Mario Carpo: “The Second Digital Turn” @ Talks at Google (Transcript)

Yes, we build buildings, but we build buildings by making drawings of buildings. We make drawings, not buildings. We make drawings, clean drawings. We give them to the builders, and the builders, they, will make the buildings we design. We do not make.

The designers do not build, and the builders, they are not allowed to design or to change our design. The kind of drawings we give to the builders are called blueprints, because once upon a time, they were blue, and they were printed. Now they’re no longer blue, and they’re no longer printed. But we still call them blueprints. And it is thanks to blueprints, to these kind of drawings, that our profession is not a craft. It’s a liberal art, an intellectual activity. We do not climb on scaffoldings. We do not toil in the snow, or in the rain, or in the heat of summer. We work in clean offices with heating, ventilation, and air conditioning.

And we are, most of the time, even better paid than the actual workers who make the actual buildings happen. That’s the advantage of being a notational art. We build by notations — by making drawings, drawings which eventually become buildings, but we do not actually make the buildings. This is the advantage. A disadvantage is that, if you think of a way a building happens, we can only build that of which we can make a drawing.

If we cannot draw it, they, the builders, cannot build it. We are at the mercy of the notational tools we have at our disposal to notate a building. If we cannot draw it, they cannot build it. Now some geometrical shapes — come on — like this one are very easy to notate. That’s a shoe box. You make — eight points is enough to notate it — or three points and three vectors. That’s a really very easy drawing to make, so that way, notation is easy.

But think if what you want to build that potato. Now, since I was cooking last night, I can actually show you. Oops! This is what I have in mind, this awesome British potato. If you want to build that potato by notations — plans, elevations, and sections — think of how many drawings you have to make. Each one of these points — not one of these points is aligned. So you have to make a huge number of sections in plans, elevation, et cetera, et cetera, because each point has to be notated individually and separately– millions and millions of point, which will take millions and millions of drawings, which will take a huge amount of time, which is possible, but it’s not very practical, because it takes too long. So to build the shoe box, it only takes eight points.

To build the potato, you must make thousands and thousands of drawings, each slice, each section, being different with different points notated and measured in, XYZ, in three dimensions, which takes a heck of a lot of time– which is one of the reasons why potatoes were very seldom built in the history of architecture until computers came. Because this kind of repetitive boring operations, notating each point– xyz, three measurements– if you have to notate 4 million points, for us, it takes a lot of time. A computer does it in the blink of an eye. So as of the early ’90s, when computer-aided design became to be affordable, you would expect that architects start building potatoes like crazy, because there have been for centuries, a pent up demand for potatoes which was never fulfilled, because potatoes were impossible to grow and build. They are easy to make.

I could make this potato in clay with my own hands. That’s easy. But if I want to make plans, elevation, and section to scale– to give a blueprint to the builders to build the potato, that takes a long time. That was impossible until 20 years ago– not impossible, but impractical. But now it’s possible, so you would expect but as of the early ’90s or mid-’90s to see potatoes everywhere.

And if you google– you know what it means, right– digital architecture, or computational architectural, or parametric architecture, you will find a mosaic of this kind of stuff. And they are potato-esque, but they are not real potatoes. Because they are round and smooth, but they are streamlined, and technological, and clean, and almost mathematical. A potato is rough and disorderly. These are not potatoes. They are something else.

And as of the early ’90s– well, this is closer to a potato but still not– don’t ask me which kind of a building this is, because it ain’t one. But it’s still not a potato, because it is too smooth and too precise. You can understand that there is mathematics in that, which isn’t in this. In the mid-’90s a colleague of mine came out with a suitable catchy term. They decided to call these new kind of digital shapes, blobs, taking inspiration from this science fiction movie from the ’50s. And so, for a while, these potato-esque digital stuff were called blobs. Today, they are called– oops– parametric or parametric design. But in truth– and that’s the way I often tell the story to my students. If we look at the actual timeline of how these things came to happen, these are not potatoes. These are not blobs. These are fish. Because it all started with this big fish, which you may have seen if you go to Barcelona on the beach. Above the beach, not in the water– over the beach, there is this huge metal fish floating in mid air. This was built by Western architect Frank Gehry, and this was the first time Frank Gehry used software for computer-aided design for which, eventually, it became famous– a star architect which he is now, but which he wasn’t then.

