On AI and Human Creativity

Date
30. November 2023
Time4 min read

As an artist and technologist, I find this letter and its reading quite stirring, more than a little conflicting, and not entirely wrong but not entirely right either.

The fact is, AI and LLMs are here to stay and are only going to get better and faster and more versatile in what they do. Working _with_ them is our only realistic option. There's no fight we can win in the end for the human soul in the arts, whatever that means.

However, what we find meaningful in art is somewhat intrinsically tied to the human labour that produced it, that separates the virtuoso from the novice. That imbues it with a perceived value, or rareness. AI art can be quite beautiful and even indistinguishable from the artists it imitates, but, aside from the effort of creating the initial algorithm and the (largely stolen) training data it fed upon, that same effort isn't there. Writing a prompt is not participating in art in nearly the same way, and does cheapen it.

That said, if you want to produce art in today's day, you would be unwise to not work with and explore the possibilities in using AI as one of your tools, and thinking of it as such. If you don't, others are and will, and they'll be the ones to discover new boundaries where human intuition and AI can combine to produce something truly new, and you won't.

I think the place where we'll land is a hybrid where live performance likely has the largest buffer, and other arts will quickly incorporate AI in invisible and pervasive ways. Humans are drawn to other humans doing impressive things, and I believe there will always be an appreciation for the years of practice and development that go into a great piano performance or dance, a clever improv joke, or a great rap lyric.

And fortunately, if you've ever tried to get ChatGPT to write rap lyrics, you'll discover it's pretty much crap at it (like most of the drivel that comes out of it still). LLMs are not the proverbial 100 Monkeys Typing yet, but they already have the seeds of Shakespeare's sonnets tucked away in their training sets to slowly be teased out under the guise of actual creative work.

I'm reminded of the Star Trek TNG episode "Data's Concert" and how Data works to develop more human-like nuance. He's a much different vision of what an AI would be capable of and mimesis in his scenario is a combination of sensory and motor functions and not simply a next token predictor like we have today. He's an allegory that speaks to a future actually sentient AI attempting to understand what it is to be alive. What we have today are useful parlour tricks by comparison.

In the end, the fool will be fooled by the trick just the same, and we're all fools for a clever trick in our own ways. And that may well be Good Enough for most purposes.