The End of the ‘Photoshop Fail’
In 2017, Rihanna posted a photo of herself on Instagram in which she appeared to have an extra thumb. It was, in retrospect, the thumb-shaped canary in the coal mine. Although far from the first celebrity “Photoshop fail,” it just so happened to predict the era of faux-finger drama we now live in: AI image generators are universally, horrifically bad at rendering human hands. Today, an extra finger is a telltale sign of digital manipulation.
Flaws aside, faking it has never been easier. Advances in generative AI mean that anyone can spin up a faux picture of the pope wearing a chic white puffer, no design skills required. New AI image creators such as Midjourney and Stable Diffusion use sophisticated technology to let users conjure entire worlds from just a few words. Instagram is rolling out AI-editing features; with a couple of taps, an everyday user can place their dog at the foot of the Eiffel Tower. We are living in the world Adobe Photoshop first teased 34 years ago—but it is no longer defined by the enterprise software.
Yet there is a reason Photoshop is the Kleenex of photo manipulation, the rare piece of software so powerful that it became a verb. To Google is to search, and to Photoshop is to sneakily slim a waistline, or to make it seem like something happened that didn’t. The software has been blamed for all kinds of societal ills, such as spreading misinformation and reinforcing unattainable body standards. Our entire modern cultural understanding of photographic trickery is closely entwined with the history of this application. It should be uniquely well suited for this moment.
[Read: AI is about to Photoshop your memories]
From the get-go, Photoshop had ties to unreality. The software was developed in the late 1980s by a pair of brothers, one of whom worked in Hollywood special effects. In early demonstrations, one of the co-creators, John Knoll, would pull up a picture of his then-girlfriend sitting topless in the sand in Bora Bora, looking out over perfect blue waters. He’d select and clone her, creating a second topless woman. When it formally launched in 1990 after being acquired by Adobe, Photoshop quickly differentiated itself from previous photo-editing software. As Walter Scheirer recounts in his book A History of Fake Things on the Internet, the key innovation was that it allowed creatives to make and share “plug-ins” to augment the software, such as Aldus Gallery Effects, a popular set that came with tools including “Splatter” and “Neon Glow.” These helped foster an online community around the software.
Photoshop grew alongside the rise of the consumer internet; people traded presets and add-ons on sites such as DeviantArt. Digital photography was becoming more and more popular, and a cheaper version of the software, Photoshop Elements, offered the average consumer the ability to tweak their digital photos. In the early 2000s, publications began using Photoshop as a verb; by 2008, usage was so widespread that Merriam-Webster had added the word to its dictionary. At this point, Photoshop was at the apex of its cultural power: It popped up in discussions of misinformation and cultural conversations about whether the technology had gone too far. The blog Photoshop Disasters documented hundreds of hiccups among celebrities and advertisements. “Indeed, in a world where so many images of the beautiful and famous are enhanced, ordinary people sometimes believe they need to prettify pictures of themselves just to keep pace,” Alex Williams wrote in The New York Times in 2008.
In the 2010s, smartphones and social media moved photo editing onto apps such as Instagram and later, Facetune. Many professionals continue to use Photoshop, and Adobe still makes a ton of money, but the application is less central in our culture: People have a lot of competing options on their smartphones. Adobe, meanwhile, has continued to innovate. Long before today’s generative AI, Photoshop introduced algorithmic functions that transformed the work of image editing: “Content-Aware” features, for example, react to what is depicted in an image, allowing a user to, say, seamlessly remove a person from a scenic forest view or add additional fluffy cumuli to a cloudy sky.
Photoshop was well positioned for the AI arms race, though it lagged behind image-generation tools such as DALL-E. By the time ChatGPT snapped the world to attention, Adobe was already moving toward putting smaller generative-AI-image features into its products. After that, the company seemed to pick up the pace: The next spring, it launched a full text-to-image model, Firefly, in beta form.
[Read: Kate Middleton and the end of shared reality]
When I asked Ashley Still, the senior vice president and general manager of the Creative Product Group at Adobe, about whether Photoshop is threatened in this new age of fake imagery, she told me “not at all,” and that the program has been consistently growing. Subscriptions, she said, were up 30 percent last year. “What’s happening is, more and more images are being created—more and more people are being creative. And that increases the market for us, too,” she explained. (When asked, representatives for the company did not provide revenue numbers for Photoshop specifically, instead directing me to Adobe’s overall earnings, which are at a record high.)
The company has tried to be “thoughtful about how generative AI can help creatives and make their work more productive,” Still told me. Indeed, the company’s previous algorithmic features saved people a lot of time. (“No one is proselytizing about the dangers of the content-aware tool, when it’s essentially the same technology,” Jordan Wannemacher, a freelance graphic designer, told me.) Whereas other image generators have been drawn into copyright lawsuits, Adobe advertises Firefly as safe for commercial use.
And there’s the challenge: Photoshop, the most professionalized fake-image program, now has to walk a middle path. It needs to offer generative-AI tools to stay cutting-edge, but it also cannot risk alienating its clientele by taking off the guardrails. Adobe is a technology company that serves artists, and in the art world, AI is deeply polarizing.
Still, people have always manipulated photos and created outright fakes, even long before the age of personal computers. “We’re sort of always yearning for authenticity, and it’s sort of also forever eluding us in some ways,” Michael Serazio, the author of The Authenticity Industries: Keeping It “Real” in Media, Culture, and Politics, told me. For the first generation of internet users, Photoshop was the primary battlefield for debates around fake imagery. Now AI is the focus of our anxieties. Surely there will be another thumb-related celebrity scandal in the future—but this time, it will almost certainly be the fault of a bot.