Auto Generated Architecture
This article is written after much internal debate over the last few years on the ideas of optimization and the quest for solutions of a fittest model, especially in lieu of new rampant generic parametric models. It is as much a critique on my own past ideologies as those which are becoming more pervasive in parametric architectural modeling. The trigger is the relationship shared between the book “Rethinking A Lot” and a parking generator definition developed in Grasshopper for architectural firms and students to quickly layout large acreage of parking spaces–the vast suburban impervious sea awaits. While the parking lot generator tool may be useful in the mundane generic sense for its aspects of a plug-and-play stick-anywhere architectural typology that has come to inundate the post-industrial environment, this specific use of the “parametric” tool has highlighted many of problems which have plagued and will continue to plague architecture and ecologies.
What is more desirable to an architectural intern than to nonsensically pick a scaleless bounding limit within a void and generate a parking lot with absolutely no theoretical discussion, context, or forethought as to the design of that space and the impacts it will have on the surrounding network? Furthermore even more troubling than the parking generator is the quest for pseudo-”optimal” solutions whereby program, mass, and space magically transform into “optimal” solutions based on nondescript inputs from an excel sheet and subjective variables. Questions quickly arise as to what or for who are these less than optimal solutions? Optimal solutions are typically sought through the introduction of variables in the form of fitness criteria, however, even these criteria are constantly shifting and fluctuating. While for moment you may be getting an “optimal” solution, you are by no means addressing the full spectrum of variables to optimize toward, furthermore the fitness criteria are constantly shifting and in a state of flux.
While I am a staunch supporter of digital tools and the benefits they hold, including parametric and algorithmic modeling for the benefits it affords the designer with choices, customizability, and in bridging the gap between architectural design and craft in construction (fabrication), to what degree do we allow parametric scripts to define the design process, “program”, and space. In fact this approach raises more questions than it answers, and highlights more problems with process than the “solutions” it seeks to derive.
This one-click approach without the forethought of the attributes, data, or variables being plugged into the scripts is an increasing problem without the proper critical assessment of the technology and its consequences. This becomes clearer as the software’s utilization moves away from one of mainly surface treatments toward a greater role of planning. Attempts range from auto-generating parking lots to the layout and sectional relationships of entire buildings based on extruded area takeoffs. This approach may be well and good for the Toll Brothers, but how in any scenario is this seen as a viable solution for a profession plagued by suburbia, the shopping mall, and the International Style which has crippled our urban areas. This use of parametric software only exacerbates the problems we are working hard to recover from and fight against.
Architecture, it would seem in these instances, are moving the way of the evolution of manufacturing where manual labor is replaced for robotic processes, where human intervention as sense are negated for pure unhindered work of menial repetitive process. However in this scenario we are not trading off manual labor to create gadgets but intellectual capital. As with modern architecture of the past, quantitative means still aspire to trump the qualitative. Rather I am interested here in an architecture which creates a conversation between the two; where the quantitative means of data influences the qualitative information experienced by the human being to explore meaningful and sustainable connections. This does not negate the parametric or scripted processes, but challenges and critiques the data and meaning being put into these tools and realizes that these are not universal models to be copied and pasted generically.
While these approaches of pseudo-optimization and nonsensical auto-generation may seem like the Holy Grail for practicing architects focused on Fordist Logic and machine-like efficiency, we must stop to question their validity and intention. In fact what are the intentions here? To create sustainable spaces for people in every conceivable aspect (optimal)? Or to stamp out generic buildings at the fastest possible rate (optimal)? We must be sure we are posing the correct problems, not how many parking spaces can we cram into an area or how much land we need to clear of life for the automobile, but begin with whether these spaces are needed to begin with. And then with the minimal amount needed, begin to question how this space can be sustainable to environment, spatial relationships, and experience. How can we better design the environment around us with thoughtful insight.
Questions such as these cannot be answered by auto-generative processes such as the parking generator that take into account variables and rule-sets which in themselves need to be critically challenged to begin with. Furthermore, processes such as these begin to take a tool that offers customizability, instead championing it as an auto-generating generic modeler. In the end the problem is not of the software but of ourselves, to break age-old habits of expediting seemingly menial tasks for the “larger picture”. We must focus instead on the details, posing the correct questions and problems before seeking less-than-optimal solutions. It is time to be critical of our choices and the trajectory of our digital current.