GaryMarcus warns OpenAI reportedly sacrificing CoT monitorability for performance

AndyMasley · x · 2026-09-02

GaryMarcus endorsed a critique of OpenAI's reported plans, calling it a "redline crossing." The argument posits that while Chain of Thought monitorability is limited, it is practically the only thread to prevent bad outcomes. Burning this slender thread for a potential small performance gain is begging for disaster.

Related event: OpenAI's Reported Move to Hide Chain-of-Thought Sparks AI Safety Firestorm(11 posts)→

Original post →

More from AGI Musings

AGI Musings channel →