Can AI be Creative? A Conversation with Matthew Putman, CEO of Nanotronics

We have some pressing questions for Matthew Putman.

Machines in many ways already surpass human capabilities—but can they compete with us in the creation of music, art, and literature? Can an AI ever be truly objective? Can we? These are questions that Matthew Putman, the founder and CEO of New Lab member Nanotronics, has spent some time thinking about.

We recently sat down with Matthew in advance of our event “Every Bot is a Critic”—an evening on art, AI, and creativity—to discuss the impulses that are uniquely human.

Here, Matthew answers questions posed by New Lab’s Director of Communications, Molly Erman.

Can machines be creative, or is creativity inherently ours?
It is important that we hold onto our creativity, because it is what makes us human. To create AI, you must give it something to optimize for—my feeling is that we don’t even know what questions to ask an AI to inspire it to become truly creative. The way one pianist can play Chopin better than another: that’s something that can be determined by a critic, and debated by another with a different opinion. This kind of thinking cannot be understood by a machine—even humans don’t understand the nuanced reasons some art is perceived as “better” than others.

So, we’re talking about objectivity.
We are. And that is where this gets complicated. Art is subjective. I am making the claim we have reached the border of what machines are capable of: non-human creativity.

Are you coming into this panel discussion, which includes experts in the tech and art worlds, with a formed point of view? And is that POV that machines can’t be creative?
Yes, specifically, it is my opinion that machines cannot be objective, and therefore can’t be truly creative. But—here’s another thing about me that’s inherently human: I don’t optimize only for my opinion. I can adjust it based on arguments, and creative comments from others. It’s possible that over the course of this conversation, my thesis will completely break down, and I’ll come away from it with an evolved perspective.

But in the meantime…
As of now, I’m wondering how you would even set the utility function for an AI making art. Would a machine sample all the music in the world, and all the art in the world? Does it set out to make a sculpture better than Michelangelo’s David? Who can even say what’s “better?”

That’s not what art is, somehow. A human, though creative, can’t understand what art is. And we humans are incapable of writing the utility function that will get AI there. There are plenty of things we don’t know how to do.

On that note, let’s talk about something light: the limits of human ability.
People talk a lot about general intelligence versus narrow intelligence. I don’t think humans possess general intelligence. There are other creatures on earth that do things better than we can—a squirrel can remember where it buried an acorn better than us; a cheetah can run faster. We can’t fly. We don’t know how to do a lot of things—we are narrow. And AI, which is also narrow, reminds us of this.

However, AI will have narrow intelligence that we humans will be jealous of. The one thing that we will hold on to—or that we should want to hold on to—is this creative intelligence. And even if that’s the only thing we have, that’s a lot.

Would you say there’s an urgency, then, in human creativity?
I don’t think it’s urgent for us to learn to play piano, no. We started out as inherently creative—we don’t have to fight to be creative. We have to fight to have the freedom to be creative. And that freedom is something I think machines can help with.

For more from Matthew, follow him on Medium.