PeopleofSuperintelligence

Researcher

Philip Ball

Also known as Philip J. Ball, Phil Ball

Philip Ball is a machine-learning researcher whose work covers reinforcement learning and world models. He studied for a PhD at Oxford under Steve Roberts, earned master’s degrees in machine learning from Cambridge and engineering science from Oxford, and held earlier roles at Accenture, Microsoft Research Cambridge, Waymo and UC Berkeley’s RAIL Lab. As a research scientist at Google DeepMind, he contributed to Genie 2 and Genie 3, interactive world models generated from prompts. Since August 2026, he is a co-founder of Emulate, a UK world-model startup; in September 2026, the company was in advanced talks to raise up to $700 million.

Now at
Google DeepMind and Emulate
Born
1962
Works
2
Badges

Years link to their sources. Publication numbers from OpenAlex, as of 22 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2021

    Oxford training and offline RL

    At Oxford, he studies for a PhD under Steve Roberts in the Machine Learning Research Group and holds master’s degrees in machine learning from Cambridge and engineering science from Oxford. Earlier roles include management consulting at Accenture, an AI residency at Microsoft Research Cambridge, internships at Waymo and Microsoft Research, and a visiting-student position at UC Berkeley’s RAIL Lab with Sergey Levine. In a 2021 preprint, “Revisiting Design Choices in Offline Model-Based Reinforcement Learning,” he works with Cong Lu, Jack Parker-Holder, Michael A. Osborne and Stephen Roberts to compare uncertainty heuristics and tune key hyperparameters with Bayesian optimization.

  2. June 2022

    A benchmark for visual RL

    In June 2022, he and Cong Lu, Tim G. J. Rudner and coauthors post “Challenges and Opportunities in Offline Reinforcement Learning from Visual Observations.” The paper establishes simple continuous-control baselines and introduces a suite of benchmark tasks for offline reinforcement learning from visual data.

  3. July 2022

    A-LIX for pixel-based RL

    In July 2022, he coauthors “Stabilizing Off-Policy Deep Reinforcement Learning from Pixels” with Edoardo Cetin, Steve Roberts and Oya Celiktutan; the paper is a short presentation at ICML 2022. The authors call a failure mode catastrophic self-overfitting and propose A-LIX, which outperforms prior state-of-the-art methods on DeepMind Control and Atari 100k.

  4. February 2023

    Efficient online RL

    In February 2023, he posts “Efficient Online Reinforcement Learning with Offline Data,” a first-author paper with Laura Smith, Ilya Kostrikov and Sergey Levine. It appears at the 40th International Conference on Machine Learning in 2023; minimal changes to off-policy methods yield a 2.5× improvement over existing approaches across competitive benchmarks.

  5. March 2023

    Synthetic Experience Replay

    In March 2023, he coauthors “Synthetic Experience Replay” with Cong Lu, Yee Whye Teh and Jack Parker-Holder; the paper appears at NeurIPS 2023. SynthER uses a diffusion-based approach to upsample an agent’s collected experience.

  6. December 2024

    Genie 2’s promptable worlds

    In December 2024, Google DeepMind introduces Genie 2, a foundation world model that generates action-controllable, playable 3D environments from one prompt image. As a research scientist there, Ball contributes to the project; Jack Parker-Holder leads it, with Stephen Spencer in technical leadership.

  7. August 2025

    Genie 3 in real time

    In August 2025, Google DeepMind unveils Genie 3, a general-purpose world model that generates text-prompted worlds people can navigate at 24 frames per second and 720p. As a research scientist there, Ball contributes key research and engineering to the model.

  8. Since August 2026

    Co-founding Emulate

    Since August 2026, he is a co-founder of Emulate, a UK AI startup pursuing world-model research, with Jack Parker-Holder and Matthew McGill; all three worked on DeepMind’s Genie models. In September 2026, Emulate is in advanced talks to raise up to $700 million in a seed round, with Index Ventures and Lightspeed Venture Partners expected to lead.

Career

  1. Since 2026EmulateCo-Founder
  2. NowGoogle DeepMindResearch Scientist
  3. Microsoft ResearchAI Resident
  4. WaymoIntern
  5. University of California, BerkeleyVisiting Student

Education

  1. University of OxfordPhD
  2. University of CambridgeMaster's degree in machine learning

From the biography.

  1. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  2. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  3. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Google. 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.

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 Philip Ball yet. Send us a free-licensed photo.

  1. Philip J. Ball | Research Scientist @ Google DeepMind philipjball.github.io. Checked 21 October 2026.
  2. Revisiting Design Choices in Offline Model-Based Reinforcement Learning arxiv.org. Checked 21 October 2026.
  3. Challenges and Opportunities in Offline Reinforcement Learning from Visual Observations arxiv.org. Checked 21 October 2026.
  4. Stabilizing Off-Policy Deep Reinforcement Learning from Pixels arxiv.org. Checked 21 October 2026.
  5. Efficient Online Reinforcement Learning with Offline Data arxiv.org. Checked 21 October 2026.
  6. Efficient Online Reinforcement Learning with Offline Data proceedings.mlr.press. Checked 21 October 2026.
  7. Synthetic Experience Replay arxiv.org. Checked 21 October 2026.
  8. Genie 2: A large-scale foundation world model — Google DeepMind deepmind.google. Checked 21 October 2026.
  9. Genie 3: A new frontier for world models — Google DeepMind deepmind.google. Checked 21 October 2026.
  10. Former DeepMind world-model researchers’ startup Emulate in talks to raise $700M thenextweb.com. Checked 21 October 2026.
  11. Month-Old UK Start-Up Achieves Nearly $4bn Valuation | Silicon silicon.co.uk. Checked 21 October 2026.