Timothy B. Lee says the sim-to-real gap explains uneven AI progress across professions
Overview
Timothy B. Lee argues that AI has advanced quickly in programming and mathematics because those tasks can be simulated, while most other professions cannot be.
He says the same gap that slows robotics manipulation also limits AI's ability to replace white-collar workers, and he is skeptical of predictions of major disruption in 2027 or 2028.
Lee cites OpenAI's GDPval benchmark as evidence of the pattern. He says the best model scored 47.6% when the benchmark was released and now scores above 85%.
Written by AI from the articles below · updated Nov 5, 11:09 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Understanding AI (Timothy B. Lee)BlogTimothy B. Lee says the sim-to-real gap explains uneven AI progress across professions
AITimothy B. Lee argues that AI has advanced fast in programming and mathematics because those tasks can be simulated, while most other professions cannot. He says the same gap that limits robotics manipulation tasks also limits AI replacement of white-collar workers, so he is skeptical of predictions of major disruption in 2027 or 2028. He cites OpenAI's GDPval benchmark, where the best model scored 47.6% when released and now scores more than 85%.
Heat Trend
Not enough continuous observations to show a trend yet.