Minds and machines

The debate over the ‘brain-as-a-computer’ metaphor often conflates two distinct questions: whether the brain’s functions are computable, and whether computation alone is sufficient to produce a conscious mind.

1. The Beast Machine: Biological Naturalism

Cognitive neuroscientist Anil Seth rejects the classical computationalist view of the mind as substrate-independent software. Drawing on cybernetics and predictive processing, Seth’s “Beast Machine” framework argues that:

• Cognition serves homeostasis: The brain did not evolve for abstract symbol manipulation or disembodied logic; it evolved to regulate the physiological integrity of a living organism (allostasis).
• Simulation is not instantiation: A computer simulation of a weather system does not get wet, and simulating the molecular and neural dynamics of a brain does not necessarily generate subjective experience (qualia).
• Substrate matters: Information processing in biological wetware is continuous, metabolic, and deeply intertwined with the chemical and physical constraints of living cells.

2. The Physical Church-Turing Thesis: Functionalism in Practice

Demis Hassabis approaches the brain from a computational and physical perspective:

• Turing computability: Unless the brain relies on unproven, non-computable physics, every neural mechanism—from synaptic plasticity to perceptual inference—falls within the scope of the Church-Turing thesis.
• Distributed computation over GOFAI: Modern AI does not rely on rigid, classical symbol manipulation. Deep neural networks perform non-linear transformations across continuous vector spaces, operating much closer to biological neural dynamics than classical Turing tape caricatures suggest.
• Functional equivalence: If an artificial neural network replicates the input-output mappings and representational structures of human cognitive functions, there is no empirical reason to treat biological tissue as uniquely privileged for intelligence.

3. The Core Divergence

The disagreement between Seth and Hassabis highlights a fundamental philosophical split:

Dimension	Biological Embodiment (Seth)	Computational Functionalism (Hassabis)
Primary Driver	Homeostatic self-preservation	Algorithmic optimization & learning
Substrate	Essential (living wetware)	Fungible (hardware-agnostic)
Phenomenology	Inseparable from physiological embodiment	Emergent from complex information processing

Critics of computationalism often target classical symbolic AI, creating a strawman that overlooks modern connectionism. However, the fundamental tension remains: while Hassabis focuses on whether cognitive processes can be computed, Seth questions whether computational simulation alone can bridge the explanatory gap to conscious experience.