Free Daily Podcast Summary
Machine Learning Street Talk (MLST): Daily Summaries Delivered
by Machine Learning Street Talk (MLST)
We engage in fascinating discussions with pre-eminent figures in the AI field. Our flagship show covers current affairs in AI, cognitive science, neuroscience and philosophy of mind with in-depth analysis. Our approach is unrivalled in terms of scope and rigour – we believe in intellectual diversity in AI, and we touch on all of the main ideas in the field with the hype surgically removed. MLST is run by Tim Scarfe, Ph.D and features regular appearances from MIT Doctor of Philosophy Keith Duggar.
Latest Episodes
The most recent episodes — sign up to get AI-powered summaries of each one.
- 2 weeks ago55 min
Why a Nation Can't Outsource Its Frontier AI - Alistair Pullen (Cosine AI)
This episode is sponsored by Notion. Learn more about Notion's Developer Platform today at https://notion.com/mlstBritain's most capable coding model can't be exported, and that ban is the whole reason Cosine set out to build one from scratch. Alistair Pullen, CEO and co-founder of Cosine, sits down with Tim Scarfe to explain how a frontier system he calls Fable, locked behind US export controls, became the founding case for a UK sovereign model trained on the Isambard supercomputer in Bristol.The bet underneath it is economic. Pullen argues that an inference company, rather than a training-first lab, doesn't need billions to compete: millions, a national compute allocation, and a consortium feedback loop can be enough. From there it gets into the machinery, why open-weight models still trail the frontier on size, active parameters and data, the mixture-of-experts versus dense trade-off and why active params dominate how a model actually feels, and the edge that real coding trajectories confer.The back half is about making agents trustworthy. Pullen makes the case for beating "slop" by rewarding the process instead of the final answer, reframes code review as runtime proof (spin the bug up in a VM and force the agent to actually exploit it), and walks through Swarm, Cosine's system running hundreds of sub-agents in one shot. It ends on why memory is still an unsolved hack, how synthetic graders let you run RL on tasks with no built-in test, and why Pullen reads US export controls as an accidental gift, with a supply-chain sting in the tail.---TIMESTAMPS:00:00:00 The sovereign mandate and the Fable ban00:04:02 Millions vs billions: the inference-company model00:07:19 The consortium feedback loop00:07:40 Why open models lag the frontier00:14:59 MoE vs dense, and why active params matter00:16:29 Trajectories: the process-data advantage00:19:48 Beating slop: reward the process, not the answer00:26:06 Reusable abstractions and the epistemic wall00:29:56 Code review becomes runtime proof00:37:32 Do agentic harnesses still matter?00:40:35 Swarm: orchestrating hundreds of sub-agents00:45:14 Why memory is still unsolved00:48:25 Synthetic data and graders for RL00:53:09 The US export gift and supply-chain risk---REFERENCES:organization:[00:01:15] Cosinehttps://cosine.sh[00:04:14] Mistral AIhttps://mistral.ai[00:05:50] Anthropichttps://www.anthropic.com[00:07:42] Coherehttps://cohere.com[00:08:36] DeepSeekhttps://www.deepseek.comtool:[00:02:52] Isambard-AIhttps://isambard.ac.uk[00:05:56] Colossus (xAI)https://en.wikipedia.org/wiki/Colossus_(supercomputer)[00:07:52] GLM (Z.ai)https://z.ai[00:11:52] NVIDIA B300https://www.nvidia.com/en-us/data-center/dgx-b300/[00:15:37] gpt-oss-120bhttps://huggingface.co/openai/gpt-oss-120b[00:15:52] Devstral 2https://mistral.ai/news/devstral[00:16:01] Llama 70bhttps://www.llama.com[00:17:05] Claude Codehttps://www.anthropic.com/claude-code[00:26:23] ARC-AGI (Francois Chollet)https://arcprize.org[00:40:38] Swarm (Cosine)https://cosine.sh[00:40:50] OpenAI Codexhttps://github.com/openai/codex[00:41:16] Lumen Outpost (Cosine)https://cosine.sh[00:41:18] Kimi K2 (Moonshot)https://huggingface.co/moonshotai/Kimi-K2-Instruct[00:49:55] SWE-benchhttps://www.swebench.com[00:52:40] SystemVeriloghttps://en.wikipedia.org/wiki/SystemVerilogperson:[00:23:40] Andrej Karpathyhttps://karpathy.aipaper:[00:27:10] GRPO (DeepSeekMath)https://arxiv.org/abs/2402.03300[00:27:13] GSPOhttps://arxiv.org/abs/2507.18071Incompressible Knowledge Probes, Bojie Lihttps://arxiv.org/pdf/2604.24827Estimating the Size of Claude Opus 4.5/4.6https://unexcitedneurons.substack.com/p/estimating-the-size-of-claude-opus---ReScript:https://app.rescript.info/session/5852d2b884c4ce4b?share=10b9799160845bb11779f8ac6cd3124f
- 4 weeks ago1h 24m
The Benchmark With No Instructions — ARC-AGI-3 (winning team!)
