Someone sent me an e-mail with a long transcript of trying to use ChatGPT to figure out how to build some program in a public git repo of mine.
(The program has makefile; "make; sudo make install")
ChatGPT's responses can be summarized as "I have no idea how to install that, but here is this similarly-named, completely unrelated project and its installation instructions".
Uh, that's nice, thanks for sharing.
This seems like foreshadowing; users are now going to show up at your door armed with lengthy AI chat transcripts, whose failures are somehow your fault. "If you had better documentation, AI would have found it, scraped it and helped me, instead of bullshitting in circles."
What’s even worse is crypto meaning cryptocurrency instead of cryptography now. Now I can’t talk about “crypto” without sounding like I’m going to fill up someone’s spam folder.
Reminds me of working at a big company where everyone gets openly cc'd on an email and people start using reply-all and soon everyone is bombarded with ridiculous replies. But then the funniest part comes when higher-ups start replying all to say "stop hitting reply-all!", as if that somehow helps. I understand the temptation to say something, but sometimes you just have to let it blow over because anything you do will just make it even more obnoxious
One of the hardest parts of learning to write for print is learning what's worth writing, especially when doing criticism. It's always helpful to ask who will be reading it, what they want, and what style best conveys it to them.
A lot of our professional norms contradict this. I get unnecessarily long emails every day, I sift through byzantine project structures as a profession, and I attend meetings that could have been an email -- all from people more senior than me.
If there's a message in what I'm saying, it's to question everything, especially in writing. If it makes more sense than what you're doing, write short emails, write short blog posts, write nothing at all.
I agree there's lots of 'puff', even hysteria, etc, around AI, but there really is a 'there' there, IMHO, and just the volume of new, often quite astounding, applications shows that.
If you find this article boring, you might want to watch this show instead, where somebody asks live AI replicas of Jordan Peterson and Ben Shapiro stupid questions: https://youtu.be/QXsnvPRwNss?t=560
(The program has makefile; "make; sudo make install")
ChatGPT's responses can be summarized as "I have no idea how to install that, but here is this similarly-named, completely unrelated project and its installation instructions".
Uh, that's nice, thanks for sharing.
This seems like foreshadowing; users are now going to show up at your door armed with lengthy AI chat transcripts, whose failures are somehow your fault. "If you had better documentation, AI would have found it, scraped it and helped me, instead of bullshitting in circles."