Still Higher and Subtler Arts

Somewhere in college I encountered this passage from Michel de Montaigne, part of a longer argument about the difference between acquiring knowledge and making something of one’s own. A student could read widely, memorize the words of accomplished thinkers, and become extraordinarily good at repeating what others had discovered. But Montaigne had something else in mind. Bees gather nectar from many flowers, yet what they produce is honey – something materially different from what they’ve collected, with properties of its own. I’ve always liked that distinction, because it suggests genuine creative accomplishment was something more than synthesis. Transformation requires something to happen to the ingredients themselves in the process – an alchemy that applies to matter as it does to knowledge and ideas.

It’s a curious property of invention, this moment when familiar things, brought together in a particular way, become something that can no longer be understood as just a sum of its parts. Consider the printing press. Gutenberg didn’t conjure his press from nothing. Press mechanisms, metal casting, inks, and even movable type had histories of their own, some extending back centuries. But the system he developed in Mainz around 1450 was something materially different from its ingredient technologies. Their integration produced a transformation: the ability to reproduce written material with a consistency, speed, and economy that was unimaginable before.

For centuries – millennia – much of what skilled people knew about making things had been transmitted through other people. A craftsperson acquired their abilities through apprenticeship, observation, repetition, and years of working alongside others who had already mastered a trade. Expertise was often inseparable from the environment in which it had developed. If you wanted to learn a particular technique, you might need to find someone who practiced it, travel to where they worked, and persuade them to teach you. There were exceptions, of course. Craftspeople traveled, hand-written manuscripts did circulate, and techniques passed between workshops long before Gutenberg. But the reproduction of practical knowledge among medieval artisans was constrained by the difficulty of transmitting it, particularly when learning depended on demonstrations or visual examples that were difficult to find or reproduce.

The printing press began changing those conditions. By the early sixteenth century, European printers produced religious texts, classical literature, scholarly works, and a growing variety of books intended for people who made things. Craftspeople contributed to this expanding culture, documenting techniques, publishing designs, and assembling collections of patterns that others could put to work. These books weren’t necessarily meant to be read from beginning to end. They could be collections of ornamental motifs for goldsmiths, patterns for embroiderers and weavers, lettering for engravers, architectural details for woodcarvers, or illustrations of technical processes for people working with metals and machinery. Some explained how to perform particular operations. Others simply presented accomplished examples from which the reader could work. The Metropolitan Museum of Art’s collection of Renaissance ornament documents an astonishing variety of them, intended for everyone from armorers and furniture makers to jewelers, potters, bookbinders, and textile workers.

What’s interesting is many of the surviving copies are incomplete for a rather telling reason: they were consumed. Their owners tore out useful pages, pinned them to workshop walls, traced designs, punctured them to transfer outlines, and worked them over until the paper was no longer serviceable. The books were being consumed in the process of making things. There’s something pleasingly literal about that. Knowledge developed in one workshop had become material for use in another.

The consequences of this change in knowledge transmission went beyond the reproduction of established designs. An accomplished metalworker who lacked exceptional drawing ability could now work with ornamental compositions developed by sophisticated artists elsewhere. A ceramic painter might encounter figures conceived for a painting or engraving and translate them into an unfamiliar material. A practitioner didn’t need to acquire their skills with another master to benefit from some of that master’s discoveries. And because these publications could circulate widely, craftspeople could encounter examples from traditions that might otherwise have remained remote from their own. The repertoire of what an individual could draw upon began to expand beyond the knowledge available in their immediate surroundings.

By 1538, this practice had become familiar enough that a German artist and printmaker named Heinrich Vogtherr published a small volume called the Kunstbüchlin. It contained designs useful to painters, goldsmiths, sculptors, armorers, and other trades. His publication was part of a flourishing culture of printed craft knowledge. What distinguishes his book is the ambition he expressed for what others might accomplish with it. Vogtherr hoped that the designs would help other practitioners bring forth “noch vil höher, und subtiler Künsten”—still higher and subtler arts.

What I find remarkable isn’t that Vogtherr published his designs, but what he hoped others would make of them. By then, plenty of craftspeople were doing much the same. He wasn’t simply distributing examples to be reproduced; he hoped they might become the starting point for work that surpassed them. And he was partly responding to what he perceived as a decline in craftsmanship, rather than assuming that progress would happen on its own. This is a profound change from the culture of 100 years before. Techniques and designs that once distinguished an individual workshop could become part of the working vocabulary of an entire trade. The dissemination of excellence could make excellence, at least in its established forms, more ordinary.

Less than a century after Gutenberg, the printing press’s ability to circulate sophisticated working knowledge was familiar enough within this world of craft publishing that Vogtherr could concentrate on what other people might do with that knowledge. The marvel of the technology had become, for his purposes, an assumption underlying a new kind of ambition.

You can see that possibility in the Italian ceramic tradition known as maiolica. During the sixteenth century, ceramic painters drew on printed engravings, including compositions derived from Raphael and other accomplished artists. Sometimes they selected figures from different compositions and rearranged them into new narrative scenes. A figure that had begun in one artist’s composition, circulated through an engraving, and emerged as part of the decoration on a ceramic vessel had undergone several transformations. The final object embodied the contributions of people who may never have met, working in media that demanded different skills and decisions. The results weren’t necessarily original in the sense of having been conceived independently of everything that preceded them. But neither were they always simply reproductions. And perhaps that is where Montaigne’s bees become especially interesting.

The flowering of printed craft knowledge didn’t automatically produce better or more inventive craftsmanship. These books also enabled faithful copying, the repetition of fashionable motifs, and the spread of conventions that might otherwise have remained local. One could argue that the faster circulation of technique tended to make different workshops more alike, even as it expanded the range of sophisticated designs and techniques available to them. I wonder whether this is one of the tensions inherent in accelerating the transmission of knowledge: what was once exceptional becomes familiar, even as unfamiliar combinations become possible. More imitation and more invention needn’t be mutually exclusive outcomes.

