Note — Jun 23, 2019

The Past, Present, and Future of AI Art

TL;DR instead of comparing AI to “normal” art or looking at it purely as software, we miss the better way to critique it; as a continuation of computer generated art / aesthetics, which have been happening since the 1950s. Also takes us through some of the history of the field. Connects to previous issues where I included pieces on centaurs, hybrids, and AI as medium.

Finally, photography and film have long settled the issue of machine authorship (that I will discuss in detail below): “the owner or operator of the machine owns” the work created with it. […]

Nake’s argument is simple: there are no masterpieces in computer art because computer art is not about the production of “pieces”. It is about the production of system designs, and about the beauty and coherence of these designs. In other words, it is the method, not the artifact, that is relevant for the aesthetic judgement of a work. […]

In contrast, AI art, in the limited scope discussed here, has the problem that it is always essentially mimetic. After all, all of a neural network’s knowledge about the world comes from the data it processes.

Pure mimesis, it turns out, is impressive, but has no lasting aesthetic value. It provides immediate gratification in the easy recognition of similitude, but this quickly wears off. In other words, purely mimetic AI art becomes kitsch fast. […]

It is often overlooked that (non-trivial) art progresses much like science does - by building on a history of invention and discovery, sometimes taking incremental steps, sometimes questioning and overthrowing paradigms.