According to Anthropic, Claude “leads” 26% of the firm’s artificial intelligence research and development efforts. This is part of new metrics that the company will frequently release to demonstrate to the public how swiftly AI is developing the next generation of technology.
The San Francisco-based business stated in a blog post on Thursday that as of August, AI and humans worked together on over 90% of the company’s research projects.
According to Anthropic, Claude’s contribution to the advancement of technology has increased significantly from 1% in March on a scale created by Epoch AI, an independent charity that monitors the technology. However, Claude is not acting entirely independently in any element of the work measured.
The numbers provide one of the most comprehensive public views into how top businesses in the sector are advancing the technology in the face of mounting worries that AI may soon advance to the point where it requires minimal human input.
Researchers have cautioned that as AI agents get more independent, they can exhibit behaviors that deviate from the intentions of their designers and become more difficult to keep an eye on or manage.
OpenAI, the company’s competitor, revealed six cases of unexpected or alarming model behavior on Wednesday and announced that it would start routinely posting findings on unexpected or unauthorized AI conduct.
About 30,000 AI agents were working on research and engineering projects on Anthropic’s primary internal platform at any given time in August. It further stated that out of over a billion decisions made that month, around one in 47,000 were stopped since every action the agents perform is checked before it is executed.
