we are seeing a massive shift where AI is moving from “something you run on Linux” to “something that runs the Linux system.” I know that Redhat RHEL 10 introduced command-line assistant with lightspeed, what about other distro’s. I am curious here answers from you. Thanks in advance..!
Ouch! I hope not. That would be a nightmare come true 
If that becomes commonplace I will run to Gentoo…screaming!
6 Likes
I think for sure before 2030 we will have a few distros available with AI CLI. 
What will be interesting to see is the reception. One of the many reasons we love Linux is transparency and system visibility into what every line, config, path, file, and dir does and also having full control over what’s being done.
No “Almost Done. Please wait just a few mins” with no info on what’s happening on our systems behind that message. So AI integration should not change that, at least for those of us who value this about Linux.
Of course, some of us with lat our reject AI while others will adapt. All part of what makes Linux great!
EDIT: also it should be local only, with remote connections being completely optional.
5 Likes
Regarding the above industrial trending, I have strong opinions which are reflected by a couple of quotes from “Dune - The Butlerian Jihad”, that I will share with you:
-
“Logic is blind and often knows only its own past.”
-
“We learned a negative thing from computers, that the setting of guidelines belongs to humans, not to machines.”
-
“In the process of becoming slaves to machines, we transferred technical knowledge to them … without imparting the proper value systems.”
-
“The intelligent machine is an evil genie, escaped from the bottle.”
-
“When humans created a computer with the ability to collect information and learn from it, they signed the death warrant of mankind.”
… and, as a prediction regarding how things will evolve unless we can put a clamp on the AI technology before it gets completely out of hand,
- “Humans tried to develop intelligent machines as secondary reflex systems, turning over primary decisions to mechanical servants. Gradually, though, the creators did not leave enough to do for themselves; they began to feel alienated, dehumanized, and even manipulated. Eventually humans became little more than decisionless robots themselves, left without an understanding of their natural existence.”
… and that pretty much sums it up!
When I first read the various Dune books decades ago, I never imagined that the technology’s rate of advance would explode the way it has recently. I never imagined that ubiquitous AI would happen in my own lifetime.
Now it is upon us, but in a way that makes me fearful about how much longer we have, as a species, on this planet!
Human-compiled rule-based expert systems: good!
Artificial Intelligence, without a leash: BAD! BAD! BAD!
4 Likes
Present Ubuntu attitude towards AI is outlined here:
7 Likes
Thanks for the update. I seems that I really have to find an alternative distro if faeces hit the machina venti 
(btw, i added a reply of concern to that post, which was immediately flagged down. I’ll forward that post to you.)
5 Likes
How Linux impacts AI revolution.
1 Like
Or maybe, “how AI impacts Linux” 
2 Likes
Do you wanna know the real impact ?
soon all the linux distributions will ship with open source 2-4-bit Quantization models that hear you talks to you do actions monitor you 24/7 and it’s gonna be the normalized standardized even linux main file system tree will change by including AI system directions I know people right now not ready for this
but no escape and for me I’m big AI fan and willing to contribute
Do you want to complex ffmpeg commanding ?
ask AI
Do you want to check feed on linuxcommunity.io
ask AI
Do you wanna send emails ?
Yes ask the AI
Do you want to Sudo install package
just ask the AI
Need to shutdown you machine
no, way you gonna click gui or type commands just ask the AI & it will do it for you 
4 Likes
If the AI is trained on facebook answers, it might do the shutdown by issuing:
rm -rf $HOME
8 Likes
Sad thing is that there is a real possibility it might just do that!
6 Likes
My expectations be like :
AI models will have it’s own package manager that very module
for example linux core-util AI package is very tiny model that trained only by offical developers to do the right stuff same goes for other thousands of cli and system software for example
the pm is artificial intelligence model (aimod) & daemon
aimod install neovim
>>https://aimod.com/models/neovim.tar.zst<<
%[>>>>>>>>>>>>>>>>>] 100%
16 MB downloaded, installing...
/intel/models/neovim
~>
~ > aimod -pipewire-input
⠕ how to exit vim ?
⸫ press escape then type ":q!"
⠕ I don't have keyboard right now can you exit it for me ?
⸫ PID check "604"
⸫ neovim quit ,done
⠕ thank you,
⸫ you are welcome,closing
~>
I’m just wondering which organization gonna initiate that.
4 Likes
It’s worth separating cloud AI services, from local LLMs. They are very different beasts.
I think local LLMs that are not web aware, can really be empowering and time-saving tools on Linux desktop and CLI. I can probably write about some examples of this later.
The local option keeps everything on your machine, no telemetry, no API calls. For anyone curious and seeking hands on understanding, three solid starting points are: Ollama (easiest), LocalAI, and llama.cpp.
More here:
But again the beauty of Linux is that whether it’s snapd or systemd uses will always have the freedom to fully embrace or completely avoid. 
4 Likes
Nothing with high quality, assistants only cover the mediocre cases only. Days ago, on a test machine for Code Development configuration, I install node 2.0 from source, nodejs and install with the classic curl -fsSL ``https://claude.ai/install.sh|bash
Not great answers, great cash for anthropic-ai.
It’s not cost-effective. It’s like a generic cut-and-paste template that ignores the context and goals of your application under development. This development that you have shared with the model, and which it will reuse if it enters parameters for other responses once debugged. Nothing cash for you, Cash for this business with your code (of course, you accepted the license, is not only your code, but you have to pay for AI model reading and answers via token cash.
Same for Copilot, ChatGPT Pro,…a great tool, but you must know what it’s for, and recognize what it’s not for, and what the developer who creates the application is for.
Soon, even to enter commands, we’ll waste more time convincing the LLM model and its agents of what we want, and finally, the model will graciously give us what’s generally accepted, but not correct or efficient.
Regards.
8 Likes
Sounds quite similar to the useless results of the regular search engines of today. 
4 Likes
Trying to bring this topic back to the forefront to stimulate activism amongst the younger crowd who will need to live with (suffer?) the consequences of poorly managed design/roll-out of such Artificial non-Intelligence tools.
I’ve submitted a rather lengthy post in response to what I perceived as a challenge from Alan Pope on the Ubuntu Discourse site. I would encourage as many as possible to review that response to help frame your own stance and position regarding what should be the industry’s, and Canonical’s, next steps in pursuing what is recognized as a very dangerous technology by many people of sound mind and body!
1 Like