Cosponsored by the Black Hole Initiative and the Harvard Department of Physics
February 17, 2026
2-3 pm
Harvard Science Center A
1 Oxford St., Cambridge, MA
Large language models are rapidly becoming useful partners in scientific work: they can now reason through complex problems, carry out long computations, and even provide valuable insights. These emerging capabilities are beginning to compress the research loop from idea to verified result. I will give some examples of what frontier models can now do in practice, focusing in particular on a novel theoretical-physics result that was very recently obtained with the help (and formal co-authorship) of an internal OpenAI model. Finally, I will discuss the growing role of AI in scientific discovery, and what it could mean for the pace of science over the next few years.