Tim Scarfe travels to Zurich to sit down with the Tufa Labs ARC-AGI-3 team — founder Benjamin Crouzier, with Jeroen Cottaar, Dries Smit, Stefano Viel and Michal Tesnar — to work out what their leaderboard-topping system does and what the benchmark is really testing.The cut opens on the games: a walkthrough of the Locksmith game, where you read the rules of an unfamiliar world straight from raw frames. ARC-AGI-3 makes ARC interactive and agentic, so the model has to *discover* the goal rather than transduce a static grid. It stays easy for humans and breaks LLMs, and it runs through everything that follows. Dries traces his StochasticGoose preview win — brute force that only searched actions which changed the frame — and why it collapsed once the organisers added action-efficiency scoring and unseen games.Induction and transduction run through the middle of the conversation — how much of an answer is really priors leaking back the moment a model recognises a maze. The abstraction mountain, and Tim's case that LLMs reach the right answer through fractured, entangled representations — performance, not competence. Whether transformers plan at all or just fake it well enough. Why the score really measures action efficiency, not games solved, and why agents lock onto the wrong goal and cannot climb back out.Crouzier closes on the Tufa Labs thesis — a small lab against the giants, the bitter lesson against hand-built harnesses, and safety — and Tim ties it back to Kenneth Stanley, deep constraints, and creativity as competence.Disclosure: Tufa Labs sponsors MLST. ---TIMESTAMPS:00:00:00 Meet the Tufa team and what makes ARC-AGI-3 hard00:02:11 Locksmith game: reading the rules from raw frames00:03:10 Why build an independent research lab00:04:11 StochasticGoose: a preview win, then the hardened games00:07:58 Induction, transduction, and priors inside LLMs00:10:31 Curiosity, world models, and exploring by frame change00:14:32 Understanding debt and losing sight of your own code00:15:53 Requirements-based agents and human-AI co-creativity00:19:22 Why auto-research misses the big picture00:21:54 The abstraction mountain and fractured representations00:27:36 Constraints and making LLMs act as if they understand00:34:51 Human difficulty calibration, esports priors, and emergence00:41:35 Agency, goal acquisition, and two kinds of planning00:47:31 Harnesses, the 36% number, and wrong-goal loops00:52:33 Rewards, goals, and why ARC-AGI-3 resists brute force01:00:46 Would solving ARC-AGI-3 prove AGI?01:07:53 Stripping language away, then priors leak back01:14:06 Representation and whether language is necessary01:18:04 The bitter lesson versus specialised harnesses01:22:20 Capability research, safety, and the software singularity---REFERENCES:organization:[00:02:11] ARC-AGI-3https://arcprize.org/arc-agi/3[00:03:10] Tufa Labshttps://tufalabs.ai/team/[00:04:20] ARC-AGI-3 Preview Agent Competitionhttps://arcprize.org/competitions/arc-agi-3-preview-agentstool:[00:04:55] StochasticGoose ARC-AGI-3 solutionhttps://github.com/DriesSmit/ARC3-solution[00:07:42] ArcGenticahttps://github.com/symbolica-ai/arcgentica[00:07:49] RGB-Agenthttps://github.com/alexisfox7/RGB-Agent[00:14:38] Claude Codehttps://www.anthropic.com/claude-code[01:03:42] Qwen 3.6 27Bhttps://huggingface.co/Qwen/Qwen3.6-27Bpaper:[00:13:03] On the Measure of Intelligencehttps://arxiv.org/abs/1911.01547[00:27:42] DreamCoderhttps://arxiv.org/abs/2006.08381[00:43:55] On the Biology of a Large Language Modelhttps://transformer-circuits.pub/2025/attribution-graphs/biology.html[01:18:46] ImageNet Classification with Deep CNNs (AlexNet)https://proceedings.neurips.cc/paper_files/paper/2012/file/c399862d3b9d6b76c8436e924a68c45b-Paper.pdfother:[01:18:16] The Bitter Lessonhttp://www.incompleteideas.net/IncIdeas/BitterLesson.html---https://app.rescript.info/share/463d7f031349b4b9db428553eed88230
- Jun 28, 20261h 2m
The First Thermodynamic AI Computing Chip - Thomas Ahle
