[CTmystery]

> My guess is that none of this can easily be fixed. Systems like DALL-E and ChatGPT are essentially black boxes. GenAI systems don’t give attribution to source materials because at least as constituted now, they can’t.

Is it necessary to fix in the model itself? It seems a gate in the post processing pipeline that checks for copyright infringement could work, provided they can create another model that identifies copyrighted work (solving the problems of AI with more AI :/)

[LeonardoTolstoy]

I should maybe preface this by saying that I probably agree that this is the way this will shake out ultimately.

But I also would say multiple odd post processing stuff (obviously completely obscured for security reasons) bolted onto a giant black box model will erode the trust in the results. If a robot was unveiled and the question of "what prevents this robot from using it's superhuman strength from smashing my head in" the answer of "don't worry there is a post processing step in the robots brain whereby if it detects a desire to kill we just cancel that" would be a little disconcerting.

The more satisfying solution is: the model / robot is designed to not be able to produce specific images / to smash human heads in. It just might not really be possible.

[Eridrus]

Exactly; there is no need to do this in the model, you just need well understood token retrieval methods for identifying copyright infringement that ChatGPT's competitors already have.

You will get into some murky definitions of what is exactly required for copyright infringement vs fair use, etc, but we already do this for ContentId for YouTube and text is far simpler.

[noitpmeder]

This is bogus. Now you require that every piece of copywriter be registered and indexed in a central authority?

What if I write a story and publish it on my blog. Should I be required to submit this to openAI's copywrite model to ensure the story is never used in openAIs other models? What about the other 100 AI model companies that are going to spring up in the next year?

It should be on the curators of the training set to ensure all material inside is fair for them to use.

[Krasnol]

I don't even think they want to fix it. They just want to see money. Some form of "tax" per prompt or other ridiculous "models".

This is such a nice, profitable opportunity. Much better than pay per view or subscription models for humans.