With just a few sentences, designers can now generate phantasmal architectural renderings for their clients. Artificial intelligence may help us create beautiful, sustainable and cost-effective architecture projects. But, this emerging technology also comes with a lot of risks including: intellectual property violations, errors caused by bad data, creative job loss and more. Whether we like it or not, AI in architecture is here to stay, and how we use it will define our future built environment.
How AI Works
In simple terms, AI uses computer programs that mimic human cognition in order to generate images or text and/or solve problems. To do this, it typically utilizes machine learning, a technology that allows AI to recognize and react to a stimulus (like existing photos and designs). With machine learning, the AI system learns, remembers, and improves as it is exposed to more data, algorithms, and images.
AI for Architecture
Currently AI in architecture has the ability to analyze an entire internet’s worth of imagery and generate beautiful architectural designs and even floor plans. For some of these programs, users only need to write a few descriptive sentences and the AI program will generate a design. However, AI is still limited in its capacity to solve complex architectural problems. And it “lacks the ability to derive buildable schematics from images and compelling imagery from buildable schematics.” This essentially means that current AI architecture technologies do not yet have the ability to generate designs of buildable structures. They can create beautiful fantasy designs or simple, unimpressive, but buildable designs.
How You Can Use AI for Architecture
While AI cannot yet make a buildable project on its own, it does have many beneficial uses that are already being employed at architecture firms.
Early design assistance
In architecture, it can be difficult and time consuming to work and rework designs in the early stages. AI programs like Midjourney can quickly generate ideas that essentially replace the “sketching phase” of a project. Like a sketch, these programs generate designs that capture the essence of how the final design might look without taking into account actual engineering feasibility. This can be helpful with brainstorming and reducing the “thought-to-execution delay.” That is, the delay that can occur when showing clients initial designs because it can be difficult for them to visualize a final result. A quickly generated, detailed AI “sketch” can help reduce this delay.
Increase design to build speed
While AI cannot yet make a buildable project on its own, it can greatly increase the speed of an architecture firms’ design to build process. For example, in China, the company XKool and its English equivalent LookX, developed an AI system that has helped design, create, and produce large scale architecture projects. The goal of the company is to limit repetitive tasks to let designers focus on the creative aspect of a project. “LookX develops algorithms to quickly generate, evaluate, and recommend schemes for architects that take account of local regulatory requirements while providing real-time cost analysis.”
Zoning and Regulation guidance
AI tool Maket has the ability to sift through zoning data and answer questions about these regulations. This can be very helpful in the development phase of a project, where architects begin to take their designs and apply zoning laws and engineering capabilities.
An AI tool called Cove.tool can help architecture firms optimize the performance of their designs. For example, the tool has the ability “to analyze how building designs can improve their energy and carbon consumption, daylighting levels, cost structures, and more” It does this through machine learning that allows the program to hypothetically alter building materials and building position to produce the best possible design result.
AI Still Has a Lot of Problems to Work Out
Despite its usefulness, AI has serious potential problems and implications for architecture the industry.
In the realm of technology, AI is still a baby. As such, the law, particularly pertaining to intellectual property, hasn’t quite caught up yet. Although AI generated content may appear as new content, it is actually an amalgamation of a lot of different data. The question thus becomes, who owns what AI produces? And what data should AI be allowed to access? We still don’t know how the law will handle these tough questions, but we do know from the advent of the internet, which sparked similar concerns, that technology to identify instances of plagiarism and violations of copyrights will likely follow suit.
While many may boast about AI’s potential to enhance sustainability, the truth is that AI has a massive carbon footprint. In fact, it is estimated that by 2040, the Information and Communications Technology industry will account for 14% of carbon emissions worldwide. And, in another study, researchers at the University of Massachusetts found that training an AI model “can produce about 626,000 pounds of carbon dioxide, or the equivalent of around 300 round-trip flights between New York and San Francisco – nearly 5 times the lifetime emissions of the average car.” This is because it takes a lot of energy and space in the form of large factories and many devices to create and power AI.
Hazards and Biases From Bad Data-
AI really is all about the data. So it is imperative that that data be free from bias and error. Unfortunately, AI is prone to both error and bias. For example, in architecture, AI may favor images and data only from Western countries, or that are only written in English. This leaves out a lot of styles and knowledge about architecture from around the world. Other serious errors may include bad data about safety and engineering standards that could create serious hazards. Moving forward, it is vital that AI companies control their data for accuracy- especially in a crucial industry like architecture.
One of the biggest fears when it comes to AI is that it will replace our jobs. Or, that it will replace all the creativity in our jobs and leave us to act as fact checkers. Unfortunately, some of these fears might be legitimate. While early design assistance AI programs might save time and be convenient, they do sometimes replace more creative jobs like generating initial sketches. However, AI still doesn’t have the capability to completely replace any job. Midjourney might create a design worthy of a fantasy world, but cannot create a buildable design. LookX and XKool may be able to create some buildable designs, but they still have to be reviewed and reworked by humans. Still, it would be naive to say that the field of architecture will stay the same. More likely than not, AI will fundamentally change architecture.
Time Will Tell…
Only time will tell what impact AI has on architecture and engineering. However, as this technology advances it seems increasingly likely that it will profoundly change how we build and work.
If you have any comments, questions or suggestions, please use the comment section on the bottom of this page. And don’t forget to check out our blog for more interesting content!