Thomas Ahle wants Normal Computing to be the Lovable for chip design: type your intent, and a swarm of agents carries it from design through optimisation, formalisation and verification to tape-out. To get there, his team at wrote their own open-source Verilog simulator, 580,000 lines in 43 days, because commercial EDA verifiers run about $10,000 per core and there are no decent open-source compilers to build on.That sets up the question Tim keeps pressing: if an agent can produce a chip design, a proof, or a working program, how do you actually know it is correct? Passing 70% of tests is not the same as being right, and a single fabricated bug can cost a company a fortune. They dig into ProgramBench (rebuild a program from its tests, roughly 0% success), the difference between structure and competence, and the "understanding debt" you take on when nobody reads the code.From there: auto-formalisation in Lean and the AlphaProof trick of training on prove-or-disprove; why there is no single true representation of a spec (Petri nets, TLA+, Erik Curiel's "math does not represent"); and thermodynamic computing, where Normal Computing's CN101 chip is built so that its physical noise *is* the computation, settling a stochastic differential equation in hardware to invert a matrix. Plus Bayesian uncertainty, specialisation, the Chomsky hierarchy, AI slop, and whether performance is all that matters.Recorded in Zurich.Disclosure: Normal Computing paid our production and travel costs for this show. We retained full editorial control. They did not see the video before publication, and we did not show it to them or discuss it with them beforehand.---TIMESTAMPS:00:00:00 Meet Thomas Ahle: the Lovable for chip design00:03:41 Why hardware needs formal verification00:06:36 Ten thousand dollars per core and a six-month agent run00:07:40 Rebuilding programs from tests: ProgramBench and zero percent00:12:15 Structure vs competence: can you learn a program from behavior?00:15:27 Continual learning, abstraction, and Claude as an ecosystem00:23:17 Autoformalization and the AlphaProof trick00:29:31 No single true representation: specs, Petri nets and TLA+00:34:43 Thermodynamic computing: when noise is the computation00:37:32 Bayesian uncertainty in the age of token streams00:41:12 Hybrid compute: vibe-coding loops, binaries and Stockfish00:44:44 Co-design, central-AI apps and API pricing00:49:45 Chain of thoughtlessness and the Chomsky hierarchy00:53:40 AI psychosis, slop and the broken social contract00:57:34 Typing it yourself, teamwork and performance vs competence---REFERENCES:person:[00:00:10] Thomas Ahlehttps://thomasahle.comorganization:[00:00:27] Normal Computinghttps://normalcomputing.com/paper:[00:11:21] ProgramBench: Can Language Models Rebuild Programs From Scratch?https://arxiv.org/abs/2605.03546[00:31:55] Autoformalizing Memory Device Specifications with Agentshttps://arxiv.org/abs/2605.00058[00:35:20] Thermo AI and the Fluctuation Frontierhttps://arxiv.org/abs/2302.06584[00:36:40] Thermo Comp System for AI Applicationshttps://arxiv.org/abs/2312.04836[00:37:05] Thermodynamic Linear Algebrahttps://arxiv.org/abs/2308.05660[00:44:50] An efficient probabilistic hardware architecture for diffusion-like modelshttps://arxiv.org/abs/2510.23972tool:other:[00:01:00] Building an Open-Source Verilog Simulator with AI: 580K Lines in 43 Dayshttps://normalcomputing.com/blog/building-an-open-source-verilog-simulator-with-ai-580k-lines-in-43-days[00:02:55] Normal Computing Announces Tape-Out of the World's First Thermodynamic Computing Chip (CN101)https://www.normalcomputing.com/blog/normal-computing-announces-tape-out-of-worlds-first-thermodynamic-computing-chip[00:32:02] DRAMBench: Autoformalizing DRAM Specifications with Timed Petri Netshttps://www.iese.fraunhofer.de/blog/drambench-autoformalizing-dram-specifications/---ReScript: https://app.rescript.info/share/ff9684a112ab37744096adaeb097a263
- Jun 22, 202653 min
He won a Nobel here for AlphaFold. Then he left. - John Jumper
This episode is sponsored by Notion. Learn more about Notion's Developer Platform today at notion.com/mlstProtein folding stalled biology for fifty years. A sequence of amino acids dictates a three-dimensional shape, but reading that shape meant a year and roughly $100,000 of crystallography per structure. Then AlphaFold 2 won CASP14 so decisively the organizers called the problem essentially solved.In this documentary cut, John Jumper, who shared the 2024 Nobel Prize in Chemistry and has since left DeepMind for Anthropic, walks Tim Scarfe through what the system did and, more interestingly, what it did not. The architecture gets a proper dissection: MSAs, the Evoformer, invariant point attention, the FAPE loss, and Jumper's correction of the equivariance story, which ablations valued at roughly 2.5 of 30 GDT points rather than the whole win. He is blunt about the limits. AlphaFold predicts one experiment extraordinarily well; it is not a model of the cell, it does not capture dynamics, and on a given drug target it is "wrong nine times out of ten."From there: the AlphaFold Database of 200M+ predicted structures, AlphaFold 3 and ligands, Isomorphic Labs, and Jumper's quarrel with the bitter lesson, where finite data and human hypotheses still matter. Emmanuel Nji of BioStruct Africa closes the film on what changes when work that took years now takes months, and on training the next thousand structural biologists across Africa.