Company
ActiveAnthropic
San Francisco · Since 2021
Still an independent lab, selling model access rather than a finished industry of its own.
A later timeline chapter projects a role for this organization. That chapter is a projection. This page is about the institution as it exists, or as it existed.
Anthropic was founded in 2021 by people who had left OpenAI, including Dario and Daniela Amodei. The company trains large language models and sells access to them. It describes its reason for existing as safety: the model should be less likely to do damage, and the company should be structured so that a rush for customers cannot silently overrule that.
It is a present-day employer, small beside the firms that buy its models, and expensive to run because the training runs are expensive. The 2030 chapter calls this shape a model lab. Anthropic is one of the labs that sentence refers to.
What the staff actually does
Researchers and engineers train and evaluate models. Policy and trust staff write the rules the model is supposed to follow and test whether it follows them. There is also the ordinary company: recruiters, lawyers, sales. The distinctive job is evaluation, the person paid to show that a fluent answer is wrong. That job exists inside the lab and is beginning to exist, under other names, inside the companies that deploy the model.
Not a future firm
Anthropic is operating now. Any chapter that predicts which lab will still be independent in a decade is a projection, and the timeline marks it as one. This page does not choose a winner. It records a company that exists, sells a capability rather than an industry, and employs the people who try to say when the capability is not ready.
The work it organized
- Research scientist
- Evaluator
- Policy researcher
- Applied engineer
On the timeline
Related
- OpenAIActive