Illustrating AI

🏭 After a year of AI shaping illustration, can illustration shape AI?

This is a letter of The (Im)posture — the newsletter from Julien Posture. If you like it and want more like it in your inbox, consider subscribing.

Hello all ! It’s been a couple of months since my last post, apologies. As it turns out, the first year of a PhD is quite intense. I’ve been working on a history of the concept of style for a while and have fully integrated issues of artificial intelligence to my research project, which I will tell you about soon. In a nutshell, I’m now interested in how illustrators, art directors and machine learning look at the same object, illustration, from different perspectives and with different goals, shaping its value in various ways. But for now, here is a short reflection on visual repertoires of AI. After a year of reading on how AI is redefining what the creative industry looks like, I felt the need to think about how creatives can redefine what AI looks like.

As you may have heard in the past year, investors with half a computer science degree and other long-termist billionaires have once again decided that instead of putting their immense resources to solve actual social issues they would come for artists’ livelihood and trivialize the work behind creative labour by making models that can make images from text. With the horizon of a creative industry without creatives, and a media discourse that’s stripping them from their agency, it’s hard to remember that illustrators have still have a some cards in their game. One thing I’d like to do in this essay is taking a break from the list of ways AI has been shaping illustrators’ lives and ask instead, how can illustrators shape the public opinion on AI by doing the one thing we do best, illustrating?

Freelancing has made it increasingly hard for illustrators to organize, unionize, and have a common voice to defend their interests, it is a problem. Yet, there has been some exciting development in terms of organizing and resistance in the past few months. One of them is a class action lawsuit against Stable Diffusion by the John Savery law firm along with artists Sarah Ander­sen, Kelly McK­er­nan, and Karla Ortiz. The goal of such action has been to create a legal precedent that would substantiate the claim that generative AI diffusion models create “derivative” works and therefore constitute copyright infringement. Another one has been a more short-term solution to the scrapping of “data” (or theft of copyrighted work) from artists. A group of researchers from the University of Chicago, led by Prof. Ben Zhao (along with artists such as Ortiz), has developed a tool, Glaze, to shield images from AI gaze. The program adds very small elements, barely visible to the human eye, to an image before it’s posted online. Thanks to these micro changes in the digital file, an AI wouldn’t be able to extract the formal patterns that constitute an artist’s style.

These strategies are amazing examples of artists’ resilience and ability to organize and collaborate with other experts in both legal and computing fields. Yet here I’d like to draw your attention to another way of actively shaping the generative AI debate in the favour of artists using our own expertise, illustration. One of the greatest powers of illustrators has been to subtly, but surely shape people’s visual worlds. In our image-saturated environment, our images convey to people key informations on what to think of a publication (is it trustworthy or not?) or a brand (is it high-end or affordable?) or a book (is for adult or chilren?) etc. In editorial context, illustrations can stir a text in one direction or another, and frame the reading in such a powerful way it can even override the intent of the original text. The choices we make in depicting a subject through a visual metaphor rather than another are crucial in the formation of cultural representations of that subject. AI, as a culturally specific, historically situated, human made thing also has a visual existence that have been shaped by illustrators. So what are the metaphorical and graphic resources used today to visualize AI ? What does AI look like? And more importantly, what can illustrators do about it?

Blue robots and black boxes

AI systems seem to always be blue. Blueish robot hands or humanoid profile floating in a digital void of teal and ultramarine. The hue is commonly associated with intelligence and trustworthiness, two things tech companies are eager to make us believe about machine learning. The robot we often envision as the embodiment of AI is also part of a culturally situated narrative. Some researchers from Cambridge have traced the long history of AI narratives from ancient Greece machines made by gods to seventeenth century automatons. They show how consistently, these stories portray anthropomorphized visions of AI, as if we’re only able to recognize intelligence if it looks human. But anthropomorphizing AIs doesn’t mean to account for the humans behind these systems, quite the opposite.

Machine learning has been visually defined by the absence of human presence and labour. Instead, the way we talk about and depict it is more like a force of nature over which none of us has any power, a computational wave that is as inevitable as future itself. Similarly, AI research seems to be framed more as a form of Renaissance-type discovery of a new territory, with computer scientists as explorers more than as makers. The “black box” of AI for example, this unknowable space to explore, shares many similarities with the Terra Incognita (lit. unknown land) of 15th century maps, and just like the latter, the “exploration” of the unknown justifies relentless colonial extractive objectification of humans and nature alike.

Of course, these are convenient fictions far from reality. As Kate Crawford argues in her book Atlas of AI :

We need a theory of AI that accounts for the states and corporations that drive and dominate it, the extractive mining that leaves an imprint on the planet, the mass capture of data, and the profoundly unequal and increasingly exploitative labour practices that sustain it.

What AI can look like

Drawing from Crawford’s call for different theories of AI that include its infrastructure, I’d like to suggest that we need a different visual and narrative repertoire to illustrate stories about AI than the one I just outlined. And who better than illustrators to do so? At a moment where tech companies are shaping the way the public sees what creativity and creative work is, there’s a chance for editors, art directors and illustrators to work together to paint a more accurate picture of what AI is and does.

From abuse of underpaid labour in post-colonial settings to the objectification of people’s lives and work as “data” to the extraction of rare minerals to power increasingly demanding computational needs, AI runs not on sci-fi futuristic utopia but on the appropriation of today’s world’s material and human resources. Our illustrations can, and should, reflect these aspects too. The moodboard I’m imagining to direct accurate art on AI includes images of toxic lakes of lithium and other rare minerals mined in lands made sterile by human greed. It also has secretive, energy demanding data centres as well as underpaid crowd-workers and data labellers in global south countries. The colours of this mood-board are not white and blue of futuristic, digital spaces but the yellow and pink of lithium pools and the grey and browns of concrete buildings. And most importantly, it always includes people, from the workers made invisible by the idea of automation to the ones whose work has been stolen to train models.

Unlike Dall-E or Stable Diffusion, we’re not bound to average the previous representations we’ve already seen, as artists, we get to subvert and unsettle commonly accepted ways to represent things in new ways. Of course, illustrators can’t be the only ones bearing the weight of this shift, it has to be embraced by editors and art directors as well, to open up the realm of possibilities of what AI looks like. What I’m suggesting is far from being a solution, it’s merely another brick to the collective efforts to reclaim a sense of agency as a community in a moment when some people seem to think we can have valuable work without workers, creativity without creatives, art without artists.

You can also follow me on Twitter or Instagram. This letter is the start of a conversation, don’t hesitate to share it to someone who might like it, or to like and comment below.

P.S. I cancelled the paying option of the newsletter, it was stressing me out too much! maybe it’ll be back when I’ll have more time to dedicate to my writing, but for now, I just want this place to be for everyone. Cheers!

P.P.S. My PhD is about visual culture as a whole and within that I have many interests and I’m wondering how much of this should I showcase in this newsletter. I would really appreciate if you’d let me know what you want to see more of !