Three years after it was initially announced at E3 2013, No Man’s Sky has officially gone gold. Few games in recent memory have created so much buzz. The near-infinite universe that No Man’s Sky offers is plenty reason to be excited but the way Hello Games is creating that universe using procedural generation is equally inspiring.
Procedural generation is a type of design methodology in which components or elements are handcrafted and each given unique properties. These components are then fed into a database which generates forms based on a series of rules and regulations put into place by the designers that dictate what components and elements are used, combined, and generated. Hello Games has created its own periodic table of elements for the universe of No Man’s Sky and assigned properties unique to each element. This allows them to generate 1.8 quintillion planets that are all different from one another and built governing the inherent logic set forth. For example, a certain planet may be in a solar system with two suns, those suns may cause the planet to have an atmosphere composed of a certain gas. That gas may reflect the visible light spectrum different than our own, causing this planet’s sky to be red. The atmosphere would then dictate the planet’s temperature and radiation levels, which in turn affect the flora and fauna that can survive on the planet.
If procedural generation can allow the small team at Hello Games to make what should be the biggest videogame ever in terms of experiential space, what implications could it have on other design fields? Architecture as a discipline has struggled to manage and understand virtual design. As a result, architectural pedagogy has been undergoing a massive paradigm shift for what has seemed to be the last 40 years. Even now, in 2016, if you attend graduate school at a more traditional East Coast university your introductory studios will revolve around analog architecture, forcing students to sketch and draft by hand. Attend graduate school at a more contemporary university and you can make it through three years of school without ever putting graphite to paper. While computer-aided design software has been around for decades, programs like AutoCAD, SketchUp, and Rhino did not drastically change the nature of architectural design but rather sped it up, made it more efficient, more precise.
The true paradigm shift that occurs in architectural pedagogy comes from the proliferation of parametric design. Parametric modeling has a lot in common with coding and consists of interdependent parts so that one change creates a rippling effect. The two are so similar, in fact, that John May, a faculty member at Harvard Graduate School of Design, believes that coding, not physics and calculus, should be required for all architecture students. In this way, parametric design has a lot in common with procedural generation. In both instances, the designer is not designing form so much as she is designing regulations and constraints that then dictate form. While the two methods share some similarities, they are not the same. The most notable difference comes in their classification, parametric design versus procedural generation.
At its heart, parametric design is still design, it creates one intentional form that adheres to all the fundamental principles of design that have been taught to architects for centuries, only the toolset has changed (albeit, that change requires a new pedagogical approach). A script is designed and it in turn creates a form, but that script is always in service to a particular form. Procedural generation on the other hand functions quite differently. Forms are not designed, they are generated, meaning that there is no longer a strict intentionality behind form nor is there one specific creation. The script and constraints are designed, but they are designed to serve themselves, not the final form. Procedural generation relies on a series of inputs that are capable of generating multiple outputs.
A former classmate of mine, Iman Ansari, who is founding principal of the firm AN.ONYMOUS, lecturer at the USC School of Architecture, and a PhD candidate at UCLA School of Architecture and Urban Design, is able to talk about the implications of No Man’s Sky and procedural generation on the field of architecture. He mentions how architects have always been inspired and influenced by nature and the laws of the natural environment, and how this is slowly beginning to change. “What is exciting about the time we live in now,” Asari said, “is that, beyond the availability of all kinds of alternative or artificial materials and fabrication techniques, we are also able to rethink and reconfigure the rules and conditions that enable and inform them in the first place.” In other words, architects are no longer beholden to conventional rules meant to maintain a natural order. “It is no longer about designing a final object or a product, but about designing or configuring the system or the process of their formation—the underlying code, algorithm, or procedure that can generate not just one but multiple outcomes.”
While Hello Games may have designed certain elements (leaves, ears, tails, horns) the true design came in the properties assigned to those elements and the ways in which they work in tandem with all other components of the game. Yet, as Ansari points out, this isn’t drastically different from traditional architectural design. “Design is a dirty word in architecture. Typically, the word implies an arbitrary move or decision-making process primarily concerned with form-making. In my view, ‘design’ is the outcome of a process or procedure, generated from a set of predefined rules and operations—whether defined by the architect, the city, the site, the program, or other factors.” Traditional notions of design have changed according to Ansari. What had been considered a harmonizing mathematical process in the Renaissance and Early Modern Era became an algorithmic process in the last 40 years. As a result, choices and rules set out at the beginning impact every subsequent choice.
Often, the most difficult design projects in architecture schools are the ones that give the fewest restrictions. Architecture is not meant to be free from constraints, it is meant to adapt to them, reason with them, and respond to them. Ansari notes that videogame design will always be unique because it is liberated from many of the rules and restrictions that the physical world relies on. With games, there are no pre-existing conditions, no real world context, and no site, they belong in the realm of the imaginary. The creative process relies just as much on self-imposed constraints as it does on real-world constraints. In many ways, designing a videogame can be more difficult than designing a building because game designers are only left with their own self-imposed rules. This same reason is why it can also be much more appealing to designers. Videogames can produce environments and buildings that are impossible to conceive in the real world.
