Timothy Alouani-Roby: What do you think of as the potential dangers of AI in its current form?

Gerald Matthews: I think the biggest risk of AI as we currently experience it the assumption that it is intelligent, when it really isn’t. Most of the current AI toys that are doing the rounds right now are language processing systems that re-churn already existing content into new content, more or less as a complicated ‘paraphrasing tools’ (or the visual image equivalent). There tools are not making aesthetic or value judgements about the quality of what they themselves are producing, so I don’t think they should be considered to be artificial intelligence. 

They may be valuable tools, but they are processing, not thinking. Alan Turing’s famous thought experiment, about whether a machine that could fool us into believing that it is thinking, assumes that a close enough imitation of something is equivalent to the original.  With a digital view of the world this feels very true. But with an analogue view of the world it falls far short. Perhaps a better term for the kinds of tools we are seeing are ‘intelligence imigators’.  The science and technology historian, George Dyson, said that “any system simple enough to be understandable will not be complicated enough to behave intelligently, while any system complicated enough to behave intelligently will be too complicated to understand”.  I think he’s probably right – which is perhaps why analogue aspects of computing are making a strange comeback in the creation of machines that are becoming too complicated to be predicted, let along understood.

Our pre-occupation at present seems to be trying to get AI, or what Judea Pearl called ‘Opaque Learning Machines’, to do things that we ourselves can do – only faster, cheaper and sometimes better.  I think that’s a low bar for what may still evolve to become sentient entities that do not think or act anything like us, or like any other living thing for that matter. It is perhaps no wonder then that so much of the current fear is about making ourselves collectively redundant. This is not just another step in the improvement of tools.  When hand-tools developed into power tools they still required thought in how to utilise them; they were just faster and took less human effort. What we now have in development are potentially tools that could decide their own purpose and require nothing of us to achieve it. We’ve hypothesised an enormous number of dystopian or apocalyptic outcomes regarding AI – hopefully some of these will serve to inoculate us from such mistakes. I doubt it, if our reaction to climate change is anything to go by.

I think that it is time to turn our minds and these tools to the kinds of things that AI might be able to do, create or solve that we could never achieve. Interstellar exploration, for example, or perhaps overcoming our own inadequacies to create global peace, health and happiness.

I have wondered many times what lays ahead for architects with the rise of AI. This first phase of opaque learning machines is likely to change my profession, but unlikely to diminish its importance. It will probably mean that fewer architects are needed worldwide (there it goes taking our jobs again), but it hopefully will mean that the best (i.e. remaining) architects around the world will have access to more powerful tools to deliver more valuable outcomes.

I can imagine AI design generators spitting out thousands of potential designs to suit an individual based on their music playlists, liked movies, relationship history, medical information, cultural background, wardrobe choices, DNA sample, etc. But I very much doubt that any time soon we will have developed an AI that can actually say which design out of the almost infinite possibilities will result in that person being the happiest.

I believe this because the more people I get to work with in my career as an architect, the more I am in awe of the diversity, complexity and sheer unpredictability of people.  Sometimes what people like isn’t what they need, or indeed even what they really like at all. All of the best architects I’ve ever met understand the value and importance of human contradictions and I think some of that is beyond the comprehension of anything that does not share the human condition.

TAR: How might we protect against the worst case scenario of AI?

GM: We protect children from getting hurt or from hurting each other through supervision and through boundaries. It should probably be the same for AI. Giving any AI unrestricted access to the internet might help it learn quickly but not necessarily develop appropriately. Perhaps we need a global authority that regulates and monitors AI welfare and development, and certainly asking whether any group that wishes to should be allowed to create and develop an AI. We should be taking it just as seriously as we do the development of nuclear weapons because the implications could be even greater. 

It honestly astounds me that as a species, after all this time, we have never managed to peacefully form a functioning global government. Without it we will never solve any global issues such as climate change, genetic engineering, nuclear technologies, or address disparities of wealth, health, education or resources. Perhaps AI (or our potentially justified fear of it) will draw us together to work as one.

TAR: Is there a danger of overlooking AI as a potentially valuable design tool?

GM: Computer interfaces really haven’t evolved much in the last forty years – we still use mouse and keyboard – but one of the potential benefits of machines that can learn is that they might learn to interpret us in new ways. I imagine an architect standing on a street corner looking at an empty site and with a wave of their hand, like a conductor, the structural columns of virtual light rise into place, followed by floors and walls and all of the complex internal systems that make a building function. Being able to not only rehearse the construction of a building in augmented reality (what William Gibson called ‘Virtual Light’) but to use it as a means to shape and refine design ideas will require that we interact with our tools in far more fluid ways than tapping on bits of plastic. I think that AI could play a valuable role in learning how to interpret an individual person’s hand gestures, breathing, eye movement, facial expression and posture as well as our voice to enable us to use computers in more expressive and nuanced ways.

