Letting Claude Cook

Harvard FAS

Can a team of AI agents design an experiment, write and run the code, interpret the results, and then decide what to do next—all on their own? This is no longer a thought experiment. In this seminar, we will explore what happens when we give AI systems genuine autonomy to do science. We begin by building a working understanding of how large language models function—not just what they produce, but why they produce it. Drawing on perspectives from both physics and the emerging field of…