Today’s Tech Thoughts is a simple enough endeavor in theory, but feels like it has layers when it comes to the various wrinkles that may occur in practice.

I wanted to read up on and discuss the issues with generative artificial intelligence when it came to generating images that would, presumably, infringe on the copyrights of its holders.

The issue gets murkier, however, when I consider how much large companies hold sway over protecting the creations they presumably “own.” This is an important distinction to consider when it comes to allowing companies versus the creators of said visually identifiable intellectual property to benefit from their creations.

The problem with ‘visual plagiarism’

In this 2024 opinion piece by Gary Marcus and Reid Southen on the problems with generative AI and how it engages in “visual plagiarism” of copyrighted works, the authors noted how “it was in fact easy to generate many plagiaristic outputs, with brief prompts related to commercial films.”

In their piece, they used Midjourney V6 at the time for the purpose of testing their ideas. They also cited a related work on visual images on Stable Diffusion that they said “converged on similar conclusions, albeit using a more complex, automated adversarial technique.”

Basically, image-generating AI models can be made to inadvertently produce plagiaristic outputs based on copyrighted material, even if the prompts are indirect, or do not mention a specific brand or property. For instance, the prompt “popular 90’s animated cartoon with yellow skin” yielded Simpsons-like images, while “black armor with light sword, movie screencap” yielded Darth Vader-likes.

Must Read

Getty argues its landmark UK copyright case does not threaten AI

What do these studies, thus, point to?

One the one hand, some generative AI systems may be cribbing off copyrighted work. On the other, some people may be using generative AI to circumvent a lack of talent at creating art, and getting seemingly plagiarized results without meaning to.

As Marcus and Southen explain, “These results provide powerful evidence that Midjourney has trained on copyrighted materials, and establish that at least some generative AI systems may produce plagiaristic outputs, even when not directly asked to do so, potentially exposing users to copyright infringement claims,.”

Must Read

[Tech Thoughts] My issues with generative AI and AI-made images have a social dimension

Companies versus creators

This leads to one other wrinkle that deserves discussion: Companies that own copyrights to specific visuals versus the people who actually made them.

When I think of visual images in this case, my go-to idea is of Marvel characters, which Disney now owns, so let’s start there.

Disney (as well as Universal) is actually fighting Midjourney over copyright infringement. According to a Reuters report, the companies claim Midjourney pirated its libraries, “making and distributing without permission ‘innumerable’ copies of characters such as Darth Vader from Star Wars, Elsa from Frozen, and the Minions from Despicable Me.”

The Motion Picture Association (MPA) supported the suit and called for AI companies to respect copyright.

Said MPA chairman Charles Rivkin, “A balanced approach to AI that both protects intellectual property and embraces responsible, human-centered innovation is critical for maintaining America’s global leadership in creative industries.”

It’s all well and good that companies get to uphold their copyrights, but do artists — especially the ones who created the characters that made Marvel and Disney brand names — get any of the money from such legal proceedings?

According to the US Copyright Office, it seems that — at least in the comics world where Disney and Marvel operate— there are work-for-hire situations, where the publisher generally owns the rights. Then there are creator-owned endeavors, in which a project is brought forward to a publisher, and the creator signs off publishing rights but keeps the intellectual property rights.

One site, Comics for Beginners, mentioned that if a work-for-hire (whether on retainer or some other arrangement) makes something new for a comics publisher, then the publisher will likely keep — and fervently defend — maintaining their new IP as theirs through legal means.

That seems unfair.

What would be adequate versus what would feel just

In a more just world, we wouldn’t need to worry about a generative AI or corporations and people without scruples, but we are in the real world, so we adjust.

What might be adequate would be corporations getting compensated for breaches of copyright by generative AI companies, and then creators get properly compensated as well.

What would be just, however, would be creators maintaining the rights to their creative work, and having financial and moral support from their publishers to sue generative AI companies for breaches of copyright.

Of course, with rich companies — whether they’re in the entertainment industry or the generative AI industry — holding the money that keeps things rolling, it seems the worst will normally come to pass, and it’s the creative minds who make art who’ll suffer in the long run. – Rappler.com