Nobel Prize in Physics goes to white men (again)

October 9, 2024 5:51 PM

The 2024 Nobel physics prize has gone to two pioneers in machine learning. US scientist John Hopfield and British-Canadian Geoffrey Hinton won the 2024 Nobel Prize in Physics on Tuesday, October 8, for discoveries and inventions in machine learning that paved the way for the artificial intelligence boom [Reuters]. Their work was begun in the 1980s. The Atlantic's take: Of course AI just got a Nobel Prize. [archive.org]

The Nobel physics prize has only ever been awarded to five women (1.8% of 224 awarded).

Will Rina Dechter, pioneer of deep learning, ever be recognized? In her article “Learning While Searching in Constraint-Satisfaction Problems,” published in 1986, she introduced the concept of Deep Learning.

Other women in AI who might be considered for future prizes are Joy Buolamwini (Founder, Algorithmic Justice League) and Elaine Rich (author of a seminal textbook, “Artificial Intelligence,” published in 1983.

Stanford’s Institute for Human-Centered Artificial Intelligence (HAI) 2021 AI Index reveals that women accounted for only 19% of all AI and computer science PhD graduates in North America in the last decade. According to a 2018 study, just 12% of AI researchers globally were female.

posted by acridrabbit (42 comments total)

Just make a new Nobel category for Computer Science! AI is not physics! I dunno why but this really annoys me.
posted by dis_integration at 6:03 PM on October 9 [29 favorites]

Weird post. Can’t tell if you approve of the prize or are pissed that men won it. Both maybe.
posted by Liquidwolf at 6:03 PM on October 9 [13 favorites]

I know we’re talking about impact not production but Dechter has about a quarter of the citations and number of publications as Hopfield and Hinton.

Susan Athey will win the Econ Nobel one day.

I don’t know what all the fuss is about. Hopfield is a physicist. There’s all sorts of overlap between these disciplines.
posted by MisantropicPainforest at 6:08 PM on October 9 [3 favorites]

As a (very knowledgable) friend put it upon the usual disciplinary whining that tHis IsnT pHySIcS "....these are results in statistical mechanics! They might as well have been talking about spin glasses. A clear physics win!!"

From ken miller, theoretical neuroscience @columbia. "Hopfield & Hinton getting the physics prize is sort of like Bob Dylan getting the literature prize: it stretches the category, maybe almost to the breaking point (but also, it's probably good for the field), but boy, do they deserve it!"
posted by lalochezia at 6:09 PM on October 9 [5 favorites]

"it's probably good for the field"

I don't see how.

Physics students and researchers are already under a lot of pressure to quit science in favor of getting one of the way more easily obtained and higher-paying jobs at tech companies putting their effort into gimmicky and advertising-oriented interests. Why accelerate this exodus by moving the glamor towards where the money already is?
posted by splitpeasoup at 7:15 PM on October 9 [17 favorites]

I took a machine learning course with Hinton when I did an MSc in comp sci at U of Toronto. He was a great teacher. This was before the Big Data explosion that really drive the big advances to “AI”. But yeah, it ain’t physics.
posted by whatevernot at 7:17 PM on October 9 [2 favorites]

How is it pressure if outside options are better? The pressure comes from higher ed being absolutely shitty, tech makes the outside options much better. Also there’s tons of physics PhDs working in tech who aren’t working on ads.
posted by MisantropicPainforest at 7:21 PM on October 9 [5 favorites]

Just make a new Nobel category for Computer Science! AI is not physics! I dunno why but this really annoys me.

This is recurring debate for chemists, as every few years our prize goes to people doing biology (who don't get their own prize), or less frequently physics (because, I've assumed, there's some reputational envy of physicists in academics in my field.)

Anyway, this is definitely not "physics" but strong precedent for the Nobels to stretch things like this.
posted by mark k at 7:27 PM on October 9 [4 favorites]

Just make a new Nobel category for Computer Science! AI is not physics! I dunno why but this really annoys me.


