Dreaming While Machine pt.1
On AI art, activism, and the failure of our collective imagination.
Open curtain to a half-filled room of Web 3.0 enthusiasts and a handful of cynics there for free wine and something to do on a Thursday evening. I’m drinking room-temperature still water and wondering what I’m doing here. Various Mint 0s flash across a too-large screen at 180 House. The occasion is a preview of a show headed to Paris the following week for the 2024 NFT festival. After the gallerist gives an introduction to the show, each artist is invited to describe their work and process. The works are pretty unremarkable, and the rationale does little to convince me (one of the aforementioned cynics) of its merit. Both practice and curation suffer the STEM-native affliction of focusing on the “can” rather than the “should”. All of this to say, I did not like the show.
My interest was piqued by the introduction of a toy that I hadn’t seen before — an AI image generator that translates voice prompts into images on the big screen. They called it the “Dream Machine”. How novel! I thought. After an eternal five minutes of images of Obama and Batman hanging, morphing into one another, drinking coconut water on the beach while embracing, the allure of sophistication dissipated. Behind the single microphone, a queue of bros had crowded, they giggled to themselves about the hilarious images they could conjure. Hobbits playing basketball. Harry Potter in hot pink Hogwarts robes. Darth Vader at the pub. Two Darth Vaders at the pub. Darth Vader sharing a pint at the pub with hobbits in Hogwarts robes. The possibilities were endlessly boring.
I hated it. I could not look away. Already my mind ticked over with all the hot takes I could write. But more than anything I felt sad for them. Is this the best they could come up with? A half-baked recreation of a film that already existed? A malformed centaur that tricked the parental controls into showing tits?
A brief interlude as the next promising millennial mind takes the stage to whisper secrets into the Dream Machine. In these moments between prompts, the computer continues to generate similar images on the same theme. In an unfortunate transition, an ape leaping through a cityscape morphs into a Black boy with a backpack. I wince at the thought that a computer stumbled upon a centuries-old racist trope.
Joy Buolamwini borrows the vernacular of social scientists and feminist theorists to propose the “coder’s gaze”. By this, she means the ways in which the biases and weltanschauung of the engineers and programmers become embedded and enshrined in the program itself. On one level, it is an inevitability of the process of creation — in everything that is created, there is a microcosm of its creator. This is acutely true for AI trained on datasets that exclude masses of information about the world deemed “irrelevant” to the project at hand. Buolamwini noticed that the facial-recognition technology she herself was working on did not effectively recognise her as a person. She attributes this primarily to the demographics of the people who create the technology, a workforce that heavily skews male and non-black.
The consequences are indeed racist, misogynistic, and transphobic. For example, Giggle, a social media app marketed as a women’s only safe space that used facial recognition to verify users’ identity. The backlash was immediate and justified, particularly as TERF ideology was rearing its ugly head with its eugenicist approach to who could exist within the category of “womanhood.” More dangerous still, is the use of facial recognition by militarised wings of the state. For example, if AI is inefficient at distinguishing between individual black people, how can it be reliably used in policing without increasing the rate of wrongful convictions for Black people in an already unjust legal system?
The solutions presented range from bog-standard DEI to a technologically evolved DEI.
More women in coding! More black women in coding! Make the next CEO of Microsoft a Black woman! We will herald in a DEI’d world of queer bombs and anti-racist climate collapse!
At the same time, there are direct interventions in the models to force them to be more inclusive. A viral tweet showed the sad result of one of these attempts. A benign prompt to reimagine the “guy with swords pointed at him” meme as Homer Simpson rendered Homer as brown rather than Simpsons yellow with a name tag that reads “ethinically ambigaus” (image generators are notoriously bad at writing). The tweeter notes that they did not ask for this. This undermines the very purpose of an AI generator as a user-facing tool. Instead, it generates what the creators have instructed rather than the user, even more acutely than before the correction. Furthermore, the promise of machine learning is that it develops based on its interaction with users becoming more sophisticated, moving closer to the ideal of human intelligence to serve human needs and desires.
