Artificial intelligence can now create compelling images of buildings


There’s a new craze in town. Recently, designers typed prompts into a broadcast-based artificial intelligence (AI) platform and waited for images of never-before-seen buildings, logos, products, and more to materialize in seconds. Platforms like Midjourney are built on datasets of billions of existing images pulled from the web. In this vast library you will find images of buildings, birds, balloons and beaks, so if a building in the shape of a bird with a beak made out of balloons is something you are looking for, type it in and Midjourney will take you will deliver it. But beware, it’s addictive. In less than a month of using AI, I created 11,515 images.

Midjourney (or DALL-E 2, Disco Diffusion, Imagen – there are many versions and more to come) is a text-to-image AI. Simply put, it’s a web platform accessible via Discord (think of a chat room) where you type “/imagine” followed by a prompt, which is a description of what you would like the AI ​​creates. The limit is your imagination. For example: “/imagine a small house made of dinosaur fossils”. Hit return and the AI ​​analyzes your prompt, searches its database for images to pair with your text, then constructs four completely original images from a random pattern of dots. You have the option to scale (add resolution) and edit any or all of the four images or run a new iteration of your prompt. The AI ​​fills in whatever you left out of the prompt with items related to objects and parameters that would typically be associated with the content you provide. So if you forget to include “doors” and “windows”, the AI ​​will, in all likelihood, add them for you. On the other hand, if you want to exercise a little more control or replace the doors with beehives, add more details to the prompt (color, material, surround, mood, view, lighting, aspect ratio or even style) and run again. Do this over and over (I’ve found the more scaling and variation the better) until you get something that impresses you. Or, if that particular thread doesn’t do it for you, type in a few more words, and off you go for your next co-creation.

Soft and Hairy Architecture (Virginia San Fratello/Midjourney AI)

These images are designs co-written (if you want to call it that) with the AI, so I didn’t do the heavy lifting. We’ve designed houses on a lake, skyscrapers in Manhattan, hotel lobbies for a future where 3D printing and robotic manufacturing will be ubiquitous, apartment blocks shaped like letters, cities made of ingots , and even the background for our school’s back-to-school poster 2022 Lecture Series. With this AI, there’s room for conventional design, cutting-edge speculative projects, and world-building utopian (or dystopian). It is difficult to find the limits of its design capacity.

I have no doubt that this will be a complete game-changer, not only for architecture but for all creative disciplines. AI is already deeply embedded in our lives (targeted marketing, self-driving cars, facial recognition), so it was only a matter of time before it found its way into architecture. Soon it will be in every office, every school and every smartphone, playing a part in the design process. The entry threshold is minimal. For the first time, we have state-of-the-art design technology that experts and non-experts have immediate and equal access to. It may seem that such ease of access will contribute to the devaluing of expertise that affects many professions, including architecture, but I don’t think so. Artificial intelligence is surprisingly good at composing and cross-referencing a complex web of architectural stories, styles and contexts – things that even experts in our field often lose sight of. And when a vast majority (some say 75%) of buildings are not designed by architects, then not having AI capable of designing in the hands of others would be much worse: the proof is everywhere.

an architectural image produced using a broadcast-based AI platform
Models of adaptive reuse, suburban densification and urban agricultural hybrids (Jose Sanchez/Midjourney AI)

Of course, not everyone agrees. The images posted on social media received a fair amount of criticism. While the number of users is rapidly reaching gold rush proportions, there are plenty of skeptics. The most common refrain is a variation of “It’s just a picture; when will he draw sections? It takes a lot of work to go from a two-dimensional image to a fully three-dimensional building, but that’s what architects do. It may be the photorealistic quality of the images or the perceived existential threat to our livelihoods, but these are sketches, and what could be more stereotypical than an architect sketching on napkins in the first meeting with the client? What if our towel sketch was now a highly detailed and realistic representation of a fully formed building proposal? It’s still just a sketch.

an architectural image produced using a broadcast-based AI platform
Timber Pavilion Series (Andrew Kudless / Midjourney AI)

Like Midjourney, the architects guide us through various resolutions. We go from a concept to a schematic set of drawings, which we then develop until we finally complete a set of construction documents. Those worried that AI won’t give us blueprints, sections or a 3D model (yet) might be unsure about the enduring role of architects or just looking for a reason to ignore it, but that’s not the first time that the practice of architecture has been completely transformed, and it will not be the last. We are always experts, and our expertise has value, even in the face of AI. You shouldn’t hope or be afraid that an AI will solve everything; we should be excited about the opportunities and creative tangents this will provide along the way.

an architectural image produced using a broadcast-based AI platform
Merging nature and housing (Rasa Navasaityte/MidjourneyAI)

Technology has the ability to transform the work of architects, and as a community we need to approach its offerings with openness, perseverance, optimism and, yes, skepticism. We need to be lucid about the pitfalls and ethical issues surrounding AI as we move forward. There will be new issues around labor, energy, authorship, copyright, representation and appropriation, all of which will need to be addressed. But if we move forward with the goal of building the capacity of architects to contribute to contemporary material and cultural discourses, then AI has the potential to expand our influence and help us be effective agents of change. We must participate in the development and use of AI to ensure it meets the needs of the profession and those we serve, otherwise we risk losing even more ground to those who prioritize profits. to progress.

Kory Bieg is director of the architecture program at the University of Texas at Austin and director of OTA+.

Previous Deep Learning for Quality Assessment of Optical Coherence Tomography Angiography Images
Next Can innovative industrial properties survive the rest of 2022?