from Hacker News

I feel like ChatGPT would point out gaps within our understanding of math and science before being able to take over our software jobs.

There are many gaps within our mathematical understanding and absorbing propositional logic and spitting out more theorems or proving more theorems seems more aligned with what an AI would be good at finding.

Oppose to software jobs where the inputs are product problems and half (way more than half in my experience) of the problems aren't even defined well enough.