"Pre-training is now just a warmup": post-training RL and inference take over compute
joeddav · x · 2026-10-07
Vincent Weisser shared a chart showing post-training/RL and inference taking over compute, and Greg Cook agreed that "pre-training is now just a warmup phase"—reflecting the industry-wide shift of compute budgets from scaling pre-training toward reinforcement learning and inference-time compute.
More from AGI Musings
- Gary Marcus, Ed Zitron and guests spend 90 minutes dissecting the AI bubble — GaryMarcus · 2026-10-07
- NeuroAI manifesto: applying AI scaling laws to BCIs, the bitter lesson for the brain — w1kke · 2026-10-07
- Peter Diamandis: AGI will NOT end human purpose — PeterDiamandis · 2026-10-07
- Pedro Domingos Dismisses AI Doom Fears: LLMs at Inference Are Just Ordinary Programs — pmddomingos · 2026-10-07
- Scott Alexander pens open letter to Steven Pinker on AI risk debate — ZeroStateReflex · 2026-10-07
- Not enough mathematicians for the AI proof deluge—centaur collaboration as the fix — bradneuberg · 2026-10-07