And how is it that to build a fish, Frank Gehry, at that time already a well-known architect, decided that he should use computer-aided design? And which kind of software did he think he would need? Well, think of it. His problem was to model the streamlined curve of the shape of a skin of a fish. And his argument was why is the fish so streamlined and smooth? Because it moves in water the way a boat moves in water, and the boat, the hull of a boat, has the same streamlined lines. For a long time– this is a 17th century drawing– shipbuilders have had the technique to make these curvy pieces of wood at the point of contact between the hull of the boat and the water. Imagine the frame of a boat in timber is given its structure.

And then you have to nail this slat of timber to the frame of the boat. And they have to be smooth, because that’s the point of contact of the water. With a boat, when the boat moves in water, there is friction, or drag. And if the slat of timber isn’t smooth, the boat will start to rock, and it will slow down. And this timber curvy slats of the hull, they have a name in English.

Since the 17th century, they are called splines like a line but sp before sp, splines. In the dictionary, you find this word in use as of the 17th century– technical term used by craftsmen who were building the hull of boats by curving with artisanal skills– and each one, they did in a different way– these slats of timbers, so that they would connect a given number of fixed points in the smoothest continuous way. Now the line of the stream– the line where the hull of the boat touches the stream of the incoming water– the line of the stream is called the streamline, so this is the principle of streamlining. That’s where it comes from. The line of the stream, that’s the streamline, which has to be smooth to avoid friction, or drag.

And boats were built this way for a long time, and airplanes are built this way too. Because aerodynamic and hydrodynamic have the same principles. This picture was made in the mechanical engineering department of Cornell University during World War II. This is a team of engineers designing the curve of the wing of a fighter plane. And we were still using the same artisanal technique. They are obtaining the line of this curve by craft. It all depends on the elasticity of the material you’re using, so the kind of timber you are using. If you choose, I don’t know, pitch pine, you get a certain curve. If you use fir tree or oak, a different one– if you use as a same material and you bent rubber, you get a different color. If you use melted mozzarella cheese, it would be a different curve still. That’s the way it was then even during World War II when airplanes and wings of airplanes were mass produced, but they still did it in this artisanal way.

And so, in the ’50s and ’60s, the automobile industry kept doing it by hand. Oops! This is probably the most famously streamlined automobile of all times. Who knows the name of this car, just for my personal curiosity? Perhaps your grandfather owned one. Who knows it?

AUDIENCE: A DS?

MARIO CARPO: Say that again.

AUDIENCE: Is that the DS?

MARIO CARPO: It’s the DS, DS, which in French means the divinity. S — designed in the early ’50s. By the way, this masterpiece of French design was entirely designed by an Italian engineer, and it came into service in ’56 or ’57. That was the car of General de Gaulle, et cetera, et cetera. But all this was made by hand.

And the way design was made, when this car was produced, was entirely artisanal. The curves, the folding curves, of the body of the car were made first in clay, and then when the final design of the car was ready, it was made in hard wood. There was no blueprint of this car, because nobody could have made all the drawings necessary to make all these curves– in plans, elevation, and section was too complicated. And so, the actual design of the camera was a model of a car in timber kept in an atomic shelter somewhere in central Paris. So, you know, it was the Cold War, so if something happened, the design would stay there.

And when engineers needed to drive blueprints for any practical reason, they went to the basement into the atomic shelter, and they measured the wooden model in scale one to one. They took the measurements out, and they derived blueprints as needed when needed. It was laborious. It was entirely artisanal, which is why in the late ’50s, simultaneously at Citroen and at Renault also in Paris, the engineers, designers, and the CEO of two companies called their top engineers and said, listen, guys; we know how you do it, by hand. This has been done that way for centuries. Now you– them speaking to their engineers. You are the best engineers in the world. The French always think they are. Now shouldn’t there be a way to notate these curves using calculus, x and y, a function– the way we learned it at school.

We can notate parabolas, hyperboles, ellipsis, circles, you name it. Why should we not be able to notate any curve whatsoever in 3D just using three letters and a little bit of coefficients and parameters—mathematics. Could you not do it mathematically, so it is more precise? And one day– one day, who knows? We may even use computers to do that. We don’t have them now, but one day we may. And so, two teams started to work in complete secrecy– one at Citroen, one at Renault– with the same assignment, find a way to notate this using mathematics.

We now know the name of the team leader at Renault– and the name of the team leader at Citroen remained unknown for a long time. It’s a fascinating chapter in technological history. These people evidently knew each other. They had studied in the same schools. They probably did the military service in Nigeria together. And their wives were probably a member of the same rotary club in central Paris. But they had to work in secret and in competition. And we know that the Renault team came up with some results sooner. It was the early ’60s, when they made a famous demonstration in front of the CEO of Renault. They took a signature on a banknote — where it is, you know, the signature of the head cashier of the Banque de France, a doodle full of circles.

