Discussion about this post

User's avatar
Rajveer Kapoor's avatar

The recurring story about Geoffrey Hinton's 2016 prediction on radiologists captures a pattern that matters more than the specific timeline. The prediction was not wrong about AI capability, AI can read scans, it was wrong about the speed of organizational and institutional substitution. That gap between technical capability and deployed impact appears to be wider and more persistent than most economists expected, and it is showing up clearly in the current AI cycle too, where controlled experiments show large productivity gains while the macro data shows almost none.

No posts

Ready for more?