The National Council of Architectural Registration Boards (NCARB) has announced a consensus opinion regarding the regulation of artificial intelligence technology and its future impact on the architectural profession.
This announcement emerged from NCARB's 2024 Annual Business Meeting in June, where 150 licensing board members convened. The meeting built on earlier discussions held during the March 2024 Regional Summit and the inaugural Futures Symposium in December 2023, which explored AI's impact on the profession. NCARB stated, "Going forward, NCARB will continue to monitor the expanding use of AI across practice and its potential interface with the regulatory mission of NCARB and its members. NCARB remains committed to collaborating with its licensing board members to ensure the best interests of the public are served through the effective and reasonable regulation of architectural practice."
NCARB emphasized that while advancements in AI and computational design are providing architects with new labor-saving tools, it is essential that architects maintain responsibility and accountability. They noted that AI is a tool, not a replacement for professional judgment, and that regardless of the AI tools used, it remains the architect's duty to conform to the standard of care and remain in responsible control of all technical submissions under their seal.
The organization also highlighted key points of consensus from their regulatory community:
In addition, NCARB plans to explore guidance and best practices for architects using AI, including reassessment of responsible control parameters and identification of best practices for ethical AI usage.
You can explore the evolution of AI and its role in practice in further detail as part of our recent Archinect In-Depth: Artificial Intelligence series here.
7 Comments
As a profession, both architects and other building design professionals... I think AI can be a useful tool. AI like these evolved "ELIZA" programs adapted to various uses are tools.... AI Office Assistant (Virtual Receptionist). For example, we can have an AI Assistant (a bot or such) help us as professionals being our assistant. Fielding common questions and data (like a smart interactive form) where it helps prospective clients setup appointment and begin inquiring questions that can help with putting together our the things that would assist us in getting to know the client's project.
Ultimately, an AI is really in major core part a complex multi-tier nested and concurrent conditional logic. This kind of AI (an Architect's "Alexa" --- don't say her name out loud, she just might respond to you). In addition, the AI can be used in full robotics and that can be possibly useful for us. The point is we are in charge and ultimately in control. These are tools or assistants to us.
While certain clerical roles might change in the workplace of firms. Some jobs like traditional receptionists maybe replaced by the AI but their role be morphed into something more than the person answering the phone initially. If done well, such an AI can be useful and nicer to work with than what we have seen. There maybe something from which we can also use the tools of AI and modern technologies as design or engineering assistance tools. As computers are used do energy analysis. Adding an "Alexa"-like touch to these tools and then it does analysis as it does, it can make more human friendly feedback.
These tools are really just doing what they been doing but with maybe a human friendly touch by adding natural language processing and voice recognition and basically a voice to text process on top of natural language processing. We been doing stuff like natural language processing with keyboard input since the 1960s with ELIZA and the likes.
We must as professionals not just rely on the input as fact but verify. The tool is a tool, an assistant. Like that intern assistant who doesn't know the job of the profession to be in charge maybe tasked to do a task and needs us to verify, correct, or otherwise help it "learn" but also help it do its job better for us but as an assistant but we actually in charge of design decision making. It can help with generative processes but we may need to make correction. While "machine learning", a part of this concept behind artificial intelligence, natural language processing, could lead to it learning and doing a better job at assistance. Kind of like JARVIS in the Iron Man movies. The actual decision making rests with us. For small firms who might not be able to hire a comprehensive office assistant staff, we may be able to use such AI tools to help.
There is a number of things to consider with AI and its applications and various ethical issues.
Let that be known, as an image maker, AI is already used widely in architecture.
The strong keywords, are "new labor-saving tools".
Ethics, what ethics? Old white man's ethics? Or, the corporate copycats' ethics?
All of these sound like NCARB is testing the waters to see if it can be a party to the regulatory changes.
There can be ethics issues but I think more important than ethics issue being legal issues that are associated but there might be a point where some issues may fall in either category like the AI may generate an image but is the image actually delivering on the criteria of the design. That may be an issue but if AI tools are adequately "trained" then it might not be an issue especially if the "input" information isn't garbage. The issues that one may need to worry most about is copyright issue. Is there already an image that already exists that the AI's generated image is too similar that it might actually be an AI generated derivative. How did the AI arrive at the generated image?
There can be a number of issues but if NCARB is talking about the AI matters as part of this profession, then maybe it is a good thing than if they are keeping their head in the ground. NCARB is a party to regulatory changes because they are essentially a body that is constructed as a vehicle for regulation by the regulatory bodies in charge of the regulation of the practice of architecture.
They already have a legally mandated role and so they have a responsibility to address issues (not only have the individual regulatory bodies but also as the collective regulatory standards maker) to be addressing the issues that effect the practice of the profession and AI is a newer issue now just like the discussion about how we deal with use of computers in the profession. This is part of that scope as an ongoing part of the use of computing technology in the profession as that evolved.
If I allow a computer to procedurally generate building plans, eventually it would generate so many plans/elevations that reflect nearly every conceivable design to the point that there are so many iterations and variations produced that there isn't any more room to make a new design that doesn't violate the copyrights. You can run into the issues eventually.
If a powerful enough computer say a TOP 500 supercomputer was given a year or so, to procedurally generate designs of buildings, could it possibly generate so many designs in 1 year that is would be 300-500 TIMES the number of human designed buildings since the first human made building/shelter structure for habitation? If it did that, would we literally no longer be in the business of designing buildings because all the generated designs produced leaves no room for unique characteristics left between all those iterations. What if it had 10 years at that rate. Would it even be possible to make a unique design that the courts would recognize as unique sufficient for copyright status or will that effectively break the copyright laws as it pertains to designs of building rendering the law pointless to exist and thus remove such works from being copyrightable. If there is some means to regulate and limit such from being able to be used for such massive procedural AI generation at such prolific pace, we may actually have a design profession career. Why we can still create new designs is because we aren't that prolific. It takes us this quantity of time to produce a design that we just can't even as entire species produce that many designs. Can a computer just procedurally produce designs at a rate that can outpace what you can produce in your life time in 3-4 hours with a single desktop PC? There can be reasons to be concerned about how the tool can be used to effectively render the end of the architecture profession as we know it as designers.
