One thought: humans train on other humans work too

+, I'm curious after seeing companies specializing on it, and trying to understand if they have future. Make sense if data is public, but what do you think about protected data, like NYT, where people pay to get access to content?

Taking a step back - and related to my first comment - I find it helpful to think of AI systems relative to humans. So, AIs training on data owned by others is OK to me since that’s what humans do.

For data behind a paywall, if you agree with me above, assuming the AI company pays for it, then that’s fine if they train on that too. If I read an NYT article and told you about it, no one would care.

Yeah, but that's a thing they don't pay for it. And even if you'll read an article and tell other people, we cannot "regenerate" similar content in a journalist style (or it's takes too much effort), but ai can make it immediately.

scalping the web and data without consent is unethical imo. there is so much nuance here but i don’t trust any big corp with this. the model you brought up with NYT is an idea. but we’ve dug ourselves a deep whole, having opted into these “free” spaces for decades now...

agree that it's complicated, but do you feel big corps will be willing to pay royalties to artists? In my mind, if they need this artist's type of work, why not just hire him? It only works if they need a massive amount of similar content

I’ve torrented too many gigabytes of creative works for zilch in the name of “culture wanting to be free”, that I simply don’t feel like I have a leg to stand on criticizing this

As a person from ukraine 🤝🏻. But now being in us and having more resources makes me think differently