Artificial intelligence could potentially help us streamline the more rudimentary and tedious aspects of design to free up more time for creative problem solving and response to human needs. It would be unreasonable to say that any AI that we could conceive in our lifetime would be able to intimately familiarize itself with the breadth of human experience to allow it to make accurate determinations about the things we need. Oftentimes the things we need are rooted in aesthetics that are meant to facilitate our emotional well-being or they are otherwise rooted in cultural lineages and traditions that are difficult as yet to quantify.
It is obvious that with more precise data, and with the use of more intelligent software, architects and designers will be able to enhance their work in many predictable ways such as optimizing building maintenance and security, streamlining BIM workflow, and increasing the sustainability of a design through more accurate environmental analysis. Having said this, I feel that the notion of having better tools allowing you to do your job better and faster is not necessarily worth discussing. What is of interest is whether or not improving these tools will eventually make their very users obsolete by becoming the users themselves. From this, it's important to consider what makes humans particularly useful (as users of these tools).
The threat is actually posed not by artificial intelligence itself but by users who deem AI to be a cheaper, more efficient means to an end.
What makes a thing intelligent is not only access to data but also the ability to identify connections between disparate pieces of information and use those connections to solve problems — to develop intuition. The way we identify connections between bits of data as humans is by working within certain logical frameworks that allow us to create a relationship between things that would otherwise feel random and/or arbitrary. For example, one such framework is cause and effect. You know not to look directly at the sun because it hurts your eyes when you do that. A connection was drawn between the action and the subsequent pain that informed the newly-developed behavior that followed.
Architects spend several years developing skills that give them a heightened degree of spatial awareness. Beyond that, many are also skilled at bridging gaps between their lived experience, their ability to identify sociocultural cues and traditions (humans are particularly skilled at this because we are the only species that places sentimental value onto objects, as far as our current understanding goes) and their technical skills in order to come up with clever solutions to a specific set of problems. For better or for worse, a designer’s individualism seeps through when they make decisions that influence the emotional impact of space using their lived experiences — their memories — as a basis.
For the same reason that it is difficult to imagine fully relating to another human being because your lived experiences are fundamentally different, it is difficult to imagine how digital intelligences would begin to develop a sense of ego and then use that to make aesthetic choices about a space. If that were to happen, the point would essentially be moot because we would come to a point where it would be redundant to distinguish between supposedly "real" humanity and a synthetic one. It wouldn’t be a matter of asking what the future of architecture and design looks like under the influence of AI because AI would not exist — human intelligence and artificial intelligence would essentially be the same, and so any useful distinction (those that look beyond the pedantic notion of humans being made from organic matter as opposed to AI, which is created with synthetic material) will probably cease to exist. Would AI eventually reach a point where designers become obsolete? Probably not until we reach a level of sophistication with AI that is indistinguishable from our own complexity, but in that case, it wouldn’t be a matter of AI vs. designers, it would simply mean that there are more designers.
Architects and designers should aim to liberalize pertinent data so that anybody can have access to them.
An example of how AI is already integrated in architecture can be seen at The Bartlett, wherein their space syntax software "depthmapX" can generate accurate spatial analyses that remove the need to actually visit the site. Granted, there is as yet no way for such a software to tell you, for instance, how a certain place "feels" or how culturally significant certain elements at a site are, but any physical or spatial data that can be quantified is still perfectly fair game. This not actually limited to just environmental analysis. In much the same way that analytics companies gather our social and behavioral data to essentially generate profiles on us to create more successful marketing campaigns, in an architectural setting, this data can be used to democratize development. With this data, software may be able to prioritize certain projects, calculate population growth and categorize streets or neighborhoods by usage and density (and then further categorize those things into time of day).
Still more interesting integrations of AI in architecture can be seen in an installation called Ada as part of Microsoft's Artist in Residence program. Ada is a pavilion that incorporates AI to generate a performative environment based on analyses of its users. It collects data from facial expressions and vocal tones and translates that data into certain colors and materials based on specific sentiments that it perceives from this data. What it becomes is this vehicle for a uniquely responsive architecture that allows designers to expand their conceptualization process to encompass not only what a certain building or space must be but also what it could be. The question that arises here is how this data is being perceived and translated by the AI and who programs it to perceive things in this way — and these things are determined by a variety of cultural and social biases. Perhaps the challenge will come from attempting to get the AI to understand certain illogical human behaviors that are rooted in cultural stigma such as Americans' preference for private vehicles over robust public transportation networks and infrastructure. Logic isn't standardized because culture and experience inform a person's idea of what is logical.
