ARTIFICIAL INTELLIGENCE
Insights: Architecture & AI

Stephan Schütz, Executive Partner, gmp Architekten von Gerkan, Marg und Partner: We have been following the rapid development in the field of AI over the past year and a half and are continuously evaluating its potential for our work processes together with our IT team. The specialists involved are testing the various applications and making the corresponding configurations available to the design and planning teams.

MVRDV NEXT: Increasingly, yes. While we have been exploring AI models for image classification and generation for several years already, the recent developments around diffusion models and large language models truly accelerated both the research and the application of AI within MVRDV.

Kasper Rønn von Lotzbeck, Founding Partner Norm Architects: Until now we have only been testing the photo editing tools, as well as a few web based tools, but we have yet to implement it in our work. As for text editing, like Chatgpt, it is the same. As visual AI tools are now, they are tools that can make a lot of creative suggestions and variations in no time, but the decisions that ultimately shape and create the work must come from the person using the tool. In this way, it's no different to the software we already use. Only high-quality input creates high-quality output.

Davide Angeli, Deputy Managing Director AMDL CIRCLE: In our studio we use AI software, such as Mid Journey and Dall-E, to generate variations on design details, to work on the photography and atmosphere of images. This technology represents an opportunity for us as architects and designers to redefine our approach and enrich our creativity. It is not just a research tool, but an additional voice that sits with us at the discussion table, fuelling and inspiring an innovation-oriented approach.

Stephan Schütz, Executive Partner, gmp Architekten von Gerkan, Marg und Partner: We use AI-based programmes to generate images from sketches and 3D models, which produce a fascinating atmospheric density in a matter of seconds. At present, we still have to realise that these images are often imprecise random products that cannot be precisely optimised, meaning that they can only replace conventional renderings to a very limited extent. As an internationally active office, we also use AI for lip-synchronised speech and are planning to use it for our various training programmes in just a few months' time.

MVRDV NEXT: We anticipate a diverse spectrum of uses for AI within MVRDV and architecture at large. Text-to-image generators such as Stable Diffusion or Midjourney are used widely across our office for testing design directions in early concept stages, or variations in massing or materiality in later stages. Following MVRDV’s long-standing ambition of public participation and making architectural design processes accessible to a non-expert audience, we also use AI extensively for creative brainstorming in workshops and public events.

Within our research and development unit MVRDV NEXT, we develop experimental applications and fine-tune MVRDV-specific AI models. Such experiments range from energy modelling with the support of ChatGPT to embedding MVRDVs design language in custom diffusion models. We even trained an image classification network to categorise images of our rapidly expanding servers and encourage creative reuse of archival material. It is important to note that we maintain an excited yet critical position towards the rapidly developing landscape of AI tools and therefore verify results if they are used in actual ongoing projects.

Kasper Rønn von Lotzbeck, Founding Partner Norm Architects: The true benefit is the same as the benefit of non-ai software: Saving time on less important tasks which then gives us time to focus on the more important tasks.

Stephan Schütz, Executive Partner, gmp Architekten von Gerkan, Marg und Partner: In view of the breathtaking speed of development, we can only speculate about the opportunities and risks at present, although we can assume that AI will extend to almost all parts of our planning processes in the coming generations. In my opinion, the complex development of autonomous driving is a good example of the regulatory and safety-related fields of action that will be affected by this. It seems clear to me that we as architects need to develop a creative and technology-open understanding of the utilisation of AI.

Davide Angeli, Deputy Managing Director AMDL CIRCLE: Artificial intelligence presents before us the responsibility to reflect on our role, defining it, evolving our practice and adopting new tools. This form of intelligence operates on the reworking of pre-existing elements in surprising ways; however, the architect's or designer's perspective lends a unique element to the creative process: the challenge is to strike a balance between the efficiency of artificial intelligence and human creativity. This is likely to be a paradigm shift, one that we must be prepared to address, welcoming the stimuli it provides and keeping the essence of human creativity at the centre of our work.

Kasper Rønn von Lotzbeck, Founding Partner Norm Architects: The great opportunity for the creative industry is then to spend that time improving the quality of the creative work and making better projects. The risk is of course that some will choose to increase the number of projects they produce and be less critical about quality, but again, there is nothing new to that. In other words, though AI is definitely a revolutionary technology in many areas like medicine and robotics where AI can help test millions of scenarios in no time, AI is in my opinion not a revolution in our industry, it is merely a tool. Hopefully it will help us improve and become even more quality conscious.

MVRDV NEXT: Due to the rapid development of AI models and tools, it is hard to predict what impact they will have in the future. However, considering the current state of these tools, speed, as their strongest advantage, might also be their greatest risk. It is incredible how fast AI can transform a sketch into a photorealistic rendering or how ChatGPT can conduct extensive research into a design’s operational energy consumption in just 30 minutes. When you factor in other practical applications like image and project text editing, it appears architecture firms could become a lot smaller in the future. This would likely entail a significant loss of architectural jobs on one hand, but on the other hand, could help architecture firms to reduce costs and survive in an increasingly efficiency-oriented market.

However, over-reliance on AI tools and a lack of rigour in the selection or verification of the results can lead to unfeasible design proposals. Thinking more long-term, the ‘mirroring’ nature of AI can stagnate the evolution of the design vocabulary of a practice. How could an AI contribute to the development of new design approaches if the model is trained exclusively on MVRDV’s past images, and consequently, reproduces them? Similar to the impact of the digital turn, parametric design tools, BIM workflows, the internet or social media, it is certain that AI will transform the profession of the architect. We try to refrain from frantic technological optimism and dystopianism in equal measure. We look hopefully into the future, following the development of technologies with excitement and applying them with curiosity. In addition, we try to contribute to the advancement of new tools through in-house research and openly sharing our findings. Rather than a competitor, we hope and expect AI to become an architect’s powerful collaborator or even a friend.