{"content":"\u003cp\u003eI remember the exact moment I became disillusioned with the tech industry.\u003c/p\u003e\n\u003cp\u003eI was sitting in a meeting with our design team. It must have been sometime around the fall of 2019. After working at the company for over a year, we were finally doing focus group testing. I felt strongly (and still do) that observing users\u0026rsquo; real world behavior was essential to good design; and as a developer, I often felt removed from the people who used the software I wrote by an opaque bureaucracy, not knowing if that bureaucracy even bothered to ask its users what they wanted. I had to settle for hearing our user\u0026rsquo;s thoughts second-hand; still, I was thrilled that we had actually spent the time and money on what every company said was most important.\u003c/p\u003e\n\u003cp\u003eAs far as tech companies go, they were hardly evil: they didn\u0026rsquo;t make weapons; they didn\u0026rsquo;t make software to figure out who to point the weapons at; they didn\u0026rsquo;t even provide logistical support to the people who point the weapons. Not saving the world, but not setting it on fire either. They were, as they put it, in the business of \u0026ldquo;Customer Experience Management\u0026rdquo;. Their core product was a SaaS application which ingested all the different surveys customers fill out, from comment cards in hotel rooms, to google reviews, to the link at the bottom of your email receipt (I was told they sent emails to literally a billion people; you probably have one sitting in your inbox right now). They then did a bunch of natural lanuage analysis and statistical transformations to figure out the right action to take to \u0026ldquo;close the loop\u0026rdquo; with that customer. If they were happy, \u0026ldquo;closing the loop\u0026rdquo; might involve sending them a short email thanking them for their business. If they were unhappy, it might mean leaving an apologetic reply on a google review. If they were \u003cem\u003every\u003c/em\u003e unhappy (and a big enough spender), it might mean the manager of the hotel personally calling them to grovel and offer some free perks.\u003c/p\u003e\n\u003cp\u003eLike all enterprise software products, it had its quirks and deficiencies, but mostly it worked, and worked well for an industry with very tight margins, one that needs to cut costs anywhere it can without making the experience feel cheap: hospitality. At the time, the company had the bulk of the largest hotel chains for its customers, and it kept them year after year because the product improved customer loyalty in a directly measurable way.\u003c/p\u003e\n\u003cp\u003eThe company\u0026rsquo;s focus was on their \u003cem\u003ecustomers\u0026rsquo;\u003c/em\u003e customers; figuring out what they wanted, why they were dissatisfied, what would keep them coming back. But those customers rarely, if ever, touched our actual product; if it was working well, they, like you, probably didn\u0026rsquo;t know it existed. The people who \u003cem\u003eactually\u003c/em\u003e used the app were our customers\u0026rsquo; \u003cem\u003eemployees\u003c/em\u003e, who\u0026rsquo;s continued employment depended on making the customer happy, sometimes even (we hoped) by doing what our tool recommended.\u003c/p\u003e\n\u003cp\u003eAs a frontend engineer, what \u003cem\u003eI\u003c/em\u003e cared about was the user of \u003cem\u003emy\u003c/em\u003e app. Did it load fast? Was it intuitive? Did they use it how we imagined they would, or did they want something completely different? And finally, after a year of trying to learn everything I could about what our \u003cem\u003eusers\u003c/em\u003e, not our \u003cem\u003ecustomers\u003c/em\u003e, wanted from the app, I was finding out how they \u003cem\u003eactually\u003c/em\u003e used it (albeit secondhand).\u003c/p\u003e\n\u003cp\u003eIt was only an offhand comment, an incidental detail to everyone else in the room; but for me it completely changed the stakes of what I was building.\u003c/p\u003e\n\u003cp\u003eOne of the features of the app that made it so \u0026ldquo;sticky\u0026rdquo; (industry jargon for making the product indispensable) was the ability to make specific employees accountable to \u0026ldquo;close the loop\u0026rdquo; (\u003cem\u003eread: grovel and beg for forgiveness\u003c/em\u003e) with an unhappy customer. Theoretically, this was supposed to be the person who had let the customer down in the first place, often determined by natural language processing and other statistical trickery. I learned in that conversation that it was common for our customers to tie \u003cem\u003ebonuses\u003c/em\u003e to how often they were able to turn a bad score into a good one, and that having too many bad scores to your name often meant termination.