The legibility as optimization framing is especially clarifying here. Once scale, agency, and information outran narrative capacity, repair was structurally impossible rather than morally deficient. Seeing the present as a phase transition toward divergence rather than breakdown feels exactly right.
Venkat proves yet again in this trilogy that he's practically the only one awake. I absolutely love his thinking, his thesis, and his approach here. That said, I have two critiques.
First is that The Modernity Machine is not an especially machinic project; to the contrary, this is an overridingly humanist effort (possibly because Venkat is a philosophe in a technologist's body). To wit, a machinic approach to explaining modernity might have a Vaclav Smil gloss, with the hermeneutic center of gravity being energy flows, and the explanatory mechanism being the forces modulating and ramifying these flows. Or a Blaise Aguera y Arcas approach, with information processing as the driving force, and functions transforming information (whether instantiated as bits or atoms) the explanatory mechanism. Either way, a legitimately machinic lens will provide quantifiability and legibility.
Explaining history as the product of a modernity machine or a divergence machine locates the explanatory mechanism in the realm of ideas, however. I have no objection to this - I rather prefer it - although I admit it makes Venkat's analysis perhaps less quantifiable, less falsifiable, and ultimately less machinic.
The problem with locating the motive force of modernity in a conceptual contraption is that you then have to contend with your fellow travelers. For example, while Venkat mentions postmodernism in part II, he seems to be defining it in his own terms rather than negotiating the adequate (if inscrutable) world of definitions that have already accrued to it. If we're gonna walk down this road, it's only fair to engage with those who walked down it before, particularly Habermas and Lyotard. Heck, even Thomas Berry wrote (in 1988) "The old story that sustained us no longer functions; the new story has not yet emerged."
My second critique is that the sloptraption approach did not 'stick the landing' better than Venkat could have, at least in the case of part III. I admit that the AI-assist yielded a phenomenal synthesis which moreover transmuted the group's takeaways into a pleasingly linear narrative. Indeed, it reads easier than most of Venkat's Venkat-authored work. But - the drivers of deep value got lost. If, like Venkat does, we were to depict the ideas in this text as an image, it would look more like a washed-out jpeg than a lossless tiff. In other words, good human-authored text has a fractal character that admits of great richness: innuendo, allusion, metaphor, anecdote, etc. - techniques that say a lot with a little, or as Venkat might say, have high implicit information density. These amplifiers of meaning are not present here.
I think you’re confusing materialistic and machinic. A materialistic explanation rests on scientific phenomenology, empiricism, and boundary conditions like Smil’s. That element was somewhat present in the horse book and 1493. A machinic explanation necessarily rests on ideas because machines are primarily ideas. A clock can be made of many materials. What makes a clock is the configuration of components shaped by ideas like “escapement” or “gear.” These are defined by near-platonic design primitives. A computer could be implemented in any design language from paper and water pipes to silicon that can express the idea of a Turing machine. Quantifiability is central to materialistic explanations but a red herring in machinic ones. Legibility is closer to a proper tell of a machinic explanation, and I think we made quite a bit of progress in legibilizing 1200-1600 in machinic terms. We saw a lot of machinic primitives in the readings, such as in the governance structure of Venice or the idea of “fixity” (a memory mechanism) in the impact of print book. Ibn Khaldun’s sociology is mechanistic. Utopia read like the architecture description of a computer. I don’t know if you’re an engineer or been around engineers much, but in my experience, unlike scientists, engineers tend to be very idea driven rather than observation or measurement driven. They learn just enough science and empirical design to allow them to manifest the Rube Goldberg contraptions in their head. We did that in this book club, as you’ll see if you look at the chat threads. Actually rather like Giordano Bruno — who went both wrong or right in very engineering ways. Engineers are “magicians” in the Bruno sense (see my old essay, As Above, So Garage). I suspect scientists as much as true humanists would be aghast at the cavalier way we traipsed over this historical territory. But this is more than a critical perspective in reading the history. It’s also a hypothesis about how the history emerged. A machinic perspective implies you think the history took shape primarily through design. I think I demonstrated that to my satisfaction. 1200-1600 was a machine being constructed by design, even if it was designed via chaotic emergent coordination among many designers. Humans making crackpot notions in their head real by rearranging reality in deliberate ways rather than helplessly surrendering to natural events.
