• Log In
  • Sign Up
    • You may have read last month that an AI created print sold for $432,500 at Christie's in NY, far exceeding the estimated $10,000 it was expected to fetch. Obvious Art in Paris produced the work using an adversarial generative network, an AI technique that is being used for creative experiments. I have to say that I didn't care for the piece much, though I recognized the accomplishment.

      Yesterday, I saw a Ken Weiner blog posting in Scientific American that showed more AI created stuff, some of which I quite liked. Actually, to be more accurate, the stuff shown was a collaboration between AI and human artists, though I'm not entirely clear on what the process involved.

      The article raises a number of questions that don't have easy answers. Some are philosophical: if it's created by an emotionless machine, can it really be called art? Some are legal: who is entitled to the copyright? The Edmond de Belamy portrait sold at Christie's had this whimsical signature:

      This is one of the formulas used in generating the piece. So does the work belong to the programmer? The program itself? To nobody? Then there are social questions: are human artists going to become as endangered as truck drivers?

      My personal view is that yes, a machine can create art. Any artistic creation should be judged on its own inherent merits, not on its history or the merits of its creator. I have absolutely no idea how we should deal with rights management. I suppose that until an AI can sue its programmer or producer, it's probably a moot question. As for artists themselves, I find it hard to believe that they will ever be displaced. But perhaps some of them will end up as editors, sorting through masses of AI produced stuff, throwing out the garbage and selecting whatever is worthwhile. Maybe in art, as in other areas, what we will eventually see is a collaboration between AI and humans with AI enhancing human activities but not replacing them.

    • My own personal standard by which I judge art is: is this image a meaningful metaphor? (i.e. Is it a surprising, or an insightful, or an unusual, or an honest, or a pleasing, etc. metaphor?) This is just my own personal standard. It’s not meant to be a universal measuring stick. Does the AI-generated stuff delve into the realm of metaphors at all?

    • Interesting idea. I wonder whether metaphor is inherent in the work itself or if it's something that the observer brings to the table. Certainly, an artist may have a metaphor in mind while creating the work. But the artist may fail to convey her intent, so ultimately it must reside in the work, or at least the potential to evoke a metaphor must be there. In principle, I don't see why an AI created work couldn't have that same potential, even if the intent is lacking. I think it's also possible to train an AI to produce metaphors, though I'm guessing that the majority would seem very strange to us or not even be recognizable as such.