PeopleofSuperintelligence

Lab leader

Greg Yang

Greg Yang is a mathematician and AI researcher whose Tensor Programs theory describes how very wide neural networks behave and how their training settings carry across model sizes. After earning a mathematics degree at Harvard in 2018, where his undergraduate thesis won the Hoopes Prize, he began the Tensor Programs series in 2019. The work produced the maximal update parametrization, or µP, and in 2022, as a senior researcher at Microsoft Research, he showed with OpenAI collaborators that hyperparameters tuned on a small model transfer to a 6.7-billion-parameter GPT-3. In July 2023, he was named one of the 12 founding members of Elon Musk's xAI. He stepped down from xAI after a Lyme disease diagnosis in January 2026 and is an informal adviser to the company.

Timeline

  1. 2018

    Mathematics at Harvard

    He earns a bachelor's degree in mathematics from Harvard University in 2018, and his undergraduate thesis wins the Hoopes Prize. In 2018, he also receives an honorable mention for the Morgan Prize for outstanding undergraduate research in mathematics, for work on a homological theory of functions.

  2. 2019

    The first Tensor Programs papers

    In February 2019, he introduces tensor programs, a way to write most neural network computations and derive their limits when the tensors are large and random. His paper Tensor Programs I, presented at NeurIPS 2019, shows that wide random networks built from RNNs, convolutions, attention, normalization and other modern layers behave as Gaussian processes. He is an invited speaker at the International Congress of Chinese Mathematicians in 2019.

  3. November 2020

    Feature learning and µP

    In November 2020, he and Edward J. Hu post Feature Learning in Infinite-Width Neural Networks, the fourth Tensor Programs paper, which appears at ICML 2021. It proposes simple changes to the standard parametrization so that infinitely wide networks learn features, and the analysis yields the maximal update parametrization, known as µP.

  4. March 2022

    µTransfer with OpenAI

    In March 2022, as a senior researcher at Microsoft Research, he publishes Tensor Programs V with Edward Hu and researchers at OpenAI. Its µTransfer method tunes hyperparameters on a small model and transfers them to the full-size one; tuned from a 40-million-parameter model, it beats published results for the 6.7-billion-parameter GPT-3 at a tuning cost of 7% of total pretraining. The team releases the method as the PyTorch package mup.

  5. July 2023

    Founding member of xAI

    In July 2023, Elon Musk announces xAI with 12 founding members, among them Yang, a former researcher at Microsoft Research. At xAI he works as a mathematician on Tensor Programs, aiming for a theory of everything for large-scale deep learning that shows the optimal way of scaling neural networks and guides safety and alignment efforts.

  6. October 2023

    Depth-µP in Tensor Programs VI

    In October 2023, he coauthors Tensor Programs VI with Dingli Yu, Chen Zhu and Soufiane Hayou. For residual networks with one layer per block, the paper identifies Depth-µP, an extension of µP that lets hyperparameters transfer across depth.

  7. Since January 2026

    Informal adviser to xAI

    In January 2026, he steps down from his role at xAI after a Lyme disease diagnosis. He becomes an informal adviser to the company while focusing on his recovery.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI from 1961 to 2025, with a year-by-year chart.
  2. The United States employs an estimated 53% of people in our data with a known employment countryOur estimates of where AI careers are concentrated now and how countries’ shares of dated positions changed over time.
  3. 46% of living people in our data with known birth and employment countries work outside their birth countryOur estimate of cross-border careers in AI, from country of birth and the country of each current organization.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionOur estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

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 Greg Yang yet. Send us a free-licensed photo.

  1. Greg Yang | Large N Limits: Random Matrices & Neural Networks | The Cartesian Cafe cartesiancafe.podbean.com. Checked 19 October 2026.
  2. Morgan Prize - Wikipedia en.wikipedia.org. Checked 19 October 2026.
  3. [1902.04760] Scaling Limits of Wide Neural Networks with Weight Sharing: Gaussian Process Behavior, Gradient Independence, and Neural Tangent Kernel Derivation arxiv.org. Checked 19 October 2026.
  4. Tensor Programs I: Wide Feedforward or Recurrent Neural Networks of Any Architecture are Gaussian Processes - Microsoft Research microsoft.com. Checked 19 October 2026.
  5. [2011.14522] Feature Learning in Infinite-Width Neural Networks arxiv.org. Checked 19 October 2026.
  6. Greg Yang | Professional page thegregyang.com. Checked 19 October 2026.
  7. µTransfer: A technique for hyperparameter tuning of enormous neural networks - Microsoft Research microsoft.com. Checked 19 October 2026.
  8. [2203.03466] Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer arxiv.org. Checked 19 October 2026.
  9. Elon Musk Announces xAI and Its 12-Man Team: Here’s Who They Are | Observer observer.com. Checked 19 October 2026.
  10. [2310.02244] Tensor Programs VI: Feature Learning in Infinite-Depth Neural Networks arxiv.org. Checked 19 October 2026.
  11. XAI co-founder Yang steps down following Lyme disease diagnosis theheadandtale.com. Checked 19 October 2026.
  12. xAI Co-founder Greg Yang Steps Aside Into Informal Advisory Role Following Health Concerns officechai.com. Checked 19 October 2026.