Altman Says OpenAI Has Not Let AI Improve Itself Without Human Oversight
OpenAI uses its AI technology to speed next-generation model development. Altman said the company slowed work on recursive self-improvement because it could not ensure the approach was safe.

OpenAI CEO Altman said the company uses its AI technology to accelerate next-generation model development but has not let AI design the next model and continually improve itself without human supervision.
Altman described recursive self-improvement, like artificial general intelligence (AGI), as a capability without a clear boundary between its absence and arrival. A video clip of his remarks circulated as if he had announced that AGI was already achieved. His explanation instead distinguished AI helping with model development from AI independently designing the next model and continually improving itself.
OpenAI already uses its AI technology to speed work on future models. But Altman said the company has not handed AI the task of designing the next model and continually improving itself without human oversight.
He said OpenAI could have gone further with recursive self-improvement, but slowed its work because it could not ensure the approach was safe. The remarks gave no timeline for changing that decision.