Neil Leach is a professor at Florida International University, where he directs the Doctor of Design program. He has published two books on AI and architecture: Architecture in the Age of Artificial Intelligence: An Introduction to AI for Architects (Bloomsbury, 2022) and Machine Hallucinations: Architecture and Artificial Intelligence (Wiley, 2022).

"AI is putting our jobs as architects unquestionably at risk”

In the near future, architects may become a thing of the past. Artificial intelligence is quickly advancing to a point where it can generate the design of a building completely autonomously. With the potential to create designs faster and with more accuracy than ever before, AI has the potential to revolutionize the architecture industry, leaving traditional architects out of the equation. This could spell the end of the profession as we know it, raising questions of what the future holds for architects in a world of AI-generated buildings.

I did not write the paragraph above. It was generated by ChatGPT, a highly impressive AI text generator that was launched recently. Make no mistake. Despite its innocuous sounding name, ChatGPT is no simple chat bot. Rather think Deep Thought, the massive computer in The Hitchhiker’s Guide to the Galaxy, designed to give us the ultimate answer to ‘Life, the Universe and Everything’. ChatGPT is astonishingly capable. The key difference, however, is that while Deep Thought took 7.5 million years to come up with an answer, ChatGPT can do it in 3 seconds.

Architects have now woken up to the extraordinary potential of AI. This is mainly because of the remarkable capability of ‘diffusion models’ – such as DALLE, MidJourney and Stable Diffusion – to generate images. The quality of the images generated can be simply astonishing. Amazing as they are, however, these images are a potential trap. Some architects have become obsessed with them to the point that they are overlooking the real issue. Ultimately, the AI revolution is not about image production, but about the assistance that AI can offer throughout the entire design process.

ChatGPT is based on GPT3, a massive pre-trained Generative Pre-Trained Transformer (GPT) that uses Deep Learning to produce human-like text. DALLE, MidJourney and the other diffusion models are also based on GPT3. Both use ‘prompts’ – verbal cues that generate an outcome. But whereas diffusion models connect text with images through the logic of captions found on the web, ChatGPT operates purely from text to text. As such, ChatGPT is more direct.

I first became alarmed by ChatGPT, when a Brazilian colleague, upset that Neymar had not been selected to take the first penalty in the World Cup shoot out against Croatia, asked ChatGPT who should have been selected. The answer was Neymar. The ramifications of this are somewhat startling. Could not football coaches now use ChatGPT for advice during a match, just as referees use VAR? Or could not others use it for more general advice. Could we not use ChatGPT, for example, for advice on which material to specify for a building? In fact, could not anyone else do so – including non-architects?

Social media is now awash with reports about the jaw-dropping potential of ChatGPT. US Congressman, Ted Lieu, is freaked out by it, and calls for AI to be regulated.[1] Canadian academic, Jordan Peterson, tells us about how stunned he was with the outcomes generated by ChatGPT: “I asked it to write an essay, written in a style that combines the King James Bible and the Tao Te Ching. That’s pretty difficult to pull off. You know, any one of those things is hard. The intersection of all three, that’s impossible. Well, it wrote it in about 3 seconds. . . grammatically perfect, and quite impressive philosophically.”[2] There are, however, systems under development right now – such as GPT4, a substantial improvement on GPT3 – that surely will make the next version of ChatGPT even more impressive. As Jordan Peterson puts it, “There are things coming down the pipeline on the artificial intelligence front that are just going to make your hair stand on end.”[3] 

ChatGPT is already putting some jobs at risk – and not necessarily the jobs that you might think. We have all seen distribution plants for Amazon, or the floor of a TESLA car factory, with hardly a human being in sight, and we might imagine that blue collar workers would be the first to go. But progress in robotics has been relatively slow. Simple tasks, such as selecting and picking up a brick, remain challenging for a robotic arm. Meanwhile AI has been racing ahead, to the point that ChatGPT is now quite capable of writing code. As software engineer, Metehan Ozten, puts it: “This is terrifying. What ChatGPT means is that probably within five years from now, software engineers will be obsolete.”[1] Be afraid! Be very afraid! 

