A.I. Books and Art
A.I. Books and Art
Thoughts on creativity, copyright, and artificial intelligence (A.I.)
Hey there, friend.
We've weathered a storm the last couple days that knocked out power to many people in the region—fortunately we weren't impacted other than taking a detour on the way to the doctor due to work being done on downed lines. Power had been restored to most people yesterday thanks to the hard work of all of those out there in nasty conditions putting things back together.
I've been thinking a lot this week about creativity, copyright, and the questionable practices around A.I. art. Much of it came about because the topics keep showing up in my feeds. It's an interesting topic for me as an information scientist, data analyst, librarian, writer of science fiction, artist, and coder. I've gone from reading about A.I. potential when I was a kid to seeing it widely deployed in all sorts of ways, obvious and non-obvious. I've read about all the jobs that A.I. could replace. The ways A.I. could improve our healthcare. The dangers of relying too much on A.I., and the growth of A.I. in areas formerly thought beyond the reach of the technology. Namely, artistic pursuits.
Machine learning coupled with computing power was always going to open these doors. Now we have A.I.s that write, compose, paint, and create videos. Whatever shortcomings they have right now, the technology is only going to improve. So let's speculate a bit, and I'd love to hear your thoughts in the comments.
Suppose an A.I. writes a novel in the style of a favorite author, say Stephen King. A book that is, by all accounts, on par with his best work. A hit with readers and critics. A major movie adaptation is in the works.
• Would you read the book?
• What if you did, loved it, and discovered afterward that it was written by an A.I.?
• How would it affect your view of the book if Stephen King wrote a one-page outline used by the A.I. to write the book?
• How would it affect your view of the book if someone working for the publisher wrote the outline used by the A.I.?
• What would you think if the publisher claimed all rights to the book, saying that the creation of the book was covered under existing contracts as a 'derivative work', and paid Stephen King nothing for the book?
• Suppose that neither Stephen King or the publisher were involved in creating the book, but some other company produced books 'in the style of' famous authors, using the text of the books by those authors to train the A.I., without permission from the author or publisher. They claim that the A.I. writes books for the 'fans of' famous authors—but that the books produced don't use any copyrighted text, characters, settings, etc., and don't violate any trademarks. Would you feel different about those books?
None of this is imaginary. These questions exist right now. A.I. models are being trained this way already. They produce blog posts, news articles, non-fiction and fiction content. It raises a number of issues for copyright laws around the world that haven't been answered yet. Is it a copyright violation to train these machine learning models using copyrighted material? Many of the A.I. art models have done exactly that by pulling copyrighted images from the web for training the models. The companies behind these works are well aware of the potential legal issues and have gone to lengths to protect themselves with hybrid non-profit / profit corporation structures. The non-profit corporation creates the data set for "research" to avoid copyright restrictions, but is funded by the associated for-profit corporation that then licenses the dataset to use with a commercial project.
In 1955, Clifford D. Simak wrote a story titled "So Bright the Vision." In it, Kemp Hart is a 'writer' longing for the latest computerized machine—a "yarner"—that he can't afford.
"Behind the glass the machines were shining-wonderful, the sort of merchandise that belonged on this svelte and perfumed street. One machine in the corner of the showroom was bigger and shinier than the others and had about it a rare glint of competence. It had a massive keyboard for the feeding in of data and it had a hundred slots or so for the working tapes and films. It had mood control calibrated more sensitively than any he had ever seen and in all probability a lot of other features that were not immediately apparent.
"With a machine such as that, Hart told himself, a man could become famous almost automatically and virtually overnight. He could write anything he wished and he would write it well and the doors of the most snooty of the publishers would stand open to him." — THE BIG FRONT YARD: AND OTHER STORIES.
The details of the machine aren't as important as what it represents in the story. Editor David W. Wixon describes 'Cliff's' intent.
"The machine, Cliff argued, was being allowed to set the norm for literature, which had the effect of setting up a pattern that would be deadly to good fiction."
It's easy enough to see how this scenario could play out. A major e-book retailer might offer A.I.-generated books, even custom A.I. books created by a reader's prompt. Those books could be available to subscribers, sold, and a percentage of each sale offered to the prompter. The e-reader devices can track all the data on how people read the book—feeding that back into the machine learning model to train it to generate more compelling books.
Successful prompters become the influencers of the platform. The platform owns the A.I. generated content, allowing fans of a world or setting to generate endless new prompts for the A.I. to use to generate books. As it grows and builds a bigger and bigger following, it's also using all of that data to build new A.I.s until there's no need for anyone to enter prompts because the machine learning is creating new titles designed specifically for the subscribers.
In that sort of system there's no need or desire from the platform to promote books written by humans.
The same sort of scenario is playing out for music, videos, art, etc., created by disregarding copyrights as often as possible. Not that it matters. If the A.I. is trained using content provided with or without consent doesn't change the end result that eventually it gets to the point where it doesn't need our creative input, only our consumption.
I don't believe in that scenario. Not because it can't happen. I've experimented with auto-narration to produce audio books read by A.I. and it's already something you could listen to and understand, if still a bit flat and not expressive (for now). That's a nice tool for people with visual impairment and those that simply want to listen. It expands the availability of the book to more people. I've seen amazing illustrations created by A.I. models. I've heard music composed by machine learning. I believe that everything in the scenario described can and likely will happen.
I think we have our favorite authors because we want to hear their stories. Not a story generated by an A.I. flawlessly executed 'in the style of' our favorite authors.
And, maybe I'm kidding myself, I think people will still want books, art, and music created by human writers, artists, and musicians. I think people want human expression. I think we have our favorite authors because we want to hear their stories. Not a story generated by an A.I. flawlessly executed 'in the style of' our favorite authors. It might be a terrific book, but fundamentally I think it makes a difference if it is an A.I. creation or not. I think there will be new markets, new ways to share human-created artistic works.
It'll take laws time to catch up with the technology. I don't think companies should be using copyrighted material to train their models without permission. I think works created by A.I. should be identified as such. Or if a writer or artist creates a work with the 'assistance' of an A.I. tool. People are already using these programs to create a 'first-draft' that they edit. I think we should identify these works co-created by A.I., though it's entirely possible that won't happen. Maybe A.I. will take over the 'ghost writer' market first without us knowing it is happening.
Comment. I'd love to hear what you think.
Thank you for reading. I hope you enjoyed this latest issue of READINARY. I'll be back soon with more stories, recommended books, and other thoughts.
Best wishes, always—Ryan
Drive-By Stories on READINARY
I learned recently that I could create new sections for publications on Substack. I've added one for my DRIVE-BY STORIES and moved all of my short short stories into that section. I hope you'll check them out.
Reading This Week
I'm still listening to the new Discworld audiobooks. I've moved on to Equal Rites now, sticking to publication order.