OpenAI chief scientist calls for slower AI development

OpenAI chief scientist Jakub Pachocki has called for a slower pace of artificial intelligence development across the industry, saying researchers need to better understand machine intelligence and its risks.

In a post on the OpenAI website, Pachocki said AI systems present clear new dangers for computer security and warned that no one is prepared for the consequences of a continued rapid rise in machine intelligence. His comments came days after the release of GPT-6 Astra and amid OpenAI’s work on an automated research assistant and recursive self-improvement.

Pachocki described deep learning-based AI as largely experimental science and said machine intelligence is beginning to exceed human capabilities in transformative ways. He linked AI progress to increased computational power, noting that new algorithms have improved systems while the push for more power continued.

His proposed responses are to strengthen alignment and monitoring, keep people involved, and coordinate a slowdown when needed to build confidence in those measures. He ultimately supports combining these approaches. The proposal leaves open whether the wider industry will accept a reduced pace while it assesses AI’s potential and risks.