---TIMESTAMPS:00:00:00 Cold open: predicting nature with a button press00:01:03 The protein folding bottleneck and CASP00:04:39 The Nobel, the database, and the move to Anthropic00:07:39 Sponsor and framing: what AlphaFold does not claim00:07:39 Proteins as self-assembling nanomachines00:12:24 From structures to biology: drug discovery and Midnolin00:17:37 The humility of AlphaFold: a narrow predictor00:22:18 Inside the architecture: Evoformer, IPA and FAPE00:30:20 Ruthless empiricism: ablations and 100x in data00:35:20 Predict, control, understand00:40:00 Against the bitter lesson; AlphaFold 3 as diffusion00:45:07 Intelligence, representations and AGI00:49:23 Epilogue: AlphaFold in Africa00:52:16 Closing: the case for hybrid science models---REFERENCES:organization:[00:01:55] Critical Assessment of Structure Prediction (CASP)https://predictioncenter.org/[00:04:39] The Nobel Prize in Chemistry 2024https://www.nobelprize.org/prizes/chemistry/2024/summary/[00:05:18] BioStruct Africahttps://www.biostructafrica.org/[00:18:03] Isomorphic Labshttps://www.isomorphiclabs.com/paper:[00:03:09] AlphaFold Protein Structure Databasehttps://doi.org/10.1093/nar/gkab1061[00:17:25] Accurate structure prediction of biomolecular interactions with AlphaFold 3https://www.nature.com/articles/s41586-024-07487-w[00:22:18] Highly accurate protein structure prediction with AlphaFoldhttps://www.nature.com/articles/s41586-021-03819-2[00:23:10] Midnolin promotes degradation of substrates independent of ubiquitinationhttps://doi.org/10.1126/science.adh5021[00:27:00] Improved protein structure prediction using potentials from deep learninghttps://www.nature.com/articles/s41586-019-1923-7tool:[00:03:09] AlphaFold Protein Structure Database (EBI)https://alphafold.ebi.ac.uk/[00:45:55] AlphaEvolve: a coding agent for designing advanced algorithmshttps://deepmind.google/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/other:[00:39:40] The Bitter Lessonhttp://www.incompleteideas.net/IncIdeas/BitterLesson.html---ReScript: https://app.rescript.info/share/d8cde5c221fb71e2c0f5aafe94f90dfa
- May 31, 20261h 20m
When AI Decides You're a Threat — Brad Carson
Brad Carson was the Army's General Counsel, served two terms in Congress and was Acting Under Secretary of Defense for Personnel and Readiness. He now heads Americans for Responsible Innovation, the AI-policy advocacy group he co-founded. Keith Duggar spends roughly eighty minutes pushing back.SPONSOR:---Cyber Fund built the Monastery to help founders ship products that were impossible a year ago. Applications for Batch 1 are now open.Apply now: https://cyber.fund---Carson's whole case rests on one line: the genie is not out of the bottle. We have pulled dangerous tech back before. Asilomar halted recombinant DNA in 1975, and the West still controls the chips AI runs on. Calling it unstoppable, he says, is the most dangerous idea in the room.Then Keith drags him somewhere darker. A Palantir heat map scores you 0.73 on whether you are a combatant, and a strike follows. The model is wrong some accepted share of the time, and when it is, nobody answers for it. You cannot court-martial a model, and not even the interpretability researchers can say why it picked you.—Note: after recording, we learned that Americans for Responsible Innovation is backed by EA-aligned philanthropy (not sponsored)---TIMESTAMPS:00:00:00 From the Pentagon to AI governance00:04:52 Regulatory capture vs Silicon Valley networks00:07:56 Transparency and the Claude tier changes00:09:40 Tort liability when AI tools cause harm00:13:40 AI is a product, not a person00:16:01 Children, suicide, and the suicide business00:19:59 Opaque neural nets and the law of war00:25:54 Probabilistic targeting and the death of accountability00:28:47 The arms race fallacy: Asilomar and restraint00:34:02 Talking to China: track 2 talks and chip leverage00:39:45 Air power never wins: capital