AI is Inherently Evil: Part 3
An essay so long I had to break it up into multiple posts
Link to Parts 1 + 2 < - If you haven’t, I recommend reading these before coming back to this post.
3. Virtual Information Architecture
Technology does not exist unless someone produces it, and in modern digital societies, the producer is always a corporation. In societies in which capitalism determines survival, markets cannot stabilize. As competitors use up resources and consumers1, all producers must increase profit just to maintain scale; faster growth requires a still larger increase. The only ways to increase profit are to reduce labor costs, reduce production costs, or increase market price2. When a producer has the option to replace labor or materials with cheaper alternatives, without overcharging relative to competitors, they must do so3.
The production costs of any product include both the cost of the physical materials incorporated into the object and that of any information it stores. Computers are expensive because they contain complex metal hardware4. Newsletters and Patreons are cheap because you rarely receive physical objects from them; you're paying for artwork, music, or short essays5. In addition to the combined cost of materials and information, products also incur labor costs. To use the same analogy, Apple pays an army of designers, developers, and manual laborers, while content creators might have one or two friends helping them. The labor costs per product for Apple are much higher, though so are their profits.
Material costs can only fall a fixed amount before the only savings come from reducing the amount of material used. No matter how efficient the mining industry becomes, a real diamond will always cost more than a plastic fake. The ultimate savings limit occurs once the product consists of no materials, meaning it consists only of reference to those materials6. Because companies need technology to transfer data, the substitution of data for materials necessitates the incorporation of technology into labor. Since labor costs depend on the quantity and duration of labor, and technology exists independently of the laborers who use it, a company can reduce quantity by using singular technologies to finish many tasks, requiring a few skilled laborers, or by automating singular tasks that technology can perform, reducing the time any worker need spend training or applying their skills7.
In other words, capitalist technology tends to become universal, capable of performing all labor functions; autonomous, requiring no labor input to function; and immaterial, demanding no material expenses and no material output. These are the three key attributes of technocapitalism8; the extent to which any technology maximizes them is called efficiency.
Computers are, by definition, universal machines9; they can store and transfer any information in valid symbolic logic. This allows them to represent any idea in information (immaterial), including representations of labor functions (universal)10 and of the operation of those functions (autonomous). Notably, computers cannot quite simulate all material events11, nor can they fully operate without prompting. However, their efficiency has no definite limit.
AI is the closest to perfect efficiency that technology has come so far. Theoretically, AI models can perform any tasks for which they have training samples, they require little labor to function, and they handle exclusively immaterial information. Of course, weaker AIs produce unacceptable information or require labor-intensive prompts, and will naturally be replaced by stronger AIs. Hence, AI is not just a temporary fad like Bitcoin or NFTs12, nor is it just another change to the information industries, like photography or typewriters. There is no limit on how efficient AI can get except for the size of the training data, which is still expanding. It is in the best interests of anyone doing any kind of information labor for profit, whether your boss or your parents, to use AI as much as possible to perform as many tasks as possible13.
Because computers can model any other technology, and those models incur reduced initial costs relative to their analog counterparts14, users will choose virtual models over analog when given the choice. In other words, computers and AI do not coexist with other technology; they imitate and replace them. Data stored on old technology is either digitized or ceases to exist. Plenty of illustrators still use a pen and pencil, but they scan those drawings and edit them in Adobe. Publishers still produce physical books, but Amazon controls over half of the book distribution market15, including eBooks for Kindle, including a slew of books generated by AI16. It doesn't happen immediately, but either computer technology replaces that which cannot compete, or the more costly option becomes a premium product, which is just another, slower way of removing it from market.
More broadly speaking, all companies wish to eliminate products from which they do not profit. I noted earlier that material and labor costs are the only essential factors on profit, but these imply a third: the profit lost to competitors. As such, corporations consider competition wasteful, and, in a global economy that already favors monopolization17, a company producing optimized technology will not only passively, but actively, destroy alternative technology.
This reveals the fourth key attribute of technocapitalism: it is totalizing. The faster technology optimizes the other three attributes, the faster it makes other technology unavailable.
Because computers seek to produce any possible representation of any commodity with ultimate efficiency, a goal only achievable using information they already have, we must examine the kinds of information they already store before discussing what kinds they generate. This set, the total information shared across computers, is the internet.
Part 4 will be published tomorrow.
Monopolies face the pressure to continually grow profits because, as they expand operations, they expand the range of potential competitors. This is why all the tech giants are racing to create the "best" AI, even though Apple sells computers and Meta sells platforms—all want the best data and the best consumers first.
This last factor is irrelevant because market price is just a consequence of how different competitors manage the first two variables. Even monopolies have to keep prices below the threshold at which consumers revolt or starve.
Greed, sadism, and the hunger for power are arguably bigger factors in the drive to continually grow profit beyond the point of survival, but make for a weaker argument.
And because Apple and Microsoft are unregulated monopolistic corporations.
Products which are traditionally devalued anyway.
Of course, information is stored on materials, so the next major savings come from converting those materials to information. This is why, historically, economies switched from barter to currency representing bartered objects, then to plastic cards representing currency, and now to Venmo representing card transactions.
The only other agent a company can offload labor to is the consumer, but such offloading is only possible with either assistive labor or automation, since consumers only consume because they cannot produce for themselves as effectively. The most obvious example is automated checkout kiosks in grocery stores.
I'm stealing this idea from Jacques Ellul, but I find his arguments lacking in ambition—he does not back up his astute observations with mechanical explanations, and his writing predates the modern internet.
Cory Doctorow, "How Big Tech Got So Damn Big," Wired, 7 Sep. 2023, https://www.wired.com/story/the-internet-con-cory-doctorow-book-excerpt/.
These are called programs.
The reason for this is that all data exists in material containers. You can replace your apartment with a Metaverse apartment or replace your Metaverse apartment with a brain implant hallucination, but the VR headset and brain implant are still physical, material objects.
And Bitcoin and NFTs are hardly fads to the few who got rich off of them.
This applies just as much to consumers as to laborers and producers, incidentally. For one thing, consumers want cheaper, convenient products, just like producers, incentivizing both to eliminate materials and labor. More importantly, as technology replaces labor and materials, it becomes harder and harder to avoid efficient technology. Google automatically shows AI results even to users who have never touched GPT, and Adobe promotes AI images within their store of stock photos.
Your laptop may be overpriced, but it comes with a free typewriter, television, telephone, and postal service.
Porter Anderson, "US Publishers, Authors, Booksellers Call Out Amazon's 'Concentrated Power' in the Market," Publishing Perspectives, 17 Aug 2020, https://publishingperspectives.com/2020/08/US-PUBLISHERS-AUTHORS-BOOKSELLERS-CALL-OUT-AMAZONS-CONCENTRATED-POWER-IN-THE-BOOK-MARKET/
Robert Evans, "AI is Coming For Your Children," Substack, Shatter Zone, 20 Jun. 2023. https://shatterzone.substack.com/p/ai-is-coming-for-your-children
Cory Doctorow, "Big Tech Got So Damn Big," Wired, 7 Sep. 2023. https://www.wired.com/story/the-internet-con-cory-doctorow-book-excerpt/.