AI Formulates Its Own Hypotheses and Drives New Discoveries... Who Will Receive the Nobel Prize? [Capturing 125 Years of the Nobel Era]②

AI Moves Beyond Tool to Hypothesis and Experimentation
The 125-Year-Old Concept of the "Discoverer" Put to the Test
"Humans Remain Responsible for Judging and Taking Ownership of Discoveries"

Editor's Note125 years have passed since the first Nobel Prize was awarded in 1901. Science has evolved from an age of discoveries made by lone geniuses to an era where thousands of people and artificial intelligence (AI) conduct research together. The Asia Business Daily is running a seven-part series examining how the main agents and methods of scientific discovery have changed over 125 years of Nobel Prizes, and to what extent the Nobel Prize has reflected these changes.

Among the three winners of the 2024 Nobel Prize in Chemistry, two used artificial intelligence (AI) to solve long-standing challenges in the scientific community. Demis Hassabis, CEO of Google DeepMind, and John Jumper, then senior research scientist (vice president), developed "AlphaFold2," which accurately predicted the 3D structure of proteins based only on their amino acid sequences. This was a problem that the scientific community had grappled with for over half a century. The Royal Swedish Academy of Sciences awarded the Nobel Prize in Chemistry to the researchers who developed AlphaFold2. David Baker, professor at the University of Washington in the United States, who contributed to computer-aided protein design, was also recognized alongside them.

Demis Hassabis (left) and John Jumper, winners of the 2024 Nobel Prize in Chemistry, are attending the Nobel Prize press conference held on December 7, 2024, at the Royal Swedish Academy of Sciences in Stockholm, Sweden. They jointly received the Nobel Prize in Chemistry for developing the artificial intelligence (AI) model 'AlphaFold2,' which predicts the complex three-dimensional structure of proteins. Photo by Jennifer 8. Lee/Wikimedia Commons

AI That Formulates Hypotheses and Designs Experiments: The Emergence of Partnership

Until now, AI has been a ‘tool’ utilized by humans, and the Nobel Prize has been awarded to individuals who achieved groundbreaking results using that tool. However, as AI begins to independently generate hypotheses, design experiments, and even seek solutions, the boundaries between the agent and the tool are breaking down. A new challenge is arising: how to distinguish the contributions of AI developers, scientists who utilize it, and researchers who verify results through experiment. As AI moves from a stage where it merely answers human questions to one where it formulates both the questions and the answers, the concept of the ‘discoverer’—predicated on being human for 125 years—has also come under review.

There is growing evidence that AI is already stepping beyond being merely a tool that supports human research activities and is now deeply involved in the process of scientific discovery. A notable example is the AI system "Robin," which was published in the international journal Nature. Robin consists of several AI agents working together: it searches academic papers to generate hypotheses, proposes verification experiments, and uses data analysis to suggest follow-up research directions.

Autonomous Chemistry Laboratory operated by the University of Liverpool in the UK. Research on 'autonomous laboratories,' which combine artificial intelligence (AI) and robotics to perform experiments, analyze results, and explore new substances, is underway. As AI moves beyond data analysis to actual experimental procedures, new questions arise about how to delineate the roles of humans and AI in scientific discoveries. Provided by University of Liverpool

The research team tasked Robin with identifying new treatments for dry age-related macular degeneration. Robin proposed treatment strategies and candidate drugs, analyzed the experimental results, and even suggested subsequent experiments. In this process, it identified a new drug candidate that had not previously been proposed as a treatment for this condition. However, the actual experimentation was conducted by human researchers. This demonstrates that AI is moving beyond merely calculating data created by humans—it is now participating in the full research process, connecting hypotheses, experiments, and analyses.

Dashun Wang, professor at Northwestern University in the United States, said in an interview with The Asia Business Daily, “AI is moving beyond the role of a simple tool by exploring vast domains humans cannot reach alone.” He described the future of science not as AI replacing humans, but as a "human-machine partnership." He predicted that the model, where humans and AI share different capabilities and jointly produce a discovery, will become increasingly common.

Professor Wang pointed out, “If the contribution of a machine becomes essential to a scientific breakthrough, a new challenge will be how to recognize that contribution.” He explained that existing systems of scientific recognition are designed on the assumption that only humans make intellectual contributions, so they are unable to properly reflect this new era of change.

He further explained, “While large-scale collaborative research involving thousands of human researchers already raises the issue of dividing credit among people, AI takes this challenge a step further. Now, the premise that the subject of intellectual contribution must necessarily be human is itself being called into question, going beyond merely dividing credit among humans.”

Curiosity and Responsibility for Recognizing Serendipity Still Belong to Humans

Of course, experts note that even if AI can perform hypothesis generation and experiments, it cannot yet be considered an independent ‘discoverer.’ This is because many major discoveries in science have not always emerged from simply finding answers to predetermined problems.

Tobias Erb, Director of the Max Planck Institute for Terrestrial Microbiology in Germany, emphasized that numerous great discoveries began not from specific goals, but from “serendipity and pure curiosity,” as shown by Alexander Fleming’s discovery of penicillin in 1928. Fleming discovered that bacteria died around mold growing in a dish he was using to culture staphylococcus. From the perspective of his original experiment, the contaminated petri dish was close to a failed sample that should have been discarded. But he did not overlook this unexpected phenomenon. The human insight to find meaning in serendipitous events laid the foundation for the era of antibiotics.

Director Erb said, “Many fundamental discoveries begin with pure curiosity,” highlighting the ability to pursue questions without knowing where observations may lead. He explained that it remains uncertain whether AI can realize "scientific curiosity," the ability to draw significance from unexpected phenomena beyond set goals.

The more possibilities that AI generates, the more crucial the human role in identifying which outcomes actually constitute discoveries may become. Jinho Jung, President of the Korean Academy of Science and Technology, emphasized, “As AI generates thousands of results, human discernment in picking out the gems will become ever more important. Ultimately, humans must define what is a valuable discovery for humanity, and only humans should take responsibility for those results in their own names.”

Peter Brzezinski, Professor at Stockholm University and Secretary General of the Nobel Committee for Chemistry, likewise stated that, at present, AI is regarded as just one among many ‘tools’ utilized by researchers. He explained that the criteria for evaluating researchers who achieve Nobel Prize-worthy discoveries or inventions do not depend on which tools were used. In other words, it is not the AI itself, but the insight and outcomes produced by the researcher using it, that matter most.

This content was produced with the assistance of AI translation services.