With image generators and conversational agents, artificial intelligence (AI) is putting artists and writers to the test, but scientists believe that it could also revolutionize research and even feature prominently in the works of the Nobel Prize winners. .
In 2021, Japanese scientist Hiroaki Kitano is launching what he calls the Nobel Turing Challenge. It challenges researchers to create a “AI scientist” capable of autonomously carrying out research worthy of a Nobel Prize by 2050.
Some researchers are working hard to create such an artificial colleague and a hundred “scientific robots” are already at work in science, explains Ross D. King, professor of artificial intelligence at Chalmers University in Sweden.
The specialist published an article in 2009 in which he presented, with other researchers, a scientific robot called “Adam”the first machine to produce scientific discoveries autonomously.
“We built a robot that discovered new scientific ideas, tested them and confirmed that they were correct,” Mr. King told AFP.
The robot was programmed to formulate hypotheses autonomously, design experiments to test them and even program other laboratory robots to carry out these experiments and finally learn from these results.
“Not trivial” discoveries
“Adam” was tasked with exploring the inner workings of yeast and discovered previously unknown “gene functions.”
These discoveries are “modest” more “not trivial” however, the authors estimated in their article.
A second scientific robot called “Eve” was later created to study drug candidates for malaria and other tropical diseases.
With such robots, “it costs less money to conduct research and they work 24 hours a day,” explains Ross D. King, adding that they are also more rigorous in monitoring processes.
The researcher concedes, however, that AI is far from being up to the task of a Nobel-deserving scientist. This would require robots. “much smarter” capable of “understand the big picture” to compete with the Nobels.
“Not about to be replaced”
“The scientific tradition is not about to be replaced by machines”, agrees Inga Strümke, associate professor at the Norwegian University of Science and Technology, to AFP.
“That doesn’t mean it’s impossible.”she adds, believing that it is “certainly” clear that AI has and will impact the way science is conducted.
A good example of this is the Alphafold AI model, developed by Google Deepmind, which can predict the three-dimensional structure of proteins based on their amino acid.
“We knew there was a relationship between amino acids and the final three-dimensional shape of proteins and that we could use machine learning to find it”explains Ms. Strümke.
However, these calculations are too complex for humans and “the machine thus did something that no human could do”, she continued.
Alphafold at the same time highlighted the weakness of current AI models such as neural networks, according to it.
They are very good at processing massive amounts of information and coming up with an answer, but not able to explain why that answer is correct.
So, if the more than 200 million protein structures predicted by Alphafold are “extremely useful”, Mr. Strumke explained.“they teach us nothing about microbiology”.
AI-assisted
For her, science seeks to understand the universe and not simply to “make the right guess”.
Yet the groundbreaking work done by Alphafold has led experts to place its designers among potential candidates for a Nobel Prize.
Google DeepMind Director John M. Jumper and CEO and Co-Founder Demis Hassabis have already received the prestigious Lasker Prize in 2023.
They appear in the tablets of the analysis firm Clarivate, which anticipates potential scientific Nobel laureates on the basis of citations in research articles.
David Pendlebury, director of the firm, admits that if the article by Jumper and Hassabis published in 2021 has been cited thousands of times, it would be unusual for the Nobel jury to reward a work so quickly after its publication.
The Nobels usually reward discoveries dating back several decades.
But he believes AI-assisted research will soon feature prominently in Nobel-winning work.
“I’m sure in the next decade there will be Nobel Prize winners who will be aided by computing in one way or another, and computing these days is increasingly AI,” declared Mr. Pendlebury to AFP.
