AI is the word of the month/season/year right now and the council is trying to come up with some basic guidelines for the project around AI/ML, but in order to do that, we need to know what our community would like and perhaps even need from AI capabilities in Fedora. So we would like to survey our community on the subject of AI, and I need your help coming up with questions to ask. The questions are completely undefined, so feel free to propose any kind of AI-related questions you think should be included in the Fedora & AI survey.
As a side note, there will be an annual contributors survey during the summer too with a section on AI, but this survey is intended to be a precursor to that and a source of information the council can use when definiing what AI guidelines are for Fedora.
Please pitch me the questions you think are important to ask our community on the subject of AI so I can get this survey created and out as soon as possible!
What would Fedora require to highlight contributions that involve AI? Code comments, etc. (If you canât even tell if something was AI generated then asking about it is pointless.)
How does Fedora define AI?
How does Fedora delineate between solving problems with automation and solving problems with AI?
How does Fedora ensure there is a human in the loop for both AI and general Automation?
Should Fedora use existing contributions or new contributions to create or fine-tune their own AI models to solve Fedora problems (e.g. AI chatbot for Discussion FAQs, AI-infused zodbot to make it better at capturing online meetings and producing actual readable summaries, AI assistant for generating spec files, AI models for OpenQA work).
What AI, Where? Do we need AI? Another copilot integration? Discussion only? Inside Fedora? Can we just leave AI opt in/out? What is general purpose of AI? So many questions so little answers
aI is the boom and trend, but at least leave it out of operating systems those who wants AI have options to use it still there
What are some potential applications of AI in Fedora that you would be interested in?
Developing AI on Fedora
Utilizing large language models for local tasks and code generation
Generating images using AI
Using a local AI assistant
I am not interested in anything AI related
I am not that good with creating the choices for this question because I am not that interested in AI. If someone knows better options or wants to add anything feel free to edit this wiki post.
I think having a more general survey might also be interesting. Would be interesting to see what people use Fedora for and what they would like to see improved. Or does something like that already exist?
Would Fedora consider using AI for content moderation as a potential tool for keeping up with moderating potentially AI-created spam in community spaces if that starts to become a significant drain on the time of the individuals/groups that handle that?
Are more natural-sounding TTS voices desirable or undesirable? See this video from a user who is visually impaired. This user prefers robotic voices as they are more accurate, faster, and just more usable.
On the other hand, I think Mozilla has found a great application for AI with auto-generating alt text for images that donât have them. Something small that could make a huge difference for visually impaired users. This is a problem that could only be solved by AI, even though itâs a very small, self-contained issue.
Do you think Open Source can make AI/ML technology better and more accessible?
I want to better understand the feelings that someone might have when we talk about âOpen Sourceâ and âAI/MLâ at the same time. I suspect survey respondents will have strong opinions about both topics. But what happens at the intersection of these topics? Is our contributor community more optimistic or pessimistic toward the outlook? I would like to know this.
I think this is a good question, but probably not a good general survey question. Itâs something we should ask of our users and contributors who need or benefit from voice interfaces.
I donât use these technologies for that (at least, not at this point in my life) but I do for home automation, and one thing to note is that the state of the art in pre-AI open source TTS is quite poor compared to proprietary offerings.
An advantage of using open source AI for this is that we have the opportunity to tune such voices how we like. We can offer more ânaturalâ voices for those who like that, and for those who need something else, adjust parameters so the voice is more like the âroboticâ screenreader â fast, but with an even cadence, without trying to interpret tone. And, we can do this in many more languages with better pronunciation than we possibly could with, say, festvox.
But, whatever we do, we should do it in conversation with our users and contributors who rely on the tech.
Thank you very much for sharing your ideas on the kinds of questions we need to be asking our community on AI in and with Fedora! Next steps will be to create a survey based on your feedback, and get it out to our community as soon as possible.
For an idea of timelines, as I am travelling to Brno this week to attend Devconf.cz, please allow a little flex time, but my goal will be to have this survey ready and launched by Friday 21st June. Your responses to the survey will be instrumental to how Fedora approaches and adopts AI so when it becomes available, please do take some time to fill it out.
Thanks again everyone, I really appreciate your time and ideas and look forward to our next steps
Thank you for contributing to our discussion on what kinds of questions are useful for us to ask our community on the subject of AI in the Fedora ecosystem. Weâve created a short survey based on the questions that were proposed that we think are general enough for everyone to feel comfortable answering, and from the answer we receive, the council and other governing bodies of the project can start formalising some AI-related focus areas and develop solid guidelines for the use of AI in the project ecosystem.
The deadline is July 16th, it shouldnât take more than a few minutes of your time to complete, and I will be sending some reminders between now and the closing date too.
