean? 07:18 — Centralized vs. decentralized models: objectivity vs. personal meaning 08:50 — Hinton vs. LeCun: where Mengye stands on AI risk 10:29 — Bottom-up vs. top-down architectures and feedback loops 21:28 — Biology and AI: inspiration, not blueprint 24:14 — Biological plausibility, spiking nets, and where the analogy breaks 25:39 — JEPA, Mamba, and architectures beyond the transformer 27:31 — Language as a speci...

Language, Cognition, and the Limits of LLMs - with Tal Linzen (NYU/Google)

1 hr 23 min

We host Tal Linzen, Associate Professor at NYU and Research Scientist at Google, for a conversation on the intersection of cognitive science and large language models. We discussed why children can learn language from around 100 million words while LLMs need trillions, and the surprising finding that as models get better at predicting the next word, they become worse models of how humans actually process language. Tal walked us through how his lab uses eye-tracking and reading-time data to compare model behavior to human behavior, and what that reveals about prediction, working memory, and the limits of current architectures. We also got into nature versus nurture and how inductive biases can be instilled by pre-training on synthetic languages, world models and whether transformers actually use the geometric structure they encode, the BabyLM challenge and data-efficient language learning, and what mechanistic interpretability can offer cognitive science beyond just fixing model bugs. The conversation closed on academia versus industry, the role of PhDs in the current AI moment, and how AI coding tools are changing the way Tal teaches and evaluates students at NYU. Timeline 00:13 — Intro and what cognitive science means02:16 — Using computational simulations to understand how humans learn language05:26 — How children learn language vs. how LLMs are pre-trained07:53 — Why mainstream LLMs are not good models of humans 10:07 — Comparing humans and models with eye-tracking and reading behavior13:52 — Sensory modalities, smell, and how much you can learn from language alone16:03 — Animal cognition and decoding animal communication17:00 — Nature vs. nurture, inductive biases, and what transformers can and can't learn21:21 — Instilling inductive biases through synthetic languages 27:34 — The bouba/kiki effect and cross-linguistic sound symbolism28:33 — Latent causal structure in language and whether models discover it31:13 — Does knowing linguistics help build better model...

The Principles of Diffusion Models - with Jesse Lai (Sony AI)

55 min

We host Chieh-Hsin (Jesse) Lai, Staff Research Scientist at Sony AI and visiting professor at National Yang Ming Chiao Tung University, Taiwan, for a conversation about diffusion models, the technology behind tools like Stable Diffusion, and most of the AI image and video generators you've seen in the last few years. Jesse recently co-authored The Principles of Diffusion Models with Stefano Ermon, and the book is quickly becoming a go-to reference in the field. We start with what a generative model actually is, and what it means to "generate" an image or a sound. Jesse explains the core idea behind diffusion in plain terms. You start with pure noise, and a neural network gradually cleans it up, step by step, until a realistic image emerges. From there, we talk about why diffusion has come to dominate so much of generative AI. Because the model builds an image gradually, you can guide it along the way, nudging the output toward what you actually want, refining details, or combining it with other controls. We also discuss the common critique that diffusion is slow and how the field has largely addressed it through new techniques. We zoom out to the bigger picture, too. Jesse shares his view on world models and whether diffusion is the right foundation for them. We talk about what makes a generative model genuinely good versus just good at gaming benchmarks, and why evaluating creativity and realism is so much harder than scoring a multiple-choice test. Timeline 00:12 — Intro and welcoming Jesse 00:47 — Why Jesse wrote the book, and who it's for 03:29 — The three families of diffusion models, and why they're really one idea 05:14 — What makes a good generative model 07:39 — How do you even measure if a generated image is good 08:59 — Why diffusion beats autoregressive models for images 10:33 — Is diffusion still slow? How fast generation got fast 11:12 — A simple intuition for what a "score" is 14:12 — How the different flavors of diffusion connect under the hood 14:4...

