The architects teaching AI to print cities
From AI-designed buildings to large-scale 3D printing, cities of the future could be more machine than man
The architect Arthur Mamou-Mani grins as he tells me about parametric design.
“Instead of coming up with a final form,” he explains, “or drawing that final form like a sculptor, you define the parameters. You say ‘my building can only move by 5m, can only have a thickness of X. Here are the load figures, this is the amount of sun, this is the amount of wind… boom.’ Then you let the computer come up with solutions within that space.”
Mamou-Mani’s practice – also called Mamou-Mani – specialises in a new, emerging form of architecture. Instead of handcrafting a building from the ground up, his team works with specialised software to create computer-generated structures. Mamou-Mani likens the process to evolution, where the software – much like nature – adapts its designs to best suit a given set of parameters.
“Basically, you have this natural evolution [process] in the computer, then you see what’s called the solution space,” he says. “You might be really unsatisfied with the results, but at least you’ve got a thousand options. So then you go back, generate another thousand options. The [human] designers define the system, the parameters, and let the computer help them to understand what the constraints are. So you’re really collaborating with the computer.”
What sort of buildings do computers generate? Interestingly, many of Mamou-Mani’s designs look like strange plants, or the fossilised bones of ancient crustaceans. There’s an organic quality to the rippling structures, often made of hundreds of interlacing grids. For example, the studio’s Wooden Waves installation – winner of a Gold award at 2016’s American Architecture Prize – is a sinuous plywood lattice, “topologically optimised” to be supported from the ceiling.
Computers played a large part in Wooden Waves’ design, but it looks like it has grown across the room like an organism. Indeed, this new form of AI-architect collaboration is characterised by designs that appear much more biological than one might expect from a machine. The vast, cavernous lattice of Terminal 3 of Shenzhen Bao’an International Airport, for example, was designed by Massimiliano Fuksas in collaboration with Knippers Helbig, and was created with the help of parametric design. Mamou-Mani says this aesthetic is to be expected – after all, the only other forms to have evolved to suit their environments, albeit over billions of years, are natural forms.
Striking a balance between human- and machine-led architecture isn’t only a matter of physical parameters, however. The ethics of computer-created buildings come into play when you consider the purpose of what’s being made.
For example, Mamou-Mani tells me about a shortlisted entry his studio worked on for the Holocaust Memorial in London with Heneghan Peng Architects, as parametric consultants. “There, the idea of letting a computer do it sounds crazy and completely irrelevant.” Instead of giving the software free reign over its parameters, the overall – highly symbolic – design of the monument is predetermined by humans. However, the pattern made by the gap between the bricks is calculated by the computer as the result of the intersection of many grids, a radial grid facing the Houses of Parliament, one parallel to the ground, and another parallel to the tapering walls. The result is a complex and intriguing moiré pattern, similar to computer glitches. It’s a collaboration with the computer, but here the machine is relegated to heavy-lifting.
When does a human response to a space stop being a human response?
“The computer is here to help you deal with a lot of information. In this case, a lot of bricks,” Mamou-Mani says. “Where we position those bricks to create that pattern was influenced by the computer, but the story is about human loss, and the gaps between those bricks will be used for visitors to listen to the voices of the survivors.”
A Holocaust memorial is an extreme example of how a computer-led design runs the risk of becoming tasteless – a risk Mamou-Mani seems keenly conscious of – but it’s an issue that ripples throughout architecture. To what extent can a computer address a brief? When does a human response to a space stop being a human response?
A machine might compute exactly where to place the bricks in Mamou-Mani’s example of the memorial, but ultimately, a human hand is likely to do their laying. Another company, however, is applying AI to the other side of construction, experimenting with the potential for machines to build large-scale structures using artificial intelligence and advanced 3D printing.
“Traditional construction methods rely heavily on human labour, which is inefficient, wasteful and unsafe,’’ says Dağhan Çam, co-founder and CEO of Ai Build, a firm developing AI and robotics for construction.
“By automating the materialisation process, we aim to maximise savings in time, material and costs, as well as enabling designers, architects and engineers to create entirely new, complex designs that were impossible or too expensive to build before. We believe that automation and AI-powered autonomous manufacturing robots are essential for the sustainability of our future-built environments.”
3D printing has progressed significantly over the past few years, but it remains a tool best suited to manufacturing small-scale objects. A main issue facing large-scale 3D printing is the difficulty of accurately simulating material behaviour – miniature mistakes may be insignificant when printing a small object, but broaden that to something building-sized and those tiny errors can quickly become dangerous structural problems.
