SCIENCEDECRYPT
Researchers Tried Letting AI Do Science. It Failed
A multi-institution study found that current frontier AI agents can perform research mechanics but cannot generate original scientific work meeting the standards of a top AI conference.
Related Signal
Adjacent reporting
- Human scientists trounce the best AI agents on complex tasks
- AI Boosts Research Careers but Flattens Scientific Discovery
- AI ‘scientists’ promise to accelerate research — how do they work?
- The uncritical adoption of AI in science is alarming — we urgently need guard rails
- Why AI cannot do good science without humans