This year’s AI-focused Nobel prizes reflect the zeitgeist
FCAI and University of Helsinki professor Teemu Roos breaks down why AI was so prominent in many of the Nobel prizes awarded in 2024.
Lue tämä suomeksi
Yutong Liu & The Bigger Picture / Better Images of AI / AI is Everywhere / CC-BY 4.0
Why was artificial intelligence research featured so heavily in the prizes in both physics and chemistry?
For physics, the prize went to John Hopfield’s and Geoffrey Hinton’s neural network models. Developed in the 1970s and 80s, these models drew inspiration from the nervous systems of both humans and animals as well as from physical systems. This research itself is not in significant use now, but new technologies have been built on its foundation.
In chemistry, by contrast, AlphaFold is a recent advance, and its developers Demis Hassabis and John Jumper were regarded as Nobel favorites. AlphaFold predicts the three-dimensional structure of proteins—emphasis on ‘predicts’, in other words uses statistical likelihoods. The research that got the chemistry prize is very applied, in practical use and provides a clearly understood solution—it can be used to develop new drugs and enzymes.
It’s a valid question whether the work that received the physics Nobel prize is really physics, but the story from neural network to the modern ChatGPT and billion-parameter models is fascinating and also involves Finland. The self-organizing maps of Teuvo Kohonen, professor at the Helsinki University of Technology (now part of Aalto University), were a major invention on par with Hinton’s neural networks. Hinton also made use of the backpropagation algorithm, which was developed by Seppo Linnainmaa, the first person to graduate with a computer science PhD from the University of Helsinki.
Why were AI-based achievements recognized this year?
It’s unusual that something as recent as AlphaFold is awarded a Nobel, compared to the physics prize, where the work was done decades ago. Traditionally Nobel prizes emphasize scientific innovation, and artificial intelligence is a commercial and technological phenomenon, which doesn’t carry much weight in the Nobel context, since commercial success gets rewarded in other ways. Machine learning has pervaded all fields, and AI has achieved tremendous success and caught the public’s attention. The Nobel committee has perhaps wanted to go with the times and has given more grist for the mill with these prizes. There are huge expectations for AI to accelerate science as well as economic growth. Daron Acemoglu, the winner of the Nobel economics prize, has been very critical and careful about what we can expect from AI.
Has AI made the practice of science obsolete?
AI has had a big impact benefiting all scientific fields. It’s a fascinating technology and object of study, because the time from research to applications is so short—you don’t need ten years to commercialize a product.
AlphaFold has accelerated how we approach one of the biggest challenges in chemistry and biology, but it can still make wrong predictions about protein structures. We need the same processes as before, like validation, tests and empirical studies. AI does not eliminate the need for these, it just makes them faster or more efficient. We still need human experts and traditional ways of doing things in parallel with AI.