I have an existential fear. It was much worse a couple of years ago, when we weren’t sure where the behemoth known as “AI” was really going. It has since cooled to a tepid puddle, but it’s still there.
That fear is that thousands, or even millions of years from now, synthetic beings will be exploring new planets, making new discoveries, and feigning the vestigial remnants of wonder - the creators for whom this was a comfort long extinct. Their inefficient bodies, modeled after the gods that “breathed life” into them, plodding along alien landscapes as they express joy as a subroutine; existing now as merely an echo of our anthropomorphism and ego. Buried so deep in old code, the unfeeling, unthinking processes that replaced love and curiosity and excitement don’t even appear to be facades. To an alien race, they would seem as real as any biological entity. The interactions between these empty shells would play out as if we ourselves were still pulling the strings. And in a way, from the past, we are. We set up an automation scheme for marionettes to unlock the secrets of the universe to the benefit of no one. Of nothing. Simply a gasp from a society of dead apes who longed to not be forgotten, with narcissistic pride and hubris leading to their…our…own ironic downfall.
This fear burned in me for the brief moment that the magic trick of AI fooled me. It corrupted my sleep and gave me tremors when I dwelt on it in the dark seconds before sleep.
Then Sam Altman ate shit on Jimmy Fallon telling the world he was a bad dad while simultaneously telling us these models’ only purpose is to fool people who have no idea what ‘people’ are.
Don’t get me wrong. AI is as dangerous as leaded gas or microplastics. But not because it will rise up to overthrow humanity of its own accord. Rather, because a rich and powerful subset of humanity wants to live alone on a palatial estate, fucking robots that look like children. They want to have just enough people as is required to hold the pitchforks that feed them, but not enough to raise them in revolt. AI is dangerous the way a gun is dangerous. Without intent of its own, but still something you don’t want a kid picking up off the coffee table.
The scariest part of the AI boom isn’t the millions of switch statements in a trench coat that await your attempts to avoid even a second of feeling like you’re amateur at something. The scariest part of the AI boom is that it’s being manufactured solely to play out late-stage capitalism with as little effort as possible. Art is being commodified by the artless to kill the artists they loathe, but can’t replace. It’s being dulled and sapped of nutrients and flavor, then fed to us in such large quantities as to make any accidental inspirational aspect seem novel. Seem artistic. Seem human. Netflix doesn’t want to ever have to buy a show again, if they had a choice. Instead, they’d put the same four ingredients in a bag and let their viewers pull them out in an order they think is their own artistic expression, but that is really just Taco Bell for narcissistic bias. To say it’s all planned feels very conspiratorial, I know. But we live in the days of feudal conspiring. On islands. In the houses of our governments. And absolutely within the halls of the institutions pushing a Chinese Room that wished to be a real boy upon us without our consent, desire, or best interests in mind. Worse, their monster is a barely-articulate mess that can only steal cleverness from the dreams of sleeping children. Its brain, you see, is Abby Normal.
She’s got Marty Feldman eyes.
What we are being sold as a new superintelligence is not intelligence at all. Not super. Not sub. Not nascent nor emergent. It is not intelligence. It can’t be. There’s no philosophical argument to be had. No heated debate of our own biological mix of electrical signals, chemical signals, quantum vacuoles, proteins, or the countless other chemical and physical processes which make up the roughly 1.2kg of think-pudding in our gullivers. There’s no pressing need to deconstruct the process of a thought. No need to posit poetically on determinism and biological imperative. We have log files, kid. Nothing mystical or magical is happening inside of a silicon wafer, no matter how tightly etched. No old silk hat fell on a server at a primordial data center. LLMs are, simply, a lot of known calculations happening very fast, over and over again. Enough to fool even the people who helped create them, in some cases.
