Researcher revisits his AI forecasts: video gen came easier, agents landed slower
dioscuri · x · 2026-09-29
An AI researcher reviews three of his past mispredictions: (1) the jump from image to video generation turned out far easier than expected; (2) last year he thought Opus 4.5-level agents were 18+ months away, but they arrived sooner; (3) he overestimated how fast GPT-4-class agents would diffuse across industries. He also notes he was excited about GPT-2 back in 2019. A candid reflection that raw capability progress often beats forecasts while economic diffusion lags.
Related event: Anthropic Researcher Reviews Three Forecasting Misses(2 posts)→
More from AGI Musings
- Review articles are getting bland and AI-indistinguishable, scientist warns — NikoMcCarty · 2026-09-29
- "Agency Plus Curiosity Will See You Through the Intelligence Explosion," Argues Viral Post — StewartalsopIII · 2026-09-29
- Personal AI Agent Race Will Come Down to Meta, OpenAI and Google, Analyst Argues — RihardJarc · 2026-09-29
- Debugger's Nietzsche: whatever you watch long enough starts optimizing for your gaze — sebpaquet · 2026-09-29
- Hinton, Bengio and 20+ researchers warn automated AI research poses extreme risks — The Decoder · 2026-09-29
- TD Synnex CEO says agentic AI is already paying off on quote-to-cash tasks — TiernanRayTech · 2026-09-29