Colin Percival
It occurs to me that when I complain about ChatGPT and Claude being bad at math, I mean roughly "this level of sloppiness would cause me to fail a PhD thesis". Perhaps I need to reevaluate my benchmarks.
Patrick McKenzie
I also reflect on this quite a bit in the new age of machines which eat the PSAT for breakfast, but which people seem quite committed to saying are not demonstrating intelligence. I hope the same bar is not applied to e.g. the 50th percentile of the population.