Newsletter - Links - Advertise - Contact Us - Privacy

Artificially Intelligent: AI or Naw?

30 Apr 2024

Multiplying utopian and apocalyptic headlines may give the impression that we are snowballing toward a new reality but is that the case? Andrew Strachan of Stallan-Brand delves behind the media hype with a broader look at the impact of AI on architecture. Images by Pixlr.

Multiplying utopian and apocalyptic headlines may give the impression that we are snowballing toward a new reality but is that the case? Andrew Strachan of Stallan-Brand delves behind the media hype with a broader look at the impact of AI on architecture. Images by Pixlr.

A.I Artificial Intelligence was once just a science fiction film. Now, AI is either lambasted having expedited the end of the world, or alternatively hailed as the 21st Century’s answer to the Industrial Revolution. Thankfully, the reality is likely to lie somewhere in between. While esteemed AI scientists have sounded the death knell for all humankind (“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war”) its impact on architecture has not yet reached those levels.

That said, by the time Lord Foster or Renzo Piano pen a similar letter to the architectural profession, it will be more eulogy than warning. As a profession which moves at a comparatively glacial pace, the rapid development of AI as a creative tool may be sounding alarm bells for many. For others, it’s simply the next step in the evolution of the profession; the next logical step in our drive to automate and digitise the world around us. Computer-aided Design arrived in the 1970s. Building Information Modelling arrived in the 1990s. The latest, greatest evolution is arguably overdue. We are working more efficiently than ever, thanks in no small part to AI. Where once pen and paper ruled all, we are now drawing, writing, meeting, corresponding, presenting, and everything-in-between digitally. Typically met with little resistance or concern, menial tasks are conducted by computer power through various pieces of software – many of them employing AI.

Whether acknowledged or not, AI tools are already critical to the design process. Environmental simulation and analysis, virtual and augmented reality, rendering software, and building performance analysis all feature heavily. None of this is deemed a concern. Why? Because we understand how it works in tandem with human interaction and inputs. We’re still in control of these aspects. The next generation of AI software must be utilised similarly. Increasingly frequently we are seeing Architects transition to software development and working with AI software. Companies such as ARK, Spacio, LookX and Finch are all founded by Architects or those who have worked in architectural practices. Their primary objective is – as you might expect – to revolutionise the architecture industry in some shape or form. They have all identified similar inefficiencies in our tried-and-tested processes; enough so that they believe AI can improve these immeasurably.

Watching their promotional material is quite mesmerizing; building blocks appear, shift, bloat, shrink back down, stretch, and rotate with minimal input or effort. Flat layouts appear on floor plans, moving around, growing, adding (or removing) bedrooms as they develop, whilst simultaneously generating alternative kitchen layouts and bathroom configurations. On screen, it makes for compelling viewing. Other AI software such as Blocktype is developed by a Planner-turned-developer team. The aim here is to stress-test sites for viability before an Architect need get involved. To varying degrees of success, each of these pieces of AI software employ generative design in an attempt to automate the earliest stages of a project. A common stereotype used disparagingly towards the profession – whether warranted or not - sees Architects consumed by the visual; the aesthetic; the superficial. Text-to-image software such as DALL-E and Midjourney now let anyone conjure up increasingly realistic looking images without formal training (or imagination, for that matter).

Conceptual art can be automatically generated by AI to expedite the design process and convey design intentions. LookX takes this image generating technology one step further and focuses solely on architecture and the built environment. It allows users to use hand sketches and photographs alongside text inputs as a prompt to generate more specific – and accurate – AI-generated visuals. Further again, OpenAI’s Sora now generates increasingly convincing videos in a similar text-to-video format. AI alone cannot currently produce the type of project-specific images that are required to accurately portray the design intent or specifics of a project, but the speed with which it is learning likely means it’s more a case of ‘when’ and not ‘if’ that can be done. In the same way that the advent of rendering software shepherded in the end of the classic photomontage, so too might advancements in visual AI technology mark the end of the physical model or hand drawn sketch. Let us hope not. Until AI can truly learn and understand human behaviour, compassion, and emotions, we still have the upper hand when it comes to fully considered design solutions. Whether that’s still desirable in the future is out of our control.

