SNOLAB and the 2024 Nobel Prize in Physics

October 09, 2024 — Astrophysics News

By Stephen Sekula

The Royal Swedish Academy of Sciences awarded the 2024 Nobel Prize in Physics jointly to two physicists: John Hopfield (Princeton University) and Geoffrey Hinton (University of Toronto). It was given “… for foundational discoveries and inventions that enable machine learning with artificial neural networks”. What do these words mean and how do they connect to SNOLAB?

Nobel Prize in Physics 2024
Nobel Prize in Physics 2024

The prize is awarded for answering a very basic question. “If I make a machine that mimics a biological brain, how do I teach the network to hold and use information?” A person can take in stimuli (sight, sound, touch, taste, smell) and process that through a biochemical network to generate a response. Understanding bits of how such living networks function inspired the creation of artificial neural networks from the 1940s onward. However, it was in the 1980s that key breakthroughs happened, both in constructing a network that can store information (a memory) and how to then train that network. Hopfield and Hinton earned the prize for their contributions to these two areas.

How is this physics? The modern understanding of neural networks is based on the same approaches that physicists employ to study systems of atoms and the energy and order those systems can possess. In a sense, this prize is given for asking a biology question, “How do I get this network to learn and even to think?” and answering it with, “But have you tried physics?” That answer turns out to be profound and enables the current generation of neural networks, including the highly popular GPT4 (ChatGPT).

How does this connect to SNOLAB? Most of us are using ChatGPT (or something similar) to enhance our daily work, be that writing prose, writing software, or just getting semi-reliable answers to complicated questions. The impact of these ideas, however, goes far deeper. Experiments at SNOLAB are always trying to use machine learning to improve their experiments or the way they explore their data. For example, SNO+ has investigated the use of graph neural networks to improve event reconstruction. PICO has explored machine learning to improve its understanding of acoustic and camera data. There have been efforts to use data from PI Vision, along with neural networks, to predict what will happen next to equipment in the laboratory.

We are at the beginning of the age of AI. As with all new technologies, things can go either way. The good news is that physicists like Hinton, and many other people engaged in this space, are also deeply engaged in the ethics of AI. As the Nobel Prize committee said this morning in its announcement, let us strive to use this discovery for the betterment of all humanity.