Artificial Intelligence and Creativity
Artists and architects are beginning to embrace AI to do some of their “creative” work. Anecdotally, it seems to be a game of proposing off the wall combinations of things—“prompts”— to an app that then generates credible images of the wacky hybrid. Like most forms of artificial intelligence, the work of the app mimics the human mind, but with impressive capacity, beyond the human. For some, this seems like progress. To me, it seems illusory, and problematic. Let’s deal with the illusory part first.
In the late thirteenth century, Raymond Llull invented what he called his ars magna, or ars combinatoria, as a method or mechanism—effectively an algorithm—that could answer complex philosophical questions and also generate new information. It was presciently based on a mechanistic understanding of the mind which, if true, could be replicated in an actual mechanism or system, but free of human error or prejudice. Every form of artificial intelligence presumes two things: that the mind is rationally knowable and reducible to a mechanism or system, and that that mechanism or system can be recreated outside of a human person.
Llull’s combinatory art was remarkably advanced for its time, and yet inevitably reductive with respect to human reason (and imagination). AI is lightyears ahead of Llull, but still only mimics human imagination, albeit impressively. I’m interested in the Renaissance understanding of creativity, which they called invention (because, they would say, only God can create, which means to make something out of nothing). Invention means discovery, finding, even unearthing. It depends on the memory, a storehouse of ideas from which the imagination draws. That sounds like what AI is doing, but there is a difference.
The creative person in the Renaissance tradition built that storehouse of ideas through long, hard work, choosing what to learn and remember. As much as artists could have impressive visual memories, they couldn’t know everything. So, what they drew on was personal, selective, and theoretically coherent. It was not geared to generating absurdities, although Giorgio Vasari would describe novel inventions as bizzarrie.
AI is not selective, it is absolutely comprehensive—or, at least, that’s its ideal. It has an infinite memory and infinite capacity graphically to synthesize disparate things into new images. Which, at one level, means it is superior to human capacity. But it lacks two things.
One is the coherence that comes from deliberate personal selection—an individual mind with its own experiences, interests, and capabilities. The other is motivation: it requires a prompt, a list of often disparate attributes, to initiate its ars combinatoria. The AI generated mashups are funny, sometimes fascinating, exercises in combining images, which is a kind of caricature of the creative process.
It seems to me that artists who depend on AI don’t trust their own inventive capacities. But it also seems that they are surrendering the challenging fun of invention to the desire for appearing preternaturally creative. There has long been an argument that technology is only a tool. But when it is also the guiding hand there’s not much left for us. And I’m not convinced the results are inherently superior, even if they are impressive.
Read a postscript on Memory and Imagination