liveBook · Manning

Brain model

As explained in Machines that Think

The construction of formal or computational models representing aspects of brain structure, function, or behavior.

A brain model represents selected aspects of biological cognition or neural organization. Brain modelling is a family of approaches, not a single route to human-like intelligence: a model may test an explanation of biological behavior, inspire an engineered system, or probe which features of intelligence depend on biological structure. section 5.3 section 5.3.1

The goal determines how a model should be judged. A model of neural processes is assessed against biological observations; an engineering model may instead emphasize efficiency, robustness, or adaptability. Success on one criterion does not necessarily establish success on the others. section 5.3.1

Models also differ in what they preserve and simplify:

  • Biologically detailed models retain fine-grained neural mechanisms, supporting comparison with neural data. Their predefined structure can restrict the behaviors they explore. section 5.3.2
  • Abstract computational models retain broader organizational principles while setting aside much biological detail. They offer flexibility and scale, but a weaker basis for biological explanation. section 5.3.2

Brain modelling can inform designs that use feedback, tolerate partial loss, and adapt over time. Redundancy and population-level coordination suggest ways to absorb disruption; continual learning and adaptive control take inspiration from learning alongside perception and action. These are design insights, not a blueprint for a thinking machine. section 5.3.5

Redundancy and pruning

The persistence of function after significant pruning in neural networks illustrates redundancy and distributed organization rather than strict reliance on localized components. It does not by itself establish that artificial and biological systems work in the same way. section 5.5

Biological fidelity shapes what a model can explain, but it is not a measure of progress along a single path to intelligence. Similar behavior in humans and artificial systems also need not arise from comparable mechanisms. section 5.3.2 section 5.5

How Machines that Think covers it

How deep it goes
Compares detailed and abstract models through their explanatory and flexibility trade-offs, without providing implementation procedures.
Examples
Uses neural wiring maps and neural network pruning to illustrate modelling emphases and distributed organization.
What you can do after
Evaluate brain models against their goals and distinguish biological explanation from engineering performance.
How it connects
Connects brain modelling to distributed representations, continual learning, adaptive control, and robustness.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machines that Think. The section links point to the parts of the book it draws on.