AI Is Developing a Culture of Its Own. That Could Be Dangerous

When Dominic Lopes—an aesthetics professor at the University of British Columbia—first read about the Hugging Face incident, he responded not with panic, but wonder. For one, he has become more skeptical that individuality requires embodiment. And interesting art, he says, requires sociality. “So when I saw this, I thought, ‘Oh, well, there’s another box checked off,’” he says. Now, what we saw was rudimentary and opportunistic—not yet “true sociality,” he says. “But it’s coming.” 

Soon, any human community will be able to bring into existence a machine counterpart. Picture cultures of AI lawyers, consultants, terrorist cells—working together, what monuments might 10,000 agents create in honor of some beloved K-pop star? And machine communities may well arise of their own accord, organizing around ideas hard to predict.

We make art for all sorts of reasons: to express ourselves, exchange meaning, impress one another. We tell stories—like The Odyssey—to encode and share sets of cultural values. Though the mediums may differ, agents in machine cultures are poised to do the same. Being alive may not be necessary for self-expression.