Kirsten Crawford, architectural designer at INFORM, shares her experiences and views on the use of AI as a tool in architecture. The use of AI can drive innovation while also making design more accessible, thus shaping a more inclusive and forward-thinking industry.
INTRODUCTION
Over the past few years, we have seen a large influx of influence in the dominance that Artificial Intelligence (AI) plays in our daily lives. From image recognition features on our smartphones to advanced healthcare diagnoses, AI has acted as a strong tool that augments humans’ capabilities. AI is a technology that learns from human intelligence, observing patterns and harvesting this data to enable humans to make informed decisions, automate tasks, and solve complex problems more efficiently. This large influx has been equally prominent in the architecture industry as AI has provided new tools to architects and designers that allow for them to ponder the question: “What if?”
AI IN THE ARCHITECTURE INDUSTRY
In the architectural industry, AI has found itself making its way in through several different means, whether it is through augmented reality, energy modeling, or powered visualization softwares, it has become an invaluable tool for architects and designers. For Kirsten Crawford, architectural designer at INFORM, she has found powered visualization softwares, such as Veras or Adobe Firefly, that function as a design catalyst to drive innovation, to be the most prominent in the day to day workflow.
AI AS A SKETCHING TOOL
Many of the AI-powered visualization softwares that exist today enable designers with the ability to integrate images of quick hand sketches, simple massing geometries, or robust 3D models in tandem with thoughtful directions on what the desired output should look like. The set of instructions that are fed to an AI engine are referred to as the “prompt”, which is of similar nature to giving a chef a recipe. The quality of the prompt determines the quality of the output and therefore requires the skill of “prompt crafting”.
AI is the vehicle that drives underdeveloped thoughts to the finish line.
The outcome of the image heavily depends on the instructions that are given to the engine. Most of the AI-powered visualization softwares Crawford has tested have features that allow you to test the strength of your prompt against the original image. A lower strength prompt suggests that the image will only change slightly, while a higher strength prompt will promote the engine to create a more profound transformation.
When using software such as Veras, which Autodesk defines as an “AI-powered visualization app, that uses your 3D model geometry as a substrate for creativity and inspiration”, there are opportunities to adjust a “Geometry Override” and a “Material Override” slider bar that function similarly to the prompt strength. Having the ability to determine how much your geometry and material changes will depend on the goals for the final outcome of the generated image.
To further explain, if you were in the schematic design process and were pondering different design outcomes for the overall geometry of your design, through utilizing the Veras Revit or Rhino plug-in, it may be recommended that you have a higher range selected on both the Geometry and Material overrides to give you insight on further design possibilities within a matter of seconds.
To represent this, Figure 1 represents a basic building that was modeled in Revit and exported to Veras to explore the design possibilities. For the sake of this experiment, the prompt that was used was: “modern design with large windows, timber building, during autumn”.
Figure 2 represents the outcome of a high geometry override and a minimal material override. It is evident that the software has begun to venture off the tracks of what the original geometry is as it is now making suggestions for adjustments to the overall form with suggested window placements. Due to the reduced material override, the software is heavily dependent on the original material format. This explains why the object remains gray instead of the specified timber, indicating that the software has effectively disregarded the material mentioned in the prompt. Additionally, the software has failed to acknowledge our request for an autumn scene, as the low materiality heavily influences the background of the rendered image.
Figure 3 showcases the scenario of having a low geometry override, but a high material override. Evidently, this comes in handy for situations where the primary focus is on experimenting with different materials without altering the underlying geometry. The geometry stays consistent in this case, but represents new possibilities for new material explorations.
Lastly, Figure 4 demonstrates a high prompt strength, moderate geometry and material overrides. These settings are ideal for exploring material and form possibilities during schematic design. As illustrated in the figure, both the geometry and materials have been adjusted, allowing us to effectively evaluate various ‘what-if’ scenarios.”
DESIGNING FOR TOMORROW
In a recent experiment, INFORMers, Erik Knauss and Kirsten Crawford, questioned “How can AI be used to strengthen communities, design for inclusion, and build for tomorrow?” The resulting experiment ultimately became “how can design and construction professionals better see what is in the imagination of the people they are serving?” This is just one means to the end goal of designing for change in a design and construction sector that constantly wrestles with the pressures of change.
Designer or not, many of us ponder what our dream house would look like if we had the ability to hire someone to make it a reality. Our thoughts and dreams live within our heads and it oftentimes takes a talented architect to listen and empathize with our comments. But what if there was a tool that could take exactly what you are thinking of and make it a reality?
The experiment involved four diverse participants who were given the simple task of sketching their ideal home. To ensure a broad range of perspectives, Knauss and Crawford selected individuals with varying ages, professions, and backgrounds. The volunteers included: a 7 year-old boy, an automotive lighting engineer, a remote worker, and a professional architect, who served as a control for the experiment.
DEMOCRACY WITHIN DESIGN
Carter, a 7 year old boy who has minimal design experience, but has an imagination that runs wild, has always enjoyed sketching out his dream castle-like residence. Through feeding his sketch into AI generators, his sketch becomes a reality (Figures 5-6). Carter shared with us that this made him feel very excited and happy. The transformation from a simple sketch to a realistic image was a moment of empowerment for Carter. It showed him that his ideas, even if they seemed fantastical, could be realized in a tangible way.