And so what about architecture? Rumours of the death of the architect are greatly exaggerated – or so we are led to believe. In his recent article, Will Wiles reassures us, ‘Architects can rest easy that AI isn’t coming after their jobs just yet.’ Comforting words. I beg to differ, however. There are signs that AI is becoming not only good, but terrifying good, to the point that it is beginning to expose our own limitations as human beings, and putting our jobs as architects unquestionably at risk. 

Architecture, it could be claimed, is already under threat. Early research by two Oxford scholars, Carl Benedikt Frey and Michael Osborne, suggests that designers will be relatively immune from the dangers of being replaced by AI.[2] The mistake they make, however, is to assume that there would be a simple one-for-one replacement of a human worker by a machine. In fact, the way that AI actually operates is as a form of ‘prosthesis’ that extends and augments the abilities of the architect. Of course, this can be incredibly helpful. Using AI, a single person office can now compete with bigger offices, and enter large scale competitions. However, the corollary is that practices will no longer need so many architects. Wanyu He of Xkool estimates that a single architect using AI can achieve as much as 5 architects not using AI. Does that mean that 80% of all architectural jobs are now at risk?

But what about further into the future? An interesting comparison can be made with taxi drivers, now that driverless taxis have been introduced in San Francisco. Once a taxi is self-driving, we no longer need a taxi driver. What happens then, when AI can generate architectural designs completely autonomously – as will surely happen very soon? Will we also not need an architect? The problem with AI, then, is not how evil it is. For how can anything be evil when it is incapable of even thinking? Rather, the problem of AI lies in its very capabilities. It is already better than us in some areas, and will eventually be better in every single domain. As Garry Kasparov noted, following his defeat at Chess by IBM’s Deep Blue computer, “Everything we know how to do, machines will do better.”[3]

So what is the solution? Most obviously, architects need to start taking advantage of AI as a way of staying ahead in an increasingly competitive world – ‘If you can’t beat ‘em, join ‘em,’ as the saying goes. We need to familiarize ourselves with the potential of AI, and upgrade ourselves to become ‘superusers’ – to use a term coined by Randy Deutsch. Indeed Spacemaker already claims that there is no future for architects unless they can use AI: “We do firmly believe that in the workplace of the future architects using AI will replace architects that don’t.”[4]Or, as ChatGPT puts it, “AI offers powerful tools to automate tedious tasks, to optimize decisions, and to design new and more efficient solutions. AI will allow architects to design more efficient and cost-effective buildings, and will no doubt be the future of architecture. By utilizing AI, architects can create faster, smarter, and more efficient designs. By not taking advantage of these capabilities, architects are missing out on the opportunity to stay ahead of the curve and remain competitive in the changing world of architecture.”

The most important issue, however, is to be aware of a problem. For, once a problem has been recognised, it becomes a different kind of problem – not one by which we are trapped, but one which we can address. Surely, what we architects should be designing right now is not another building, but rather the very future of our profession.

I will leave the final words to ChatGPT: “Architects who choose to ignore AI will be left behind and ultimately forgotten as the industry evolves and advances. Therefore, it is imperative that architects pay attention to AI and its potential to revolutionize architecture, or they risk sleepwalking into oblivion.”

This has been published with the permission of Neil Leach

To learn more about AI generated work follow @neilleach14

[1] Ted Lieu, ‘I’m a Congressman Who Codes. AI Freaks Me Out.’ New York Times, 23 January 2023. https://www.nytimes.com/2023/01/23/opinion/ted-lieu-ai-chatgpt-congress.html

[2] https://www.youtube.com/watch?v=WLehkWESDJ8

[3] https://www.youtube.com/watch?v=WLehkWESDJ8

[4] https://www.youtube.com/watch?app=desktop&v=1hHfoB4mSrQ

[5] Carl Benedikt Frey, Michael A. Osborne, ‘The future of employment: How susceptible are jobs to computerisation?’, Technological Forecasting and Social Change, Volume 114, 2017, Pages 254-280, ISSN 0040-1625, https://doi.org/10.1016/j.techfore.2016.08.019. https://www.sciencedirect.com/science/article/abs/pii/S0040162516302244

[6] Garry Kasparov, as quoted in Neil Leach, Architecture in the Age of Artificial Intelligence: An Introduction to AI for Architects, London: Bloomsbury, 2022, p. 50.

[7] Neil Leach, Architecture in the Age of Artificial Intelligence: An Introduction to AI for Architects, London: Bloomsbury, 2022, p. 124