I’d have picked very different books for a humanist exploration. On art history, architecture, religion, theology, probably a history of France, a bunch of poetry etc. I’d say the only truly humanist picks were The Monkey King and Montaigne.
As for whether I’d have done a better synthesis, if the goal were fractal texture, yes probably. I haven’t yet gotten good at prompting that out of LLMs with less effort than just doing it myself. But that’s only one possible goal. The reason I felt this stuck the landing better than me is that it solved for a different, more important criterion here much better than I could — thorough coverage of the explored space. I’m bad at marshaling hundreds of points into a few rough clusters and themes. This did an exceptional job. These rather flat lines can be developed to explore fractal depth, but going the other way would be hard.
You make a compelling case for your definition of machinic, and given that definition, I would concede that you're right.
That said, to decompose your definitional claim, it seems to have (1) an ontological component (machines are configurations of ideas instantiated in matter); (2) an epistemic component (machines are best understood as compositions of legible design primitives); and (3) a normative component (machinic explanations should privilege said primitives). I won't quibble with (1) and (2), but I do wonder about (3).
I agree that clocks are the perfect test case here.
By way of background, I am trying to build a mechanical clock with the lowest-ever kolmogorov complexity. In so doing, I have abandoned composability as a design criterion, and instead am leaning on (non-fungible) material intelligence to do the heavy lifting. Less abstractly, I'm 3D printing the clock out of a variety of materials (with gradients between them), where each material confers some desired property, but not at the expense of mechanical complexity.
It has been slow going (argh, you asked me to write about this over 5 years ago for Ribbonfarm). In the best case scenario, this is because few have attempted building a machine along these lines, so there is little prior art to work off of, and every design element has to be empirically validated from scratch. (Indeed, maybe material affordance can replace design primitives as the organizing principle of a certain class of machine!) In the less charitable scenario, it's because material-continuous optimization of a machine cannot outcompete discrete conceptual decomposition (or, it can, but that doesn't mean it should).
To return to the original argument, I worry that design primitives may be contingent, not necessary. Or to take a different tack, ideas are often the coin of the realm in historiography - maybe they need to be, but maybe that's simply a reflection of the fact that ideas are sticky when optimized for legibility, composability, and transmission.
The legibility as optimization framing is especially clarifying here. Once scale, agency, and information outran narrative capacity, repair was structurally impossible rather than morally deficient. Seeing the present as a phase transition toward divergence rather than breakdown feels exactly right.
Venkat proves yet again in this trilogy that he's practically the only one awake. I absolutely love his thinking, his thesis, and his approach here. That said, I have two critiques.
First is that The Modernity Machine is not an especially machinic project; to the contrary, this is an overridingly humanist effort (possibly because Venkat is a philosophe in a technologist's body). To wit, a machinic approach to explaining modernity might have a Vaclav Smil gloss, with the hermeneutic center of gravity being energy flows, and the explanatory mechanism being the forces modulating and ramifying these flows. Or a Blaise Aguera y Arcas approach, with information processing as the driving force, and functions transforming information (whether instantiated as bits or atoms) the explanatory mechanism. Either way, a legitimately machinic lens will provide quantifiability and legibility.
Explaining history as the product of a modernity machine or a divergence machine locates the explanatory mechanism in the realm of ideas, however. I have no objection to this - I rather prefer it - although I admit it makes Venkat's analysis perhaps less quantifiable, less falsifiable, and ultimately less machinic.
The problem with locating the motive force of modernity in a conceptual contraption is that you then have to contend with your fellow travelers. For example, while Venkat mentions postmodernism in part II, he seems to be defining it in his own terms rather than negotiating the adequate (if inscrutable) world of definitions that have already accrued to it. If we're gonna walk down this road, it's only fair to engage with those who walked down it before, particularly Habermas and Lyotard. Heck, even Thomas Berry wrote (in 1988) "The old story that sustained us no longer functions; the new story has not yet emerged."
My second critique is that the sloptraption approach did not 'stick the landing' better than Venkat could have, at least in the case of part III. I admit that the AI-assist yielded a phenomenal synthesis which moreover transmuted the group's takeaways into a pleasingly linear narrative. Indeed, it reads easier than most of Venkat's Venkat-authored work. But - the drivers of deep value got lost. If, like Venkat does, we were to depict the ideas in this text as an image, it would look more like a washed-out jpeg than a lossless tiff. In other words, good human-authored text has a fractal character that admits of great richness: innuendo, allusion, metaphor, anecdote, etc. - techniques that say a lot with a little, or as Venkat might say, have high implicit information density. These amplifiers of meaning are not present here.
