D&D General DMs Guild and DriveThruRPG ban AI written works, requires labels for AI generated art

[Reynard]

A few thoughts -
Agreed on the practicality of banning AI Art. The risk is too large at this stage.

Even if Art was copyrightable, there still wouldn't be money because someone can always get it to make new art for free or at least very cheap.

And while AI Art itself isn't coprightable, using it in conjunction with a copyrightable work could be a value add for that work. If you make 2x as many sells by selling a book with ai art as a book with no art then you would potentially be willing to pay something for the ai art.

None of that explaisn why OBS is banning AI written works outright and not AI art.

 

[Charlaquin]

I think AI art is amazing, and would happily use it, and in many cases, in my opinion, I doubt a human could duplicate what it does in anything like the time it can churn it out.

Not the least, Sci-Fi Pantheon on DMS Guild uses a lot of AI art - which in my mind is an invaluable tool for those wanting to minimise costs.

Theft certainly does minimize costs, yes.

 

[Stalker0]

Fair use, the crux of the issue, tends to be complex to adjudicate. For it to be a thing, it needs to be checked on many criteria. Being used for education is one such criteria. (And it is education, not just "teaching". One could argue education has having a human destinatary as a necessary condition) However, there are other conditions. The most important being that it is transformative and it doesn't replace the original. Using a given artist's work to train an AI, not only allows the AI to draw, but also allows the AI to replace the artist, not just on the generic, but on the specific. An AI trained on a given artist, can perfectly mimic that artist's style, making the artist themself superfluous. It will have to reach the courts, but from existing precedent, it seems artists have a case with existing laws. All of them, not just artists putting their work behind paywalls.

Thank you for the response!

On education, it will be an interesting question of whether AI learning would be considered analogous to human learning, and therefore if that clause would apply.

The transformative clause is tricky, because ultimately the AI is still creating original works. If I put in a prompt to an AI that says give me the mona lisa....I'm never getting the mona lisa. I'll get something close, but it will never be a copy, and pretty much always recognizable as something else. The counter here is that an artist of sufficient skill can replicate another artist's style, often to the point of being "unrecognizable" as different except by art experts. Art AIs in many ways are doing the same thing, but less exact than what a skilled human artist can do.

The last interesting question will be.... is the AI creator liable or the content maker? If I buy a john deere tractor and run over someone....john deere isn't at fault (most of the time, lawyers will of course try to sue anyone they can).

So if an AI is capable of replicating another artist's style...well is the issue that the tool exists or that someone used it maliciously? Just because a tool can potentially do a thing doesn't mean the tool maker is liable....its the person that used the tool to create the content that ultimately violated the artist's rights. That is one argument we may see made as well.

 

[Rogerd1]

Theft certainly does minimize costs, yes.

Who says it is theft.

 

[Charlaquin]

Who says it is theft.

It’s a predictive software that works by filling in the blanks with information gathered from the work of human artists, the majority of which was gathered without notice, let alone permission. That’s art theft.

 

[Rogerd1]

It’s a predictive software that works by filling in the blanks with information gathered from the work of human artists, the majority of which was gathered without notice, let alone permission. That’s art theft.

Proof?

 

[Charlaquin]

Proof?

Proof of what? How these algorithms work? Any rudimentary amount of research will confirm as much.

 

[Rogerd1]

Proof of what? How these algorithms work? Any rudimentary amount of research will confirm as much.

I am under no obligation to do so.
And quite frankly, I like AI art, so I am glad it is here as it does it better, and quicker than a human artist.

 

[Voadam]

The transformative clause is tricky, because ultimately the AI is still creating original works. If I put in a prompt to an AI that says give me the mona lisa....I'm never getting the mona lisa. I'll get something close, but it will never be a copy, and pretty much always recognizable as something else. The counter here is that an artist of sufficient skill can replicate another artist's style, often to the point of being "unrecognizable" as different except by art experts. Art AIs in many ways are doing the same thing, but less exact than what a skilled human artist can do.

It is not as straightforward as not being an exact copy gets you out of copyright issues. There is also the issue of a copyright holder's copyright over derivative works of copyrighted material. This is impacted again by the concept of fair use.

Writing a sequel to a copyrighted novel, for example, has copyright implications even though it is not an exact copy.

 

[Reynard]

It’s a predictive software that works by filling in the blanks with information gathered from the work of human artists, the majority of which was gathered without notice, let alone permission. That’s art theft.

That is not the definition of theft. At worst, it is a copyright violation. (That doesn't make it right, but I feel like we should be clear when labeling things.)

Also, if it WAS trained only on works in the public domain, it wouldn't even be a copyright violation.