Logic isn't standardized because culture and experience inform a person's idea of what is logical.
The threat is actually posed not by artificial intelligence itself but by users who deem AI to be a cheaper, more efficient means to an end. As we are encouraged to indulge in our consumerist tendencies, we become less concerned with creating spaces that we can emotionally connect to and see ourselves in and more about acquiring material things. In this case, it is about acquiring four walls and a roof as quickly and efficiently as possible. While it is indeed possible for architectural firms to adapt to this and begin implementing AI technologies to help them fill in gaps in their output (such as Ada or depthmapX), larger companies that have an edge in gathering data (especially if that data is deemed proprietary) will have negative influences on the competitive environment of the field.
In order to prevent the consolidation of an immense amount of decision-making power in the hands of a small group of already resource-rich entities, architects and designers should aim to liberalize pertinent data so that anybody can have access to them. Data sets should be available for public use and perhaps managed by an international body. We see this occurring more and more frequently in the design world through the emergence of open-source programs, plans, and data such as Wheelmap, which is an urbanism platform designed to help people identify and share accessible spaces around the world. Decentralizing design in this way may prove beneficial to society as a whole by giving more people greater access to quality design that is most often reserved for people with the capital to access the finest pieces.
It would be prudent for architects to reflect on how they can synthesize the more intangible aspects of their skill sets in order to be more equipped to navigate these rapidly shifting environments.
Sebastian Errazuriz has a rather bleak — albeit realistic — perspective regarding the impact of AI on the architecture industry. Approaching it purely from a brass tacks perspective, architects are largely expendable mainly because they take a lot of time and resources to get equipped with the skills needed to become architects. Beyond that, the level of coordination between all these different entities makes it so that it's normal for projects to take 2, 3, or even 10 years to finish. How could any of that possibly compete with a program that is unbiased and unburdened by ego, that can learn anything in a matter of seconds, and that can communicate and coordinate with other equally egoless programs with complete fluidity (more fluidly than we can even communicate with our own selves). His suggestion is that architects should take their advanced spatial awareness and apply it in a tech landscape wherein they would apply their skills more abstractly to design other kinds of systems.
As with almost every other profession, architects are on the precipice of a reckoning with their roles in society moving forward. This is mainly due to the fact that we recognize that AI isn’t just a tool — it has the potential to eventually surpass our ability to do anything. What is particularly new about this is that it will cause us to fundamentally re-evaluate our relationship with our labor, and what our role in society will be if our ability (or even our need) to work is taken away. While it may not necessarily be a matter of the utmost urgency, it would be prudent for architects to reflect on how they can synthesize the more intangible aspects of their skill sets in order to be more equipped to navigate these rapidly shifting environments.
An aspiring designer interested in understanding our built environments through practice, documentation and archiving architectural heritage. I design interiors and objects and occasionally write about things with the intention of furthering the discourse surrounding political, sociocultural and ...
15 Comments
Vitruvius said well building consists of firmness, commodity, and delight. If you're inclined to agree, there's nothing to fear from AI, it'll just be another great tool. If you don't believe delight is necessary and commodity a matter of inputting data, then better look for another profession +/-.
I agree. There are certain typologies that lend themselves to some advanced automated design and planning tool - say, Amazon warehouses or hospital layout. But even there's still going to be a lot of human control and input given the sheer amount of conflicting goals and coordination issues that are inherent to such large projects - not to mention compliance and local code matters. As far space planning and "design options" go, Autodesk and friends are already incorporating that into their BIM suites, for better or worse.
the majority of my work consists of negotiating between parties who have conflicting desires and expectations for a project. The truth is even now things like room layouts or making things 'to code' is sort of just a commoditized part of the work that gets done expediently once the hard issues are settled.
