Draft:Sam McCandlish
Machine learning researcher and Anthropic co-founder
From Wikipedia, the free encyclopedia
Sam McCandlish is a machine learning researcher and theoretical physicist who co-founded Anthropic.[1] He previously worked at OpenAI, where he co-authored research on large-batch training and scaling behaviour in neural language models.[2][3]
 | Review waiting, please be patient.
This may take 8 weeks or more, since drafts are reviewed in no specific order. There are 3,271 pending submissions waiting for review.
Where to get help
How to improve a draft
You can also browse Wikipedia:Featured articles and Wikipedia:Good articles to find examples of Wikipedia's best writing on topics similar to your proposed article. Improving your odds of a speedy review To improve your odds of a faster review, tag your draft with relevant WikiProject tags using the button below. This will let reviewers know a new draft has been submitted in their area of interest. For instance, if you wrote about a female astronomer, you would want to add the Biography, Astronomy, and Women scientists tags. Editor resources
Reviewer tools
|
Sam McCandlish | |
|---|---|
| Alma mater | Brandeis University Stanford University |
| Known for | Neural-network scaling laws; Anthropic's Responsible Scaling Policy |
| Scientific career | |
| Fields | Machine learning Theoretical physics |
| Institutions | OpenAI Anthropic |
Education
Career
Theoretical physics
McCandlish co-authored work in high-energy theory and holography, including papers on integral geometry and AdS/CFT.[6][7] He was also a co-author of Modular Berry Connection for Entangled Subregions in AdS/CFT, which was published in Physical Review Letters.[8]
OpenAI
At OpenAI, McCandlish was credited as an author of the company's 2018 post How AI training scales describing empirical results on training efficiency and batch-size effects.[9] He also co-authored OpenAI research on large-batch training and scaling behaviour in neural language models.[2][3]
Anthropic
In 2023, VentureBeat interviewed McCandlish about Anthropic's Responsible Scaling Policy (RSP), a framework intended to manage catastrophic-risk thresholds for frontier AI systems.[10]
In 2024, Business Insider described McCandlish as Anthropic's chief technology officer and said he served as the company's Responsible Scaling Officer.[11] In October 2024, Anthropic announced an updated RSP and said that co-founder Jared Kaplan would serve as the company's Responsible Scaling Officer, succeeding McCandlish in that role.[12]
In October 2025, TechCrunch reported that McCandlish moved from CTO to a new role as chief architect when Anthropic hired Rahul Patil as chief technology officer; Anthropic also announced the transition.[1][13]
Selected works
- McCandlish, S.; Kaplan, J.; Amodei, D. (with the OpenAI Dota Team) (2018). An Empirical Model of Large-Batch Training. arXiv:1812.06162.
- Kaplan, J.; McCandlish, S.; Henighan, T.; Brown, T. B.; et al. (2020). Scaling Laws for Neural Language Models. arXiv:2001.08361.
- Czech, B.; Lamprou, L.; McCandlish, S.; Sully, J. (2018). "Modular Berry Connection for Entangled Subregions in AdS/CFT". Physical Review Letters 120 (9): 091601.