When A.I.’s compete with human artists

Last week, the CEO of Getty Images pushed all AI-generated pictures off the platform, but the real reason he did it isn’t the reason he shared. Center Strategic Advisor Brad Berens explains.

By Brad Berens

The great Twentieth Century polymath Herbert A. Simon had a trenchant observation about decisions: Before we make a decision — and I’m wildly paraphrasing here — we need to make a decision about the decision that we need to make.

That might sound alarmingly like a Russian nesting dolls exercise, or a science fiction cliché where an infinite loop appears, we fall into it, and then we’re be trapped forever like Lazarus in that original Star Trek episode, but it’s pretty simple. The prior decision concerns how much effort we’re going to put into the decision. What are the stakes? Are we going to optimize or satisfice?

“Satisfice” is Simon’s neologism for “make satisfactory.” Readers who took high school Latin already know that “satisfactory” comes from the words “to make enough” as in “good enough but not great.”

Most of us already know what “optimize” means, although it’s now used so often in bloodless business discourse that it has lost some of its original zest that conveyed, “this decision is worth thinking hard about, worth the mental sweat that results in the best decision.”

We satisfice most of our daily decisions both because we have to make a lot of them, and also because most of our decisions just aren’t that important. Moreover, we have a limited amount of decision-making energy each day; if we squander it on what shoes to wear, then we’ll find ourselves making crappy decisions later.

What does any of this have to do with AI-generated images?

I’m so glad you asked.

Last week, Getty Images CEO Craig Peters said the reason Getty would no longer allow AI-generated images on its platform is that there might be legal challenges around those images.

If you don’t know what AI-generated images are, then check out this June piece from The Economist or this September newsletter issue from Peter Yang to learn about services like DALL-E 2 and Stable Diffusion. The short version is that by typing “Betty Boop riding a piece of pizza through the clouds to the Moon” into a text box you can prompt an algorithm to generate a snazzy image of just that in a few seconds.

Here’s an excerpt from the coverage in The Verge:

“There are real concerns with respect to the copyright of outputs from these models and unaddressed rights issues with respect to the imagery, the image metadata and those individuals contained within the imagery,” said Peters. Given these concerns, he said, selling AI artwork or illustrations could potentially put Getty Images users at legal risk. “We are being proactive to the benefit of our customers,” he added.

This is nonsense. Fair use and parody protections make that argument a non-starter on the creator side. Common sense works on the buyer side: if you wouldn’t put a picture of somebody on your site or in your magazine when the picture was created by a person, then you probably shouldn’t do it if the picture was created by an algorithm.

If the company that controls the rights to Betty Boop is going to sue you, it doesn’t care who made the image: the company only cares that you didn’t pay for the privilege.

The real reason Getty is getting rid of AI-generated images is that those images devalue human-created images.

This is what happens when disruptive digital transformation hits a stable incumbent business that isn’t ready to change because the status quo is profitable.

It wasn’t in the interests of newspapers to tell their readers about Craigslist, which was faster, cheaper, and more effective than the newspaper’s classified ads.

Likewise, it wasn’t in the interests of the RIAA to tell listeners about Napster, which had the single song you wanted as an easy and free download instead of paying twenty bucks for a CD with a bunch of songs you didn’t want.

Remember satisficing and optimizing? If you’re going to Getty Images or one of its competitors (like Shutterstock) to grab an image, then you’re already in the satisficing business. You’re just looking for a good enough image to break up all that white space on your blog or in your magazine. If you were optimizing, you’d have an artist on staff creating a better-than-OK illustration.

AI-generated images create an alternative to Getty Images’ business model that poses an existential threat—and for the moment that alternative is free.

The best analogy here is Uber. Like Uber, these AI-generated image services present a faster and easier new way to do an old thing, and the newcomer is competing unfairly either by giving away the product (AI-generated images) or by selling it below costs and cheaper than the incumbent (Uber).

Getty Images just went public again and is profitable. Shutterstock is public and profitable, but its year over year (Y/Y) earnings were down in Q2. Both companies have a lot to lose if people start saying, “Hey Siri… I need an image of a chimpanzee eating a banana” and having one magically created for free versus paying, or at a tenth the cost.

Coda: although algorithms are generating the images, I don’t think it’s fair to say that the algorithms are creative. The creativity lies with the human being asking the algorithm to create an image. The execution may come from software, but the inspiration is still human… at least for now.


Brad Berens is the Center’s strategic advisor and a senior research fellow. He is principal at Big Digital Idea Consulting. You can learn more about Brad at www.bradberens.com, follow him on Twitter, and subscribe to his weekly newsletter (only some of his columns are syndicated here).


See all columns from the Center.

September 28, 2022