Cover Photo by Solen Feyissa on Unsplash
A decade ago, I saw a Tweet that made me rethink my career in creative technology.
I was working for a company called PublishThis, a content curation app to help media companies scale their work. Among the efficiencies we introduced was the “Digest Post”, which curated existing online articles into a blog post, with the Digest Poster adding their own commentary. We even worked with an SEO firm to confirm that these patched-together content pieces would rank as well as a piece of truly original content.
The day we signed Fandango, we were so excited! We had plenty of excellent companies already but we loved having an entertainment industry leader like them. Then I saw this tweet (paraphrased because “X” doesn’t keep history too well):
“Laid off today. Thanks, PublishThis.”
My heart cratered. The platform that I worked on and genuinely believed in was putting creative folks out of a job. I spent my long L.A. commute home that night thinking that I was a Turncoat Creative, selling out my fellow creatives for my love of technology solutions that were making things more efficient and streamlined.
Fair Use to the Rescue
At PT, we didn’t call it stealing because we insisted this content format was just “Fair Use.” In case you aren’t familiar with the term, it’s a legal one that helps US courts determine if someone can use another person’s work without permission. There are a variety of factors to consider when citing Fair Use:
1. Purpose and Character: Are you doing something non-commercial (for personal use, say), as a parody (hello, Weird Al), or for educational purposes (teachers need all the support they can get). If so, there’s a reasonable chance that you’re talking about Fair Use.
2. Nature of the Copyrighted Work: If the work you are quoting, referencing, or repurposing a piece of for your work is from a factual work (real news, not conspiracy theories), you have got a better shot at calling it Fair Use than if you are using something that is a creative work (movies, novels, artwork, music). One of the factors here is whether you are replacing the value of the original work with your own. Will people read your book instead of the one you cite, you know, rather a lot?
3. Amount and Substantiality of the Use: Did you just use a snippet of that Grant Lee Buffalo song in the intro for your presentation? Just a quote from that Maya Angelou book? A smidgen of that artwork in a collage filled with other things? You might be okay by Fair Use’s standards — although if you used the true ‘heart’ of the work, maybe not.
4. Effect on the Market Value of the Original Work: Does your use of the work change the value of the original work? Will your work compete with it or possibly prevent the creator of the original’s ability to earn money? If so, Fair Use isn’t likely to be here to save you.
At PT, we were confident that even if we got a cease-and-desist letter from someone, we’d be okay.
Not Stealing — We’re Helping!
Even back when we said this to PublishThis customers, many of them were not so sure. They worried they’d get sued if they did these Digest Posts. We didn’t think so because we were helping the original creators in a few ways:
- Direct Traffic: We included a link to the curated content to drive traffic to the original work.
- Quality Links: We were working with reputable publishers, creating Google-valuable links to them from another trusted site.
- Syndicated Content: We only used snippets the original creator put into RSS feeds to promote their work online through syndicated outlets and feed readers.
We felt that Fair Use was going to protect us from anyone that got litigious. But even then, we weren’t 100% sure.
So, how does an AI tool that does none of those things for the original creators hide behind the same shaky and inconsistent tenet we did?
Unfair Use
Some AI companies are still shamelessly trying to unfairly use Fair Use because of how murky the concept can be. The ugliest bugaboo in the Fair Use principles is the term “transformative.” Per the rules, if you are truly adding something new and not creating a direct substitute for the original work, Fair Use may save the day for the “re-user.” Unfortunately for creators, this definition is fuzzier than a Maine Coon.
AI tools that claim Fair Use when using — sorry — stealing your content for training their models are quick to point out that GPT means “Generative Pre-Trained Transformer.” See, they say, we’re like those car-robot toys that Michael Bay makes movies about — we’re transforming all that old stuff into new stuff! Fair Use FTW?
Yeah, no.
Unlike how we used it at PublishThis, content that is used for training by a LLM gets no respect. They aren’t enshrining the work inside of another piece of content created by another human who is speaking about their work with an original voice. I mean, at least Google used to send other sites traffic before their dubious AI Overviews.
No, instead, LLMs grind your content — be it articles, novels, music, artwork, short-form videos, or memoirs — into mystery meat content that an algorithm has pieced together from what it has gained through statistical correlation learning across vast amounts of possibly incorrect or ironic, or even just outdated data.
Unlike people, who learn in a variety of ways, LLMs are limited in their intake and learn models, despite the brilliance of the neural network structure. As a creative writer, if I used ChatGPT or Claude to write something, I could look at the output with the critical eye of someone who has been writing professionally for thirty years. I will find flaws, the cringy use of some of the telltale AI terms, and end up rewriting everything because the work of the AI is not up to snuff by my standards.
But what about someone who just needs a piece of content done? What if they look at these articles, stories, videos, pieces of art, anything that AIs can spit out and say, “Looks good, let’s go!” We see that now and we’ll see it all the more in the future as AI models get better and better. People can be more efficient with these tools. No amount of Luddism will convince some people whose lives get simpler with the use of AI to abandon them. Well, maybe when BigAI gets good enough to not need them anymore they might feel differently.
“Fair” Use Means Paying Creators
Let’s talk about what is actually fair. That involves paying creators when you extract value from their copyrighted work. No amount of ‘transformation’ removes the fact that newly-minted AI music, books, and artwork could not exist without the very much human, hard work put into the content that is being analyzed for its construction.
