In a previous blog post about the industrial revolution as an analogy for AI, I wrote a bit about how someone might end up with a monocausal narrative about what happened between 1750 and now: one that starts with the invention of the steam engine and other technologies and ends with everything else, “revolutions of all kinds”, the world as we know it, etc. Perhaps, the industrial revolution is a powerful idea, in part because it combines two words we associate with a large level of explanatory power: Industriousness (as in, efficiently creating a particular output) and Revolution (as opposed to, say, a revolt: The night of July 14, 1789, after news reached Versailles of the storming of the Bastille, Louis XVI famously asked the Duc de La Rochefoucauld-Liancourt: “Is it a revolt?” And La Rochefoucauld replied: “No, Sire, it is a revolution.”).
Many historians and philosophers have – in my view – provided overwhelmingly strong arguments for the limitations of this kind of technological determinism. Heidegger, for example, wrote excellent, but not very accessible, things about the topic, the collection of articles in “Does Technology drive History” is more understandable, and I highly recommend it, but it’s still written clearly for an academic audience. Richard Sclove’s analysis in “Democracy and Technology” is very readable, but their reasoning is shortened and seems somewhat motivated by the “new and better vision” of technology policy that they propose in the later chapters. These arguments are often understood as a high-brow pedantic game of “gotcha” – there are clearly many understandable, non-ridiculous reasons why people might intuitively gravitate to such stories, even if they are not as frequently mentioned – or explained – in critiques of various technological determinisms as they could be.
To me, technological determinist views seem intuitive for two reasons:
Firstly: When I look for an answer to the question how something could happen, technologies provide a very salient and satisfying clue. Technologies are built to enable an activity, an event, a process and they’re a tangible physical object. These objects can then be associated with their intended purpose. As a result, it’s easy to imagine counterfactuals around technologies, Markus Aurelius making a call on an I-Phone, or Benjamin Franklin talking to Claude about electricity. (even if Markus Aurelius would not have had anyone to call, or Benjamin Franklin wouldn’t have needed to ask Claude about electricity). In many ways, blaming technologies as an enabling factor for how large scale change also feels kinder than blaming a person.
Secondly, We might find it relatively easy to pinpoint the time when a technology “started”, when they were invented, or built, or plugged in. It’s much harder to pinpoint the beginning of other parts of our world – an idea, a phrase, an activity, or a story, or a symbol. 1
These reasons combine in interesting ways: Technologies seem to have an intuitive “beginning”, and are associated with the people and activities they “enable”, so it’s understandably intuitive to divide up the past into different chapters, the results of different technological revolutions, starting with the “birth” of a tool that enables the proceeding events.
Needless to say, this kind of super intuitive history has selection-effects, and they’re not always consciously chosen. We might consider the time of the industrial revolution, from 1750 to 1850 to be primarily about steam engines and power looms – and suddenly end up thinking the founding of the United States or the French Revolution or the Napoleonic Wars had to be either “downstream” of these inventions – or simply unimportant. In the same “less important” category would be the history of what led to the invention, production and diffusion of technologies, why they were used the way they were, and that these things were contingent on other factors. Cutting up our retelling of the past into eras is a rhetorical instrument to get people to pay attention to a particular aspect of a story about the past. This is legitimate, but it can be easy to forget that time doesn’t actually have eras, it’s just one thing after another, and the “damn things overlap”. In the same way Taylor Swift divides up her career into different albums, and Karl Marx divides up the past (and future) of human life into a feudal, capitalist, socialist and communist “stage” separated by revolutions, the “technological” story emphasises the importance of technologies that “revolutionised” one era into the next. Actual events aren’t “enabled by technologies” in the same way that I can “enable notifications” on my phone – A meteor impact can happen, without a particular actor being “enabled” to intentionally bring it about.2
- One of my former history professors used to say that “Most of what historians do is argue about how something is actually much older, or far newer than everyone thinks it is.” This exercise of asking about the “beginning” is useful for historians, and arguably, everyone else. The question “When did the United States start” is a prompt for people to think about what we mean by “the United States”. Is the constitution a necessary condition? If yes, which parts of the constitution are necessary? What about the constitution’s application in practice? How did either of those things happen? It’s not necessarily a question that’s valuable because there is an objective right answer, but because it prompts us to think about the world as concretely as possible. ↩︎
- We can argue, of course, that technologies actually also explain the meteor impact, because technologies would have enabled dinosaurs to stop a meteor. These kinds of counterfactuals are, as noted above, easily imaginable: A device that would automatically shoot down meteors before entering the atmosphere would have protected the dinosaurs. We can, indeed, explain any unintended event by referencing the absence of an imaginary automated “protective”-device, but the utility of this explanation is pretty limited, very often. Of course, The meteor impact happened, because noone (or nothing) stopped it from happening. Similarly, we can argue that future events depend solely on what people want and the tools they have at their disposal. That an advanced AI tool, for example, once deployed, will shape the universe exactly according to the “value function” of its user. The difficulty of finding a technical “solution” to the alignment problem provides a clear counterargument to the idea that our values are as fixed or internally consistent as the reward function of an AI system. Another counterargument can be found in the currently observable, undeniably chaotic scramble for which AI companies or governments hope to define the value function of state of the art AI systems. ↩︎
larathurnherr.blog 
Leave a comment