The AI games are on…
The fastest way to improve AI now is to let agents learn in arenas that reward long-horizon planning and social reasoning. And that’s the plan.
We’re collaborating with
@Princeton,
@UTAustin,
@Meta,
@Radboud_Uni,
@nyuniversity,
@MIT_CSAIL and the Agency for Science, Technology and Research to bring a unified arena to test strategic, multi-agent, social reasoning for LLM agents (think deception, cooperation, coalition-building, and conflict).
🧵 Read more about the competition we’re hosting
@NeurIPSConf and why we’re excited about AI playing games.