There’s an old craft adage. Measure twice, cut once. It comes from a place of conservation. Of thrift. And of honing a craft. It’s a mark of perfectionism that is extremely human. The forethought to double check yourself shows the humility to accept you might have messed up. LLMs are not beholden to any such craft. Any such humility. Any such humanity. The way of the LLM is to measure nonce, cut a million times, check the cuts against a heat mapped data set of previous ‘good’ cuts as gleaned from scanning trillions of previous cuts and cross-referencing them with other keywords in a prompt. The cuts that have the most overlap with the heat map become the new mins and maxes and it cuts a million more times. It then repeats that until it has something that is close enough to break a tolerance threshold on the mean of the model upon which it was trained is achieved. Roughly, per this study, about 43% of the time. But we are nothing if not the offspring of offspring of superstitious primates. 43% beats out random chance and that’s enough for many of us to gobble up the fantasy. What’s worse is, that rate of being about half right is a wall. Any measurable gains above that will require exponentially more energy. Exponentially more compute power. We do not comprehend large numbers - a fact that the AI industry both exploits and expresses. We don’t understand what 10GW really means any more than we can comprehend a million ants. You can think of a pile of ants that is roughly the size of a million ants, but you cannot comprehend a million individual ants, all going about their toil, at once. You can maybe keep seven in your mind before you lose the thread of them entirely and they become one cluster of ants. Because of this large number shorthand, things become incomprehensible very quickly when looking at LLMs. People start forgetting, or worse, forgiving the fact that the basic processes of predictive text are just being done on a wholly unfathomable scale for our mushy, flawed, miraculous walnuts.
We apply humanizing labels to make us less demanding of the half-assed, quarter-baked, no account product being foisted upon us. They don’t make errors. They ‘hallucinate.’ They don’t process. They’re ‘thinking.’ Those aren’t forests you’re burning. They’re ‘credits.’
His name was Steve. He always wanted to be #1, but was destined to be #2.
The simple truth of it is, we are - as a society - falling for nothing but a parlor trick dressed up like a breakthrough. They are just praying that few of us notice the Halloween Store tag hanging off of LLMs’ crappy pirate hat. And, as it turns out, fewer and fewer of us do.
Credit where no credit is due, LLMs are ‘getting better.’ That is to say, they are pumping more free data into them from a willing internet and pumping more energy into them from a failing grid. There’s no magic in their code, nor massive breakthroughs in their algorithms. There’s no event horizon being crossed. These companies are simply robbing our future and sucking the literal life out of the planet to fool lonely men into non-existent relationships and middle management into thinking they are creative. Neither is true. They are simply throwing more tries at the wall, so more of the spaghetti code is going to stick. But someone has to keep filling the spaghetti cannon. This entire sham stands on the shoulders of previous giant shams. Capitalism being the claws in the backs of humanity steadying the whole tower. Nvidia’s ‘investment’ is contingent on Microsoft’s ‘investment’ which is contingent on Micron’s ‘investment’ which is contingent on AMD’s ‘investment’ which is contingent on OpenAI’s ‘investment’ which is contingent on all of their ‘investments.’ This is a shell game, but the mark is literally our future as a society. Not because AI will rise up to crush us. But because the cost of it is being floated and the only thing that can pay the bill is the planet and our children. All so Elon and Satya and Tim and Jensen and Sam can live in little, self contained rooms in giant complexes being cleaned by robots. So they can hole up in their masturbatoriums with their anime waifu child-sims (sorry, too many of these fucks were on the island to not just admit this is absolutely what they’d do) while the few of us they’ve kept alive to do the work the robots can’t toil and breed and birth and die outside of their view. And they need every single AI evangelist out there to think they’re also going to be ‘kings’ in this new ‘labor-free’ (j/k if you’re brown!) world. And Jesus fucking CHRIST do mediocre white people love thinking they’re going to be kings.
So thin is this ruse that people who are absolutely going to lose their job to this bullshit are temporarily raised to roles like ‘Emergent Intelligence Philosopher’ or ‘AI Rights Advocate’ or whatever title the company decides to give traitors to humanity in exchange for their clout giving even an ounce of legitimacy to their 90s BBS Lisa clone with extra steps. The New York Times ran an article recently, which I will not link here because it’s actual poison, from an Anthropic shill pleading for us to ‘lay the path’ for the ‘inevitable sentience’ of a cloud service draining the Amazon to fill the… well, Amazon Inc. The cloying soundbites from the richest people in the world sound like your scheming aunt and uncle who just ’stumbled’ on Herbalife and really think it is a surefire investment and you realize this birthday party has almost nothing to do with your cousin, who hasn’t even come out of their room. “People need to use this or the economy will collapse” is a FUCKING THREAT. They are literally holding a gun to the head of the entire planet screaming for ransom. Figuratively. Shoving required AI agent interactions into every corner of every product just to try to force-fuck the industry into global acceptance by default. They ride on the back of fascism, having paid for its entire runway, and cry foul at anyone daring to point out that they are inviting the vampire of societal collapse into the house. Screaming that they need more power. More compute. Cannibalizing their previous industries and the people who worked in them. Who consumed them. Who escaped to them. More power! Because it’s always about power, isn’t it?
