[Satam]

A professor might be able to iterate to a correct answer but a student might not.

And ChatGPT is definitely able to get improve its answer by iterating, it just depends on the toughness of the problem. If it's too difficult, no amount of iteration will get it much closer to the correct answer. If it's closer to its reasoning limits, then iterating will help.

[wruza]

But if you stop them just there, an error persists. A professor is “multi-modal” and in a constant stream of evebts, including their lecture plan and premeditated key results. Are you sure that at some level of LLM “intelligence”, putting it into the same shoes wouldn’t improve the whole setting enough? I mean sure, they make mistakes. But if you stop-frame a professor, they make mistakes too. They don’t correct immediately, only after a contradiction gets presented. Reminds me how LLMs behave. Am I wrong here?

Edit: was answering to gp, no idea how my post got here

[Jensson]

Asking the LLM to correct itself doesn't improve answers since they will happily add errors to correct answers when asked to correct it. That makes it different from humans, humans can iterate and get better, our current LLMs can't.

> But if you stop them just there, an error persists

But humans doesn't stop there when they are making things that needs to be reliably correct. When errors aren't a big deal humans make a lot of errors, but when errors costs life humans become very reliable by taking more time and looking things over. They still sometimes makes mistakes that kills people, but very rarely.