Listen to the episode:

---

Welcome to today’s What is The Future For Cities podcast and its Research episode; my name is Fanni, and today I will introduce a research paper by summarising it. The episode really is just a short summary of the original paper, and, in case it is interesting enough, I would encourage everyone to check out the whole paper.

Our summary today works with the article titled Steering into the Skid – Arbitraging human and artificial intelligences to augment the design process from 2020 by Geoff Kimm and Mark Burry, presented at the 40^th Annual Conference of The Association for Computer Aided Design in Architecture and published in their proceedings Volume 1: Technical papers, keynote conversations. Since we are investigating the future of cities, I thought it would be interesting to see how artificial intelligence can be noted as an ally instead of a risk, at least in the design processes. This article investigates how can architects arbitrage an ever-changing gap between maturing AI and mutable social expectations.

Technology and artificial intelligence seems to threaten the survival of humanity, largely in science fiction. This paper investigates a smaller aspect: whether AI can usurp humans in the architectural practice. The possibility of machines to automate design has been pursued for over a half a century. There have been two main developments: the increased ease and efficiency of modelling, drafting, and detailing, and a degree of automation in the evaluation and generation of design possibilities as options. This paper focuses on the design evaluation and generation and the conflict with the architect’s traditional creative role.

There is a history of machine’s increasing agency in design. Algorithmic thinking and design have existed since the 1960s at least, probably even earlier with Gaudí’s evolving parametric design for the Sagrada Família Basilica. This approach relies on the architect formalising rules and connections, probably causing the failure of AI because it was supposed to replicate the human intelligence rather than focusing on logical if-then statements. The next phase can be when AI displays a navigation across many layers and can drive design proposition for optimal solutions, which might be within the designer’s assumptions. In this sense, the machine creates a range of possible solutions based on its research, beyond any traditional notion of a singular design outcome. This can be evolutionary computing or similar methods which are primarily directed to optimization, and the designer still must formalize the rules of what is good.

Artificial neural networks, abbreviated to ANNs have been around since the 1940s, but gained more attention when Google’s AlphaGo beat a world champion Go player. The GANs, generative adversarial networks are currently positioned as one of the most promising AI technologies for design. Generative built environment design with the use of such generative adversarial networks has already been used, even though in small scale for bedroom scenes. However, it proved that AI can be available for the architectural practice. Thus, the architect moves in to a new mode of control through providing example and precedent rather than the formal, setting of rules and limitations.

Despite its inefficient historical connection to the design profession, AI is increasingly used in architecture. There are three phases of automation from the workers perspective. First, the machine is dependent and only used to serve as a labour-saving device, possibly undertaking the bulk of the work. Second, the machine becomes the labour-enslaver making the worker assist its processes but there is no or little intellectual stimulation in the exchange. Third, the machine replaces labour phases, thus the human producing the actual work. The human may be liberated to new levels of productivity, or devalued and displaced. With respect to AI, contemporary architects may just be in the first phase and progressing towards the second.

In the first level, design computing is dependent on the architect. The architects is in control, and the computer undertakes little or no exploration or creation of novelty. In the second level, design computing becomes autonomous. Decisions show a level of novelty or surprise which may not be easily explainable. The architect is at risk of being devalued unless they can use the output to augment their own creative processes. In the third level, design computing becomes autonomous in its agency. The computer exploits new opportunities and understands nuanced contexts. The architect may potentially be replaced in significant parts of the design process within traditional architectural practice.

Of course, architecture is not as simple as any factory job, and AI replacing the architect’s creativity could seem fanciful. Nevertheless, it is worthwhile to investigate its effects seeing the long history of machines replacing human work. There are no special protections for the architect, but fundamental reasons why the traditional role of the designer may be threatened by AI. Architecture has many areas that can be readily automized due to its logic, such as scenario modelling. The AI can exploit the previous plans and designs to learn from them and create new designs based on those.

Additionally, the subjectivity of the value of design introduces vulnerabilities for architecture. The architectural output can be seen as manufactured output, and many AI techniques increasingly excel in creating such solutions with optimization through researching and recognising patterns in contemporary alternatives. Architecture is also vulnerable because it is a product to be sold, and the architect and the client can have different meanings for good results based on their price points. In a world of open-source images and ideas, a generative AI can easily create superficial novelty at a lower price.

Interestingly, creativity is also a vulnerability. Creativity is said to be at the crux of making variations and can be a result of cultural and social information. Art, design or cultural output is always rooted in what has gone before. Any image created that truly does not already exist would be entirely unrecognisable, in a literal sense, as art, architecture, or indeed anything else. Therefore, there is prevalent and general misunderstanding of how people should relate to AI and this is also present in architecture.

Accurately forecasting the future is difficult, including AI as well, but protecting the status of human intelligence may be a losing strategy: the ultimate outcome may be a very narrow definition of what it means to be a designer. The use of AI need not become a zero-sum game if the essential differences between human and artificial intelligence are properly appreciated and utilised. To enable this, architects, and people indeed, need to reassess AI’s role. Machines do not yet understand the human sense or experiences which drive architectural design. Counterintuitively, a closer, more equitable collaboration between artificial and human intelligence could progressively exploit their difference to the benefit of the architectural profession. As the generality of AI increases, architects can generate and embed new subtleties and expectations of outcomes. Architects may also change the very rules of the game at halftime. Therefore, architects are challenged to arbitrage the gap between the actuality of AI and fluid dynamic social expectations.

Effectively exploiting that gap requires an ongoing assessment of what AI is today and its challenges while keeping the architects’ longstanding agility to respond to the pressures and opportunities. The architects can take three roles in the automation’s three phases. The architech is the master builder of AI frameworks that will complement their designs, providing understanding of cause and effect and accommodate cultural nuance. The arbiter would create the social obligations and direct the AI towards fair and desirable outcomes, instead of limiting and enforcing solutions. The shaman can the interpreter for the AI recommendations and outputs to provide meaning for incorporating the methods and solutions into the design processes, mapping references, emotions, intuition and biases. As the shaman, the architect could design the rich narratives around AI contributions to the design process.

With appropriate foresight, the architect’s distinctive role in the nuanced synthesis of social and emotional understanding with emerging technology may allow them to sidestep the obsolescence AI may impose upon other professions. The three roles – architech, arbiter and shaman – are facets of the same response: the future architect need to collaborate with AI to offer new value to their innate skills. Whatever the future of AI-enabled architectural practice, it is crucial for architects to work together with the available best innovation and technologies. Otherwise, would architects as a profession have performed better than the insentient generic algorithm?

As the most important things, I would like to highlight 3 aspects:

1. Artificial intelligence has been emerging but it is not yet at the human intelligence level yet.
2. Automation has three stages: labour saving, labour enslaving, and labour replacing.
3. Architects have the opportunity to grasp the opportunities with AI and collaborate with instead of fearing it, and the authors recommended three roles for that: the architech, the arbiter and the shaman.

Additionally, it would be great to talk about the following questions:

1. Why is there a fear of artificial intelligence instead of curiosity? Why do we keep ourselves from investigating the opportunities the collaboration can provide?
2. Why are architects afraid of technological evolution? Why aren’t designers concentrating on the opportunities it can provide?
3. What do you think of AI? How do you approach the future with AI in it? Do you fear it or are you interested in the possibilities in AI collaboration?

---

---