Digitalisation in design is one of four scans that forms part of the Technological Innovation theme for RIBA Horizons 2034.
Arthur C. Clarke’s third law states: “Any sufficiently advanced technology is indistinguishable from magic.” [1] Fifty years on, this adage is truer than ever before. The advent of artificial intelligence (AI) and its ‘black box’ operations have challenged not only our understanding of technology and how it works, but also our place in this brave new world (thank you, Mr Huxley!). [2]
These science fiction references are apt. After all, many sci-fi authors have written about the transformational potential of technology with uncanny accuracy, decades before reality caught up with them. RIBA Horizons 2034 provides a similar opportunity to predict, occasionally warn, but more often to dream about the profession’s future.
The past and present
Technological advancements that have transformed the architecture, engineering and construction (AEC) sector are not new. Ever since Ivan Sutherland’s Sketchpad was released in 1963, working practices in the sector have undergone unprecedented change, with pen and paper gradually giving way to mouse and screen.
This rite of passage from analogue to digital meant that designs could be realised not just by hand sketches, but also by using, first, Computer-Aided Design (CAD), then Building Information Modelling (BIM), and now, finally, artificial intelligence.
These technological breakthroughs have helped to augment, automate and re-invent the way the architecture profession approaches design, collaboration, and creativity. Very soon, though, they will likely be replaced solely by systems capable of creating architecture from language – the ultimate interface between humans and machines.
I have always been fascinated by the fact that the focus of the AEC sector’s digitalisation efforts has for the past 20 years been about BIM. This is understandable: BIM has allowed stakeholders to collaborate using a single, shared source of truth from conception to completion – a long-cherished objective. And yet, the disruptive technologies that are powering the AEC sector today are much more diverse, developed first for the games, films, and computer science industries. Their potential to revolutionise the way architects design the built environment has only recently been properly recognised.
The near future
Performance-driven design powered by two technologies – graphics processing units (GPU) and distributed computing – are transforming the design process, simulation and evaluation of key performance indicators. What once took hours and days has become a real time experience.
Both technologies were initially developed for gaming and rendering and are now used in the AEC sector to, for example, run optimisation studies on an urban scale, capable of creating hundreds of thousands of options in a matter of hours – a solution pool for architects to start working from.
Open formats such as Pixar’s USD (Universal Scene Description) are breaking down both software and disciplinary silos, allowing data and design changes to flow between different software applications and devices, connecting people and locations in real time.
This newfound interoperability and API-first (Application Programming Interface) approach allows the AEC sector to customise design workflows with various SaaS (Software as a Service) components. We can now seamlessly connect project workflows across different disciplines at a fraction of the time it took doing it the traditional way.
Augmented, virtual and extended reality technologies with headsets first developed for the military are now transforming how people can appreciate spaces in and around buildings. The virtual worlds they create are hosted in what are called metaverses, a term first coined by Neil Stephenson in his seminal sci-fi novel Snow Crash [3] and later popularised by Meta.
By enabling people to work together to customise virtual spaces on the fly, these immersive experiences not only improve clients’ understanding of the design of 3D spaces long before they are built but also provide collaborative interfaces.
Physical space also finds a digital counterpart in digital twins, a technology introduced decades ago in the aviation industry and now on the rise in the built environment. Using information generated from conception, design and construction, digital twins are live, as-built representations of completed assets in operation.
The information they yield feeds back to the project team, completing the data loop. Digital twins help their stakeholders monitor and regulate how well buildings (and cities) perform and, importantly, identify ways to improve their occupants’ experience. Buildings’ performance can be optimised with Machine Learning (ML) and the resulting knowledge can be used to inform future projects for better outcomes.
Digital technologies are transforming building design from a labour-intensive, time-consuming, occasionally fragmented and often repetitive process into a real time, immersive and collaborative experience. They automate mundane tasks and enhance designers’ creativity and problem-solving capacity.
The software underpinning these advances, particularly in artificial intelligence (AI) and ML, emerged because of the exponential growth of computing power, the creation of continuously growing datasets, and the ease with which they can be adopted.
In the years leading up to 2024, growth in computing capacity was doubling every six months, entirely overtaking Moore’s Law, which predicts a doubling of chip capacity every 20 months. [4] This allows AI models to be created and trained against vast amounts of data. To top it all, the rate of adoption of these systems is unprecedented: whereas Facebook took 10 months to reach one million users, ChatGPT reached that number in just five days. [5]
Speculations on the (not so) distant future
So, where do we go from here?
There is little doubt that, of all the technologies, AI and ML will be the ones to revolutionise design processes and ultimately change the architecture profession as we know it.
