AI in Linux? Yes or No, Comment! 🐧 πŸ€–

#linux #AI

@itsfoss

Stress free NVIDIA drivers

@itsfoss Of course! Is a must. You canΒ΄t stop advancement. Open Source must be involved. What do I want? Pretty much anything MS do, but well guided

@itsfoss

Nooo!

Currently i don't really see "AI", but bullshit...

@itsfoss I think Linux just has to. There is no real choice. AI is forced upon all of us, refusing to implement it is not a option, but showing a alternative more privacy focused and ethical way of implementation is the task, I think Linux (and FOSS as a whole), has.
@itsfoss No Ai into any of my devices.
@itsfoss I would like to have ChatGPT implemented directly to the kernel.
@itsfoss - what do you mean by AI? If Things like ChatGPT, Copilot+ Recall and all that other tech bro LLM shit, then NO.
As long as "AI" is not defined there is no valid answer to your question.
Just my 2ct.
@itsfoss
Fortunately we'll always have a choice. Even if gnome and kde both implement AI in their main design, environments like xfce, lxqt and especially window managers will not. Personally, I'd rather that AI stays out of Linux unless it's a part of an optional feature for some app. I just don't see the benefit of integrating it in the core desktop experience.
@itsfoss AI will make Linux irrelevant pretty soon with bare-metal hardware support for AI infrastructure hosted remotely, or in some kind of a hybrid ecosystem. Add to this natural interfaces based on voice, sight, touch and human-like interaction in real time and that's it. It's inevitable and it's already happening. People will love it, for sure. And will forget about openness and independence in software creation, most people do not care about it and never cared.
@greblus @itsfoss counterpoint: (the lack of) AI is making linux even more relevant now. Which is actually untrue, as AI can and usually runs on linux based servers. Linux gets the ability to run AI locally and no comercial crap by default, so i see it really hard to lose any relevance

@itsfoss Tools to run and develop ML models? Sure. Integrating it as part of the OS? Absolutely not. I don't need my OS guessing at what I want to do, or making things up for me.

At best, there may be an ecosystem of apps that use related technologies - maybe for accessibility, or searching/summarising documentation (or files) perhaps?

But if these are exposed as anything but optional packages/features (or worse, as features that train themselves by aggregating user data collected over time*, rather than pre-trained, local models) then the distro or desktop environment in question would likely see a mass migration of users away from their platform for sure.

* Maybe personalised training could be a benefit for accessibility-related use cases? Or is determinism and well-defined behaviour better? I feel I don't know enough about it so would like to hear ideas and thoughts on that from people who know more about it...

@itsfoss Please God, no! I promise I'll be good!
@itsfoss What is the point of installing a software that will steal your data without understanding it ? (I don't want glue in my pizza).

@itsfoss
Yes.

LLMs are too important a development to ignore.

Yes, there are problems associated with how training data are acquired.

Yes, there are problems with the behavior of some organizations promoting LLMs.

Yes, there are privacy issues.

Nevertheless, there are potential benefits to using LLMs as a tool for various tasks and the Linux community should not stick it's head in the sand and ignore this.

It is important to address these concerns and Linux is best situated to do so.

@itsfoss not against, just not the big tech ones

@itsfoss I want to finally see accurate, multi-language voice typing that has zero setup other than installing a single package (or preferably, out of the box).

And I want to see an assistant like Siri/Alexa/Cortana that works locally on-device.