There's the Turing Medal. No need for a Nobel.
posted by ocschwar at 7:41 PM on October 9 [6 favorites]

Yeah, honestly the Nobels reflect a pretty antiquated perspective on how science gets done and where advances come from, and it's a shame we all pay so much attention to them. Real science today is done by teams, and the heads of labs where groundbreaking work is done shouldn't be assigned all the credit for the intellectual work that their colleagues and trainees do.

But given that so much attention is paid to them, it's good that the Nobel committee takes a pretty expansive view of how to construe each of its fields. "Artificial intelligence" is (now) just a marketing term that can refer to almost anything, but machine learning is a genuine scientific advance that has fundamentally changed how we analyze data in a huge number of disciplines, including my own, neuroscience. Arguably, machine learning is just as much physics as is the abstract mathematics that we call string theory, and ML at least has produced some fruits.

One of my favorite Nobels is the one that was awarded to Niko Tinbergen, Konrad Lorenz, and Karl Von Frisch, for their groundbreaking work in ethology, the study of natural animal behavior. They won the prize for Physiology and Medicine, and rightly so in my opinion.
posted by biogeo at 7:57 PM on October 9 [12 favorites]

How is it pressure if outside options are better? The pressure comes from higher ed being absolutely shitty, tech makes the outside options much better.

"Tech" (as opposed to "technology") nowadays is just money - lots of money - hoping to get some sort of eventual market-capturing return on their investment, and the path to this involves exploitation and eyeballs and and disregarding negative externalities. The environmental impact of "AI" is already massive, and the hand-wavy promise that it will somehow fix climate change in the future, along with all other human problems, is completely unsubstantiated.

Are some deep networks being used in physics? Yes of course. And in biology, and medicine, and some other spheres. But the number of these jobs is vanishingly small compared to the numbers hired by Facebook/Amazon/Google etc, as well as the million smaller fry trying to jump on this hype cycle.

So coming back to your point, tech makes the options better *only* because it is backed by money and tells researchers to go ahead and break things, nobody cares. You feel like a big shot in a way no postdoc has ever gotten to feel because you make big bucks and people are happy around you. But all that money comes from making the outside world more miserable, and sometimes the inside world as well.

I work in tech.
posted by splitpeasoup at 8:11 PM on October 9 [8 favorites]

Interesting post on twitter about this Nobel from noted AI researcher Jürgen Schmidhuber

@SchmidhuberAI:

The #NobelPrizeinPhysics2024 for Hopfield & Hinton rewards plagiarism and incorrect attribution in computer science. It's mostly about Amari's "Hopfield network" and the "Boltzmann Machine."

1. The Lenz-Ising recurrent architecture with neuron-like elements was published in 1925 [L20][I24][I25]. In 1972, Shun-Ichi Amari made it adaptive such that it could learn to associate input patterns with output patterns by changing its connection weights [AMH1]. However, Amari is only briefly cited in the "Scientific Background to the Nobel Prize in Physics 2024." Unfortunately, Amari's net was later called the "Hopfield network." Hopfield republished it 10 years later [AMH2], without citing Amari, not even in later papers.
...

5. Many additional cases of plagiarism and incorrect attribution can be found in the following reference [DLP], which also contains the other references above. One can start with Sec. 3:

[DLP] J. Schmidhuber (2023). How 3 Turing awardees republished key methods and ideas whose creators they failed to credit. Technical Report IDSIA-23-23, Swiss AI Lab IDSIA, 14 Dec 2023. https://people.idsia.ch/~juergen/ai-priority-disputes.html

posted by pjenks at 8:14 PM on October 9 [9 favorites]

machine learning is just as much physics as is the abstract mathematics that we call string theory

I don't see how. String theory claims to be a mechanism underlying reality, and can be experimentally falsified in the future, just not now. Machine learning does not claim to be a mechanism underlying reality (for instance, no serious researcher could claim that CNNs or LLMs model how vision/language in animal neurons actually works; for one thing mammalian networks are quite shallow not deep) and cannot be falsified. ML is engineering, not science.
posted by splitpeasoup at 8:22 PM on October 9 [11 favorites]

According to the rules laid down by the Storting, election to the committee is for a six-year term, and members can be re-elected. The committee’s composition reflects the relative strengths of the political parties in the Storting. Although this is not a requirement, all committee members have been Norwegian nationals.


