Op-Ed: Can AI save the world from unspeakably dull architecture? Maybe. Or make it a lot better?

Anything more boring, lifeless, and drab than “modern architecture” would be pretty difficult to achieve. LEGO bricks have more style and character.  Everything looks like a clinic or just another minimalist desert. Stick a door in a KFC bucket, and you’ve taken care of the aesthetic requirements.

Let’s be clear. Architecture knows these problems all too well. It’s more or less stuck with them. A lot of this astonishing underachievement is of course cost-based. Some genius has saved $1 per square metre on the construction of a multi-purpose future slum and sentenced the world to looking at it for the next 40 years.

AI is inevitably getting involved in architectural design from the ground up. There’s an irony here in that the AI may be more architecturally literate than whoever or whatever it is that designs these lifeless eyesores.

News about AI designed architecture is the predictable mix of techno-bafflement and the inevitable confusion between AI architecture and AI designed architecture. Nothing like lousy contextual searches to really help you research the subject.

What’s emerging from this terminological straitjacket is at least a bit encouraging.

Architects please excuse some very basic background here:

Modern architecture is designed using AutoCAD. This is the mental equivalent of being attacked by a deranged physiotherapist during a car crash at the training level. Five seconds of exposure to AutoCAD will explain why “brutalism” isn’t so much a form of architecture as an unavoidable mental state.

The construction sector is no charity and no admirer of architectural inspirations. These buildings all look the same because they are the same, coming off the same palette and the same bottom line.  There will be no Gothic bus stops or palatial Georgian kiosks from this direction.

The current generation of monotonous monstrosities share an architectural heritage with the whole of human history. The main difference is size and scale. They’re the same basic ideas as the endless rows of terrace houses of the 19^th-century or 20th-century tenements. You’re not going to see a conical pyramid in this design hamster wheel. Your architect is more likely to be Frank Lloyd Might than Frank Lloyd Wright.

Architecture vs AI is an all-embracing major issue. There’s a particularly useful article in Common Edge which cites architecture’s rather staid approach to emerging tech and the major issues with “tech regression” that sum up the problems, particularly the creative issues and human vs AI design. You may not agree with all the points, but it’s a good summation.

Exploring the unknown with AI? Not quite yet.

It’s debatable whether any technology since the wheel has come with as much utterly useless metaphysical dreck as AI. The technology is actually being slandered by its advocates. It’s being presented as simple when it’s anything but. “Anyone can do it” couldn’t be more inappropriate or more misleading. You’ll note that it’s the duly accredited experts that have to undo the damage when “simple” AI makes a mess.

That’s not helping architects to get a grip on what could be a very useful tool. Architecture is a very demanding practical discipline. Verbosity isn’t. If it’s not practical, it’s not much use when you’re designing a building. Try getting approval for a metaphorical building. Then try explaining that to the builders.

The definitive denouement of AI hype could well be when architects tell the hype where to go, which they’re already doing to some degree. There’s a surprisingly good reason for that. To explore AI design possibilities, the output has to stick to verifiable guidelines and specifications.

Now the good news:

AutoCAD is cross-platform. It can level the hype and slot into any AI configuration and there’s not a lot of LLM-fodder in the core design dynamics. How do you misinterpret a building design, or hallucinate the almost maniacal level of depth and detail in specifications in building design? Every component in modern architecture is effectively audited before it enters plans. AutoCAD will win that round with ease.  

Architects are already demanding drastically upgraded AI design policies. The Australian Institute of Architects has made a very unambiguous call for “reforms designed to steer AI towards broader public value outcomes” in a submission to the Australian Senate. They mean social outcomes. That’s a long way away from just sticking up future barrios. If you know Australia’s weird, expensive, Monopoly-like approach to the building sector, that’s a true revolution. It’s also a very commendable effort on the part of the architects, calling a spade a spade.

Even the architectural competitions are amping up their criteria and looking ahead. By way of example, check out the CartesianForms: Architecture Reimagined – Global AI Design Competition on the Parametic Architecture website. This competition is in conjunction with OpenAI, and the theme is multigenerational architecture, no less, with a timeframe of 2030 to 2080.  

Where architecture has got AI drastically right

Architecture is an incredibly ferocious competitive field. It’s a portfolio market. You’re as good as your latest building. Scale defines your business. You’re competing with nothing but experts.

It’s no coincidence that architecture has clearly gone straight for AI’s strongest suits. Never mind chatbots. Precision, depth, and high-value data at all levels are exactly the areas in which AI excels. A personable AI agent doesn’t deliver the approvals for construction or sign the contracts. Architecture as a business is all about delivery, not verbosity.

This is where architecture and AI are instantly on the same page. AI is much better at generating product than meaningless drivel. That combination will work and work well.

There’s another critical issue here which has gone totally under the radar.

AI is a good “what if” tool. It can chase ideas. It can make them visual and tangible. You can doodle with it. You can use it as scientific AI is used, exploring possibilities and combinations of ideas. You can test anything and everything. You can even cost it accurately.

Say you wanted to build a city that looks like a coral reef. You want to build in a local ecology in an environmental context. You want support systems, water, power, etc. You want multiple designs and strictly compliant town planning criteria, line by line.

Now imagine doing an architectural draft of all of that in less than a lifetime on a realistic budget. AI can. As a creative tool, AI isn’t the enemy. It’s the answer.