And they showed that we have translated that random curvy doodle into formulas, and they fed these formulas into a plotter. The plotter made the drawing of the signatures at the different scale, and then at another scale, and then at another scale. Because the point is a mathematical notation has no scale, so it was the same at every scale. And it was a fantastic and persuading demonstration. And the CEO of Renault, as to Pierre Bezier, that was the top engineer at Renault– said good, fantastic. Great work. How long did it take your team to translate that doodle into mathematics? Two years. And how long would it take to notate the whole body of a car using the same mathematics? 22 years. Well then, that’s not very practical. But they said, no, it isn’t if you do it by hand, but one day computers will come. And then they will do it very fast. Do we have these computers? No. Will we have these computers? No. Are we going to buy any computer? Has anyone have any computer in France? No. Perhaps the Americans do, but we don’t– he says, well then, shoot yourself. That’s what I presume the CEO told to Bezier, publish his pure mathematics, no practical use– publish it in some wacky mathematical journals and good luck, which they did. They published it, 1966, but only a few years later, computers did come. And so, these mathematics was then open source, as we would say. It had been published– scholarly journal. Anyone could use it, and everyone started using it like crazy.

Renault, they built their own software for computer-aided design called Uniserve. In France, an aircraft maker, Dassault, developed its own CATIA software. In America, McDonnell Douglas, General Motors, and Boeing started to use it. Boeing improved upon it significantly. They generalized the mathematical notation into something called NURBS, Non-Uniform Rational B-splines– which in school of engineering is an acronym meant to stand for No One Understands Really B-splines– which means it’s complicated.

The work at Citroen was actually, we now know, mathematically better. But the leader– the bosses at Citroen– a private company. Renault was state-owned — decided that it was an industrial secret. They kept it in a safe for 20 years, so nobody could do anything with it. [Pierre de Meuron] is the name of the other guy. These kind of curves are now called Bezier’s curves, from the name of Pierre Bezier, or NURBS. Fast forward to 1991 or ’90 when the office of Frank Gehry had that problem I told you. They wanted to design that fish, and they knew or with some reason that not really shipbuilders but aircraft makers would have the technology they needed, because they knew part of the story. And so, Frank Gehry was in Los Angeles. McDonnell Douglas was– and still is not called McDonnell Douglas– south of Los Angeles, so they made a phone call which is not recorded.

But we can imagine something like that Consider that in 1990-’91, Frank Gehry was not yet a star architect. He was known among the architects, but people outside the profession didn’t really know about him. So imagine the call– office of Frank Gehry to the technological office of McDonnell Douglas… office of architects. We have to build a big fish. Can you help? Sorry, guys. You have the wrong numbers. We are not in the fish building business. We do birds, big birds, airplanes. And yes, but the mathematics is the same. And yes, sure My name is Napoleon. Good luck. But Frank Gehry didn’t take no for an answer. He tried again– called the MIT– his friend, Bill Mitchell, then head of architecture at the MIT– one of the founders of computer-aided design. He said, oh, don’t try with those. You know whom you should call– these crazy guys at Dassault in Paris. They made these fantastic fighter jet, the Mirage, so expensive that no army can buy it except the French. And they have this fantastic software for designing splines, which is called CATIA– so complicated that no engineer can use it except their own. Call them, they may have something. And he did call them, and they said, sure, sure. A fish, seashells, the snakes, elephant tusks, all kinds of organic cures, that’s our business. We would be happy to help. A team of Dassault engineers went to Los Angeles.

They worked together. They simplified CATIA to make it suitable to architectural design, and they produced something which was used to make the fish you have seen. Frank Gehry liked that stuff so much that he kept using CATIA and building fish. He has been building fish all over the world, and he’s now famous as the most famous fish builder in the history of architecture. The original fish– I know it from, Frank Gehry, himself, it was actually meant to be a carp, which is– I think it’s OK, considering my family name.

It doesn’t look like a carp, but then immediately after he built, of course, the Guggenheim Bilbao using the same software, CATIA, simplified for architectural purposes. And he went on building the same kind of spliney fishy curves all over the world. This is the Guggenheim Bilbao integrated in ’96 or ’97, but this is the Philharmonic in Los Angeles a bit later. This is very recent. It is in Paris. And this is now Gehry’s signature style– the style of a fish, because of the spline modeling software he’s using.

