Consciousness is self-awareness, and a machine can be made self-aware

The argument

This argument proceeds from a definition. If consciousness is self-awareness — a system's having information about its own states and about its position in an environment, and being able to act on that information — then nothing in the definition mentions what the system is built from. Proponents argue that the definition is the right one, since it is what we actually appeal to when we judge that a creature is conscious: it knows where it is, it knows something is happening to it, and it behaves accordingly. Machines, on this view, already do a weak version of this and can in principle do a strong one. A computer can be given models of its own internal states, can monitor whether those states are functioning as intended, can represent itself as an object among other objects, and can revise its behaviour in light of all three. Proponents note that each advance of this kind was dismissed in advance as impossible for a machine, and that the history of computing is a sequence of capacities once held to require a mind being implemented without one — which they take as a reason to distrust confident claims about the next frontier rather than to accept them. The argument concludes by shifting the burden. If a system has information about itself, acts on it, and reports on its own states, the sceptic must name the further property it lacks and explain why that property is required. Appeals to flesh, proponents argue, name a material rather than a function, and materials are not what awareness consists in. Because consciousness is self-awareness, and a machine can be built to be aware of itself and its environment, this argument concludes, a machine can be conscious.

Premises

Counter-arguments

Opponents reply that the argument wins by choosing a definition that omits what is at issue. A system that holds and acts on information about its own states has met a functional description, but whether anything is felt in doing so is precisely the question, and every criterion offered is behavioural. A machine that reports being self-aware has produced a string of symbols, and that it produced them because it is self-aware is what remains to be shown rather than what the report establishes.

Rejecting the premises

[Rejecting P1] The definition is the whole argument and it is offered rather than defended. Holding information about one's own states and acting on it is something a thermostat does in small measure and a flight computer in large; what the opposing case means by consciousness is that there is something it is like to be the system, and this account omits it by construction. [Rejecting P2] True of the account as given, and the premise inherits whatever is wrong with it. Independence from material follows from the definition rather than from anything established about consciousness itself. [Rejecting P3] Self-modelling, self-representation and behavioural revision are capabilities the premise supplies, and it does not address whether a system possessing them is aware of anything or is processing representations of itself unaccompanied by any experience.