Artificial Insanity

First, we need to be very clear on something that companies like Anthropic and OpenAI really don't like hearing in public discourse: frontier AI is still just a highly augmented stochastic parrot. A stochastic parrot with sufficient firepower can be a fearsome creature, and it can be useful in many different contexts. However, this parrot can't hope to emulate any serious form of consciousness, and therefore can't ever be capable of intelligent judgment. No matter how the data this parrot operates on is refined or how its selections are augmented, this parrot will always be prone to unpredictable and extremely undesirable token selections. You'd get better judgment from training an actual parrot, because it operates on a spectrum of data that the artificial parrot has no hope of operating on unless someone figures out how to make light move faster.

This brings us to a big underlying problem with the way AI is being sold to us: it's being exaggerated to sound like way more than it is, and it's being sold to people/governments as something that will be capable of exceeding human "intelligence" in the near future when it's not capable of intelligence at all. Indeed, it's possible to exceed basic human memory and storage limitations through this implementation of AI, but it's already a physically demonstrable dead end when it comes to approaching the reasoning capacity of a human brain, or even an insect brain.

There is no danger of a "rise of the machines" scenario with this technology, and while I feel a little bad for having to say it, people like Jacob Coxon are very much serving as "useful idiots" for American AI companies. They're taking away all the wrong messages and all the wrong dangers, and accidentally or otherwise, inflating the technology's perceived scope/importance by spreading a false narrative around its potential dangers. The only way this parrot can endanger humanity is if someone says "let's put this chaotic token predictor in a position where its output can kill people." Are people stupid enough to do that? Well, yes, of course they are. Are they stupid enough to do it at scale? Maybe, but the danger posed here is not by AI, it's by people, same as it's always been.

The dangers posed by this technology on its present trajectory are best understood as simple accelerants for the existing dangers humans pose to themselves. Engineering a deadly new pathogen, for example, is already a danger - the barrier to entry is mostly in the hardware required, so being able to rapidly auto-generate and/or analyze variants isn't really adding much to the danger, it's just an assistive mechanism for state-level actors already engaging in this dangerous practice. Mind you, hardware availability/scalability in this realm stands a very good chance of indirectly resulting in our extinction in the not too distant future, but AI anywhere on its present trajectory is not likely to change that much.

This paradigm is apparent in pretty much every category of "species-level" danger posed by the current generation of AI technology and its trajectory, it is at most an accelerant for an existing "misalignment" in the species. Our global socioeconomic order fails to align to our species-level needs and embraces a ruthless, sociopathic ethos which will almost certainly result in a large percentage of the global population dying horrible, premature deaths at some point in the not too distant future. AI has nothing to do with that, it's just helping speed it up a bit. AI is also helping authoritarian regimes inflict various forms of fascism in ways Nazi Germany only dreamed of, which poses a far more imminent danger to our collective quality of life than anything I've heard uttered in American media.

So why are many of the very same sociopaths who are driving us toward climate destruction now suddenly so concerned about the "danger" posed by AI, and seemingly acting in defiance of their fiduciary duties to warn us that we must slow down development and seek regulation? Of course the reasons are entirely self-serving, which you should almost exclusively be able to determine from their history of narratives! They're combining this "threat to the species" narrative with a "we must stay ahead of China for the sake of our babies and all of Western Civilization" (this is incredibly ironic, I'll explain why in a bit) narrative, while milking the shit out of every perceived accomplishment (especially their highly publicized and highly distorted mathematical "victories"); it's important to understand that these narratives are directly linked to each other, in order to understand the reasons behind them.

First you must realize that these companies have pulled off the greatest theft in all of human history. They've stolen every digitizable thing they can get their hands on with little to no respect for license or ownership (except where already forced by large-enough entities) under the assumption that they would be important enough for governments to save them from their crimes after the fact. They are completely dependent on the narrative that they've developed a revolutionary technology. Is a sufficiently-augmented parrot truly revolutionary? In some ways, in some industries, kind of. It's definitely going to make a lot of jobs more annoying and less satisfying in the coming decades. Will its economic gains offset its economic costs? Debatable, especially when you factor in the yet-to-be-realized (but predictable) consequences on future generations of labor. In fact, the current financial models behind these companies are unlikely to even be sustainable unless the companies succeed in their attempts at regulatory capture. Which brings up another important point.

