In Scientific Culture and the Making of the Industrial West Margaret Jacob argues that the Industrial Revolution started in Britain, as opposed to France, the Netherlands, or Germany, due to its culture welcoming or even encouraging tolerant public debate, mechanical and physical reasoning, experimentation, and entrepreneurship. Jacob repeatedly throws shade at “simplistic” or even “economic” historiographical assumptions that some combination of knowledge and capital are all you need for homo economicus to industrialize, as if following a textbook chemical reaction devoid of any motivated human beings. Culture, she argues, is significantly more important. This is kind of obvious if you have read Bitcoin Is Venice but it doesn’t hurt to be independently validated again …
This isn’t the kind of post in which I am inclined to quote Jacob at length. I recommend reading it if you are into the sort of thing just outlined. Still, I bring it up to introduce the context of the Industrial Revolution on the one hand to be returned to later in the post, but on the other hand to introduce the argument by singling out a point she makes in a handful of places: public promoters of novel scientific understanding and technical progress in Enlightenment Britain would go to great pains to emphasize that machines create work for the poor specifically as opposed to replacing it. Jacob even implies on occasion that this was sneaky rhetoric – verging even on propaganda – promoted in public discourse for the selfish purposes of the newly minted capitalist class.
Jacob may be performing a kind of apologetics around this sleight of hand on the part of figures she is mostly otherwise championing. But of course, in a purely economic analysis outside of the social and historical lens Jacob adopts, both could be true.
It depends on your perspective. From a static perspective, sure, steam-powered industrial machinery “does the job of 100 men”. But from a dynamic perspective, we must recognize that opportunity costs are real and that any job such a machine could do almost certainly couldn’t be economically done by 100 men prior to it existing, so the static comparison is irrelevant. It is only even conceivable after the fact. We retain linguistic traces of this once-obvious truth in how we talk about cars. A 200-horsepower car implies dramatic improvement, not equivalence, because it is not physically possible to perform the same tasks with 200 horses. There are not 200 horses obsoleted by the existence of a car, there just newly viable economic tasks.
And so, since nothing is ever new, we find ourselves in the midst of only the twentieth or thirtieth major technology revolution since the birth of industrial steam power, and yet another moral panic over “the end of work”. I am talking, of course, about LLMs. While “AI will do all the jobs” is both a stale, decades-old meme at this point and is so vague and general as to be difficult to debunk, this “AI” cycle has thrown up the neologism of “vibe coding”. I like vibe coding, as a concept but also as a standard-bearing meme, as it presents a kind of memetically powerful black humour to the effect that “the code is coding itself ermegheeeerd!” that is easier to pin down but also funnier to mock.
The AI doomers would have you believe that the ability to vibe code spells the end for coding as a profession, which recursively spells the end for all valuable work relying on human thought, and that all the time spent learning any skill, coding or otherwise, is now a sunk cost to be covered only by working at McDonalds, or playing video games on UBI, or whatever.
But fear not, for I am here to explain to you that this is extremely foolish. p(doom)=0, as the cool kids might say. This time isn’t different, and the reason this time isn’t different is the same as the reason every other time wasn’t different. Technology is always and everywhere a tool, not an autonomous agent. It requires humans to operate, which they will do in order to satisfy human desires. When we say technology “saves labor” what we really mean is it leverages labor. It gives to labor powers that were previously impossible – inconceivable, even. Only in the aftermath of witnessing its novel capabilities do we articulate that a car gives to a man the (literal, physical) power of 200 horses; we do not start by imagining squishing 200 horses into a metal box.
As Steve Jobs eloquently put it, “the computer is the bicycle of the mind,” to which I once less eloquently added,
“This is usually understood as commentary on technology, but I think it is more about capital. A less romantic framing, perhaps, than “economic potential energy,” capital is tools. Jobs’ point is just how potent software is as a form of capital. But ultimately all capital magnifies the desirability of the output of some exertion of time and energy. A bicycle is one example, and a computer is another. At the risk of mixing metaphors, capital is the bicycle of labor; of time, of effort, and of the grind to produce value by hand.”
