Post #2087074

2026-05-06 13:36 UTC

#Ethics #AI Given that 1. Machine-learning (neural nets and shit, idk) are a thing, 2. We, humans-who-gaf, want to train them ethically (no abuse of humans, no theft), is there a way we can do so without using 1. Data centers, 2. Big Tech? Is the idea of a (Folding-at-home) Training-at-home distributed system a foolish idea? I ask because there&#39;s a weird sitch in the evil-ai-models rn: (afaik) they are *very* good for disabled (blind/deaf) people. How do we keep the good and ditch the evil?

Replies (3)

@dbat@mastodon.gamedev.place 2026-05-06 13:37

Actual irony is that toxic Big Tech llm shit is vital to disabled people. Can we bust this? Can we take over?

Open ##2314912
@johnoestmannmusic@mastodon.social 2026-05-06 13:38

@dbat@mastodon.gamedev.place You should definitely check out AllenAI's work. They are the only ones I have seen building models out of completely open data sets: https://allenai.org/

Open ##2314913
@Agonio@social.vivaldi.net 2026-05-06 13:41

@dbat@mastodon.gamedev.place as with most powerful tools, the issue lies in who controls them. And that goes for data centres. Sure we can have crowd-trained models but that's less efficient than a data centre, even in terms of heat and chip usage. Data centres aren't evil by themselves, the thing is that their owner have a very high purchasing power and end up eating up resources that people need for other more essential needs. If we had any form of democratic control over resources and how they are used (read: the economy), then we wouldn't need to fight tech, we would enjoy its positives.

Open ##2314918