… so you could type anything at the terminal and the artificial intelligence would provide documentation + suggestions for corresponding commands. Of course, the a.i. should never be able to run any commands by itself.
Of course, the a.i. should never be able to run any commands by itself.
Oh you sweet summer child, that’s like handing a toddler a gun and telling them not to shoot anyone.
If recent “advancements” in products have taught us anything, it’s that adding AI to things that don’t need it always makes it better! Right, guys? …Right??
Eew, no. Multiple issues with this.
- Performance. When I’m booting my computer I want it to turn on ASAP. I don’t want it wasting time loading such a heavy system like a “language” model.
- Security. Yes, someone will make it run commands. And even if it doesn’t, simply suggesting what you should type is already bad.
- Discouraging people from reading the manpages, and understanding what the command does, before typing it out. Unless your LLM gives you a comprehensive explanation of the command, but then you’re just reinventing the manpages.
(Upvoted as unpopular.)
Thanks for the upvote, i expected it to be most unpopular ever 😋
Now about the performance issue : please see my other comment in here where i suggest how it could be … using 2 computers+ USB link.I’ve seen it. Using a second machine to perform actions is inefficient IMO; plus as someone else said, this already exists - you could simply ask some chatbot the question through your second machine, like writing
!ai how does the Linux command "ls" work?in DDG.
I don’t agree with this opinion, because it seems very not-well-thought-out to me.
If your computer is having trouble booting up then how are you supposed to get an AI model running there?
If you can get the AI model running, you’ve got at least a mini kernel working, but that means you need to load that mini kernel, and oh look, where are we now? GRUB-ception?
It seems like a lot of engineering for something that maybe one person per mille might use. The vast majority of users don’t do things that would cause GRUB to break and those that do likely already know how to get out of it.
Hilarious. Also we have that already it’s called using google on your second device.
Heck yeah. Just yesterday I solved a gnupg key issue the good old fashioned way; digging through dozens of forum posts from various websites until one of them worked…
The good old days. Where’s the fun in letting a robot do all the work. You don’t learn anything.
As long as it’s a small, local “AI” that’s basically just regurgitating man pages and not sending everything off to servers for data scraping and actual processing power.
Anything more than simply feeding it man pages would get gross very fast both for hallucinations and for the ability to keep it up to date with any specific system.
They may not have seen that documentary TRON.
The nice thing about Open Source is that the very moment some distro decides to stuff A.I. into itself, that shit will immediately be forked so that we don’t have to deal with it.
I’m happy to visit a website where I can ask an AI to grammar check my code or something similar. But There is no way in Hell I’m allowing one to live on my system accessing all my things.
… and something similar exists:
https://www.aiterm.net/
