The future of AI in Ubuntu

I think it would be good to mention the specific usecases you have in mind for AI, as the technology has rightfully earned a bad reputation. I already see tweets about this decision. First good usecase I can think of is improving the Help application since I doubt users actually use it when they encounter a problem. Or maybe a utility that renames New File (2) into a name based on its content.

Having all the models running as confined snaps is really cool in my opinion. No need to worry about one spying on the entire operating system.

2 Likes

I think I was an early reader of the document that opened this topic. I do not think anyone had made a reply at that time. I wondered if I could say anything in reply. Here is my response:

Many, many years ago when I was young I read a small book by Albert Einstein. In the book he explained his theory of General Relativity and his other theory of Special Relativity. The book was written in English. Yes I am certain it was written in English. I did not understand a word of it!

I feel the same way about the document that opened this topic. Even so, I would like to remind the readers of this topic of the little I did understand.

This year Canonical has begun a more deliberate push toward education and developing competence with AI tools. We are not setting shallow metrics on token usage, or percentages of code written with AI, but rather incentivising engineers to experiment and understand where AI tools add value. Rather than force a single early-choice AI stack, we’re incentivising teams to each pick ‘something different’ and go deep, so we learn more as an org in the next six months.

Do we want Canonical Software engineers to be taking Ubuntu into the future ignorant of what is happening around us?

I also see this:

Part of our role as long-standing members of the open source community is to stay at the forefront of what can be achieved with AI tools, then lead by example. We should be demonstrating what can be achieved through responsible and thoughtful use, and guiding new contributors toward better practices that will see them using these tools to amazing effect, and contributing to the next wave of open source for years to come

I think that is a worthy purpose.

Regards

4 Likes

How’s a local model running on your machine, a bigger threat than google?

3 Likes

Google is a choice, and it is safely on the other side of my internet connection. I, for instance, don’t want google integrated in my systems.

(btw, I don’t use google)

5 Likes

i dont use google either.

duckduckgo, cloudflarewarp and vivaldi browser for one simple reason no AI.

best of luck Steve ..

4 Likes

I’m not against LLMs ( I even have ollama installed on a separate machine ). But integrated in an OS is for me only acceptable if I can curate c.q.cherrypick the trainingmaterial in advance. I must be able to trust the LLM output.

2 Likes

using AI to construct the OS and find flaws to improve security, stability and reliability not a problem but not within the OS, that is user choice. MS have made that mistake hopefully Ubuntu wont.

because if they do other OS’s dont have or will have AI within the OS such as Zorin.

we shall have to wait and see which way ubuntu jumps

best of luck Steve ..

4 Likes

And how’s something that you only use if you want, and that doesn’t leave your PC threat to privacy?

1 Like

I don’t exactly understand what you mean by that. Would you be so kind to rephrase that ?

1 Like

I will preface this by saying that I’ve never worked on developing an operating system, nor do I know the scale or depth of the AI integrations intended, how intertwined they will be with the underlying operating system, so I do not know if this is a possible or even feasible idea, or if doing so would involve a tremendous amount of additional overhead.

That said, looking at how polarized the responses in this thread so far have been, I think it’s worth at least mentioning in hopes it can be done: Perhaps it would be wise to release two streams of post-AI Ubuntu, one with AI integration and one without? Those who like Ubuntu but dislike AI can continue to experience the operating system without it, and those who like Ubuntu and are open to AI can enjoy the two being integrated.

I can’t speak for others, but I can definitely say that with a lot of people I know who’ve expressed a curiosity to trying Linux in the last few years, and particularly those who have moved from “A curiosity to trying Linux” to “Actually daily driving Linux,” a common reason for wanting to leave Windows 11 is to find refuge from the heavy AI integration that Microsoft is aggressively pushing in its products. Another common reason is to breathe new life into old, outdated hardware that is not able to run the latest versions of Windows, but IS able to run Linux, giving a second life to perfectly functioning hardware that would otherwise become e-waste.

While not universal, these are fairly consistently two of the top reasons that people I know who were never interested in switching to Linux and open-source software before are interested in and actively making the switch. I don’t imagine I need much explanation why AI integration flies in the face of people seeking to keep their digital space AI-free, and while I do appreciate and agree with @jnsgruk ‘s point about smaller models eventually being able to close the gap with bigger, more powerful models, that tension does still exist as of now. That puts you in a Sophie’s choice of either integrating more demanding local models at the expense of availability and compatibility, or integrating smaller, lighter models at the expense of quality and performance.

One of the great assets that Ubuntu has in my opinion is that it is one of if not the most accessible way for someone who’s new to Linux or not particularly tech-savvy to get started. It’s easy to set up, involves minimal tinkering to get rolling that might scare off someone who’s new to the world of Linux and open-source. While I understand the need to keep up with the times and integrate newer and better technology where it can be beneficial, I can’t help but feel like it’s a bit of an own-goal for a distro that’s well-positioned as the gateway into the world of Linux to integrate AI in a way that isn’t optional given the reasons that a lot of people are making that first dive right now.

