News for hackers. Stuff that matters.

A comment on Anthropic publishes the 'system prompts' that make Claude tick. Parent comment

Re:Anthropic publishes the 'system prompts' that make Claude... by TeMPOraL on Tuesday August 27, 2024 @07:59PM (#41374607)
Specifically, "hallucinations" are very common in humans; we usually don't call it "making things up" (as in, intentionally), but rather we call it "talking faster than you think" or "talking at the speed of thought".

Which is pretty much what LLMs do.

[ Parent | on HN ]
Re:Anthropic publishes the 'system prompts' that make Claude... by adastra22 on Tuesday August 27, 2024 @09:30PM (#41375135)
Yeah an LLM is basically doing what you would do with the prompt “I’m going to ask you a question, give your best off the cuff response, pulling details entirely from memory without double-checking anything.”

Then when it gets something wrong we jump on it and say it was hallucinating. As if we wouldn’t make the same mistakes.

[ Parent | on HN ]
Re:Anthropic publishes the 'system prompts' that make Claude... by kortilla on Tuesday August 27, 2024 @10:11PM (#41375332)
It’s not like that at all. Hallucinations are complete fabrications because the weights happened to land there. It has nothing to do with how much thought or double checking there is.

You can trick an LLM into “double checking” an already valid answer and get it to return nonsense hallucinations instead.

[ Parent | on HN ]
Re:Anthropic publishes the 'system prompts' that make Claude... by adastra22 on Wednesday August 28, 2024 @03:40AM (#41376984)
That’s how your brain works at the base level too.
[ Parent | on HN ]