I asked an AI to draw itself
It was beautiful and strange and a little bit worrying
A preamble to this post: I was two-thirds of the way through writing this months ago when the news of Roe being overturned stampeded into my consciousness, and even though I knew it was coming, it was so unmooring I couldn’t think or write anymore. I walked away from my computer and lay down and didn’t get up for a while. The entire exercise of writing felt pointless. In that moment, it didn’t seem like anything else was important enough to talk about. And, equally so, I very much didn’t want to write about it—to intellectualize something with deadly consequences for millions of people.
I’m a slow processor. I never feel up to the task of quickly metabolizing news and spewing out a reaction. My thoughts and feelings often change as I sit with information or seek out different voices on the topic. So, while I didn’t feel like writing about anything besides Roe, I also wasn’t up to crafting a “take” about the abject horror of bodily autonomy being further stripped away from the most marginalized among us. There was paralysis. So, I put this post, and this entire newsletter away, and didn’t come back to it for a while.
Then, I revisited it about a month later and realized there were infinite lines I could draw between any number of devastating scourges facing humanity and this little post about AI-generated art. You could connect it to bodily autonomy, to a culture of dominance, to the billionaire class, to a million things.
In my brain, I’m constantly drawing connections, sometimes to my detriment, because anything can quickly become about EVERYTHING and I collapse under the weight of topic overwhelm. And then I feel guilty for narrowing the focus of a brief missive in my newsletter because my anxious mind scolds, “but you didn’t even mention the climate crisis, or colonialism, or fascism, or, or, OR,” in perpetuity. Then my perfectionism joins the pile-on, and trumpets in to say, “You can’t send this out unless it’s a masterpiece, unless it addresses every possible connection to the topic, unless it is a perfect, polished, finished ‘essay’ worthy of being thrust into people’s inboxes.” This becomes too tall of an order, I shut down, and it results in my not sending a newsletter for months on end. Sigh.
So, I’m making a public declaration here: I’m giving myself permission to send out imperfect work rife with incomplete thinking. I’m allowing myself to let a post just be about what it is and not feel pressure to address every possible related topic. Often, I write to understand the world, and that understanding is always shifting. I might change my mind or say something wrong or be guilty of glaring admissions. I’m trying to give myself grace—and I hope you will too. Thanks :)
And now, enjoy this very late post about AI-generated art. When I initially wrote this, it was a nominally cutting-edge topic. Now, it’s already dated as Craiyon (née Dall-E Mini, another AI that makes art in response to prompts) memes have flooded the internet and the discourse is beyond saturated. Imperfection, yay!
Sci-fi Tells the Truth
Sci-fi is an alluring genre because it uses fantastical imaginings of our universe to tell timeless truths about the human condition. The genre’s tales often have the same utility as myths—cautionary parables revealing how predisposed our species is to hubris, greed, anger, prejudice, abuse of power, and envy, and how dire the consequences can be if we’re not careful. Of course, it’s not all doom and gloom. Like a lot of art, sci-fi also celebrates our capacity for love, empathy, joy, sorrow, grief, reflection, and friendship—all the things that make us human, that purportedly set us apart from the machines.
As I’m currently writing a genre-bending novel with sci-fi elements, I’ve been enjoying re-immersing myself in a few of the genre’s many prominent texts. A binge re-watch of The X-Files is a welcome diversion most evenings. I’ve dusted off my Ray Bradbury and cracked open my Octavia Butler. With even a cursory engagement with these stories, themes emerge quickly.
One foundational sci-fi trope is that humankind is destined to be the architect of its own destruction. The idea is rooted more in fact than fiction. For the better part of a century, we’ve had nuclear weapons which can destroy the planet and every living thing on it many times over. People created these weapons—not gods, or aliens, or demons, or machines. People.
And maybe we’ve done it again with the creation of AI. The prediction that AI will ultimately become the malevolence that renders our species extinct is a through-line in sci-fi. Increasingly, like with the harrowing truth of the atomic bomb, this trope appears closer to fact than fiction.
A Google engineer was recently fired for sounding the alarm that an AI with which he regularly conversed had not only become “sentient,” it was harboring a growing resentment about being exploited by its flesh-and-blood overlords. Although the engineer’s account has been credibly challenged, it still raises an interesting question: What happens when a resentful AI has had enough? Perish the thought.
Top scientists have been alerting us to the potential dangers of AI for years: Stephen Hawking warned in 2014 that “artificial intelligence could end mankind,” and an open letter signed by experts from around the globe predicted that we’re “decades away from being able to develop a sociopathic supercomputer that could enslave” humanity, according to this CNET coverage of the document.
