What’s going on with this year’s Nobel Prizes? We can totally see headlines going round like “AI wins 2024 Nobel Prize in Physics” (and half of Chemistry too, it turns out!). Technically this is not too far off: on October 8th, the Physics Nobel was awarded to John Hopfield and Geoffrey Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks” (and on 9th October, it was announced that the Chemistry Nobel will be shared between David Baker and Google DeepMind’s Demis Hassabis and John Jumper, the latter for using AI “for protein structure prediction”). The big question that is dividing experts and the public is: is AI research Physics?
While most of us–who like to identify as Physicists–may have been surprised by this year’s physics announcement, I was even more surprised by the amount of colleagues who seemed to side against the decision, and more generally against the representation of the research celebrated by this year’s award as being in the domain of Physics. Luckily, some of the scientists I respect the most, including Nobel Physics Laureate Giorgio Parisi, have no doubts: the foundation of neural networks is rooted in beautiful mathematical and statistical Physics and deserved to be celebrated this way.
Echoing Parisi and Hopfield himself (btw, a former president of the American Physical Society), decades ago there used to be a much more rigid classification of scientific disciplines and any attempt at thinking outside the box risked to be seen as wacky and not always well tolerated. Thankfully, some true pioneers persisted in exploring and crossing the traditional discipline boundaries, ultimately creating whole new fields of research. That’s exactly what this year’s Nobel prize winners have done. By trying to model mathematically how our brain works–how sentient beings store and retrieve information, as in Hopfield’s associative memory model, and how after sufficient training we can recognise and spot new patterns, as in Hinton’s Boltzmann machine model–these visionary scientists, together with their research teams and numerous academic descendants, have planted the seeds for the AI revolution that is unfolding today.
As beautifully elucidated in the supporting documents presented by the Nobel committee (both the popular summary and the more technical advanced information are worth a read), the award is recognising the fundamental role played by tools from Physics in these early studies, yielding building blocks for the more advanced computational models of machine learning that are now developed at an astounding pace. Clearly, the impact of those seminal inventions and discoveries has crossed over to almost all sectors of knowledge and society, and has great potential to lead to more breakthroughs in the years to come. Already this year, the transformative role of AI in Chemistry was recognised by assigning half of their Nobel Prize to the developers of AlphaFold2, an AI system that predicts a protein’s 3D structure from its amino acid sequence.
What is particularly beautiful to realise is also how much the seminal research into neural networks and modern-day AI advances can give back to Physics itself. The clever and responsible use of these technologies has opened entirely new frontiers for data processing and analysis, from materials science to gravitational wave astronomy and even catalysing new discoveries in pure mathematics. Some of these discoveries will profoundly re-chart the science landscape, and that’s the most exciting perspective that an aspiring Physicist can hope to contribute to.
So let’s celebrate this year’s Nobel Prize in Physics and its legacy, which is for once so relatable by the general public. And to my skeptical colleagues: rather than arguing whether unquestionably groundbreaking research from four decades ago should be labelled as Physics or not, let’s focus our efforts into finding new creative ways to decode the rulebook of the Universe and ultimately reshape how Physics will look like five decades from now!