for labour00:43:29 Anthropic vs the Department of War00:51:29 Concentration, open source, and brain drain01:00:18 DeepSeek, Chinese culture, and AI as diplomacy01:12:25 Upskilling Congress and why public trust matters---REFERENCES:organization:[00:02:45] ICRC position on autonomous weaponshttps://www.icrc.org/en/law-and-policy/autonomous-weapons[00:05:22] Americans for Responsible Innovation (ARI)https://ari.us[00:07:20] Andreessen Horowitz (a16z)https://a16z.com/[01:16:05] Office of Technology Assessmenthttps://en.wikipedia.org/wiki/Office_of_Technology_Assessmentother:[00:03:35] Beneficial AGI 2019 Conference (Future of Life Institute, Puerto Rico)https://futureoflife.org/event/beneficial-agi-2019/[00:18:30] Section 230 of the Communications Decency Acthttps://en.wikipedia.org/wiki/Section_230[00:19:59] Lethal Autonomous Weapons (LAWS)https://en.wikipedia.org/wiki/Lethal_autonomous_weapon[00:31:35] Strategic Arms Limitation Talks (SALT)https://en.wikipedia.org/wiki/Strategic_Arms_Limitation_Talks[00:32:28] Asilomar Conference on Recombinant DNA (1975)https://en.wikipedia.org/wiki/Asilomar_Conference_on_Recombinant_DNA[00:39:45] The New Iron Triangle (ARI policy byte)https://ari.us/policy-bytes/the-new-iron-triangle/[00:48:05] Defense Production Acthttps://en.wikipedia.org/wiki/Defense_Production_Actperson:[00:03:35] Anthony Aguirrehttps://en.wikipedia.org/wiki/Anthony_Aguirre[00:06:48] Dean Ball — Hyperdimensionalhttps://www.hyperdimensional.co/[00:23:13] Neel Nanda — mechanistic interpretabilityhttps://www.neelnanda.io/[00:36:02] Jack Clark (Anthropic) on Conversations with Tylerhttps://conversationswithtyler.com/episodes/jack-clark/[00:39:15] Robert Trager — Centre for the Governance of AIhttps://www.governance.ai/team/robert-trager[00:41:55] Giulio Douhethttps://en.wikipedia.org/wiki/Giulio_Douhet[01:15:05] Don Beyer (US Congress)https://en.wikipedia.org/wiki/Don_Beyertool:[00:22:19] Phalanx CIWShttps://en.wikipedia.org/wiki/Phalanx_CIWS---ReScript:https://app.rescript.info/public/share/9405ff35c0215b7cdae6402d41284171https://app.rescript.info/api/public/sessions/0a6c081b8e5fe413/pdf
- May 21, 20261h 17m
Intelligence is collective, not artificial — Prof. Michael I. Jordan (UC Berkeley / Inria)
Michael I. Jordan, described by Science magazine as the most influential computer scientist alive, has never thought of himself as an AI researcher. In this conversation he explains why that distinction matters.SPONSOR:---Cyber Fund built the Monastery to help founders ship products that were impossible a year ago. Applications for Batch 1 are now open.Apply now: https://cyber.fund---Jordan trained as a statistician and cognitive scientist, and his career has been spent building machine learning systems that work in the real world: supply chains, commerce, healthcare, and large economic systems. When the field rebranded itself as AI and then AGI, he did not follow. Instead he argues that the framing is wrong. AI is better understood as a collective economic system than as a race to build a disembodied superintelligence.We talk about why AGI is mostly a PR term, what machine learning achieved before the LLM hype cycle, and why the assistant-on-your-shoulder vision may be less compelling than it sounds. Jordan explains why explanations need to be actionable, not merely mechanistic; why AlphaFold's missing error bars matter; how prediction-powered inference changes the picture; and why drug discovery is an incentive-design problem rather than a pure pattern-matching problem.ERRATA: Science magazine ranked him the most influential computer scientist, not Nature---TIMESTAMPS:00:00:00 Cold open: A demoralizing message to young builders00:02:04 CyberFund sponsor read00:02:50 From symbolic AI to machine learning systems00:05:42 Why AGI is mostly a PR term00:08:48 A collectivist, economic perspective on AI00:11:33 Why LLMs need system design, not hype00:14:50 Predictability beats faux understanding00:17:55 AlphaFold, bias, and prediction-powered inference00:21:48 Stop anthropomorphizing intelligence00:27:44 Drug discovery as an incentive problem00:32:29 The three-layer data market00:38:07 Social knowledge, markets, and culture00:45:39 Creator economics beyond Spotify00:48:30 How science-fiction AI narratives mislead young builders00:51:45 AI should improve humans, not replace them00:56:42 Safety is a property of the whole system00:58:12 Silicon Valley gurus and the cream off the