PeopleofSuperintelligence

Researcher

David Balduzzi

David Balduzzi is a machine learning researcher whose work on game dynamics in neural networks earned an ICML 2018 Best Paper Runner Up award. His PhD at the University of Chicago was in representation theory and algebraic geometry, and he then turned to computational neuroscience at UW-Madison, where he and Giulio Tononi published PLOS Computational Biology papers on integrated information in 2008 and qualia geometry in 2009. In machine learning, he worked at the MPI for Intelligent Systems, ETH Zürich and Victoria University Wellington, coauthoring papers on Domain-Invariant Component Analysis in 2013 and the shattered gradients problem in 2017. At DeepMind he researched game theory and machine learning, coauthoring 'Open-ended learning in symmetric zero-sum games' in 2019 and 'Real World Games Look Like Spinning Tops' in 2020. He works on machine learning at XTX Markets and serves as an action editor of Transactions on Machine Learning Research.

Badges

Timeline

  1. By 2008

    Integrated information at Wisconsin

    After a University of Chicago PhD in representation theory and algebraic geometry, he turns to computational neuroscience at UW-Madison. In June 2008, he and Giulio Tononi, both in the university's Department of Psychiatry, publish 'Integrated Information in Discrete Dynamical Systems' in PLOS Computational Biology, which introduces a time- and state-dependent measure of integrated information called φ.

  2. 2009

    Qualia as integrated information geometry

    In 2009, he and Giulio Tononi publish 'Qualia: The Geometry of Integrated Information' in PLOS Computational Biology, which describes a quale as a shape in a qualia space that characterizes the quality of a conscious experience.

  3. 2013

    Domain generalization with DICA

    His machine learning work spans the MPI for Intelligent Systems, ETH Zürich and Victoria University Wellington. In 2013, he coauthors 'Domain Generalization via Invariant Feature Representation' with Krikamol Muandet and Bernhard Schölkopf, which proposes Domain-Invariant Component Analysis (DICA), a kernel-based optimization algorithm that learns an invariant transformation by minimizing the dissimilarity across domains.

  4. 2015–2017

    The shattered gradients problem

    He sits on the NIPS 2015 program committee as a member from Wellington. In 2017, he and five coauthors, including Marcus Frean, identify the shattered gradients problem, in which the correlation between gradients in standard feedforward networks decays exponentially with depth.

  5. 2018

    A runner-up ICML paper

    ICML 2018 names 'The Mechanics of n-Player Differentiable Games', by David Balduzzi, Sebastien Racaniere, James Martens, Jakob Foerster, Karl Tuyls and Thore Graepel, a Best Paper Runner Up. The paper splits the second-order dynamics of games into a potential-game part and a Hamiltonian-game part, which motivates Symplectic Gradient Adjustment (SGA), a new algorithm for finding stable fixed points in general games.

  6. 2019

    Symmetric zero-sum games at ICML

    In 2019, with DeepMind coauthors, he publishes 'Open-ended learning in symmetric zero-sum games' at ICML 2019, proposing the rectified Nash response algorithm PSRO_rN and applying it to nontransitive resource allocation games.

  7. 2019–2020

    Game theory at DeepMind

    In December 2019, while working on game theory and machine learning at DeepMind, he gives the invited talk 'Composition, learning, and games' at the NeurIPS 2019 workshop on bridging game theory and deep learning. In 2020, he coauthors 'Real World Games Look Like Spinning Tops', which investigates the spinning-top geometry of real world games including Tic-Tac-Toe, Go and StarCraft II, and in December 2020 he gives the NeurIPS tutorial 'Designing learning dynamics' with M Garnelo and W Czarnecki.

  8. As of October 2026

    Machine learning at XTX Markets

    He works on machine learning at XTX Markets and serves as an action editor of Transactions on Machine Learning Research. In September 2026, he posts the single-author arXiv paper 'Differential Functionalism: Jacobians Shape the Structure of Phenomenal Experience', which introduces two measures of Jacobian structure, transparency and integrity.

Career

  1. NowXTX Markets
  2. 2019–2020Google DeepMind
  3. Until 2008University of Wisconsin–Madison

Education

  1. University of ChicagoPhD

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes David Balduzzi. 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.
  2. 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.
  3. 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.
  4. 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.

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 David Balduzzi yet. Send us a free-licensed photo.

  1. Smooth Games Optimization and Machine Learning Workshop: Bridging Game Theory and Deep Learning sgo-workshop.github.io. Checked 22 October 2026.
  2. Integrated Information in Discrete Dynamical Systems: Motivation and Theoretical Framework | PLOS Computational Biology journals.plos.org. Checked 22 October 2026.
  3. Qualia: The Geometry of Integrated Information ncbi.nlm.nih.gov. Checked 22 October 2026.
  4. [1301.2115] Domain Generalization via Invariant Feature Representation arxiv.org. Checked 22 October 2026.
  5. 2015 Committees nips.cc. Checked 22 October 2026.
  6. [1702.08591] The Shattered Gradients Problem: If resnets are the answer, then what is the question? arxiv.org. Checked 22 October 2026.
  7. ICML 2018 Awards icml.cc. Checked 22 October 2026.
  8. [1802.05642] The Mechanics of n-Player Differentiable Games arxiv.org. Checked 22 October 2026.
  9. Open-ended learning in symmetric zero-sum games proceedings.mlr.press. Checked 22 October 2026.
  10. [2004.09468] Real World Games Look Like Spinning Tops arxiv.org. Checked 22 October 2026.
  11. balduzzi sites.google.com. Checked 22 October 2026.
  12. Transactions on Machine Learning Research jmlr.csail.mit.edu. Checked 22 October 2026.
  13. [2609.09306] Differential Functionalism: Jacobians Shape the Structure of Phenomenal Experience arxiv.org. Checked 22 October 2026.