Abstract: The Bayesian Revolution of the last two decades has transformed cognitive science, enabling us to tie together a vast array of disparate phenomena under the rubric of evidence gathering in the face of uncertainty, in an elegant mathematical framework that just happens to connect to the technological advances in AI. It contains, however, a crucial gap: while it tells us how new experiences ought to change our beliefs, it gives us no systematic way to account for the mental load of actually doing so—which means that our most crucial questions appear as ad hoc corrections. I’ll present recent progress towards unification, which draws on foundational results—and novel conjectures—in computational complexity, cryptography, and formal theorem proving.
Speaker
Simon DeDeoSFI External Professor & Associate Professor at Carnegie Mellon University