🔊CEOInterviews

Tom Brown on costs

From First Public Talk on GPT-3 by Tom Brown · · Tom Brown

“GPT-3 is incredibly expensive and inconvenient: you have to spend a lot of money to train it, you have to spend a lot of money to do inference on it, and you need a crazy distributed system to run it.”

Tom Brown
Cofounder, Anthropic
Policy Impact costsinfrastructureAI research

On , Tom Brown, Cofounder at Anthropic, spoke about costs during First Public Talk on GPT-3 by Tom Brown on Tom Brown.

First Public Talk on GPT-3 by Tom Brown
Watch on YouTube
First Public Talk on GPT-3 by Tom Brown
Tom Brown
Watch on YouTube
Recorded in August 2020 during the early days of the COVID-19 pandemic, this is the first public talk given by Tom Brown – author of GPT-3 and now cofounder of Anthropic, building advanced AI like Claude – on the past, present and future of this groundbreaking AI system. Tom outlines the scaling laws that enabled GPT-3's improvement in performance, shares insights into its limitations and calls for responsibility in using and shaping such powerful models. He emphasizes the need to do "public goods" research on how to make models that are interpretable and robust - a goal Tom is pursuing now at Anthropic. Join Tom Brown on this fascinating tour inside the making of GPT-3 and the frontiers of AI - where we've come from, where we are today, and where we may be headed next.
Tom Brown

About Tom Brown

Cofounder · Anthropic

Tom Brown, cofounder of Anthropic, participated in a January 2025 conversation with other Anthropic co-founders about the company's history and approach to AI safety. During the discussion, Brown referenced the company's Responsible Scaling Policy (RSP), describing it as a framework that "creates healthy incentives" by aligning internal team incentives with safety and making safety-related decisions more legible externally. He noted that if the company cannot make a model safe at a certain capability level, the RSP provides a preexisting framework for that decision. In a 2020 talk on GPT-3, Brown described the model as "incredibly expensive and inconvenient" to train and run, requiring significant resources and distributed systems. He stated that "the stakes are becoming high" as more resources are directed toward AI, predicting that AI would "permeate every aspect of our lives within the next five to ten years." Brown called for public goods research on making models more interpretable and robust, a goal he said he was pursuing at Anthropic.

Profile compiled from Tom Brown's verified public interviews and appearances. See all quotes & transcripts →

More from Tom Brown Full Transcript Explore All Executives