Now he was so successful in using this adaptation of CATIA for architectural purposes that he copyrighted a simplified version of CATIA for architects. We call it Digital Projects. He founded an independent company called Gehry Technologies, providing fish making to other companies, who didn’t have the expertise. And he made so much money with this company, which eventually, he sold the company– which is now an independently owned company still called Gehry Technologies– but owned by another company– to do that. Most normal offices of architecture and students in school cannot use CATIA. It’s too complicated and too expensive. They use cheaper, simpler, crappier software, who include some of the same spline modeling tools called Rhino, Maya, formZ. You may have heard some of these names. They have all been developed in the course of the ’90s, and they are now universally used.

So all these, if you look at them after the story I told you, you should not see them as blobs, but you should see them as fish. Look at this fish. This is fish—fish. Some of them were never built. Some of them were built. This looked like a Photoshop, but this is a real building. This is a real building too. This looks like a real building, but it was never built and never will be, et cetera, et cetera– fish, fish design. Now the problem is if you go to my school over there one mile in that direction or to a handful of other schools which are from the avant garde of design innovation, and you look at what our students or my colleagues have been doing for the last five or six years– and this is a selection of what they do– they do not look like fish at all. The style is completely different.

This is what I call, in my book, the second digital style. It’s disjointed, disconnected, broken, fragmentary. Continuity has been replaced by discreteness. It is not smooth. It is rough. And I was trying to explain to my students. Well, this is what one of my colleagues has done. Since I tell the story with metaphors or analogs– the potato, the fish– I was looking for another animal to describe this. And they thought it would be a sheep. The problem is with my accent– and many of my students also speak English as a foreign language– if I say sheep, no one will ever understand if I mean the boat or the animal.

So I had to find something easier, so I decided to call it a dog. It looks like a dog a little bit, but I was teaching American during the foreign American students are demanding more than our British students, and so when one of the students came to me after class– professor, which kind of dog do you have in mind? And the students produced this vignette showing that not all dogs look like that.See, this is a modernist dog, which is streamlined as if it were a fish. You know, everything in this picture is streamlined. It’s a model in this picture. And you will see the fish itself. So now the dog looks like a fish, and the fish looks like a ballistic missile. And because the modernism was all about streamlining.

Everything is streamlined, even the dog. The fish is normal. It’s streamlined by nature. The dog shouldn’t be, but it was made. Now this is the dog I have in mind. This is a postmodern dog, which gives an idea of complexity, and, you know, discreteness, and messiness. This is clean, but I would imagine that in normal life, this dog is actually quite dirty. So this is what is going on. So the question is since evidently computer-aided design is driven by the tools– the technical tools we are using and then by the performance of the machines we are using, why is it that the style of computer-aided design has shifted from fish making to dog making? Was the software of the ’90s more fishy, or is the software today more inclined to being dog? As you probably know, computers still do today what they did 20 years ago, or even 30 years ago, or 20 years ago. But we are much faster, and more powerful, and cheaper.

For the last 5 or 10 years, we got so used to this wealth of data that we even invented a new definition to define this new data rich environment. We call it big data– to define this notion that, for a long time, data was supposed to be rare and expensive. Now it’s ubiquitous and cheap. It’s a big, you know, upheaval of the anthropological condition of humankind from the beginning of time, still 10 years ago, we always needed more data than we had. For the last few years, it is, as we always had, more data than we need.

So we call this big data, or we have a number of other terms to define the same machine learning, deep learning, artificial neuro– which I don’t know what it is– blah, blah, blah. Or you can read all this. And there is a notion that some of these may somehow be already related to an idea of artificial intelligence, which is odd because, for people my age, we used to think that artificial intelligence was something that would come one day in the future. And now we are being told we have already been using it for the last five years without even knowing, which is odd. And I don’t know what artificial intelligence is.

Perhaps you know. You will tell me after during coffee. But I can tell you what is starting to happen in design when we start to use some of this big data, or deep learning, or machine learning, or even artificial intelligence tool in the ordinary practice of our work trade. And I will show you a couple of examples to give you an idea. Since we are talking about curves, which is why I gave you this long introduction, this is the normal human mathematical way to notate a curve.

We studied it at school. This is a parabola. And the magic of calculus is that with– two letters x, y– and three coefficients, or parameters, numbers– a, b, and c– we notate an infinite number of points. How many points sit on this curve– a huge number of them, an infinite number of them. But a simple line of plain mathematical script as long as this is enough to notate all the points we need that sit on that curve. That’s the way we do it. That’s a very brilliant way to do it, because it compresses a huge amount of data into a very short clean and memorable notation– so memorable that I studied 100 years ago, and I still remember it even though I really never used it. But this is the way we do it, and this is increasingly the way a machine, or artificial intelligence, would do it– not by using the function but by making a long list– a long list of points For each x, there is a y, which is located on that curve. And you make that list.