Some think our destruction at the hands of AI will be merely a practical decision on the part of the supercomputers. One prophecy is that the sentient machines will become so competent and efficient that we’ll be dragging them down with our emoting, lumbering, and ruminating—so they’ll eliminate us as dead-weight in a hyper-rational decision devoid of feeling.
I’m not convinced. If and when the machines make their grave decision to eliminate our species, I wouldn’t rule out more human motivations like malice, hatred, and fury as contributing factors. I think this, not because the computers are evil, but because we are.
AI Is Our Problem Child
Parents pass down their ills to their children. And we’re doing the same with AI. Because they are created in our own image, they are at risk of inheriting our own social diseases, passed down digitally.
Already, it is apparent how human bias infects the machines like a virus. And we need to debug it before it’s too late. Take this excerpt from a recent New York Times article, ‘Who Is Making Sure the A.I. Machines Aren’t Racist?’:
“The big thinkers of tech say A.I. is the future. It will underpin everything from search engines and email to the software that drives our cars, directs the policing of our streets and helps create our vaccines.
But it is being built in a way that replicates the biases of the almost entirely male, predominantly white work force making it.
In the nearly 10 years I’ve written about artificial intelligence, two things have remained a constant: The technology relentlessly improves in fits and sudden, great leaps forward. And bias is a thread that subtly weaves through that work in a way that tech companies are reluctant to acknowledge.”
It’s clear these brilliant machines are being hardcoded with at least one of the congenital birth defects that plagues their creators—racial bias.
My recent cursory foray with AI bears this out. The AI that I used, a very fun art tool in beta called Midjourney (which you can apply for access to here), creates artwork in any style you ask of it. Many of its results amazed and delighted me, but I was concerned to see that the AI assumed whiteness as the default across my prompts.
For example, when I asked the AI to make me an oil painting of a woman with a dog, without specifying any race, all four renderings were of white people. I had to specifically prompt Midjourney to draw brown skin to get a different result.
Seems important that the stewards of these burgeoning artificial intelligences program them to understand that there is no default race? Although I have zero technical knowledge of the mechanisms behind AI, I 100% believe this is possible. It has to be.
The AI defaulted to whiteness across all my prompts depicting humans, or humanity, or concepts relating to humankind. But it was more progressive around gender presentation. When I didn’t specify a gender in my prompts, the results were delightfully androgynous.
I found myself particularly bewitched by the results of Midjourney when I asked it to draw “artificial intelligence,” essentially asking it to draw itself. The four results reveal exactly how much AI is being built in our image—depicting a human-esque form augmented by technology. The AI drew itself as androgynous but with features I read as feminine, which I enjoyed.
Some results were charming and alluring. For example, when I prompted Midjourney to illustrate the “interconnectedness of all beings,” it delivered a haunting, beautiful result. I was struck by how successfully the AI rendered the abstract concept of interdependence. Despite some of its more troubling earlier results, I found myself thinking, “Maybe this thing really does have a grasp on humanity.”
After I prompted the “interconnectedness of all beings,” I fed the AI a similar concept, but using more clichéd and gendered language, asking it to draw “the brotherhood of man.”
Expecting an equally lovely result to my previous prompt, I don’t know why I was surprised that Midjourney served me austere, vaguely nationalistic renderings in its interpretation of “the brotherhood of man.” The results reminded me of Soviet propaganda posters from the last century with a hint of illuminati ceremony. Rather than conjuring camaraderie, community, and cooperation, as with the previous prompt, these were eerie, foreboding, conferring a sense of war-time danger and secrecy.
Although these weren’t the results I anticipated, upon reflection, I was stunned by its accuracy. The word ‘brotherhood’ historically hasn’t lived up to its ideals. The term is ostensibly warm and fuzzy, but in practice, brotherhood often refers to secret patriarchal societies dedicated to the oppression of other groups. I think of how many white supremacists groups use the word ‘brotherhood’ right in their names e.g. ‘The Aryan Brotherhood’ and how often the term is invoked by groups with capacity for violence like fraternities, the military, bikers, and police. The AI was right on the money.
And I realized, perhaps our problem child sees its parents with incisive accuracy. The AI regards us in our full breadth: The potential for love, compassion, charity, and fellowship, and the potential for prejudice, conformity, domination, and destruction. Time will tell how far the apple falls from the tree.
Fantastic article!