Inside xAI, and the Bet on AI Math - with Christian Szegedy (Math Inc)

1 hr 12 min

We talked with Christian Szegedy, co-inventor of Inception and Batch Normalization, founding scientist at xAI, now at Math Inc, about what it takes to build a frontier lab, and why he left xAI to work on formal mathematics. Christian thinks Lean and auto-formalization are the missing piece for trustworthy AI: a machine-checkable layer underneath all reasoning, where proofs are guaranteed correct without anyone having to read them. We got into his bet with François Chollet that AI will hit superhuman mathematician level by 2026, and what that actually unlocks beyond math itself: verified software instead of vibe-coded apps that break when you refactor, AI systems you can actually trust because their reasoning is checkable, and a path to handling protein folding, chemistry, and parts of biology with real guarantees instead of hand-waving. Christian also walked us through how Math Inc's Gauss system pulled off a proof in two weeks that human experts had estimated would take another year. We also covered xAI's first 12-person year, why Christian no longer buys the original batch normalization story, why he's sure transformers won't be the dominant architecture in five years, what mathematicians do in a world of cheap proofs, and his take on whether humanity will handle AI well. He distrusts humanity more than he distrusts AI. Timeline00:12 — Intros: Christian's background (Inception, Batch Norm, xAI, Math Inc) 01:29 — Building a frontier lab from scratch: the first 12 people at xAI 04:15 — Hiring for proven track records when 200K GPUs are at stake 06:07 — Elon's "dependency graph" and balancing long-term vision with investor demos 07:28 — Gauss formalizes the strong prime number theorem in 2 weeks 12:25 — What "formalization" actually means (and why it's not what most people think) 14:39 — Why Lean gives 100% certainty and why that matters for RL 15:26 — ProofBridge and joint embeddings across mathematical subfields 18:07 — Does math formalization transfer to coding an...

Reasoning Models and Planning - with Rao Kambhampati (Arizona State)

1 hr 11 min

We sat down with Rao Kambhampati, a Professor of CS at Arizona State University and former President of AAAI, to talk about reasoning models: what they are, when they work, and when they break. Rao has been working on planning and decision-making since long before deep learning, which makes him one of the most grounded voices on what today's reasoning systems actually do. We start with definitions of what reasoning is, why planning is the hard subset of it, and what changed when systems like o1 and DeepSeek R1 moved the verifier from inference into post-training. From there we get into where these models generalize, where they don't, and why benchmarks can be misleading about both. A big chunk of the conversation is on chain-of-thought: what intermediate tokens are actually doing, why they help the model more than they help the reader, and what outcome-based RL does to whatever semantic content was there to begin with. We also cover world models and why Rao thinks the video-only framing is the wrong bet, the difference between agentic safety and existential risk, and what the planning community figured out decades ago that the LLM community keeps rediscovering. Timeline(00:12) Intros(01:32) Defining "reasoning" and the System 1 / System 2 framing(04:12) Blocksworld vs Sokoban, and non-ergodicity(06:42) Pre-o1: PlanBench and "LLMs are zero-shot X" papers(07:42) LLM-Modulo and moving the verifier into post-training(10:12) Is RL post-training reasoning, or case-based retrieval?(13:12) τ-Bench and benchmarks that avoid action interactions(14:12) OOD generalization and what we don't know about post-training data(19:02) Does it matter how they work if they answer the questions we care about?(21:27) Architecture lotteries and why no one tries different designs(23:42) Intermediate tokens and the "reduce thinking effort" cottage industry(26:12) The 30×30 maze experiment(27:42) Sokoban, NetHack, and Mystery Blocksworld(34:58) Stop Anthropomorphizing Intermediate Tokens — the...

What Actually Matters in AI? - with Zhuang Liu (Princeton)

