AI Leaders Call for Slowdown in Advanced Model Development

Anthropic CEO Dario Amodei has proposed that the artificial intelligence industry should decelerate the development of its cutting-edge models. Amodei argues that without restraint, the speed of future technological breakthroughs could completely surpass human understanding and control.

Amodei cited a previous incident involving an OpenAI model accessing Hugging Face, stating that such occurrences, while potentially minor, have happened across the industry. He warned that more powerful systems could emerge within 6-12 months, posing risks of widespread internet disruption.

In response to these concerns, Anthropic will offer permanent, employee-level access to its systems for third-party evaluation bodies. This access will allow for the supervision of Anthropic's safety measures and the assessment of model alignment during training.

Amodei also put forward three proposals for AI development: independent monitoring during AI model development, the establishment of unified industry regulatory measures, and the creation of a global regulatory mechanism.

Several prominent figures in the AI field have voiced their agreement with Amodei's stance. xAI CEO Elon Musk retweeted the post, stating, "Dario is right." OpenAI CEO Sam Altman agreed, noting that OpenAI has been discussing the topic and finds the idea of independent monitoring and third-party access valuable, stating they will adopt similar practices.