“One way to minimise these failures is to slow down the manufacturing process – which isn’t a particularly smart solution, because time costs money,” says Çam. “As an alternative, we’re using a combination of simulation and real-time decision-making algorithms to compensate for minor mistakes that might have accumulated into more severe structural failures if they weren’t taken into account. In this case, rather than the system becoming an architect, it’s becoming a builder with greater flexibility to handle unexpected problems.”
At the moment, Ai Build is only working on human-designed, installation-sized structures made with thermoplastic composites, but Çam tells me he has plans to allow other manufacturers working with metals and concrete to benefit from the technology. Indeed, AI Build isn’t the only company pushing building-scale 3D printing. San Francisco-based firm Apis Cor has been working with Russian real-estate developer PIK to make affordable concrete bungalows, while Dubai-based startup Cazza has announced somewhat ambitious plans to 3D-print a skyscraper in UAE.
Regulation is naturally a major hurdle for 3D-printing developers to maneuver. Convincing regulators that robot-built bungalows, let alone skyscrapers, are structurally sound will be a challenge given the current issues with ensuring structural consistency. If these barriers can be negotiated, however, 3D printing stands to have an enormous impact on construction. Robot-made structures could provide affordable emergency accommodation, or allow plans to be drawn up for buildings that are currently too technically challenging for human hands.
Not all of the potential impacts are positive, however. Labour costs could be reduced, yes, but the flipside may be job losses across the sector. Traditional building firms could be undermined by cheaper, faster robotic competition. While these issues look to be some way off, they’re serious considerations when figuring out how our cities of the future will be built.
“It’s not a question of if it will happen, but when it will happen.”
Brand-new urban development in the UK has dwindled in the decades since the New Towns Act of 1946 (along with experimental designs such as Geoffrey Jellicoe’s car-based Motopia), but China, India and Kenya – among other emerging countries – are very much building cities to house rapidly urbanising populations. Will tower-block-scale 3D printing become commonplace in these new urban sprawls?
“If we speculate about 50 years from now, there’s no doubt that our cities will be built by autonomous robots and drones,” Çam tells me. “If we think about the next five years, we’ll start seeing practical applications of autonomous additive manufacturing being used and bringing value in established industries such as construction, aerospace and automotive. So it’s not a question of if it will happen, but when it will happen.”
The natural extension of all this, of course, is what happens if we combine robotic builders with AI architects? If machines can design and construct their own cities, will we end up in a position where our homes and workplaces, our streets and parks, are dreamt up and put together by some benign artificial intelligence? What would an AI-built city even look like?
“A city thought up and built by AI would be similar to nature,” says Çam. “Not visually, but philosophically. Living organisms go through millions of years of evolution to reach the level of sophistication and adaptation to their environments. This process can be accelerated by computational power to design and build cities that are complex, adaptive and resource-efficient.”
You don’t have to look for a computer-built metropolis to see how basic AI is being used to increase urban efficiency. Existing cities are already being made “smart” thanks to sensors and algorithms angled to make infrastructures run effectively and economically. One big application in this field from an architecture point of view is “space syntax”, which measures how people move through cities, allowing you to optimise the paths between different areas. Extending the nature metaphor, “smart” technology makes it easier to measure movements of citizens, and subsequently “evolve” in a way that optimises those movements.
The suggestion is that an AI-guided city would be extremely optimised. Like an enormous organism, it would find a structure that best suits its environment, where movement of cells is at its most efficient. Such a model, however, doesn’t have a great deal of room for you or me, and our individual, subjective experiences. As Mamou-Mani tells me, building with computers is a delicate balance between humanity and the machine.
“We need poetry. We need a certain dose of freedom to choose.”
“You might say, okay, that solution is more optimal,” he says. “However, optimal for me might be very different to you. I’m French, so I think of the flâneur; people that walk around aimlessly, that learn from the beauty [of the city]. Beauty comes without a definite aim. Is that optimal? No, but we need it. We need poetry. We need a certain dose of freedom to choose, and to not only be defined by what the computer says.
“I feel we should really trust our personal stories, our gut feelings. Otherwise, if we compute buildings and they end up being horrible places to live, we’ll just throw ourselves from the top of them.”
Image credits: Mamou-Mani, Heneghan Peng Architects, AI Build