Another reason behind the desperate attempts to inflate the publicly perceived importance of AI is regulatory capture. These companies want to carve out as many exceptions for themselves as possible and quickly pull the ladder up. They can perform mass-theft, but no one else can! I also expect these companies to begin behind-the-scenes attempts to bury/hinder open-weight models, if that isn't already happening. Some of the dynamics at play here are very similar to the dynamics behind software engineering and Open Source which I outlined not too long ago, and in fact AI is also accelerating many of the existing harms in that ecosystem. Being open is only a good idea for a sociopathic entity (e.g. a corporation) for as long as it profits more by the devaluation of associated labor than it profits from the labor itself. By contrast, it's hilarious to see these companies internally panic about the practice of distillation; oh no, they are stealing from us, after we stole from everyone! For that reason, these companies have been very careful about how that panic is presented to the public, because they are very aware of the irony inherent in their objection to having their data "stolen".

Now in order to protect their own interests (and I can't say it enough, those interests are all derived from the greatest theft in history) and to cover their own crimes, these companies have every incentive to try to reinvent copyright law. They have to contend with other big players in the American economy here, because some very powerful companies are very vested in existing copyright law. This is where you start to taste the irony in that "we must defeat China" narrative. These companies have already stepped all over individual property rights. For example, even some of the most permissive source code licenses tend to carry some loose terms and requirements, or simply request that authors receive attribution in derived works. All of that source code has been slurped up with no shits given. Every token from every given work and every weight that token influences (and every weight that weight influences, and so on) should reasonably be considered derivative of the work, and should be required to carry forth the terms of the license. By that logic, virtually every output from any of these LLM's is completely illegal.

Of course, the industry is appealing to pragmatism where it suits them here; if your source code is public, you should want it to be used for the "common good", and of course that's what this is! It's not consolidating the vast power represented by decades of human labor into the hands of a small number of people! IT ISN'T! Oh, and it's not a derivative work, because like, people can read that source code and the license doesn't transfer to their neurons! That's totally exactly the same thing! YOU KNOW IT IS! IT'S THE SAME THING! I'M TELLING YOU IT IS, NO, SHUT UP. Of course, there is a fun parallel there: the incredibly disingenuous practice of paying an intern to "rewrite" (e.g. do everything short of copy-paste) an open-source library because of incompatible licensing isn't exactly unheard of across various software industries, and AI is a convenient means of fully automating that theft. No intern necessary! So you have a system built on theft which also enables more forms of theft, which is funny, but those sub-thefts barely register in contrast to the primary theft.

Maybe at some point these companies will get around to revising a "permission-only" data set, but they aren't too incentivized to do that unless they lose their bid to reshape the entire western concept of intellectual property. That irony should be very salty now; companies pulling off mass theft, protecting themselves by leveraging western notions of property and ownership, then (by necessity of their theft) attempting to invalidate those notions of ownership for everyone else... except for other companies large enough to fight to preserve their own rights. This is effectively mandating a "state ownership" model like they warn you about in those "communism is evil" propaganda pieces, except the state is just a collection of powerful-enough corporations. Like some kind of corporate oligarchy. Gosh, where have I heard that term before?

Another big reason for the "AI will kill us" narrative is that these companies can't deliver on their promises, at least not on their present trajectory. They've had to over-promise in order to over-inflate the perception of the thing they're making. This links directly to their economic sustainability, and will soon enough link to the promises they make to public investors. Regulatory "slowdown" is the perfect narrative for investors. The narrative turns from "we aren't under-delivering because we're actually running out of new ideas for cyborg parrot attachments, we promise" to "we had to slow down for the good of the species, but the wait will be worth it, we promise." Perfect!

Anyway, capitalism being what it is, these companies are desperate to justify their business model and their valuations. Those valuations being what they are, at this point it seems to me that these companies are either going to "succeed" or scrape a chunk so big out of American GDP that its economic fluids can be felt spattering all the way over in China. The Chinese government has to think this whole situation, and the western hand-waving over "AI-driven annihilation", is fucking hilarious; yet they understand the reasons to avoid piercing the narrative. So we already have a system of very perverse incentives in place for the American government to embrace all of these lies about what AI is, what it can be, what it needs to be, and so on.

So let's summarize a bit here. They've stolen from our entire species on one of the deepest levels imaginable. They appear to have repeatedly misled the public and embraced false narratives about what they've done and what they're intending to do. They're desperate to avoid accountability and will reach for any possible deception in order to do so. It's easy to demonstrate that they're primarily serving private interests. They've trapped us in a lose-lose proposition for western values on the whole; this applies to countless social foundations, from property ownership to democracy. Their longer-term strategy is something I'd like to describe as a form of passive genocide. Oh right, let me get to that next!