I think you’re confusing materialistic and machinic. A materialistic explanation rests on scientific phenomenology, empiricism, and boundary conditions like Smil’s. That element was somewhat present in the horse book and 1493. A machinic explanation necessarily rests on ideas because machines are primarily ideas. A clock can be made of many materials. What makes a clock is the configuration of components shaped by ideas like “escapement” or “gear.” These are defined by near-platonic design primitives. A computer could be implemented in any design language from paper and water pipes to silicon that can express the idea of a Turing machine. Quantifiability is central to materialistic explanations but a red herring in machinic ones. Legibility is closer to a proper tell of a machinic explanation, and I think we made quite a bit of progress in legibilizing 1200-1600 in machinic terms. We saw a lot of machinic primitives in the readings, such as in the governance structure of Venice or the idea of “fixity” (a memory mechanism) in the impact of print book. Ibn Khaldun’s sociology is mechanistic. Utopia read like the architecture description of a computer. I don’t know if you’re an engineer or been around engineers much, but in my experience, unlike scientists, engineers tend to be very idea driven rather than observation or measurement driven. They learn just enough science and empirical design to allow them to manifest the Rube Goldberg contraptions in their head. We did that in this book club, as you’ll see if you look at the chat threads. Actually rather like Giordano Bruno — who went both wrong or right in very engineering ways. Engineers are “magicians” in the Bruno sense (see my old essay, As Above, So Garage). I suspect scientists as much as true humanists would be aghast at the cavalier way we traipsed over this historical territory. But this is more than a critical perspective in reading the history. It’s also a hypothesis about how the history emerged. A machinic perspective implies you think the history took shape primarily through design. I think I demonstrated that to my satisfaction. 1200-1600 was a machine being constructed by design, even if it was designed via chaotic emergent coordination among many designers. Humans making crackpot notions in their head real by rearranging reality in deliberate ways rather than helplessly surrendering to natural events.
I’d have picked very different books for a humanist exploration. On art history, architecture, religion, theology, probably a history of France, a bunch of poetry etc. I’d say the only truly humanist picks were The Monkey King and Montaigne.
As for whether I’d have done a better synthesis, if the goal were fractal texture, yes probably. I haven’t yet gotten good at prompting that out of LLMs with less effort than just doing it myself. But that’s only one possible goal. The reason I felt this stuck the landing better than me is that it solved for a different, more important criterion here much better than I could — thorough coverage of the explored space. I’m bad at marshaling hundreds of points into a few rough clusters and themes. This did an exceptional job. These rather flat lines can be developed to explore fractal depth, but going the other way would be hard.
You make a compelling case for your definition of machinic, and given that definition, I would concede that you're right.
That said, to decompose your definitional claim, it seems to have (1) an ontological component (machines are configurations of ideas instantiated in matter); (2) an epistemic component (machines are best understood as compositions of legible design primitives); and (3) a normative component (machinic explanations should privilege said primitives). I won't quibble with (1) and (2), but I do wonder about (3).
I agree that clocks are the perfect test case here.
By way of background, I am trying to build a mechanical clock with the lowest-ever kolmogorov complexity. In so doing, I have abandoned composability as a design criterion, and instead am leaning on (non-fungible) material intelligence to do the heavy lifting. Less abstractly, I'm 3D printing the clock out of a variety of materials (with gradients between them), where each material confers some desired property, but not at the expense of mechanical complexity.
It has been slow going (argh, you asked me to write about this over 5 years ago for Ribbonfarm). In the best case scenario, this is because few have attempted building a machine along these lines, so there is little prior art to work off of, and every design element has to be empirically validated from scratch. (Indeed, maybe material affordance can replace design primitives as the organizing principle of a certain class of machine!) In the less charitable scenario, it's because material-continuous optimization of a machine cannot outcompete discrete conceptual decomposition (or, it can, but that doesn't mean it should).
To return to the original argument, I worry that design primitives may be contingent, not necessary. Or to take a different tack, ideas are often the coin of the realm in historiography - maybe they need to be, but maybe that's simply a reflection of the fact that ideas are sticky when optimized for legibility, composability, and transmission.
Not my own theory. Check out Brian Arthur’s Nature of Technology for a systematic treatment.