News for hackers. Stuff that matters.
Re:GPT-4 Can't Reason by ChatGTP on Tuesday August 08, 2023 @06:14PM (#37056185)
And using chain of thought, asking to think it out, I'm sure it could get much trickier reasoning done.

When I read comments like this I just translate it too, “I could coax an answer I like better out of it”

Doesn’t really sound like the system is reasoning but you are.

Re:GPT-4 Can't Reason by adamgordonbell on Wednesday August 09, 2023 @09:19PM (#37070668)
You can get a lot more out of LLMs by asking them to spend more tokens thinking it out. It's a tool, and that's just a technique to get more out of it.