The impact of artificial intelligence: why the Nobel Prize in Physics and Chemistry was awarded to psychologists and mathematicians.
The awards in both disciplines share a common denominator: artificial intelligence (AI) played a central role in the discoveries. However, despite AI's recognition at the highest scientific level, this year's choice by the Royal Swedish Academy of Sciences raised several questions.
In particular, Hinton from the University of Toronto initially focused on experimental psychology before turning his attention to artificial intelligence, and Hassabis and Jumper, who work at Google DeepMind on the AlphaFold2 project, are mathematicians rather than chemists or biologists.
Moreover, some physicists have expressed doubts about the appropriateness of awarding the prize to Hopfield and Hinton specifically in the field of physics, as their achievements are more related to computer science. This decision has led to the notion that the Nobel Committee was influenced by the popularity of AI when determining the winners in the field of physics.
Despite these controversies, the committee's recognition of the significant contribution of artificial intelligence may pave the way for future awards to scientists for discoveries in which AI becomes an integral part.
Machine Learning and Protein Structure Prediction
The physics laureates Hinton and Hopfield laid the groundwork for the field now known as machine learning back in the 1980s. This branch of artificial intelligence involves algorithms and rules used to perform specific computational tasks.
"Artificial intelligence is a broad concept that includes machine learning, and a part of it is deep learning, which uses mathematical models of artificial neural networks. In particular, transformers, which are the basis for all generative AIs, are one of the architectures of neural networks," explains Alexander Krakovetsky, co-founder and CEO of DevRain.
Hopfield and Hinton made significant contributions to the development of artificial neural networks by applying principles of quantum and statistical physics. The scientists created specialized mathematical models that simulate brain function. These models are called artificial neural networks. They consist of many small elements working together like neurons in the brain.
Hinton developed a neural network based on the Boltzmann machine, which learns from examples and can classify or generate images. His discoveries contributed to the development of large neural networks that are now widely used in science and technology.
This year's chemistry laureates shared the award for achievements in protein prediction and design, opening new horizons in biology and medicine.
One of them, David Baker, participated in the CASP bioinformatics competition in 1998, where scientists attempted to predict protein structure as accurately and quickly as possible based on its sequence. The goal of the competition is to understand how complex protein molecules are formed from "bricks" – amino acids. This process can be likened to assembling a puzzle without having the picture.
At that time, Baker decided to take an unconventional approach. Instead of making assumptions about what the final result should look like, he started with a ready-made "picture" and selected the corresponding pieces to fit it. In 2003, this approach allowed for the creation of entirely new proteins that did not exist in nature. This led to the emergence of a new field in bioinformatics – protein design.
Other Nobel laureates in chemistry, Hassabis and Jumper from Google DeepMind, developed AlphaFold2 – an AI that has fundamentally changed the approach to predicting protein structures.
Why was the Nobel Prize in Physics awarded for Artificial Intelligence?
The awarding of the physics prize to Hinton and Hopfield for their work on machine learning has sparked a debate among scientists. In particular, the Nobel Committee's decision has been criticized for the research not meeting the prize's criteria. Moreover, the scientific background of the laureates does not closely align with the categories in which they won. For instance, Hinton did not work in physics – he studied experimental psychology.
"I have no words. Despite my commitment to machine learning and artificial neural networks, it's hard to recognize this as a breakthrough in the field of physics. It seems that the Nobel Prize has been overtaken by the wave of enthusiasm for AI," – wrote astrophysicist Jonathan Pritchard from Imperial College London on X.
According to theoretical physicist Sabine Hossenfelder, who works at the Munich Center for Mathematical Philosophy, the scientists' research is more related to computer science than physics.
"The Nobel Prize is a rare opportunity for physics and physicists to be in the spotlight. It's a day when friends and family remember they know a physicist and might reach out to ask about this year's prize. But not this time," – emphasizes Hossenfelder.
As indicated by the arguments of the Nobel Committee, the laureates' deep knowledge in physics, particularly regarding the behavior of small particles within materials (e.g., atoms), helped the scientists create artificial neural networks. However, if the physics prize is to be awarded for everything related to materials and processes within them, then nearly all discoveries could be attributed to physics, as everything around us is made up of atoms.
On one hand, the discoveries of this year's laureates represent a breakthrough in science, while on the other hand, it is difficult to determine to which field of knowledge they belong more.
AI is primarily computer science. Since, for obvious reasons, there is no Nobel Prize in computer science or informatics, the field of physics has become the only option for scientists working in this domain. This leads to a rhetorical question: how much has the current popularity of AI influenced this year's decision by the Nobel Committee?
However, not all scientists are skeptical about this year's physics prize laureates. "The research by Hopfield and Hinton was interdisciplinary as it combined physics, mathematics, computer science, and neuroscience. In this sense, their work pertains to all of these fields," notes Matt Strassler, a theoretical physicist at Harvard University.
AI is now part of science
Nobel Prizes are often awarded for discoveries that occurred a long time ago. One of its main criteria is the significant impact of research results on a particular field, and this can only be assessed over time. This year's Nobel laureates in physics, whose work is around 40 years old, only confirm this rule.
Although the emergence of ChatGPT has made AI another area dominated by tech giants, structural changes in artificial intelligence began long before that. According to data from Stanford University, in 2014, academia was the driving force behind AI. At that time, the academic field had only three machine learning models, while by 2022, businesses had created 32 models.
The 2024 Nobel laureates in chemistry exemplify the importance of interaction and data sharing between scientific research and business in the development of AI. The principles of neuroscience, physics, and biology were used to create modern AI models. Data obtained by biologists facilitated the development of AlphaFold software, which simplifies the determination of protein structures.
Professor of Molecular Biophysics at King's College London, Rivka Isaacson, notes that traditional methods of deciphering protein structures are labor-intensive, but this data has become the foundation for training AlphaFold. She adds that AI technology has allowed scientists to make strides in deeper investigations of protein function and dynamics, posing various questions and potentially opening new areas of research.
Overall, the use of AI in scientific research, such as information gathering or data analysis, can significantly accelerate the emergence of new scientific discoveries. "Now, with generative AI, I could write my dissertation in two months," says Krakovetsky from DevRain, emphasizing how much time AI can save in the most resource-intensive