On AI:AM

Peter Jansen

Research Scientist, Allen Institute for AI (Ai2); Associate Professor, University of Arizona

Peter Jansen is a research scientist at the Allen Institute for AI (Ai2) and an associate professor at the University of Arizona. He builds the benchmarks and systems — ScienceWorld, DiscoveryWorld, CodeScientist, and Theorizer — that measure whether AI agents can actually carry out scientific discovery. His Theorizer project generated roughly 3,000 organizing theories by reading ~14,000 papers; his cautionary findings on CodeScientist (a claimed discovery that turned out to be analyzing a random-number generator) make the case that verification, not idea generation, is the bottleneck for AI science.