Image makers are stuck in a pattern.
Illustration: The Atlantic. Source: Getty
August 16, 2024, 6:05 PM ET
This is Atlantic Intelligence, a newsletter where our writers help you make better sense of artificial intelligence and the new machine age. Sign up here.
At this point, AI art is about as noteworthy as an email offering you 10 percent off a new pair of jeans. On the one hand, it's miraculous that a computer program can synthesize an image based on a text prompt. On the other, these images are so commonplace that they've become a new kind of digital junk, polluting social media feeds and other online spaces without providing any particular benefit to users.
But it's not just a matter of volume that makes spam so energetic: These images tend to look pretty similar, too. As my colleague Caroline Mims Nice writes in a new article for The Atlantic, “Two years into the generative AI boom, these programs' work may seem technically advanced, but it's stuck in a peculiar aesthetic sensibility.” By default, these models tend to produce images with bright, saturated colors, beautiful, almost cartoonish figures, and dramatic lighting. Caroline spoke with experts, who gave her four theories as to why.
Ultimately, her report suggests that while tech companies are competing to offer more attractive image generators, in reality the products aren't all that different in the end. The situation is more like Pepsi vs. Coca-Cola than Toyota vs. Mercedes. People will probably just use the image generator that's most convenient for them. That may be why companies like X, Google, and Apple are eager to incorporate these models into their existing platforms. Image generators are no longer magic, but a feature to check.
Illustration: The Atlantic. Source: Getty.
Why does AI art look the way it does?
Caroline Mims Nice
This week, X launched an AI image generator to allow paid subscribers of Elon Musk's social platform to create their own art. Unsurprisingly, some users quickly created images of Donald Trump flying into the World Trade Center, Mickey Mouse holding an assault rifle, and enjoying a cigarette and a beer on the beach. Some of the images created with the tool are deeply unsettling, others strange and kind of funny. They depict completely different scenarios and characters. But somehow, they all feel similar, bearing the unmistakable hallmarks of the AI art that has emerged in recent years thanks to products like Midjourney and DALL-E.
Read the full article.
What to read next
Trump Discovers a New Benghazi: Earlier this week, Donald Trump falsely claimed that Kamala Harris had “AI-enhanced” photos of crowds at her rallies—in other words, that the images had been doctored or outright fabricated to inflate the numbers of Harris cheering. As Matthew Kirschenbaum writes in The Atlantic, Trump may have used the term less to do with the technology itself, but more to give his supporters something to post, “as a way of giving them permission to follow Trump's example by filling in a text box on their screen.”
P.S.
Maybe AI art works best with an audience of one. “Approaching a generative image creator to produce a desired result could push its potential in exactly the opposite direction,” Ian Bogost wrote in The Atlantic last year. “AI can give form outside of your mind, quickly and cheaply. Any thought can be visually output in seconds. The result is not an image to be used as a medium, but an idea recorded in a photograph.”
Damon
