What are the key takeaways from this Machine Learning Street Talk episode?

Key insights include: **Perception as Inference:** The brain does not directly perceive sensory input but constructs models of reality and tests them against evidence. Visual illusions demonstrate this - you cannot see a duck and rabbit simultaneously because the brain renders one simulation at a time, explaining why we cannot unsee illusions once perceived.; **Mental Simulation in Rats:** Hippocampal place cells in rats activate not just in current locations but along potential future paths during decision-making pauses. Researchers observed rats imagining foregone choices in the orbital frontal cortex after making regrettable decisions, demonstrating model-based reinforcement learning in simple mammals through measurable neural activity.; **Agranular Prefrontal Cortex:** Layer four of the neocortex atrophies in mammalian frontal cortex during development because this region primarily generates intentions rather than processes sensory input. This architectural difference supports active inference theory - the brain fits behavior to its model of goals rather than constantly updating goals based on sensory feedback.

What did Max Bennett discuss on Machine Learning Street Talk?

Max Bennett explains how the brain evolved through five breakthroughs, from basic steering to mental simulation, revealing how the neocortex functions as a generative model that enables planning, imagination, and social cognition through 600 million years of evolution. Key topics include: **Perception as Inference:** The brain does not directly perceive sensory input but constructs models of reality and tests them against evidence. Visual illusions demonstrate this - you cannot see a duck and rabbit simultaneously because the brain renders one simulation at a time, explaining why we cannot unsee illusions once perceived.; **Mental Simulation in Rats:** Hippocampal place cells in rats activate not just in current locations but along potential future paths during decision-making pauses. Researchers observed rats imagining foregone choices in the orbital frontal cortex after making regrettable decisions, demonstrating model-based reinforcement learning in simple mammals through measurable neural activity..

How long is this episode of Machine Learning Street Talk?

This episode is 197 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

Machine Learning Street Talk

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

December 30, 2025

197 min episode · 2 min read

Max Bennett

Episode

197 min

Read time

2 min

Topics

Product & Tech Trends, Psychology & Behavior, Science & Discovery

AI-Generated Summary

Published Dec 30, 2025

Key Takeaways

✓Perception as Inference: The brain does not directly perceive sensory input but constructs models of reality and tests them against evidence. Visual illusions demonstrate this - you cannot see a duck and rabbit simultaneously because the brain renders one simulation at a time, explaining why we cannot unsee illusions once perceived.
✓Mental Simulation in Rats: Hippocampal place cells in rats activate not just in current locations but along potential future paths during decision-making pauses. Researchers observed rats imagining foregone choices in the orbital frontal cortex after making regrettable decisions, demonstrating model-based reinforcement learning in simple mammals through measurable neural activity.
✓Agranular Prefrontal Cortex: Layer four of the neocortex atrophies in mammalian frontal cortex during development because this region primarily generates intentions rather than processes sensory input. This architectural difference supports active inference theory - the brain fits behavior to its model of goals rather than constantly updating goals based on sensory feedback.
✓Primate Social Intelligence: Neocortex size in primates correlates directly with social group size, not with other mammals. Chimpanzees demonstrate theory of mind by distinguishing intentional from accidental actions, choosing experimenters who can see them, and engaging in multi-level deception - abilities emerging from uniquely primate brain regions like granular prefrontal cortex.
✓Self-Supervision Principle: Transformers and the neocortex both achieve generalization through self-supervised learning - predicting masked or future inputs without explicit labels. This shared principle suggests that generative models trained on prediction naturally develop rich internal representations, though transformers lack the autonomous neuron-level agency present in biological systems.

What It Covers

Max Bennett explains how the brain evolved through five breakthroughs, from basic steering to mental simulation, revealing how the neocortex functions as a generative model that enables planning, imagination, and social cognition through 600 million years of evolution.

Key Questions Answered

•Perception as Inference: The brain does not directly perceive sensory input but constructs models of reality and tests them against evidence. Visual illusions demonstrate this - you cannot see a duck and rabbit simultaneously because the brain renders one simulation at a time, explaining why we cannot unsee illusions once perceived.
•Mental Simulation in Rats: Hippocampal place cells in rats activate not just in current locations but along potential future paths during decision-making pauses. Researchers observed rats imagining foregone choices in the orbital frontal cortex after making regrettable decisions, demonstrating model-based reinforcement learning in simple mammals through measurable neural activity.
•Agranular Prefrontal Cortex: Layer four of the neocortex atrophies in mammalian frontal cortex during development because this region primarily generates intentions rather than processes sensory input. This architectural difference supports active inference theory - the brain fits behavior to its model of goals rather than constantly updating goals based on sensory feedback.
•Primate Social Intelligence: Neocortex size in primates correlates directly with social group size, not with other mammals. Chimpanzees demonstrate theory of mind by distinguishing intentional from accidental actions, choosing experimenters who can see them, and engaging in multi-level deception - abilities emerging from uniquely primate brain regions like granular prefrontal cortex.
•Self-Supervision Principle: Transformers and the neocortex both achieve generalization through self-supervised learning - predicting masked or future inputs without explicit labels. This shared principle suggests that generative models trained on prediction naturally develop rich internal representations, though transformers lack the autonomous neuron-level agency present in biological systems.

Notable Moment

Bennett describes how David Redish recorded rats literally imagining eating food they chose not to take, watching their orbital frontal cortex activate for the foregone treat. This neural evidence of regret demonstrates that even simple mammals engage in counterfactual reasoning about alternative choices they could have made.

Know someone who'd find this useful?

Episode Transcript

Think Verizon the best five g network is expensive? Think again. Bring in your AT and T or T Mobile bill to a Verizon store today, and we'll give you a better deal. Now what's to do with your unwanted bills? Ever seen an origami version of the fist? Jokes aside, Verizon has the most ways to save on phones and plans where you can get a single line with everything you need. So bring in your bill to your local Detroit Verizon store today, and we'll give you a better deal. Rankings based on route metrics whose were partated 01/2025, your results may vary. Must provide a postpaid consumer mobile bill data within the the deal. Additional terms apply. What's really interesting about about this book, Max, is, you know, obviously, I've I've read loads and loads of books in in the space, and there's, you know, people like Hinton and Hawkins and Damasio and Friston and, I mean, god. You know, you you you even like Sutton. And what's interesting is it's a bit like the blind men and the elephant. So they've all got a completely different story to tell, and I think the magic that you have pulled off with this book is somehow you've woven it together into a coherent story. Like, what what do you think about that? Well, first, I'm very appreciative of of the kind words. Yeah. I think I came from a very unique perspective just because I was a complete outsider. And I think, you know, and I didn't come to it with the objective of writing an academic book at all. I came to it from the objective. I was just learning on my own, and I just started building this corpus of notes, because I was so independently curious. And I kinda stumbled on this idea really for myself of how do I make sense of all of these disparate opinions, and really this complete lack of information about how the brain actually works. And, I had my own set of, I think, biases coming from sort of the technology entrepreneurial world where, we tend to think about things as ordered modifications. When you think about product strategies or how to roll things out, we like to think about things as what's step one, then what's step two, what's step three. So I think I did have sort of a cognitive bias to when presented with an incredible amount of complexity to try and make sense of it in a similar type of way. But, yeah, I think as an outsider, I felt very free to sort of explore and cross the boundaries between fields. I mean, I I look at the book as a merging of three fields. One is comparative psychology, so trying to understand what are the different intellectual capacities of, different species, evolutionary neuroscience. So what do we know about the past brains of humans and the ordered set of modifications of how brains …

Get the full transcript (37,334 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all Machine Learning Street Talk transcripts →