Anthropic’s AI biology finding raises questions about scientific discovery
Anthropic says its molecular biology lab produced its first discovery using 950 Claude agents. The agents spent 21 hours examining biological data and identified a repeating pattern around a known enzyme. The company said the pattern had not previously been catalogued and described it as reminiscent of features associated with CRISPR. The finding was not a new DNA sequence, and human scientists conducted the experiments based on the agents’ reports.
Some biologists dispute calling the result a discovery. Lucas Harrington said that identifying an unusual cluster of genes and repeats may be relatively easy, while determining what the system does is the more difficult and important step. Mario Rodríguez Mestre of the University of Copenhagen said his team had already found the same pattern, according to The New York Times. Anthropic denied learning it from Mestre’s conversations with Claude; Mestre said he would stop using the chatbot.
The case has renewed debate over whether AI systems discover knowledge independently or contribute to research alongside people and other tools. The article notes that narrowing 200,000 candidates to a small group for further study can be valuable scientific work, even when humans guide the process and run the experiments. It also compares the dispute with questions surrounding OpenAI’s claim that its agents solved a million-dollar mathematics problem.
Written by Sakhanda Wire AI from reporting by MIT Tech Review. Facts, quotes and full context are in the original article below.