A quick reminder that this is not our annual contributors survey, and that survey will be launching soon with a section around AI too. I want to offer an advance apologies for any survey-related fatigue, but when it comes to creating AI guidelines for the project and understanding how our community is doing and feels about the project, this AI survey and the contributor survey is truly the best way to get that feedback and create actionable stuff ânâ things from your direct responses.
Hi - Iâd like to point out one issue I see with the survey. Particularly when it comes to the question âWhere do you feel like AI/ML has the best fit in Fedora?â where you rank the answers from top to bottom and have to rank all 5 suggestions. I think this question is going to give you a heavily biased answer towards AI/ML inclusion for the reason that you did not provide any way for someone to remove an item from the ranking. The reason I ask is because of these two items:
âDeveloper Assistance (e.g. copilot)â
âAI in the operating systemâ
Both of these are items I feel uncomfortable giving any response to that isnât a clear âNo. please do not do this at all in any capacityâ (especially in the wake of the fallout from microsoftâs Copilot+, I was surprised this was mentioned at all). Iâd like to make sure Iâm getting across to surveyors that tools like this are things that should actively be kept away from the community due to the enormous PR risk they carry. Otherwise it makes it feel like the only option thatâs being given is one that feels like itâs saying âwell, I donât think we should keep these things out of Fedora - I just feel theyâre less important.â
Thank you for sharing that feedback Lyude. We deliberately tried to keep the tone of this survey positive about AI as it is very easy to find negatives for the use of AI (and for good reason!) and we didnt want to take that route. We wanted to approach the questions and uncertainty around AI and its potential uses in Fedora from a positive view and useful application, but I will reassure you that just because we are asking you to rank your preference of AI in certain areas of the project, does not mean we will be introducing AI into all of these areas.
We are striving to understand peoples preference only and any AI introductions into Fedora will always be done in the Fedora way - an open conversation about intent, community feedback, and transparent decision-making and/or planning that may follow after.
If you have not submitted your responses yet, I would suggest ranking those areas you mentioned as last/2nd last, and in the final question feel free to share your concerns and preference to keep AI out of the OS. That way the response is recorded formally.
I appreciate the insight as Iâm sure others may have felt similar, so I hope this workaround helps you and others to still complete the survey but in a way that you can express your honest feelings about AI and Fedora.
Would you like a simple âopt out of AI on my Fedoraâ setting (during install / usage).
If Iâm coding or creating something, I don´t want my content being sampled to train AI, so I for one would like an easy to use setting to keep my work private from that thanks.
I filled out the survey. Mostly uncertain for everything.
I tend to use AI for small, self-contained situations. Like when Iâm proofreading, and I know something doesnât sound right but Iâd like to see it rephrased in a few different ways to see what direction clicks better. Sometimes itâs even so good that I use it without editing. And a grammar checker is good to have, of course, even if there are a lot of false positives that do more to eliminate style or distort meaning than make writing more professional.
Iâve found summaries to be a really useful application for AI, especially where correctness is secondary.
I donât know what would be useful at an OS level that is also something I donât need to trust with my private information or that Iâm not particularly concerned about being incorrect. Microsoftâs Recall is a vaguely interesting concept overshadowed by the privacy and security concerns.
RecommendationsâŚmaybe? In Windows, Recommendations on the start menu are paid advertisements which are rarely useful. If you indicated to Fedora that you were interested in art, say, it would be useful if you were pointed toward applications like Krita and Aseprite.
But then, are recommendations generated by an algorithm for a particular âprofileâ worth anything compared to hand-picked ones like Flathubâs App of the Day? Discoverability is a real problem especially for newcomers, but I think solving it with LLMs brings more problems. For example, accidentally recommending proprietary software.
In my experience it seems like the biggest thing people are worried about with AI is their content being used/stolen by AI systems (as has become such a popular thing with the big image generation models)
I think theres several distinctions at play here that may be useful to tease apart into distinct survey questions to help frame the AI discussion productively when evaluating AI policy in fedora. Notably:
What data was the AI trained on? was that content misappropriated or used according to its terms/permission of those who are in the dataset? is that data available for review? (this essentially targets the âis it trained on my stuffâ concern)
Is this an AI that self-learns as it is used? (this addresses a similar, but different concern regarding whether user content will become part of the AI model during use)
what is the purpose of the AI? does it have creative input or is it simply a translator from one form of content to another (text to speech, image to text/OCR, etc)? (This seems to be a category of tools that are VERY easy to objectively measure the performance/quality of and have very little room for bias, which could be a good easy way to start introducing AI into fedora in a caution-first way)
Kinda - this plus consequences.
Encarta eating quality Encyclopaedias didnât cause too much pain if you didnât sell, contribute etc to the latter for a living.
AI eating a whole bunch of quality service professionals is going to be horrendous.
20% garbage is going to be sold as âwe did trials, humans make errors too.â
Once the livelihoods are destroyed the training, quality control etc pathways are gone.
Edit: privacy issues are also underrated - âItâs not electronic surveillance, itâs just scraping for AI trainingâ.