Given all that, I go back to my original point: Might the way forward be to have an AI-free “Lite” version and an AI-integrated “Max/pro/etc.” version? Again, I don’t know how feasible it is or how much extra demand this would place on contributors or if it’s even possible, but it seems like a reasonable compromise that would let those who want AI integration to have it without alienating users who explicitly don’t want it. (At the very least, making it an option that’s uninstalled by-default and requires explicit opt-in would also be a good option imo)

3 Likes

Greetings from the Moderators.

We have engaged Slow Mode on this topic due to the large number of low-quality posts in the past few hours.

We strongly advise everyone to read and understand the entire first post, and subsequent clarifications, before posting in this topic.

This topic is for technical opinions and questions.
Leave your flamethrowers behind.

5 Likes

I truly understand that, at first glance, AI offers huge benefits that any sane person would want to pursue.

The issue with AI, in the current context of explosive growth without necessary controls, is that it is too dangerous a technology.

The floodgates have been opened wide!

Praying that it isn’t too late to make the attempt to “corral the wild stallions” and “tame the beasts”, I invite you to consider some quotations that I have posted in this other forum, on this very topic.

I thank you in advance for your willingness to reflect on those insightful quotations (not mine).

3 Likes

Im going to be honest, thats kinda hard to do as it all feels very marketing speak which is very common for any AI introductions into products.

It would go a very long way to clarify if this if this is going to be opt-in or opt-out. Opt-out is always the definitively worst option to choose and especially with such a hot button topic as this the ubuntu folks are asking to shoot themselves in the proverbial foot if they got that route.

I would simply ask that there be a clear determination on if this is going to be something users with have the option of installing themselves or its going to be something that is installed without consent and the user will have to go out of their way to remove it.

There is already a fair amount of animosity towards canonical and ubuntu when you bring up the topic of snaps and i think overall ubuntu is a good distro but if you dont want to have to get more fire retardant gear on for the deluge of vitrol fire that is bound the come if this isnt handled well then it might be best to be far clearer as ive gone through the thread and read the entire OP and still am confused as to the exact intent here.

8 Likes

Houston, we have a problem.

Redhat plans to add AI to Fedora and Gnome

Redhat infects Gnome and Gnome infects Debian and Debian infects Ubuntu. Before you know it you are using software written by unethically trained AI. Is that worse than using software written by unethical humans?

Principles are the immovable object being hit by the AI irresistible force.

5 Likes

With all due respect, the idea of integrating AI features directly into Ubuntu is going to be a controversial decision that is already drawing backlash and skepticism, and disappointing to me personally as someone who sought out Ubuntu specifically to find an operating system that respects users’ choices and avoids poor decisions like forcing controversial features like AI. Personally, the better option would be to leave AI features, especially the “explicit” AI features you mentioned as either optional or as a separate distribution or set-up.

I have no issue with users who do choose to user AI programs like LLMs, but I and many others do not care for these features for a variety of reasons, especially ethical ones. MIcroso. If an AI killswitch like the one Mozilla added into Firefox isn’t on the table, then having a version of Ubuntu that does not include these features would be the best way to respect users’ agency and choices in what is included (or not) with their operating system.

6 Likes

I am shocked that the privacy (& so security) implications seem to have been mostly ignore. If a computer isn’t powerful enough to run an LLM, that means it must use a ‘cloud’ LLM service - and basically all of those are known to keep permanent records of all user input for training, etc. And a lack of privacy is just one cloud security breach (or hacked online account) away from becoming a target of hackers yourself.

To be frank, I don’t want my computer’s logs (and whatever else might be sent) permanently recorded in the cloud, no matter what their opaque ‘privacy’ policy appears to claim.

An “AI kill switch” is absolutely essential, just from a privacy & security point of view. It should have at least 3 options:

  1. Disable all AI features.
  2. Only use local AI compute (and if that’s not available then the feature is not available).
  3. Full AI features, which requires logging into your chosen LLM provider that you trust.
7 Likes

A mature post that I think will age well.

I can see AI having another big impact on open source projects which large communities, and that’ in issues and PRs. A project I like is now drowning in them. github has no real clue on how to help. The project has bug tracker agents, but they are conflicting: an agent to more broadly find duplicate issues to manage the influx is keeping old issues fresh when they should be allowed to go stale, for instance.

I suppose Ubuntu already gets a lot of low quality bug reports, so it has mechanisms such as issue voting to help, and a lot of the patches happen upstream so it’s someone else’s problem, but I still worry about this aspect of gen AI getting out of control.

At least Ubuntu has bug reporting tools, maybe they can deploy AI well to catch duplicates and low quality issues at the source.

2 Likes

Imagine being able to ask your Linux machine to troubleshoot a Wi-Fi connection issue, or to stand up an open source software forge that’s pre-configured, secured, and reachable over TLS.

This is a terrible idea. Asking a black box of any description to configure security-critical software is a massive red flag.

8 Likes

You (they/we) likely already are running code written by AI agents, and reviewed by human meatbags. Indeed this very forum software we humans are using to conduct this civil discourse has AI steering.

(I linked to the markdown files in repos that are good indicators that the project is amenable to AI-based contributions. I have not audited either project to determine if they have accepted such contributions, but it’s a good indicator)

Not passing judgment, just adding data.

1 Like