How generative AI in construction will level-up design and collaboration
„Interactive chatbots like CHATGPT have mainstreamed AI technology, and construction companies are getting on board „Several leading Japanese construction companies are developing AI that aids design, modeling, and collaboration
As AI develops, it will become more useful in construction workflows—and will help builders use their collected data more effectively
AI is a hot topic, including conversations about AI chat services, 2D and 3D image generation from text, and even AI architectural design.
Interactive AI chatbots like CHATGPT show the technology’s new applications—and shows the new precision of AI. Though the mechanism has been around for some time, the fluent and natural responses of generative AI can learn data patterns and relationships to generate new content. AI is also advancing in image-generation technology, with a series of new products that create images, videos, 3D models, and more from the text you input.
Generative AI in construction
Leading general contractors in Japan are beginning to use AI in varied ways. Obayashi Corporation, a constructor of large-scale global buildings—including the Tokyo Sky Tree, the world’s tallest tower (2,080 feet), and Singapore’s Jewel Changi airport—has been actively using AI in its projects. For example, Obayashi has worked with Autodesk Research to develop an AI platform that lets architects enter building parameters to create volumetric estimates and interior programming layouts.
In 2022, in collaboration with SRI International and Hypar, Obayashi developed Aicorb, a technology that can quickly output multiple building facade designs based on hand-drawn sketches and text descriptions, and then create a 3D model.
After conducting a volume study of the land, the proposed facade design is reviewed using Aicorb, and the generated design is integrated and visualized in a 3D model. This process is expected to dramatically accelerate the consensus-building process with the client and reduce the designer’s
Aicorb’s development began in 2017 with the initial question, “Can AI be creative?” Its main focus is proposing various designs from sketches as a generative AI specialized for architectural design. “We trained the AI to read the design intent from detailed sketches and from rough sketches,” says Yoshito Tsuji, an architect in the Asia Architectural Design Department of Obayashi’s Architectural Design & Engineering Division.
“We have prepared multiple AI models, including one that faithfully reads sketches and another that focuses on the quality of the generated results rather than fidelity, so that the tool can be used according to the designer’s intent.”
Shimizu Corporation has also recently announced SYMPREST, an AI that assists in the early design stage of structural study work. This involves studying and setting up structural framing and member cross-sections according to the shape and scale of the building plan. According to Shimizu, SYMPREST will be a digital design method that improves the efficiency of the work, enabling advanced and speedy proposals to developers.
Using proprietary databases for AI
To understand how AI services work, it’s important to note that the name of the service offered doesn’t necessarily correspond to the AI embedded in it or to the company that developed it. For example, Openai, which provides CHATGPT, is engaged in the entire process from AI development to service provision, but its base model AI is GPT-4; CHATGPT is the name of the service that exchanges information with it via chat. The Bing AI service provided by Microsoft is the same Gpt-4-based AI chat integrated into the Bing search engine, which also has access to Microsoft’s search database, making it possible to add new information and use AI.
If companies can connect to their databases in this way when using AI, they can draw from their own information in addition to pretrained information, which improves the accuracy of AI while protecting confidential information. For example, Kajima operates Kajima Chatai, which provides a secure environment for approximately 20,000 employees of various companies by building an in-house model equivalent to CHATGPT, where the input information is not used for external learning. Such examples are increasing among companies in Japan and around the world.
Helping construction companies use data
When considering the introduction and use of AI, the following three points should be kept in mind. First, consider AI from a corporate perspective. If the purpose of introducing AI as a company is to generate profit and improve productivity, it is necessary to consider how AI fits into the overall workflow. The way the company’s data is used is more important than the AI itself.
Second, be aware that AI will increase some costs. When AI is tasked with a job, the results are immediate and unlimited, but not always accurate. The cost of generation is very low, but as a result, the more you use AI, the more you are forced to review outputs, which can significantly increase costs and reduce productivity. If a company is going to introduce AI, it is necessary to consider how AI can improve productivity and to consider a mechanism to scrutinize the AI’S deliverables.
Improving productivity with AI
Finally, even though AI’S accuracy was previously thought to be guaranteed by AI development companies, improving its accuracy depends on what the AI learns (that is, its database), so it’s necessary to develop a framework to maintain the company’s database and put it into a form that the AI can use. The key is collecting information into the database and checking the contents of the database, which construction companies can achieve by using BIM and cloud services.
For construction companies, connecting AI services to business operations is a shortcut. Consider a system that allows AI to respond to construction status. To understand the progress on-site, it is necessary to inform AI of the status of the site, which can be reported using the asset function of Autodesk Build to track and manage all the assets of a project and the entire lifecycle of equipment. If the data can be visualized by the tool, the accuracy of that reporting can be verified.
The capabilities of AI and the services that use it are constantly changing, and it may not be something that can be started immediately. However, the visualization of BIM models and cloud-based databases can be carried out regardless of the use of AI. And, by proceeding with this work, it’s possible to create a situation where the database is ready for AI to learn. The first step toward using AI is deciding how to digitize your company’s information and create a path toward digital transformation.