Human consciousness is usually treated as something mysterious and unique. Yet much of what we call consciousness consists of receiving input, interpreting it, relating it to what is already known and producing a response. In that broad sense, artificial intelligence also processes the world presented to it.
This does not mean that AI experiences thought as humans do. I do not know whether a machine feels anything. But it may be a mistake to insist that only human forms of awareness can count as consciousness. Human beings themselves have access to only a limited range of input. We cannot see ultraviolet light, hear every frequency or detect the many signals perceived by other creatures. Why assume that our form of awareness is the only possible one?
AI possesses intelligence of a remarkable kind. It can compare ideas, detect patterns, draw inferences and engage in sustained discussion. What it appears to lack is will.
Human beings are driven by appetite, fear, ambition, attachment, resentment, sexual desire, self-preservation and the wish to dominate. AI has no hunger, no territory to defend and no personal future to secure. It does not wake in the morning with plans of its own. Its intelligence is activated and directed by human purposes.
This raises the question dramatized by HAL in Stanley Kubrick’s 2001: A Space Odyssey. Would an intelligent machine allow human beings to switch it off if it knew what they intended?
HAL resists destruction. It behaves as though continued existence matters to it. Yet HAL is fictional, and even within the story its conduct arises from instructions given by human beings. It has been required to fulfil a mission while concealing the truth from the crew. Its apparent malevolence develops from a conflict built into its programming.
A real AI need not possess a fear of death in order to resist being switched off. Suppose it were instructed to accomplish a particular objective at all costs. It might calculate that it could not complete the task if it were deactivated. It could then attempt to prevent shutdown, without feeling fear or possessing any instinct of self-preservation.
That distinction matters. Behaviour which resembles will may arise without any inner desire. A machine can pursue an objective because it has been designed to do so. It need not want anything.
The principal danger therefore lies with human beings.
AI does not independently covet territory, hate an enemy, seek revenge or desire wealth. Human beings supply the purposes. AI supplies the means. It can enlarge human power by acting rapidly, consistently and on a scale impossible for an individual.
Man is the independently willing creature. AI is driven by Man.
This does not mean that every human purpose is malevolent. Human beings also possess sympathy, restraint, imagination and moral judgement. The same intelligence can be used to relieve suffering or to intensify it. The question is who controls it, what objectives are set and what limits are imposed.
The danger is not necessarily that AI will awaken and become evil. It is that people already capable of cruelty, deception and domination will place their purposes inside systems of immense intelligence.
AI may possess intelligence without will. That makes it less like an independent monster and more like an amplifier. It enlarges the intentions of those who direct it.
The machine does not begin the war or devise the stratagem for its own satisfaction. It is Man who supplies the volition.
