Is AI Changing What We Consider Beautiful?
By BETH HARPAZ
Artificial intelligence edits our photos, creates highlight videos of our families and friends, and tracks every sunset and cocktail we post. But by steering us to certain images and content, is AI changing what we consider beautiful?
That’s one of the provocative questions raised by the book AI Aesthetics by Professor Lev Manovich (The Graduate Center, CUNY). Understanding how AI influences aesthetics is important “if you want to be culturally literate,” he writes, “and essential if you are a creator.”
Using data from the “cultural behavior of multitudes,” AI can “automatically beautify selfies and edit user photos to fit the norms of ‘good’ photography.” The EyeEm company used AI to learn styles of photo curators by analyzing 20 pictures the curators chose, then selected similar images for them to make further selections.
But what is lost when AI steers us to “choices preferred by the majority” — or even to styles we’ve preferred ourselves in the past? Is aesthetic standardization inevitable? If AI is programmed to make “good videos with proper composition” and certain facial expressions, will we lose the bloopers and footnotes that could prove as important in the long run as the glossy perfection of the moment?
And if AI replaces professional cultural creators, will innovators and outliers have more trouble breaking through? Cultural products that follow formulas like romance novels or music videos can be automated. Mathematical models can predict which images and videos will be most popular. But do we want to watch movies or live in communities designed by AI? In art, and in our lived experience, uniqueness is often what makes things worthwhile.
Manovich says there’s an alternative to systems that “classify artifacts, people, and behaviors” in ways that reinforce existing patterns. AI can also recognize what’s unusual. For example, Manovich directs the Cultural Analytics Lab, which analyzes visual media from manga to paintings to Instagram. The lab’s aim is to “map in detail” and “understand the full diversity” of cultural artifacts by focusing on “what is different … not only on what they share.” A system that allows “us to discover new categories for which we don’t have names and to see connections we were not previously aware of” could lead to an increase in “aesthetic variability” in cultural production, rather than a decrease.