top01:00:47 Game theory, mechanism design, and contracts01:04:39 Conformal prediction, e-values, and anytime inference01:08:11 A new liberal arts triangle for the AI era01:11:30 The Bayesian duck and markets as uncertainty reductionReScript (transcript, PDF, refs etc) - https://app.rescript.info/public/share/fb68f94af29d3745c6cf6125e01328b5---REFERENCES:person:[00:02:50] Michael I. Jordan (homepage)https://people.eecs.berkeley.edu/~jordan/paper:[00:06:01] A Collectivist, Economic Perspective on AIhttps://arxiv.org/abs/2507.06268[00:18:09] AlphaFoldhttps://www.nature.com/articles/s41586-021-03819-2[00:20:36] Prediction-Powered Inferencehttps://arxiv.org/abs/2301.09633[00:33:47] On Three-Layer Data Marketshttps://arxiv.org/abs/2402.09697[01:04:39] Conformal Prediction with Conditional Guaranteeshttps://arxiv.org/abs/2107.07511[01:04:51] A Tutorial on Conformal Predictionhttps://www.jmlr.org/papers/v9/shafer08a.html[01:06:00] E-Values Expand the Scope of Conformal Predictionhttps://arxiv.org/abs/2503.13050[01:08:23] Computational Thinkinghttps://www.cs.cmu.edu/~CompThink/papers/Wing06.pdfother:[00:28:20] How Should the FDA Test?https://rdi.berkeley.edu/events/sbc-assets/pdfs/Summit%20session%20speaker%20slides%20submission%20form-s1-5%20%28File%20responses%29/Slides%20in%20PDF%20%28Please%20name%20the%20submitted%20file%20as%20_firstname_-_lastname_-slides.pdf%29.%20%28File%20responses%29/27-Michael%20Jordan-Session%20V.pdf#page=15[00:28:40] Michael I. Jordan Session V Slides<truncated, see ReScript link or YT VD>
- May 4, 20261h 53m
The AI Models Smart Enough to Know They're Cheating — Beth Barnes & David Rein [METR]
Beth Barnes and David Rein on the one graph that ate the AI timelines discourse, and why the two people who built it are the most careful about how you read it.**SPONSOR**Prolific - Quality data. From real people. For faster breakthroughs.https://www.prolific.com/?utm_source=mlstInterview: https://youtu.be/cnxZZTl1tkk---Beth Barnes and David Rein from METR on the one graph that ate the AI timelines discourse, and why the people who built it are the most careful about how it gets read.Beth founded METR after leaving OpenAI alignment. David is first author on GPQA and co-author on HCAST and the METR Time Horizons paper. Together they built the measurement Daniel Kokotajlo called the single most important piece of evidence on AI timelines: the log-linear line of "how long a task a frontier model can complete at 50% reliability" vs release date.The conversation opens on reward hacking. Current models can articulate in chat why a behaviour is undesired and then execute it anyway as agents. From there: construct validity, Melanie Mitchell's four-problem taxonomy, and the ARC-AGI 1-to-2 collapse as a worked example of adversarially-selected benchmarks regressing once labs target them. Beth's counter: METR deliberately does not adversarially select. David's: models do not have to do the right thing for the right reasons.Methodology, then specification — David's compiler analogy, Beth on four-month tasks as expensive to evaluate rather than unspecifiable. Then the SWE-bench reality check, the METR finding that half of passing PRs would not be merged, and Beth's horses-versus-bank-tellers analogy for the labour market.The close: monitorability, the coin-spinning boat, two-year recursive self-improvement, and Beth's line that "overhyped now" and "big deal later" are not correlated claims.---TIMESTAMPS:00:00:00 Intro00:02:06 Sponsor break: Prolific human-feedback infrastructure00:02:33 Welcome and the scalable oversight motivation00:06:02 Construct validity, benchmark pathologies and the Chollet worry00:15:45 Time Horizons: human time, HCAST tasks and the 50% logistic00:24:50 Is human difficulty really one variable?00:33:05 Agent harness evolution and the inference-compute dividend00:40:00 Scaffolding bells, token budgets and the credit-assignment problem00:44:15 Look at the damn graph: regularisation bug and reliability nuance00:50:00 Why 50%? Reliability, reward hacking and pizza-party transcripts00:55:20 Extrapolation risk and straight lines on graphs00:59:25 Software engineering as a specification acquisition problem01:07:40 Compilers also made ugly code: vibe-coding quality and Claude on METR Slack01:15:15 Strongest defensible claim, Carlini's compiler swarm and AI 202701:23:45 SWE-bench merge rates, the bank-teller analogy and horses01:31:45 Scheming, alignment faking and the mentalistic vocabulary problem01:40:45 Reward hacking, monitorability and chain-of-thought faithfulness01:45:25 Recursive self-improvement, knowledge vs intelligence and closingReScript: https://app.rescript.info/public/share/de3bb40cc02ee39fdf36e2c60366eb4d(PDF, refs, transcript etc)