\u003c/p\u003e\n\u003cp\u003eMy mother worked in a hotel. She made minimum wage ($7.25 in Pennsylvania, then as now), worked long hours, and was treated like trash by people who felt entitled to view any service worker as their personal whipping boy. She came home exhausted every day, and her chronic health issues were the worst I ever remember them being. Hotel work is some of the hardest, worst paid, and most demoralizing work there is. Those bonuses are a significant part of the money they make.\u003c/p\u003e\n\u003cp\u003eAnd there I was, working hard but not egregeously so, making \u003cem\u003e10 times\u003c/em\u003e what my mother used to make, to write software which made it easier for managers to squeeze more value, more indignity from their already overworked and degraded employees.\u003c/p\u003e\n\u003cp\u003eAll I could think was: \u003cem\u003ewhen will they figure out how to do this to me?\u003c/em\u003e\u003c/p\u003e\n\u003ch1 id=\"software-engineer-as-a-service\"\u003eSoftware Engineer as a Service\u003c/h1\u003e\n\u003cp\u003eI would work other tech jobs, but my heart wasn\u0026rsquo;t in it anymore. When tech work shifted remote in 2020, I took advantage of the flexibity and the massively overinflated pay to do some soul searching. I wandered around, getting involved in online debates and offline struggles, trying to answer the question: why me? Am I really so smart, is my labor really so valuable as to deserve making twice the median income in the richest country in the world (and many times that when you factor in stock awards)? Why did I make it out of the post-industiral rural poverty death spiral?\u003c/p\u003e\n\u003cp\u003eFast forward to 2026. I have fully left the tech industry; I work in a soup kitchen. My exit from the industry, which once felt like a principled choice, is starting to feel like I simply fell behind. AI has transformed the industry, and continues to do so at an unbelievable pace. My leftist friends are largely dismissive toward and ignorant of everything AI, seeing it as a last ditch effort to prop up a dying social order; while my tech friends have either transformed their entire way of doing their job, or are holding down their place in what they believe to be a temporary eddy in the coming tide of agentic chaos.\u003c/p\u003e\n\u003cp\u003eI think that the reality is more complicated than either. But one thing seems clear to me: whatever happened between 2015 and 2020 that allowed me to go from being a homeschooled, ex-religious shut-in with no social skills or will to live, to a Marxist anarchist who just so happens to own a house and a car, is over. I can still program, a fact I have to remind myself of from time to time with the occasional all-night coding binge; but I am no longer willing or able to work in the way that is required to keep up with the direction the industry is heading.\u003c/p\u003e\n\u003cp\u003eAt the end of this post, I will provide some resources I have found helpful for understanding the present state of AI; resources which are neither purely hype nor purely dismissive. For now, I\u0026rsquo;d like to use a concept from Marx\u0026rsquo;s Capital to articulate what I believe is currently happening to the software industry. Such a dramatic reduction in the value of the labor power of a class of workers, value which they formerly enjoyed almost exclusively out of the mass of workers, demands explanation. According to the logic of the stock market, all is explained by the creation of a massive pile of new wealth in the form of AI. However, I argue it is possible both for this valuation to be grossly inflated, and for AI to have brought about a real and permanent change in the economic situation.