In Wednesday’s Future Perfect newsletter, my colleague Dylan Matthews wrote about the case for skepticism about this year’s Nobel Prize in Economics winners. His argument was that while their theories are interesting, there is ample reason to doubt the validity of those theories.
For a number of other Nobel Prizes this year, however, my skepticism goes in the opposite direction. This year’s Physics Nobel Prize was awarded to John J. Hopfield and Geoffrey E. Hinton “for fundamental discoveries and inventions that enable machine learning with artificial neural networks.”
Sign up here to discover the big, complicated problems facing the world and the most efficient ways to solve them. Shipped twice a week.
The award undoubtedly reflects serious, impressive and world-changing work on their research topics, almost certainly some of the most impactful work in existence. The hotly debated question is whether this Nobel Prize should actually be awarded to physics physics.
Together, Hopfield and Hinton did much of the foundational work on neural networks, which store new information by changing the weights between neurons. The Nobel Committee states that Hopfield and Hinton’s background in physics inspired their fundamental AI work, and that they reasoned based on analogies to molecular interactions and statistical mechanics in developing the early neural networks.
That’s cool, but is it physics?
Some people don’t buy it. “At first I was happy that they were getting such a prestigious award, but when I read further and saw that it was about physics, I was a bit confused,” Andrew Lensen, an artificial intelligence researcher, told Cosmos magazine. “I think it’s more accurate to say that their methods were inspired through physical research.”
“I am speechless. I love ML [machine learning] and ANN [artificial neural networks] as much as the next person, but hard to understand that this is a physics discovery,” physicist Jonathan Pritchard tweeted. “I think the Nobel Prize was affected by the AI hype.”
Resentment over AI stealing the spotlight only intensified when the Nobel Prize in Chemistry was announced. It went partly to Google DeepMind founder Demis Hassabis and his colleague John Jumper for AlphaFold 2, a machine learning predictor of protein structure.
One of the most difficult problems in biology is anticipating the many molecular interactions that influence how a protein printed from a given sequence of amino acids will fold. A better understanding of protein structure will dramatically accelerate drug development and basic research.
AlphaFold, which can reduce the time needed to understand protein structure by orders of magnitude, is a huge achievement and very encouraging about the eventual ability of AI models to make significant contributions in this field. It is certainly Nobel worthy – if there were a Nobel Prize for biology. (That’s not the case, so chemistry had to do it.)
The Nobel Prize in Chemistry seems to me much less complicated than the Nobel Prize in Physics; To the extent that it gave rise to resentful grumbling, I suspect it was mainly because, along with the physics prize, it was starting to look like a trend. “Computer science seemed to complete its takeover of the Nobel Prize,” Nature wrote after the chemistry prize was announced.
The Nobel Prizes bet on AI, declaring on one of the world’s most prestigious stages that AI researchers’ achievements in machine learning constituted serious, respectable, world-class contributions to the fields that had loosely inspired them. In a world where AI is becoming increasingly important and where many people find it overhyped and extremely annoying, that is a loaded statement.
Overhyped is a bad way to think about AI
Is AI overhyped? Yes, absolutely. There’s a constant barrage of obnoxious, exaggerated claims about what AI can do. There are people raising absurd amounts of money by applying “AI” to business models that don’t have much to do with AI at all. The enthusiasm for “AI-based” solutions often exceeds any understanding of how they actually work.
But all that can – and does – go hand in hand with the fact that AI is really a very big problem. AlphaFold’s achievements in protein folding took place in the context of pre-existing competitions to better predict protein folding, because there was a clear understanding that solving that problem really mattered. Whether or not you have any enthusiasm for chatbots and generative arts, these same techniques have brought cheap, fast, and effective transcription and translation to the world – making all kinds of research and communication tasks much easier.
And we’re still in the early days of using the machine learning systems for which Hinton and Hopfield first established the framework. I really think some people who position themselves as “against the AI hype” are actually leaning against the wall of an early 20th century factory and saying, “Have you gotten electricity yet to solve all your problems?” No? Hmmm, I guess it wasn’t that big of a deal.
At the turn of the 20th century it was difficult to anticipate where electricity would take us, but it was in fact quite easy to see that the ability to transfer large parts of human labor to machines would be of great importance .
Likewise, it’s not hard to see that AI will matter. So while it’s true that there is an annoying and eager group of ignorant investors and dishonest fundraisers eager to label everything with AI, and while it’s true that companies often systematically exaggerate how cool their latest models are, it’s not “hype” to To see AI as a huge problem and one of the most important scientific and intellectual contributions of our time. It’s just accurate.
The Nobel Prize Committee may or may not have tried to ride the hype train – they’re just regular people with the same motivation as anyone else – but the work they’ve identified really matters, and we all live in a world that is enriched by it.
