@julia@fedora did you genuinely read the post? Can you share how pushing for ML-powered improvements to accessibility, local translation, and organizing your files with completely free software entirely under your control is a bad thing?
@cassidy@julia@fedora agreeing with cjames here, i don’t see what’s wrong with AI/ML as long as it’s stuck on the user’s machine and not doing any data collection
liking this concept, though I hope it’s removable (for performance reasons)
@cassidy@fedora@julia though i don’t really like the “we can generate package descriptions w/ AI”, AI has been notably tricky with providing accurate information, and I presume smaller ones will only grow this issue for themselves
basically, I’m OK with AI as long as they implement it responsibly (especially in the backend/construction parts)
I saw the ideas from Fedora this morning and i think it's a pretty moderate plan to integrate AI into the OS and process. They are right, there is very little actual open AI and making Fedora a premiere platform for it is a good idea. I think personally my main sticking point is the package descriptions, those are a tad questionable and I would personally steer away from AI generated text.
As for Ultramarine's intentions, we're not sure yet. We've discussed this at length and haven't come to much of a conclusion yet. We want to use more traditional ML in @tau to provide suggested actions and the like, but haven't figured much use for UM (yet). We haven't really discussed the use of AI in Releng and development beyond GitHub's copilot and will never use AI in lieu of human artists.
tl:dr, it's a big question and we need more time but whatever we do, it will not be something like Windows Copilot.
Add comment