Gabrielle, a 28 year old woman who is a remote worker with minimal design experience, has always imagined that she would live in a large white house with large windows, modern charm, and a long and luscious front yard. Feeding this exact prompt to an image generative AI software can allow for Gabrielle to see the possibilities of her dreams and through the power of architecture, her dreams could become a reality through the powerful sketching tool of AI (Figures 7-8). Gabrielle shared that she was impressed with the results as she usually has a difficult time sketching, and only finds herself doing so if she is playing games such as Pictionary. She feels that her drawings always come out rather flat, so the ability for her to compare her drawing to the 3 dimensional nature of the AI renderings was a very helpful tool for her to express her ideas.
Michael, a 62 year old automotive lighting engineer has years of experience in drafting and sketching, however only as it pertains to the automotive industry. Michael wanted to be an Architect when he was younger and always imagined that he would live in an expansive home with large windows. Again, through the incorporation of thoughtfully crafting his desires into a prompt in tandem with Michael’s sketch, AI will generate several images in the matter of seconds, showcasing the several different outcomes of Michael’s imagination and dreams (Figures 9-10). Michael shared that he liked how the rendered image was exactly what he pictured in his mind. He shared, “there was no way I would have been able to communicate that vision in any sketch I could have made.” He also included that he felt that there was a level of “magic” to it as it effortlessly transformed basic sketches into photorealistic images, even for those without a strong design background.
Eric, a 37 year old Architect acted as our control for this experiment as he has had several years of architectural education and experience. Represented within Figures 11-12, it is evident that the final results of his sketch are of similar nature to the other examples listed above who have little to no design experience. Regardless of one’s ability to sketch, the AI powered visualization software is capable of outputting a robust and thoughtful image that encapsulates one’s vision. Eric shared that this experiment surprised him through how close the final image was to his thinking, but also how far away it was from his vision. He shared, “there were certain things that showed up in the image that deviated from my thoughts while I was drafting the sketch. These deviations opened up possibilities to think about the opportunity of ‘the dream home’ in a different way and challenged me to explore a broader spectrum of possibilities.” He also expressed that he found the image to be extremely realistic which simultaneously could be described as dangerous. He raised an important point of consideration in which the high quality of the images could create an illusion that alludes to a design that is much more resolved or considered than it actually was. This is an important reminder that utilizing AI-generated images in the architectural industry should mostly be utilized in preliminary or schematic stages of design.
Through comparing the original sketch of the selected individual, to the final outcome of the AI generator, it is evident that there are several discrepancies between the two, however the core ideas remain intact. These discrepancies are what fuel the iterative design process as they promote latent concepts to blossom into innovative solutions. For example, in Eric’s sketch, he did not illustrate skylights on his roof, however the AI generated image suggests the implementation of them.
CONSTRUCTING FUTURES
This experimentation performed by Knauss and Crawford was inspired by a recent call from the Constructing Futures Exhibition for AI generated images that envision the future of construction technology.
Constructing Futures will highlight the future of construction and design for the built environment, with a particular focus on robotics, additive manufacturing, computational design, and AI. The installation, exhibition, lectures, and workshops will serve as a platform to facilitate crucial conversations around these emerging technologies, bringing together stakeholders from cultural institutions, universities, and industry partners.
Crawford’s personal experience of using AI as a sketching tool made it exciting to encourage others with a range of design experiences to partake in this process. In Crawford and Knauss’ exhibition entry, more agency is given back to the building user by leveraging the original “digital” medium – referring not to the numerals in a computerized display, but rather the digits on the human hand. Hand sketches are shown being taken into an artificially intelligent engine to produce an interactive holographic model, where the user is presented with multiple options in which to select basic priorities of their design and narrow in on a comprehensive solution, which then provides the design professional an efficient concept prototype to develop into a buildable information model.
Due to the exhibition’s call for futuristic construction techniques, we wanted to take our democratized hand sketches and holographic models one step further to imagine a future of construction that leverages programmable labor force assets. The data that began with the building user can be guided by the design professional through to human craftspeople wearing AR goggles which aid in their own skilled handiwork. For repeatable tasks, we imagine drone fabrication, robot installers, and autonomous 3d material printing to complete the physical product safely, efficiently, and in a timely manner. Our imaginative thoughts were quickly brought to life in a similar manner to the photorealistic images that we explored above: through plugging the same hand sketches into an AI-generated image software with prompts that highlight these future construction possibilities.
WILL AI REPLACE ARCHITECTS?
A prominent discussion that has been raised given the recent significant influx of AI appearing in the architectural industry is the conversation revolving around whether or not AI will replace Architects. It can be understood that this is a valid concern, however it is extremely unlikely that AI can replace the human ability to conceptualize, collaborate, experience intuitive thinking or contextual observations, etc. You cannot take the human out of humans. AI will continue to empower architects, enhancing their creative abilities. Instead of viewing AI as a potential threat, it could be embraced as a tool that augments their capabilities and drive innovation.
CONCLUSION
The future of architecture fosters a symbiotic relationship between human creativity and AI’s computational power. AI’s ability to transform rudimentary sketches or 3D models into sophisticated, realistic renders empowers individuals with limited design expertise. This democratization of the design process fosters a level of inclusivity that does not exist today. With this level of technology, users have the access to personalized design options that contribute to the facilitation of efficient communication amongst design professionals. Designers can leverage AI-generated insights to create more personalized and accessible environments, and to ultimately design for a better tomorrow.
AI is a vehicle that drives underdeveloped thoughts to the finish line and is a tool that will only continue to strengthen the architecture industry.