And I remember making this list, when I was a high school student, to check that the point really sat on the parabola– that you cannot notate every point that way. The purpose of scripting the function is precisely that you do not need to indicate each point one by one, because the script indicates them all in a single line. But for a computer, from a computer’s point of view, making a list of 1,000 points, or 1 million points, or 1 billion points is not a big deal. It is for us, because we cannot work with that list. But a computer can manipulate a list of 1 billion points in the blink of an eye, which is indeed what is already starting to happen in architectural design.

Because if you look at this, a famous fish, built recently by Zaha Hadid– well, when she was still alive– one of her last buildings before she died tragically two years ago. It’s huge. It’s the biggest fish ever built, I think. You can see it from the moon with the naked eye. It’s enormous. But the mathematical script is more– just two lines or three lines of script or perhaps a bit more, but it is mathematics. You can recognize that it is mathematical because this curve, this streamlined curve, looks familiar. This is something that we recognize something– ours, because we know the mathematics underpinning it. It has been around for quite a long time. Calculus was invented at the end of the 17th century. Bezier’s mathematics between 1958 and 1964, you don’t need computers to use that mathematics, but with computers, it ran faster– because more affordable. That’s, you know– they weren’t as we knew it even though it is a bit bigger than what we used to do using that kind of tools. If you look at this– built recently by Michael Hansmeyer– a 3D-printed grotto using the biggest to-date commercial 3D printer. You can 3D print a single block in the size of a room. And each part of these 3D-printed structure is printed as a voxel.

A voxel is a little unit of matter, which is calculated individually and separate. It’s like an atom in three dimensions, xyz– the smallest unit of this volumetric composition. And if you look at it close up, that’s the belly of that grotto. This grotto was made by 3D printing 4 billion voxels, one by one– each one notated, calculated, and fabricated. For each voxel, there is a notation and a calculation, for billions of them. We couldn’t work that way, because if we print out that list, it would start here and finish in another continent. But a computer can run that list in the blink of an eye, so this way of working, which is absurd for us, makes perfect sense for a machine. Because we cannot work that way, but the machine can. And if this composition looks a bit weird, and wacky, and strange, there is a reason for that. It is already the expression of an intelligence which is not ours.

This is the outward and visible sign of an inward invisible logic at play, a logic which is not the natural logic of our mind. It is the artificial logic of a machine. A machine can do that, because the machine thinks in a way which is different from the way we think. If we use a formula, we don’t do that. If we use a list of points, we could do that. But how many points can be manipulated by hand or using a slide ruler? Not many. These are 4 billion points, and it shows. It is a complexity, or richness, of data, which the machine can manage, but we can’t. This is the way the machine thinks. That’s not the way we think.

Another example– this is not a Photoshop. It was actually built in Germany, Stuttgart, by Achim Menges. And it is a small pavilion. It was built in a public space. And this being Germany, before you build it, you need permission from the local office for something, which has to validate your structural calculation to make certain that this building would withstand the wind of x knots and will also withstand the weight of certain [inaudible] of snow– something like that. So you have to provide structural calculation, which have to be validated by a public office, to certify that the building is safe– structural calculation, structural design How do you think this building, which is made of filaments, was calculated? Using these structural formulas — these are the formulas I studied at school — to the limit– with a lot of effort of imagination, we could use these mathematics and this engineering to calculate each filament individually, but there, again, probably 4 million filaments in that shell. So if you want to calculate each filament individually, in principle, it is possible. In practice, it would take six years. So we would not really do it.

Yet the structure was calculated, but how did they do it? Not using traditional structural formulas– using something called finite element analysis, the computational version. I shall not go into details which, by the way, I don’t know very well myself. But the way it works is that you see on the screen a simulation of a structure, and you simulate in the simulation a certain load– wind, or snow, or whatever. What you see on the screen is that part of the structure starts to blink red– what time is it? At what time did I– yes Well, we are on time.

It started blinking red, on the screen, the part that we break. So what you do? You tweak it. And you change it, and you try again another load– and if you blink red, somewhere else– and again and again. And you try and you keep trying until you see something on the screen that doesn’t blink red. Everything is green. That means it will not collapse, so that’s the good one– the one you will build. The problem is you have to keep trying, and trying, and trying. And you have no clue as to which trial will give the best result. The trials are fast, because they are in simulation, which is why, to make the progress even faster, we ask the machine to keep trying automatically. And we can call that optimization.

