James Owen Weatherall, UC, Irvine

Title: John von Neumann and the Prehistory of Artificial Intelligence

Abstract: John von Neumann is well-known for his foundational contributions to many fields, from pure mathematics, to physics, economics, computer science, operations research, and nuclear policy.  Among the most impactful was a report, drafted in Spring 1945, sketching a design proposal for an electronic digital stored-program computer that would become the blueprint for virtually all electronic computer projects developed in the years immediately following World War II.  In this talk, I will track two separate strands that led to and shaped von Neumann's interest in computation: one, through his conversations and collaborations with David Hilbert, Kurt Godel, Max Newman, and Alan Turing during the 1920s and 1930s; and the other through his work on operations research and partial differential equations during World War II.  These two strands intersected at a surprising juncture: a series of discussions, beginning in January 1945, with Warren McCulloch and Walter Pitts, who had developed the first mathematical model of a neural network in 1943.  I will describe how von Neumann took from McCulloch and Pitts the idea that both biological neurological systems and electronic computers could be seen as Turing machines, and how this led him to investigate the connections between automated computation, agentic systems, and ultimately the possibility of self-reproducing systems.