The Architectural Formula - Inspiring Property Design at Pole Lap Pace
- Rob Anderton
- Jun 23, 2024
- 5 min read
Updated: Nov 14, 2024
Unlocking the value of AI in architecture through a Father’s Day conversation, Formula 1, and shared observations of where AI can accelerate creativity and the confirmation of client requirements.
If you had told me that Father’s Day morning would start with a Colombian coffee, an architectural journal, and a pot of berried muesli with greek yoghurt, I’d have thought we were playing an inverse game of two truths and a lie. Now I’m a muesli convert. Whilst at one point I had myself down for following in my father’s architectural footsteps, the knowledge that there was more to it than simply colouring in drawings of university campuses soon kiboshed that fantasy. But there we were, chatting over an article in the RIBA journal, about how AI is impacting the landscape of architecture services - a sort of first for blending my Dad’s profession and mine.
When talking about AI in my role, a few topics frequently surface. Loss of intellectual property and data privacy are certainly towards the top of that list. The article my Dad slid over the kitchen table was discussing how a renowned practice was looking at using AI for design, setting parameters to maintain their signature style whilst inspiring new concepts. What was interesting about this was that much of the AI conversation to date centred around productivity was now shifting to creative consistency, with particular focus on customer delivery. Many organisations have quickly acclimatised to the administrative easing afforded by such tools as Microsoft 365 Copilot, but few have yet to really tap into how these tools transform and relay value to the customer.
There I was, brew in one hand, article in the other, my eyes hooked on this fascinating visualisation. A beachside mansion in pearl white concrete curiously shaped to look like a classic acoustic guitar - my interest was piqued. On face value, you would have struggled to tell it was AI generated. On closer inspection, staircases leading to walls, the odd E-string too many, and even the musically unaware could spot some of these comical errors.
Still impressed by the overly strung building, my Dad’s curiosity centred on how this technology transformed text to image, and at what pace this could be achieved. What I hadn’t fully appreciated was the different techniques he had deployed in his line of business to accelerate design inspiration. He cited that one technique involved collating words and seeking synonyms around a topic, pooling the individual and collective experiences, memories and knowledge of architects in the practice. For example, designing a golf clubhouse could invoke words such as manicured, pristine, crisp, and immaculate. For me, I’d throw in sandy. The idea was that collating these words would provide the jumping-off point for possible designs and encompass all the emotions they wanted to evoke within the client. The evolution of that particular process led us both to think about how other applications of the technology may encourage architectural firms to consider bringing outsourced and previously laboursome tasks back in-house, saving both time and money.
Seeing is believing though so a few moments later, and with Chat GPT 4o open, I brought another of our shared passions into the mix - Formula 1. In a time that would have the best pit stop crew in awe, I’d drafted this prompt:
“Create an image of a state-of-the-art property for Lewis Hamilton, that is three stories, by a beach and inspired by racing. There must be capacity for three F1 cars.”
We got this response.
Not bad right? I’m happy for you to forward this to Lewis if he’s in the market for some renovations, or a garage extension for a new Ferarri or two. The eagle-eyed amongst us may have noticed that the property featured well in excess of the three car parameter I’d given. Nevertheless, this brought us quickly onto the topic of hallucinations, or the fact that the output we get isn’t always in line with what we request. Not ideal in times where there is little room for error, and for maintaining professional integrity, it’s definitely recommended that we take a second look. For times of creativity and giving a client early visibility, having models that push the boundaries of possibility can make the odd hallucination a light bulb moment for new ideas. This thought reminded me of the Edison quote:
“I have not failed. I've just found 10,000 ways that won't work.”
With AI, we may not always get exactly what we expect. Yet experimentation and imperfect responses might just be what we need to think differently and more critically.
The next bit surprised me. I knew from an architect's perspective that being able to show a design from different angles could make all the difference between a client’s acceptance or rejection of a concept. However, from experience, prompting a model to make a change to an image previously created had always led to a completely different visualisation. Could this time be different? For example, I wanted an aerial view of our design so I prompted:
“Could you show me this exact image but with five stories, a helipad and from an aerial view”
Aha! Different, yes. Massively different, no. But we had created a moment. A moment where what my Dad and I did collided - AI meeting architecture.
Whilst we both agreed that neither of these would be submitted directly to a client due to apparent inconsistencies, the speed in which we created these ideas was a game changer. Rapidly creating a mood board to meet or better understand what a client likes and dislikes not only would help to refine requirements but could also be key to winning that next bid or framework.
This was a first for me in creating a use case demonstration in the field of architecture. It highlighted that not only are there many different uses for AI within the field, but the way in which we think of deploying it should not solely be about employee productivity but an ability to better meet the needs of our clients in real-time.
What this experience reinforced in my mind was that applying AI to augment familiar day-to-day actions led to a quicker understanding of the technology and an increased appetite to continue experimenting. It indicated that tailored demonstrations could create exciting ‘aha’ moments, moderated by an appreciation that there is still a need to critique AI generated outputs. Importantly, it also spotlighted that I may need to seek further AI inspiration for breakfast suggestions to go along with my coffee…
Credits
Author: Rob Anderton
Images: OpenAI. (2024). ChatGPT 4o [Large language model]. https://chat.openai.com/chat
