ChatGPT co-creator: LLMs are built for assistance, not automation
SimplyAnnisa · x · 2026-11-12
The post quotes what many called the most first-principles talk at #AIECode, from someone who co-created RLHF/ChatGPT at OpenAI and now builds Jev.
His argument: today's LLMs are optimized for assistance with a human in the loop, not automation. "95% Einstein, 5% Mr. Bean" is fine for a chatbot but fatal for back-office decisions. His "bitterest lesson" hierarchy: compute beats algorithms, data beats compute, and the bitterest layer goes further (truncated). The poster's takeaway: reliability beats everything.
Related event: RLHF Co-creator: LLMs Are Built for Assistance, Not Automation(4 posts)→
More from AGI Musings
- FT: AI agents supercharge start-ups — no employees, prototypes within days — nordicinst · 2026-11-12
- Erik Voorhees slams The Economist over Jensen Huang criticism: 'must go' — beffjezos · 2026-11-12
- AI Circle Debates: Would One Company Winning Break the Global Socioeconomic System? — teortaxesTex · 2026-11-12
- Paul Christiano's three-night Hinton Lectures on governing transformative AI earn rave reviews — S_OhEigeartaigh · 2026-11-12
- Philosophers argue 'humans have no plot armor' in case for near-term AI risk — AaronBergman18 · 2026-11-12
- Linear CEO on agents: give a man a way to build his own tools, and building tools becomes his work — jiayuan_jy · 2026-11-12