Copyright exists to encourage more creative output from people. Before this law, it was so easy to steal the work of other people and it was done all the time. From copying the popular fiction of Charles Dickens to news stories being copied word-for-word into other publications, patents being ripped off, music being performed with no compensation to the composer — the list goes on and on — creative work was being stolen every day. There was simply no meaningful way for a creator to protect their work and make a good living in true invention. That’s why we need to make sure this AI kleptocracy ends…and NOW.
OpenAI Signals a Change of Heart(?)
While this may be generous, industry leader OpenAI has actually helped validate this point recently. Although they are likely motivated by the myriad lawsuits they are facing from the likes of the New York Times and The Center for Investigative Reporting, OpenAI have started forging pacts to train their models on licensed content by paying money to major publishing entities like The Atlantic, Vox Media, Time, and Newscorp. YouTube is following suit, now trying to work more closely with the recording industry to license work for AI training as well.
While it is encouraging to see that most of these pacts are multi-year deals rather than one-time payments, one hopes that this isn’t just about getting access to the new output from the media company and it takes into account the continued use of past work to train the ChatGPT models…and the value of the work in the model.
That’s the only way licensing of creative content for AI makes sense. As long as the model is in use, the creators of the training material MUST be paid for their work. To do otherwise is just a time-delayed theft of someone else’s Intellectual Property. If the value is still being used, then the creator should still be paid.
Imagine a publisher just ignoring agreements and ceasing to pay royalties on a book or a studio that owns an old television program just stopping with residuals while still airing the classic show. Media companies have had this right for a long while and it’s time to make sure technology companies adopt the same paradigm.
This needs to be firmly enforced by media companies, especially since OpenAI’s buddies at Microsoft still seem think that content on the Internet is still free for the taking.
Inclusive Capitalism or Universal Basic Income?
Earlier this summer, AI elder statesman Geoffrey Hinton noted at the AI for Good conference in Geneva, Switzerland, a point that has become a clear and common belief among the AI-erati: The coming rise of AI will not be just like all other major technological changes in history. In the past, the rise of the Industrial Revolutions, the Information Age, the Digital Revolution — all of these massive developments that changed the job market because of new technologies ended up creating different jobs, often more than they took away.
Not this time, said Dr. Hinton. This is why he, OpenAI CEO Sam Altman, and many other technology leaders have called for more study of Universal Basic Income. They and most futurists see that we need to figure out now how we will steady a global society that will need less human work.
While mostly successful experiments in UBI have been happening all over the world, this concept gets in the craw of many people who cannot abide the idea that there is value in a society where everyone can eat, has shelter, and healthcare even if that means not everyone does the same amount of work or contributes just as much money.
But this is the reality we are facing. Perhaps those who simply don’t believe in the comity of humankind can get behind the idea of a more inclusive type of capitalism. What if, instead of allowing billion and trillion dollar companies to steal the value of work forged by creative people, we PAY the people for their work when it is reused to transform society and, you know, destroy hundreds of millions of jobs?
What if all these amazing artists, musicians, writers, crafters, inventors, and just people who love to make things to enrich humanity actually got paid recurring revenue for their contributions to humanity and their part in enabling technology companies to make society more automated?
Maybe if BigAI actually pays for more human-created content, they won’t ‘run out of Internet’ to train on again. Maybe they won’t have to create ‘synthetic content’ that risks worsening their AI tools with lower-quality, less authentic content. Maybe they’ll see how this approach actually seeds creativity for the future in a way that’s good for their business while also being great for the human race.
Perhaps seeing that compensating the people who empowered their AI models is actually good for business will inspire more tech titans to act like the best billionaire around, MacKenzie Scott, and spread the wealth. She keeps giving away her billions to worthwhile organizations, charities, and public educational institutions that don’t have their own billion-dollar endowments. She’s smart enough to realize that one of the best investments for both economic growth and society as a whole is to invest in people.
Yet, creatives should not have to rely on the generosity of billionaires because this isn’t about charity or donations. This is about paying artists and makers for using their work. It’s time for creatives to work directly with AI tools and LLMs to grow AI in an ethical fashion that rewards all parties involved.
Sure, as a creative and a technologist, I want to reward both sides. With this partnership, we can transform our world into a place where people are empowered rather than exploited and made obsolete by AI technologies.
Join the Movement — Sign Up at Credtent.org
Credtent.org is a Public Benefit Corporation devoted to empowering creators to control their content in the Age of AI. We are also devoted to helping AI tools make better decisions about using credible and licensed content to train their models.
AI training without compensation is simply Unfair Use and we should call it out as such whenever we encounter it. That is why we create Credtent, a Public Benefit Corporation devoted to empowering creators to manage their content in the age of AI. Now, not just the big media companies can make these deals. Credtent allows every creator and maker to license to or EXCLUDE their work from AI companies.
Register your work with Credtent and let us know:
- Do you want your work excluded from AI training? Great — we’ll handle the opt-out process for you at every AI tool/LLM we can find, all for a reasonable annual fee of $20. Sign up before the end of 2024 and you can lock in our introductory pricing until 2034.
- Do you want to earn money from your content being used for AI training? We will license your content at a fair-market rate to reputable AI companies who will pay an annual fee for use of your work.
- Do you want more control over your content’s pricing? We’ve got you covered. Just sign up for our Premium License and you can set your content’s pricing values yourself. More control, more passive income.
Before more AI tools get a hold of your work without your consent, sign up now and start the process of taking back control of your content. Credtent is on your side.
As you can see from above, this article was written without any help from AI tools. It’s all human-created and, occasionally, a bit gonzo. Feel free to use the Credtent Content Origin logos without cost at credtent.org/blog.
