When can we say AI made a scientific discovery?
Executive Take
AI vendors are now marketing research output like product launches, and that invites scrutiny their claims can't survive. Leaders evaluating AI for R&D should judge the actual work done, not the press release framing.
Executive Summary
Anthropic said its 950-agent AI system found a new biological pattern near a known gene-editing-related enzyme in 21 hours, calling it a discovery. Biologists disputed this, saying flagging a pattern isn't the hard part, and a Copenhagen researcher claimed his team found it first. OpenAI faced similar pushback on a math claim.
Why It Matters
AI leaders need to separate genuine AI capability gains from marketing hype before making R&D investment or partnership decisions. Technology and HR leaders training staff on AI tools should also note the growing credibility gap between vendor claims and expert review.
Bizquad Perspective
The real story isn't whether Claude "discovered" anything, it's that AI labs are now competing on scientific PR instead of letting independent verification build trust slowly.