Human-level play in the game of Diplomacy by combining language models with strategic reasoning.

Autor: Bakhtin A; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Brown N; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Dinan E; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Farina G; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Flaherty C; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Fried D; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.; Language Technologies Institute, Carnegie Mellon University, Pittsburgh, PA, USA., Goff A; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Gray J; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Hu H; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.; Department of Computer Science, Stanford University, Stanford, CA, USA., Jacob AP; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.; Computer Science and Artificial Intelligence Laboratory, Massachusetts Insititute of Technology, Cambridge, MA, USA., Komeili M; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Konath K; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Kwon M; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.; Department of Computer Science, Stanford University, Stanford, CA, USA., Lerer A; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Lewis M; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Miller AH; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Mitts S; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Renduchintala A; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Roller S; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Rowe D; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Shi W; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.; Department of Computer Science, Columbia University, New York, NY, USA., Spisak J; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Wei A; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.; Department of Computer Science, University of California, Berkeley, Berkeley, CA, USA., Wu D; Meta AI, 1 Hacker Way, Menlo Park, CA, USA., Zhang H; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.; EconCS Group, Harvard University, Cambridge, MA, USA., Zijlstra M; Meta AI, 1 Hacker Way, Menlo Park, CA, USA.
Jazyk: angličtina
Zdroj: Science (New York, N.Y.) [Science] 2022 Dec 09; Vol. 378 (6624), pp. 1067-1074. Date of Electronic Publication: 2022 Nov 22.
DOI: 10.1126/science.ade9097
Abstrakt: Despite much progress in training artificial intelligence (AI) systems to imitate human language, building agents that use language to communicate intentionally with humans in interactive environments remains a major challenge. We introduce Cicero, the first AI agent to achieve human-level performance in Diplomacy , a strategy game involving both cooperation and competition that emphasizes natural language negotiation and tactical coordination between seven players. Cicero integrates a language model with planning and reinforcement learning algorithms by inferring players' beliefs and intentions from its conversations and generating dialogue in pursuit of its plans. Across 40 games of an anonymous online Diplomacy league, Cicero achieved more than double the average score of the human players and ranked in the top 10% of participants who played more than one game.
Databáze: MEDLINE
Nepřihlášeným uživatelům se plný text nezobrazuje