Hi Pierre, thanks for writing this,
I usually do my best to not jump on the keyboard, but when I see the long thread on -private that I will never have the time to read, I think now is my only chance to contribute. I will not post further in this thread. Let's keep it open to as many people as possible. Debian can, and should, contribute to the writing of entirely Free generative models that satisfy our needs and our values. We have data. We have time to curate it. We can team up with other organisations that share our values. I can not imagine that if we join forces we would not be able to produce something useful. We have patience when we need and we have impatience to get things done. Generative language models empower billions of people to use the priviledge language of the day. Think of your life if in 20 years everthing important goes in a language far from the ones you speak today and you have no computer assistance to learn it and use it. Think of our life if we could, in an ethical, social and ecologically responsible way, ask to a language model "do you think I will hurt someone by posting this email ?". [ Copilot gave me a warning about taking a specific example about languages, and I fuzzied the "think of your life in 20 years" example to milden the issue ] The LLMs that I use know well enough about Debian to give me access to old knowledge that is new to me (hello, libdpkg-perl) and that I would not acquire otherwise because I live on a small island in a timezone far from where Debian's heart is beating. I want to keep access to that knowledge, in a more socially and ecologically responsible way. AI made with Free Software for Free Software can make us truely global and help us survive and rejuvenate. Have a nice day, Charles
