This site is supported by our readers – content may include affiliate links.
At this point, if you have read my past articles on AI, you probably think I’m very bullish on generative Artificial Intelligence, but let’s separate principle from practice. I am a fan of the technology, but I have vast reservations about implementations as there are two thorny issues here, one moral and one ethical.
The moral question is inescapable and as old as Photoshop – what will people do with this? Just as the explosion of actresses’ faces copied and pasted onto nude models accompanied the growth of bitmap editors, so we are now seeing original images created to look like popular figures having sex. I use this example as it feels particularly invasive, but it is fundamentally the same issue as any form of disinformation. Indeed, North Korea, after getting caught photoshopping a “successful” missile launch, might welcome an AI that’s even more detail-oriented than their human cattle. Editors. I mean, editors.
And it is going to happen. Whether it is wanting to see an actress spit roasted or spreading disinformation about a political figure, bad actors will always exist and they will use whatever resources are available to them. Long term, the answer to how we address this will be the same that Fox Mulder came to – trust no one and bookmark Snopes. Okay, Snopes probably came around later [yep, two years later – 1993/1995] but he would have.
The Question of Legality
The ethics question comes paired with the question about legality and is all about training. No, I’m not talking about a gym montage with Eye of the Tiger playing. I’m talking about all the data fed into a generative AI’s database so it can find those patterns and pump back out an infinite supply of cat girls in garters, occasionally with extra arms thrown in for free.
At least two class action lawsuits are floating around with authors accusing ChatGPT of training on their works with illegal copies downloaded from websites. For this purpose I’m going to bypass the questions about procurement – if they should own a copy and the issues of bulk downloading off the web – they are valid questions, but they distract from the core issue which is copyright. For this purpose I’m going to discuss it in the context of American copyright as it is what I’m most familiar with, so be warned some countries may vary in specifics.
When you ask ChatGPT to write a story in the style of Arthur Conan Doyle, it is using texts that are in the public domain. Regardless of the categorization of those works, they are legal to create. However, let’s imagine you create a work in the style of Stephen King. Well, he is still alive and there is no doubt his works are covered by copyright. The act itself of feeding his books into an AI engine is legal. What it creates might not be.
What Makes Work Derivative?
Only the copyright holder of the original work (e.g. a book) has the right to create derivative works of their intellectual property. Since generative AI creates new work based on an analysis of original pieces it has been fed or hoovered up online, the result would not exist without the source works created by legions of human artists. But can these new AI-generated works pass the threshold to be considered derivative under copyright law? Certainly, if you cannot tell exactly which original source a new piece of work is based on, it becomes hard to argue a case of copyright infringement. Unsurprisingly, this is the stance the defendants of the class action suits have taken. They do not believe that their works can be considered derivative.
I should note that there are also exceptions to the rules regarding derivative work, such as when it is adequately transformed, and it will be interesting to see which tack the lawyers take here. These legal questions are very important, and there are many more to follow. What about AI-created work where you can’t trace the sources? If you can’t recognize something as derivative – is it?
These lawsuits attack the training of the systems, something that we in the creative fields should have been more aggressive about years ago, but the genie is out of the bottle now. This is the first real chance to legally parse the issues. My hope is that indirectly from the conversation, legislation will be written that establishes expectations about the rights of creators to have their content ingested and used in these systems. I expect friend-of-the-court filings to be copious and interesting reading. I suspect US courts will take very slow motions in this direction, concerned about undermining copyright but also concerned about not understanding the core issues. I also suspect some kind of measure will be given to protect creators, though I’m not sure what it will look like in intent or scope.
Does that mean the creation of badly written novels in the style of Stephen King will disappear? Of course not. Some bad actors will do it anyway, and host it in countries that don’t recognize copyright, but they will be cheaper, less useful and may move onto onion networks with other questionable services.
However, all of this avoids one really important topic that I will pick up next time for my final article on Artificial Intelligence for Writers: What about the created works themselves? It’s a bit complicated, so join me in four week’s time for that discussion.
ABOUT LETO ARMITAGE
Leto Armitage was born in America under a set of circumstances that prophesied that he would one day unite the lost tribes and return the Ever Summer. Somewhere around twelve, he realized he had been left unsupervised and binged too many Arthurian movies in his formative years and that he was just another kid who accidentally got an education while reading above his age level.
By the time he turned old enough to get a passport, he started finding excuses to travel determined to find out what culture, food and women there were to experience. After learning to grill in Oaxaca, do kinbaku in Japan, and being banned from several former Soviet block countries, he returned home to settle down and see what damage he could do locally.
After working jobs including being a short order cook, bodyguarding strippers and professionally doing reader’s advisory for erotica he realized the most reasonable path forward was to become a writer. Today he lives with cats, dogs, and humans who seem to like him despite actually knowing him. He prefers to sit on his back deck, listening to the birds and Barry the Bumblebear bee, while he writes cozy, uplit romance and raunchy erotica.