Concern around AIs intrusion into the technical aspects of architecture and construction has greater founding than those associated with simply treading on the toes of human creativity, and rightly so. There is plenty to suggest that the more complex coordination processes are in a far more precarious position. Many technical-focussed duties have already been deemed menial enough to warrant being automated by AI. The visual and aesthetic is not where AI can – or will – have the greatest impact upon the Architecture profession. That unenviable position is reserved for the technical and coordination aspects. We’re not quite at the stage where – with the single click of a button - RIBA Stages 0 through 7 are taken care of, but AI is changing how we work in practice at a faster rate than ever before. Clash detection is one of the more common AI-driven software processes used in practice.

Simply, AI is used to detect clashes between services, structure, FF&E, architectural and other assorted elements within coordinated BIM models. Constantly developing AI software can instantly suggest solutions to the clashes, before automatically correcting the offending models to avoid risks and resolve errors. The automation of these processed removes both human error, and the time implications of correcting issues manually. Automated Design Optimization (ADO) already leverages AI algorithms to generate intelligent design solutions based on selected parameters, predefined criteria, and selected data inputs. While not yet optimised for the construction industry, ADO will eventually be used to rapidly test options which will, for example, minimise the volume of concrete needed for construction, or minimise the volume of waste material generated for each AI-generated design option it produces. Predictive analytics aims to answer the question of “What happens next?”. It learns from historic data (such as project costs, programmes, timelines, performance criteria, costs, supply chains, weather patterns etc.) to predict future outcomes. This information can be used to improve efficiencies in project programmes, highlight potential areas of concern (such as delays caused by supply chain issues or predicted inclement weather), and can enable pro-active project risk management. AI software does not yet understand the nuances associated with all local Building Standards and Regulations, but it is learning.

UpCodes, and CodeComply are the first clear attempts to have AI learn and apply the data associated with specific US-based Building Codes to BIM models. Ultimately, the desired outcome will be to use AI as we do clash detection software to highlight contraventions of Regulations in coordinated BIM models before drawings are issued. Eventually, AI might be automated to analyse your BIM models for compliance, before automatically correcting any issues. The common requirement for all the above listed AI applications is the need for it to first learn before it can be productive. How these AI models are trained, what data they are learning from, and how they are implemented needs to be carefully controlled.

Quite how Duty of Care laws will impact those managing the development of AI software remains to be seen, while the social and ethical implications of AI generally is an article for another time. Doomsdayers will argue that there is a timeline in the not-too-distant future where we’ll be pitching to Clients (over Teams, of course) using a script generated by ChatGPT, sharing images generated by LookX and videos generated by Sora, showcasing a scheme massed in Spacio, populated with layouts in Finch, all laid out using a presentation-generating AI software like Pitch.

If the AI-powered analysis of our submission is deemed successful versus the other AI-assisted bidders, then we’ll simply click a button, do a rudimentary check of the AI-generated drawing packages, and issue them for construction. That construction might then be completed by robots with materials delivered to site by AI-driven vehicles is surely a concern for others. Some will argue we’re already in the age of building-by-numbers, so why not let AI do the hard work? As with every step in the evolution of our profession, we will always remain accountable for our outputs regardless of how that information is produced. Control over the data used to develop AI, how that AI technology is applied, and reviewing the outputs of AI-generated material will become increasingly important as its influence and integration increases.

The databases from which one piece of AI software has learned will become just as – if not more – important than its outputs. The governance and control of AI learning will become critical to its integration in practice. While AI looks unlikely to eradicate the need for Architects in the short term, our role will undoubtedly have to evolve to accommodate it. Whether that sees Architects harness the positive capabilities of AI as we have (largely) managed so far, or whether understanding big data, machine learning, and programming becomes the new norm remains to be seen. Finally, it wouldn’t be an AI-related article without some actual AI input. As ChatGPT has so succinctly summed up AIs impact upon the architecture profession; “While AI offers numerous opportunities for the architecture profession, it also raises ethical and social considerations.

Architects may need to adapt to new roles, incorporating AI tools as valuable resources while maintaining a balance with human creativity, empathy, and ethical considerations. Additionally, ongoing professional development and a willingness to embrace new technologies will be crucial for architects to thrive in an AI-driven future.” I couldn’t have said it better myself…

Back to April 2024

For more news from the industry visit our News section.
Subscribe to Urban Realm Magazine
Features & Reports
For more information from the industry visit our Features & Reports section.