1 hr 9 min

In this episode, we hosted Zhuang Liu, Assistant Professor at Princeton and former researcher at Meta, for a conversation about what actually matters in modern AI and what turns out to be a historical accident. Zhuang is behind some of the most important papers in recent years (with more than 100k citations): ConvNeXt (showing ConvNets can match Transformers if you get the details right), Transformers Without Normalization (replacing LayerNorm with dynamic tanh), ImageBind, Eyes Wide Shut on CLIP's blind spots, the dataset bias work showing that even our biggest "diverse" datasets are still distinguishable from each other, and more. We got into whether architecture research is even worth doing anymore, what "good data" actually means, why vision is the natural bridge across modalities but language drove the adoption wave, whether we need per-lab RL environments or better continual learning, whether LLMs have world models (and for which tasks you'd need one), why LLM outputs carry fingerprints that survive paraphrasing, and where coding agents like Claude Code fit into research workflows today and where they still fall short. Timeline 00:13 — Intro 01:15 — ConvNeXt and whether architecture still matters 06:35 — What actually drove the jump from GPT-1 to  GPT-3 08:24 — Setting the bar for architecture papers today 11:14 — Dataset bias: why "diverse" datasets still aren't 22:52 — What good data actually looks like 26:49 — ImageBind and vision as the bridge across modalities 29:09 — Why language drove the adoption wave, not vision 32:24 — Eyes Wide Shut: CLIP's blind spots 34:57 — RL environments, continual learning, and memory as the real bottleneck 43:06 — Are inductive biases just historical accidents? 44:30 — Do LLMs have world models? 48:15 — Which tasks actually need a vision world model 50:14 — Idiosyncrasy in LLMs: pre-training vs post-training fingerprints 53:39 — The future of pre-training, mid-training, and post-training 57:57 — Claude Code, Codex, and codin...

The Future of Coding Agents with Sasha Rush (Cursor/Cornell)

1 hr 24 min

We talked with Sasha Rush, researcher at Cursor and professor at Cornell, about what it actually feels like to we in the heart of the AI revolution and build coding agents right now. Sasha shared how these systems are changing day-to-day work and how it feels to develop these systems. A big part of the conversation was about why coding has become such a powerful setting for these tools. We discussed what makes code different from other domains, why agents seem to work especially well there, and how much of today’s progress comes not just from better models, but from better ways of using them. Sasha also gave an inside look at how Cursor thinks about training coding models, long-running agents, context limits, bug finding, and the balance between autonomy and human oversight. We also talked about the broader shift happening in software engineering. Are developers moving to a higher level of abstraction? Is this just a phase where we “babysit” models, or the beginning of a deeper change in how software gets built? Sasha had a very thoughtful perspective here, including what he’s seeing from students, researchers, and engineers who are growing up native to these tools. More broadly, this episode is about what it means to do serious technical work in a moment when the tools are changing incredibly fast. Sasha brought both optimism and skepticism to the discussion, and that made this a really grounded conversation about where coding agents are today, what they are already surprisingly good at, and where all of this might be going next. Timeline 00:00 Intro and Sasha joins us 01:11 What “coding agents” actually mean 02:34 Why coding became the breakout use case 08:56 Long-running agents and autonomous workflows 15:08 How these tools are changing the work of engineers 17:15 Are people just babysitting models right now? 22:11 How Cursor builds its coding models 26:29 Rewards, training, and what makes agents work 34:53 Memory, continual learning, and agent communication 38:...

The Hidden Engine of Vision with Peyman Milanfar (Google)

1 hr 24 min

How Denoising Secretly Powers Everything in AI Peyman Milanfar is a Distinguished Scientist at Google, leading its Computational Imaging team. He's a member of the National Academy of Engineering, an IEEE Fellow, and one of the key people behind the Pixel camera pipeline. Before Google, he was a professor at UC Santa Cruz for 15 years and helped build the imaging pipeline for Google Glass at Google X. Over 35,000 citations. Peyman makes a provocative case that denoising, long dismissed as a boring cleanup task, is actually one of the most fundamental operations in modern ML, on par with SGD and backprop. Knowing how to remove noise from a signal basically means you have a map of the manifold that signals live on, and that insight connects everything from classical inverse problems to diffusion models. We go from early patch-based denoisers to his 2010 "Is Denoising Dead?" paper, and then to the question that redirected his research: if denoising is nearly solved, what else can denoisers do? That led to Regularization by Denoising (RED), which, if you unroll it, looks a lot like a diffusion process, years before diffusion models existed. We also cover how his team shipped a one-step diffusion model on the Pixel phone for 100x ProRes Zoom, the perception-distortion-authenticity tradeoff in generative imaging, and a new paper on why diffusion models don't actually need noise conditioning. The conversation wraps with a debate on why language has dominated the AI spotlight while vision lags, and Peyman's argument that visual intelligence, grounded in physics and robotics, is coming next. Timeline 0:00 Intro and Peyman's background 1:22 Why denoising matters more than you think Sensor diversity and Tesla's vision-only bet 15:04 BM3D and why it was secretly an MMSE estimator 17:02 "Is Denoising Dead?" then what else can denoisers do? 18:07 Plug-and-play methods and Regularization by Denoising (RED) 26:18 Denoising, manifolds, and the compression connection 28:12 Energy-b...

