1 min read
Add as a preferred source on Google

Scientists Question Anthropic’s Safety Commitments Over Self-Improving AI

The signatories argued the company’s development push conflicts with warnings that such systems could escape human control.

A blank statement page buckles beneath a steel clasp / TokenPost.ai
A blank statement page buckles beneath a steel clasp / TokenPost.ai

In a joint statement, scientists, legal scholars and technology figures questioned Anthropic’s safety commitments, arguing its push to develop self-improving AI conflicts with warnings that such systems could escape human control.

The signatories said Anthropic has warned that recursive self-improvement could put AI beyond human control even as the company competes to build such systems. They also said Anthropic had weakened earlier safety commitments.

The signatories included a Harvard professor, a Nobel laureate and a Turing Award laureate.

Simon Yoon

Reporter

Simon Yoon reports on blockchain technology for TokenPost. Send corrections or tips to info@tokenpost.com.

Loading…