Note, by the way, that bicycles do not ride themselves. Humans ride bicycles, and while they consume 3x fewer calories and go 3x as fast as running humans, they nonetheless do not replace 9 humans. This is so painfully obvious it might just justify all these confusing metaphors in allowing us to recognize that all capital ultimately rests on human capital. Technologies may obsolete other technologies, but they never obsolete people because people can learn to use them.
This may seem curious in the scheme of people losing their jobs because they are replaced, if not by AI, then by literally any major technological shift in history. Consider, for example, that in Britain in 1650 there were roughly 2.1 million agricultural workers whereas today there are more like 700 thousand. That’s a lot of unemployed farmers!
The trick here is once again a static versus dynamic view. 1850 is a much more relevant marker (rounding to 50 years) of the Industrial Revolution having taken hold in Britain. Consider, for example, that the growth in non-agricultural jobs even from 1800 to 1850 (never mind from 1650) exceeded the entire population at 1800! This was on the order of ten to one hundred times the gradual loss in jobs in the agricultural sector in the same period and was entirely the product of the same mechanical advances that were ever-so-slightly reducing the agricultural workforce. It is worth noting that none of this is to dispute the likes of Wendell Berry in arguing that agriculture shouldn't be valued. Conflating work with value is an unfortunate modernism that is outside the scope of this post to evaluate. To get back on track, we might offer that being freed from the necessity to work in agriculture allows, but does not require, the space and time to value it more deliberately and conscientiously than otherwise.
At the dawn of the steam engine, surveying the economic landscape with an eye to what might change, one sees pretty much nothing but agriculture. One imagines such improvements as pumping water for irrigation or driving mills. Eventually, the engines become portable and start to work on tasks like threshing. I tactically left out above, by the way, that the expression “horsepower” was coined by none other than James Watt as a way to articulate the capabilities of his new machine. So, in a sense, the origin of the phrase really was to communicate the ability to “replace horses”, but this only makes sense for agricultural or early industrial enterprises with perhaps four horses, or eight at a stretch.
But, of course, twenty-horsepower engines and above performed tasks that were unfathomable before their existence, and they created orders of magnitude more jobs than they destroyed. Even that is inexhaustive because it assumes a direct replacement as opposed to engines of merely one horsepower taking up less space or eating less hay and taking fewer shits than a horse. And so “horsepower” quickly becomes useless as a real equivalence and more like a term of awe (or a "unit of account" lolololol). To recap, then, yes, we do lose some jobs as machines leverage human labour. This is inevitable and can cause local hardship about which we shouldn't be callous.
But given humans are not obsolete machines but rather the malleable operators of an unbounded array of mechanical advances, we gain significantly more jobs than we ever lose, and we get significantly wealthier in aggregate the more capital we accumulate in this manner.
It’s worth noting a slightly different perspective on the statistics also: in Industrial Revolution Britain, the decline in absolute numbers was relatively gradual, especially when smoothed out over longer and longer periods, whereas the decline in the proportion of the overall population was much more dramatic, from roughly 65% in 1650 to less than 1% today. This is leverage. This is power. 64% of the population is now free to create things other than food, and more to the point, are perfectly capable of doing so! They are not unemployed but compounding away, creating orders of magnitude more wealth than you could even begin to explain to a seventeenth-century Brit.
Jumping ahead from the Industrial Revolution but not quite all the way to today just yet, I strongly, strongly encourage the reader to watch +this documentary on YouTube+ in which a 1982 British television crew was sent to Japan to report on the daunting prospect of … robots!
The video is unintentionally hilarious because while the entirety of the surface-level content can be reduced to “capital accumulation creates improved employment opportunities”, there is another layer of interpretation available in recognizing that the British narrator seems incapable of recognising what he is literally seeing with his own eyes and even accurately reporting on. Clearly profoundly pwned by socialist inclinations and demoralised by the malaise of British industry produced by exactly such inclinations in others around him, he acts as if he alone has the wisdom to predict that the entire situation is paradoxical and can only end in misery.
It is tempting to transcribe the entire video, but I will restrain myself and pull out some choice quotes:
“Japanese car manufacturers can afford to pay high wages and still undercut the competition. This one Nissan plant turns out nearly half a million cars a year, equivalent to 50% of the British car industry. It’s volume production like this that helps keep costs down. and every Nissan worker has at his elbow several times more capital equipment than is available to a British car worker.”
