Noam Brown presents the test-time compute thesis at TED AI
Presented the test-time compute thesis at TED AI — cited the poker case in which letting a bot think for just 20 seconds gave the same performance gain as scaling the model up 100,000 times and training it 100,000 times longer
Why it matters
Shifted the axis from pretraining scaling to inference-time scaling. Later became the line cited as grounds in the reasoning-model discourse
Source — Primary sources first; secondary reports are labelled “Press”
Pressventurebeat.comhttps://venturebeat.com/ai/openai-noam-brown-stuns-ted-ai-conference-20-seconds-of-thinking-worth-100000x-more-dataAdded in Week 35, 2026 · See that week