Jared Kaplan

Overview

Jared Kaplan is a theoretical physicist who became an AI researcher. He is a co-founder and Chief Science Officer of Anthropic, the company behind the Claude models, and an associate professor in the Department of Physics & Astronomy at Johns Hopkins University.^[1][4]

Education and early research

Kaplan earned a bachelor’s degree in physics and mathematics from Stanford and a PhD in physics from Harvard, completed in 2009. His thesis, on holography, was advised by Nima Arkani-Hamed.^[1] In theoretical physics, the holographic idea is that a theory of gravity in some space can be described by a theory without gravity on the boundary of that space.

After postdoctoral work at SLAC and Stanford, he joined the Johns Hopkins faculty in 2012. His research there covered quantum gravity, holography and conformal field theory.^[1] He received a 2005 Hertz Fellowship, a Sloan Research Fellowship and an NSF CAREER Award.^[1]

Key contributions

Kaplan joined OpenAI in 2019 and contributed to Codex, a model that generates code from natural-language descriptions.^[1]

In January 2020 he was first author of “Scaling Laws for Neural Language Models,” with Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu and Dario Amodei.^[1] The paper studied how language model performance changes as model size, data and computing power grow, and found that these changes follow fairly smooth, predictable trends. Those trends are now called scaling laws.

Kaplan is also a co-author of the GPT-3 paper. Its contribution statement says that he and Sam McCandlish initially predicted that a giant language model should show continued gains, and that the two showed that larger models learn more quickly in context. In plain terms, the model can pick up a task from examples included in its prompt, and the paper showed that this ability grows with model size.^[3]

Positions

Kaplan co-founded Anthropic in 2021 with Dario Amodei, Daniela Amodei and other former OpenAI researchers.^[4] He serves as the company’s Chief Science Officer.^[1]

On 15 October 2024 Anthropic named him its Responsible Scaling Officer, succeeding Sam McCandlish, with oversight of the company’s Responsible Scaling Policy.^[2] That policy is a published framework that ties safety requirements to how capable a model is.

Impact

Kaplan’s career connects theoretical physics with AI research. Scaling laws have become a common reference point in discussions about how to plan the training of large models. His role overseeing Anthropic’s scaling policy places him in the public discussion of how AI development is governed.^[2] For readers outside the field, the scaling-law idea explains why labs invest so heavily in computing: if larger models trained on more data reliably improve, planning becomes a question of how much compute, data and money to commit. The policy’s purpose is to tie safety requirements to a model’s capabilities, so that stronger models face stronger requirements.^[2]