News for hackers. Stuff that matters.
Re:Claude 2 Internal API Client and CLI by deadmutex on Friday July 14, 2023 @04:50PM (#36729743)
How do you know if it is correct or a hallucination?
Re:Claude 2 Internal API Client and CLI by paxys on Friday July 14, 2023 @05:37PM (#36730265)
That's what prompting is all about. Ask it to prove its statements. Ask it to quote passages that support its arguments. Then double and triple check it yourself. It isn't going to do the work for you, but can still be a pretty great reference tool.
Re:Claude 2 Internal API Client and CLI by luma on Friday July 14, 2023 @05:30PM (#36730193)
Investigate, same as you would a paralegal etc. If it makes an assertion, contest it and ask where it found supporting evidence in the document for the claims made. Ask it to make the counter-argument, also with sources. Verify as needed.
Re:Claude 2 Internal API Client and CLI by ronsor on Friday July 14, 2023 @04:53PM (#36729789)
Presumably one can test it with documents one has already read and knows before. If the summaries of the test documents are good, future summaries will probably be OK too.
Re:Claude 2 Internal API Client and CLI by lambdaba on Friday July 14, 2023 @05:17PM (#36730051)
Maybe having it summarize a fiction book (outside of training data)?
Re:Claude 2 Internal API Client and CLI by zmmmmm on Friday July 14, 2023 @06:00PM (#36730549)
> If the summaries of the test documents are good, future summaries will probably be OK too

But that is exactly what is problematic with hallucinations. It's a rare / exceptional behaviour that triggers extreme departure from reality. So you can't estimate the extremes by extrapolating from common / moderate observations. You would have to test a lot of previous documents to be confident, and even then there would be a residual risk.

Re:Claude 2 Internal API Client and CLI by sitkack on Friday July 14, 2023 @07:18PM (#36731245)
As with anyone, if you are going to take action on something, double check it.