PeopleofSuperintelligence

Researcher

Jonathan Gray

Jonathan Gray is an AI researcher at Anthropic who earlier worked on Diplomacy-playing agents at Facebook AI Research and Meta. After studying electrical and computer engineering at the University of Toronto and working at AeroFS and Magic, they were welcomed to OpenAI in January 2017. In 2020, they were first author of a Facebook AI Research paper on a no-press Diplomacy agent that ranked in the top 2% of human players on a popular Diplomacy website, and in 2022 they were among the authors of the Meta team's Science paper on CICERO. Their 2022 paper on human-regularized reinforcement learning and planning, which trained the Diplodocus agents, received an Honorable Mention at ICLR 2023.

Now at
Anthropic

Timeline

  1. By 2017

    Toronto, AeroFS and Magic

    They study electrical and computer engineering at the University of Toronto and, in high school and during their undergraduate years, build software for cancer treatment at Sunnybrook Research Institute's Focused Ultrasound Group. A peer-to-peer file sync tool at AeroFS and early engineering work at Magic, an SMS-based virtual personal assistant, follow.

  2. January 2017

    Joining OpenAI

    OpenAI's team update of January 2017 welcomes Gray as a full-time hire, when the team reaches 45 people.

  3. July 2019

    A Minecraft assistant

    In July 2019, they publish CraftAssist: A Framework for Dialogue-enabled Interactive Agents as first author, with Kavya Srinet, Yacine Jernite, Arthur Szlam and colleagues. The paper describes a bot assistant in Minecraft, with tools that let players interact with the bot and record those interactions, to support the study of agents that complete tasks specified by dialogue.

  4. October 2020

    Equilibrium search for Diplomacy

    In October 2020, they are first author of Human-Level Performance in No-Press Diplomacy via Equilibrium Search, written at Facebook AI Research with Adam Lerer, Anton Bakhtin and Noam Brown and published at ICLR 2021. The agent combines supervised learning on human data with one-step lookahead search via regret minimization, and it ranks in the top 2% of human players on a popular Diplomacy website.

  5. October 2022

    Diplodocus and DiL-piKL

    In October 2022, they coauthor Mastering the Game of No-Press Diplomacy via Human-Regularized Reinforcement Learning and Planning with Anton Bakhtin, David Wu, Adam Lerer and others. The paper introduces DiL-piKL, a planning algorithm that regularizes a reward-maximizing policy toward a human imitation-learned policy, and uses RL-DiL-piKL to train an agent named Diplodocus. In a 200-game tournament with 62 human participants, two Diplodocus agents score above every other participant who played more than two games and rank first and third in an Elo ratings model.

  6. 2022

    CICERO in Science

    In 2022, the Meta Fundamental AI Research Diplomacy Team publishes Human-level play in the game of Diplomacy by combining language models with strategic reasoning in Science, with Gray among the authors. Its agent, CICERO, integrates a language model with planning and reinforcement learning algorithms; across 40 games of an anonymous online Diplomacy league, it achieves more than double the average score of the human players and ranks in the top 10% of participants who played more than one game.

  7. 2023

    An ICLR honorable mention

    The 2022 paper on human-regularized reinforcement learning and planning receives an Honorable Mention at ICLR 2023, where it is also listed as a top 5% paper.

  8. As of October 2026

    Working at Anthropic

    They work at Anthropic, with research interests in artificial intelligence and machine learning systems.

Career

  1. NowAnthropicAI Researcher
  2. From 2022Meta
  3. From 2020FAIR (Meta)
  4. From 2017OpenAI

Education

  1. University of Toronto

From the biography.

  1. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Jonathan Gray. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.
  2. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Jonathan Gray. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  3. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Anthropic. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  4. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Anthropic. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Jonathan Gray yet. Send us a free-licensed photo.

  1. Team update (OpenAI, January 30, 2017) web.archive.org. Checked 21 October 2026.
  2. [1907.08584] CraftAssist: A Framework for Dialogue-enabled Interactive Agents arxiv.org. Checked 21 October 2026.
  3. [2010.02923] Human-Level Performance in No-Press Diplomacy via Equilibrium Search arxiv.org. Checked 21 October 2026.
  4. Human-Level Performance in No-Press Diplomacy via Equilibrium Search (HTML version) ar5iv.labs.arxiv.org. Checked 21 October 2026.
  5. Human-Level Performance in No-Press Diplomacy via Equilibrium Search | ML Anthology mlanthology.org. Checked 21 October 2026.
  6. [2210.05492] Mastering the Game of No-Press Diplomacy via Human-Regularized Reinforcement Learning and Planning arxiv.org. Checked 21 October 2026.
  7. Jonathan Gray - Google Scholar scholar.google.com. Checked 21 October 2026.
  8. Source: DeepMind Cicero research | Longterm Wiki longtermwiki.com. Checked 21 October 2026.
  9. ICLR 2023 Awards iclr.cc. Checked 21 October 2026.