Machine Learning Street Talk (MLST)

Intelligence is collective, not artificial — Prof. Michael I. Jordan (UC Berkeley / Inria)

The AI Models Smart Enough to Know They're Cheating — Beth Barnes & David Rein [METR]

AI evaluation methodologies currently struggle to capture the true capabilities and risks of rapidly advancing models. Traditional benchmarks often suffer from distributional leakage and over-reliance on headline accuracy, failing to measure generalization or robustness. The Time Horizon Graph, developed by METR, addre...

When AI Discovers The Next Transformer - Robert Lange (Sakana)

Evolutionary algorithms combined with Large Language Models (LLMs) offer a transformative path for accelerating scientific discovery through autonomous, sample-efficient exploration. Robert Lange, a founding researcher at Sakana AI, details the "Shinka Evolve" framework, which utilizes LLM-driven program mutation and c...

"Vibe Coding is a Slot Machine" - Jeremy Howard

Evolution "Doesn't Need" Mutation - Blaise Agüera y Arcas

Life functions as embodied computation, where the emergence of complexity arises from symbiogenesis rather than simple Darwinian mutation. By modeling life as a self-replicating, Turing-complete system, the "BFF" experiment reveals that random, non-living components undergo a phase transition into complex, functional r...

VAEs Are Energy-Based Models? [Dr. Jeff Beck]

Abstraction & Idealization: AI's Plato Problem [Mazviita Chirimuuta]

The podcast explores the philosophy of science, particularly focusing on abstraction, idealization, and the limitations of mechanistic explanations in understanding the brain and cognition. Mazviita Chirimuuta, author of "The Brain Abstracted," discusses how scientific models often simplify or misrepresent reality due ...

Why Every Brain Metaphor in History Has Been Wrong [SPECIAL EDITION]

We Invented Momentum Because Math is Hard [Dr. Jeff Beck]

Your Brain is Running a Simulation Right Now [Max Bennett]

The neocortex functions as a generative model that enables "learning by imagining," allowing organisms to simulate future outcomes and engage in model-based reinforcement learning. This evolutionary breakthrough, supported by structures like the basal ganglia, facilitates flexible adaptation by pruning search spaces th...

The 3 Laws of Knowledge [César Hidalgo]

The discussion centers on understanding knowledge, its growth, diffusion, and value, as explored in Cesar Hidalgo's book, "The Infinite Alphabet and the Loss of Knowledge." Hidalgo argues for a scientific study of knowledge, governed by laws affecting its growth, diffusion, and value estimation, emphasizing its non-fun...

"I Desperately Want To Live In The Matrix" - Dr. Mike Israetel

Making deep learning perform real algorithms with Category Theory (Andrew Dudzik, Petar Velichkovich, Taco Cohen, Bruno Gavranović, Paul Lessard)

The podcast explores categorical deep learning as a potential unifying framework for neural networks, addressing the current lack of theoretical foundations in deep learning architectures. It highlights how category theory can bridge the gap between constraints and implementation in deep learning, offering a universal ...

Are AI Benchmarks Telling The Full Story? [SPONSORED] (Andrew Gordon and Nora Petrova - Prolific)

The Mathematical Foundations of Intelligence [Professor Yi Ma]

In this interview, Professor Yi Ma discusses his book, "Learning Deep Representations of Data Distributions," and his mathematical theory of intelligence based on parsimony and self-consistency. He clarifies common misunderstandings about intelligence, differentiating between compression and abstraction, and memorizati...

Pedro Domingos: Tensor Logic Unifies AI Paradigms

The podcast explores Tensor Logic, a new language for AI, with Pedro Domingos, a computer science professor at the University of Washington. Domingos argues that Tensor Logic unifies symbolic AI, deep learning, kernel machines, and graphical models by deeply merging tensor algebra and logic programming. The discussion ...

"I Co-Invented the Transformer. Now I'm Replacing It." & Continuous Thought Machines - Llion Jones and Luke Darlow [Sakana AI]

The podcast features Llion Jones, one of the inventors of the Transformer model, and Luke Darlow, a research scientist at Sakana AI, discussing the current state and future directions of AI research. Jones expresses concern about the field being oversaturated with Transformer-based research, advocating for more explora...

Why Humans Are Still Powering AI [Sponsored]

The podcast features a discussion about the importance of human data and expertise in the field of artificial intelligence. The speakers highlight the often-overlooked role of human input in AI development, from labeling data to evaluating model performance. They discuss Prolific, a human data infrastructure company th...

The Universal Hierarchy of Life - Prof. Chris Kempes [SFI]

This podcast features an interview with Professor Chris Kempes from the Santa Fe Institute, who discusses his interdisciplinary research background spanning physics, biophysics, ecology, and astrobiology. The conversation explores the reconciliation of different scientific cultures—variance, exactitude, and coarse-grai...

Google Researcher Shows Life "Emerges From Code" - Blaise Agüera y Arcas

In this interview, Blaise Agüera y Arcas discusses his new book, "What is Intelligence?" and his work at Google as CTO of Technology and Society, including his research group, Paradigms of Intelligence (PAI). He delves into the computational nature of life, drawing parallels between DNA and computer programs, and explo...