Evrim Yazgin
Cosmos science journalist
The worlds of physics and computer science are abuzz as this year’s Nobel Prize in Physics has gone to two trailblazers in artificial intelligence (AI) and machine learning technologies.
John Hopfield and Geoffrey Hinton share the prize.
Hopfield, an emeritus professor at Princeton University in the US, invented in 1982 an artificial network that can save and recreate patterns. This has become known as the Hopfield network.
Hinton, emeritus professor at the University of Toronto, Canada, developed upon the Hopfield network to create the Boltzmann machine. This advance saw an artificial neural network which, like the brain of an animal, can recognise characteristic elements in a dataset.
It is thanks to these developments in the 1980s that today’s AI innovations are able to make the impact they are – and will.
The announcement of the winners of the prize has been met with a buzz of excitement… and a little confusion.
Enthusiastic response
Diep Nguyen, a computer scientist at the University of Sydney tells Cosmos the award is “very well deserved and I am not surprised at all. Given the boom of AI applications, I have been questioning when the inventors of the area will be recognized.”
In an email to Cosmos, Monash University data scientist Geoff Webb says he is “delighted.”
“AI is transforming our lives,” he adds. “Hopfield and Hinton were major contributors to the development of the Deep Learning technologies that underly many recent advances and are most deserving of humanity’s highest academic recognition.”
Similarly, Mostafa Azghadi, a James Cook University electrical engineering researcher, says he is “thrilled” that the award has been awarded to Hopfield and Hinton who is known as the “godfather of AI.”
“I was initially a bit surprised,” says Fiona Panther, an astrophysicist from the University of Western Australia. “But on reflection I think awarding the prize to Hopfield and Hinton represents the importance of computational methods in modern physics.
“I think the award also highlights that rather than being some sort of science-fiction magic with computers ‘thinking for themselves,’ neural networks and artificial intelligence are based on physical laws relating to entropy and how information can be stored, retrieved and transformed.”
A lasting impact in many fields
“The fundamental work that was done in the ‘80s, at the time, didn’t have any particular practical application,” explains La Trobe University biochemist Dave Winkler. “But I and lots of others could see how it could potentially grow. It underpins all the paradigm-shifting stuff we’re seeing in almost every field of science, technology and medicine with AI and machine learning.”
Nguyen echoes this, saying that he and others are “standing on the shoulders of giants” when it comes to innovations in AI.
“Their works laid the foundation for presenting physical objects/information like images, data by mimicking how human neural network works,” Nguyen says. “These are the foundation for what you see in smart cameras, autonomous cars.”
He even says that developments in AI originating from the work of Hopfield and Hinton could help tackle the biggest problems like climate change.
“I use Hinton’s research daily,” says Azghadi. “Their contributions have profoundly shaped the evolution of modern neural networks, making lasting impacts on everything from machine learning research to practical applications in various industries such as Agriculture and Aquaculture, to which I apply AI.”
“I have been interested in neural networks for about a decade now, when astronomers started to use machine learning more widely and graphics processing units (GPUs) made designing and operating artificial neural networks feasible even on small desktop computers,” explains Panther.
She highlights the importance of artificial neural networks (ANNs) in physics research and its close connection to the physics fields.
Because of popular media, you ask the general public what they think an ANN is, or what artificial intelligence is, and we get this idea of an almost ‘Skynet’-like artificial intelligence that can formulate ideas and think for itself,” Panther says. “Or a machine that makes random guesses and occasionally succeeds in hallucinating the right answer depending on the information it was trained on.
“However, the principles on which ANNs and artificial intelligence are based are the same as those we use in statistical mechanics. So, as well as sometimes being a useful tool in physics research, ANNs themselves are actually based on physics – or more accurately, statistics.”
Not your average Nobel in physics
Others in the scientific community, however, aren’t quite convinced.
“Initially, I was happy to see them recognised with such a prestigious award, but once I read further and saw it was for Physics, I was a bit confused,” says Andrew Lensen, a senior lecturer in artificial intelligence at Victoria University Wellington in New Zealand.
“We already have our own “version” in Computer Science (the Turing Award), which Hinton has already received for his work in deep neural networks,” Lensen adds.
“I think it is more accurate to say their methods may have been inspired by physics research,” he says. “I know that many in the physics community are upset about this, and I can’t blame them. Given that we already have a Turing Award, I would’ve preferred to see the Nobel Prize go towards a contribution to the Physics community. It does feel like the committee may have gotten a swept up in the AI hype!”
Geoff Webb, too, says he was “amused that they received the Nobel Prize for Physics as I do not believe that AI research is part of Physics. It illustrates the problems of applying disciplines from the start of the 20th century to modern research. AI is part of Computer Science, a discipline that only really started to emerge in the 1940s.”
Despite their initial surprise, some researchers have come to accept Hopfield and Hinton’s receipt of the award for Physics.
“I was initially surprised, given that AI isn’t traditionally seen as part of physics. But on reflection, the Physics category is quite fitting,” says Azghadi. “The recognition highlights how AI research, while rooted in computer science, has had wide-reaching implications, even beyond its original domain.”
And, some experts note, it is not unheard of for a Nobel Prize to be contentious for its scientific field.
A celebration of broad understanding
“The awarding of the Nobel Prize in Physics to AI research again proves the necessity of interdisciplinary research where various fields may converge,” Nguyen says. “Most of the time, breakthrough research or applications is rooted from this convergence where scientists apply the knowledge from one field to another field.”
Winkler agrees, saying it is “not so uncommon” for the prize to be awarded in a field that may leave some scratching their heads.
“Sometimes the Nobel Prize for Chemistry, you might say, well, it really was medicine, and some of the Nobel Prizes for Medicine could really be classed as chemistry. There’s much more team work being done to solve difficult problems, and there’s more multidisciplinary stuff brought to bear on the problems, which has been huge advantage.”
Hopfield was a theoretical physicist, after all, says Winkler.
“He asked: How does biology do these amazing things with memory?” Winkler says. “Can we emulate that by a simple statistical physics model. He did with the Hopfield networks.”
“I think we’re seeing the tip of the iceberg of where this is going,” Winkler says. I’ve worked in this field for over 30 years. It was mainly small neural networks, and they were incredibly useful for doing specific things. But now we’re getting to the point where we’re getting close to what might be called general AI. Programs like ChatGPT can do many different things.
“And I think everyone working in the field, even the experts, have been kind of a bit amazed at just how well these things can generalise.”
