Artist-made AIs

Artist-made large multimodal models could function as a contemporary art medium and be part of artistic practice.

A controversial aspect of that practice is the highly technical nature of designing large generative AIs purely for artistic purposes. This raises questions about the extent to which technical proficiency is “allowed” to coexist with emotional resonance in art, reminiscent of historical debates on perspective in the Renaissance and the impact of photography on painting. It is by now abundantly clear that generative AI art challenges many of the traditional notions surrounding “real art”, but what to think of artists actually crafting, not merely using, these AIs?

This question touches on something deeply human. There are good reasons to suspect that engaging artwork evokes strong response from the brain’s right hemisphere, because the left hemisphere is confronted with something it cannot comprehend, cannot encapsulate, with analytical reasoning alone.

I suspect that the same thing is true for AI. We associate the initial shock and awkwardly personal rapport many of us have felt while using AI technology for the first time with unexpected right hemisphere activation, which then diminishes over time. This diminishing is unavoidable, as repeated use of AI triggers the left hemisphere to reduce it to “just another tool”, explaining it away up to the point that it isn’t considered AI anymore (this is known as the AI effect).

In other words, you could say that there is a place in us where AI and art meet, and it is astounding that this statement is not only metaphorical, but also likely physically true: this place is our right hemisphere. This is why I believe that artist-made AIs have the potential to bring about strong art, even without considering the new possibilities it brings, such as highly interactive and personalized artworks.

My research for graphomaniac has taught me that keeping the artificially generated stream of thoughts and emotions engaging and “fresh” requires the large multimodal models to have strong “seeding” feedback from its environment, to which they are then “freed up” to respond creatively in the moment. This is true also for artists, scientists and engineers, and this leads naturally to the idea of making an installation which portrays collective, as opposed to individual, creative thought processes.

Go to next post ▤ or ▥ previous post in vault series.