causal

Brute-Force Pattern Recognition Is Brittle

Pure statistical pattern recognition without reasoning about what an object is for is brittle and not general; it gets things right on average but misses important cases (e.g., classifying coffee cups as a cat), errors no four-year-old would make because human reasoning about objects' purpose and nature is far more sophisticated than surface statistical features.

causalpending

Speaker

David Autor

Evidence Quote

“Well, it gets it right on average and misses all the important cases”

Source

David Autor on the Future of Work and Polanyi's Paradox— EconTalk
Created: 6/13/2026, 12:29:27 AM

Your notes

Loading notes...