Google and Facebook trained their neural networks on Stormfront, Kiwifarms and pirated ebooks, amongst other things. Because of course they did.
https://www.washingtonpost.com/technology/interactive/2023/ai-chatbot-learning/
@mathew I don't believe those I'm not a robot pictures where you have to find traffic lights can be safe against AI. If they are it isn't (AIn't?) intelligent
@JohnLoader6 @mathew AIs are not intelligent.
While I don't believe there's a theoretical reason why we couldn't build thinking machines one day, we're nowhere near building them today, and none of the applied research done today is leading toward that end.
AI is a misnomer applied to any research models that came out of an AI lab, if even that. It basically means "complex automation" these days, just anything that does pattern matching and/or prediction.