Tom Brown

Tom Brown is an AI researcher and entrepreneur who co-founded Anthropic. He studied computer science and cognitive science at MIT; the available sources do not establish a completed degree. His career includes software entrepreneurship and technical work on large-scale machine-learning systems before his involvement in the development of major language models.

At OpenAI, Brown was the first author of Language Models are Few-Shot Learners, the 2020 paper presenting GPT-3. The research evaluated a large language model on tasks specified through prompts and examples rather than separate task-specific training. Its results helped draw attention to the changing capabilities of models trained on broad collections of text.

Brown also co-authored research on language-model scaling laws. In 2021, he co-founded Anthropic with colleagues who had worked at OpenAI. The company's research and Claude products depend on substantial computing infrastructure, making the acquisition, organization, and effective use of compute an important part of model development.

Anthropic identifies Brown as its chief compute officer. His responsibilities include securing computing capacity, scaling it, and using it effectively for the company's systems. His contribution connects published research with the practical resource decisions behind advanced AI: experimental plans only become executable when computing arrangements, training systems, and operational processes can support them at the required scale.