Related: How will AI be used in architecture?

TAR: Do you see parallels between the transition from hand-drawing to CAD in the emergence of AI? What makes AI qualitatively different from just another medium of expression?

GM: Architecture has a long history of embracing new ideas, like mechanical pencils and indoor plumbing. I’m sure AI will be no different.  Architects will embrace AI – not only as a tool to create better projects with more efficiency, but also as an integrated part of all of our future lives. I don’t think it will replace us, I think it will enhance what we can do and how we do it.

TAR: A design challenge: create the ideal architectural education. How would you design it?

GM: I have worried for a long time about how architecture should be taught. As far as I’m aware, almost every school of architecture around the world prides itself on a ‘design thinking’ focus – and why shouldn’t they? Perhaps because design thinking can’t be taught. Design thinking can be nurtured and developed, but only through practice and through evaluating the success of the finished outcome (which isn’t possible if the design is never built).

Placing such a strong emphasis of the surface level appearance of concept designs and ‘pretty picture’ architecture has resulted in a widely held belief that architects are capable of nothing more than pretty pictures, eroding the respect our profession once commanded as well as the genuine skillsets we once possessed. 

In a time when buildings are become more and more complex, it is unfortunate that architects are becoming less knowledgeable and less interested in the technical aspects of design. For this reason, at our practice, we begin training someone who is at the start of their architecture career by building a technical foundation. This includes a deep understanding of a range of building codes, design standards, guidelines and legal requirements, as well as learning about a range of structures, mechanical systems, hydraulic systems, electrical systems, fire services, acoustic design, cost comprehension and other design parameters. It also involves learning to communicate at a high level through precise drawings and clear specifications.

Our philosophy behind this approach is simple: before you can begin to creatively design something as complex as a building you need to first understand a lot the ‘rules’ and how buildings go together.  Typically it takes a new graduate three to five years to build a meaningful technical foundation – only once they have done this can they successfully deploy any creative design thinking. Very few universities seem to provide a robust education relating to building codes, structures or building services, meaning that when graduates begin their careers they often find it either terrifying (knowing how much they still have to learn) or shocking that they aren’t given design opportunities (not understanding that they lack the required information to successfully design beyond surface level appearance).

If I were to craft a process to train someone to become a good architect, it would start from day one with parallel training in codes and standards, engineering and building services, as well as graphic communication, artistic expression and sculpture. Learning about the history of architecture and the celebrated designs of great architects is important but there are far more important things, especially when you consider that you can’t really appreciate any great work of architecture by just looking at photos or overly simplified drawings. If you wanted to learn to paint as well as Da Vinci, you can’t just look at finished paintings by great artists – at some point you have to watch someone with skill while do it and be guided by them while you practice. 

So, I think an apprenticeship should be at the core of any architectural education. Rather than ‘learning about architecture’ for years and then jumping into the profession with little knowledge or skills that can add value to real projects, I think the learning process should create a solid technical foundation. Graduates then have their whole career to explore and develop creative design talent.

TAR: You’ve spoken about this being ‘the last days of architecture’. What do you mean by that?

GM: Not every building needs to be designed by an architect. I didn’t always think that, but now I understand that not ever structure has to be a work of architecture.

When I started my career I remember hearing much older and more experienced architects complain that the profession was declining.  They talked about builders that were building design into their offer, project managers delivering projects as if they were managing a production line that just had to keep moving faster and faster even if the outcome suffered; they talked about universities teaching students how to present impractical designs well rather than teaching them to design good buildings, about appearances becoming more important than performance.  And I thought they were just grumpy old men (in my case they really were all men).

Now, however, I wonder if I’m destined to become a grumpy old architect. I hear myself talking about how too few architects understand the building code, because it’s easier to just email the certifier than to read and understand the code; I hear myself talking about drawings that are too often pretty rather than precise. I also sometimes find myself trying to explain why design matters, when no explanation would seem necessary. I’m not grumpy yet though, it’s just that I look back and begin to understand what all those older experienced architects were talking about two decades ago.

Things have changed and are continue to change. Buildings and their functions are more complex, and codes are more onerous. Architectural education is pre-occupied with appearance over practicality and functionality. Architects have less authority and influence within projects. Buildings have become assemblies of products and systems, rather than a crafting of materials. They are built for shorter lifespans. The tools of design have become more powerful but far more complex. The average experienced architect today knows less than the average architect of a generation earlier.

Is it any wonder then that architects have lost much of the respect and esteem that they once had?  For all of these reasons, I wonder if we are in the last days of architecture. I’m sure some form of the profession will endure, I just hope that it isn’t a surface level symbolic position fixated on short term aesthetic fashion trends while being the butt of every other profession’s jokes.

Matthews Architects
matthewsarchitects.com.au

Photography
Aaron Citti, courtesy of Matthews Architects

We think you might also like this Q&A on AI with Woods Bagot’s Jet Geaghan.