02/09/2026
Mathematics Research Seminar Series
Intro to Technical AI Safety and Alignment Research
David Demitri Africa (BS AMDSc 2023, M DSc 2024)
Research Scientist, Resolution
11 September 2026 (F), 5–6 PM
Online
Abstract:
In this talk, I’ll give an introduction to technical AI safety and alignment research: the problem of understanding and shaping how increasingly capable AI systems behave. I’ll discuss questions like: How do we tell what a model has learned or “wants”? Why can models behave differently when they think they are being evaluated? How can training make undesirable behaviours emerge? And, why should we care? I’ll use examples from my own research to illustrate what work in the field actually looks like. I’ll also discuss why I think this is an exciting and precipitous time to work on these problems and give some practical advice for getting started. The talk is intended for a broad technical audience in the Philippines and assumes little prior exposure to AI safety.
About the speaker:
David Demitri Africa is a research scientist at Resolution, where he studies the character and motivations of AI systems, and how we might understand and shape them as they become increasingly capable. Previously, he was a research scientist on the Alignment team at the UK AI Security Institute. David completed an MPhil in Advanced Computer Science at the University of Cambridge, graduating with distinction, and previously studied Applied Mathematics and Data Science at Ateneo de Manila University. Outside work, he was the 2023 World Universities Debating Champion, and enjoys bouldering and reading.
