Leggi in ItalianoLately, I have been coming across more and more interventions warning against artificial intelligence. We will lose our jobs, we will stop thinking, we will become dependent on machines, we will no longer be able to tell truth from falsehood. Reading some of these commentaries, one could almost believe AI is a calamity from which we should try to save ourselves. I confess that this picture leaves me perplexed. The risks exist, and some are very serious, but I have the impression that a great deal is said about what we might lose and considerably less about what we are already gaining.
In my university work I now use artificial intelligence tools almost daily. When I prepare a lecture, study a judgment, or try to reconstruct a complex legal problem, I realise how much research possibilities have changed. I can compare documents, explore questions that would previously have required far more time, test a hypothesis and follow connections I might not even have thought of pursuing. I must check everything, of course, because errors do occur — and how. Sometimes I even find myself arguing with an LLM that insists on an unconvincing legal reconstruction. But that is precisely the point: I do not feel that I think less. If anything, I manage to go deeper into far more things than before. And the point extends well beyond the university. I think of medical research, of the development of new drugs, of programming, of translations, of tools for people with disabilities, and of all those repetitive tasks that consume hours without adding much to our work. How is it possible that, when discussing AI, these benefits so often end up in the background?
The opposite risk, that of trusting too much, I can see perfectly well. As I recounted in a previous post, every now and then, talking to my daughter, I notice that she tends to attribute to ChatGPT’s answers an authority even greater than that of my own observations. The thing makes me smile, but also reflect. And I worry when I see students tempted to replace study with an automatically generated answer, or when I read of lawyers who have filed briefs containing non-existent judgments. I tell my students to use artificial intelligence like an electric bicycle, not like a moped or a self-driving vehicle. With an electric bicycle you go further, you tackle climbs that previously seemed impossible, and you reach destinations that were unthinkable in far shorter times. But you must keep pedalling. The muscles, in our case, are the brain: the capacity to reason, to doubt, to recognise errors and to critically evaluate answers. If you simply sit back and let yourself be carried along, perhaps without even knowing where you are going, you lose the most interesting part of this technology: the possibility of amplifying our intelligence, rather than replacing it.
Reading certain interventions, however, what comes to mind is a modern version of the revolt of carriages against the arrival of the train. Back then, too, there must have been understandable reasons for worry: the speed, the accidents, the trades destined to disappear, the habits turned upside down. And not everyone benefited equally from that transformation. But would anyone today seriously propose going back to carriages? Perhaps my reaction also has something generational in it. I knew research work before digital databases, when finding a judgment or an article meant consulting repertoires, journals and volumes in the library. Then came the Internet, search engines and online archives. Each step changed the way we work. I remember that world with some fondness for its rhythms, but I do not miss its limitations at all.
Then there is the geopolitical dimension, which it would be naive to ignore. The competition between the United States and China in artificial intelligence has assumed a strategic importance that recalls, in certain respects, the race for nuclear technology between the United States and the Soviet Union in the post-war period. Enormous economic and military interests are at stake, as are national security and the concentration of technological power. I find particularly troubling the prospect of ever more powerful systems deployed in the military domain without adequate controls, a subject I also addressed in a recent post on nuclear arms control. But compared with the nuclear arms race there is an important difference: AI is already finding civilian applications in a great many sectors. As far as I can observe today, the benefits of its use seem to me largely to outweigh the direct negative effects, even if it would be imprudent to claim we already know all the long-term consequences.
And it is precisely for this reason that the discussion seems to me particularly important for Europe. We have accumulated an enormous lag behind the United States and China, including in terms of investment and the capacity to develop our own technologies. If, to the difficulty of competing, we add a predominantly defensive attitude — more focused on limiting AI than on creating the conditions to develop it — we risk watching from the sidelines a transformation bound to affect our societies profoundly. I do not at all think that Europe should give up rules or the protection of rights. I wonder, rather, whether we are devoting equal energy to research, infrastructure, enterprise and training. Otherwise we risk becoming extremely good at regulating technologies developed, financed and controlled elsewhere.
I do not know where artificial intelligence will take us, and I distrust those who today make predictions that are too confident, whether optimistic or catastrophic. Continuing to discuss the dangers is indispensable. It would, however, be a pity if, out of fear of what might happen, we lost sight of everything this technology already allows us to do. Above all, I would avoid turning every discussion about AI into a premature funeral for human intelligence.