Yeah, my concern is that the bottom level employees and newcomers would suffer the brunt of such new technology. The PMs and principals are still in demand for their business development and coordination skills. The workhorses and noobs who currently produce the kind of work that could be automated would be reduced in headcount. You see a similar situation in the legal industry as discovery gets automated and young lawyers find themselves competing for a smaller amount of work.
it would definitely change the role of interns and jr staff. this might be good. i can imagine it being more similar to the model of practice 200 years ago where there were no large production teams because everything could be easily drawn by a few experienced experts. possibly one would be ready to take on projects immediately at graduation, limited only by imagination and skill at negotiating with clients.
i wouldn't assume it necessarily reduces the amount of money available to perform the work. this is fungible. doctors don't get paid so much because 1 hour of their time is so much work - just because the patients value it immensely.
Artificial intelligence is an oxymoron.
the profession will be fine, it's just some dinosaurs that will go extinct.
Dinosaouria ruled the planet for hunderds of millions of years and only went extinct because of an asteroid strike. Humans have only been around for some 300,000 years and are well on the way to extincting themselves.
We’re not extincting, we’re evolving...it’s simply a survival of the fittest.
Interesting read, come to think of it AI is threatening other professional fields such as Medicine, Accounting and most aspects of Engineering. On the long run, nothing can replace human capital and design requires 60-70% of human cognitive engagement. Humans are still winning, besides we made the AI anyway.
I wonder if the nature of work itself might evolve to suit the tools available. That is, what once required human ingenuity would be downgraded to an automated and vastly more cost effective process that produces an inferior but still competitive product. In the design business, we're already at the whims of software makers whose tools we're bounded to use.
i think we are really still very far from widely usable AI tools in design. Consider how bad predictive text on an iphone still is - 80% of the time saving on that is probably related to guessing the20 most common words (are, the, they, etc...). Anything with meaning it's totally worthless. Self-driving cars remain an obscure novelty for flat sunny cities with well planned roads and homogonous traffic conditions (ie not driving in Boston during a summer evening thunderstorm).
Something like that room-layout generator is neat - but honestly I can't see a situation where I'd everuse it. Very basic situations like applying mimimum compliant bathroom layouts to designated spaces in a shopping center for example - that would be helpful. But it's like 0.1% of the labor on a large project. It would save 3 days of work for some poor intern somewhere.
I worked at a developer doing prototype apartment units and so much of the work (all of it, almost) was research, talking to customers, looking at trends in competing developments, and creative speculation on what features matter to the people who will live there. Would AI recognize when the assumed constraints (A 2-bed unit in an L-shape) are inappropriate and better to change the overall plan, or the program requirement?
--
The above is just a small example in a very specific and well-defined niche of architecture. The bigger challenge is that architectural design though often described as 'problem solving' isn't actually problem solving in the same way driving a car or playing a game is. There are rules, but they aren't all-encompassing; most of what architects do is not prescribed in codes. Deciding which rules applyand what problems to solve is the most fundemental part of architecture work.
No rules define how broad a vision should be, how compelling a design needs to be, nor how comfortable a space should be. No AI is going to be able to forumalate an argument favoring the interests of Party A over Party B, or explain to a client why it's worth pursuing a more expensive scheme that will better support their vision. It's hard to imagine AI generating looks and feels for the space that are anything beyond rearranged copies of existing models, like picking from a menu of reference pics (which we do, but no one considers it a big part of the work effort).
--
It is interesting to speculate though, imagining an AI system could exist which would perform the routine tasks of laying out plans and essential building systems according to the instruction of an architect, and then following up by developing massing and design schemes. It would absolutely change the organization of medium and large firms and the valuation of the work stages.
It would be an interesting question for licensing too. If the design is done by an AI and code compliance verified by its algorithms (ignore the substantial area of code which is subject to interpretation) - should it be the AI that is licensed? And the authors of the AI who are legally responsible for any shortfalls in the design compliance? --NB that so much of the problems with compliance are miscommunications and construction mistakes which I'd expect to be worse without a human coordinator.
One issue that rarely gets discussed is just who will be using these newfangled software and in what markets? The architects, specialist consultants, the developers themselves, or ... some new tech company that will deliver a combined service package that encompasses CM, architect, and some consultant scope? Flux was supposed to cut off the architect altogether before it downsized into a plugin.
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.