Richard Ngo
Existing frameworks for understanding intelligent agency don't do a great job at describing multi-agent dynamics (e.g. agents recursively modeling each other, merging with each other, threatening each other, etc). Most work in my stream aims (implicitly or explicitly) to move towards a multi-agent understanding of intelligence.
Stream overview
My MATS mentees have a lot of freedom in what they work on; my only constraint is that they should aim to develop a clear conceptual understanding of something interesting. I'll be able to provide higher-quality mentorship on topics that are more related to my own research interests. Some recent projects they've taken on:
- Understanding the relationship between your past, present and future self as a mixed-motive game.
- Characterizing neural circuits using evolutionary models of multi-level selection.
- Extending existing results on Lobian cooperation to the probabilistic setting.
Standard (1-2 hours of weekly 1:1s)
SF Bay Area
Indifferent
Strong preference
Mentors
I previously worked on the alignment team at DeepMind, and on the governance team at OpenAI. I'm currently an independent researcher focusing on multi-agent intelligence. My research is in the tradition of natural philosophy; I'm trying to develop vague intuitive concepts (like trust, identity, and integrity) to the point where they can serve as seeds for new scientific paradigms.
Fellows we are looking for
No required qualifications—I'm looking for scholars who are capable of very clear and curious thinking, but I'm open to many ways that they might demonstrate that.
Project selection
I will talk through project ideas with the scholar.