It is likely that we will see more Nobel medals awarded to researchers who used AI tools. As this happens, we may find that the scientific methods honored by those Nobel Prize committees move away from the simple categories of "physics," "chemistry," and "physiology or medicine."
By Nello Christianini, Professor of Artificial Intelligence, University of Bath
The 2024 Nobel Prizes in Physics and Chemistry gave us a glimpse of the future of science. Artificial intelligence (AI) was central to the discoveries honored in both of these awards. One wonders what Alfred Nobel, the founder of the awards, was thinking about all this.
It is likely that we will see more Nobel medals awarded to researchers who used AI tools. As this happens, we may find that the scientific methods honored by those Nobel Prize committees will move away from the simple categories of "physics," "chemistry," and "physiology or medicine."
We may also see a change in the connections between the scientific background of the winners and these categories. This year's physics prize was awarded to American John Hopfield from Princeton University, and Geoffrey Hinton who was born in Great Britain and belongs to the University of Toronto. While Hopfield is a physicist, Hinton studied experimental psychology before turning to AI.
The chemistry prize was shared between biochemist David Baker from the University of Washington and computer scientists Demis Hassabis and John Jumper from Google DeepMind in the UK.
There is a close connection between the AI-based advances that have been recognized in the physics and chemistry categories. Hinton helped develop approaches that led DeepMind to a breakthrough in predicting protein shapes.
The physics laureates, especially Hinton, laid the foundation for the powerful field known as machine learning. It is a subfield of AI that deals with algorithms, sets of rules for performing specific computational tasks.
Hopfield's work is not widely used today, but the backpropagation algorithm (renovated by Hinton) has greatly influenced many different sciences and technologies. The algorithm focuses on neural networks, a computational model that mimics the structure and function of the human brain for data processing. Backpropagation allows scientists to "train" huge neural networks. While the Nobel Committee tried to link this influential algorithm to physics, the connection may not be very direct.
Training a machine learning system involves exposing it to huge amounts of data, sometimes from the Internet. Hinton's progress eventually enabled the training of systems like GPT (the technology behind ChatGPT), and AI algorithms like Google DeepMind's AlphaGo and AlphaFold. Therefore, the effect of backpropagation was enormous.
DeepMind's AlphaFold 2 solved a 50-year-old problem: predicting the complex structures of proteins from their molecular components, amino acids.
Every two years, since 1994, a competition has been held to select the best methods for predicting protein structures and shapes from amino acid sequences. The competition is called Critical Assessment of Structure Prediction (CASP).
In several recent competitions, CASP winners used some version of DeepMind's AlphaFold. There is, therefore, a direct line that can be drawn between Hinton's backpropagation and Google DeepMind's AlphaFold 2 breakthrough.
David Baker used a program called Rosetta to achieve the difficult feat of building new types of proteins. Baker and DeepMind's methods hold enormous potential for future applications.
Credit assessment has always been a controversial issue in the Nobel Prizes. Up to three researchers can share a Nobel Prize. But scientific breakthroughs are collaborative. Scientific articles can have 10, 20, 30 or more authors. More than one team may contribute to discoveries that the committee honors.
This year we may hear more discussions regarding the attribution of the research on the backpropagation algorithm, claimed by various researchers, as well as the general attribution of discoveries to a certain field such as physics.
Now we have a new dimension to the attribution problem. Things become less clear if we can always distinguish between the contributions of human scientists and those of their artificial partners - the AI tools that are already contributing to pushing the boundaries of human knowledge.
In the future, is it possible that we will see machines take the place of scientists, with the people limited to a supporting role? If so, AI tools may get the main Nobel prize along with humans needing their own category.