While a master working in 1550 didn’t necessarily possess greater individual skill than one working in 1450, an equally talented practitioner could begin with a considerably richer inheritance because of their expanded access. Discoveries and connections that might have once required years of travel or personal connections to encounter could become available as working resources. Someone else’s exceptional accomplishment could be incorporated into one’s own practice without first requiring the acquisition of all the expertise that produced it. The printing press didn’t simply make more people capable of producing things that had already been made. It expanded the possibilities for what people might conceive and produce by allowing them to work with accomplishments they hadn’t personally developed. The repertoire of ordinary capability could rise, while the frontier of exceptional accomplishment moved outward. To still higher and subtler arts. To make honey.

For a while I’ve been thinking about large language models as books that talk back. It began as a slightly humorous way of making a fantastical technology seem familiar and connected to a tradition, but the expression has stayed with me.

Books allow us to encounter thinking from people we may never meet. They separate knowledge from its originator, allowing it to travel across geography and generations. But reading has traditionally left most of the work of interpretation and adaptation to the reader. You could argue with an author in your head, scribble objections in the margins, or seek out another book offering a different perspective. What you couldn’t do was ask the book to consider your particular circumstances, entertain a hypothetical, or help work through the consequences of an idea that hadn’t occurred to its author.

Now we can do something remarkably close to that. Large language models have made an enormous inheritance of human knowledge available through an interface that can respond to questions, respond with alternatives, and facilitate in the development of ideas. They can explain a technique, propose adaptations, translate concepts between disciplines, and help someone test possibilities that might otherwise have remained speculative. That experience is sufficiently different from consulting a book that the analogy has obvious limitations. Yet I’ve wondered whether the more interesting comparison lies not in what the technologies do, but in the kinds of change they might make possible.

I find those capabilities both astonishing and unsettling. Some of the accomplishments I’ve spent years learning to value are becoming easier to reproduce, and I can’t pretend that’s an entirely comfortable prospect. Every few months seems to bring another demonstration of work that previously required considerable training, professional experience, or intellectual effort. Yet our attention to what these systems can do instead of us may be obscuring another possibility: what might people begin to imagine doing once access to this kind of capability becomes ordinary?

I don’t mean simply producing more work, or doing familiar things more efficiently. I’m thinking about something closer to Vogtherr’s ambition: the possibility that accomplished works, made more accessible to others, become the starting material for accomplishments that previously lay beyond reach. Someone with deep expertise in one discipline might be able to explore what becomes possible when it meets several others, without first spending years acquiring expertise in each. Many will produce little more than superficial recombinations of familiar ideas. But I wonder what might emerge from the attempts that would never have previously been practical to make.

We have considerably more evidence that AI improves performance on established tasks than that it expands the frontier of genuinely new accomplishments. Whether the latter will follow remains uncertain, and it would be a mistake to treat the Renaissance experience as proof that it must. The historical parallel offers a perspective, not a prediction. As Twain said, “History never repeats itself, but the kaleidoscopic combinations of the pictured present often seem to be constructed out of the broken fragments of antique legends.”

There are reasons to be skeptical. If we increasingly rely on a relatively small number of systems to interpret the world, generate alternatives, and recommend solutions, we might become more capable of producing sophisticated work while becoming less capable of producing distinctive work. And unlike the pattern books of the Renaissance, LLMs don’t simply make existing knowledge accessible. They can participate in interpreting, adapting, and combining it. The very process Montaigne associated with transformation is becoming something machines can help perform. That might encourage invention, or it might leave us with an astonishing abundance of accomplished variations on things we already know. Or it might echo Socrates’ ancient concern about writing: that access to others’ knowledge might give us the appearance of wisdom without the understanding that makes it our own.

Still, I find myself returning to Vogtherr.

He was writing nearly ninety years after Gutenberg, at a point when the practical use of printed knowledge had become a familiar part of the world around him. He wasn’t marveling at the ability to produce hundreds of copies of a design. He was imagining what other practitioners might accomplish once those designs were available to them. The technology that enabled that was by his time unremarkable.

We are still preoccupied with the marvel of machines performing work that previously required considerable human expertise. But what happens when those capabilities are no longer remarkable? What might people begin to imagine attempting once they can take them for granted? The people living in that world may be no more individually brilliant than we are. They may simply have inherited a different range of possibilities.

I find that prospect both encouraging and unsettling. Much of my own anxiety about artificial intelligence concerns the possibility that accomplishments I’ve spent years learning to value will become ordinary. Many of them seem likely to. And there’s no assurance that whatever new opportunities emerge will compensate the people whose existing expertise loses its distinction. Not a great thought for someone who takes a lot of pride and identity out of intellectual alchemy.

But thinking about the printing press has changed how I look at that possibility. When I first encountered Montaigne’s bees, I took the metaphor to be a measure of individual intellectual and creative achievement. The important thing was what a person could make of the knowledge they’d acquired. I still believe that. Yet an equally talented craftsperson working a century after Gutenberg could draw upon accomplishments their predecessor might never have encountered. Their talents needn’t have changed for the possibilities of what they could produce to expand.

Perhaps we’re encountering another such change. Not a second printing press, and certainly not a guarantee of some new golden age of human accomplishment. But possibly another transformation in what we can inherit, what we can work with, and what we can reasonably imagine making.

When I first read Montaigne, I was mostly interested in the difference between collecting the flowers and making the honey. Lately I’ve been wondering what happens when the garden itself begins to change.