Designs iterate over time. Architecture designed and built in 1921 won’t look the same as a building from 1971 or from 2021. Trends change, materials evolve, and issues like sustainability gain importance, among other factors. But what if this evolution wasn’t just about the types of buildings architects design, but was, in fact, key to how they design? That’s the promise of evolutionary algorithms as a design tool.
While designers have long since used tools like Computer Aided Design (CAD) to help conceptualize projects, proponents of generative design want to go several steps further. They want to use algorithms that mimic evolutionary processes inside a computer to help design buildings from the ground up. And, at least when it comes to houses, the results are pretty darn interesting.
Generative design
Celestino Soddu has been working with evolutionary algorithms for longer than most people working today have been using computers. A contemporary Italian architect and designer now in his mid-70s, Soddu became interested in the technology’s potential impact on design back in the days of the Apple II. What interested him was the potential for endlessly riffing on a theme. Or as Soddu, who is also professor of generative design at the Polytechnic University of Milan in Italy, told Digital Trends, he liked the idea of “opening the door to endless variation.”
The prospect of generative algorithmic design is, on the surface, relatively straightforward. There are certain “rules” that define particular objects, be they a bungalow or a Baroque cathedral. A genetic algorithm replicates evolution in the form of a computer program as a way to optimize solutions. By defining what Soddu considers to be the rules that define a particular piece of design, his algorithms can conceptualize what objects might look like if they were living entities that had undergone millennia of natural selection.
He started out designing a “species” of generative Italian medieval towns in the 1980s. The work generates infinite 3D models of medieval cities, which each one slightly different. Since then, Soddu has continued to experiment with generative design as part of his practice. “One of my last projects is a proposal for the recovery of the Notre Dame pinnacle in Paris,” he said. The design (which, unfortunately, will not be used) depicts the fire-destroyed spire as almost a twisted inverted icicle, thrust skyward.
Part of the appeal of Soddu’s designs is the element of randomness that the algorithm introduces. But it also opens up intriguing possibilities that could be applied not just to architecture, but any kind of design. For instance, a design agency could create hundreds of chairs, with each one slightly different. He refers to this kind of design as “idea-products.”
Creative technologist at work
Soddu isn’t the only person interested in using genetic algorithms to imagine buildings, either. The Lisbon, Portugal-based musician and creative technologist Moullinex recently created a music video featuring an undulating landscape of brutalist buildings. At first glance, these appear to be real buildings with a morphing effect allowing one to transition to the next. In fact, they are dreamed up by generative algorithms.
“A Generative Adversarial Network — in this instance, StyleGAN2-ada — was trained with images of real buildings and is then able to create new images based on what it’s learned,” Moullinex, aka Luís Clara Gomes, told Digital Trends. “What you’re seeing is the network’s interpretation of what the data is. As an analogy, this would be like showing a child pictures of cats and then asking them to create a new cat based on what they’ve learned. The more cats you show — [meaning] the bigger the training dataset — the more catlike the generated images will be.”
Since creating the initial music video, Moullinex has continued to work on the project, and remains intrigued by the possibilities.
“I like to look at GANs as an oven: You source and shape the clay, let the batch cook, and wait for the results,” Moullinex said. “What pieces you pick and how you present them is again up to you. It’s an exercise of giving up some degree of control and leaving some things up to chance. As someone who’s been following this area of technology for years … I find it fascinating that our source of entropy, chaos and unpredictability — all good ingredients for inspiration and creativity — is coming from technology.”
One of the big questions with any kind of alleged creativity A.I. is whether it takes away from the human designer. No one would ever credit paints or gravity as an equal author of a painting as the artist, but with A.I., it’s not quite so straightforward. But Soddu isn’t worried.
“No, on the contrary,” Soddu said. “Creativity, quoting [French mathematician Henri] Poincaré, is the ability to interpret what exists by proposing a new system of relations between the parts. It is certainly not only the search for new forms. My generative project is born … from the subjective interpretation of the past using original genetic algorithms capable of representing this subjectivity. The digital evolution is an incredible occasion for the subjective idea to be represented in its multiple and infinite variations.”
Coming to a street near you?
To date, evolutionary algorithms have remained a fascinating sidebar to mainstream design. In the same way that A.I. ethics were an abstract idea until, suddenly, the world needed it, generative design has been an interesting area to discuss, but with seemingly minimal application.
The problem isn’t that genetic algorithms produce impractical designs. NASA has utilized satellite components designed by genetic algorithms on expensive, real-world missions. Not only were these components effective, they worked better than alternatives designed by humans — and the human engineers who looked at them couldn’t figure out why. As Moullinex points out, there is therefore no reason why a genetic algorithm couldn’t be optimized to take into account requirements like functionality, aesthetics, cost, sustainability, and ethics.
Instead, the problem may be that generative algorithms were a software solution without accompanying hardware. That may be changing. Recently, Digital Trends wrote about a multiyear project seeking to use evolutionary A.I. to design robots for exploring other planets. The idea? That a 3D printer setup could be launched to another planet to print the best robot designs, no humans required.
We’re now living in a world in which 3D-printed housing exists. These houses, which can be manufactured rapidly and sold at a price below that of “regular” houses, could conceivably lead to a revolution in affordable mass housing. And you know what? Thanks to genetic algorithms, each one could be slightly (or very) different.
