AI and architecture: How is it being used and would you trust it to design your home?

0
AI and architecture: How is it being used and would you trust it to design your home?

Four words: ‘Ideal Australian suburban home’ and 10 seconds was all it took for AI to produce that image above.

Could it be a vision of your future home? You have to admit, it’s something that looks eerily familiar, like it might be taken from the cover of any design magazine.

And this can all be done on your phone while sitting on the couch. You can literally create an image of your new home between ad breaks on the telly.

To some degree, that’s the promise of AI in architecture — the ability to produce in seconds, hundreds of design options generated from your text prompts with no training, no architect, no engineer and no planner required.

The promise sounds amazing, but while AI may be helping write many of our annual reports, uni essays and shopping lists, would you trust it with your life savings to build your dream home?

If we look at the rapid integration of AI into architecture, it’s not farfetched to say AI is likely already playing a role in some new designs.

According to a report by the Royal Institute of British Architects, AI adoption in architectural practices currently stands at 41 per cent.

Given that figure is set to keep growing, it’s inevitable that AI will influence and change the way we design our homes.

The question is, how?

How is AI being used in architecture?

Architects, like many other professions, are rushing to get a handle on not just what the rapidly developing technology is capable of, but the implications (ethical and otherwise) of this powerful new player in the design field.

Indeed, most major practices already have an expert on staff and even small practices are experimenting with using everything from ChatGPT to advanced visualisation engines.

That last part — visualisation — is perhaps the most widely-used application of AI at the moment.

Whether it’s embellishing existing images by adding features like trees, people and clouds, re-sampling images using different materials to “test” design options or just creating completely new images, the bar to entry here is very low.

An AI-generated image of a backyard with a pool surrounded by native plants and the back of a well-lit modern house

In seconds AI gives you ultra-realistic images of what a future home could look like. (LookX)

To the non-designer, the ability to generate images like these allows access to a world of design they’ve never had (case in point, my new house design above).

For actual designers though, this use could be problematic.

As Mohammed Makki, senior lecturer and expert in AI at the University of Technology Sydney, told me, AI is likely going to “make the life of an architect a living hell” as clients, convinced they’ve “done the design”, seek to build what AI has produced for them.

“The illusion that it works [is] because it looks real, feels real, … it’s amazing,” Makki said.

The issue is, as anyone in the industry will tell you, there’s a lot more that goes into a design than image-making.

Will architects survive?

Believe it or not, an image of a building is not a design and architects spend 90 per cent of their time doing everything except drawing images.

Instead, time is taken up doing all the necessary work that makes an idea come together and come to life (think costing projects, sourcing potential materials, making sure it’s compliant with regulations etc).

That work requires quite a bit more intelligence and qualitative assessment than AI can currently muster.

Where AI is making an impact, other than visualisation, is in its ability to streamline the production process.

That promises to speed up the mundane and administrative work, so architects and designers can get back to what they are best at — working with people on their unique design solutions.

Like with anything, improving efficiency may also lead to fewer costs and a cheaper bill for you — but that may be a while off yet.

For now, the translation from image to reality is still a space where people will need expert guidance.

Talking to experts and practitioners alike, there doesn’t seem to be much concern about AI taking over any time soon.

“Creativity is a skill that needs to be developed over time,” Makki told me.

“AI can’t replace that … it’s really just a tool.”

Already, though, there are some big questions about how AI is used that remain unanswered.

Can you spot a good AI design from a bad one?

While an AI engine might generate an image for you, do you really know what you’re looking at? Can you tell a good design from a bad one? Does it matter?

The answer to at least the last question is yes — absolutely.

The quality of a design or image you get from an AI engine is solely based on the quality of the inputs (in this case, other images of buildings) it’s being trained on — junk in gets you junk out.

As novices, knowing the difference between a good quality output, aka something you could genuinely build, and a bad one, can be hard to decipher.

An AI-designed Australian home with a large kitchen and dining area opening out onto a backyard

The promises of AI seem amazing, but using it to design our houses raises many questions. (LookX)

And yes, that includes much more than style and looks — think function, appropriate materials and building codes. Like it or not, this takes more than browsing design magazines to learn.

The next hurdle is assessing the options AI gives you. With so many so quickly available, and so persuasively rendered, how can an ordinary person pick?

This is where expert judgement comes back in. 

That requires training and experience to understand how a drawing will become your house, and all the potential pitfalls of building in between.

Who owns the design?

On top of all of the above, AI doesn’t know what it’s mashing up to create your images, who they belong to and whether or not it’s even appropriate to use them.

This raises questions that don’t have definitive answers but are already being tested in practice, for example, around intellectual property.

Even more sinister, the assumptions of an AI engine that’s generated your image or design are well hidden.

Just like with social media, where we don’t know who or what is biasing our feeds, you don’t know who’s training the AI algorithm you’re using.

The more accessible the interface is to novice users, the more assumptions are built into the generator to make it easy to interact with.

Funny how that material keeps coming up in my designs? Isn’t it interesting that my designs keep featuring a certain brand? You see where this goes …

Too many of the choices aren’t yours.

Advertising goes underground, biases are hidden, cultural norms are entrenched in a design version of a social media bubble, and we’re being manipulated by actors we can’t see, only choosing from what’s been put in front of us.

Changes afoot

So, where do we land with all this?

While some would argue the no-cost AI approach, where we sidestep the architect and all the other professionals involved, is ideal, I strongly disagree.

In that scenario, I can already imagine an irredeemably dull built environment, not a richer one, and a pathway littered with the broken promises of images that were never going to make it to a real site, or never should have.

Instead of being people-focused, everything becomes about the image, with promises of efficiency and cost reduction leading to real problems when the physical building actually begins.

So, while the opportunity is there for people to start imagining what’s possible and get more involved in the design process — which is a great thing — it feels like architects are safe for now.

But the reality is the field is moving so fast and there are very real changes afoot.

Watch this space!

Anthony Burke is a professor of architecture and the host of ABC iview’s Grand Designs Australia, returning later this year. 

He also hosts Grand Designs Transformations and Restoration Australia which you can stream now on ABC iview.

link

Leave a Reply

Your email address will not be published. Required fields are marked *