Prior to deep-learning algorithms, audiences recognized a given piece of art as the product of its creator’s artistic vision. The Starry Night, per say, is not just a painting of some starry night: it’s the visual imagery of Vincent van Gogh’s The Starry Night; and the artwork’s different features and elements help spectators internalize van Gogh’s ownership of and labor over the piece. But complications arose since the emergence of digital art, and the near-unlimited access provided by the internet furthers these challenges.
Now, these traditional conventions for recognizing artists and their labor dissolve substantially in the presence of artificial intelligence. Not only can people (artist or not) use these algorithms to generate large quantities of creative content, the individuals who opt out to get more involved in their work face doubt from audiences: that is, any presence of AI collaboration massively distorts to whom (or to what) the original idea traces back.
To see this matter resolved requires us to know where artistic visions go within digital and multi-minded interfaces. This next installment of AI in the arts aims to find that next step forward. Here with me is James Taylor, a digital specialist with keen expertise in software and designer platforms.
This interview has been edited for length and clarity.
Smile Politely: Where is your artistic portfolio concentrated, in terms of medium, styles, etc?
James Taylor: In terms of styles, I’m all over the place. I do different things, depending on how I feel at the moment. I do a lot of studies on existing photographs or works. I work with something I call “fractal filigree”, a black-and-white medium that ends up looking like a finished artwork, but really it’s the assembly of very fine line designs; as you get closer, you see the different details, but it looks like a coherent picture of something from farther away. I also do cartoon-styled robots—those seem to be quite popular—and I do lots of procedural work: basically mathematics and software to generate a final image that looks like a thing that’s composed of symbols. I do quite a few things.
SP: Speaking of software-based art, have you ever felt pressure to use or turn to AI within your practices?
Taylor: I don’t use AI, but there have been times where it’s been tempting. There was one time I used it, but the results weren’t great. I don’t think AI as a whole is there yet. Apart from the ethical concerns of AI, and in terms of its use as a tool, using AI depends on how specific your end goal is. If you have a very specific thing in mind, trying to meet your expectations is extremely difficult. If you have a vague idea—say, Mr. Spock fighting Elvis in a boxing ring right—it can give you a picture of that idea, but the result will be fairly non-specific. You’re not going to know until you get it. Also, I think AI is not yet in a position to be super useful: it’s good mostly for idea generation. If you just want a pretty picture of an idea but don’t have a specific look in mind, then yeah, AI will spit it out and give you what you want thematically. But in terms of getting an idea out of your head, current AI is not super useful.
SP: How would you distinguish the algorithmic art you do from AI art, seeing that they’re both similar in terms of the coding aspects behind them?
Taylor: Again, I think it boils down to the specificity of what the artist is after. With the procedural stuff I do, there’s a lot of randomness involved, but I can steer that randomness towards a specific result. It will look like the thing I’m looking for, and these algorithms I work with can generate different images out of that randomness. AI does a similar thing: it uses lots of randomness to then generate imagery on top of that randomness. But AI is also less likely to produce that desired look you want, whereas my software is a lot more steerable. You also wouldn’t want to do pictures of people using the stuff that I do. It’s better for atoms and surfaces and that kind of thing. AI can theoretically give you anything, but the less photographic reference there is, the less accurate AI will be at giving you your specific thing; and since it can’t go out and take its own photograph, it’s limited to the photographs available online. Depending on what you’re asking for, it may be better, or it may be worse.
SP: What will we gain and lose by allowing deep-learning algorithms and machines to become a greater part of artistic cultures?
Taylor: You’re losing a lot of your personal outlook and vision when you use AI. I’ve seen people say, “Here’s a picture I made with AI.” They didn’t do it — the computer did. If somebody came to me and said, “Make me this,” and I did that picture, that’s my interpretation of what the person told me. But whatever that person’s vision is, they can tell me, and I can take it into account and give them the picture they wanted. Of course, they might say that it’s not quite what they wanted, but it’s still their vision: it’s just filtered through my vision. For any given idea, if I do it (i.e. the art that the idea gives life to), then I make that up. And if you ask me for it, then I will make the art, but it’s still your idea. What we see with AI is that somebody has an idea and they ask the computer to do it, and then they say, “I did that.” But if you asked me to make a picture, you wouldn’t say, “I did that.” You would say, “James did that, based on my idea.” And that’s a pretty big distinction, right?
SP: What are some proactive measures artists today can take for AI to comfortably have a place in future art culture?
Taylor: Visual artists can use tools such as Glaze/Nightshade to protect their work from being included in AI datasets; I don’t believe any such thing exists (or possibly can exist) for written/audio works. What specific measures can be taken to include AI in art workflows, right now I’m not sure I can answer that. It’s going to vary greatly for different artists and their different moral codes. An artist can include or reject AI in their workflow, but a non-artist can leapfrog that same artist’s comfort with the technology to make money on social media, or by flooding the market with prints that are essentially the same thing with 300 variations. A regular artist can’t produce that kind of volume to compete. But a discerning buyer can look at what’s available, and they can decide to support the human who worked hard on something over the corporate entity just cranking things out with the algorithm. I doubt most people would care about the difference over an end result they just like, but hopefully there’s enough who do to not make art a non-viable path to making a living in the future.
Enjoy browsing more of James A. Taylor’s work here and here.