So far, so good. Immediately followed by the following interview snippet with a factory worker:
“What happens if all of this work is robotized?” “Even if robots were to be introduced, we would be given different jobs with better status, so I don't believe the introduction of robots would mean we would lose our jobs.”
This might seem like a positive message, but I’m telling you, it’s the way he says it that is comic gold. You just have to watch it. His tone is such that he may as well preface every remark with “you’ll never believe it, but …”. Patient enough viewers will get a good chuckle in discovering around the 20-minute mark that the only Japanese they could find to say anything negative about this whole situation was … wait for it … an economist! And clearly a Western-trained one at that.
So, even though the narrator is accurately reporting on the refutation of his own ridiculous beliefs, he can’t help but end the segment with (spoilers!):
“If once the robots become so numerous that they are regarded as rivals rather than colleagues, workers and management may find themselves pulling in opposite directions. And once the full potential of Japanese robots is realized, they could threaten everybody’s jobs.”
As my friend and occasional co-author Sacha said, having brought this marvellous material to my attention, you have to keep in mind the early 80s British intended audience so as to translate these ornate yet nonsensical locutions into something more straightforward, like:
“But these robots will KILL JOBS!!! Omagad!!! In Britain, we have super high unemployment, and yet we don’t even have robots. Imagine if we had robots. They’re taking our jobs!!! Can’t have robots.”
And of course, this all comes back to culture. This illiterate nonsense is how most people thought and still think in the UK, which is why we have no real industry any more, just bullshit fiat services. Meanwhile, the countries with the highest rates of industrial robot usage also have the highest rates of industrial employment. Much strange. So surprise. Wow. From 1950 to 1980, British industrial output roughly doubled in inflation-adjusted terms, while Japanese industrial output went up 20x. If your explanation of this doesn’t include general social and political attitudes to the merits of capital accumulation (that is to say, culture) then it is wrong.
What does any of this have to do with vibe coding? That this time is not different, and not much more …
I think the first and most essential intellectual exercise is to get clarity on what vibe coding really is as a function of what LLMs are really doing. At the risk of yelling at clouds, I hate “artificial intelligence” as an expression because it is needlessly confusing. It makes this technology sound like a god, a truly autonomous agent that truly does replace people. This, of course, is retarded.
I’ve dabbled with calling them “statistical linguistic algorithms”, whereas “fancy autocomplete” has become a bit of a cliché. But I am also wary of downplaying their capabilities. I don’t at all mean to dismiss them. My point is to praise their potentially positive effect, while contextualizing it appropriately as something like leverage for the mind. Comparisons to steam engines only get us so far before becoming confusing, and, besides, we are not introducing LLMs into predominantly agricultural societies.
Let us consider a more practical comparison. I think LLMs are like calculators or spreadsheets, but for words instead of numbers (pick according to your age and formative experiences). What electronic calculators enabled was an interface for people to perform much, much more complicated arithmetic operations than they could possibly hope to have done in their heads or by their own workings with pen and paper. Spreadsheets generalized the operations and the interface by making it even easier to treat different types of data as numbers to be computed on, and returned to whatever original form they were presented in. LLMs are word calculators: they do the same thing but they let you ask in words and get answers in words as well. Since code is a form of language, hey presto, code itself can be “computed on” in this way.
The etymology of the word “calculator” is worth considering in a similar vein to “horsepower”. It was originally that of a human who calculates, at first with reference to logarithmic tables, but as time progressed, with mechanical adding machines in the eighteenth century, right up to early electronic computers in the twentieth. It is a profoundly good thing that people don’t “calculate” anymore, just as it is profoundly good that such a small proportion of the total work in agriculture.
This is not because the latest edition of Excel comes with 200 brainpowers of human-equivalent arithmetic capacity (or terabrainpowers, more like). It is because Excel can not only perform calculations that no human could in the lifetime of the universe but also comes with an interface that makes this easy; that provides a form of leverage to the limited bandwidth of humans to enable the performance of previously inconceivable tasks. And to free up these humans to intelligently and adaptively accumulate human capital in other areas that then informs their application of this immensely powerful tool. If you fail to grok this, you will slide down the slippery slope to proclaiming that computers must have already put 20 quadrillion human calculators out of work.
