Mollick: OpenAI and Anthropic hint at early recursive self-improvement in rare statements

emollick · x · 2026-09-13

Ethan Mollick notes that this week brought the clearest statements yet from both Anthropic and OpenAI that some form of recursive self-improvement (RSI) has been achieved, though early. RSI could drive rapid AI capability gains, and the first firm to achieve it may gain an unsurmountable lead.

He also points to the anxious tone of both companies' posts, linking Dario Amodei's new essay "We Must Pace the Frontier", which argues AI could cure most major diseases within 5–10 years while warning of loss-of-control, misuse, and economic-disruption risks — and calling for a "race to the top" on safety rather than a commercial race to the bottom.

Related event: Mollick: OpenAI and Anthropic Hint Recursive Self-Improvement Has Begun(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →