Why Neuroscience Got Everything Backwards
Audio Brief
Show transcript
In this conversation, artificial intelligence researcher Joscha Bach and developmental biologist Michael Levin explore how biological cells function as cognitive agents to form collective intelligence.
There are three key takeaways from this discussion. First, all biological cells possess computational capabilities similar to neurons but operate on different spatial and temporal scales. Second, evolution scales up individual cell agency into collective tissue intelligence using ancient bioelectric toolkits. Third, understanding biological systems requires an engineering stance that focuses on functional software patterns rather than physical hardware.
Every somatic cell functions as a learning agent, capable of processing information and adapting to its environment. While neurons are specialized for high-speed, long-distance communication, non-neuronal cells utilize the same ancient biochemical machinery to coordinate complex behaviors over slower timescales. This shared toolkit allows evolution to scale individual cellular activities into goal-directed, multicellular intelligence.
Adopting an engineering stance allows researchers to look past the biological hardware of cells and pathways. By analyzing the causal software patterns and control structures, scientists can better decode how living systems process information and realize specific behaviors. This interdisciplinary approach bridges the gap between computer science, neuroscience, and developmental biology.
Ultimately, shifting the focus to these shared computational principles reveals that intelligence is not exclusive to brains, but is a fundamental property of cellular networks.
Episode Overview
- This episode features a deep-dive conversation between AI researcher Joscha Bach and developmental biologist Michael Levin, hosted by Curt Jaimungal.
- The discussion explores the intersection of biology, neuroscience, and computer science, focusing on the cognitive capabilities of individual cells and how they form collective intelligence.
- It challenges traditional disciplinary boundaries between neuroscience, developmental biology, and computer science to propose a more unified framework of agency and information processing in living systems.
- The content is highly relevant to anyone interested in cognitive science, artificial intelligence, synthetic biology, and the fundamental nature of intelligence.
Key Concepts
- Cells as Reinforcement Learning Agents: Every biological cell, not just neurons, possesses the capability to process information, send and receive multiple message types conditionally, and learn from its environment. Non-neuronal cells function similarly to neurons but operate over shorter distances and slower timescales.
- The "Telegraphic" Nature of Neurons: Neurons are specialized cells that evolved to communicate over long distances at extremely high speeds. This specialized capability allows complex organisms to coordinate muscle movements and update internal world models rapidly to compete with other animals.
- Scaling Up Intelligence: Evolution utilizes ancient bioelectric and biochemical communication toolkits (dating back to unicellular ancestors) to scale up individual cell agency into collective tissue- and organ-level intelligence.
- The Engineering Stance in Science: Applying a computer science or engineering lens to biological systems shifts the focus from merely describing components to understanding the underlying "software"—the causal patterns and control structures that realize specific goal-directed behaviors.
Quotes
- At 1:04 - "As a result, this means that every cell can, in principle, function like a neuron. It can fulfill the same learning and information processing tasks as a neuron. The only difference... is that they cannot do this over very long distances." - Joscha Bach, explaining the fundamental cognitive equivalence of somatic cells and neurons.
- At 5:39 - "Evolutionarily, they are reusing machinery that has been around for a very long time... Our unicellular ancestors had a lot of the same machinery... and evolution began to reuse this toolkit specifically to scale up the computation." - Michael Levin, describing how complex nervous systems are built on ancient cellular communication mechanisms.
- At 9:30 - "What I miss specifically in a lot of the way in which neuroscience is done is what you call the 'engineering stance'... You don't really care about what language it is written in; what you care about is what causal pattern is realized and how this can be realized." - Joscha Bach, advocating for looking at biological systems through a functional, computational lens.
Takeaways
- Apply the "engineering stance" when evaluating complex biological or cognitive systems by looking past the physical hardware (cells, pathways) to analyze the functional, causal software patterns that direct behavior.
- Look for commonalities across scientific disciplines rather than respecting strict departmental boundaries; computational models from neuroscience can be successfully applied to developmental biology and vice-versa.
- Avoid the pitfall of assuming intelligence requires a brain; recognize that cellular networks, tissues, and even plant systems process information and exhibit goal-directed behavior on different timescales.
