Sam Staton

Professor in University of Oxford

Sam Staton is Professor of Computer Science at the University of Oxford. His research is on programming language theory, but with a view to applications, including quantum computing, statistics, and machine learning. He currently holds various grants, including an ERC Consolidator Grant, "Better Languages for Statistics", and UK ARIA grants "Employing Categorical Probability Towards Safe AI" and "SynthStats: GFlowNet-finetuning for synthesizing world models for safe AI". He has previously worked in Cambridge, Paris and the Netherlands.

Talk

Title: Expressive probabilistic programming, and world models for safe AI

Abstract: Many have proposed that AI alignment needs to refer to formal world models. These world models should often be complex and statistical, and so probabilistic programming is a convenient formal method for specifying them. Standard programming abstractions help in organizing and auditing such world models. I will discuss some expressive methods in probabilistic programming and possible generalizations towards robust statistics.