MIT Technology Review is reporting that Anthropic announced last Wednesday that it had launched a molecular biology lab earlier this year, where Claude agents read and conjecture about hard biology problems while human scientists run experiments on what they report. The company said the AI-powered lab made its first discovery.

The outlet said Anthropic's system of 950 agents, working over 21 hours, did not find a brand-new DNA sequence. It said the agents instead flagged a repeating pattern surrounding a known enzyme, a pattern Anthropic said had not been catalogued before. MIT Technology Review said Anthropic's announcement described the pattern as "reminiscent" of what led to the gene-editing technology CRISPR.

It said the claims have angered some biologists. A viral post from one biologist, later endorsed by the chair and CEO of drugmaker Eli Lilly, said finding a weird cluster of genes and repeats is often the easy part, and that the real discoveries come from figuring out what the system actually does. MIT Technology Review said the agents helped with laboratory grunt work, but that a discovery it is not.

The outlet also reported that biologist Mario Rodríguez Mestre of the University of Copenhagen said over the weekend that his team had already discovered this particular pattern, the New York Times reported. It said Mestre, who regularly chatted with Claude in his work, wondered whether Anthropic's team had learned from his conversations. Anthropic denies this, but Mestre said he is stopping all use of Claude anyway.

MIT Technology Review said part of the problem is that AI companies are not presenting their systems simply as tools scientists can use, like microscopes or supercomputers, but insisting the systems are making discoveries themselves. It said that approach is incompatible with how science actually works to some, with new knowledge more typically emerging from collaboration and a growing arsenal of tools.

The outlet said whittling 200,000 candidates down to a few worth exploring is legitimate scientific work, and that a general-purpose chatbot doing that work is notable. But it said once the standard becomes whether Claude itself made a discovery, the debate has only two answers: breakthrough or bust.

MIT Technology Review also noted that OpenAI said earlier this month its own team agents had cracked a million-dollar problem in mathematics, but that a Scientific American article later asked whether it was the math problem that really mattered. It said the piece did not argue OpenAI's solution was wrong, but that the result may not be the one mathematicians care most about. The outlet said a mathematician also accused the models of using some of his work without credit.

It said biologist Lucas Harrington, who wrote the post critiquing Anthropic's announcement, suggested AI companies should set the bar high now so that when an AI actually discovers a fundamentally new biological mechanism, everyone appreciates how big a deal it is. MIT Technology Review said that as OpenAI's Sam Altman and Anthropic's Dario Amodei race to one-up each other, raising the bar for scientific breakthroughs by AI might be the last thing on their minds.

Full Article: When can we say AI made a scientific discovery?