AI: Explainer ET: What about AI and the 2024 Nobel Prize in Physics?

When John Hopfield (top left) and Geoffrey Hinton (above right) received the 2024 Nobel Prize in Physics, almost everyone was stunned. Why were they two computer scientists, undoubtedly pioneers in advancing? machine learning (ML) that drives artificial intelligence (AI), receive the top prize in physics?

The short answer is that under the hood of ML lie some of the most complex and elegant concepts in physics. It is prudent to return to John Hopfield’s contributions because they are not very well understood outside the realm of AI researchers.

To really get at Hopfield’s ideas we must think about how humans come to conclusions, some smart and some stupid. Memory and intelligence are closely related. Remember how when we see a particular color/pattern, we remember someone. Sometimes, simply a t-shirt and its aroma – Heath Ledger in Brokeback Mountain – or a cake – Madeleine and Marcel Proust – transport us in time and place. This is called associative memory. There is an initial trigger and all of our brain cells collectively fish for “that memory.”

That’s what a Hopfield network does. It goes from a “high excitement” state when trying to find a match until it stabilizes into a “chills, I got this” state once a match is found. Now let us consider the contribution of Geoffrey Hinton. How do we learn something? Making mistakes and then correcting assumptions until we finally get the right answers. Hinton used a physics technique called the “Boltzman Machine (BM).”

Simple things like “it looks cloudy maybe I should bring an umbrella” don’t require too many layers of analysis. Hinton’s intelligent use of BM can address complex, multi-layered decisions. Ergo, the Nobel of 2024. — Shishir Prasad