ABAB Pedia

Dan Hendrycks

Overview

Dan Hendrycks is an AI researcher and executive director of the Center for AI Safety. He contributed to evaluations including MMLU and has worked on machine learning reliability, AI policy and education.⁠[1][2]

Contents9 sections
Key facts

Topic: From Missouri to Berkeley

Topic: Research, education and public policy

Hendrycks contributed to the GELU activation function and to MMLU, an evaluation of a model’s ability to answer questions across academic subjects. He studied the reliability of models under changing inputs and environments, helped organize ICML workshops on robustness and uncertainty from 2019 through 2021, and wrote teaching material on AI safety, ethics and society. He became executive director of the Center for AI Safety and lists advisory work for xAIxAIxAI is an artificial intelligence company founded by Elon Musk in 2023 that develops the Grok family of large language models and chatbots. It acquired the social platform X in an all-stock deal in March 2025, was acquired by SpaceX in February 2026, and as SpaceX’s AI division has since been renamed SpaceXAI and then SpaceXSI.Open the full entry, Scale AI and Meta.⁠[1][5]

Interview: No Priors

Interview: No Priors

The October 31, 2025 highlights episode of No Priors replayed his discussion of advanced AI and geopolitics.⁠[6]

Advice Shared on No Priors

He argued that closed-ended questions cover only part of academic ability and recommended separately testing whether a model actually finishes an open-ended task over a longer period. (32:25)⁠[2]

He advocated developing new evaluations as existing tests saturate so the public can still track where and how quickly capabilities advance. (33:10)⁠[2]

Related Organizations, Websites and Public Accounts

Center for AI Safety: Executive director. Official website: https://safe.ai/ (opens in a new window).⁠[1]

Interview with No Priors

He recalled asking professors and researchers worldwide for questions resembling difficult problems in their work but having definite, checkable answers for Humanity's Last Exam.⁠[2]

Sources6 sources

  1. Dan Hendrycks (opens in a new window)danhendrycks.com↑
  2. No Priors Ep. 105 | With Director of the Center of AI Safety Dan Hendrycks (opens in a new window)youtube.com↑
  3. Dan Hendrycks: Curriculum Vitae (opens in a new window)people.eecs.berkeley.edu↑
  4. Machine Learning Safety (opens in a new window)www2.eecs.berkeley.edu↑
  5. Dan Hendrycks (opens in a new window)safe.ai↑
  6. No Priors Ep. 138 | The Best of 2025 (So Far) with Sarah Guo and Elad Gil (opens in a new window)youtube.com↑