\u003c/p\u003e\n\u003ch1 id=\"machine-slop\"\u003eMachine Slop\u003c/h1\u003e\n\u003cp\u003eCompanies that make software typically spend, as a percentage of their revenue, surprisingly little on paying people to write code. The common benchmark is to spend 20-30% of revenue on \u0026ldquo;R\u0026amp;D\u0026rdquo;, the largest piece of the pie usually being sales. But even of that 20%, only a fraction has to do directly with code. R\u0026amp;D includes designers, product managers, and technical writers, among many other roles; and even of the actual engineers, around a fifth are managers who write little or no code. The engineering and product managers together spend nearly all their time trying to do just two things: to articulate clearly and accurately \u003cem\u003ewhat the software will do\u003c/em\u003e, and to predict \u003cem\u003ewhen it will be done\u003c/em\u003e. Even those who\u0026rsquo;s job is theoretically to write code spend hours out of every week in scrum meetings, sprint planning meetings, and retrospectives, trying to account for their time and to predict how long \u003cem\u003ejust their next few tasks\u003c/em\u003e will take. Software engineers, managers, and even expert consultants who\u0026rsquo;s entire job it is to predict developer productivity are constantly and hilariously wrong about how long a given task will take. There is an entire industry of books, seminars, and conferences related to predicting and controlling the productivity of software development, with competing and contradictory schools of thought about how to approach the problem. And this unpredictability has, in recent decades, been a source of bargaining power for the engineers who have found themselves in the position of being able to make untold amounts of value appear out of thin air.\u003c/p\u003e\n\u003cp\u003eNeedless to say, this is an intolerable situation for a capitalist, who wants the value generated by an enterprise to be predictable, large, and, to the extent that it \u003cem\u003emust\u003c/em\u003e involve human labor, for that labor to be simple enough that almost anyone could do it.\u003c/p\u003e\n\u003cp\u003eThe tendency of modern industrial societies has been to transform the labor process in several key ways:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003eMassively increasing the scale of production (this reduces cost, as well as allowing precise supervision and control of the labor process)\u003c/li\u003e\n\u003cli\u003eMaking human labor an increasingly simple and mechanical component of the process, so that a less skilled person can be used, increasing the pool of candidates and reducing the wage\u003c/li\u003e\n\u003cli\u003eConcentrating the means of production (tools, raw materials, knowledge) under the ownership of a smaller and smaller number of people, the better to control and tune every aspect of the process. The (unintended but advantageous) result being that ever fewer people in society own what they need in order to produce useful goods, and must work for a capitalist\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eAs a result, a ever-greater portion of the working population tends to become \u0026ldquo;de-skilled\u0026rdquo;, with work that used to require a skilled craftsman now being done by, on the one hand, a team of industrial engineers who design and continuously improve a large-scale, machine-driven process; and on the other, workers who perform the same repeatable tasks as subordinate components of that process, and neither could nor are allowed to control the process in which they are immersed.\u003c/p\u003e\n\u003cp\u003eThe classic example of de-skilling is that of machinists. It was in battles with machinists over control of the labor process that Frederick Winslow Taylor, of Taylorism fame, developed his theories of scientific management which have fundamentally reshaped all industrial labor processes. At first, he made no headway, because the machinists, being in posession of the knowledge and skill of their craft, had all the bargaining power; they banded together to work at the pace they wished to, and there was little management could do. As Taylor himself put it:\u003c/p\u003e\n\u003cblockquote\u003e\n\u003cp\u003e\u0026ldquo;[T]he shop was really run by the workmen and not by the bosses. The workmen together had carefully planned just how fast each job should be done.\u0026rdquo; In his [battle with the Midvale machinists], Taylor pointed out, he had located the source of the trouble in the \u0026ldquo;ignorance of the management as to what really constitutes a proper day\u0026rsquo;s work for a workman.