The next 10 years will see advancements that were not thought possible just a few years ago, and at a rate that will surpass everything that came before it. Trained on information collected through designers’ everyday communications, AI will perform as personal assistants. From AI agents (“autonomous intelligent systems performing specific tasks without human intervention”) [6] – that execute tasks in response to verbal commands (without the need for bespoke customisation to connect across apps), to fully autonomous AI software engineers, the designer’s role will be unrecognisable. As I discuss below, it is even possible that domain expertise will be commoditised.
Conversations about AI in the AEC sector usually mean Large Language Models (LLMs) (such as Open AI’s ChatGPT, Anthropics’ Claude, Google’s Bard and Gemini, and Meta’s Llama 2) and diffusion models (such as MidJourney, DALL.E 3, Stable Diffusion, and Firefly). The automation and augmentation afforded by AI and ML are hugely versatile, supporting surrogate modelling, knowledge dissemination, generative design, business insights, and models to assist design.
Despite these benefits, it has been image creation through prompts using diffusion models that has taken the architectural world by storm. Architects seem mesmerised by the imaginative and creative possibilities of AI generating pictures simply from descriptions using natural language. [7]
While charming and doubtless innocent, this approach to design is rather superficial, lacking in critical thinking about the essence of architecture, the creative process, and architects’ professional responsibilities to society.
Preferring prompt engineering to actual design, they see themselves as moulding new realities with words, conjuring what is in their mind’s eye using a machine’s dreamscape blended from other people’s imaginations. As such, words are important as much as they are limiting, as humans’ existing vocabulary does not and cannot yet describe styles and solutions that have yet to emerge.
Defining architectural creativity in the era of AI
So, what is coming down the road?
AI is already moving far beyond image and text creation. In a short space of time, we have moved from unimodal learning to multimodal, ‘zero-shot’ models. These are AI models that have great generalisation abilities and are capable of understanding context and connecting across different media without being given an example of how to do so. They are allowing the development of speech-to-movies or even 3D models, with many new models by the likes of NVIDIA [8], OpenAI and Meta [9] coming out all the time. Some models now integrate text, images, video and robotic kinematics.
Tools like these hold the promise of fully automated processes that take design from conception to completion. From massing, performance analytics and layouts, to automatic configuration of solutions, drawing documentation and even code compliance checks – all controlled just with language and verbal commands.
In the future, all the workflows that today are run manually would be automated, with AI agents providing the connecting glue and unifying the whole process by stitching different applications and actions together.
AI is also going to automate and augment every aspect of construction. BostonDynamics and Figure are already developing humanoid robots to replace or support site workers. AI tools developed by companies like BuiltRobotics are affecting the installation of piling systems. ML-powered software such as nPlan is already used in project management, cost estimation and schedule optimisation.
Products and services that use AI and ML are being used to support the operation of buildings, including everything from asset management, security, and diagnostics to the development of digital twins and even cognitive buildings that learn from their own data and optimise themselves.
Imagine this as a joined-up system: a project brief would ignite an AI-powered chain reaction. A feasibility report would be created automatically. Different generative AI models orchestrated by AI agents would start delivering ideation material based on verbal guidance by the human lead designer. Drawings would be fed to ML models pre-trained on company design data.
From there, options would be automatically tailored to comply with stakeholders’ KPIs. Different layouts would be generated automatically and adjusted to fit the brief. A sustainable supply chain would be engaged to suggest components, materials and timeframes. Drawings would be produced and shared automatically with consultants and contractors. All data would be accessible and updated in real time by the whole project team.
Environmental performance, structural design, material specification, contractual duties, construction details and regulatory compliance would all be dealt with and fulfilled automatically through AI processes. Robots would be valuable collaborators, and problems on-site would be identified and resolved in real time.
Beyond that, AI-powered digital twins would make buildings smart or ‘cognitive’, able to self-regulate to accommodate the needs of their occupants – or even keep going without them, like Ray Bradbury’s fictional house in his cautionary 1950 short story There Will Come Soft Rains. [10]
The tight curve ahead
If this future is realised, what role will architects play? Will they be orchestrators still firmly in charge but now free to focus mainly on creativity? Or would any such role be redundant, merely a comforting illusion that human architects entertain to quell their anxieties about ceding control?
Answering these questions requires consideration of three factors.
1. Quality of data
The first factor is data. Data about the built environment is siloed and flawed. The AEC sector lacks any universal and open standards for encoding not just buildings’ structures and systems but also how they perform socially, contextually and operationally. And because the quality of ML outputs depends on access to good quality data to train, bad quality data will result in the inference of bad solutions. As the old saying goes, "garbage in, garbage out".
For this to change, the AEC sector must improve data collection, organisation and processing across disciplines by reconfiguring workflows to collect it properly, pooling and cleaning it, and organising and managing it to industry-wide standards.