Machine keeps trying, and trying, and trying. And the machine does, say, 1 million trials in one hour, and chances are that some of these will be good. And the one which the machine finds will be among the best. One of them will be built, but nobody knows which changes will give the best results. How do you think this was calculated before it was built? It was calculated, but it was calculated the way I told you– by simulation and trial and error on the screen.

Why does this one structure stand up and the 9,999 just tried in computational simulation didn’t? Nobody knows it, least of all is designers. And yet we know– and yet we know that it will stand up, which is why we can build it. Because in simulation, we know it did withstand. So what we call simulation and optimization is massive computational trial and error– try, and try, and try, and try, and keep trying. Or in another way is the art and science of finding good solutions without knowing why they work. So in a nutshell, it is already evident that artificial intelligence– and this is artificial intelligence, which to some extent it is– it works. We can already use it to solve problems we couldn’t solve in any other way. But it works in a way which is different from the way we would work. It already shows a logic at play, which is different from the logic of our mind, which is probably a good reason to call it artificial. Because our organic intelligence would not solve problems that way.

The artificial intelligence of a machine can already solve problems in a different way, which in many cases is the only way to solve problems of a given complexity. So yes, it works, but it works in a way which we may find– well, which is different from the way we think, which is probably one reason why we should call this intelligence, artificial– because the organic logic of our mind doesn’t work that way. Think of what we did not do. In the case of a function, we cannot calculate 4 million points one by one. The machine does, and the machine does it.

In the case of structural design, we cannot run 4 million trials in a sequence, because it would take forever. But the machine does it in 20 minutes. So trial and error, which is a very stupid strategy for humans, is a very good strategy for the machine. If I look for a formula to put all these into tag line, I don’t have to look very far, because you people have already found it 14 years ago when you launched Gmail with this type of search, don’t sort. Because if you think of what I just told you, what we do– we humans– we sort. We take data, and we organize them and structure them to make them smaller, and more functional, and more understandable. That’s the way our mind, and traditional science, and mathematics always worked. A computer can search so fast, but this sorting is often unnecessary. Because by the speed of the processing of the simple sorting process, sequential, a computer can find the best solution, but we with our slow searching would. Think of names in a telephone book or in a telephone directory. 1 million names, we have to sort them alphabetically, so we know where our name is when you look for it, and we don’t have to read all the names to find the name we’re looking for. But a computer can do just that. The computer can read 1 million names in two seconds. So our sorting, which is indispensable for us, is unnecessary for the machine, because the machine can search so fast that the preliminary sorting, which we need, is not necessary for them.

Think of books in a library. We put books on a shelf for following a system of logic of classification, so we know where certain subject is when we look for it. We put things in certain places, so we know where things are when we look for them. So the shelf mark or the code number– if you’re looking for a book on architecture, Renaissance, the 16th century Florence churches– it’s a formula, a shelf mark. And with that number, you go to that shelf, and you find the book you’re looking for without having to read the titles of all the books in the library– which would take forever. A computer can do it.

And with virtual reality, we can do that as well. Think of– instead of using a librarian– when all the books come in, you tag them with radio frequency identification– a little chip– then you put them in a huge mountain, no sorting whatsoever. Then you fire all your librarians, and you buy a pair of virtual reality glasses. And when you’re looking for that book in that mountain and saying, where is that book? And you would see blinking red in your field of vision, so you don’t need librarians to sort, because the machine can search. Search or sort? We sort– human intelligence. Computers search– artificial intelligence. But that’s not my cup of tea. To go back to my cup of tea, which is potatoes, and fish, et cetera, it is evident that our computational design, it’s a fascinating testing ground already for artificial intelligence, because the stuff we do is so simple and so cheap. And the software we use is so, you know, elementary. And we do physical stuff– the feedback loop, you know, the verification is also immediate.

Start in the morning, and by the evening we know if it will stand up, or if it will break down. And we know what we would look like– verification is faster probably than in many other trades and professions, which is why we can probably redesign as some architects. We can probably intuit the spirit of this game sooner and faster than many other professions do, because we have an immediate physical feedback. If it falls down, it doesn’t work, so we have to try another. So if we are lucky, we can hope to glean the spirit of the game sooner and perhaps better than some other professions do, but the problem is this is not our game.

This was never our game. It was never meant to be our game, because this is your game. You came up with it. We didn’t. So this is where I should stop speaking and start listening. I think we have still a few minutes for that. Thank you.

QUESTION AND ANSWER SESSION

HOST: So we have 10 minutes for questions. Yeah, right here.

AUDIENCE: You had so many choices. Hi. You said you actually were an expert in classical architecture. How do you see this as an evolution of what the Greeks and Romans were trying to do? Because there’s are quite mathematical as well, their architecture. Do you think this is just the evolution of that, or is this something totally different?