\u0026rdquo; He had \u0026ldquo;fully realized that, although he was foreman of the shop, the combined knowledge and skill of the workmen who were under him was certainly ten times as great as his own.\u0026rdquo; This, then, was the source of the trouble and the starting point of scientific management.\u003csup\u003e\u003ca class=\"cite\" href=\"#cite-harry-braverman\"\u003ecite\u003c/a\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/blockquote\u003e\n\u003cp\u003eHis efforts to induce them to work faster by training up more desperate competitors, as well as various piece-rate incentives, all failed. The workers knew more about the work than management ever could; it was simply too complicated. It was only by separating the \u003cem\u003econception\u003c/em\u003e of the work from its \u003cem\u003eexecution\u003c/em\u003e that Taylor was able to break their hold on the process:\u003c/p\u003e\n\u003cblockquote\u003e\n\u003cp\u003eThe machinist of Taylor\u0026rsquo;s day started with the shop drawing, and turned, milled, bored, drilled, planed, shaped, ground, filed, and otherwise machine and hand-processed the proper stock to the desired shape as specified in the drawing. The range of decisions to be made in the course of the process is—unlike the case of a simple job, such as the handling of pig iron—by its very nature enormous. Even for the lathe alone, disregarding all collateral tasks such as the choice of stock, handling, centering and chucking the work, layout and measuring, order of cuts, and considering only the operation of turning itself, the range of possibilities is huge. Taylor himself worked with twelve variables, including the hardness of the metal, the material of the cutting tool, the thickness of the shaving, the shape of the cutting tool, the use of a coolant during cutting, the depth of the cut, the frequency of regrinding cutting tools as they became dulled, the lip and clearance angles of the tool, the smoothness of cutting or absence of chatter, the diameter of the stock being turned, the pressure of the chip or shaving on the cutting surface of the tool, and the speeds, feeds, and pulling power of the machine. Each of these variables is subject to broad choice, ranging from a few possibilities in the selection and use of a coolant, to a very great number of effective choices in all matters having to do with thickness, shape, depth, duration, speed, etc. Twelve variables, each subject to a large number of choices, will yield in their possible combinations and permutations astronomical figures, as Taylor soon realized. But upon these decisions of the machinist depended not just the accuracy and finish of the product, but also the pace of production.\u003c/p\u003e\n\u003cp\u003eNothing daunted, Taylor set out to gather into management\u0026rsquo;s hands all the basic information bearing on these processes. He began a series of experiments at the Midvale Steel Company, in the fall of 1880, which lasted twenty-six years, recording the results of between 30,000 and 50,000 tests, and cutting up more than 800,000 pounds of iron and steel on ten different machine tools reserved for his experimental use. His greatest difficulty, he reported, was not testing the many variations, but holding eleven variables constant while altering the conditions of the twelfth. The data were systematized, correlated, and reduced to practical form in the shape of what he called a \u0026ldquo;slide rule\u0026rdquo; which would determine the optimum combination of choices for each step in the machining process. His machinists thenceforth were required to work in accordance with instructions derived from these experimental data, rather than from their own knowledge, experience, or tradition.\u003c/p\u003e\u003c/blockquote\u003e\n\u003cp\u003eWhat justifies this de-skilling, this concentration of the means of production in the hands of the managers of capital? Naturally the increase in productivity, and therefore the increase in the total wealth of society; combined with the fact that only capitalists could undertake such massive investments into deconstructing the labor process in exacting detail. And yet, as this process takes off, it becomes a vortex into which all labor is sucked, each industry reducing its craftsmen to components of a system, a system which they cannot earn their living without.