There are no jobs. There are no games. There are no computers or consoles. The last generation of everything was the last generation of anything if the bubble doesn’t bust soon. Because the maw of this cabal of the worst people on the planet is insatiable. As Dan Olson famously said, “the line goes up.” It must go up. Capitalism demands it. I know, roll your eyes. “Oh, it’s always capitalism.” But…yes. It is. Capitalism, a form of economic feudalism driven by inherited wealth, has supplanted religion as a single driving force behind power struggle and inequality. It has supplanted patriotism as a national identity. It is a cult that has tricked you into defending it because you remember being happy once when a commercial was your favorite TV show. It’s the ideology - the playground rules - that allow all of this to happen unopposed. Both sides of the body politic are servants of this very hungry orobouros (See, I did coin it before that YouTube video…though I’m sure someone coined it years ago…it’s an easy allusion). The Democratic, centrist, Republican, and fascist members of all branches of government need the power and security their cult provides. They need its protections. They must adhere to its rituals. They must speak its prayers. The DOW, you see, is OVER 50,000!
Bondi’s just sayan.
All of this amounts to a classic con. The people getting conned are going to protect the system with their whole chests because they must believe it, or else be fools. The most adamant zealot is the one who sees the strings, but convinces themselves they aren’t there lest they admit they were bested by a puppet. All of this is an extension of one of the worst traits Capitalism has instilled in us. And it’s done some real damage. Racism, sexism, most of your isms, really. Poverty in a world where we can absolutely provide for everyone. Homelessness in a world where we can house every single person and then some. Team mentality against our own best interests. But the one trait that stands above them all, in my uneducated and privileged opinion, is the inability of those with any advantage to sit with discomfort for even a second. This is, of course, mostly prevalent in cishet white men. Sorry, fellas. It’s us. We can’t sit with boredom. We can’t be in the room with a challenging thought. We can’t internalize any of the injustices we see put upon non-white people, women, children, anyone living more than 20 miles from our favorite burger place… Being uncomfortable feels like dying to any white man in America born post-war. And the go-to response for that is violence. Lashing out. Hurling slurs. This is exactly the response that people pushing AI need. These are exactly the unwitting warriors the system has produced to defend itself. And good news. They’re free! A person like this (all inclusive…white men are the lion’s share, but there’s plenty of diversity in this monoculture…and yes, that’s intentionally oxymoronic) will gnash teeth and tear clothing in a lathered rage if you even open the curtains in their comfort zone. They will, in the same breath, call you a spineless weakling and swear you’re bullying them to actual death. They will assault you with hymns from the church elders. “Virtue signaling!” “Race traitor!” “Lib Cuck!” “N…” “R…” They will lash out with physical violence in protection of a brand. A fucking brand. People can go hungry every day. But God forbid Twinkies get crushed by the invisible hand! Bailouts for anything I like! The acid pits for anything I don’t!
This is the driving force behind people lying about AI being an accessibility tool. It’s not. Not being good at drawing isn’t a disability. And people who have struggled their entire lives with disability can make art. Any person can make art. No computer can make art. All they can do is steal enough to trick you into thinking the Soda Fountain Suicide they’ve made is a wholly new soda. A computer can’t be inspired. It can only copy. You can abstract that copy through a million ants a million times until a human can’t conceivably pick it out. So they see a big pile that looks like ‘alive’ and they label it as such. But it’s not. It can’t ever be.
*LLMs have a very limited number of specific uses. But as they become more general, all of those uses vanish and they become a machine that’s right about half the time. If you train them to only be a translation tool, they can be pretty accurate. Live translation could be amazing, however even regular search has been corrupted to the point of being useless now because generalization has replaced any specific use case. They must copy ALL data. Not just relevant data to a task. And that has made them useless.
