F E A T U R E S    Issue 1.04 - August 1995

Architects of change

By Hari Hunzru and Jess Search



The object that appeared on the screen was a long way from anything most people would recognise as architecture. Without doors, windows or anything else one would associate with a building, it hung frozen in its 3-D grid like an exhibit in some museum of alien natural history. But the Interactivator is among the first born of a new discipline - evolutionary architecture. Nobody will build a structure so grotesque, so impractical, but architects have much to learn from the process of its creation.

The Interactivator posits that someday, probably someday soon, architects might wish to design a building in much the same way as nature designed a snail's shell: by trying out millions of possibilities and discarding all but the few which are suited to the environment in which they are to function. True, evolutionary architecture relies on only the simulation of sun, rain and wind, and is created in the non-world of a powerful computer. But the computer can try out more snail shells - or buildings - in a few seconds than nature can manage in a century. Tapping into that fecundity may enable designers to perform unexpected marriages of the practical and the unimaginable. It may also fundamentally change the nature of design itself - from creating to curating.

The Interactivator is an experiment which enables anybody with Netscape to try their hand at evolving a shape. As the product of the tinkerings and modifications of a thousand passing surfers, it is little wonder that the Interactivator looks so weird. It can be found on a labyrinthine Web-site created by Jonathan Moberly and Tom Neville. Together they form Ellipsis, Britain's most innovative architectural publishers. But its creator is Professor John Frazer, architectural theoretician, CAD guru, professor in the faculty of Art and Design at the University of Ulster and head of a teaching unit at the Architectural Association (AA) researching morpho-genetic processes. More importantly, he is a man whose self-appointed mission is to bring evolution into the centre of the discipline of design.

Frazer lives and works on a yacht designed by Philippe Starck, sharing the space with his wife Julia, also an architect, and a collection of antique typewriters. But it is the Architectural Association bar that seems like his natural habitat. Around us, knots of fashionable young people pore over blueprints. Slides are being projected on one wall, and above our heads a monitor plays a tape of a bizarre CAD walkthrough that resembles the visuals at Megatripolis. It's worth remembering that Prince Charles once called this place "Frankenstein's Academy". Frazer's wish to give life to inanimate buildings seems eerily similar to the Baron's experiments. Maybe Charles was right after all.

The computer's cornucopia

Evolving architectural designs may be a way, says Frazer, "of harnessing the full profligacy of natural creation." Architecture, he claims, is on the brink of a revolution which will transform it beyond all recognition. The architect will no longer be required to foresee every contingency or calculate every technical requirement of a design. In partnership with the computer, he or she will simply direct the evolution of an idea, steering it in fruitful directions and taking advantage of the variations thrown up in the process.

Although the new techniques have yet to be used to produce finished buildings, they are already being used in the design of industrial components, like turbines. The mathematical engine driving all these developments is the genetic algorithm (GA). GAs are nothing less than computerised versions of Darwinian evolution, incorporating ideas of natural selection and genetic variation. Chromosomes and genes determine a creature's characteristics - inside the computer, chromosomes and genes are represented by strings of bits rather than nucleic acids. A genetic program runs by generating a population of digital creatures and testing each against a simulation of the salient features of its environment. Those that are "fit" - that perform well in the environment - are selected as the parents of a new generation. Just as in sexual reproduction, breeding a new generation in the computer mixes up the genetic material more or less randomly, creating new and unexpected combinations. In the case of turbine design, for example, the most important features of the environment concern aerodynamics. So a "fit" turbine has characteristics that make the air flow more smoothly over its surface. GAs are especially useful for this sort of problem because aerodynamics and fluid flow are more easily simulated than predicted. By simulating - selecting the "fittest" designs, breeding them and simulating again - the designer can make an evolutionary parade of forms march across his screen, culminating in the one with the least drag.

But drag and beauty are of course two very different things. In buildings, the difference between good and bad is often solely a question of aesthetics - not the sort of thing which can be expressed mathematically and incorporated into a simulation. Architects are notorious for relying on intuition in judging their designs: "It just works, that's all". So, given that quantification is anathema to most of his profession, what led Frazer to consider genetic algorithms as architectural tools?

