Artificial intelligence (AI) in the spotlight for the Nobel Prize in Physics! This is surprising. One of the laureates, Geoffrey Hinton, contacted by telephone this morning by the Nobel committee, was himself very surprised to learn that he had been awarded the Nobel in physics. However, when we look more closely, the techniques that come into play in the success of machine learning – or statistical learning – have a strong link with statistical physics: the tools used in statistical physics to describe metallic alloys – spin glasses – are the same as those used to describe neural networks.
In a recent issue of Research (quarterly n°577, April-June 2024), we have devoted a long file to AI and its use in research. On this occasion, computer scientist Yann Le Cun, who heads Facebook’s artificial intelligence laboratory, gave us an interview where he looked back on the emergence of this field of research, recounting the role of the two Nobel laureates in this field. year: the biophysicist John Hopfield for the development of completely connected neural networks in 1982, and Geoffrey Hinton, at the time post-doctoral in a group of psychologists and cognitivists interested in artificial intelligence, which puts developed the gradient backpropagation technique, fundamental to the success of machine learning.
Others, including Yann Le Cun and Yoshua Bengio – both 2019 recipients of the Turing Prize with Geoffrey Hinton – subsequently made these methods increasingly effective.
Find the entire interview with Yann Le Cun, with free access, here: “Machines lack common sense”
Crédit photo: Or. Niklas Elmehed © Nobel Prize Outreach
Related News :