AI agents could reshape scientific research beyond AlphaFold
AlphaFold’s success has strengthened hopes that artificial intelligence can accelerate scientific discovery through large datasets. The system, developed by Google DeepMind, predicts three-dimensional protein structures from experimentally measured examples. Demis Hassabis and John Jumper received part of the 2024 Nobel Prize in Chemistry for this work. However, the article argues that AlphaFold’s model may be difficult to reproduce in most fields because its training depended on the Protein Data Bank, which contains about 170,000 validated structures and took 53 years of international cooperation and an estimated $21 billion in experimental work to assemble.
Many scientific results are less consistent than protein crystallography, making comparable datasets difficult to create. The article instead highlights AI agents: systems that use large language models together with digital or physical tools to develop, test and revise hypotheses. Google’s AI Co-Scientist, for example, used several sub-agents to investigate how antibiotic resistance moves between bacterial species. Its conclusion about bacterial viruses carrying resistance genes matched research later produced by Imperial College London.
Agents can still hallucinate and have limits involving judgment, memory and autonomous operation. Their automated records could nevertheless improve reproducibility and preserve laboratory knowledge. By reducing the time and cost of testing ideas, they could also increase the pace and range of scientific research.
Written by Sakhanda Wire AI from reporting by MIT Tech Review. Facts, quotes and full context are in the original article below.