Don’t ask if AI can make art — ask how AI can be art
Debates over AI’s artistic value have focused on its generative output. But so far, interactive systems have proved far more interesting.
You are viewing a single comment's thread from:
Debates over AI’s artistic value have focused on its generative output. But so far, interactive systems have proved far more interesting.
If you’re yearning for a fistfight with an artist, one simple phrase should do the trick: AI can do what you do.
The recent explosion of chatbots and text-to-image generators has prompted consternation from writers, illustrators, and musicians. AI tools like ChatGPT and DALL-E are extraordinary technical accomplishments, yet they seem increasingly purpose-built for producing bland content sludge. Artists fear both monetary loss and a devaluing of the creative process, and in a world where “AI” is coming to mean ubiquitous aesthetic pink slime, it’s not hard to see the source of the concern.
But even as their output tends to be disappointing, AI tools have become the internet’s favorite game — not because they often produce objectively great things but because people seem to love the process of producing and sharing them. Few things are more satisfying than tricking (or watching someone trick) a model into doing something naughty or incompetent: just look at the flurry of interest when xAI released an image generator that could make Disney characters behave badly or when ChatGPT persistently miscounted the letter “r” in “strawberry.” One of the first things people do with AI tools is mash together styles and ideas: Kermit the Frog as the Girl With a Pearl Earring, a Bible passage about removing a sandwich from a VCR, any movie scene directed by Michael Bay.
Despite artists’ concerns about being replaced by bad but cheap AI software, a lot of these words and images clearly weren’t made to avoid paying a writer or illustrator — or for commercial use at all. The back-and-forth of creating them is the point. And unlike promises that machines can replace painters or novelists, that back-and-forth offers a compelling vision of AI-based art.
rt by algorithm has an extensive history, from Oulipo literature of the 1960s to the procedural generation of video games like No Man’s Sky. In the age of generative AI, some people are creating interesting experiments or using tools to automate parts of the conventional artistic process. The platform Artbreeder, which predates most modern AI image generators, appealed directly to artists with intriguing tools for collaboration and fine-grained control. But so far, much of the AI-generated media that spreads online does so through sheer indifference or the novelty factor. It’s funny when a product like xAI’s Grok or Microsoft’s Bing spits out tasteless or family-unfriendly pictures, but only because it’s xAI or Microsoft — any half-decent artist can make Mickey Mouse smoke pot.
Article