"Italian hilltowns," he says. "No-one designed them. They just grew. And there's something so beautiful about that natural evolution - it strikes everybody." Biological terms like organic and natural have long been buzzwords in architecture, as in other arts. The impulse to capture nature's process of creation has driven many aesthetic projects. "After all, everyone's fascinated by life."

Technology without terror

Frazer's professional evolution began in Cambridge in the 1960s. As luck would have it he found himself sharing late-night computer time with John Conway, inventor of The Game of Life. Maybe there was some memetic cross-fertilisation between the guru of A-Life and the young architect tinkering with form generation. Certainly Frazer has gone on to do sophisticated work with cellular automata, declaring that his work should be considered A-life, even as a form of artificial intelligence. His Cambridge experience confirmed to him that computers were the way forward to alternative ways of creating architecture.

Since Conway and Frazer shared computer time in Cambridge, the discovery that structures of incredible complexity could be generated by the repeated application of simple instructions has led to the disciplines now known as non-linear dynamics and chaos theory. Today scientists of all persuasions are mining these fields for ideas - and concepts of emergent global order, spontaneous form generation and so on are making their way into the cultural mainstream.

Yet when Frazer and Conway were sitting up late in the Cambridge computer centre, watching their results inching out of the graph plotter, their activities were considered marginal, even eccentric. In the computer world, systems scientists were then fixated by the idea of rigid top-down control; in architecture, the Modernists who dominated British architecture with their severe, functional structures were far from amenable to the ideas taking shape after dark in the Cambridge lab.

In the 1970s, however, Frazer concentrated on other ideas. He was at the intellectual - if not commercial - forefront of the revolution in computer-aided design. His company, Autographics, produced a number of pieces of software which, in the British tradition, anticipated the commercial successes of others. He has always advocated creatives getting stuck into the "hard" technical side of computer-aided work. His students at the Architectural Association are encouraged to construct their own components, from light-sensors to full-blown neural nets. "Once you've built your own computer, technology holds no more terror," he smiles.

Part of the reason Frazer first turned to computers was his dissatisfaction with paper and pencil as design tools. As time went on, dissatisfaction with mouse and screen spurred him to further conceptual leaps. "As an interface a computer terminal is so ... two dimensional," he says. To accuse someone of "two-dimensional thinking" is the architect's insult par excellence. (Frazer has, for example, been known to dismiss graphic designers as "flatheads".) Seeking to avoid this terrible fate led him to experiment with various 3-D tools for thinking. One avenue of exploration was "intelligent modelling" which, in essence, involves embedding chips and sensors into building blocks. The designer works in 3-D, and his modelling-kit keeps examining the positions of its various parts. It transmits a constantly changing record of its state to a computer, which then converts it into a CAD document.

Given his background, it was perhaps inevitable that Frazer would eventually turn to genetic algorithms in his search for an evolutionary architecture. His research groups at Ulster and the AA, which have over the years involved a large number of architects who deserve to be regarded as innovators in their own right, began using evolutionary modelling to solve relatively simple problems.

One early test of the application involved evolving columns to fit the rules of proportion laid down by James Gibbs in the 18th century. "Unless we told it to select for other things," Frazer recalls, "the programme took just a few seconds to converge on a perfectly proportioned Gibbsian Tuscan column." With evidence that genetic algorithms could work, they went on to look at the design of racing yacht hulls - something close to Frazer's heart. Using complex calculations concerning displacement, hull area, waterplane and so on, they bred a series of increasingly high-performance hull designs.

At one level, both columns and hulls can be approached in the same way as turbine design. Success or failure is easily measurable: the column either follows Gibbs or it doesn't; the hull performs well or it doesn't. But Frazer's group introduced a new type of evolution into their models, allowing unquantifiable human factors to influence the direction of their algorithms. They did this simply by allowing people to intervene directly in the process of evolution. With the yacht hulls, the designer could influence evolution so that good-looking hulls or ones which worked best with a particular deck layout, would be chosen. At last, aesthetics had fused with computation, and genetic algorithms began to look like a useful tool for all types of design.

