Pixel and Mortar: AI is Optimizing the Design Process

Before we look forward, let’s examine the reasons for pockets of resistance within the profession to embrace this changing technology. Foremost is that architects have well-established workflows that have been refined, improved, and directly curated through years of learned successes and failures. A changing architectural landscape instinctively leads one to react, and the fear of the unknown creates an aversion to change.

At first, AI technologies can be perceived as complex and unfamiliar, creating feelings of worrying anxiety related to a mixture of lack of creative control, fear of client acceptance, and fear even of perceived job security. Let’s think back to the digital revolution, where architects moved from pen and paper to computer drafting - architects were resistant to change for similar reasons. We feared a loss of artistic expression; feared we would depend too heavily on technology to succeed and were of course concerned with the impacts it would have on traditional design processes. Yet, as CAD became more advanced, more widely known, and easier to use, the benefits of using such a system began to outweigh the perceived doubts. Drafting became more efficient, precise, and revisable, and collaboration within teams greatly improved. Today, it’s an integral part of successful architectural practice and has allowed designers to create innovative and complex designs with more accuracy and efficiency than ever before.

So why then, are we so hesitant to adopt new technology like AI?

Like with all things, change takes time. Ethical dilemmas coupled with steep learning curves and challenges to established workflows nurture seeds of fear and doubt in many practices. There is the fear that maybe this whole thing is just a fad. Remember when parametric architecture was the buzzword of the future? It promised an inherently iterative design process, the creation of complex geometry, and was infinitely flexible and adaptable. That said, its application was challenging, and many architects moved towards more intuitive design methods. AI feels like the next generation of parametric design but with more applicability.

While parametric design and AI are two distinct concepts, they complement each other in a variety of ways. Parametric design uses algorithms and computational tools to generate and manipulate complex forms, while AI refers to the simulation of human intelligence in machines that can analyze and learn from data to make informed decisions. Parametric design offers high flexibility and adaptability while AI can make suggestions based on trends as its flexibility is limited to the data it has been trained on. For human involvement, parametric design relies heavily on human creativity, expertise, and input to manipulate and explore the parametric model. AI on the other hand can work independently to generate design solutions or provide design insights but it still requires human guidance and validation to ensure results. Finally, while parametric design is a valuable tool for creating innovative architectural expressions, AI has a broader scope that can be applied across various aspects of architecture. Think data analysis, energy modeling, material optimization, and design conceptualization.

While there are similarities in both tools, AI has a broader reach, and the fact that it’s autonomous or semi-autonomous potentially adds many layers of support over something like parametric design. There are already dozens of AI tools focused on architecture. Large language models like ChatGPT and Bloom allow architects to communicate more effectively and in multiple languages. Image generative tools like Midjourney and Dall-E 2 encourage designers to think outside of the box and allow them to quickly iterate through potential conceptual design options. More focused tools like Veras directly integrate into an architect’s existing 3D toolbox i.e., Revit, Sketchup, and Rhino to conceptualize masses with existing context. Exploring each of these involves a deeper dive into what they are currently used for, how they can be applicable to our workflow, and how they can improve our design thinking and process.

• ChatGPT (Generative Pre-trained Transformer) is a conversational AI language model used to generate human-like responses. Its purpose is to assist in a wide range of tasks including answering questions, providing explanations, creative writing, language translation, code assistance, and more. It can carry out dynamic conversations making it valuable for question series as it understands and uses previous questions to provide context to future ones. For architects, the uses for programs like this range throughout the entire design process. Designers can use ChatGPT to brainstorm and explore design ideas by describing project constraints. ChatGPT can then generate design concepts or suggest alternate solutions providing a range of design possibilities for consideration. It can be used to help answer building code questions. When creating presentations, architects can use language models to generate descriptive narratives and can use them to articulate concepts and explain benefits to clients and stakeholders.

• Midjourney and Dall-E 2 (herein called Dall-E) are artificial intelligence image-generation tools. They are neural networks designed to generate images from text descriptions and are commonly used to augment artistic expression and create unique compelling artwork. There is a range of ways that these exciting tools can be used. They can be used to augment existing artwork with different styles to explore various rendering and visualization techniques and can be used to generate completely new images from text prompts. These powerful visual tools can also be used to generate complex design ideas and explore different possibilities and potential directions for a project. It can be used to brainstorm concepts quickly and allow for infinite possibilities to be explored.

• Veras is a visualization add-in for Revit, Rhino, and Sketchup which uses 3D model geometry as a substrate for AI-powered iteration. It can be used to ideate and iterate faster giving multiple designs in seconds freeing up designers to either work on other tasks or broaden their sights onto multiple design options that would otherwise take longer to model and texture. Its creator, EvolveLab, also has other computational and AI-driven tools such as programs that offer Revit performance analytics, sheet generation, tagging, and dimensioning, real-time iterative space planning, and more.

Like with anything else, using these tools requires some level of honing skillsets. Acquiring useful images from Midjourney and Dall-E requires knowledge of prompt structure to get useful results. The same goes for Veras, where prompt design has a large impact on the result. ChatGPT is more forgiving in that sense since it’s designed to respond conversationally, however, as is often the case, it comes with drawbacks and caveats. One of which is telling lies with confidence. It has no way to confirm what it’s saying is true, and so in that vein, it requires constant checking and validation through trusted outside sources.

While AI presents an obvious array of benefits, designers must also cautiously approach its adoption with a clear understanding of the potential pitfalls. Ethical complexities, maintaining the human touch, clear decision-making processes, cultural nuances, and even maintaining human-centric design principles, to name a few.

We have utilized various AI tools like ChatGPT, Midjourney, Dall-E, and Veras in our design work for clients; ChatGPT has been used for brainstorming, concept description, precedent-building analysis, and generating statements. Midjourney and Dall-E were employed for creating conceptual images and brainstorming architectural ideas. Veras helped with materiality, lighting, and visualizing building styles. These tools work in tandem to expedite aspects of our design process, allowing us to focus on the finer details and project efficiency.

AI tools are increasingly finding their way into architect’s workflows and revolutionizing the design process. Their integration is emerging as a transformative force that promises to reshape the very foundations of design. Designers are no longer confined to the traditional modes of creation and are instead empowered by AI-driven tools like generative models and language processing which offer a range of benefits and challenges for designers and their teams.

AI can act as a collaborator, and enabler, and amplify innovative ideas, co-creating designs that balance aesthetic and functional pragmatism. It can automate repetitive tasks, generate design alternatives, and analyze complex data which encourages architects to make data-driven decisions. Of course, the journey is not without challenges and ethical considerations as designers grapple with ownership, authenticity, and preservation of the human touch.

However, as we continue to advance, the digital/pixel and physical/mortar worlds will continue to converge as the symbiotic relationship between artificial intelligence and designers flourishes and evolves, enabling us to utilize its full potential to advance our collective profession.