Architects should see AI as an opportunity, which is a tool to augment practice, replacing mundane tasks, not as a threat to their jobs. Rather than considering AI as a robotic alternative to human beings, AI is meant to be an accelerator that positions the computer to handle certain things that a computer is really good at.
AI in architecture at present
At present, our business probably uses the simplest form of AI by developing bespoke algorithms to understand patterns and fitness criteria within our designs but only based on a fixed set of parameters. The application of these optimisation routines is more a function of the specific problems being solved rather than the overall resolution of a complete architectural design.
On the Museum of the Future project, the profile of the steel diagrid structure was developed using computational algorithms designed with a fixed set of fitness criteria; basically ensuring that the length and spacing of the frame elements were optimized to reduce the proximity of the adjacent nodes, but still following a set of physical criteria, namely making sure there were nodes at each floor plate
and the local stresses within each element did not exceed a given material capacity. By ‘teaching’ the algorithm to identify the more favourable scenarios the tool was able to generate a series of efficient scenarios which achieved the target outcome. Without the computer-aided intelligence for deriving the optimal form, this would never have occurred, or else we would probably still be designing it at this point, rather than seeing it being built.
Benefits of AI
The example used on the Museum of the Future is very specific, and probably best represents the starting point of AI into our industry. Usually this coincides with the limits of our natural intelligence. Finding opportunities for efficiency is the starting point, as efficiency generates quality. And quality comes from the reproducibility of a process. This is basically called automation, which ultimately can give you a better and higher-quality product.
Architects should see AI as an opportunity, which is a tool to augment practice, replacing mundane tasks, not as a threat to their jobs. Rather than considering AI as a robotic alternative to human beings, AI is meant to be an accelerator that positions the computer to handle certain things that a computer is really good at.
Threats
To date, the only examples I’ve seen of designers being replaced by machines, is in the form of the small design elements; perhaps the design of a chair, or maybe a basic floor diagram. Where architecture distinguishes itself is the overall social component associated with the lengthy process that goes into building a project. Over the course of a single project’s life span, there are so many variables to consider that even if you could optimize certain aspects, the changes that invariably occur over the course of 2, 3, 4 years are so unpredictable that you still require people to respond to the evolving circumstances. This is the social component of the industry.
I do see aspects of the design process being enhanced through AI, which may cause some tasks to be replaced, but if jobs are merely simplified to singular tasks, then yes, jobs would be lost, but if certain tasks are replaced with AI, then that leaves plenty of opportunities for other tasks to be absorbed and addressed by the human designers.
AI in other branches of the construction industry
Other parts of the industry that are also seeing AI enhance and contribute to their evolution; notably the more programmatic and commercial aspects. Google recently released Jane, built to support project managers to run different scenarios of construction based on as many procurement variables as possible. Similarly, project stakeholders (clients and developers) have been running financial models in order to find funding strategies for different construction projects.
AI and clients
I’m not convinced they are driving the process, unless of course the added value is literally translated into their product. Otherwise if the value is solely contained within the business to streamline and improve the general processes, then the client’s incentives are minimized. Still, the same way we are seeing AI disrupt any number of industries, including architecture, the same debate can be probably mirrored within the developers / clients businesses. And if there were to be direct overlap between a client’s AI investments and an architect’s, in that case I could see a definite driving influence coming from above.
AI, radical designs and human touch
Looking at what has been built in the last few decades around the world, I would say that architecture has come a long way from the linear and straight buildings from 50 years ago. Based on the expanding material palette for buildings, and the everevolving building technologies being deployed, buildings are bound to become more radical, but only of course within the appetite of society. But ironically what AI will probably do best is optimize a design to point where every parameter is set to
achieve the highest efficiency and sadly this may well generate a very boxy and square solution. The human touch is actually the interesting part because capturing the subjective part of the architecture is probably the most difficult aspect to teach
AI, essentially being able to accept flaws within the efficiency of design. The flaws are basically the aesthetics of architecture that are driven by people’s opinions; which cannot always be rationalized to a binary like or dislike. It’s these appreciated flaws that AI may struggle to capture, and what some clients may ultimately desire.
I don’t see it as completely disrupting our industry, but rather how architects stand to benefit from learning about data, its applications, and how taking small steps now to incorporate artificial intelligence into practice could ensure a better business and a stronger profession.
Machine learning, automation and architecture
The last technology advent to our industry was probably the era of Building Information Modelling, which coincidentally serves as the ideal precursor to AI. BIM
is basically a data repository for the built environment. All aspects of a building are designed in the 3D environment before they even break ground. And what this is, is data, the extent of which is becoming more and more involved, down to the last nut and bolt. For a business, for a city and for an industry; to have the data already available is the perfect opportunity for AI to come and learn from what has already been designed. The same can also be said for the as-built environment. Any building that’s already been completed typically has a collection of post-documentation information available. Collecting this data and interpreting it is exactly where AI could thrive.
Equally, building life cycle analysis is becoming more and more valuable (something like the internet of things) where buildings are able to self-diagnose themselves and report their live performance. This is actual live data being reported on the building’s performance. Using this feedback loop to inform the new design will generate a better and more efficient design. It’s unlikely to happen all at once, but you can already begin to discover start-ups all over the world that are capturing the data from small sectors of our industry and applying AI to it to add value.
AI and architectural education
Architecture graduates do not have degrees in Machine Learning, Software Engineering, or Mathematics and therefore probably don’t have the skills and training required to independently implement AI. But because the added value of
AI has already begun to influence the industry, designers are beginning to expand their skill set sufficiently to at least Learn to Learn AI. This may be as simple as learning a new software, or else applying new (more effective) process, but the
technological evolution will undoubtedly arrive and whether you embrace it or reject it will definitely not change the inevitable.
Building life cycle analysis is becoming more and more valuable where buildings are able to self-diagnose themselves and report their live performance. This is actual live data being reported on the building’s performance. Using this feedback loop to inform new design will generate better and more efficient design.
Software companies & architecture
In fairness, I cannot speak on behalf of the big software companies out there in terms of their business aspirations, but to date, their business model is very strong, so I don’t see the need to change it. They sell a package with annual renewal fees, they take no responsibilities for the output of what their software can develop and have nearly no material or hardware costs as most of the software’s are deployed digitally nowadays. The major companies that could disrupt the system are probably the same ones that could disrupt any industry: Apple, Google, Amazon… And if they had an interest in expanding their product breadth to our industry I would be excited. But for the time being, they seem pretty focused on phones and watches, so maybe buildings are not on their immediate horizon.
AI in architecture in the future
I don’t see it as completely disrupting our industry, but rather how architects stand to benefit from learning about data, its applications, and how taking small steps now to incorporate artificial intelligence into practice could ensure a better business and a stronger profession. Rather than relying on a finite amount of experience, having access to historical as well as live data will allow the performance of an architectural design to be tested against actual data rather than simulated data.
AI is already being to help solve the more complex design issues. The starting point of AI could figuratively be at the limits of where natural intelligence ends. These could be very abstract designs that are perhaps too difficult to describe and resolve using conventional theories and strategies.
One of the more interesting areas of development relates to legislation; approving authorities are becoming more open-minded to demonstrating actual performance-based design rather than simulated performance-based design. Relying on actual data will not only evaluate the design, but obviously inform future design solutions, even to a regulatory level.
