News

METR reports rapid AI-development gains at Anthropic

Anthropic is using artificial intelligence to develop more capable AI systems, and a new report from the AI safety research organization METR indicates that the approach is accelerating the company’s progress.

METR conducted an 11-week investigation inside Anthropic as part of an evaluation of its AI research and development. According to the report, Anthropic’s models are improving roughly twice as quickly as they were a year ago, with little indication that the pace is slowing.

The findings provide some of the clearest evidence yet that AI-assisted development can substantially shorten the cycle for building stronger models. They also explain why researchers are concerned about the implications of AI systems that help create their successors.

The investigation was separate from a program that gives safety researchers employee-like access to Anthropic’s systems following a series of high-profile hacking incidents over the summer.

Sources