BlueDot Grant
Game theory is by definition fundamentally significant for any kind of strategic interaction. The way AI agents interact with humans and each other will define, over the next decade, the fate of our civilization. Mechanism design especially, together with formal verification, can serve as a bedrock for AI safety. Beyond that, strategic reasoning capabilities will define to what extent AI will be able to outsmart us, or on an even more fundamental level, whether AI will be predictable or chaotic.
I will spend this year working on the intersection of game theory and frontier AI. Specifically, I'll start working on reasoning post-training and systematically reaching out to collaborators.
Throughout my one-year grant period, I will:
- produce the equivalent of three papers that will be accepted to a top conference or journal,
- publish one long-form technical essay every month,
- send weekly bullet point updates.