Post B4nsxpDvxgzHD8zLSi by daya@princess.wedding
(DIR) More posts by daya@princess.wedding
(DIR) Post #B4npKuBzFsYTSrpNsO by daya@princess.wedding
2026-03-31T02:37:27.382163Z
0 likes, 0 repeats
is the world a better place with generative AI?
(DIR) Post #B4npWJgaoK7jiWaJEm by mangeurdenuage@shitposter.world
2026-03-31T02:39:35.948284Z
0 likes, 1 repeats
@daya No, and to be precise, it's not better with proprietary generative AI.
(DIR) Post #B4npc4lvApVdCQp6bg by daya@princess.wedding
2026-03-31T02:40:33.275624Z
0 likes, 0 repeats
@mangeurdenuage is it better as long as you use a free model?
(DIR) Post #B4nqg9fszI2zW8VdVQ by mangeurdenuage@shitposter.world
2026-03-31T02:52:34.896245Z
1 likes, 1 repeats
@daya According to the FSF, and I agree with them, source code is not enough for make machine learning free as in freedom.You need to train it on data that is also copyleft or compatible with copyleft, the results, the resulting model of that must also be under a copyleft or compatible license and same for the source code/file configurations.It would have been better first on a moral point as copyright still exist and it's the only way to protect ourselves against proprietary software, at least until laws become sane.Because currently all models are being trained on a lot of data that has been leaked and then that is commercialized by these entities who have no moral compass.Same problem for the resulting models, they're basically black boxes and most of them do not know how they work.Then if you add another layer of black boxing with the code.So imagine if a large amount of society becomes dependant on something they cannot even control.Well we don't have to imagine we already see those results with non free/libre computers, it's already a big negative, so with a new type of automation that can give itself new instructions you can see where it already leads.
(DIR) Post #B4nr1FPoq2NzP53H3g by sarvo@novoa.nagoya
2026-03-31T02:56:17.728Z
1 likes, 0 repeats
@mangeurdenuage@shitposter.world @daya@princess.wedding a free model can't exist unless you make a free to use dataset, which creative Commons for example doesn't even have a clause to.
(DIR) Post #B4nsK6brtxf3Y8FBJo by mangeurdenuage@shitposter.world
2026-03-31T03:11:00.156859Z
1 likes, 1 repeats
@sarvo @daya >which creative Commons for example doesn't even have a clause toCC isn't the only kind of license in existence. CC is preferable for content rather than code/data. You'd prefer GPL or compatible for that.
(DIR) Post #B4nsxpDvxgzHD8zLSi by daya@princess.wedding
2026-03-31T03:18:04.808786Z
0 likes, 0 repeats
@mangeurdenuage I see, the situation is complex. But you made me to think. But I'd like to think this is helping people right now. I hope the current situation improves and free models increases.
(DIR) Post #B4ntHJMY87eSBgJHbU by mangeurdenuage@shitposter.world
2026-03-31T03:21:42.039541Z
0 likes, 1 repeats
@daya It can help people, but in most cases it damages them, see this paper for example.Plus since you aren't in control of the services, data output will change depending on who's in charge.
(DIR) Post #B4ntUcephfTdnID28m by Paradox@raru.re
2026-03-31T03:23:28Z
1 likes, 0 repeats
@dayaI think it will be. I don't know for sure it is right now, but it will be eventually.It's like the internet. It was a real mess in the beginning, still kinda is, but we've had time to figure out how to deal with it without getting mad at everyone for how they use it.GenAI use has been around for just over three years. It's still messy and hard to tell if the current good outweighs the current bad. However, I support it because I believe we'll sort all that out. It has potential, and we're kinda good at settling on morals.There are some things so prone to abuse and damage that it'd be too risky to have them around, but this ain't one.