352 | Bing Brunton on Connecting the Connectome to the Body

Episode Summary

Executive Summary: Sean Carroll and Bing Brunton discuss the connectome as a wiring diagram of the nervous system, why cell identity and body context matter beyond connectivity, and how fruit-fly connectome data can be used to model behavior. Brunton explains their work identifying a minimal three-neuron rhythm-generating circuit for fly walking, cautions against overclaiming from simulations alone, and argues embodied, physics-based models may be crucial for neuroscience, rehab, and AI.

Main Topics: What a connectome is—and why the definition is messy (Priority: 5/5): Brunton explains that a connectome is a comprehensive map of neural connectivity, but the field lacks consensus on scale: neuron-by-neuron, cell-type, or brain-region level. She emphasizes that connectivity is only one part of brain function. Cells, cell types, and non-neuronal brain components (Priority: 5/5): The conversation expands beyond neurons to glia and other cell types, highlighting that nearly half of brain cells are non-neuronal and that brain function depends on many interacting cell populations and physical substrates. Why connectivity alone may not be enough (Priority: 5/5): Brunton argues that a connectome omits crucial factors such as neurotransmitter identity, receptor expression, biophysical properties, and body interactions; these omissions explain why some early connectome efforts were scientifically useful but incomplete. Fruit-fly connectome and the search for walking circuits (Priority: 5/5): The team leverages the newly mapped fly ventral nerve cord connectome to model locomotion, focusing on a reduced front-leg system and using simulations to identify a minimal rhythm-generating circuit. Central pattern generators and the three-neuron motif (Priority: 5/5): They discuss rhythmic circuits underlying movement, breathing, and digestion, and Brunton describes a fly walking circuit reduced from about 4,000 neurons to three key neurons: two excitatory and one inhibitory. Digital twins, model validation, and skepticism about overclaiming (Priority: 4/5): Brunton warns that models can match behavior for the wrong reasons if they rely on too much flexible fitting. She contrasts careful biology with exaggerated claims about 'uploading' brains and stresses the need for validation experiments. Broader implications for AI, consciousness, and therapy (Priority: 4/5): The interview closes with reflections on embodied intelligence, the importance of physics/constraints in artificial systems, and how embodied neural models could inform rehabilitation, injury recovery, and future therapeutics.

Key Arguments: A connectome is a wiring diagram, but its value depends on what level of biological detail is included and what questions it is meant to answer. Neurons cannot be understood in isolation: cell identity, neurotransmitters, glia, and the body all shape computation. The fly connectome may be more tractable than the worm connectome because insects have more specialized cell types and less extreme multiplexing. A large connectome can generate useful hypotheses, but simulations must be validated experimentally to avoid getting the right behavior for the wrong reason. Rhythmic behaviors like walking can be modeled as central pattern generator circuits; in the fly front-leg system, a small subcircuit may be sufficient to generate the core rhythm. Embodied, physics-based models are likely necessary for understanding how nervous systems produce behavior, adapt to injury, and recover function. Biology may inspire AI, but artificial systems do not necessarily need biological implementation details; what may matter more is embodiment, constraints, and interaction with an environment.

Data Points: Human brain neurons: ~85 billion - Used to contrast the scale of human connectomes with those of smaller organisms. Fraction of brain cells that are not neurons: About half - Brunton notes that roughly half of brain cells are glia or other non-neuronal cells. C. elegans neurons: ~300 neurons - The nematode worm is the classic fully mapped connectome example. C. elegans total cells: About 1,000 cells - The worm has a much smaller nervous system and body than vertebrates or insects. Fruit fly brain neurons: About 150,000 neurons - The newly mapped fly brain connectome provides a much larger and richer dataset than C. elegans. Fruit fly ventral nerve cord neurons: About 22,000 neurons - Added to the fly brain to form a more complete nervous-system connectome. Reduced simulation size: About 4,000 neurons - The team narrowed the fly connectome to the subset relevant for two front legs before pruning further. Minimal rhythm circuit: 3 neurons - Their pruning study suggested two excitatory neurons (E1, E2) and one inhibitory neuron (I1) are sufficient to generate the basic walking rhythm. Fly body size: About 3 mm long - Sean Carroll describes fruit flies as roughly grain-of-rice sized. Fly nervous system size: About the size of a sesame seed - Used to illustrate why a full connectome is computationally approachable in flies. Long sensory neuron length: About as tall as a human - Brunton cites a human sensory neuron extending from the big toe to the brainstem. Timescale of connectome skepticism: 15–20 years - Brunton says her skepticism about large connectome projects dated back to graduate school. Age of C. elegans connectome: About 30 years ago - The worm’s full connectivity matrix was the first major connectome completed. Evolutionary origin of nervous systems: ~500 million years ago - Brunton notes nervous systems evolved to control bodies, sense the environment, and move.

Pivotal Quotes: "The brain is not in a jar." — Bing Brunton: A central thesis of the interview: brains must be understood as embodied systems interacting with a body and environment. "If it's there, there's probably a pretty good reason it's there or it wouldn't be there." — Bing Brunton: She argues against dismissing neurons or circuits as wasteful and emphasizes evolutionary and functional significance. "We have no examples that we all agree on of agents that are intelligent and conscious, except the ones that are embodied." — Bing Brunton: Brunton’s closing reflection on embodiment, intelligence, and consciousness.

Implications: The work suggests connectomes can generate testable circuit hypotheses, but only when paired with body-aware, physics-based models and experiments. That has implications for neuroscience, rehab, robotics, and cautious interpretation of AI/consciousness claims.