\u003csup\u003e\u003ca class=\"cite\" href=\"#cite-service-work\"\u003enote\u003c/a\u003e\u003c/sup\u003e Those who wish to be involved in the conceptualization of the labor process must specialize as engineers, a specialty fewer and fewer can make inroads into, \u003cem\u003eby the very logic of the process itself:\u003c/em\u003e if a change in the labor process created as many highly specialized, highly skilled, highly paid jobs as it cut in more generalized (though not necessarily less skilled) jobs, it would have \u003cem\u003ethe exact opposite\u003c/em\u003e effect of \u003cem\u003eincreasing\u003c/em\u003e the cost of the labor process.\u003c/p\u003e\n\u003ch1 id=\"dead-intellectual-labor\"\u003eDead (Intellectual) Labor\u003c/h1\u003e\n\u003cp\u003eWhat made software engineering unique among crafts until fairly recently was that it required almost no capital; all that is required is a laptop, an internet connection, and the desire and capacity to learn. Hardware has been, since the 80s, cheap and was, until recently, getting cheaper all the time. Relative to the initial investment, the amount of value that can be created is astronomical. But where does this value come from? Setting aside tech firms that supply a tangible use value, like solar panels or hydrocells, how is it that AI firms can justify valuations in excess of companies with both more obviously useful and more immanently profitable lines of business, all by providing a purely intangible good: \u003cem\u003eintelligence\u003c/em\u003e?\u003c/p\u003e\n\u003cp\u003eAs a capitalist, your success or failure depends on your ability to increase the speed and predictability of your production process, while at the same time, if possible, reducing its cost. One of the most unreliable inputs into in the production process is, of course, human beings: besides having dreams, desires, and feelings, it tends to be considered unethical to knowingly cause them harm. Oh yeah, and the one thing they definitely \u003cem\u003edon\u0026rsquo;t\u003c/em\u003e want to do is \u003cem\u003ework\u003c/em\u003e, since it\u0026rsquo;s kind of built into the definition of work that it\u0026rsquo;s something you \u003cem\u003ehave\u003c/em\u003e to do, not something you \u003cem\u003ewant\u003c/em\u003e to do.\u003csup\u003e\u003ca class=\"cite\" href=\"#cite-marx-activity\"\u003enote\u003c/a\u003e\u003c/sup\u003e\u003c/p\u003e\n\u003cp\u003eMarx\u0026rsquo;s theory of value is premised on the idea that you can\u0026rsquo;t make a consistent, long-term profit by trading commodities for other commodities; you\u0026rsquo;ll just break even in the long run, no matter how good at arbitrage you think you are. And since the one thing everyone doesn\u0026rsquo;t want to do is work, time spent doing boring and dangerous work becomes the thing that possesses social value. If lugnuts or diamonds were abundant and easy to acquire, they would have no market value; it\u0026rsquo;s the fact that someone has to risk life and limb to find or make them that gives them value. And the profit, he argues, of a capitalist enterprise is achieved by constantly expanding the scale and efficiency of the production process beyond the socially average amount of labor required to produce that thing. Each individual capital exists on a spectrum around that average, some below it and getting fabulously rich, some above it and slowly going under; and only by the constant movement toward the frontiers of efficiency does any individual capitalist turn a profit over the long run.\u003c/p\u003e\n\u003cp\u003eThe inevitable (and unintended) side effect of each Capitalist\u0026rsquo;s incentive to make the process ever faster, more predictable, and cheaper, is that human labor (what Marx calls \u003cem\u003evariable capital\u003c/em\u003e because the value it contributes to the labor process varies greatly and decisively between capitals) becomes an ever smaller part of the production process. Increasingly, a greater share of the labor process is made up of \u003cem\u003econstant capital\u003c/em\u003e: machines, warehouses, shipping containers, technical documents, objects which make up the sum total of the wealth in society, but which cannot themselves beget new wealth without coming into contact with human labor.