Abstract Art and Artificial Intelligence

How would an intelligent software agent look at abstract art? 1 Abstract art is the conventional name for a variety of artistic endeavors starting
around 1910, and called by various names: pure painting, new pictorial realism , or objectless painting. Abstraction itself has been used since the beginning of art.

Not everyone agreed about the existence of abstract art. Picasso  once said that there was no such thing as abstract art.  You always have to start with something. You can then remove the appearance of reality, but the object would leave an indelible trace. A human is part of nature, and cannot remove oneself from it. (Picasso, Propos Sur L’Art p 33)

Previously, we demonstrated why today’s intelligent agents are unable to comprehend a representational work of art. Could an intelligent agent decode abstract art?

Here is a 1911 work by Wassily Kandinsky entitled Impression III (Concert), one of his early attempts at abstract painting.2 Images and some ideas about this work were taken from the first essay in Inventing Abstraction 1910-1925 How a Radical Idea Changed Modern Art  published by The Museum of Modern Art

Image is a 1911 painting by Wassily Kandinsky entitled Impression III (Concert),  one of his early attempts at abstract painting

Where would an intelligent agent start to look at such a work of art? Where would you start?

You might start with the title which gives you the idea that it is an impression of a concert. So what?

A concert has people, it has music, it has a performer, it happens in a concert hall. Do you see any of those things in the picture? If you do, why are they not clearly represented? What impression does it give you? Do the colors correspond to anything? Might you see some people in the audience, and a person trying to play a piano? If you do, why is this a work of art?

Would it help you to know Kandinsky and some friends attended a concert of atonal music by Arnold Schoenberg. The works played included his Opus 11, Three Piano Pieces. You can listen to it here.

Kandinsky was struck by the similarities between the idea of music without a tonal center, and a painting without a representational core. After the concert, he made two sketches:

One of the sketches of a classical music concert by Wassily Kandinsky that he used as a basis for Impression III.
Another sketch of a classical music concert by Wassily Kandinsky that he used as a basis for Impression III.

Do you see the resemblances now? Would that help you make sense of the picture knowing how it was abstracted from something real? Would an intelligent agent be able to recognize the relevant objects in the sketches and compare their shapes to the resultant work? This all seems somewhat sterile.

Kandinsky’s reaction suggests another approach. Kandinsky’s impression of absolute music – the idea that music was not supposed to represent anything – gave him the impetus to create his version of abstract art. Absolute music led to the expanded tonalities of Debussy and Wagner, and eventually to the idea of no tonality. Kandinsky thought painting should take a similar path.

Once he assimilated this breakthrough, he was able to create his first truly abstract composition without reference to representational forms, Composition V:

Composition V the first real
abstract painting by Wassily Kandinsky

The key question then becomes how do you feel about this work of art? You need to be able to relate the work of art to your own internal subjective world. What would it mean for an intelligent agent to feel something? What kind of internal world would an intelligent agent need in order to be able to relate artificial intelligence and abstract art?

 How would an intelligent agent obtain a subjective feeling for colors, for lines, for shapes? Even a human who hates abstract art, feels the lack of understanding in their internal world.  How could an intelligent agent even hate something? A human’s entire life builds up their internal world.

How would an intelligent agent build an internal world?

Probably the same way a human does, through experience. It would need, however, a learning mechanism that would be capable of internal reflection. Perhaps such a mechanism would have to evolve from more primitive mechanisms. Today’s mechanisms only copy and reformulate what already exists, they do not really evolve. Reflecting about the relationship between abstract art and artificial intelligence gives us insight into what it means to be human. Artificial intelligence is nowhere even close to this today.

  • 1
    Abstract art is the conventional name for a variety of artistic endeavors starting
    around 1910, and called by various names: pure painting, new pictorial realism , or objectless painting. Abstraction itself has been used since the beginning of art.

    Not everyone agreed about the existence of abstract art. Picasso  once said that there was no such thing as abstract art.  You always have to start with something. You can then remove the appearance of reality, but the object would leave an indelible trace. A human is part of nature, and cannot remove oneself from it. (Picasso, Propos Sur L’Art p 33)
  • 2
    Images and some ideas about this work were taken from the first essay in Inventing Abstraction 1910-1925 How a Radical Idea Changed Modern Art  published by The Museum of Modern Art