ChatGPT
I just tried asking #ChatGPT some things about my library Soukai, and I'm very surprised that it knew about it. Sure, it answered mostly what is already written in the documentation, and the code snippets weren't 100% right. But I'm surprised that it knew it at all.
ChatGPT
For context, this is a library that only has 10 stars in GitHub and I'm probably the only person using it. Certainly, I'm the only person who's written any documentation or published code using it (that I know of).
So this tells me that this thing has read everything online O.o
ChatGPT
Aha! But apparently, it hasn't read my personal website xD
Which is interesting, because I'm pretty sure more people have read my website than the Soukai documentation. So it is interesting to see that it prioritizes GitHub content, whatever it is, over any run-of-the-mill personal website.
ChatGPT
Well, now it broke completely :/
This is the biggest problem I see with these tools. It's ok if they don't know everything, but when they start making things up with complete confidence, that's a problem. I don't even think this "Okgrow" company exists.
ChatGPT
@yisraeldov I see xD
This is the first CW I use, so I'm not too familiar with how they "should be used". I recently listened to a @weareopencoop podcast where they talk about different reasons to use CWs, and one of them is to mark content that could be annoying for some people.
Also, it doesn't help that I use an instance-of-one, so there isn't really any rules of conduct I can refer to.
Anyways, thanks for letting me know. I guess I won't be using them as much as I thought.
This is an instance-of-one managed by Noel De Martin.
ChatGPT
@noeldemartin
This doesn't need a CW