When can we say AI made a scientific discovery?
Anthropic says 950 Claude agents found a DNA pattern, but biologists say that is not a scientific discovery.
Anthropic said 950 Claude agents, after 21 hours, flagged a previously uncatalogued repeating pattern around a known enzyme and likened it to patterns behind CRISPR. Biologists, including a critic endorsed by Eli Lilly's chair and CEO, said finding such a cluster is routine work, not a discovery. University of Copenhagen researcher Mario Rodriguez Mestre said his team had already reported the pattern and stopped using Claude over fears it learned from his chats, which Anthropic denies. The essay similarly questions whether OpenAI's claimed solution to a million-dollar math problem is the result mathematicians most value.
- Anthropic said 950 Claude agents flagged an uncatalogued enzyme-adjacent pattern in 21 hours.
- Critics, backed by Eli Lilly's CEO, said pattern spotting is not a discovery.
- A Copenhagen biologist said his team found it first and quit Claude.
- Essay also doubts the importance of OpenAI's claimed math breakthrough.
- Organizations
- AnthropicOpenAIEli LillyUniversity of Copenhagen
- AI models
- Claude
From the original article
3 min read at technologyreview.com
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Last Wednesday, Anthropic announced that earlier this year it had launched a molecular biology lab, where Claude agents read and conjecture about hard biology problems and human scientists run experiments on what…
By James O'Donnell · technologyreview.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.