2. Skills commoditisation
The second consideration is the commoditisation of AEC skills. [11] Studies have already shown [12] that with AI, high-skilled workers see less improvement in their skills than low-skilled workers. This levelling of the playing field means that skills premiums will eventually be lost, making it more likely that highly skilled services will be commoditised.
Just as anyone with a smartphone can now become a taxi driver (with Uber, for example), in the future, anyone could become the architect of their own house. As technologies evolve, specialised, hard-earned knowledge may no longer be a differentiator. Going back to our taxi driver analogy, thanks to GPS it is no longer necessary to have a knowledge of streets. In the AEC sector, there is no guarantee that an extensive education in any domain will safeguard any profession’s future viability.
Whether this makes the architecture profession obsolete or ushers in a redefined role (much as the advent of photography gave rise to alternative painting styles in art) will depend on how we define the idea of creativity through architects’ work and its value to their businesses. To make this transition requires a decisive stance that enshrines critical thinking and domain knowledge as key. As mentioned earlier, this strength of purpose is lacking in the profession’s engagement with AI to date.
3. Defining professional value
Finally, architects must re-assess their worth. If automation means the same jobs can be carried out faster and with fewer people, how will it affect architects’ value proposition? If AI automation reduces timeframes and resourcing requirements, the value proposition should be shifted to architects being compensated for their competence in collaboration, innovation and creativity. Architects’ survival depends on ensuring that disruptive technologies augment rather than replace their services.
Afterword: deeper into the future
If this is the not-so-distant future, where could it lead beyond 2034? Will architects have to compete with conscious, creative machines run on a hypothetical computer program called Artificial General Intelligence (AGI) for their livelihoods? Or will they coexist harmoniously, walking hand in hand into the sunset?
As most surveys place this technological singularity beyond 2034, I will not attempt to elaborate on how sentient machines will affect the AEC sector, let alone humanity as a whole. For anyone interested in finding out more, the debate on the advent of AGI is quite polarised, and the definition of ‘sentience’ in this context is contended, with syntactic or semantic points of view depending on whether you’re a fan of Turing [13] or Searle. [14]
However, one thing appears certain: humanity needs to agree on appropriate rules of engagement with these systems and to understand not only the business opportunities they hold but also their potential to reinvent our creative capabilities.
Before we anticipate introspective AIs like Cutie (the protagonist in Isaac Asimov’s short story Reason) to declare, “I myself, exist, because I think...”, we may do well to first understand our role as creators in this new chapter of humanity. [15]
Author biography
Martha Tsigkari is a Senior Partner and Head of the Applied R+D (ARD) group at Foster + Partners. Her background spans architecture, engineering, and computer science. She has two decades of experience working on projects of all scales and uses. Her work incorporates computational design, human-computer interaction, machine learning, and optimisation.
Martha has investigated the usage of deep neural networks and genetic algorithms in the design process, aiming to solve problems ranging from passively actuated micromaterials to performance-driven urban layouts. She is also an Associate Professor at the Bartlett, UCL and has lectured and published on the subject of computational design internationally.
RIBA Horizons 2034 sponsored by Autodesk
References
[1] A. C. Clarke (1973). Profiles of the Future: An Inquiry into the Limits of the Possible. Popular Library.
[2] A. Huxley (1932). Brave New World. Harper Brothers
[3] N. Stephenson (1992). Snow Crash. Bantam Books
[4] J. Sevilla, L. Heim, A. Ho, T. Besiroglu, M. Hobbhahn and P. Villalobos (2022). Compute Trends Across Three Eras of Machine Learning. ArXiv, Cornell University
[5] Statista - K. Buchholz (7 July 2023). Threads Shoots Past One Million User Mark at Lightning Speed
[6] Amazon Web Services (n.d.). What are AI Agents?
[7] House of Lords - Communications and Digital Committee (2023). At risk: our creative future: 2nd Report of Session 2022–23. HL Paper 125.
[8] NVIDIA DEVELOPER - J. Stephens (12 May 2022). Getting Started with NVIDIA Instant NeRFs. Technical Blog
[9] The Verge - J. Vincent (29 September 2022). Meta’s new text-to-video AI generator is like DALL-E for video
[10] R. Bradbury (1958). There will come soft rains, short story in: The Martian Chronicles. Doubleday
[11] Platforms, AI, and the Economics of BigTech - S.P. Choudary (2023). Slow-burn AI: When augmentation, not automation, is the real threat
[12] F. Dell'Acqua, et al. (2023). Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality in Harvard Business School Technology & Operations Mgt. Unit Working Paper No. 24-013
[13] Stanford Encyclopedia of Philosophy (2021). The Turing Test
[14] Stanford Encyclopedia of Philosophy (2020). The Chinese Room Argument
[15] I. Asimov (1950). Reason, short story in: I, Robot. Gnome Press
