AI Survey Questions - What should we be asking?

To me I prefer to use AI in a self contained environment like a web app which has no impact on my system. I would suggest creating a “Fedora AI” spin if you want to integrate AI directly into Fedora so A). This shows the development is still focused squarely on the OS and not trend chasing AI B). allows people to opt-in rather then opt-out and C). keeps the main fedora accessible to people with lower end PCs which cannot handle on device AI processing.

i would vote for fedora implement a local AI (with the possibility to turn off), and a opt-in/out for external AI like gpt. Also i see opportunity for fedora be one of the first linux distribution to have an embedded local AI easy to handle for regular users , since even implementing local Ai is not a easy task for everyone (me included)

i think a spin only for AI would not be feasible, since would need any AI variant version for every fedora out there

I cant see many drawbacks for a local AI that could be turned off or maybe any option to not install from installer

Cybercrime is emerging as a major cause of poverty (livesavings pft), AI is a wonderful attack surface (look at Microsofts proposed recall!), and once the data is off a user’s device, they mostly have no say over where it goes, how its used, how long it hangs around. (‘Sorry your identity got stolen but here’s a copy of the EULA you agreed to.’)
Ethical businesses help to protect users from risks they may not fully understand.

Microsoft also have others AI products, like copilot, recall was a bad implementation because the way AI save the information (unprotected), not because the AI itself.

Are there any “local AI” that do not rely on proprietary drivers and tools ? Fedora would not be able to ship such tooling without breaking it’s GNU principles.

Folks the survey was causing a lot more discomfort with the ranking question being styled as ‘Best Fit’ and not explicitly calling out that the lowest ranked area would be assumed to be the ‘worst fit’, so I’ve stopped the survey and will either relaunch it with the question better phrased and respondents can edit their answers, or I will launch a new survey asking this question in a better, more flexible way. The purpose of this survey is to try to capture sentiment on AI from our community to allow the Fedora Council draft good guidelines (that will likely be iterated on!) for us to have when using AI for or in Fedora, and the way the ranking question is phrased right now is not really working well for that purpose. So, thank you for sharing that feedback with us on that question, we hear you and we will make it better :slight_smile: Its more important you feel you can answer honestly in the survey than keep the one we have in place and it not be the right fit and I appreciate your patience while we improve it and your willingness to reanswer - that makes it feel like this is definitely the right decision to make.

Im looking into limesurvey settings today, and hopefully I can edit the question directly and allow respondents to edit their answers. I’ll be back to you very soon with an update and thanks again!

1 Like

There are! There are many small models that will run on a CPU (even on small devices like a Raspberry Pi), but there are others with various degrees of openness.

We have PyTorch packaged in Fedora now with support for acceleration on many AMD devices. And I’m hopeful about supporting others soon, including the new Qualcomm NPUs.

From a Fedora point of view, the models themselves are content rather than code. We could develop a special policy for them (around training data and methodology, etc.), but that would be a higher standard than we hold for, say, game levels or wallpaper.


Ok, round two is live! Fedora and AI Survey :slight_smile:

Unfortunately the data from the first iteration was lost, so if you could take a few (more) minutes of your day to fill out, or refill out as the case may be, our survey on AI/ML, that would be hugely appreciated.

To accommodate for the later start, the closing date is now 31st July to complete the survey.

The questions are very similar, but I hope the small revisions we made based on your feedback allow you to feel more comfortable providing your answers. In an effort to make the survey as general as we could, we inadvertently made it too restrictive, so thank you for your patience and understanding while we went back to make some adjustments.

Thanks again folks!


There is probably a typo in the question

Upstream code abd software

from section “Is AI/ML useful for the following software use cases:” in the survey. abd should likely be and.