NCARB should get involved when AI tools can start doing documents that are beyond schematic. Which it can't do YET...
That's reactive not proactive thinking. The problem with reactive thinking is that it is often too late at that point. So they should begin the dialogue and thinking of the issues and understand the issues of AI but not prematurely do anything. NCARB itself doesn't create or enforce laws. It member boards that makes up the owners of NCARB (the individual licensing boards) are the ones that do. They may make model laws and model rules and model policies and such aside from its role in making rules and governance policies over the administration of exams, NCARB Certificate program, and such. They should however not be lagging behind the technology too much. It is something that has importance but also worth time to be discussion and dialogue on the issues and even possible non-binding recommendations to the profession and such. It is something the profession should come to terms with AI and the legal & ethical issues that surrounds AI. So far, we just haven't completed the routines to generate technical submissions but we actually already have the software routines to generate 3d model and framing generation even from $20-$40 broderbund software from nearly 20 years ago. So it really not a far stretch for the programmers to put the pieces together for a complete AI tool for procedurally generating the design of buildings INCLUDING the technical plans.
It isn't that we can't do it with current computer hardware. We can. It is well within the means of doing so, today. The thing is, we humans (the ones in the software programming field) creates the software. The AI tools aren't exactly yet creating its own programming and extensions to itself. AI in languages like Forth would conceivably do that but most common programming languages are far more conventional and traditional than Forth even when they are decades younger programming languages. Interesting side not but it wouldn't be beyond reason for someone to actually make an AI "architect" that can produce technical submissions including doing code analysis.
Part of the hiccup to that is precisely the barriers currently in-place. The biggest thing for code analysis and such is you need to get all the laws and rules to compare the design against the laws & rules for the whole project. (Code compliance for example) The problem is jurisdictions that have antiquated web presence where they don't even have their laws online. Most likely, the AI would depend on online access to jurisdictional codes and we aren't just talking building codes and related codes but also zoning codes. It is this offline access that provides challenges with some jurisdictions. Of course, with cloud AI, we might be able to catalog and keep cataloged the website locations of the various codes and development regulations on the various jurisdictions from which services the AI "architects" app/bot that accesses that which in turn gathers the pertinent information to analyze buildings.
The AI "architect" code would need to have coded into its software coding the procedures and steps of the professional architect's delivery of services and basically model the professional practice of real life architects and figure mechanisms for determining which steps must be done in a particular sequence and which steps are someone flexible given the context of requirements. Then it must be able to figure that and learn as needed. So it would need to be largely pre-trained but able to "learn" and adapt by some degree. Can such an AI architect bot replace us?
To a certain degree, yes the means exist to do so if the AI Architect bot was created by putting the various components of AI together. We have the modules, we need to put the modules together around a central "virtual architect" AI bot that utilizes these existing components using AI technology to arrive at design all the way through code and regulatory compliance analysis and technical submission documents. It could conceivably do this today. The hardware is there to do that. The issue is still in terms of programming.
One reason it is not made is because one prime rule of wisdom, "if you want job security, you do not create the stuff that puts an end to your job". We don't create the complete AI "architect", we have job security. Don't replace ourselves with the thing that eliminates the need for our profession to exist. From what SOD said, it could be that you are no longer the architect but becomes the code compliance technical drafting specialist. It isn't a far stretch of imagination for an AI architect to eventually generate technical submissions. Analyzing code would already be adapting and applying natural language processing with logic to create structural and systems assembly that complies with those regulations and the various data points.
I see that we may have 10-20 years still before desktop/personal computers and such devices would have the computational resources to do all this and the software tools. I believe to do all this on a current gen PC would be intensive and result in possibly days to generate all that stuff from an AI "Architect" bot... not to mention, to completely create such a bot and for it to 'mature' software wise. So maybe 20-25 years before we get to the point of having the risk of losing our job entirely. It can begin to have an impact on the profession in half that time if someone starts putting together the AI "Architect" bot, today! If started today, I can see it being created in 5 years for a first version. It could take longer but 5 years is conceivable. Being ready for "prime time" could take some time longer. It is hard to predict that. Maybe 20-30 years, who knows.
The thing that will likely take longer to replace us is in the field documentation with existing buildings and such. Even though there are tools used to assist us and that can be assisted by robots equipped with Laser scan / photo-based 3d mapping and such. However, part of that role may replace us in some of the tasks but there are parts that may take a little longer and that also becomes part of the role of historic/existing building conditions assessments and such. The AI would likely be partially replacement of the human in part of the tasks. Ground based robots like fancy RC cars and drones with equipment will augment us and replace the role of assistants in some of these tasks but it won't completely replace us. I can see AI use to assists me in some of these tasks when implemented into robots that assists me in these tasks. Which then, I am also, then becoming an in-house robot designer/manufacturer using many off-the-shelf components to augment RC toys into usable robots for various task based roles using stuff like Raspberry Pi based robotics and just use these tools to assist. Have a tool that can go into the spider infested crawlspaces instead of myself.
You may notice something about the inter-relation of AI---> computers ---> robots/RC devices (RC cars, planes, boats, drones).
Block this user
Are you sure you want to block this user and hide all related comments throughout the site?
Archinect
This is your first comment on Archinect. Your comment will be visible once approved.