Linux and AI
- Thread starter Bagheera
- Start date
Yes, but I firstly used AI to crack some of the Windows issues, before I moved on to Linux. Without AI I probably would never do Linux.
Not a stupid question, just a way late for how long people have been using AI for troubleshooting of all sorts.
CaffeineAddict
Well-Known Member
Yes but only once, I had audio issue while playing game.
AI wasn't perfect but it directed to me solution, it gave to command that revealed audio sink was in use by another audio service and that was reason for no audio.
Sadly I forgot to note down steps done, but in essence it was qsynth that messes up with pipewire when Lutris is installed so solution is to purge it.
No AI is perfect. My usual go to is the gpt bot, but it failed me a few times and some other web based helped more. Always stay on top and question everything.
CaffeineAddict
Well-Known Member
You want one that is able to browse the web and generate response based on real time data.
duck.ai -> Claude Haiku 3.5 does that.
There are other models that also process images.
Terminal Velocity
Well-Known Member
When I have problem with my machine I ask the forum. Never thought to ask AI. People are still more reliable and efficient at solving Linux problems
CaffeineAddict
Well-Known Member
Pretty sure because web servers display files (serve them) rather than executing them.
Maybe some people don't care, I do sysadmin stuff for a job. I don't always care for doing the same here in my free time on the forums here as well.
Terminal Velocity
Well-Known Member
And so far I only had minor problems that are trivial for knowledgeable users here... Also the questions you make matter, often you have to dig yourself in order to make the right question
Indeed, that's probably the hard work on the human side ... formulating the question correctly.
For better or worse, I've never used AI to solve a problem with Linux.
I probably would, but I haven't had a broken system that I couldn't fix in a long time. I have played with AI and used AI as a tool, but I've never needed it for a Linux problem. Maybe that means I'm not trying hard enough? Everything just keeps on ticking along without a problem. I've long since done all the tweaking I wanted to do. Maybe I'm just lucky? Nothing really seems to go wrong these days.
I probably would, but I haven't had a broken system that I couldn't fix in a long time. I have played with AI and used AI as a tool, but I've never needed it for a Linux problem. Maybe that means I'm not trying hard enough? Everything just keeps on ticking along without a problem. I've long since done all the tweaking I wanted to do. Maybe I'm just lucky? Nothing really seems to go wrong these days.
AI can provide solutions to linux issues. Certainly, if a solution is not forthcoming through other channels, AI is very much worth interrogating.
Knowing how to ask questions helps, but AI can help with that too.
A scenario can go like this:
The user asks a question.
AI provides an answer.
The user implements the answer, but it fails.
The user informs AI that the answer failed.
AI provides an amended answer.
The second answer is implemented but fails.
The user informs AI of the failure and asks AI what further information would be helpful.
AI suggests which extra information would be helpful for it.
The user provides the information.
AI proposes another solution.
The solution fails again but this time the user has logged the error messages.
The user then provides AI with all the error messages.
AI then proposes another solution.
This sort of conversation is not unusual here when dealing with ChapGPT, when logged into it and keeping the conversation focused on the problem at hand.
The AI "responder" is not a person, so it won't tire of responding. It's programmed to provide answers. It's never come to a point here where it's given up, but that's possibly because it's been able to mostly resolve the issues.
Here is ChatGPT's answer to a query about the limits of interactions when logged into it:
Code:
Free Tier Limits (GPT-3.5)
If you're using the free version of ChatGPT:
Model access: You only get access to GPT-3.5.
Query limit: There's typically a message cap (number of prompts/responses) per hour. While OpenAI doesn’t always publish exact numbers, it’s generally around 25-50 messages per 3 hours, subject to change.
Rate limiting: If you hit the message cap, you’ll see a timer telling you when you can send more messages.In the above scenario there are around 6 user messages, which I found is not unusual, never having to go anywhere near the 25 to 50 mentioned in the quote to have issues resolved here. It's going to depend on the problem I guess and the user's elaboration of it. That all said, AI hasn't solved every problem I've presented it with in the way I wanted them solved, so some are still hanging.
Last edited:
Condobloke
Well-Known Member
I use DDG as a search engine. It provides an "AI" response at the top of each search page.
I compare the ai answer with the first search engine response on the page. This gives me a 'clue' as to how accurate the AI response is....I also make allowance for the fact that serach engine response may also be inaccurate.
So far, I have DDG's AI to be helpful....with around 85 - 90% accuracy.
I have quite a degree of trust in DDG...that trust transfers on to the their ai
eg:
That looks really useful for me to study. Thank you very much Osprey.
How long have you been 'doing' Linux? I bet it was before the AI came on the scene for everyone to use in its current form.
For us, newbies, AI is a godsend as we are completely clueless about things and not having the solid background many other Linux fluent people do, who also often learned it as a subject in computing course at school/college/university or spent their teens and following years messing with it, so they can now consider themselves Pros. For those, from before the AI times, they obviously have different habits and routes to fix things.
I personally found AI very helpful for a prompt answering of my questions, analysing sensitive outputs (not too sensitive though), and immediately offering fixes. There is stuff I would not put publicly on a forum as I am a bit paranoid hackers are on the watch.
You must be the Shere Khan of the Linux world KGill. Mind you, you aren't half tempting fate with that last sentence!!!
Brave man, beware the FIRE!
Last edited:
A long time. Well... No...
I did play with Linux, not long after it came out. We also used some Linux servers at my old company, eventually.
While I have used Windows a great deal, I haven't used Windows since Vista was still supported. I moved my own computers to Linux in 2007.
I also have a background in UNIX. I got a lot of early help from a company known as DEC, and I would then use DEC in my professional career. They helped me by providing equipment during grad school. When they were moving to Compaq, we moved into a lot of SUN equipment. So, I've used UNIX for a good portion of my life.
Amusingly, back in the day, we'd have Windows on the workstation but we'd actually be using a full-screen UNIX session forwarded by X11. Those were the days...
Meh... If it breaks, I have /home preserved. I have a /home preserved for every computer I use.
If it breaks, it'll give me something fun to do for a while. I don't preserve anything but my home folder, so I'd just do a clean install, move my /home back to where it belongs, and then install the various applications I need to get started. I'd then install the rest of the software as I needed it. Because I preserved /home, all of my settings will remain. For example, when I install a browser, I can then open it and have all the same open tabs I had before that.
It'll give me an excuse to get rid of a bunch of installed applications that I don't really use. So, that'd be cool.
