It feels like I may get virtually stoned for saying this on fedi, as people seem to like black/white approaches, but I don't think all "AI"/LLM is bad. VLC's new feature for automatic subtitles for example seems very useful for me, especially as a greek. Subtitling movies takes countless hours that could be spent elsewhere, and in many cases you just can't find any. Alt text/descriptions in images is another area where AI can be useful. As always, the question in each case is who benefits. We can also be against corporate social networks for example, but not necessarily against all social networks, like fedi. #LLM / #AI is just another technology, if it's not used for profit/exploitation it can be pretty cool.
@panos yeah, also it needs to be ethically trained!
stares at github copilot
@panos that’s a good point.
Also it’s growing on me the idea that AI is particularly good at underlying the toxicity of contexts. That means it’s extremely useful when you use it for accessibility, it’s extremely bad when you use it to make capitalistic social networks more profitable for the governance. It make easier for us to see if spaces and companies are malicious for us, if you will
@getimiskon yeah, I just see so many anti-AI memes lately, and generalizing doesn't tell the whole story. I didn't bring up fedi as an example randomly, on several occassions I have talked to people about fedi, and they are not open to trying it out because they have generalized that social media=bad. It's a way to look at things, but I think it gives a distorted image.
@panos I guess the issue is what dataset are they using and was it built ethically or not? To me that's the part that makes me write them all off as 'bad'....
@reflex one AI open source/community project I'm aware of is AI Horde ( github.com/Haidra-Org/AI-Horde ) by @db0 but I'm not sure about the dataset, so I'll let him clear that out if he wants.