So, what if this technology really delivers as promised? What if a nuclear-armed cyborg parrot is sufficient to replace almost all white-collar labor? Well, that is actually the idea after all. It does currently, even in its limited capacity, have the potential to reduce/remove plenty of white-collar labor. The realistic idea is that over the next few decades, the promise of AI along its present trajectory will deliver "good enough" results to eliminate large portions of the white-collar labor pool, meanwhile robotics catches up and becomes more viable/cost-effective at scale to replace more of the blue-collar labor. You might have noticed that the US is going out of its way to over-fund prisons, under-fund education, and re-inflate birth rates. Let's not draw any hard conclusions here, but let's speculate: you might do such a thing if you anticipated more of a need for feckless prison-labor meat sacks in 30 years, rather than a well-educated working middle class.

Now let's go back from speculation to conclusion: we know how industrial revolution works. We know that productivity wins don't go to workers, and that living standards of the average working-class citizen aren't truly commensurate (you'd have to really stretch) to any factor including or in between "surplus labor" and "GDP". Traditionally, living standards go up when the working class can muster enough leverage to demand it. If only escaping and dispersing into barbarian tribes were still an option, oh sweet lord if only. So if this form of AI does deliver and manages to spark a real industrial revolution, what happens to the working class in a society where almost all essential labor has been automated, where that entire class no longer has any leverage at all?

It's a true utopia where everyone has their needs met and no one has to work anymore unless they want to! THAT'S WHAT IT IS, NO, SHUT UP, I'M TELLING YOU THAT'S WHAT IT IS.

No, of course, you can kind of put things together here. You may have noticed that some scientists have been jumping around and screaming about climate change for some decades now, while politicians on the left promise to address those concerns and politicians on the right promise those concerns are overblown. In reality, everyone on every side (except for the genuinely clueless people) understands those concerns are valid. What is less commonly known, but is indeed the quiet part out loud, is that the timeline for climate change to start mattering lines up really nicely with the timeline for human labor to stop mattering quite as much. The hope of the Elons and Sams of the world is that this will come to fruition well enough that we only have to bother sustaining the "cream of the crop" in the labor force, who will generally be occupied full-time by stepping all over each other to keep their heads above water.

So basically the plan here is to let things go to shit, and when billions of people die as a natural consequence of that, things will keep running as desired. This may seem like crazy talk, we couldn't possibly have a world order in place which is monstrous enough to just go along on such a trajectory, could we?! BILLIONS?! But, well, I hate to break it to you, we've got enough well-charted markers on this one that we don't have to consider it conspiracy territory anymore. We crossed over from "that guy warning us about that stuff is fuckin' CRAZY" to "oh yeah that shit is almost certainly happening, whoops" territory a while ago, maybe a good decade or two depending on which markers you're willing to put trust in. I do say "almost certainly" because nothing forecasted on such a level can be considered a certainty. In some ways it's far more impossible to predict than weather, but in some ways it's more reliable; many socioeconomic systems are designed to maintain predictability in chaos. Nonetheless, anything could happen to blow it all up. But I do feel pretty confident that, despite narrative attempts to the contrary, this is the prevailing attitude and the idea of "the way forward" being implemented by the wealthiest assholes in our society.

The counter-narratives these companies can launch in the face of every point I've made here are innumerable. I could set LLM's arguing against themselves over any one of the points I've made here and generate a tome. Almost no one wants to read as much as you'd have to read, to understand as much about our root civilization as you'd have to understand, to get to the root of what's really going on here. Very few people would even be capable. Trying to sway people to my conclusions as a means of political activism is an effective impossibility, by design of our mechanisms for the dissemination of information. Thus my hopelessness, but I'm still compelled to leave my protest and forecasts here.

The debate at this point should not be so much "should I use AI, or should I abstain?" There are plenty of ethical reasons to avoid it. There are plenty of practical reasons to use it. Not everyone has the ability to choose, and many employers are inflicting it on their employees whether those employees like it or not. At this point, my only reasonable wish is for as many people as possible, whether they choose to use it or not, to realize what it really is and how it has been brought into the world. You can use a product of terrible evil without hating yourself and without abandoning your knowledge of the evil involved; I do it every day by participating in the global supply chain of food products. Do what you must do to stay alive, but try to maintain awareness and help others to do the same.

You could consider this an opinion piece, but I think of it more as a conclusion piece. I've provided the conclusions, you can take my word for it (I generally advise against doing this for anyone on the Internet, except for me of course, my conclusions are definitely sound), you can follow the trail of bread crumbs backward for weeks and see if you survive the rabbit-gauntlet without being brought down by a conspiracy theorist, or you can look to people who are more vocal and probably have more patience for talking through bullshit than I do and who might have reached similar conclusions on some of the topics covered here.

This was crudely written in under 2 hours without the assistance, analysis, or hardening of an LLM, and with no editorial treatment whatsoever. I think this is how everyone should be writing in an age of shitty, sterilized parrot regurgitation. Now, let us enjoy the end, however it may come.