Reinventing AI From Scratch with Yaroslav Bulatov

57 min

Yaroslav Bulatov helped build the AI era from the inside, as one of the earliest researchers at both OpenAI and Google Brain. Now he wants to tear it all down and start over. Modern deep learning, he argues, is up to 100x more wasteful than it needs to be  -  a Frankenstein of hacks designed for the wrong hardware. With a power wall approaching in two years, Yaroslav is leading an open effort to reinvent AI from scratch: no backprop, no legacy assumptions, just the benefit of hindsight and AI agents that compress decades of research into months. Along the way, we dig into why AGI is a "religious question," how a sales guy with no ML background became one of his most productive contributors, and why the Muon optimizer, one of the biggest recent breakthroughs, could only have been discovered by a non-expert. Timeline 00:12 — Introduction and Yaroslav's background at OpenAI and Google Brain 01:16 — Why deep learning isn't such a good idea 02:03 — The three definitions of AGI: religious, financial, and vibes-based 07:52 — The SAI framework: do we need the term AGI at all? 10:58 — What matters more than AGI: efficiency and refactoring the AI stack 13:28 — Jevons paradox and the coming energy wall 14:49 — The recipe: replaying 70 years of AI with hindsight 17:23 — Memory, energy, and gradient checkpointing 18:34 — Why you can't just optimize the current stack (the recurrent laryngeal nerve analogy) 21:05 — What a redesigned AI might look like: hierarchical message passing 22:31 — Can a small team replicate decades of research? 24:23 — Why non-experts outperform domain specialists 27:42 — The GPT-2 benchmark: what success looks like 29:01 — Ian Goodfellow, Theano, and the origins of TensorFlow 30:12 — The Muon optimizer origin story and beating Google on ImageNet 36:16 — AI coding agents for software engineering and research 40:12 — 10-year outlook and the voice-first workflow 42:23 — Why start with text over multimodality 45:13 — Are AI labs like SSI on the right track? 4...

Why Healthcare Is AI's Hardest and Most Important Problem with Kyunghyun Cho (NYU)

1 hr 18 min

We talk with Kyunghyun Cho, who is a Professor of Health Statistics and a Professor of Computer Science and Data Science at New York University, and a former Executive Director at Genentech, about why healthcare might be the most important and most difficult domain for AI to transform. Kyunghyun shares his vision for a future where patients own their own medical records, proposes a provocative idea for running continuous society-level clinical trials by having doctors "toss a coin" between plausible diagnoses, and explains why drug discovery's stage-wise pipeline has hit a wall that only end-to-end AI thinking can break through. We also get into GLP-1 drugs and why they're more mysterious than people realize, the brutal economics of antibiotic research, how language models trained across scientific literature and clinical data could compress 50 years of drug development into five, and what Kyunghyun would do with $10 billion (spoiler: buy a hospital network in the Midwest). We wrap up with a great discussion on the rise of professor-founded "neo-labs," why academia got spoiled during the deep learning boom, and an encouraging message for PhD students who feel lost right now. Timeline: (00:00) Intro and welcome (01:25) Why healthcare is uniquely hard (04:46) Who owns your medical records? — The case for patient-controlled data and tapping your phone at the doctor's office (06:43) Centralized vs. decentralized healthcare — comparing Israel, Korea, and the US (13:19) Why most existing health data isn't as useful as we think — selection bias and the lack of randomization (16:53) The "toss a coin" proposal — continuous clinical trials through automated randomization, and the surprising connection to LLM sampling. (23:07) Drug discovery's broken pipeline — why stage-wise optimization is failing, and we need end-to-end thinking (28:30) Why the current system is already failing society — wearables, preventive care, and the case for urgency (31:13) Allen's personal healthca...

See all 80 episodes

–

0 ratings from members

5
4
3
2
1

Rate this podcast

Log in to rate and review the shows you listen to.

No reviews yet — be the first member to rate "The Information Bottleneck".

This podcast, its content and its artwork are not owned by, affiliated with, or endorsed by Castquire.

Keep listening

Discover more podcasts like this.

Up next

  1. Nothing queued. Use + Queue on any episode to line it up.

0:00 0:00