\u003c/p\u003e\n\u003cp\u003eOver the course of industrial society\u0026rsquo;s development, we have witnessed the slow concretization of human skill and knowledge into a system of objects, systems which increasingly interface not with us but with each other. The science fiction of the 20th century imagined that the objects would handle all the boring and dangerous work, reserving for us the part of \u003cem\u003eintelligence\u003c/em\u003e: every citizen an educated, confident, lesuirely conductor of the orchestra of industry. While our collective disregard for the value of human life allows it to remain cheaper to consume human bodies in dangerous work than it is to automate it, the capitalists have seemingly found a way to automate the parts of the process that were supposed to be the preserve of the educated classes in the imperial core. AI in its current form seems best positioned to replace \u0026ldquo;bullshit\u0026rdquo; office jobs, leaving the educated classes to slip further down the ladder of so-called \u0026ldquo;skilled\u0026rdquo; labor.\u003c/p\u003e\n\u003cp\u003eSo this is what AI firms offer: the (mostly theoretical at this point, but increasingly real) ability to replace the army of managers, programmers, technical writers, and designers, along with the entire world of office work with a \u003cem\u003emuch smaller\u003c/em\u003e army of those same workers, babysitting AIs. The unpredictability and opacity of the bureaucratic labyrinth replaced with a tap labeled \u003cem\u003eintelligence\u003c/em\u003e, which can be opened or shut at will. How \u003cem\u003ewell\u003c/em\u003e this will work remains to be seen; but it has already dramatically eroded the bargaining power of at least one segment of the working classes who enjoyed, for a time, a relatively cushy, though insecure, position in the class hierarchy\u003csup\u003e\u003ca class=\"cite\" href=\"#cite-engineer-unemployment\"\u003ecite\u003c/a\u003e\u003c/sup\u003e. The day I was sure would arrive sooner or later seems to have arrived sooner.\u003c/p\u003e\n\u003ch1 id=\"the-view-from-alongside-the-torrent\"\u003eThe View From Alongside The Torrent\u003c/h1\u003e\n\u003cp\u003eI think accepting the premise that intelligence as a service is a novel creation of the AI boom is ceding too much to the hype. It seems to me that the difference between being able to access all human knowledge through a search engine and accessing it through a LLM is a matter of degree, not of kind. Plato, coming from an oral-rhetorical tradition, bemoaned that writing would destroy our capacity to remember. The process of externalizing our memory into physical objects has been a long and slow one, and we had already given up more of our minds\u0026rsquo; and our bodys\u0026rsquo; capacity than we can now imagine before the first computer was ever switched on.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAnd yet.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eEnough difference in degree, you cross some threshold and find yourself neck deep in a difference in kind.\u003c/p\u003e\n\u003cp\u003eI find myself in a dramatically different world than the one I grew up in. A decade ago, I had this innate confidence that whatever happened, if I could just force myself to sit in front of the keyboard long enough, I could make value come out of my fingertips. The world I find myself in now, the world of agentic AI, feels more similar to how most workers around the world have probably felt for hundreds of years now. How long until I can\u0026rsquo;t keep up with the pace expected of AI-enabled engineers by clickey-claking away, no matter how good my VIM shortcuts are? And if I did embrace AI, what then? I\u0026rsquo;d be dependent on access to Claude for the rest of my career? Watching my skills atrophie until I\u0026rsquo;m just a babysitter for a bunch of choleric, schizoid intelligences babbling to themselves, puking out petabytes of code I\u0026rsquo;m not paid enough to give a crap about?\u003c/p\u003e\n\u003cp\u003eSure, in this venture capital subsidized moment, I can have a Claude Code subscription for ~$100/month. But that can\u0026rsquo;t be anywhere \u003cem\u003enear\u003c/em\u003e the stable market price for these frontier models, and already I know people who spend a significant fraction of their salary at work on AI compute hours. Even at current prices, 14-year-old me could not have afforded a Claude Code subscription. Nor can I imagine I would have become very good at coding if I had; I would have simply become dependent on Claude to make computers do anything interesting, the same way my brain is dependent on Google to remember basic information now, only moreso.