Disney sues Midjourney

[FrogReaver]

I reject your attempt to get me to oversimplify my answer.

Hold on. You said tools that enable large scale infringement are generally found complicit. I asked about copy machines and torrents both tools enabling large scale infringement and both not found complicit in it. Your next answer elaborated but didn’t fundamentally disagree with this assessment.

 

[Umbran]

I asked about copy machines and torrents both tools enabling large scale infringement and both not found complicit in it.

I countered with three high-profile examples of tools enabling large-scale infringement that WERE found complicit: Napster, Kazaa and The Pirate Bay.

 

[FrogReaver]

I countered with three high-profile examples of tools enabling large-scale infringement that WERE found complicit: Napster, Kazaa and The Pirate Bay.

Which at best shows some tools enabling infringement can be found complicit. Not that they generally will be.

 

[Ryujin]

Copy machines and torrents aren’t generally illegal.

Photocopiers don't inherently contain the infringed works, by design. Printing presses also aren't illegal, but they can be used to infringe upon copyright.

And it is in that point that Disney will probably win against Midjourney - it sounds like Disney is brining evidence of Midjourney directly reproducing Disney artwork, and effectively advertising their services based on that ability. That, if true, will scuttle the transformative defense.

I certainly hope so. And they could always use the "educational/research purposes" argument when they were initially training LLMs but, of course, businesses don't do all that work and then give up on possible profit.

 

[MoonSong]

You are correct, but once the Disneys and the appeal courts etc, have done their inevitable thing, you will no longer be correct. This ruling doesn't stand a chance of lasting a stiff breeze, let alone a Disney-powered hurricane.

Reading the actual court documents, this ruling isn't the great win the AI people wantsto think it is. The headline is this grandiose "LLM training is fair use!", but the actual ruling is way more nuanced and influenced by the American custom to disregard sweat-of-the-brow. Many of the books involved were about factual information -something that enjoy little precious protection in American copyright tradition-, and the judge considered the training didn't reduce demand for fiction writing, so it didn't hurt the fiction authors. Also, the ruling states that the licensing market for AI training isn't a real thing and no law has provided for it as a right. Now, if this ruling stands in appeal, it can actually help set a precedent for the Midjourney suit that would end up hurting Midjourney. Factual information is a minuscule part of artwork, and Midjourney actively competes with artists -and in this case Media conglomerates-. This case actively reduces demand for official artwork and hurts the very real licensing market.

Also, the ruling only was about the actual training, the case is still open for the large scale pirating that Anthropic engaged on in order to build their library (And this same ruling anticipates that the authors can still see some windfall out of it). Did Midjourney buy every single copy of these movies and art? Or did they sail for a bit in order to get these? Odds are they infringed the DRM Act in the process of scrapping the internet.

 

[Snarf Zagyg]

Reading the actual court documents, this ruling isn't the great win the AI people wantsto think it is. The headline is this grandiose "LLM training is fair use!", but the actual ruling is way more nuanced and influenced by the American custom to disregard sweat-of-the-brow. Many of the books involved were about factual information -something that enjoy little precious protection in American copyright tradition-, and the judge considered the training didn't reduce demand for fiction writing, so it didn't hurt the fiction authors. Also, the ruling states that the licensing market for AI training isn't a real thing and no law has provided for it as a right. Now, if this ruling stands in appeal, it can actually help set a precedent for the Midjourney suit that would end up hurting Midjourney. Factual information is a minuscule part of artwork, and Midjourney actively competes with artists -and in this case Media conglomerates-. This case actively reduces demand for official artwork and hurts the very real licensing market.

Also, the ruling only was about the actual training, the case is still open for the large scale pirating that Anthropic engaged on in order to build their library (And this same ruling anticipates that the authors can still see some windfall out of it). Did Midjourney buy every single copy of these movies and art? Or did they sail for a bit in order to get these? Odds are they infringed the DRM Act in the process of scrapping the internet.

I think a lot of people aren't reading the actual complaint in this case (see my earlier post).

The most recent ruling in the other case has nothing to do with Disney's complaint.

Let's use an analogy.

Imagine an AI is a human artist. The human artist can look at Disney images all they want!

But then imagine that the human artist draws a copy of a Disney piece of art. .... that output, that copy, is a copyright violation.

The Disney suit can use the "bad facts" of the data scraping, but it isn't about the data scraping. Instead, the "bad facts" are being used to bolster the claim that the output violates copyright. In other words, the complaint isn't that the bad act was scraping copyrighted material, the bad act is that the AI is producing images that are copyright violations. And it has examples of outputs from midjourney put side-by-side with Disney characters showing that.

So the other Court ruling has no direct applicability.

 

[Ryujin]

I think a lot of people aren't reading the actual complaint in this case (see my earlier post).

The most recent ruling in the other case has nothing to do with Disney's complaint.

Let's use an analogy.

Imagine an AI is a human artist. The human artist can look at Disney images all they want!

But then imagine that the human artist draws a copy of a Disney piece of art. .... that output, that copy, is a copyright violation.

The Disney suit can use the "bad facts" of the data scraping, but it isn't about the data scraping. Instead, the "bad facts" are being used to bolster the claim that the output violates copyright. In other words, the complaint isn't that the bad act was scraping copyrighted material, the bad act is that the AI is producing images that are copyright violations. And it has examples of outputs from midjourney put side-by-side with Disney characters showing that.

So the other Court ruling has no direct applicability.

And, if I remember correctly from what I've read elsewhere on that case, the judge even mentioned in his decision that the plaintiffs likely should have tried a different angle, perhaps like addressing the output as in the Midjourney case.

 

[MoonSong]

And, if I remember correctly from what I've read elsewhere on that case, the judge even mentioned in his decision that the plaintiffs likely should have tried a different angle, perhaps like addressing the output as in the Midjourney case.

Yes, basically providing a roadmap for future plaintiffs. It's almost as if this 'victory' -cause the case is still ongoing- will turn into ia Pyrrhic victory for AI.

 

[FrogReaver]

I think a lot of people aren't reading the actual complaint in this case (see my earlier post).

The most recent ruling in the other case has nothing to do with Disney's complaint.

Let's use an analogy.

Imagine an AI is a human artist. The human artist can look at Disney images all they want!

But then imagine that the human artist draws a copy of a Disney piece of art. .... that output, that copy, is a copyright violation.

The Disney suit can use the "bad facts" of the data scraping, but it isn't about the data scraping. Instead, the "bad facts" are being used to bolster the claim that the output violates copyright. In other words, the complaint isn't that the bad act was scraping copyrighted material, the bad act is that the AI is producing images that are copyright violations. And it has examples of outputs from midjourney put side-by-side with Disney characters showing that.

So the other Court ruling has no direct applicability.

The ai produces nothing without a prompt, much like a copier produces nothing without a human directing the copier to copy. We don’t blame xerox because a user directed the copier to output a copyrighted work. We blame the individual directing the machine to do that.

 

[Snarf Zagyg]

The ai produces nothing without a prompt, much like a copier produces nothing without a human directing the copier to copy. We don’t blame xerox because a user directed the copier to output a copyrighted work. We blame the individual directing the machine to do that.

Bruh, that's a stretch too far.

You can either say that the AI is generating the copyrighted image, in which case it is liable for the infringement, or you can say that it does in conjunction with another (like an artist under work for hire) ... in which case it's the person who provided the prompt AND the AI (the company).


I get that people like to spin legal fantasies, but the idea that, if an image is infringing, it just came from nowhere doesn't hold water.

Put more simply- are you making an argument based on any actual law, or based on what you wish to be true? There is a difference.