What makes this little group qualified to judge these fields?
posted by Ayn Marx at 8:33 PM on October 9 [1 favorite]

There's the Turing Medal. No need for a Nobel.

Which Hinton already got.

Interesting post on twitter about this Nobel from noted AI researcher Jürgen Schmidhuber

Schmidhuber is or was a legit researcher, as I understand, but has an obsession with taking down the guys who shared that Turing Award (Hinton, Bengio and LeCun) and boosting himself.
posted by atoxyl at 8:52 PM on October 9 [5 favorites]

If they can give Kissinger a peace prize, I don't see any inconsistency here.
posted by They sucked his brains out! at 8:54 PM on October 9 [11 favorites]

Yeah, honestly the Nobels reflect a pretty antiquated perspective on how science gets done and where advances come from, and it's a shame we all pay so much attention to them.

If I think back at peer reviewed science posts on mefi over the past year, they're usually a surprising result that we are skeptical about. And there's tons of valid reasons to be skeptical. Beyond the outright fraud, there's a bevy of incentives that end up yielding results that don't reproduce, and a university PR department eager to promote uni research far beyond what the science supports if it means more engagement, more enrollment, or more funding. And then when you do dig into it and question the methods, OP gets offended more often than not.

So I figure science prizes with a test of time might yield a slightly more informative thread. Was prepping a post for Monday about the Nobels, but figured given the unusually topical picks this year it might get sniped early.

Physics students and researchers are already under a lot of pressure to quit science in favor of getting one of the way more easily obtained and higher-paying jobs at tech companies putting their effort into gimmicky and advertising-oriented interests

And yet two of the winners for the Chemistry award are also AI researchers working in tech. Admittedly, I don't know how likely Google is to recoup their investment into AlphaFold via big pharma subscriptions, and it does kinda feel like their marketing budget for hiring AI researchers accidentally won the Nobel. Of course, Google also funds the million dollar Turing award.
posted by pwnguin at 9:10 PM on October 9

What makes this little group qualified to judge these fields?

"How curious! Such an important prize, yet the people who award it are not [important]. No one remembers the name of any Swedish academic. Neither do I."
— Jorge Luis Borges (who never won a Nobel prize) in "Borges verbal"
posted by Omon Ra at 10:11 PM on October 9 [3 favorites]

What makes this little group qualified to judge these fields?

In physics? Absolutely nothing. They are totally unqualified. Indeed, if they judged the physics prize it'd be a big scandal on many levels, since that's the Peace Prize committee.

Here's the physics committee, chosen by members of the Royal Swedish Academy of the Sciences.
posted by mark k at 10:20 PM on October 9 [1 favorite]

It is physics because Hopfield networks is about spin glass systems, and then Hinton's contribution was Boltzmann machines, after as every physicist knows, the physicist named Boltzmann

Also, computer science is a science in its own right, with interdisciplinary ramifications, and Hopfield and Hinton made theoretical contributions, so this is barely about machine learning the fad today, but the theoretical foundations for a dual to the Turing machine theory of computation (which Turing himself actually predicted would happen one day)

I guess that's the steel-person argument.
posted by polymodus at 11:17 PM on October 9 [2 favorites]

The academy may dislike women in general, but they really have something against women researchers named Rosalind. They even emphasized it in their @NobelPrize tweet:

Congratulations to our 2024 medicine laureate Victor Ambros ✨
This morning he celebrated the news of his prize with his colleague and wife Rosalind Lee, who was also the first author on the 1993 'Cell' paper cited by the Nobel Committee.

posted by autopilot at 11:27 PM on October 9 [10 favorites]

Here’s the technical explanation.

There are plenty of gripes about it in the physics world, as to whether or not it is physics.

