PeopleofSuperintelligence

Researcher

Arthur Szlam

Also known as Arthur D. Szlam

Arthur Szlam is a computer scientist whose work spans graph learning, image generation and language-driven agents. He earned a computer science PhD at Yale, completed postdoctoral work at UCLA and Minnesota, taught mathematics at City College of New York and received a 2013 Sloan Research Fellowship. His research moved from spectral networks, memory networks and a Laplacian-pyramid GAN to Minecraft agents and BlenderBot 3. From 2024 to 2025, he coauthored distributed-training research at Google DeepMind and Google Research, then advised Decoupled DiLoCo in April 2026.

Works
139
Citations
18,997
h-index
48
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2013

    A Sloan Fellowship at CCNY

    After earning a computer science PhD at Yale, he completes postdoctoral appointments at UCLA's Institute for Pure and Applied Mathematics and the University of Minnesota's Institute for Mathematics and Its Applications, then becomes an assistant professor in City College of New York's mathematics department. His research includes applications to machine learning, and he receives a Sloan Research Fellowship in 2013.

  2. 2013

    Spectral networks on graphs

    In 2013, he coauthors Spectral Networks and Locally Connected Networks on Graphs, showing how to learn convolutional layers for low-dimensional graphs with parameter counts independent of input size.

  3. 2015

    Memory networks and image generation

    In 2015, he coauthors End-To-End Memory Networks, a recurrent-attention model over external memory trained end to end for question answering and language modeling with less supervision. He also coauthors a Laplacian-pyramid GAN that generates natural images from coarse to fine.

  4. 2019

    CraftAssist in Minecraft

    In 2019, he is a research scientist at Facebook AI Research. He coauthors CraftAssist, a Minecraft bot assistant whose platform lets players interact with it and records their exchanges to study agents that complete tasks through dialogue.

  5. 2022

    BlenderBot 3

    In 2022, he coauthors BlenderBot 3, a 175-billion-parameter dialogue model with internet access and long-term memory that the team deploys on a public web page for users.

  6. 2024

    DiPaCo distributed training

    In 2024, he coauthors DiPaCo, a modular training approach that combines path composition with low-communication DiLoCo optimization to train across poorly connected and potentially heterogeneous workers.

  7. 2025

    Scaling laws for DiLoCo

    In 2025, he coauthors a study that finds DiLoCo scales predictably and robustly with model size, and can outperform data-parallel training when well tuned.

  8. April 2026

    Advising on Decoupled DiLoCo

    In April 2026, he provides additional support and advising for Decoupled DiLoCo, which runs language-model training across distant data centers and isolates the effects of hardware failures.

Career

  1. 2019FAIR (Meta)Research Scientist
  2. 2014–2023Meta
  3. 2013–2021City College of New YorkAssistant Professor
  4. 2012–2014City University of New York
  5. 2008–2009University of California, Los Angeles

Education

  1. Yale UniversityPhD

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

  1. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Meta. 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.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Meta. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  4. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Meta. 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.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Arthur Szlam (Facebook Research) – Language and Interaction in Minecraft cs.uchicago.edu. Checked 20 October 2026.
  2. Arthur Szlam wins Sloan Fellowship math.sci.ccny.cuny.edu. Checked 20 October 2026.
  3. Spectral Networks and Locally Connected Networks on Graphs arxiv.org. Checked 20 October 2026.
  4. End-To-End Memory Networks arxiv.org. Checked 20 October 2026.
  5. Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks arxiv.org. Checked 20 October 2026.
  6. CraftAssist: A Framework for Dialogue-enabled Interactive Agents arxiv.org. Checked 20 October 2026.
  7. BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage arxiv.org. Checked 20 October 2026.
  8. DiPaCo: Distributed Path Composition deepmind.google. Checked 20 October 2026.
  9. Communication-Efficient Language Model Training Scales Reliably and Robustly: Scaling Laws for DiLoCo research.google. Checked 20 October 2026.
  10. Decoupled DiLoCo: A new frontier for resilient, distributed AI training deepmind.google. Checked 20 October 2026.