This article was originally published by
Book presentations, or book launches, are holdovers from ages long past. One could argue that the same applies to books in print themselves; but we still read and write books, never mind in which shape and form, while I do not see many reasons to keep presenting them in brick-and-mortar bookshops, or similar venues. Friends in the publishing industry tell me that a single tweet, or a successful hashtag on Instagram, can sell more copies than a book launch—and at a lesser cost, for sure. Besides, one of the most baffling aspects of book launches is that, traditionally—and I remember this was already the case when I was a student—a significant fraction of the public in attendance tends to be viscerally and vocally hostile to the topic of the book being presented. Why would readers who dislike a book as a plain matter of principle take the time to read it in full then vent their anger at its author, I cannot tell; but this is to say that having published a book last fall titled
As I only recently found out, that argument is also the theoretical mainstay of the so-called
As for artificial intelligence itself, the source of so much hype and fear and loathing today—that is far from being a novelty. The term was already widely used in the 1950s and ’60s, when the pioneers of cybernetics thought that electronic brains should imitate the way we think, and replicate the formal logics of the human mind. That project failed, spectacularly, in the sense that it never produced any usable result, and artificial intelligence was soon relegated to the dustbin of technical history. For almost two score and a few years—let’s say from the mid-1970s to more or less now—the term “artificial intelligence” was simply forgotten. If it is revived now, almost as spectacularly as it was once jettisoned, it is because AI today, or something akin to it, has started working surprisingly well. Unlike vintage AI of the cybernetic age, however, today’s AI is not even trying to imitate the logic of the human mind. To the contrary, advanced electronic computation can now solve apparently intractable problems—problems we could not solve in any other way—precisely because computers appear to have developed their own logic, their own scientific method, and their own way of thinking, which is quite different from ours. Computers do not think the way we think due to a simple but drastic structural difference between our mind and theirs: unlike computers, we were never hard-wired for big data. What we today call “big data” means, simply, data too big for us to manage—but which computers can manage just fine.
It follows from the above that computers can notate, calculate, and fabricate buildings, for example, quite differently from the way we always have. Think of geometrical notations—the measurement of the position in space of all the parts of a building, which we used to draw in plan, elevations, and sections. No human designer could conceive of a building made of, say, four gazillion different particles, each one individually notated in space—because no human mind could take in, and take on, that much information. This is why our (human) notations tend to simplify buildings, converting the messy complexity of nature into leaner geometrical figures, which we can more easily draw with lines, or script with math. Computers need none of that. If a given problem can be better solved by the robotic assembly of four gazillion different and minuscule 3D-printed particles, they can go for it. Ditto for structural engineering, when computers can optimize any given structure by simply trying, sequentially, four gazillion different solutions—among so many, it doesn’t take any degree of intelligence, either natural or artificial, to find one that will do the job, and solve the problem at hand. But we (humans) cannot work that way, because it would take forever.
Evidently, buildings conceived, calculated, and built that way tend to look very different from anything we ever designed. They also tend to be better fit to specs (i.e., stronger or lighter or cheaper or whatever specs we choose to optimize) because that’s the spirit of the game—that’s where computation outsmarts us. That does not seem to me a prospect that architects should look down upon with benign neglect. We already know what the first digital turn was—that’s history. But we can already figure out what the second digital turn is going to be. The first digital turn was about bits and atoms. The next is going to be about bits and neurons. There is more digital after the digital, whether we like it or not.
Mario Carpo is the Reyner Banham Professor of Architectural History and Theory at the Bartlett, UCL, London. His latest monograph,