Automation is a word that has become, in equal measures, intriguing and terrifying in many industries. The discussion around automation turns increasingly political and contentious as jobs in many markets begin to evaporate due to the increasing availability and affordability of automated alternatives. Despite this, architecture is one field that has thus far remained relatively unscathed by the negative effects of automation, and the reason is fairly simple: It is extraordinarily difficult to replace the effect that the human touch has on architecture and design.
That is not to say however, that automation is not making waves in the field of architecture. It is easy to imagine huge skyscrapers designed entirely by AI when you begin to associate automation with architecture, and this is something that has been explored to some degree. ArchDaily has documented several cases of artificial intelligence and machine learning being used in the design process, and the results are often… Interesting.
The pictured experiment, titled Evolving Floor Plans, took the floor plan of an existing elementary school in Maine, taking into account the square footages and general layout of existing spaces, and reconfigured them along the lines of different criteria such as ‘optimal traffic flow,’ ‘minimal material usage,’ and ‘fire escape paths.’ The results are eye-catching to say the least, but while the resulting generated floor plans may be more efficient in some metrics, ultimately, they sacrifice in many others. First of all, the organic forms suggested by the algorithm sacrifice greatly in buildability. The human experience is also basically out the window, as spaces like gymnasiums have been rounded out into circles despite being typically restricted to specific dimensions to accommodate various sports and activities. The hallways are also a complete mess of branching hallways of varying sizes, shooting off at all kinds of different angles and trajectories. Imagine actually navigating a space like that.
While there is some value to exploring ‘optimized’ layouts of spaces, ultimately, the inability of the AI to consider the human experience when generating building layouts severely hampers it from being useful at this stage of the design process. The resulting bubble diagrams may provide some useful insights, but generally speaking an experiment of this nature does not offer much more than some interesting art pieces, and a springboard for conversation about automation in architecture.
On the macro scale, as we have seen, automation is in an extremely primitive state when it comes to architecture, however many interesting strides have been made in automation at medium and small scale applications within the industry. Another interesting experiment presented by ArchDaily is an algorithm called Finch which generates multi-family unit layouts based on size/dimensional constraints.
One of the biggest hurdles in the early stages of designing multi-family projects is fitting the desired unit count and mix within the constraints of the site. Many firms use standard unit layouts and sizes at this stage of design already, and the ability to quickly push and pull these standard units could allow designers to much more quickly generate multi-family layouts that address client needs and site constraints.
The ability to automate things like unit layouts, especially at the early phases of the design process, could allow us to deliver more accurate and detailed schematic floor plans to clients, while affording us the ability to focus more intently on things like common area layouts and overall building form. Though this technology is not readily available in a working state yet, when it does become widely available it could revolutionize the early stages of multi-family housing design.
Though it doesn’t exactly make for flashy headlines, some of the most useful and realistic applications of automation thus far in architecture have primarily been at the micro scale. By eliminating much of the workload imposed by the many time-consuming and esoteric tasks involved in the creation of construction documents, such as generating door schedules, drawing connection details, and numbering doors and rooms, architects can hope to expend more time and effort doing quality design work instead of getting bogged down by busy work.
The introduction of BIM software, such as Autodesk’s Revit, as the industry standard for architectural drafting has already helped make massive strides in streamlining and simplifying the drafting process. While Revit has already made some tasks like scheduling doors, defining wall assemblies, and tagging important building elements significantly quicker and simpler than it used to be. Working within the framework of a BIM software comes with its own foibles as previously simple tasks like drawing accurate staircases and troubleshooting errors have generated new headaches of their own. However, the beauty of BIM software is that it is extensible in nature and allows the end-user to create their own solutions for approaching some of these tasks via Add-Ins and Extensions. Revit allows the creation of scripted ‘macros’ to automate some smaller tasks, and recently has also introduced its own visual programming tool called Dynamo which takes inputs and outputs from Revit and allows the user to modify data within their projects via scripts.
This is where the biggest and most significant strides have been made within the realm of automation in architecture. As designers are able to simplify and reduce their time spent on minor tasks, they can better focus on delivering quality projects and good designs to their clients. Automation is not replacing people in architecture; it is enabling the people within the industry to do their jobs better. The future is not going to be cities full of buildings designed by AI, instead, it is going to be a future where architects can ultimately provide better work because they aren’t bogged down by the menial tasks involved in design and drafting.
A quick look to the future shows what this idea could look like in practice. Recently a BIM software called Modumate became available in a limited and early access form. This software promises to be a one stop shop for building design, allowing designers to place and tweak building elements in a real-time 3D modeling space. The software claims to then automatically generate all of the drawings, details, and schedules required for construction of the building with minimal human intervention. On their own website, they claim that designers could “get back 50-60% of the time spent on documentation.” While the lofty promises of this particular software can easily be called into question, the idea behind it is still incredibly promising. What if we could design freely and cut out all the unnecessary fluff? Though we may still be a long way away from something like this becoming reality, the mere existence of a concept like this foreshadows and incredibly promising future for building design. The field of architecture may be able to survive automation yet, by simply embracing it and developing a symbiotic relationship with it, rather than letting it supplant the human element that makes our work so special.