- Mar 13, 20261h 18m
When AI Discovers The Next Transformer - Robert Lange (Sakana)
Robert Lange, founding researcher at Sakana AI, joins Tim to discuss *Shinka Evolve* — a framework that combines LLMs with evolutionary algorithms to do open-ended program search. The core claim: systems like AlphaEvolve can optimize solutions to fixed problems, but real scientific progress requires co-evolving the problems themselves.GTC is coming, the premier AI conference, great opportunity to learn about AI. NVIDIA and partners will showcase breakthroughs in physical AI, AI factories, agentic AI, and inference, exploring the next wave of AI innovation for developers and researchers. Register for virtual GTC for free, using my link and win NVIDIA DGX Spark (https://nvda.ws/4qQ0LMg)• Why AlphaEvolve gets stuck — it needs a human to hand it the right problem. Shinka tries to invent new problems automatically, drawing on ideas from POET, PowerPlay, and MAP-Elites quality-diversity search.• The *architecture* of Shinka: an archive of programs organized as islands, LLMs used as mutation operators, and a UCB bandit that adaptively selects between frontier models (GPT-5, Sonnet 4.5, Gemini) mid-run. The credit-assignment problem across models turns out to be genuinely hard.• Concrete results — state-of-the-art circle packing with dramatically fewer evaluations, second place in an AtCoder competitive programming challenge, evolved load-balancing loss functions for mixture-of-experts models, and agent scaffolds for AIME math benchmarks.• Are these systems actually thinking outside the box, or are they parasitic on their starting conditions? When LLMs run autonomously, "nothing interesting happens." Robert pushes back with the stepping-stone argument — evolution doesn't need to extrapolate, just recombine usefully.• The AI Scientist question: can automated research pipelines produce real science, or just workshop-level slop that passes surface-level review? Robert is honest that the current version is more co-pilot than autonomous researcher.• Where this lands in 5-20 years — Robert's prediction that scientific research will be fundamentally transformed, and Tim's thought experiment about alien mathematical artifacts that no human could have conceived.Robert Lange: https://roberttlange.com/---TIMESTAMPS:00:00:00 Introduction: Robert Lange, Sakana AI and Shinka Evolve00:04:15 AlphaEvolve's Blind Spot: Co-Evolving Problems with Solutions00:09:05 Unknown Unknowns, POET, and Auto-Curricula for AI Science00:14:20 MAP-Elites and Quality-Diversity: Shinka's Evolutionary Architecture00:28:00 UCB Bandits, Mutations and the Vibe Research Vision00:40:00 Scaling Shinka: Meta-Evolution, Democratisation and the Three-Axis Model00:47:10 Applications, ARC-AGI and the Future of Work00:57:00 The AI Scientist and the Human Co-Pilot: Who Steers the Search?01:06:00 AI Scientist v2, Slop Critique and the Future of Scientific Publishing---REFERENCES:paper:[00:03:30] ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolutionhttps://arxiv.org/abs/2509.19349[00:04:15] AlphaEvolve: A Coding Agent for Scientific and Algorithmic Discoveryhttps://arxiv.org/abs/2506.13131[00:06:30] Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agentshttps://arxiv.org/abs/2505.22954[00:09:05] Paired Open-Ended Trailblazer (POET)https://arxiv.org/abs/1901.01753[00:10:00] PowerPlay: Training an Increasingly General Problem Solver by Continually Searching for the Simplest Still Unsolvable Problemhttps://arxiv.org/abs/1112.5309[00:10:40] Automated Capability Discovery via Foundation Model Self-Explorationhttps://arxiv.org/abs/2502.07577[00:15:30] Illuminating Search Spaces by Mapping Elites (MAP-Elites)https://arxiv.org/abs/1504.04909[00:47:10] Automated Design of Agentic Systems (ADAS)https://arxiv.org/abs/2408.08435<trunc, see ReScript/YT>PDF : https://app.rescript.info/api/sessions/b8a9dcf60623657c/pdf/downloadTranscript: https://app.rescript.info/public/share/SDOD_3oXOcli3zTqcAtR8eibT5U3gam84oo4KRtI-Vk
Get Machine Learning Street Talk (MLST) summaries in your inbox
Free AI-powered daily recaps. Key takeaways, quotes, and mentions — in a 5-minute read.
