Google DeepMind has invested heavily in scientific AI over the years, and it paid off in 2024 when the company’s CEO and director Damis Hassabis and John Jumper won the Nobel Prize in Chemistry for AlphaFold, a special system that can predict the three-dimensional structure of proteins.
Now its rivals are working to catch up. In October 2025, OpenAI launched a team dedicated to AI for science, and Anthropic simultaneously announced several Claude features geared toward the life sciences. OpenAI specifically calls building an autonomous researcher its “North Star.” It just announced GPT-Rosalind, the first in a planned series of specialized scientific models. Google released its AI co-scientist tool last February.
Under the hood, many of these AI-for-science systems are actually multiple specialized AI agents working in concert. A Google co-scientist uses a supervisor agent, a generation agent, and a ranking agent to generate possible hypotheses and research plans in response to a goal provided by a human scientist. More recently, researchers at Stanford’s AI for Science Lab, led by James Zhou, devised a “virtual lab” composed of agents playing the roles of experts in various scientific fields. They found that their system could design new antibody fragments that bind to SARS-CoV-2, the virus that causes Covid.