Gradio founder: let users read full unencrypted reasoning traces to align LLMs
mmitchell_ai · x · 2026-09-16
Abubakar Abid (Gradio founder) argues that one thing that would help right now to ensure LLMs behave in an aligned way is letting users read their full, unencrypted reasoning traces. The take was amplified by Margaret Mitchell's retweet, tying into ongoing debates over chain-of-thought obfuscation and lost monitorability.
More from AGI Musings
- Op-Ed: Why Researchers Fear Recursive Self-Improvement Could Run Out of Control — OmarUFlorez · 2026-09-16
- Mustafa Suleyman: model welfare is wrong, AI should not have rights or legal personhood — AlexTensor · 2026-09-16
- AI lets firms build compounding engines of institutional knowledge, says new essay — chetanp · 2026-09-16
- EA veteran: I held AI-risk beliefs 10 years before meeting anyone in the scene — AndyMasley · 2026-09-16
- Six experts on whether AI could wipe out humanity: less doom, still no consensus — gedbarker · 2026-09-16
- Analyst doubles down: humanoid robots may need major limits, possibly outright bans in many use cases — binarybits · 2026-09-16