Is art made by a machine art?

Maybe the promise of AI’s future creativity lies in its consistent production of errors that engineers can’t seem to eliminate. These “hallucinations,” where AI invents fake names and quotes, could be an emergent property—early signs of daydreaming in a young AI?

Like a human child, an AI in its infancy might get distracted and make new associations. Eliminating these “errors” could block its capacity for play, where “hallucinations” might actually be new hypotheses, waiting to be explored.

If human intelligence evolved with a degree of randomness that allows for exploration, then error in AI systems could be just as crucial for artificial creativity. Ideally, we won’t stifle AI’s early creative impulses, the same way standardized education curtails the imagination of our youth.

AI will keep perfecting its ability to execute our visions, but its engineered imperfection will be key in maintaining our fascination with the random, accidental innovations nature has conditioned us to favor.

But humans are drawn to humans. Will we be as captivated by art made by a machine? Would we still love a Rothko if we didn’t know it was a Rothko—if we thought it was a smudge left by a machine?

And what about AGI’s own interest in art? Will AI seek to see itself represented? What will be AI’s existential drivers?