\u003c/p\u003e\n\u003cp\u003eFrom where I\u0026rsquo;m sitting right now, it seems like the beginning of the end of the special little segment of the working classes who, for a brief flash of historical time, were imbued with a lot of economic mobility. But a lot remains to be seen; things continue to change quickly. It might be that AI is so expensive and energy intensive that our wiggly old meat sacks are more efficient in the long run. It might be that the market will collapse as it did in the DotCom bubble, only to come back in a form more aligned with economic reality, and no less devastating to our brains and our bargaining power for all that.\u003c/p\u003e\n\u003cp\u003eBut for myself, I doubt programming will ever be more than a hobby for me again. I resisted the push toward management over my entire 10-year career. Why would I want to go back to the industry now, when the job seems to consist of managing a team of resource-guzzling, corpo-speech-regurgitating interns?\u003c/p\u003e\n\u003ch1 id=\"next-time\"\u003eNext Time\u003c/h1\u003e\n\u003cp\u003eSince this post is getting long, in part 2 I will get into more detail about why I think leftists have misunderstood the threat posed by AI, both taking it too seriously and not seriously enough. For now, I\u0026rsquo;ll provide some links to resources I think clarify the stakes of the trend toward generative AI:\u003c/p\u003e\n\u003ch1 id=\"further-reading\"\u003eFurther Reading\u003c/h1\u003e\n\u003cul\u003e\n\u003cli\u003e[Read] \u003ca href=\"https://philneel.substack.com/p/lesser-gods-labor-in-the-ai-labyrinth\"\u003eLesser Gods: Labor in the AI Labyrinth, Phil A. Neel\u003c/a\u003e Paywalled, but completely worth it. He\u0026rsquo;s working on a book-length treatment. This is the main source of my critique of the left\u0026rsquo;s misunderstanding of AI\u003c/li\u003e\n\u003cli\u003e[Watch] \u003ca href=\"https://youtu.be/RJyPVLMyyuA?si=eRilbdrZiN7IsfUb\"\u003eWading through AI, Casey Muratori, Demetri Spanos\u003c/a\u003e A programmer who refuses to use AI interviews a programmer who has worked in the field since before the current bubble about the difference between the current hype cycle and the actual technology, the ethical issues with it, and more.\u003c/li\u003e\n\u003cli\u003e[Watch] \u003ca href=\"https://vimeo.com/871803593\"\u003eAnimal Linguistics: Origins and Evolution of Language, Michele Fournet\u003c/a\u003e An example of a narrowly constrained problem using neural networks to do something really cool and novel, understanding whale song!!\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch1 id=\"footnotes\"\u003eFootnotes\u003c/h1\u003e\n\u003col class=\"footnotes\"\u003e\n\u003cli id=\"cite-harry-braverman\"\u003e \n\u003cp\u003eHarry Braverman, \u003cem\u003eLabor and Monopoly Capital\u003c/em\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003cli id=\"cite-service-work\"\u003e \n\u003cp\u003eThis also explains why service work has grown to absorb the excess educated labor in post-industrial economies: having rationalized and globalized industrial production, the mass of over-educated workers can only find work providing an \u0026ldquo;experience\u0026rdquo; of artisanship to the recipients of the excess productivity thus generated.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli id=\"cite-marx-activity\"\u003e \n\u003cp\u003eMarx makes a distinction between \u003cem\u003eactivity\u003c/em\u003e, which is work we chose to do because we are able to bring our entire being, including our autonomy, to; and \u003cem\u003elabor\u003c/em\u003e, which is either boring or dangerous or both, and which we only do when we are forced to in order to survive.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli id=\"cite-engineer-unemployment\"\u003e \n\u003cp\u003e\u003ca href=\"https://www.investopedia.com/these-37-college-majors-reveal-unemployment-rates-higher-than-all-other-workers-11941188\"\u003eCurrently, Computer Engineering has one of the highest post-graduate unemployment levels of any degree, at 7.8%.\u003c/a\u003e\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ol\u003e\n","date":"2026-07-14T12:27:53-04:00","description":"on losing faith and falling behind","draft":false,"lastModified":"2026-07-14T12:27:53-04:00","path":"/posts/ten_ex_machina","title":"Ten Ex Machina: Artificial Intelligence And Constant Capital, Part 1","weight":0}