Get Free Summaries →Free forever for up to 3 podcasts. No credit card required.
You Might Also Like
Listeners also like.
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
Explores machine learning and artificial intelligence through interviews with leading researchers, engineers, and industry experts.
Latent Space: The AI Engineer Podcast
Covers advances in AI engineering, including foundation models, code generation, and AI agents, through interviews with researchers and developers.
"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis
Interviews with AI developers and researchers exploring the transformative impact of artificial intelligence on society and technology.
Artificial Intelligence Masterclass
Examines the impact of advanced artificial intelligence on ethics, society, and the future of human labor.
The AI XR Podcast.
Industry insiders interview top founders and executives on AI, spatial computing, VR/AR, and synthetic media.
80,000 Hours Podcast
Discusses artificial intelligence and global catastrophic risks with experts and researchers.
This Day in AI Podcast
Two friends discuss artificial intelligence, sharing casual insights, personal experiments, and humorous experiences with AI tools and technology.
Me, Myself, and AI
AI leaders from top companies share real-world strategies for turning artificial intelligence into measurable business results.
Unsupervised Learning with Jacob Effron
Conversations with leading AI experts to understand current breakthroughs and future implications for technology and business.
The AI Daily Brief: Artificial Intelligence News and Analysis
A daily analysis of artificial intelligence news, exploring its creative potential, industry impacts, and ethical challenges.
Intelligent Machines (Audio)
Explores the rise of artificial intelligence through conversations with pioneers shaping the future of intelligent machines.
Everyday AI Podcast – An AI and ChatGPT Podcast
Practical AI and ChatGPT tips for professionals to improve productivity and grow their careers.
About Machine Learning Street Talk (MLST)
We engage in fascinating discussions with pre-eminent figures in the AI field. Our flagship show covers current affairs in AI, cognitive science, neuroscience and philosophy of mind with in-depth analysis. Our approach is unrivalled in terms of scope and rigour – we believe in intellectual diversity in AI, and we touch on all of the main ideas in the field with the hype surgically removed. MLST is run by Tim Scarfe, Ph.D and features regular appearances from MIT Doctor of Philosophy Keith Duggar.
Customized Recaps
AI-powered recaps with compact key takeaways, quotes, and insights.
Straight to Your Inbox
Get key takeaways from Machine Learning Street Talk (MLST) in a 5-minute read.
Save Hours Every Week
Stay current on your favorite podcasts without falling behind.
Frequently Asked Questions
What is Podzilla's Machine Learning Street Talk (MLST) daily summary?
It's a free AI-powered email that summarizes new episodes of Machine Learning Street Talk (MLST) as soon as they're published. You get the key takeaways, notable quotes, and links & mentions — all in a quick read.
How does the Machine Learning Street Talk (MLST) podcast summary work?
When a new episode drops, our AI transcribes and analyzes it, then generates a personalized summary tailored to your interests and profession. It's delivered to your inbox every morning.
Is this an official Machine Learning Street Talk (MLST) product?
No. Podzilla is an independent service that summarizes publicly available podcast content. We're not affiliated with or endorsed by Machine Learning Street Talk (MLST).
Can I get summaries of other podcasts too?
Absolutely! The free plan covers up to 3 podcasts. Upgrade to Pro for 15, or Premium for 50. Browse our full catalog at /podcasts.
How often does Machine Learning Street Talk (MLST) release new episodes?
Machine Learning Street Talk (MLST) publishes weekly. Our AI generates a summary within hours of each new episode.
What topics does Machine Learning Street Talk (MLST) cover?
Machine Learning Street Talk (MLST) covers topics including Technology. Our AI identifies the specific themes in each episode and highlights what matters most to you.
Start getting Machine Learning Street Talk (MLST) summaries tomorrow morning.
Free forever for up to 3 podcasts. No credit card required.
Free forever for up to 3 podcasts. No credit card required.