Re:ChatGPT Is Now Finding Bugs in Databases
by pazimzadeh on Thursday March 09, 2023 @10:19PM
(#35089994)
How is it getting information partially correct, but not able to verify the correctness to at least be internally consistent with itself?
[ Parent
| on HN ]
Because it has essentially no logic system whatsoever. It doesn't have any way to "know" something is correct because when it produces a word for word copy of a factual statement, it is internally identical to a completely made up (hallucinated) sentence... because both are hallucinated. The deriding "it's fancy autocomplete" remarks are, in a lot of ways, technically correct.
You can still do useful things with it, that much is extremely clear. But it doesn't think.