PeopleofSuperintelligence

Researcher

Ben Poole

Ben Poole is a machine-learning researcher whose work links neural computation with information theory, diffusion models, and 3D generation. He studied computer science at Carnegie Mellon and completed a PhD at Stanford with Surya Ganguli's Neural Dynamics and Computation lab. He coauthored Gumbel-Softmax work for categorical variables, research on mutual-information bounds, and a score-based diffusion framework that received ICLR's Outstanding Paper Award in 2021. DreamFusion applied a pretrained image-diffusion model to text-to-3D generation and received ICLR's Outstanding Paper Award in 2023. His earlier roles included Google Brain, Intel Research Pittsburgh, and NYU's Center for Neural Science; as of October 2026, he is a research scientist at Google DeepMind and leads GenMedia 3D, with work on CAT4D and generative 3D for shoppable products.

Works
67
Citations
15,399
h-index
35
Badges

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

  1. 2016

    Deep network expressivity

    In 2016, he coauthors a study of deep neural networks that combines Riemannian geometry with mean-field theory to analyze signal propagation. His training includes undergraduate computer-science study at Carnegie Mellon with Tai Sing Lee and a Stanford PhD in Surya Ganguli's Neural Dynamics and Computation lab.

  2. 2017

    The Gumbel-Softmax estimator

    By 2017, he coauthors Categorical Reparameterization with Gumbel-Softmax, an efficient gradient estimator that lets neural networks backpropagate through categorical samples.

  3. 2019

    Mutual-information bounds

    In 2019, he coauthors a framework for estimating mutual information that compares variational bounds and introduces a continuum of lower bounds with a bias-variance tradeoff.

  4. 2021

    Score-based diffusion

    In 2021, he coauthors a score-based generative modeling framework using forward and reverse stochastic differential equations to generate samples. The paper receives ICLR's Outstanding Paper Award.

  5. 2023

    DreamFusion for text-to-3D

    In 2023, he coauthors DreamFusion, which uses a pretrained text-to-image diffusion model to optimize a 3D scene from text. The paper receives ICLR's Outstanding Paper Award.

  6. As of October 2026

    Leading GenMedia 3D

    He is a research scientist at Google DeepMind and leads the GenMedia 3D team. Earlier roles include Google Brain, Intel Research Pittsburgh, and NYU's Center for Neural Science. In 2025, he coauthors CAT4D, which creates dynamic 3D scenes from monocular video, and contributes to Google's generative 3D work for shoppable products.

Career

  1. Since 2024Google DeepMindResearch Scientist; Lead, GenMedia 3D team
  2. 2017–2025Google

Education

  1. Carnegie Mellon University
  2. Stanford UniversityPhD

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. Categorical Reparameterization with Gumbel-Softmax2016 · 3,616 citations
  2. Score-Based Generative Modeling through Stochastic Differential Equations2020 · 1,950 citations
  3. Continual Learning Through Synaptic Intelligence2017 · 1,783 citations
  4. Adversarially Learned Inference2016 · 1,015 citationsWith Aaron C. Courville
  5. What Makes for Good Views for Contrastive Learning?2020 · 775 citationsWith Cordelia Schmid and Phillip Isola

From OpenAlex, as of 19 October 2026.

  1. The median h-index is 54 at universities and 19 at frontier AI labsIncludes Ben Poole. The median h-index is 54 at universities and 19 at frontier AI labs. We compare publishing careers, citations and moves to estimate how universities and frontier labs contribute to AI research.
  2. Google has the highest combined h-index among AI organizationsIncludes Google. Our estimates of combined and median h-index at AI organizations, using current affiliations and OpenAlex records.
  3. Google alumni founded or led 81 other organizations, the most among the labs we compareIncludes Google. We trace 194 organizations founded or led by AI lab alumni, and 202 career moves between the labs.
  4. 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 from 1961 to 2025, with a year-by-year chart.

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 Ben Poole yet. Send us a free-licensed photo.

  1. Ben Poole cs.stanford.edu. Checked 18 October 2026.
  2. Exponential expressivity in deep neural networks through transient chaos arxiv.org. Checked 18 October 2026.
  3. Categorical Reparameterization with Gumbel-Softmax arxiv.org. Checked 18 October 2026.
  4. On variational bounds of mutual information research.google. Checked 18 October 2026.
  5. Score-Based Generative Modeling through Stochastic Differential Equations arxiv.org. Checked 18 October 2026.
  6. Yang Song yang-song.net. Checked 18 October 2026.
  7. DreamFusion: Text-to-3D using 2D Diffusion dreamfusion3d.github.io. Checked 18 October 2026.
  8. Announcing the ICLR 2023 Outstanding Paper Award Recipients blog.iclr.cc. Checked 18 October 2026.
  9. CAT4D: Create Anything in 4D with Multi-View Video Diffusion Models openaccess.thecvf.com. Checked 18 October 2026.
  10. Bringing 3D shoppable products online with generative AI research.google. Checked 18 October 2026.