Hacker News Reader

Comment by zarzavat | Hacker News Reader

This for me does not make sense as a copyright violation. It’s like saying that Adobe is in trouble because you drew something infringing in Photoshop. If you prompt the model with the intention of creating something infringing by mentioning the name of the characters and the work, and you get something infringing out, then it’s you who have infringed the copyright, not the maker of the tool.

View on HN

> If you prompt the model with the intention of creating something infringing by mentioning the name of the characters and the work, and you get something infringing out, then it’s you who have infringed the copyright, not the maker of the tool.

Yeah but that is not the case, they never mentioned Mario and Luigi, yet, that's what the output turned out to be.

The user didn't create it, the cloud-hosted machine owned by OpenAI, that charges for access, did.

When prompted with 'futuristic robot' and 'italian plumbers'.

So the argument is that if openAI had not used copyrighted and trademarked source material, this wouldn't be happening. It's not transformative as it's reproducing these copyrighted materials and trademarks verbatim.

That's how it makes sense.

This is an interesting idea. I assume that while the protected material would be obvious in some case, in many it would not. Would the tool have to be able to identify (and properly attribute) copyrighted material in its output?

No, it's more akin to if Photoshop had a 'Mario' stamp which when used would stamp a random piece of Mario artwork from the games. Do you think this would be in violation of copyright?

What about when you prompt the model without the intention of creating something infringing, and still get those same characters out in the result?

The post shows many examples where the prompt explicitly avoids any mentions of copyrighted materials but the generated results includes them regardless.

Did you even read the post?

But also, the argument of 'user responsibility' doesn't hold up on its own regardless (imo).

If I make and sell a toy printer that can only ever produce 3 pictures, and all of them contains copyright materials, would you really say that it's fine and responsibility falls under the end user? And I could sell that printer without any issues?

Did you read the article? It shows a lot of examples when no specific names are mentioned, or even with very generic prompts producing copyrighted material.

Oh c'mon, those prompts were not generic. Italian plumbers? How many other Italian plumbers do you know? What's the most popular soda in a red can?

"futuristic robot"?

Reply on news.ycombinator.com