Jared Kaplan is a theoretical physicist who became one of the central figures in the science of large language models. In January 2020 he was first author of “Scaling Laws for Neural Language Models” with Sam McCandlish and colleagues at OpenAI. The paper found that a model’s loss falls as a smooth power law in model size, dataset size, and compute, with trends holding across more than seven orders of magnitude. That result gave the field a predictable recipe: spend more on the right ingredients and performance improves in a foreseeable way, which justified the era of ever-larger models.
In 2021 Kaplan co-founded Anthropic, where he serves as chief science officer and has helped shape the training and safety research behind the Claude models. He retains an academic appointment as a physics professor at Johns Hopkins University.
Why business readers should care: Scaling laws are the reason labs have been willing to commit billions of dollars to bigger models and data centers - the spending only makes sense if the returns are predictable. Kaplan led the paper that established that predictability.