Generative AI and the Arts: 10 Extrapolations

Generative AI software for producing visual art, like Dall-E 2 and Stable Diffusion, is having an impact comparable to that of the first cameras on painting. That comparison to the introduction of photography seems apt. The impacts will be seismic in some ways, but I think much less so in others. Below are extrapolations along 10 different paths, all circling around the visual arts.

1. Commercial Arts

This warning has been made repeatedly, but let’s get this out of the way: the commercial artist is in trouble. Some have proposed silver linings to generative AI software, for example, that it will save the artist from the drudgery of certain tasks. But there’s no getting around the fact that the net demand for commercial artists will fall dramatically. The creation of commercial visuals by generative AI will be deemed “good enough” by most business folks, who, even if they complain that they can only draw stick figures, nevertheless have an “eye”, and know what they like. The off-the-shelf commoditization of the creation of visuals will slim down, if not eliminate, most company art departments, as well as many graphic art freelancer slots.

2. Horror Show

The visual mashups created by generative AI can produce very unsettling images, especially when involving the human body. Artificial intelligence can travel the entire space of permutations on the human body, without the cognitive filters we humans unconsciously carry with us.

Each of us are experts when it comes to reading human expression, having been trained all our lives. Not only that, but human facial recognition is imprinted in our genetics, and seeing a face, or something similar to a face, will light up different parts of our brains. We expect human features and expressions to go together in a certain ways, and when that gets upended, as it does with AI visuals, the results can be unsettling. I think it’s similar to the “uncanny valley” that is approached when robots too closely mimic the look of humans without having the full gamut of human expressions and tics which we expect. With time, the uncanny valley will narrow, generative AI will get better at knowing how to simulate human features and expressions, but until then, this parameter space should be mined for some nightmare fuel.

3. Copyright and Fair Use

It’s clear that these legal frameworks will need changes to adapt to the new facilities of generative AI. Legislation follows perceived harms, so harms on living artists by the hoovering of their art into generative AI datasets will provoke some legislative redress. It’s clearer what the harms are when an artist’s livelihood goes downhill. But it will get trickier as artists begin to use generative AI tools in their own art, as questions of artistic intent and presentational context have always stymied codifications around creative practice. If one is playing in this space, it is probably advisable to make one’s trespasses before the ban hammers come down (à la Biz Markie and Public Enemy), or to have clout and money (like Richard Prince).

4. Fine Arts

The divide between fine art and commercial art is in some ways made clearer with the intervention of generative AI. The traditional distinctions between the two domains have to do with notions of artistic intention and uniqueness of vision. Generative AI does not possess these qualities: there is no intention or vision within the AI beyond fulfilling the user’s request to generate an image based on keywords. This is why I think the fine art world — read, the culture space of artists and galleries — will not suffer the adverse effects akin to what the more commercial arts will undergo, and will instead increasingly thrive.

We can speculate of a future time when a general artificial intelligence will exhibit a consciousness indistinguishable from ours. Then I think the question — is the AI truly creating consciously? — will be much more interesting. It’s difficult to speculate on at this point; we would need to kick the tires. More closely in the future, we can imagine a generative AI which canvasses the space of all artistic moves of the past, and generates a decision tree that can be used to create a new artwork. Would this be art? Sure. Would this be a threat to fine artists? I don’t believe so, since intention and vision would still be lacking in the generative AI tool. Those qualities would instead be traced back to the original creative impulse, which lies with the artist who directed the AI in the first place.

5. Romanticism

We may think of the baseline relation which forms the artistic encounter as a triad formed by the artwork, the viewer, and the artist. Romanticism is the historical movement which first placed the artist as decisive in the implied contract between artwork, viewer, and artist. Romanticism opened new valences in art — like self-determination, and individual thoughts and feelings — which persist today, though they have become relegated more to the background, and overshadowed by whatever “adolescent furies of our age”. Still, strains of romanticism in art will work to emphasize its opposition to the algorithmically generated.

6. Spontaneity in Art

A gift that an artwork gives the viewer is the potential to imaginatively re-enact the making of the work. We may fall in thrall to the chain of decisions displayed in the work, and wonder about the mysteries of the actions taken. This appreciation responds to the spontaneity in the work. The products of generative AI do not have this quality, as spontaneity depends on the embodied fact of an experience of responsiveness. At best, generative AI may reveal a clever algorithm, or a vast dataset.

7. Artist Brand

The “brand” of an artist is already a part of what we value in an artist’s works. Some may feel distaste at this characterization of what an artist offers. We can substitute the word “brand” with the nexus of fascinations which surround an artist’s touch, taste, thoughts, biography, and lifestyle. This confluence does not happen with generative AI. So at most, the products of generative AI will instill curiosity and a kind of sublime, but in my view a paltry pleasure in comparison with the feelings we have and the community we feel when we appreciate an artist’s works.

8. Political Art

Political art, or any art which proposes to explicitly stake a political position and offers the artwork as instrumental in furthering a political end, already pervades the fine arts, and will remain a dominant register of artistic production.

9. Image of Freedom

We have been told that one promise of AI and robots is that they will do all the things we don’t want to do, freeing us to do whatever we want. If asked what we would do if we didn’t have the drudgery of a job, we would choose based on our natural inclinations and interests. This is the ground for artistry — but more largely conceived beyond traditional art genres (like painting and sculpture) — to encompass all human endeavor. This utopian image of freedom is thus closely linked to the image of the artistic. These ideas of freedom, inclinations, and interests are alien to the algorithms of generative AI, and will assert themselves more strongly in the coming world of AI.

10. Is It Art?

Is the product of generative AI art? No.