You must log in or register to comment.
While SO is toxic, it’s mostly about the failure of the supplicant to do the fucking research (RTFM or equivalent these days goddammit, and is usually backed by ppl with actual knowledge, Chat GPT / LLMs., not so much.
The new model 5.2 seems to have toned down a little in term of constantly glazing you up.
Honey, we all know they’re hitting the ceiling they predicted in their OpenAI paper on AI Scaling Laws in 2020 and corrected by Deepmind’s 2022 followup paper. It’s trash now and it will always be trash, attempting to give it that infinite power and compute time to reach 94% is just an exercise in futility.
Hey there Sam. Maybe we should believe you this time?
The people with the money believe him, that’s what matters.
Chatgpt just cribs from stack overflow, which in turn just cribs their answers from documentation. Once you figure that out, they both become surprisingly useless.
Stack Overflow was the antithesis of “Just say something wrong on the internet so that someone will correct you with the real answer” because none of the negative threads actually answered the question lol.
That’s a fun thought but StackOverflow are early adopters for AI so they’re probably not much different at all.
It’s true. And neither of them back it up with reasons.
Nor care to have those reasons in the first place.
And if you press them, you get complete bullshit reasons. But SO will also ban you, while ChatGPT only cares about you paying.
glm-4.7: I fixed it
deleted by creator
If you guessed this sound (rubber duck squeak) is not like the others then you’re absolutely right!
Source: Sesame Street record from when I was a kid.
Sometimes I rubber duck with ChatGPT.
Honestly I’ve learned more in a few months of fixing its mistakes than I had in years of being on the job.
I kinda miss stackoverflow
ChatGPT business model: rephrasing stackoverflow answers to make the asker feel like a valuable human being.
You’re on to something there! Shut up and take my money!
ChatGPT: “I’m absolutely wrong”
Ahh…I see the problem.
Ime Google AI is much worse about making up wrong answers to sound right.
If I’d ask ChatGPT and Google AI to help me craft a set of Ansible tasks to do something rather simple but also something I don’t do very often, like converting PEM certs + key to PKCS12…they’d both write a playbook that’s close, but ChatGPT would be much closer.
But they both say crazy shit sometimes. The other day ChatGPT told me Fedora 40 is the latest release and 43 is still in testing.
deleted by creator
The good thing about stack overflow, (when your question isn’t marked as a duplicate of something completely unrelated), is that you accidentally learn random other things while browsing answers.
Stack overflow is great for learning from posts that are marked as duplicate! I’ve learned the most from those lol
The one very important thing I learnt on stack overflow was the XY problem. Most of my first questions were about how to make my idiotic solution work, instead of how to properly solve the problem.