“Accumulating capital in other areas” could be anything at all. The cheery trope of vibe coding is a total software noob building a mildly useful app relevant to their hobby, maybe even to their job. But it can come full circle: one can vibe code as a way of becoming a better software engineer by accumulating human capital relevant to the task of coding. And don’t need to take any of this from me, a thinkboi shitposter. You can instead take it from calle (npub12rv…85vg), a doer of actual things:
You could even push it to the borderline satirical in observing further that LLMs are themselves software, are the product of human ingenuity in utilizing other software, and hence that one’s efforts in compounding capital accumulation may be directed to improving the functionality of LLMs. In any case, the obvious extension of Calle’s argument beyond his personal reflection is that the result of all this will almost certainly be more and better coding.
One might instinctively scoff at this and think something like “but vibe coding constantly throws up mistakes, it’s nothing like a professional, and it might even be dangerous if the user doesn’t know how to identify and fix these mistakes.” This may be true, but it could hardly miss the point by more. It’s like saying that calculators make us lazy in our mental arithmetic, spreadsheets invite errors in data tabulation, or steam engines invite reckless and unpredictable application of perfectly well-understood horsepower.
A tool is only as effective as its human operator. We are very, very early in understanding how best to use these tools and being bearish on the basis that somebody somewhere used it stupidly is, well, stupid. Mistakes are the obvious price of experimentation, and the nature of experiments is that they don’t always work. But it is also their nature to be reproducible. When we experiment with creating new tools, we explore the potential to widen access to their product. We democratize leverage. We (quite literally) empower people.
I trust Calle that this is as revolutionary as it sounds. And yet the very nature of radical technological advances is that you definitionally cannot predict what will happen next because knowledge is dispersed and invention and entrepreneurship are irreducible mental phenomena. Hence, with the appropriate conceptual and historical lenses, I think you only really have two options in dealing with this. Option One is you believe it’s all a nothingburger: the latest bubble of easy money blown up by Silicon Valley, perhaps. Option Two is you believe it’s the most important thing in recent history, maybe one of the most important things ever (admittedly, there are more options, but they clearly sit somewhere between these two extremes). In the case of Option One, nobody is going to get any poorer because nothing is going to change. In the case of Option Two, a few people might get poorer in the very short term, but humanity as a whole will get much, much wealthier over every other period of time.
Although I cannot predict it, I would be very surprised if LLMs change industrial society as much as the mechanical power devices of the early Industrial Revolution. But then again, imagine if they did! Imagine a 100x increase in leverage over computation on numerical representations of language, leading to more new and previously unimaginable jobs than the entire current population, with only a slow and gradual decline in roles superficially “replaced”. Can you even imagine it? Maybe it’s not possible, but whatever it is, it’s awesome.
Which brings us back to Jacob to conclude on a question of culture. The literally unfathomable benefits of technological advances will accrue to those cultures that best inculcate their adoption. The act of creating capital isn’t a spontaneous physical event. It takes a mind, and minds exist in social settings. The more a culture encourages tolerant public debate, what today we would likely broaden to “conceptual reasoning”, experimentation, and entrepreneurship, the more likely its constituents are to generate worthwhile industrial advances that give more tools to more people and ultimately benefit everybody.
So, what do you want to do, anon? Do you want to bitch and moan about how people are finally obsolete because this time is finally different? Do you want to tell people they are stupid and redundant and pay them to do nothing to placate them? Do you want to go around destroying physical capital to “save jobs”? Or do you want to embrace empowering people to do things they couldn’t previously do? Do you want humans to flourish? Do you want to vibe capital accumulate?
Is p(doom) 100% or is it zero, just like every other time?
- Allen, #910,209
Thanks to calle (npub12rv…85vg), Gigi (npub1der…xzpc), PABLOF7z (npub1l2v…ajft), and jack (npub1sg6…f63m) for the discussion that stimulated these thoughts.