Healthy AI will allow better care, either. But what are the ethical issues behind this progress?
Jean-Gabriel Ganascia: The first ethical vigilance to have, in my opinion, concerns the possible calling into question of the competence and the responsibility of the doctor. Will the automation of certain medical practices infringe the free will of the practitioner or the researcher?
Investigations or “challenges” opposing machines to humans have demonstrated this: certain artificial intelligence systems sometimes make more precise diagnoses than a doctor. As soon as the machine becomes more efficient than him, will the doctor dare to contradict it? Ultimately, the question may arise. Are we going to live in a world where machines make decisions for humans? If it comes down to deciding whether to turn right or left, like a GPS does, still pass. But here we are talking about health …
Would this be a step towards overcoming the human by the machine?
J.-GG: That’s not what it’s about ! When it comes to AI, there is always a misunderstanding. Nourished as we are by literature and films, we let ourselves believe that machines will become autonomous, take power and control us.
We are not there ! Still not. But a social system that comes to delegate certain decisions to automatons reduces the freedom of individuals.
Don’t you believe in a machine that will overtake man?
J.-GG: It’s a fantasy that’s been around for millennia. This is the myth of the Golem (in which an earth statue comes alive, editor’s note), that of the sculptor Pygmalion, falling in love with his work and asking the gods to bring it to life. It is the idea that the machine becomes autonomous, has a consciousness. It’s the old dream of taking yourself for God. There is something of transgression there.
But with AI, you don’t make a creature. AI is not autonomous, it is in fact only an automaton. In one of his books (1), written with Marcel Duchamp, the writer Michel Carrouges spoke of “Single machines”, which evolve alone, without concern for men or the common good. In health, machines should not be single, but partners.
How to ensure this partnership?
J.-GG: AI must remain a tool, an external element of information. A bit like a doctor consulting his Vidal dictionary. The final decision, however, must remain human. The more sophisticated machines we have, the more powerful AIs, the more we need doctors to give them meaning, precisely.