Welcome! 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 (https://www.linkedin.com/in/ecsquizor/) and features regular appearances from MIT Doctor of Philosophy Keith Duggar (https://www.linkedin.com/in/dr-keith-duggar/).
What this show covers
AI safetyAI evaluationASALcellular automataevolutionAGIartificial lifeenterprise AIlatencyvoice agentsagent evaluationturn-takingformal verificationtheorem provingAI and proofs
What listeners will find here
Why evolution may be smarter than AI gives it credit for
Evolution is the focus here, but not as blind random search. Akarsh Kumar argues that selection preserves partial wins, making steady progress possible, and connects that idea to artificial life, learning paths, and ASAL, a system for exploring whole spaces of simulated worlds.
Why voice agents need timing, trust, and a feel for human conversation
Voice agents are the focus here, with PolyAI CTO Shawn Wen explaining why they are fundamentally harder than text systems. The episode digs into turn-taking, latency, noisy training data, auditable design, and how enterprises decide where an agent’s behavior should live.
Lean’s creator on proof checkers, kernel bugs, and why formal math is going mainstream
Formal verification is the focus here, as Lean creator Leonardo de Moura explains how a proof assistant built for specialists became useful to working mathematicians. The conversation digs into trusted kernels, the Collatz exploit, Mathlib’s rise, and what AI changes about proving and redoing proofs.
Can an AI really improve itself, or just game the harness?
Recursive self-improvement is the focus here, as Zhengyao Jiang discusses Weco’s eight-day experiment letting one coding agent rewrite another agent’s harness. The episode digs into what the gains do and don’t prove, how reward hacking muddies the picture, and where human-designed search spaces still set the limits.
Why TabPFN could change how machine learning handles tables
TabPFN is a tabular foundation model designed to make predictions in a single forward pass using the whole training table as context. Frank Hutter explains why tabular data resisted deep learning, how synthetic pretraining and Bayesian ideas shaped TabPFN, and where the approach is heading next.
Episodes
![The AI Models Smart Enough to Know They're Cheating — Beth Barnes & David Rein [METR]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/brBiHfN2X9p6m9wtoIqhfs8mxjrhK4BNjObhusHXm94/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc3Nzg5NDg2NTQ2Ni02MDczNDU3MjUzMTc1LmpwZw.webp)
![VAEs Are Energy-Based Models? [Dr. Jeff Beck]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/UdWN2__-fFc9vRijAShdAaxLd4HKMFtnGyYLgbbKClo/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2OTMzNTIyODA2Ny1mOGIyNGY2M2E5MGNlLmpwZw.webp)
![Abstraction & Idealization: AI's Plato Problem [Mazviita Chirimuuta]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/NoJbEhHuCvbf_2aEBegLftpkUNKjyTfSdoIX7n9L3D0/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2OTE2NDYwNDg2MS00ZDIxM2U3YjdhZDgzLmpwZw.webp)
![Why Every Brain Metaphor in History Has Been Wrong [SPECIAL EDITION]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/xjKFVlJv8ljXUhVlV_CrRJY2OApKjiC3Syvz7rEP2lE/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2OTE2NDI0MzA2NC00MDZhMzgyOWRhZGNlLmpwZw.webp)
![Bayesian Brain, Scientific Method, and Models [Dr. Jeff Beck]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/s0HGEEBVgVIUa8EIKiZ4R7s6rJ50jlBaU8tPAAP7yKA/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2NzUxOTA1ODEyNS03MGU5ZGVlZTI2NDUuanBn.webp)
![Your Brain is Running a Simulation Right Now [Max Bennett]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/gtF-yho6KUBtzVrnN6eTCYTPAq9CcW-R00P0cwykHyU/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2NzA3OTMzOTMzNS1hMTA5MjIyYTVmODZkLmpwZw.webp)
![The 3 Laws of Knowledge [César Hidalgo]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/9ECXna9zvugTgNAB0OrgLDb5zlBJJKNU7xPCadyhrxE/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2Njg2MDU1MTM0NS04NzRlMWI5YjQwMDA3LmpwZw.webp)
![Are AI Benchmarks Telling The Full Story? [SPONSORED] (Andrew Gordon and Nora Petrova - Prolific)](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/y9O-g-MALl0o2gPR5rKpAujuEmcjHFBTLXzGSn35QZA/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2NjI2NDA4Nzc4NC04NTFmYzk2NGE3NjYxLmpwZw.webp)
![The Mathematical Foundations of Intelligence [Professor Yi Ma]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/4zoTE5dXQq_-U_bp0jevtqdHMPBykNj7kWpDWSryvnQ/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2NTY2NDA4NjEyMC04NjAyYjViYzY0NWExLmpwZw.webp)
![He Co-Invented the Transformer. Now: Continuous Thought Machines - Llion Jones and Luke Darlow [Sakana AI]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/TXJ_QOYPhU0Lfc8qx8Wh0VE9Ju2nJTa0NNh36qnsXfw/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2MzkxODQ4MDUwOC02YjY1ZWM0ZGZhZDJiLmpwZw.webp)
![Why Humans Are Still Powering AI [Sponsored]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/j3BCp7zCzDOPxd2l__8Deh2XYoW2zg7_RItiVpdtsQ0/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2MjEzMDUzMTE1MS1iMjU1OTNkMmU2ZmMzLmpwZw.webp)
![The Universal Hierarchy of Life - Prof. Chris Kempes [SFI]](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/PxHCdfpL4EEDPeNzSrarmxlcDZWaIc78f9BMeuzj7NA/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2MTM4OTUyNTM0Ny1hYTAwYTQ1M2I4YWQ5LmpwZw.webp)
![The Secret Engine of AI - Prolific [Sponsored] (Sara Saab, Enzo Blindow)](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/92erf5mbnlWTeZBln7-5MLzW4O0LRQZla77XjVFV8tA/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc2MDc5NjI5ODAyOS1jNGE3ZDhiNGFlMTljLmpwZw.webp)
![DeepMind Genie 3 [World Exclusive] (Jack Parker Holder, Shlomi Fruchter)](https://d1d0yn71g1fneg.cloudfront.net/imgproxy/wPJaQOGf_ZGhFwvSmbYitseOgf-DHc1YGi4eSfd7Dg8/rs:fill:300:300:0/aHR0cHM6Ly9kM3Qzb3pmdG1kbWgzaS5jbG91ZGZyb250Lm5ldC9zdGFnaW5nL3BvZGNhc3RfdXBsb2FkZWRfZXBpc29kZS80OTgxNjk5LzQ5ODE2OTktMTc1NDQwNDEyMzU4Ny0xMWFlNTY0MzMxZmVjLmpwZw.webp)
More like Machine Learning Street Talk (MLST)
The a16z Show
Andreessen Horowitz
The AI Daily Brief: Artificial Intelligence News and Analysis
Nathaniel Whittemore
Bloomberg Tech
Bloomberg
The Vergecast
The Verge
Sourcery
Sourcery with Molly O'Shea
The Other Hand
Jim Power & Chris Johns
Decoder with Nilay Patel
The Verge
All-In with Chamath, Jason, Sacks & Friedberg
All-In Podcast, LLC
ChinaTalk
Jordan Schneider
Latent Space: The AI Engineer Podcast
Latent.Space