Anthropic touts AI-led biology discovery

Anthropic, an artificial intelligence startup based in the United States, recently announced its first discovery from its molecular biology lab. The breakthrough was made by its Claude model, which identified an unknown enzyme system. This discovery is being hailed as evidence that AI can expedite fundamental research in biology. Enzyme systems found in nature have historically led to significant breakthroughs, such as the foundation of the biotech industry and the DNA-copying method utilized in many medical tests.

To uncover this discovery, approximately 950 Claude-powered software agents spent 21 hours examining a database of DNA sequences. This process yielded over 200,000 genes associated with one type of enzyme, from which 20 were selected for further investigation. The AI agents worked autonomously, determining which leads to pursue based on initial instructions provided by Anthropic researchers. The involvement of human researchers was limited to the initial setup and subsequent laboratory experiments.

One of the agents detected a unique arrangement of short, repeating DNA segments next to two genes - one coding for a known enzyme and the other for an unknown protein. This configuration resembles CRISPR, a natural bacterial defense system that scientists have adapted into a gene-editing tool now used in medicine. Anthropic's CEO, Dario Amodei, who has previously expressed concerns about the risks associated with AI development, discussed the discovery in remarks to the U.N.

Security Council via video link. He noted that the finding is preliminary but may represent a new gene editing mechanism with potential applications in gene therapy. Amodei predicted that profound progress could be made in the near future regarding diseases that have plagued humanity for centuries.

The function of the newly identified system, dubbed ART, remains unclear at this stage. Anthropic mentioned that their initial experiments suggest the system might be programmable, similar to CRISPR. The results of this discovery have been published as a preprint and have not yet undergone peer review. Feng Zhang, a renowned pioneer of CRISPR and a professor at MIT and the Broad Institute, called the identification of the repeated DNA fragments "genuinely intriguing" and emphasized the need for further investigation.

Anthropic's lab team, formed in the spring, operates in the San Francisco Bay Area and does not handle pathogens capable of infecting humans. All experiments conducted by the team are carried out by human scientists. Amodei has not ruled out the possibility that Claude, Anthropic's AI model, may eventually operate lab equipment itself, provided appropriate safeguards are in place. The company has shared this summary generated by an LLM (Language Model), and a full disclaimer is available for readers.