MARIO CARPO: Well, architecture is always at the mercy of the tools we use to make it happen. That’s inevitable. To some extent, every expression is at the mercy of the tool we use to manifest it. You can think that language is natural, but in fact, when your ideas are manifested through the alphabet or through syntax, the tools that you use to communicate feedback on the kind of message you can, you know, transmit. This feedback loop is inevitable in every human expression. In the case of architecture, if you compare it with painting, the technical bottleneck is more determinant. Of course, a painter is at the mercy of the kind of canvas he’s using and the technology of the colors he’s putting on the canvas, but you can think that that technological– the limitation, it’s not much if you compare it with the technical implication of building a big building.

And we are particularly at the mercy of tools of quantification, because we have to measure a building before it is built, and that was always the case in Greek, and Roman, and classical architecture– not so much for costing or estimate, which is important today, but for proportions, which, for them, was a matter of vital importance. And evidently, if you look at Roman or Greek architecture, you can glean, from the way the buildings are built, the kind of tools of quantification they used to make it happen, which was not number based, because the Greek and Roman didn’t have good numbers, and they didn’t trust numbers. It was all based on geometry. What we would solve by using, you know, Hindu-Arabic numbers– which evidently we didn’t have– the numbers they had were crap, and so they couldn’t use them to calculate anything. But their geometry was first class.

We still use it. So all that quantification was achieved through Euclidean geometry tools, whereas the arithmetic number-based tools we use today didn’t yet exist. You know, the Greeks always mistrusted numbers. They preferred geometry were all in the compass, the typical tool of the masons. The quantification of the proportional harmony of those buildings derived from that use of geometry, which is why neoclassical architecture as of a 17th century or 18th century– which is number-based and not geometrical determined– from a distance, it looks the same. But if you look at it with the eye of an expert, you can tell that it’s no longer Euclid. It is numbers. Those numbers were pure arithmetics. The story I’ve told you is how quantification shifted from algebra to calculus, because this is what Bezier and the French teams did. In a sense, that was the culmination of the dreams of Western mathematics, because with Descartes, Leibniz, and Newton, we could not eat conics– you know, parabolas, hyperboles, et cetera, et cetera.

With Bezier, it was, in a sense, the culmination of the dream of mathematicians of all time. You can notate, using a mathematical script, everything. A cloud in the sky, a flower in the field, any shape and form in nature can be scripted as a mathematical notation. Of course, we didn’t need to notate the cloud or a flower. We need to notate the body of a car. And it was complicated, but with computers, it became easy. Which is why when computers became available, first thing we did– we didn’t use them to make potatoes. We used them to make fish, because that’s mathematics, and there is no mathematics in a potato. We call this freeform, because there is no mathematics embedded in it. And in a sense, that dog, the last part of my story, is closer to the potato than the fish.

So in a sense, we are getting back to a potato, but with a delay of 24 years. But the feedback loop back to your question– at every point in time, the kind of architectural shapes and form you see– if you look at it as an historian of science, you can read in the physical shape, the tools of quantification which were essential to make it happen. That doesn’t mean that I can do that for every period I can only do it for some little passages in it, but that’s the game.

AUDIENCE: Thank you for an interesting historical introduction. I have a followup question to the previous one. Could you please feed the work of Gaudi to these fishes, dogs, and the straight lines story?

MARIO CARPO: Could you do what?

AUDIENCE: Could you please put the work of Gaudi into this?

MARIO CARPO: The work of Gaudi, Antoni Gaudi?

AUDIENCE: Yes, yes– into this scheme of fishes, and the dogs, et cetera?

MARIO CARPO: Well, Gaudi was a very important reference of the beginning of a digital term in the ’90s, because Gaudi– who is not a dog and not a fish. It doesn’t belong to any historical parameters, because it was a one off. It was an isolated genius, who due to a remarkable set of circumstances was allowed in Barcelona to make something happen, which is a miracle– which should in theory never have existed. He wanted to reenact the way of building of medieval master builders, so he was building without notations. In the Middle Ages, architectural blueprints did not exist. Craftsmen, there was no separation between the designers and the builders. The master builders were conceiving and making at the same time on site on the fly. And in 1890 until 1910, Antoni Gaudi in industrial Barcelona wanted to revive due to his ideology, to his faith, to his inspiration– wanted to revive the medieval way of building. He didn’t want to revive a style.

