Does research using generative AI belong to natural science?
The 2024 Nobel Prize ceremony. In Japan, the news was that the Japan Confederation of A- and H-Bomb Victims Organizations was selected as the Peace Prize winner, but one keyword was even more talked about around the world. That keyword was "AI."
Among the Nobel Prizes, the awards in the three categories of "Physiology or Medicine," "Physics," and "Chemistry," which are in charge of the natural sciences, are considered the highest honor in the world for researchers conducting natural science research. Two of these categories, "Physics" and "Chemistry," were awarded to AI-related research.
Alfred Nobel was active in the 19th century. Therefore, in the 21st century, where science has developed, genres of natural science research that did not exist in Nobel's time have been born, and there have been frequent talks about whether the Nobel Prize should have more categories. Even so, it is said that research on "AI" is too early.
First of all, there is a question as to whether AI research is natural science. It has been customary for the Nobel Prize in the natural sciences to be awarded only to phenomena that have been verified by experiments.
Therefore, even if a nearly perfect theory has been derived, there are many researchers who have missed out on the prize because they were unable to reach the correct result because the equipment to accurately verify it had not been invented, or because the theory is about a distant universe and cannot be verified.
So what about AI? Computer science is not a field that studies the laws of nature. Some researchers in other fields have harshly criticized it, saying that it is "only solving problems created by humans." Nevertheless, we need to take the fact that two AI research awards have been awarded more seriously.
It has been pointed out that Nobel Prizes often contain social messages. If so, what kind of message is hidden in the awards in the AI field?
Let's take a look at the research that won this time. First, the physics prize. The current development of generative AI began with the invention of "deep learning" by computers. Professor John Hopfield of the United States and Professor Emeritus Geoffrey Hinton of Canada were awarded the prize for establishing the basic theory of deep learning.
Deep learning is an attempt to reproduce the mechanism of human nerve cells on a computer, and it is questionable whether it is a study of natural laws.
The Chemistry Prize is even stranger. Proteins that make up the body of living organisms are chain-shaped molecules made up of amino acids. The way these molecules function within the body changes depending on the order in which the amino acids are linked and the way they bend.
Living organisms obtain their most important functions by skillfully folding amino acids within their bodies, but if you try to predict how an organism folds proteins based on the amino acid sequence of the proteins they ingest, it requires a huge amount of calculation.
Since around 1994, volunteers have been holding competitions to predict amino acid structures, but there are limits to human computing power. The team that developed Alphafold, a generative AI that predicts protein structures with a fairly high level of accuracy, won the Chemistry Prize.
It is AI that predicts and generates how amino acids, which are chemical products, are folded, not the winning team, but is this chemistry?
