In conversation with art director and 3D artist Misha Gurovich

“Using AI in my work doesn't make it easier, just different. From a pragmatic point of view, you could say that AI accelerates a large part of the creative process. Getting the hang of AI is time-consuming and requires research. Like any new skill, it takes time and effort to learn. When you open a new program, you don't automatically know how it works. I have a background in 3D and also work part-time as a Digital Innovation Lead at Mijksenaar, a wayfinding design agency in Amsterdam. Technology has always been a big interest of mine, so for me, using AI in all this kind of work is a way to get out of my comfort zone and learn something new. The fun part is: the more experience you get, the more control you have over the results. Every tool you use affects the outcome. You are an artist and curator at the same time.

“Fishing for beauty: that's what using AI in design comes down to.”

The whole process is like fishing for beauty, but with the help of an assistant. You're still not sure what you'll end up with. To limit the range of possible outcomes, I train my own AI models. An AI model is a software program that has been trained with data to recognize patterns or perform specific tasks. I make sure that they specialize in a certain subject and are good at one thing. For example, I have a model that makes faces the way I want, another model that makes flowers the way I want, and a third model that combines those flowers and faces just the way I want it to be.”

I then trained an AI model with that selection of my own work, and described the works on the AI model. The model then drew the works as I would make them myself. That way, I use my own artistic vision as much as possible, and not works by another artist that someone downloaded from the internet and included in a generic dataset. Of course, there is still a part of the model that was created by someone else. For example, the way in which it identifies subjects such as dogs and houses is pre-conceived. But when I train it myself, it doesn't look at millions of Google images of dogs; instead, it looks at the few images I took and fed to the model.

No matter how you choose to apply such models, your starting point should always be the purpose of the model. What do you want to use it for? Deciding what data to put in is also important, and it helps to have some experience to anticipate the results. The technology is new and solves a number of problems, but, like all assistive technologies, it has its limitations. So it's not necessarily more efficient. Still, the design industry is on the verge of something new and exciting. For some, that's a scary thought. But I'm not afraid of the technology, maybe just a little concerned about how people use it. For now, it will mainly help industries and enable individual artists, small indie teams and also large studios to create new exciting things. Like I said, it's a tool!”