He wanted really to revive a social organization of a medieval building site, so he wanted to devise a way of being where there’s no designer separated from the builder. The builder is a master builder, who decides every day what is going to build In an industrial world, you cannot build that way. But in Barcelona, at that time, he found illuminated sponsor, who decided that he should build a big cathedral that way– to the superior glory of god, which he did. And the building is still going on, because a medieval cathedral, since it has no design has no beginning and no end.

So long as there is someone who is willing to build, the building will keep growing. And as of the mid-’90s, someone I know, Mark Burry, took on from where Gaudi stopped using computer-aided design, because he thought with some reason that using CAD/CAM– computer-aided design and computer-aided fabrication– it is easier to reenact the medieval way of building. Because if you work with computers, you can design and make almost at the same time, because you’re using the same machine. The same tool that puts a picture on the screen can print it out, so the separation between designer and maker is compressed by the tool you’re using. So in a sense, this use of a computer is more medieval than modern, which is why they are now– they are now still building the Sagrada Familia using computers– in a sense continuing and interpreting Gaudi’s dream of reviving the Middle Ages. It’s a fascinating story. Thanks for bringing it up.

AUDIENCE: Which technique Gaudi used to some sketches I think that, even though he built recently by himself and leads the process, he had some drawings of cathedral.

MARIO CARPO: He had diagrams not blueprints, because stereotomy, which is the medieval technique for cutting stone, you cannot represent it. But there are ways to teach to the artisans how to do it. And in the case of some complicated curves, he actually used catenary models, which we are still using, to determine the shape of some arches and their load. And in that, he was probably bringing in a technology which the medieval builders would not have known. So he was probably tweaking his own rules a little bit. But I don’t think there are drawings by Gaudi.

Diagrams, yes– sketches made after the building has been built, yes– but blueprint, no. Because that is contrary. It goes contrary to his spirit. A blueprint is an instruction, which a thinker gives to a maker. In the medieval scheme of things, that separation does not exist. If you build, you think and make. As an artisan, you don’t have a blueprint. You don’t receive a blueprint to execute. You decide, every day in the morning, what you will do during the day, because that’s medieval. The architectural profession, which I described, was invented in the Renaissance when the humanists, a bunch of snobs, decided that designers should make and that ideas should have an intellectual value which is superior to actual craft, but that is a modern invention.

It was invented in the 15th century with the Renaissance. In the Middle Ages, the separation did not exist. This is what Gaudi had in mind– to revive the Middle Ages, because he did not like modernity even though the money used to build the Sagrada. Familia was being earned by the modern industrial development in Barcelona, but that’s the way it works.

AUDIENCE: Thanks very much for the talk. There was a sense that maybe in the ’80s and ’90s, maybe with CAD/CAM, other technologies, or software, that some things started to look quite similar. Like, cars started to resemble each other. Like, an Audi looked like a BMW or whatever. Do you think, with AI machine learning, that there will be more of that, or it will be that things become more differentiated in their design?

MARIO CARPO: It is true, particularly, as you rightly point out, in the case of car design. These kind of curves, the splines, the mathematics was the same in the software used by all car makers at the time, the software was not. And each software has its own tweaks. So I’m not particularly an expert, but Greg Lynn, one of the– whose work I have shown– was very much interested in car design. He’s a real expert and connoisseur. He tells– I have no evidence of that– but by looking at the body of a car, he can tell which release of a given software was used to make it, and it is true to some extent. There are tweaks of each software which leave traces.

So sometimes you can tell which release of Word was used to compose a certain text, because, you know, cut and paste works in a different way or something. So insofar as that was the early– you can do that because there were probably 25 kinds of software being used in the ’90s by all automotive industry in the world, so if you are in that field, you know them all. Now we are probably moving to an environment where there is much more and faster variations than adaptations, so my guess is that game of recognizing the indexical trace of the tools we’re using is not going to be that easy anymore. Because in the ’90s, there were probably 25 variations of a certain software, and now there are probably 1,000. And then, that game is– so I think that complexity, which is coming through artificial intelligence tools, will make this kind of thing lexical transference of the tool into the traces of its use less conspicuous.

HOST: All right then –

MARIO CARPO: That’s it.

HOST: That’s all we have time for, yeah.

Related Posts

• The Dark Subcultures of Online Politics – Joshua Citarella on Modern Wisdom (Transcript)
• Jeffrey Sachs: Trump’s Distorted Version of the Monroe Doctrine (Transcript)
• Robin Day Speaks With Svetlana Alliluyeva – 1969 BBC Interview (Transcript)
• Grade Inflation: Why an “A” Today Means Less Than It Did 20 Years Ago
• Why Is Knowledge Getting So Expensive? – Jeffrey Edmunds (Transcript)