Why Kids Outlearn AI: The Data Efficiency Gap
AlisonGopnik · x · 2026-08-25
MIT Technology Review explores the massive gap in language learning between children and LLMs. LLMs require vastly more data—up to 100,000 times more words—than children to achieve fluency. This "data efficiency gap" challenges AI architects and cognitive scientists: how do kids learn so efficiently? Understanding this could lead to better models and insights into human development.
Related event: MIT Review: Children Learn Language Far More Data-Efficiently Than LLMs(2 posts)→
More from AGI Musings
- What to Do When AI Takes Jobs Building Things — arieljalali · 2026-08-25
- Frontier Labs gaining power via controlling unreleased model access is unhealthy — michael_nielsen · 2026-08-25
- Mass Turn to AI for Connection Highlights Modern Loneliness — soumitrashukla9 · 2026-08-25
- People turn to AI for advice and connection as loneliness rises — soumitrashukla9 · 2026-08-25
- Fastest AI Adopting Hospitals See Lowest Deaths in 2026: Study — luisdans · 2026-08-25
- Why Americans Really Hate Data Centers — Don't Worry About the Vase (Zvi) · 2026-08-25