It looks like OpenAI trained Sora on game content

The number of layers is irrelevant. It’s transformative. The author had to work hard just to get it to make non exact videos of similar videos.<p>Where it’s going to get interesting is in terms of service where companies start using AI to copy SaaS products by having an AI login and copy competitors features. This breaks not just copyright law but is a breach of contract.

I don&#x27;t think there&#x27;s anything new here. It&#x27;s the same <i>unanswered</i> question of can you train AI on copyrighted material.<p>The fact that it&#x27;s video instead of photos or text doesn&#x27;t seem like it should make any difference.<p>We need to wait for the legal system to decide.

I think it&#x27;s clearly wrong to use generative AI, or any tool, to produce what are basically copies of a work, and then profit off that as if you were the original creator.<p>I think it&#x27;s just as clearly okay to use content that other people produce as part of learning to create your own stuff.<p>Generative AI can be used for copyright infringement - just like any creator technology could be.

Is AI generated content any different from a human reading a bunch of books or watching movies and being inspired by them to create something similar but not identical?

lol it&#x27;s going to be much harder to launder the training data for videos than for text and images. not that it matters legally but like who&#x27;s going to be that impressed by an app that so obviously produces facsimiles