Tired of the AI ​​hype? Why the Nobel Prize in Physics went to an AI scholar.

In Wednesday’s Future Perfect newsletter, my colleague Dylan Matthews wrote about skepticism about this year’s Nobel Prize winners in economics. His argument was that while their theories were interesting, there were many reasons to doubt how true they were.

But for several of this year’s other Nobel laureates, my skepticism points in the opposite direction. This year’s Nobel Prize in Physics was awarded to John J. Hopfield and Jeffrey E. Hinton for their “fundamental discoveries and inventions that enable machine learning with artificial neural networks.”

There is no doubt that this award reflects serious, impressive and world-changing work on their research topics, making them arguably some of the most impactful research. A hotly debated issue is whether this Nobel Prize in Physics should actually count as a physics prize.

Together, Hopfield and Hinton did much of the fundamental work on neural networks, which store new information by changing the weights between neurons. The Nobel committee claims that Hopfield and Hinton’s background in physics inspired fundamental AI research, reasoning from molecular interactions and statistical mechanics analogies in developing early neural networks. I am doing it.

Some people don’t buy it. “At first I was happy to see them recognized with such a prestigious award, but when I read more and found out it was a physics award, I was disappointed,” said artificial intelligence researcher Andrew Rensen. , I was a little confused,” he told Cosmos magazine. “I think it’s more accurate to say that their method may have been inspired by physics research.”

“I’m speechless. I love ML (machine learning) and ANN (artificial neural networks) as much as the next person, but it’s hard for me to understand that this is a discovery in physics.” academic Jonathan Pritchard tweeted. “I think the Nobel Prize got caught up in the AI ​​hype.”

When the Nobel Prize in Chemistry was announced, the outrage that AI had stolen the spotlight was further intensified. Part of this project was contributed by Google DeepMind founder Demis Hassabis and his colleague John Jumper to AlphaFold 2, a machine learning protein structure prediction tool.

One of the most difficult problems in biology is predicting the many molecular interactions that influence how printed proteins fold from a given string of amino acids. A deeper understanding of protein structure will dramatically speed up drug discovery and basic research.

AlphaFold’s ability to reduce the time required to understand the structure of proteins by orders of magnitude is a major achievement and is very encouraging about the ability of AI models to ultimately make major contributions to this field. If there is a Nobel Prize winner in biology, it is definitely worthy of the Nobel Prize. (There is no such thing, so chemistry had to do it.)

A Nobel Prize in Chemistry seems to me to be as reasonable as a Nobel Prize in Physics. To the extent that it caused angry complaints, I think it was mainly because it was starting to look like a trend with the physics prize. After the chemistry prize was announced, Nature wrote, “Computer science appeared to have completed its takeover of the Nobel Prize.”

Nobel laureates bet on AI on one of the world’s most prestigious stages, and AI researchers’ machine learning achievements are a serious, respectable, world-class addition to the field that loosely inspired them. declared that it was a contribution. This is a difficult statement to make in a world where AI is becoming increasingly important and where many people find it overhyped and deeply annoying.

Is AI overhyped? Yes, absolutely. There is a constant barrage of disgusting and exaggerated claims about the capabilities of AI. There are people who incorporate “AI” into business models that have nothing to do with AI and collect exorbitant amounts of money. Enthusiasm for “AI-based” solutions often exceeds how it actually works.

But all of these things can and do coexist with AI, which is really, really important. AlphaFold’s protein folding achievements come within the context of an existing competition for better protein folding predictions. Because it was well understood that solving the problem was really important. Whether you’re passionate about chatbots or generative art, the same technology has brought cheap, fast, and effective transcription and translation to the world, making all kinds of research and communication tasks much easier.

And the use of machine learning systems, for which Hinton and Hopfield first built a framework, is still in its infancy. Some people who position themselves as “against AI hype” are effectively leaning against the wall of an early 20th century factory and asking, “Do we already have electricity to solve all our problems?” I think there are some people who say that. no? Well, I guess it wasn’t that big of a deal. ”

At the beginning of the 20th century, it was difficult to predict where electricity would take us, but the ability to hand over large parts of human labor to machines was actually very important. It was easy to understand.

Similarly, it’s not hard to see how AI will become important. So it’s true that there’s an obnoxious and rabid bunch of ignorant investors and dishonest funders who want to tag everything with AI, and how cool their latest models are. Although it is true that AI is often systematically exaggerated, it is not “hype” to think of AI as follows. This is a huge milestone and one of the leading scientific and intellectual contributions of our time. That’s exactly accurate.

The Nobel Prize committee may or may not have been trying to buy into the hype – they are ordinary people with the same range of motivations as anyone else – but the work they identified was really It’s important and we all live in a world where: It has made me rich.