LLMs are generally OK, if you can craft an unbiased question that demands facts. Never seen ChatGPT get it wrong. But it’s stunning how easily you can manipulate them just a couple of prompts deep.

Thing is, most people, in America anyway, didn’t get the science training I got in 70s elementary school, and even though I’m barely above average IQ, I was a star science student. Imagine those people who don’t understand empiricism using LLMs. The mind boggles.