AI, Bernabé warns: 'Large IPOs serve to spread risk'

What if large IPOs on artificial intelligence were a way of passing on the risk of outsized investments to small retail investors? This is the doubt that Franco Bernabé, President of the University of Trento, leaves us with on the sidelines of the panel 'To trust artificial intelligence or not'. "In the world of AI," the banker confides to our notebooks, "a total of 850 billion dollars have been invested. This is money that will never come back, in terms of return on investment. And so we need to broaden the pool of those who invest in such a way as to reduce the risks for those who have invested before. In this sense, perhaps the oxen park is a somewhat daring metaphor, but we are certainly faced with reducing the risk for those who have invested and do not have a business model that allows a return on that size of investment. Sure, the revenues are there, the prospects are there. But on that size it is frankly difficult to say that investments will return'.

The panel, moderated by the editor-in-chief of La Stampa, Andrea Malaguti, in the beautiful setting of Palazzo Geremia, was an opportunity to take stock of one of the most topical issues when it comes to AI: trust. And from university didactics to cybersecurity, passing through the Constitution, data profiling and the risk of an AI built on investments that are difficult to sustain in the long term, the positions that emerged gave visions that were in some respects very different, but united by one point: AI is penetrating the mechanisms of society much faster than the ability to control it.

Andrea Zoppini, lecturer at the University of Roma Tre, brought the topic inside the universities, where AI has already entered teaching, and it would be useless to think of stopping it. The point, if anything, is to change the way students are trained. 'Taking a critical attitude towards AI responses is a cultural fact'. That is why, he adds, the university's task becomes teaching students to control what AI produces, turning them into 'verifiers' of machine-generated responses. For Zoppini, the road of prohibition (explicitly citing the case of Italy's Garante Privacy that blocked ChatGPT in the past) is not viable. Better is an approach based on regulation. With one alarm: the underuse of data in the European context. A limitation that, according to Zoppini, has already turned into a competitive advantage for the United States and China.

The theme of technological dependence also returns in the words of Franco Bernabé. The former manager and president of the University of Trento uses very sharp tones on the issue of profiling and data control. He says he does not use WhatsApp, questioning the real security of commercial messaging platforms and pointing out how even institutional representatives use tools that, in his opinion, expose huge amounts of personal information. 'We are in the middle of two systems, the US and China, where profiling is very high,' he notes. Bernabé also cites the case of the US ICE and the regulations that would allow access to very large databases for control and monitoring activities. In the background is the idea that AI is accelerating a model of concentration of technological and information power in a few large global platforms.

On the legal side, Giovanni Maria Flick, President Emeritus of the Constitutional Court, recalls instead the role of the Constitution as a compass to deal with the impact of AI. The reference is to Articles 21 and 41, i.e. freedom of expression and economic freedom. For Flick, the real point concerns the way in which thought is formed in a society traversed by systems capable of producing increasingly sophisticated content, information and simulations. "Finding a key that can distinguish the true from the verisimilar is necessary," he explains, adding that it becomes essential to also establish a limit beyond which to stop.