For the species that invent tools, they do so either out of need or out of desire. Each individual arrives at the programme yoked to the twin beasts of lust and yearning. Without a foray into Freudian or Hobbesian domains, the fault was never in the code. AI has neither motive nor free will. The result is based on what the user asks of it. On its own, it is too disinterested to have a perspective on creation, to even desire creation. What is spat out by these models is, for the most part, meaningless drivel that fails as art. It is art to the extent that shadows on a wall are people. This isn’t just true of AI but of all tools used carelessly – the camera lens, the paintbrush, the keyboard. What is produced is an unthinking, unfeeling approximation of life. Uncanny, I suppose, is the word.
I return to these groundless musings several months later. A few days ago an AI image went viral – potentially a first-of-its-kind evolution from 2020’s infographic epidemic. It features row upon row of rectangular forms arranged in a grid. At first glance they resemble tents, and then bodies wrapped in shrouds. Above which, rendered in white blocks the command “All Eyes on Rafah.” That image has been shared 47 million times. If you don’t know what is happening, and has been happening in the southern city of Rafah in the Gaza Strip, I’m not sure you would make out the same figures from these pseudo-realistic forms. The absurdity of the viral post is that it offers no context into what the people in Rafah are experiencing and what they have been experiencing at the hands of Israeli terror which is nothing short of genocide. In fact, the post makes no visual or linguistic reference to people – only the suggestion of corpses and disembodied eyes. At best it is soulless, at worst it is genocidal – you cannot kill what was never really alive.
In the same format, followed images calling for “All Eyes on Sudan” and “All Eyes on Congo.” In the case of Sudan and Congo, the global north is even more switched off to the daily atrocities taking place there. We have to question what these images do for us. What do we want them for? What do we need them for, beyond appeasing our guilty conscience in the face of our peers? I am not the first to criticise the logic of such an empty campaign:
“I think it's precisely the lack of context that makes this image politically safe for liberals to signal their "concern" for Gaza. Palestine is not named. Zionism is not named. It conveys urgency and awareness without naming the historical actors that have produced this genocide.” – @tsengputterman
“Palestinians have to record and document everything they’re enduring in Rafah just so the world believes them so to reshare an AI-generated template is sooooooo weird actually.” – @tashteshtish
Moreover, resorting to AI in this moment seems like escapism. It removes us from both history as well humanity. It fictionalises the very real horror. A falsehood that fails to be art is nothing other than a lie among many other lies in a moment when truth is more urgent than ever. That is not to say that art about horror is not possible (Guernica by Pablo Picasso remains a pressing example), or that AI cannot be used to create genuinely awe-inspiring art. But in this case, what do we need these AI images to do for us?
I get it, we’re all tired of living through “unprecedented times.” To be a cog in a machine is an easy but soul-destroying fate; worse still when that machine manufactures mass death. However, it is imperative that we question our coping mechanisms. Discussion of AI imagery might seem trivial in the face of genocide – but why we crave pacification reveals truths about those of us who are currently safe from it. After all, the present moment is only uncanny to those who ignore history and fail to imagine it repeating.
If humans survive another thousand years on our decaying planet, future generations could look back on our times with unprecedented clarity. They could bear witness to the granular details of our individual lives through what we share – like what you had for dinner on the 29th May 2024, and even how that made you feel. The dataset would be overwhelming for any individual to go through unless they had infinite time. The promise of AI is that it can do this for us – it can offer an amalgamated archive of our time, the sum total of all of our desires hidden in between everything that we say and do. What is left out is an important question, why is an even better one – today we write the who and the when.
Perhaps our cynical distaste for AI conceals, in part, what we hate about ourselves and our reality. I, for one, hate that for a few hours every day I don’t have my eyes on Palestine, Sudan, or Congo. I hate myself for the luxury of forgetting. I hate living at the intersection of knowing better and doing little. I hope you hate it too. I hate that, given a Dream Machine, I cannot imagine otherwise. I stumble, and fail, and fall in line. I hate that all I offer is a handful of rhetorical devices that amount to very little in the face of history. But I want to dream bigger than more inclusive technology of death, or art that isn’t self-insert fan-fiction. We owe ourselves better imaginations in every aspect of this miraculously detestable life.