Current evolutionary models fall into two distinct types. "Convergent" models allow the computer to run its course, gradually optimising a design where the criteria for success or failure are known. "Divergent" models produce a wide range of mutations, allowing the designer to choose from a limitless range of ideas. In divergent evolution the computer becomes an aesthetic prompt, an ideas factory which the designer uses as a basis for creation. Genetic design uses aspects of both types. In the architectural practices of the future, quantifiable aspects of a design such as earthquake resistance or ventilation flow might be convergently modelled, while the shape of an arch or the layout of a roof may originate in a divergent model. The partnership of human and computer is, says Frazer, one of the most exciting aspects of evolutionary architecture, the almost symbiotic relationship giving the lie to the worries about "dehumanisation of creativity" that some designers still express about computers.

Let the sun shine in

Frazer's group has already tested convergence in some areas of architectural design. With the yacht project under its belt, his group turned its attention to the surprisingly complex architectural problems caused by the sun. Creating the right balance of light and shade is a primary consideration in any design, but it can be far more than a question of aesthetics, especially when building skyscrapers. Anyone who has sat by a plate-glass window on a summer's day or attempted to use a computer screen with the sun streaming in through the window will know why. For many years architects have used complex diagrams to plot the course of the sun over their sites, then decide where to place the shades and windows. Unfortunately these two-dimensional representations are confusing, and are frequently interpreted incorrectly. Even giants like Le Corbusier and James Stirling have made serious sunshading mistakes - arising from basic errors, such as confusing the type of projection on their plan. Failure can land the architect with a permanent reminder of their arithmetical deficiencies.

Having studied the problem and built a number of conventional computer simulations, Frazer's team turned to an evolutionary model to design efficient and accurate methods of sunshading. A circular form was allowed to grow, responding to the movement of the sun over its simulated surface. Protecting an inner core, the structure's surfaces moved outwards in the direction of greatest solar radiation, with edges and planes curving up to shade exposed areas. The results were beautiful, functional and very different from the shade shapes a human designer might have chosen intuitively. This simulation software, and other sunshading tools developed by Frazer's team, are being assessed by architectural practices throughout the world.

Although these results are promising, modelling for sunshading is one of the easiest tasks evolutionary architecture might be expected to perform. The eventual goal of Frazer's team is to produce a complete and accurate simulation of the environment in which a building is going to exist. Designing such a simulation will be an incredibly complex task, since a building's environment can encompass everything from the climate to the number of people who are going to use it.

Still, the more visionary in Frazer's circle see their models eventually taking into account the social, economic and political context in which a structure will "take place". This seems, to say the least, a tad ambitious. "We're a long way off from that," Frazer admits. "But we're gradually increasing in sophistication. I see no reason why such things should not, in principle, be possible."

Bottles and other monsters

Frazer is not alone in recognising the potential of evolutionary applications for design. Computer artist William Latham (see Wired 1.03) has been using genetic algorithms to create his trademark organic forms for some years. At IBM he worked in partnership with programmer Stephen Todd to develop the Mutator interface, a general-purpose GA tool which may soon become the first commercially available piece of evolutionary software. The Mutator can run evolutionary processes on any type of data, from 3-D form generation to stockmarket figures. Recently Latham and current programming partner Mark Atkinson have been using it to solve simple design problems. Their ambitions are less grandiose than those of Frazer's team - there is no attempt to output anything more complex than sketches to form the basis of designs. Yet the simplicity of the Mutator (it runs on a PC) and the speed with which it generates evolutionary sequences indicate that GAs may soon be used, not just in esoteric professional applications, but as part of standard software.

Latham demonstrates a sequence involving the design of a shampoo bottle. It's a simple point and click operation, so simple in fact that it is easy to overlook the radical new way the user is working with the computer. From an initial wide-bellied, slim-necked shape familiar to users of anti-dandruff shampoos, we are soon presented with a vast array of interesting options. Latham starts breeding the weird ones, and soon we're faced with something that probably ought to be housing a genie instead of helping combat those flaky scalp problems. But we get the picture. When he gives an impromptu workshop in how to breed profits by mutating corporate spreadsheets, we begin to understand why he is being courted by business, scientific and military interests worldwide.

