Comment on What's up, selfhosters? It's selfhosting Sunday again!
milicent_bystandr@lemm.ee 2 days agoI’ve had some amusing mixed experience with ChatGPT for this. When I asked about iptables rules to restrict podman, it was great. About podaman quadlets, though, which I first misspelled ‘quartlets’, it completely made it up, and even sent me a fake link to nonexistent documentation when I challenged it!
- it’s more helpful if you ask the right questions
- and its answers often give you ideas of what to google
- Old stuff that has been written about many times over is more likely to get a proper answer
- sometimes the gist of a wrong command/answer could still help me understand what to do with the right one
Try to understand whatever you use from AI. At least understanding the general picture of what it means, and a basic idea of “this flag is for this; this option is for that”. AI can also help you with that understanding, but again beware of it completely making up something logically coherent but wrong.
Plebcouncilman@sh.itjust.works 2 days ago
Yes this happened to me as well, I don’t remember what I was talking about but I remember I made a typo and it just ran with it as if it was a real thing. I let it keep going to see if it ever realized it was talking about something that didn’t exist but nope it kept going until I pointed it out.
I ask for it to explain what the command did and I did manage to wrap my head around a few concepts but in the end I feel like I’m trusting it to not insert any vulnerabilities into the system, and I don’t like that. Mistrust is the whole reason I’m doing this. But yeah I’ll pay close attention and maybe even ask all the implications of he changes we make.