“To those who asked him why he had returned to the United Kingdom after having been for a while in the USA with the best mathematicians and scientists of his time, Alan Turing replied that he was not interested in solving problems with brute force. He preferred the beauty and elegance of mathematics”. Mario Rasetti, professor emeritus of Theoretical Physics at the Polytechnic of Turin, is on the same side as the English scientist who popularized the idea of artificial intelligence: “By using mathematics appropriately, much better things can be done”. Rasetti, who is also president of the Scientific Committee of Centai, will speak at the Italian Tech Week on September 27, at the opening of the session entitled “The Age of Artificial Intelligence”.
The announcement of Italian Tech Week 2023, there will also be Sam Altman, the “father” of ChatGpt: from 6 September registrations will start to listen to him 04 September 2023
And when he talks about “much better things” he refers, for example, to AI systems that use less computing power or pollute less. But couldn’t we try to ask a machine to look for more beautiful and elegant solutions? “For some reason, GPT is not so good at math. It would take a little effort to solve it, but maybe there isn’t the same market as there is for language processing”. So artificial intelligence, at least in its best-known incarnation, i.e. ChatGPT, will continue to miscalculate, yet it is already designing chips for machine learning: “If it is about optimisation, i.e. minimizing the amount of material needed and increasing as possible the speed with which the signal passes from one component to another, these are very easy processes for a machine”.
Matter of feeling
Rasetti, 81, is a member of the Princeton Institute for Advanced Science, but he is also among those who collaborated in the development of GPT4 and the next version, GPT5, of which very little is known yet: “The intelligence has not yet arrived artificial capable of feeling feelings, even if he knows how to feign them, in the sense that he knows the words to express them. And with a new function, called attention, an LLM (Large Language Model) can better understand what interests its human interlocutor”. It will be increasingly difficult to understand whether our interlocutor is a man or a machine, also because the system will have access to all human knowledge, including everything found on the internet, in real time.
An impossible comparison
Even so, machine intelligence will not be comparable to human intelligence: “A few years ago I conducted a study on the semantic map of the brain together with neuroscientists in California. We have calculated the number of different semantic structures that a human brain is capable of realizing, and it is practically infinite, 10 to the 700th power. Do you think that the number of protons in the universe reaches 10 to the 78th, that is, in comparison it is negligible. And the second point is no less important: we are certain that a human brain performs functions that cannot be defined according to the Turing machine model. All the computers we know, including quantum ones, however, adopt that architecture. We will never be able to create artifacts that faithfully simulate the brain, perhaps on condition that we use technologies that exploit living matter, so we must not be afraid of being overwhelmed by an artificial intelligence”.
Have we seen too many sci-fi movies? “Yes, we expect there are good or bad machines, in reality they are just objects that do what we tell them to do”. Even so, however, we must not fall into the cliché that technology is always neutral: “It is a thesis that can and must be discussed, because when we design artificial intelligence we inevitably introduce knowledge, prejudices and ideologies that are ours alone ”.
In the afternoon of 27 September at OGR the focus will be on “The Age of Artificial Intelligence”. In addition to Mario Rasetti, Caroline Yap, Managing Director, Global AI Business of Google Cloud, Karthik Narain, Group Chief Executive – Technology of Accenture and Filippo Rizzante, Chief Technology Officer of Reply will also be speaking with a keynote. There will be two panels: one dedicated to Italian startups that focus on artificial intelligence (with Maria Zollo, Co-Founder & CEO Mio.ai; Fabio Petroni, Co-Founder & CTO Samaya.ai; Gabriele Venturi, Founder PandasAI) and another to the ethical foundations of artificial intelligence, with Noémie Elhadad, associate professor and president of Biomedical Informatics, Columbia University, and Rickard Brüel Gabrielsson, MIT PhD and Co-Founder of the research laboratory Unbox AI. These meetings, like the entire Italian Tech Week programme, will also be available in live streaming on Italian.Tech and on the websites of La Stampa, Repubblica and Secolo XIX.