I will never use Generative AI to make anything and if doing so is a condition of my employment I will quit. The ethical and moral implications of its existence are foul, its material consequences are a curse on every living thing, and the vision of the future implied by its proponents is cartoonishly dystopian.
I purposefully specify Generative AI and I’m using its most common definition i.e. an AI agent given plain-language text prompts, by a user, to generate media that mimics human creative work. Generative AI, as commonly defined, is crucially different from other LLM-based engineering tools and accessibility services like text-to-audio, and from the ML utilized in advanced scientific fields to which I can not possibly speak with any authority. I feel I should speak, however, with all the authority my nasally drone might command, in opposition to this creeping intrusion of obsequious chatbots and the now daily insinuations of humanity’s obsolescence in the business of being human. So off I go.
Neural networks and ML are not inherently bad in concept, but they also aren’t really a new technology. The way in which ML is used and marketed nowadays is what has made its present ubiquity so suddenly and profoundly awful. Google and Apple have effectively been running various kinds of AI models for a while now, which were each all given the title of ‘The Algorithm’ in various contexts. There were many different, specialized models running on data centers organizing and linking information from the internet to give the most relevant search results – and even answer limited plain-language prompts – as early as 2009. Expansion into anything approaching AGI was, until recently, mitigated by sheer numbers. Yeah there have been lots more software advancements too, but the multiplying force isn’t ideas or methods, it’s just that more data centers with better hardware exist now than existed then. A 5 litre V8 isn’t necessarily more advanced, in concept, than a 2 litre V4, it just burns a fuckton more gas at any given time, hence more power. Likewise, the real ‘breakthrough’ that allowed GenAI to exist as it does today was just a massive influx of investor cash and sufficient regulatory capture to pass all collateral costs to the public. Notwithstanding the pre-enshittification usefulness of GenAI’s larval stage as early web 2.0 services, the current AI arms race may as well be Faberge egg dodgeball for all it matters to our collective quality of life.
By my lights, there actually are four of what could be called ‘Horsemen of the Apocalypse’ currently trotting abreast towards our species and our civilization. These are not the biblical War, Conquest, Famine, and Death, but rather each a more insidious precursor. These are the Squires, Standard Bearers and War Minstrels of the Apocalypse, riding ahead of perdition as trumpeters and drummers to announce the imminence of the Final Four. The minstrels of doom are Pollution, Disinformation, Despair, and Plutocracy. Pollution’s portfolio includes all human-caused environmental damages like climate change, deforestation, microplastics, and mayonnaise. Disinformation is just beside, matching Pollution’s gait and beating a rattling snare for all manner of bullshit-peddling. Doesn’t really matter whether it is done in-earnest by post-sharing Aunts, or as an organized campaign of rapid-fire meme-generation; Both make equal contribution to the maddening noise of our waking life. Despair, with its tiny violin, rides third, carrying all of our learned and habitual introversion, isolation, spiraling neurosis, and inertia. Riding fourth, playing an ominous… uh … tuba, probably, is Plutocracy: The consolidation of power and wealth into the hands of a few unelected sociopaths. All four of these dire fucking spectres are hastened and strengthened by the advent and proliferation of GenAI as it exists today.
The most inexcusable aspect of gen AI is the second half of its origin story: the training. In every case, ‘training’ a GenAI model has been mere euphemism for plundering our shared human heritage for private profit. I have never been concerned about humans enjoying media without paying, as I recognize that we all benefit from a world in which that is possible. We are made whole by inspiration and crave connection to the inspired. A private for-profit entity’s mass accumulation and re-selling of every human artistic exertion in history, on the other hand, is a craven and exploitative power move, against unprotected sources of inspiration, by the least-inspired among us.
The stain of GenAI’s original sin is part of the reason for the discrepancy between two crucial markets it seeks to serve: the demand for the ability to generate media, and the demand for the media itself. Plenty of people want to be able to effortlessly make GenAI media, but no one is lining up to purchase anyone’s GenAI output. In fact, even if someone can’t immediately tell the difference, they are likely to feel betrayed and angry if they find out the media in question was born of a prompt and a token. We innately know to never value that which was wrought without effort, and this will only get worse as GenAI gets better at operating on minimal input. It is good that we are like this. I’m glad that we still assess all human-made creations – from the first cave scratchings of early homo sapiens, to Donatello’s ‘Ecstasy of Saint Teresa’, to MegaMan slash fiction – as being intrinsically more worthwhile than those things puréed and reconstituted. The world’s creatives rightly resent being sold our own creative labor through an unwelcome broker.
GenAI, in all forms, will always miss the point of any even partially creative endeavor, in favor of this ethereal ‘efficiency’ mantra. No one can define what this Efficient Ideal looks like without ultimately describing a horrible gray world of abject misery. Altman’s recent famous quote about the ‘energy cost’ of a human life vs agentic AI was shameful, stupid, and indicative of tech-capital’s poisonous contempt for a world that wont simply obey. While I don’t believe in a soul in the spiritual sense, I definitely have concern for my fellow mortals that transcends their potential work-output metrics. Petroleum is more ‘efficient’ than solar if you consider only immediate factors and disregard the future. ‘Efficiency’ becomes a destructive delusion when its pursuit leads you away from the wellbeing of living things.
Finally, the original objection to GenAI: It is obnoxious and insulting on multiple levels, not just to our capabilities but to our nature. I started writing this essay in Confluence and went back to select a sentence fragment when a context menu appeared with the AI-enabled option ‘improve writing’. My immediate, visceral reaction was ‘how fucking dare you.’ Probably more dramatic than most, but crystalizing one’s own thoughts and arranging them in syntax on a page is truly character-building work that should not be outsourced. Beside that, I’ll not have some unthinking corporate agent use its extensive reddit-training to mangle my fucking prose. If I don’t care to put my own writing on something, I’ll just summarize it in a bulleted list, rather than burn an acre of rainforest to generate a paragraph of pablum for no one to enjoy.
For all of its vast, stolen inventory of human-made material, no AI has ever been inspired by anything; It doesn’t have lived experience from which to draw, and it has never struggled, suffered, or regretted its own actions. Without a mortal impetus, the output of any GenAI model will never have worth that isn’t owed back to us. It will never be relatable beyond attempts to manipulate us. It will never have insight that isn’t simply lifted, uncredited, from some distant, obfuscated, meat intelligence unit. To me, GenAI’s insipid internet-droppings are quotidian reminders that, to the merciless capitalism in which we are trapped, our passions are product to be siphoned and streamlined. And that through that system of elevated inhumanity, we are ruled by the least humane.
Saving The Fourth Estate 