These AI training data discussions are always about artists and their work being stolen.
But you know who is really being scammed? The army of thousands of coomers like me, we have spent the last 20 years properly tagging all the galleries on sites like Danbooru and Exhentai.
Do you know how they really train those image generation models to know what they are looking at? Using my work!
@bruh_momento@shota.house I have nothing against the technology but I'd just prefer it if the people making these models actually published them under a free license instead of keeping them behind paywalled proprietary APIs.
@Yoruka Facebook hasn't ever pirated a single book, as they don't have boats.
Facebook did in fact torrent a huge collection of books for the purposes of intentionally infringing copyright and used them to train a LLM and they leeched too (to avoid seeder anti-leech defences, they set the upload speed quite low and let some seeding happen, but only the smallest amount possible and stopped the torrent as soon as downloading was done).
@SuperDicq it's also just hypocrisy these companies always are so much against piracy but then you see them scaping the web like this and allegedly Facebook pirated books to train AI