TracketPacer 🚀May 9, 2023

i asked chatGPT to explain a few concepts to me directly out of the 802.3 standard & it 100% lied, completely confidently—only backtracking when i called it out on the lie.

both versions 3 & 4 did this. it has no fact checking. pls stop using it to just google things

Show threadpiegames
@tracketpacer No it doesn't "lie", it doesn't "backtrack" or "double down", it has no concept of " facts" or "truth". It's a fucking language model and not an "AI". Implying human-like agency or treating it as such is a big part of the problem.
May 9, 2023 at 10:41pmWeb
Show threadTracketPacer 🚀May 10, 2023
@piegames chill the fuck out dude
Show threadpiegamesMay 10, 2023
@tracketpacer Good job on making a borderline rage-bait post and then telling people to chill 🤷‍♀️ (also doing that makes it conveniently easy to ignore that some of them may have a point …)
Show threadTracketPacer 🚀May 13, 2023

@piegames my intent was absolutely NOT to create rage-bait. i already mentioned that im tired of people answering questions i post that are meant for real people with “here’s what chatGPT says,” implying it had the answers to everything without fact-checking.

i did my fucking best to communicate & people like u are the ones being drawn into what u perceive as “rage-bait” bc you’re identifying with a fucking LLM. im so sorry i didn’t use the perfect most completely specific syntax to communicate what i meant to, but if you really cared about doing anything other than shutting down valid criticism of people’s use of LLMs like this, you’d be able to see i even posted screenshots of what i said & got responses to.

maybe ask yourself why my critique of people using this tool TO GOOGLE THINGS is so upsetting to u & consider that i myself may also have a valid point here.

Show threadTracketPacer 🚀May 13, 2023
@piegames here, let me hand you the twitter thread with screenshots so you don’t have to do any more work, just in case you’re truly as averse to exercising your brain / critical thinking as you seem to be: https://twitter.com/tracketpacer/status/1655943472644493312?s=46&t=26EjD6dkvOVwB3DbmDD9nQ
TracketPacer on Twitter

“i asked chatGPT to explain a few concepts to me directly out of the 802.3 standard & it 100% lied, completely confidently—only backtracking when i called it out on the lie. both versions 3 & 4 did this. it has no fact checking. pls stop using it to just google things”

Twitter
Show threadpiegamesMay 13, 2023

@tracketpacer To make a point clear: I totally agree with you that one should absolutely not use LLMs as a knowledge database. I do not "identify" with that technology at all, and you inferring that from my comment tells me that you spend too much time arguing with randos on social media ^^

And I'm really sorry, but reading things like "it just starts deflecting" about an LLM output makes me cringe. This is not "syntax", this is fundamental to how we approach and think of this technology IMO.