Is anyone surprised? By definition LLM can’t be 100% correct and LLM hallucination poses significant challenges in generating accurate and reliable responses. ChatGPT Answers Programming Questions Incorrectly 52% of the Time: Study. To make matters worse, programmers in the study would often overlook the misinformation. https://gizmodo.com/chatgpt-answers-wrong-programming-openai-52-study-1851499417
#TIL that @JasonPunyon curated and compiled a whopping archive of answers from #StackOverflow and assorted #StackExchange Q&A sites in a minimal sqlite format, where they can be downloaded and analyzed offline:
Amazing effort and great idea. Reminded me of the archives that #Kiwix kept of it (alongside Wikipedia and similar projects), but more streamlined and cross-platform. Nice.
Logged in to Stack Overflow for the first time in 10 years to check exactly what have I left there.. Now, since the few answers I've given are CC-BY, so they must attribute them to me, if I make a GDPR data subject request for them to delete all my data, logic dictates that whatever they think about text as personal data, they must delete the answers, because they can't attribute them to me as a deleted account. Right? #stackoverflow
Sometimes I think #StackOverflow should let us upvote an answer again. Some of these answer I find them again after few years and they still work perfectly.
À quel point c’est hypocrite et incohérent de demander la suppression de son compte et données de #StackOverflow, tout en continuant d’aller y consulter les contributions qui y subsisteront ? Je demande pour un amis.
Funny how the "frequently asked questions" for the #StackOverflow#OpenAI thing does not list the question "How do I opt out of this?" (nor the answer to it).
Stack Overflow bans users en masse for rebelling against OpenAI partnership — users banned for deleting answers to prevent them being used to train ChatGPT
Zawsze, ale to zawsze jak wchodzi z finansowaniem VC (Vencure Capital) to serwisy, nawet zakładane przez zaufane osoby ze społeczności dają d... wróć, dają ciała...
Czy można wiedzę IT zbieraną przez społeczność od 2008 roku sprzedać, nie będąc jej właścicielem?
Jak jest się białym chłopem z krzemowej doliny, to pewnie, że można :/
Not all things are meant to last forever. Their impact was HUGE to developer space. It made programming so much easier and the model worked for over a decade plus! An incredible place to go learn and find answers when you ran into some most ridiculous problems at work.
Now, their traffic has been dying. They didn't jump to AI, they tried to resist it. But developers left them, and have been leaving them. What are they supposed to do? lay down and die?
Eine sehr gute Toot-Reihe von @bkastl über das kommerzielle Ausnutzen der kostenlosen Inputs von User*innen auf Stack Overflow über ChatGPT von OpenAI:
This is what you get when you try to delete your content from #StackOverflow: they treat you like a criminal defacing their site, rather than as a human being with privacy rights.
Will be saving this conversation and sharing publicly. Your move, Stack Overflow…
#StackOverflow bans users en masse for rebelling against #OpenAI partnership — users banned for deleting answers to prevent them being used to train #ChatGPT
It's hard to imagine #StackOverflow made their decision without considering the fate #Twitter and #Reddit suffered. That they willingly chose to burn the remaining goodwill they had with their users is bewildering.