Your Brain Doesn't Command Your Body. It Predicts It. [Max Bennett]

2025-12-30 · Source: Machine Learning Street Talk · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Life Sciences & Biology, Social Sciences & Behavioral Studies · Depth: Advanced, extended

Summary

Max Bennett's book synthesizes disparate theories on brain function, comparative psychology, evolutionary neuroscience, and AI to offer a coherent narrative on how the brain works. Bennett, an outsider to academia, built a corpus of notes to reconcile conflicting opinions and a lack of information on brain mechanisms, applying a technology-entrepreneurial bias for ordered modifications. He highlights the challenge of limited data in comparative psychology, citing the lamprey fish's unexamined map-based navigation. The discussion explores the gap between neuroscience theories and AI practice, questioning what AI successes like Transformers reveal about the brain. Bennett argues the neocortex enables mental simulation and model-based reinforcement learning, emphasizing its role in building rich world models for imagining outcomes. He also delves into the concept of perception by inference, where the brain constructs explanations of sensory input, and the unique human capacity for language, which facilitates the sharing of imagined actions and the accumulation of knowledge across generations.

Key takeaway

For AI Researchers and Machine Learning Engineers developing advanced models, recognize that current AI systems, like GPT-4, may exhibit "theory of mind" in puzzle-solving but lack the human brain's data efficiency and mechanistic synergy for generalizing intent in novel, real-world scenarios. Focus on integrating mechanisms for hypothesis testing and active intervention, rather than solely relying on observational data, to build more robust and trustworthy AI agents capable of understanding and aligning with complex human intentions beyond mere statistical prediction.

Key insights

The brain synthesizes diverse information into a coherent model, enabling perception by inference and mental simulation.

Principles

• Perception is an inference, not direct sensory experience.
• The neocortex enables rich mental simulations for planning.
• Language facilitates sharing imagined actions, driving cumulative knowledge.

Method

The "in-group, out-group, and stem-group" conditions infer behavioral abilities in ancestors by comparing descendants, non-descendants, and adaptive value in ecological contexts.

In practice

• Consider brain evolution to understand current brain function.
• Use AI systems to test neuroscience theories for practical grounding.

Topics

• Evolutionary Neuroscience
• Mental Simulation
• Theory of Mind
• Language Evolution
• AI World Models

Best for: AI Researcher, AI Scientist, Machine Learning Engineer

Related on AIssential

• Models of the Mind: How physics, engineering, and mathematics have shaped our understanding of the brain — Mathematical models are crucial for explicit hypothesis testing and understanding the brain's complex, evolved systems.
• Clever Hans Has a Brain Scanner — Brain-encoding models' high prediction scores may reflect shortcut learning, not true brain understanding, due to noise-free synthetic data.
• The Predictive Mind: How Human Brains and AI Speak the Same Language — Human brains and large language models both operate on predictive processing, continuously generating expectations and updating based on new information.
• AI and Theory of Mind: an interview with Nitay Alon — Theory of Mind is critical for navigating deception, but over-mentalization can lead to paranoia in both humans and AI.
• "Imaginative AI" with Mohamed Elhoseiny — AI imagination can enable machines to perceive, create, act, and feel beyond their training data.
• The scary question for me isn't whether AI is conscious ... it's whether we were ever as deep as we assume — The debate over AI consciousness may reveal human selfhood is less profound than commonly assumed.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning Street Talk.