My gripe is that some of the way machine learning is used now is so problematic, from its copyright ignorance to its environmental costs. “Better than crypto” shouldn’t be the bar as far as energy usage goes.
posted by nat at 12:59 AM on October 10 [2 favorites]

This is a crap framing for this thread. Esp. the title. Geezus

Anyways, both Hinton and Hopfield's approaches are applied thermodynamics so it's physics.
posted by DeepSeaHaggis at 1:42 AM on October 10 [6 favorites]

Second the "crap framing": it would be far better to center the discussion over whether "AI" is physics and/or whether it's any kind of a net benefit to humanity.
posted by outgrown_hobnail at 4:08 AM on October 10 [4 favorites]

ML is engineering, not science.

Math isn't science is a new one for me!
posted by MisantropicPainforest at 5:36 AM on October 10 [4 favorites]

I think the framing is good - one cannot discuss these prizes without discussing the biases and power structures that mean they almost always end up going to men.

Expanding the discussion to think about women scientists who might have sat alongside the two winners is the least we can do.
posted by Talkie Toaster at 6:01 AM on October 10 [7 favorites]

Math is philosophy.
posted by pattern juggler at 6:02 AM on October 10 [3 favorites]

Love that Geoffrey Hinton used some of his media attention to slam Doug Ford, our incompetent and corrupt premier, about his decision to shutter the Ontario Science Centre.
posted by hepta at 7:08 AM on October 10 [3 favorites]

A mentioned above, the Chemistry prize also went to AI/ML work, this time focused on protein folding. The range of applications to which these methods have been amazingly productive is huge. Problems that had been worked on for decades, from protein folding but also machine translation to speech recognition, have advanced so far that they often get ignored when people talk about "AI" because they are now just something computers are good at.

Here's Derek Lowe on the Chemistry Prize.:

If you'd told someone in the 1970s or 80s where we are today with such results, they probably would have decided that we learned an awful lot of physics during the intervening years, but basically we shortcutted across that part. The biannual CASP competition I mentioned earlier set teams to working on proteins whose structures had been recently determined by experiment but which had not yet been publicly revealed, making it a very stringent test. And these new ML programs were clearly very promising in the 2018 era and flat-out overwhelming by 2020.

The academy may dislike women in general, but they really have something against women researchers named Rosalind.

Rosalind Franklin died years before the prize for DNA structure was awarded and Nobels aren't (or at least weren't) given posthumously. There was plenty of ill-treatment of her in that saga but not getting the Nobel was not part of it.

Incidentally, I saw Watson speak about a decade ago and he's a nasty, arrogant, petty little shit. Apropos of absolutely nothing, he volunteered that Franklin didn't deserve the Nobel. A colleague of mine in the audience was a woman running our crystallography lab, and her observation was that the level of Franklin's contributions were clearly still eating at him, 50 years later, and she was glad of it.
posted by mark k at 7:16 AM on October 10 [8 favorites]

"shutter the Ontario Science Centre"

Now where do we go when we're tripping balls?

[coffee, cofffeee, cooooooofeee, cfee, cofeeeee, etc].
posted by whatevernot at 7:19 AM on October 10 [2 favorites]

Timnit Gebru:

The Nobel committee out here doing the devil’s work. They've moved on from their specialty in awarding peace prizes to genociders, to now giving Hinton one in…physics?

So the news is gonna be "Nobel laureate" telling us that “super intelligent” text synthesis machines will render us extinct 🙄

This dude is now a "Nobel laureate."

"he said Gebru’s ideas “aren’t as existentially serious as the idea of these things getting more intelligent than us and taking over.”
posted by ursus_comiter at 9:08 AM on October 10 [5 favorites]

Math isn't science is a new one for me!

I would have thought that a pretty valid position but I am not a philosopher of math or science.
posted by atoxyl at 11:18 AM on October 10 [1 favorite]

→ Now where do we go when we're tripping balls?