But Frazer, at least, sees no reason to stop with simply evolving better business plans and architectural blueprints. Why, he asks, should evolution stop at the drawing board? If architects really can understand the key factors that make a building "fit" its environment - and, through simulation, also understand the sorts of changes which make it a better or worse fit - then why can't architects build into their constructions the ability to keep changing and evolving even after they are built? Frazer has written excitedly of "a new form of designed artefact interacting and evolving in harmony with natural forces, including those of society". The holy grail (and here he shows the influence of heroes such as Buckminster Fuller and Nicholas Negroponte) is to produce buildings which will evolve in use, adapting themselves to their environments and the needs of the people who live and work in them. If your building can go on perfecting itself after you have built it, then there is less pressure on the architect to anticipate every contingency, to foresee the unforeseeable. This is nothing less than the transference of intelligence from human to building, a vision of houses and office-blocks as artificial life.

Sensitive skin

This vision, it must be said, also owes more than a little to Cedric Price's never-constructed Generator project, on which both John and Julia Frazer worked as computer consultants. Price was asked to produce a building which would meet the changing needs of the Gilman Paper corporation. On a site in Florida, Price proposed to build a grid of foundation pads and a kit of parts to enable enclosures, screens and gangways to be arranged and rearranged with the help of a permanent on-site crane. Electronics were to be embedded in every component, so the building, like Frazer's electronic modelling kits, could transmit an image of its state at any given time. The design even incorporated a concept of "boredom". If the Gilman employees did not change it themselves, the structure would suggest possible reconfigurations, based on previously successful versions of itself.

The Generator, with its modular components and microprocessors, produced a host of new ideas - and, through the massive media coverage it spawned, gave the world the phrase "intelligent building". Since then, however, little progress has been made. When asked what architectural developments show promise in this area, Frazer points to the work of Malaysian architect Ken Yeang, a former student at the AA who has become known for building "bioclimatic skyscrapers", eco-buildings which "interact" in a primitive way with the environment around them.

Instead of air-conditioning, his buildings collect wind, and channel it through wall and floor ducts. They are often clad in semi-permeable skins which regulate heat loss. Atriums, openable windows and "skycourts" encourage people to interact with the world outside. One spectacular tower, the Menara Boustead in Kuala Lumpur, is almost entirely covered by green plants, which give off oxygen and disperse CO2 from the polluted street below.

A genuine evolutionary architecture will require a new generation of materials, which, like the plants with which Yeang covers his buildings, can themselves adapt to changing conditions. Structures built from inert matter are doomed to remain more or less static. Frazer's students are working on various projects, including light-sensitive skins and flexible membranes that will enable future buildings to react to their environment. But for now they have just a collection of prototypes and working models - and high hopes.

Assume for a minute that those high hopes are justified, and that the practical difficulties can be overcome. The idea of evolutionary architecture still begs a basic question: if buildings evolve themselves, what is left for architects to do? Welsh design radical John Chris Jones argues that the very idea of designing will change fundamentally. The idea of a final perfected product, be it a toothbrush, a block of flats or a motorway interchange, will be abandoned in favour of a conception of dynamic process. The designer will be a facilitator rather than a controller. It's no wonder Jones is Frazer's favourite architectural theorist.

Evolutionary architecture is the digital embodiment of this dream of an unfettered, unegotistical creative process. Just as architecture defines people's environments, so too do people define architecture's. There seems no more logical way to achieve harmony than to set the two evolving together.

Hari Kunzru (hari@juju.demon.co.uk) is a London-based freelance writer with interests in art, technology and culture. Jess Search (jess@soul-pp.demon.co.uk) is a journalist and television researcher. The Interactivator is at the Ellipsis site - www.gold.net/ellipsis/ John Frazer's (100415.1704@compuserve.com) An Evolutionary Architecture is published by the Architectural Association.