Brains and computers work in fundamentally different ways

The argument

This argument rests on a contrast in architecture. A computer receives input, encodes it, and writes it to discrete locations from which it is later retrieved; the stored item sits inert until some process addresses it. A brain does nothing of the kind. Incoming signals are distributed across enormous numbers of connections which are themselves altered by the signals passing through them, so that memory is not a filed copy but a change in the structure that does the processing. There is no separation between what stores and what computes. Opponents of machine consciousness argue that this difference is not incidental. The brain's activity is continuous, chemical as well as electrical, and shaped by a body whose states it is always registering; it is active in the sense of going out and selecting what to attend to, not passive in the sense of accepting whatever arrives. Consciousness, on this view, is what that kind of organisation does, and nothing in a system of discrete storage and retrieval is positioned to do it. The machine can model the output of such a system without instantiating it, in the way a weather simulation does not get wet. Opponents accept that machines may eventually match or exceed human performance on any particular task, and deny that this bears on the question. Performance is behaviour, and behaviour is what consciousness explains rather than what it consists in. A system that produces the right answers by a wholly different route has reproduced the route's output, not the route. Because consciousness is a product of the brain's distributed, self-modifying biological organisation and computers process information in a fundamentally different way, this argument concludes, a machine cannot be conscious.

Premises

Counter-arguments

Proponents reply that this describes one computing architecture rather than all of them. Systems whose connections are weighted, revisable and massively interconnected, and in which what stores and what processes are not separate, are precisely what the objection says machines cannot be, and they exist. The difference cited is therefore a fact about current engineering rather than a limit of principle, and the argument identifies no property of neural tissue that could not in principle be implemented in other components.

Rejecting the premises

[Rejecting P1] This describes one architecture rather than computers as such. Systems in which storage and processing are not separated, and in which the connections are themselves modified by the signals passing through them, are designed and run, so the contrast is not a property of machines. [Rejecting P2] The description of the brain is accepted, and nothing in it establishes that such an organisation cannot be instantiated in other materials. The premise sets out a difference from one design where the conclusion needs a difference in principle. [Rejecting P3] That consciousness is what this organisation produces, and that reproducing outputs is not instantiating it, is the point at issue rather than a premise. Whether a system reproducing the organisation rather than only the outputs would be conscious is the question the argument was supposed to answer.