Last semester, I wrote an article cautioning against overzealous accusations of using Artificial Intelligence (AI) in the creative process. I still stand by the broad strokes of the concerns I raised, but the longer I thought about it, the more I felt I missed a necessary point. The reality is that it’s quite easy to identify AI realism. There are a lot of tells – hands are stereotypical ones, as are confusing spellings or sudden disruptions in what should be a consistent stroke (such as a stripe or a necklace), among others. Generally, the details of the image become worse the longer you look. The issue is primarily in what I’m going to refer to as “artistic non-realism,” and I want to explore this relationship.
First of all, what is artistic non-realism? For the sake of this article, I’ll be using it to refer to art that is intentionally and actively not realistic. This includes some formal artistic movements – cubism, abstract, and surrealism come to mind immediately – but also includes things like caricature or chibi or many styles of 2D animation. These aren’t just a series of flats that mimic body forms but lack realistic shading or an artist who struggles with anatomy. These are active choices to distort forms as part of the art.
While there are certainly a fair number of people who use this style to be pretentious, there are also a lot of real reasons it’s popular. Caricature can highlight unflattering features, making it popular in satire, parody, and other forms of art that mock a target. Chibi does almost the opposite, exaggerating cute and almost childlike qualities. Surrealism can be a good way to explore themes like mental illness or invisible problems in visual media. There are a lot of potential reasons an artist might gravitate towards artistic non-realism.
Okay, but what does this have to do with AI? Let’s talk about an artist I knew in high school. They’ve asked to remain anonymous, due to previous incidents where they were falsely accused of AI work and not wanting to relive those experiences. Their work uses surrealism to talk about their mental health experiences by making it look like their art is “falling apart”. This can include breaking lines or patterns, causing body parts to bend unnaturally, or messing up background details (like spelling on a sign). The goal was to make something that at first glance made sense, but the longer you stared at it, the more looked wrong. Remember how I listed some common tells for use of AI? All of those were strategies they’d use for intentional, artistic purposes in their work.
Because I’ve seen them use these tells since well before AI image generation became common, I feel confident their work is real. Set aside any trust I place in them due to knowing them personally. As an objective matter, I don’t see a world in which they could have been using AI early on. However, as AI tells have become more recognizable, they’ve had to ditch much of their surrealist style. It was to do something more realistic and therefore less likely to be AI or deal with endless false accusations.
At the start of the article, I referred to AI creating an uncomfortable tension in artistic non-realism. But I wasn’t referring to the usage of AI directly. The artistic community generally is of one mind about that. It’s not causing a lot of tension on its own. The tension is in how artists then react, being forced to abandon their styles or accept false accusations.
The concerns I heard from this artist aren’t isolated. A lot of artists are leaning either more heavily into realism or other hard-to-impersonate styles (AI still struggles with Western 3D animation, for instance, making that a safer area). Personally, the last time I wrote on this, I spoke about how my use of parametric modeling was labeled as AI. I’ve since had to pull some of these examples from my portfolio and I don’t use parametric design as often anymore, as these concerns haven’t gone away. I have other architecture major friends who used experimental rendering styles, and have more or less ditched those because they were labeled AI.
The reality is, mainstream artistic movements are not under serious threat from AI. I feel fairly good about saying that. At least for now, AI has too many tells for it to be a meaningful threat for the foreseeable future. And the reality is that – as several professional artists have told me before – any skilled person who makes a living off of their art will be able to draw in multiple styles (and mimic most styles effectively), so artists have an option in theory. Most artists I spoke to while researching this didn’t seem seriously concerned about that right now. The concern seems to have shifted to displacing people who work in non-realism styles because people, understandably, don’t want to bank their careers on a false accusation, but don’t want to give up their method of communication ideas, either.