Re:Stack Overflow Moderation Strike Update
by npsomaratna on Sunday June 11, 2023 @10:22PM
(#36287758)
Made me curious, so I tried that exact prompt in the GPT-4 playground. Response:
[ Parent
| on HN ]
Re:Stack Overflow Moderation Strike Update
by JoelEinbinder on Sunday June 11, 2023 @10:44PM
(#36287877)
This is slightly wrong. build/ will ignore the build directory at every level. /build will ignore the build directory at the top level.
I also have no idea what it’s trying to communicate with its /*/temp/ examples
> This is slightly wrong.
A perfect representation of my git usage then!
Re:Stack Overflow Moderation Strike Update
by donmcronald on Sunday June 11, 2023 @11:25PM
(#36288115)
> This is slightly wrong.
I wonder how it does if you tell it it’s wrong. I’ve tried a few times and it seems to get confused pretty easily.
Re:Stack Overflow Moderation Strike Update
by pwdisswordfishc on Monday June 12, 2023 @01:32AM
(#36288814)
"I apologize for the confusion. [Repeats the same wrong thing]"
I once had a project with a "bin-name" in the gitignore, to exclude the "/bin-name" that "go build" will produce. Then I added a "dir/bin-name/file.go" which of course got excluded. The CI/CD builds failed, but it worked locally, and great confusion ensued. Since then, I've always added that leading "/" and generally try to be as specific as possible in gitignore files.
This is a great example where many people providing different answers and comments based on their different knowledge and experience provide value, opposed to just ChatGPT telling you stuff. Simply "correct" is nice, but "correct, complete, and with experience and context" is better.
Re:Stack Overflow Moderation Strike Update
by Tijdreiziger on Sunday June 11, 2023 @10:46PM
(#36287887)
A priori, you can’t know whether this is accurate or a hallucination.
Re:Stack Overflow Moderation Strike Update
by ahahahahah on Monday June 12, 2023 @04:31AM
(#36290150)
I really think the whole "hallucination" thing is a terrible framing. All that gpt-4 does is spew out bullshit. Now, it applies a bunch of statistics to attempt to spew out believable bullshit and those statistics cause it to sometimes accidentally be correct. The base assumption with this should be that it's wrong, and so we should instead be framing it as like sometimes accidentally producing "actualities" or something.