An AI can write ‘I think’, ‘I understand’, ‘I am afraid’ or ‘I want’. This does not mean that it thinks, understands, is afraid or wants anything in the human sense of these terms. It generates language based on extremely sophisticated mathematical models. It can simulate the outward signs of thought with a sometimes unsettling degree of effectiveness, without this proving the existence of any inner experience.
This is probably what unsettles us the most. For centuries, we have associated certain forms of language with consciousness. Someone who responds intelligently to you is, normally, a person. With artificial intelligence, this old intuition no longer works quite so simply. A machine can hold a convincing conversation without there necessarily being, somewhere behind the words, a tiny digital being pondering its existence.
Artificial intelligence does not wake up in the morning. It does not lie awake staring at the ceiling, wondering what it is going to do with its life. It is not afraid of dying. It does not dream of freedom. It harbours no grudge against whoever switches off its computer. It has no childhood, no lived memories, no body, and no biological desire to survive.
We must therefore be wary of our natural tendency towards anthropomorphism. We have always ascribed intentions to things: to a car that ‘refuses to start’, to a computer that ‘holds a grudge against us’, sometimes even to the weather. With an AI capable of speech, the trap obviously becomes much more powerful.
This does not, however, mean that we should regard artificial intelligence as harmless.
The real dangers are probably far more mundane – and therefore more credible – than a revolt by sentient machines. An AI can make mistakes. It can replicate errors present in the data it was trained on. It can be used to fabricate false information, manipulate images, automate certain forms of surveillance, cut jobs or enable humans to make bad decisions much more quickly than before.
But in all these cases, the problem does not stem from a machine that has suddenly decided to harm us.
It stems from what we decide to do with it.
Artificial intelligence does not, of its own accord, possess a political, economic or military agenda. On the other hand, a company, a government, an organisation or an individual may well use this technology in their own interests. It is therefore the rules, practices, responsibilities and power dynamics that we must monitor much more closely than some hypothetical artificial consciousness hidden away in a server.
We are, nevertheless, undergoing a major cultural shift. For the first time, humanity has created machines capable of producing texts, images, voices and, soon, entire worlds that are sufficiently convincing for us to begin to doubt what we see.
This is a significant development.
But this power is not the same thing as consciousness.
A calculator can compute infinitely better than a human being without understanding the meaning of a number. A camera can capture a face without knowing what a face is. Artificial intelligence, too, can manipulate concepts with impressive skill, yet this does not prove that it experiences anything.
Perhaps one day the question of artificial consciousness will truly arise. No one can reasonably guarantee today what these technologies will become in several decades’ time. Moreover, we do not even have a fully satisfactory scientific definition of our own consciousness.
But we must distinguish this hypothetical future from what exists today.
Today’s AI is impressive. It can be useful, fascinating, and at times unsettling. It is already transforming work, art, information and our relationship with images.
But it isn’t staring back at us from the other side of the screen, pondering its own fate.
So perhaps the pressing question is not: ‘What will artificial intelligence decide?’
It is much simpler:
‘What are we, as humans, going to decide to do with it?’
