What does an imaginary fluid that isn’t made up of atoms have to teach chemists? Perhaps not much—but the community studying that fluid offers a warning about the risks and opportunities of adopting AI for research.
In September, a team from OpenAI announced that it had found the solution to a Millennium Prize Problem in mathematics. This question asks whether a system of equations describing the motion of fluids could ever break down and incorrectly predict those fluids’ acceleration to impossible speeds. According to OpenAI, AI agents running on a new model took just days to find that.
