Will AI Win a Nobel Again? The Prediction Lists Say Probably Not

Two years ago, the Royal Swedish Academy of Sciences handed its two biggest science prizes to work that runs on neural networks, and one of the winners was a University of Toronto professor. This week the Academy announces again. The Nobel Prize in Physics is scheduled for Tuesday, Oct. 6, at 11:45 CEST at the earliest, which is 5:45 a.m. in Toronto. Chemistry follows on Wednesday, Oct. 7, at the same hour.

So: will AI win again? Going by the most-watched prediction list, the answer is probably not. That is worth knowing before the alerts start buzzing.

What 2024 actually rewarded

The 2024 physics prize went to John Hopfield and Geoffrey Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks”. The committee framed it as physics, pointing to Hopfield’s use of atomic spin to build an associative memory and Hinton’s Boltzmann machine. “The laureates’ work has already been of the greatest benefit. In physics we use artificial neural networks in a vast range of areas,” Ellen Moons, chair of the Nobel Committee for Physics, said in the Academy’s announcement.

Chemistry was split. Half went to David Baker “for computational protein design”, and the other half was shared by Google DeepMind’s Demis Hassabis and John Jumper “for protein structure prediction,” the work behind AlphaFold.

Look at that second citation closely. It never says AI. The prize was for a scientific result, and the neural network was the instrument that got there. Hold onto that distinction; it matters for Wednesday.

Then 2025 snapped back to tradition. Physics went to John Clarke, Michel Devoret and John Martinis “for the discovery of macroscopic quantum mechanical tunnelling and energy quantisation in an electric circuit”, and chemistry to Susumu Kitagawa, Richard Robson and Omar Yaghi “for the development of metal-organic frameworks.” No neural networks in sight.

How a laureate gets picked

Nobody campaigns, at least not officially. Nominations come by invitation only: academy members, past physics laureates, Nordic professors and chairs at selected universities. Forms are due by Jan. 31, and “no one can nominate himself or herself.” A Nobel committee screens the field, the Academy’s physics class weighs in, and the full Academy chooses by majority vote in early October. The decision is “final and without appeal.”

Then everything goes in a vault. Nobel Foundation statutes keep nominations confidential for 50 years, so any list you read this week, including ours, is an educated guess.

The guessers have a track record, and no AI picks

The most cited guesser is Clarivate, which mines citation data to name “Citation Laureates.” Its 2026 list, released Sept. 17, says 89 past picks have gone on to win Nobels, often years later. Swissinfo puts that against 487 Citation Laureates named since 2002, a hit rate of about 18 per cent. Not a crystal ball. Not nothing.

For physics this year, Clarivate named Chihaya Adachi, Stephen Forrest and Mark Thompson for phosphorescent OLEDs, the light-emitting materials in phone and TV screens; Nicola Spaldin of ETH Zurich for theoretical work leading to room-temperature magnetoelectric multiferroics; and Qikun Xue for the experimental observation of the quantum anomalous Hall effect.

Chemistry picks: David Allara, Ralph Nuzzo and Jacob Sagiv for self-assembled monolayers; Harry Gray and Jay Winkler of Caltech for long-range electron transfer in proteins; and Harvard’s David Liu for base editing and prime editing, two precision gene-editing methods.

Not one of those citations is about machine learning. Neither are the physics and chemistry suggestions collected by The Scientist, which point to metamaterials and to the lipid nanoparticles that made mRNA vaccines possible. The closest Clarivate gets to tech is in economics, where Stanford’s Susan Athey and Hal Varian, Google’s former chief economist, are named for “pioneering analysis of the economics of information technology.”

Bettors agree, loosely. A small-stakes Manifold market asking whether any 2026 science prize motivation will contain “artificial intelligence,” “machine learning,” “neural network” or “deep learning” sat at 10 per cent on Oct. 2. Its rules note that the 2024 chemistry citation would not have counted, since it never used those words.

Our read on what to watch

Speculation, labelled as such: the 2024 pair looks like a deliberate statement about where science was heading, and committees rarely repeat a statement two years later. The more plausible route for AI is the AlphaFold route. A prize for a discovery in materials, biology or physics where machine learning did some of the heavy lifting, with the word “AI” nowhere in the citation.

That makes Tuesday and Wednesday a reading exercise. Check the motivation text on nobelprize.org, then check the laureates’ methods sections. A win for OLED chemistry or gene editing would be a win for human-designed science, full stop. A win for a materials discovery found by screening millions of candidates computationally would be the 2024 story again, just quieter.

Hinton accepted his prize as a Toronto professor. If the Academy calls a machine-learning name at 5:45 a.m. on Tuesday, a lot of people in this city will be awake for it. The citation data says they can probably sleep in.