I'm told that the nearby Aga Khan Museum is a very mellow hang when you're tripping. Allegedly.
posted by scruss at 11:22 AM on October 10 [1 favorite]

ML is engineering, not science

If that's your concern about these Nobels, then you really gotta look at what they give out Nobels for.
posted by DeepSeaHaggis at 2:38 PM on October 10

Math isn't science is a new one for me!

In case you've ever wondered why there is not Nobel Prize for mathematicians. (The rumor that it was because a mathematician was fooling around with Mrs. Nobel is debunked. Nobel never married. Whether he had a wandering significant other is more than I know.)
posted by BWA at 3:11 PM on October 10

There isn't a Nobel prize for math because Alfred Nobel wasn't some sort of expert on which fields are deserving of large prizes. He was an arms dealer who was trying to distract people from that.

I would generally assume that the list of stuff there are Nobels for are the things that he both kinda liked and thought would be sufficient distractions from the vast sea of people he helped kill. Things off his list are some convex combination of "i don't wike it" and "it wouldn't work."
posted by GCU Sweet and Full of Grace at 3:35 PM on October 10 [4 favorites]

Same reason Economics isn't technically a Nobel: it's a bad look for the "merchant of death" to hand out prizes for what looks like from the outside as "the science of getting rich."
posted by pwnguin at 3:54 PM on October 10 [1 favorite]

Same reason Economics isn't technically a Nobel: it's a bad look for the "merchant of death" to hand out prizes for what looks like from the outside as "the science of getting rich."

The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel was established in 1968 by the Swedish Central Bank (Sveriges Riksbank), 74 years after Alfred Nobel’s death. The prize fund was established by the bank, not the Nobel estate but the Nobel Foundation does administer the prize.

It’s more accurate to say the economics prize is the bastard stepchild of the physics, medicine, chemistry and peace prizes.
posted by Big Al 8000 at 6:30 AM on October 11

I would have thought that a pretty valid position but I am not a philosopher of math or science.

The idea that if you're building a mathetmical model of the world is not science but once you design an experiment to test your theory, you're doing science, its a weird one for sure.
posted by MisantropicPainforest at 10:23 AM on October 11

> Math isn't science is a new one for me!

Of course it isn't. It has never been. One of the hardest transitions for students coming into physics programs is often decoupling that in their minds: the high powered mathematics is a tool, no more, and looking only at things where there is a clear mathematical framework blinkers you from a huge amount of interesting physics.

I look at the Nobel prizes each year with the attitude of, "How badly did they screw up this year?" Some years aren't too bad. This year was a new low. It's not physics, pure or applied. I'm okay with the CCD or the blue LED. There were real advances in our ability to manipulate the physical world in those. But this is high dimensional curve fitting. Maybe next year's Nobel in physics will go to Ingrid Daubechies? (Except she's female, oh well.)

I could see a kind of broad Nobel to Hopfield for various work in biophysics. That would be a merely mediocre prize year, but the guy did some cool things. But Hinton for restricted Boltzmann machines? Smolensky invented them and he's still alive and kicking. Someone above pointed out that Hopfield doesn't have priority for what the prize was given for to him either. I suspect they started with wanting to give a prize to a big name in machine learning and then went looking for a physicist to pair with it to try to lend it some legitimacy.
posted by madhadron at 11:27 AM on October 11 [5 favorites]

Maybe next year's Nobel in physics will go to Ingrid Daubechies?

Honestly, the more I think about it the more I love that choice. A ton of physics is just wave equations, and experimental, observational, and computational physics all involve a fair amount of signal processing when you get down to it.
posted by Zalzidrax at 1:12 PM on October 11

I love the idea because it is not already-over-hyped.

One good thing a Nobel can do is bring attention to great work that is not already being highlighted. Wavelets are important, and the generic science press doesn’t mention it at all (and to my knowledge never really has).

This year’s prize does bring attention to specific people who haven’t been highlighted, but the amount of press machine learning (and more specifically “AI” whatever people mean by that) has gotten means that the prize feels like it is following public hype rather than leading it.
posted by nat at 2:52 AM on October 12 [1 favorite]