I have found my permanent distro, Manjaro. I turned off aur and use only flatpak. I use an AI to solve any problem and for cli solutions to modify kernel and os. I use to spend hours on Google searching for solutions. Having used Linux since the 90’s using AI has made Manjaro a sweet experience.
Bad idea. I know at least several users that followed AI instructions and broke their system.
4 Likes
These Language Models cannot even do basic math.
At best they may be able to aggregate some amount of common responses as found via a search engine and distill them into conversational answers.
But therein lies the problem - it has no concept of correctness, and certainly not as pertains to your specific linux distro.
It might give you decent answers in some cases, but there is no guarantee.
( See other examples of asking a chatbot about tech questions and getting fix it steps including “Kill yourself” ala its high frequency of use in responses on reddit. )
You use such utilities at your own peril.
10 Likes
well… you have to use a little critical thinking. Dont take everything it says as facts. I tried several and so far ChatGPT seems to be the best so far. However, they are evolving so fast, what may trip it up one week, is completely solid the next. I admit i absolutely suck at programming. I needed a script to do an offline backup of my 60TB server to spare hdds with different sizes. It took quite a few iterations and got a script that while not perfect, works quite well.
Also, it was invaluable migrating from ext4 to zfs for my pc. It even helped getting zfsbootmenu working correctly. It wasnt perfect. The biggest problem is memory between sessions. The free tier doesnt have much. I have been playing with deepseek locally and not sure what to do with it yet. There is no memory between sessions and no internet access.
One thing it does very well is syntax searching or recommendations for optimizations.
tl;dr:
do not take the advise for face value. Test in a safe environment. But it can help.
There is a common misconception about AI.
Many people think that AI can solve problems.
But AI only looks for solutions that others have already used more (or less) successfully and documented on the Internet. No(!) thought process is involved. That is why AI can easily “solve” problems that have already been solved many times. But it cannot solve problems for which no examples exist.
The rarer the problem occurs, or the more specific it is, the less likely it is that an AI can offer a solution.
The main problem, however, is that AI lacks judgment. In particular, judgment about its own abilities. This means that the less AI finds out about the problem, the more likely it is to start hallucinating. It does not know the limits of its own abilities.
Human experts sometimes say that they do not know something. That is a sign of competence.
AI has no competence!
4 Likes
Good advice (generally). This also extends to much of the advice many take for granted on forums. 
Always verify information when possible.
Regards.
1 Like