While the tools that videogame designers and architects use have begun to overlap, the use of those tools will never be quite the same because of the inherent nature of the physical world architects must respond to. “Despite all the possibilities that digital tools provide, the architect always has to scale, trim, or orient the object to accommodate and situate it in within the real physical realm,” said Ansari. “This is something that architecture—or at least built architecture—can never fully renounce. In other words, for architects, there is always a given: a physical, material, legal, and functional environment that architecture has to respond to.”
But what would happen if, one day, procedural generation could respond to pre-existing environments? What if one day Mr. and Mrs. Smith could pull up a website, enter a few restrictions and regulations (square footage, climate, lighting, dead load), and out pop 20-30 different homes that all fit their desired constraints? The truth of the matter is that this is a reality which really isn’t that far off. The idea of a Do-It-Yourself home has been around for decades. Domestic architectural catalogs allowed people to pick and choose suburban home designs for communities like Levittown, Pennsylvania. Today, sites like floorplanner.com and plygem.com allow users to design spaces without the help of an architect. I asked Ansari if he thought procedural generation had the potential to be a part of architectural design. “The ‘design’ phase of any project is really about five to 10 percent of the whole work of an architect,” he said. “The rest includes research, development, documentation, representation, engineering, fabrication and all kinds of different tasks that fall outside what we would consider to be the design process.” These extra tasks are time-consuming and make up the bulk of any project. Ansari seemed less interested in procedural generation’s role in design and more in its ability to automate many of these tasks that architects are less inclined to want to do.
He is right, even in school when details and logistics are less important, the most “fun” part of any project is the formulation of the idea: the creative theorizing that leads to the final concept. That is what architects want to do, conceptualize, theorize, and rationalize form, and leave the rest of the stuff up to the engineers and contractors. Very few architects enjoy the process of figuring out exactly how a wall meets the floor, or what the proper window thickness should be. Even fewer enjoy the process of finding those windows in real life. If procedural generation could automate these tasks, architects could focus more of their time and energy on the creative process.
No matter how many rules are set in place or how many regulating constraints are used, procedural generation always implies a certain level of randomness. This randomness can be mitigated to a certain extent but the truth of the matter—in fact the very point of procedural generation—is that the result will be unpredictable. In a universe with 1.8 quintillion planets, a certain degree of randomness and unpredictability is a good thing, it is necessary. But in a singular building? Well, that would be a bad idea according to Ansari. “Randomness is a relatively contemporary concept in design and architects use it a lot in designing facades, floor or ceiling patterns, or other parts of buildings,” he said. “But to use randomness for the building itself is a whole different story.” Buildings always have a clear logic—a programmatic logic, a circulation logic, a structural logic, an environmental logic, a material logic. Randomness can never fully serve as an organizing principle in architecture. “To use language as an analogy here, you can arrange words randomly on a page but they will almost never produce or generate meaning,” Ansari added. “My concern is that if randomness is taken as a design approach in architecture it would produce the same result: it would generate a lot of objects, or at best even buildings, but no meaning, and therefore, no architecture.” Architecture is more than just walls, doors, and windows, just as a novel is more than words. Not only must all the disparate elements come together in a cohesive manner but they must also be produced with purpose and intention. “Maybe one day, computer software would be able to overcome this and produce meaningful forms,” Ansari said. “But at that point, I assume, they would also be able to compose music, write poetry, make films and do everything else. It is not going to be just the architects they would replace. We would all be out of jobs.”
Just as a building must relate and respond to its site, so must it too respond and relate to itself. There is a difference between a building and architecture. It’s one of those “all architecture are buildings but not all buildings are architecture” kinda things. Architects often use the term “Architecture with a capital A” to differentiate between a building and architecture. Architecture has meaning and that meaning often comes from a human element, an intentionality, a sense of purpose, and most importantly an authorship that is necessary in all forms of art. A computer can write a novel, but is it any more than a string of coherent words positioned on a page? Even if, someday, computers are able to compose better songs, write better books, and design better buildings, would we want them to? Art is all about authorship. Computers may one day be able to imitate consciousness or even meaning, but as long as they do not actually have consciousness, their work will always be devoid of intent and true authorship, for that is something unique to the human mind.
When I asked Ansari if he thought we’d ever see the day when architects could be replaced with software he echoed a sentiment that I feel deeply as well. “To be honest I don’t want to live in a world without architects,” he said. And why would he? For a world without architects is a world without painters, musicians, filmmakers, sculptors, and poets. That is a world I hope to never see.