Published in 2/2024 - Matter and Intelligence

Article

Architects Need to Learn to Embrace the Complexity of AI

Pia Fricker, Toni Kotnik

The future of architectural education demands a radical reimagining in how AI is integrated beyond its current role as a mere tool.

When examining the intersection between artificial intelligence (AI) and architectural education and research from a historic perspective, one can discern recurring patterns concerning the evolving relationship between computational technology and architecture.

As early as in the 1960s, the first AI-driven design methods were explored within architecture by Nicholas Negroponte, Arata Isozaki, Yona Friedman and others, and at schools like the Hochschule für Gestaltung in Ulm. Common to all these explorations was an understanding of the computer not as a machine for making drawings and artifacts but as a tool for investigating behaviour. The investigations were driven by a theoretical discourse with emerging fields like cybernetics, information theory or systems theory, resulting in radical new ways of thinking about the role of the designer and the representation of operational and relational design knowledge.

These explorations were conceptually very progressive but limited in their impact on the discipline due to the unavailability of computers for architects. This situation did not change until the development of affordable personal computers in the late 1980s, which kickstarted a comprehensive process of digitalization of workflows and the adoption of digital design technologies as almost universal means of production in architectural practice.

This ubiquitous integration of digital technology has resulted in a large amount of specific data related to the architectural design process, such as plans, sections, drawings and sketches, complemented by a huge amount of public data generated by the internet and the widespread use of social networks. Both developments, the availability of computational power and the availability of large amount of data, are driving the recent interest in AI-based applications within architecture. This brings the question of AI-based design methods and novel concepts of teaching and design thinking to the forefront once again.

AI is a large academic discipline that is focused on perceiving, synthesizing and inferring information by machines. An important topic within AI is machine learning, which aims at the development and evaluation of algorithms that enable machines to learn patterns from a dataset. Deep learning is a subfield of machine learning, with algorithms defined by neural networks, which is a type of algorithm that works very well with image-based data like floorplans, drawings or photography.

Deep learning applications, such as Midjourney, Dall-e, Adobe Firefly or Autodesk Forma, have attracted a great deal of interest in architectural practice, but they may distract from more significant conversations that are necessary for advancing the discourse within architecture and architectural education. Rather than getting caught up in surface-level discussions about sustainability or merely “greenifying” aesthetics, we must prioritize reimagining how we teach architecture in order to address complex social and environmental issues.

Many current tool-based approaches aimed at solving complex challenges often prioritize performance metrics over genuine architectural quality. Consequently, interventions tend to be disconnected from the intricate underlying dynamic systems, focusing on local optimization strategies rather than fostering creative and radical innovations that could lead to the emergence of new architectural typologies.

Rather than getting caught up in surface-level discussions about sustainability or merely “greenifying” aesthetics, we must prioritize reimagining how we teach architecture in order to address complex social and environmental issues.

As demonstrated by the pioneers of the 1960s, the computational design approach offers a great opportunity for exchanges across the disciplines and the integration of various fields of knowledge into the design process. Computational design provides the opportunity to approach the huge challenges of our time, such as climate change, loss of biodiversity or population growth, with a new mindset.

However, computation does not mean the use of computers or the automatization of processes but refers to a specific way of solving design problems by formalizing relationships and establishing spatial communication between data. To flourish in today’s world, computational thinking has to be a fundamental part of the way people think and understand the world. 

We are currently standing at the threshold of developing entirely new didactical and pedagogical concepts for academic education in the area of computational design thinking that go well beyond mainstream application-oriented topics and the teaching of tools and software for representational purposes. At the Department of Architecture at Aalto University, the collaborative initiatives spearheaded by us have cultivated a distinct educational paradigm which aims to redefine computational design thinking by integrating sustainable strategies that honour the planet’s ecological boundaries, applied and tested within research-led design studios.

By analysing natural phenomena across scales and employing computational methods, the curriculum delves into the intricate and dynamic patterns that characterize our environment. It underscores the importance of not merely applying AI but actively researching and pioneering novel ways of AI integration.

We are not facing an intellectual revolution with AI, but an availability of digital tools and workflows for the broader market. Within a short time, the quality and usability of these tools have increased enormously. It is important to embrace technological innovation and push the field towards being an active agent, not merely applying existing tools. 

The term “black box” metaphorically represents the opacity of certain automation systems, like AI-enhanced design generators, as inputs and outputs can be observed, but the processes that transpire within are largely unknown. This lack of transparency often raises issues of trust, accountability and ethics in AI deployment. As architects, we are asked to set our focus on the black box to develop intellectual knowledge for “opening” these black boxes, for improving the interpretability and transparency of AI systems and for further developing generative AI tools to support our field.

This does not refer to the need for instruction on how these tools are programmed but rather to resetting our focus from the question of “how to do it?” towards the “what is to be done?”. We are encouraged to recognize complex relationships and see specific problems as part of a larger-scale system and to solve them in this way. This goes far beyond a discussion on technological innovation, towards an understanding that technical innovations have always been and will always be part of our field.

As the current discussion on AI-enhanced design tools demonstrates the importance of rethinking the current discussion on authorship, creative design processes and ethics, it also demonstrates the underlying need to go beyond the isolated viewing of singular components and to understand the larger context and intersection points of elements within a dynamic global structure. Or as Georg Vrachliotis, Professor of Theory of Architecture and Digital Culture at TU Delft, has formulated it: “what does it mean to design in a society that seeks its balance between Artificial intelligence and the datafication of all areas of life, increasingly rapid global migration, and urgent environmental and societal issues?” ↙

PIA FRICKER is a Professor of Computational Methodologies in Landscape Architecture and Urbanism at Aalto University. 

TONI KOTNIK is a Professor of Design of Structures at Aalto University.