AI Architect Esther: The Effectiveness of Manual Modeling is Overrated
Esther (@SjwEsther), a professional architect now transitioning to the AI field, posted on September 5, stating that while using the GPT-6 Astra model, she found herself in prolonged contemplation and reflected on the traditional training methods in architectural education.
Esther recalled that during her seven years of studying architecture, she spent a significant amount of time and effort creating physical models to explore spatial relationships; however, she later reflected that the part of the manual process that truly contributed to spatial exploration might only account for about 10% of the overall workload, with the majority being repetitive manual labor rather than genuine spatial thinking training.
She mentioned that as modeling capabilities have advanced to accurately handle complex spatial geometries and structural relationships in architectural design, she has re-evaluated the time she invested back then, concluding that the intense manual training phase had limited actual value in enhancing her design abilities.
However, Esther also stated that despite the reassessment of the value of manual modeling training, the "true thinking, exploration, and research" process in architectural training—finding a balance between practicality, structural integrity, and aesthetics—remains applicable even after her transition to the AI field at the age of 30.
Following her viewpoint, discussions arose in the tweet interaction area regarding whether architectural education should adjust its training focus due to the upgrade of AI tools, with some responses focusing on the boundary issues between manual training and tool dependence in design education.
Source: Public Information
ABAB AI Insight
The replacement of manual labor in the architecture industry by tools is not a new phenomenon—AutoCAD's popularity in the 1980s directly ended the traditional hand-drawing profession; after Autodesk acquired Revit in 2002 and established it as the BIM industry standard, the "drawing" process for architects was replaced by 3D data modeling; in 2007, Grasshopper, a parametric design plugin for Rhino, allowed leading firms like Zaha Hadid Architects and Foster + Partners to generate complex surface forms algorithmically, replacing the traditional method of repeatedly testing spatial relationships through manual models.
In 2020, Autodesk acquired the Norwegian generative design company Spacemaker for approximately $240 million, directly integrating AI-driven urban planning and site generation capabilities into its BIM product line; in recent years, generative architectural design startups like Testfit and Hypar have received venture capital investments, with these tools aiming to automate the preliminary design phase that previously required architects to manually test and refine volumetric relationships—capital investment is directed towards "reducing the labor hours in the early design phase."
This follows the same technological substitution curve as CAD replacing hand drawing and Photoshop replacing hand-drawn renderings—after each tool leap, debates arise within the industry about whether past training was in vain; the current stage of architectural design is transitioning from "parametric modeling" to "generative AI directly outputting feasible solutions," similar debates have also occurred in the structural engineering industry during the shift from manual force analysis to finite element software calculations.
Essentially a form of "technological substitution," but the level of substitution is worth distinguishing—at the mechanism level, AI and previous tools have always eliminated the repetitive labor at the "execution level" (drawing, modeling, rendering), while the "judgment level"—the design decision-making ability to balance practicality, sturdiness, and aesthetics—has never been truly replaced by tools, which is why those who remain in the industry after each round of tool upgrades are always those who spend their time on "judgment" rather than "execution."
ABAB News · Cognitive Law
- What is eliminated by tools is always execution, what remains is always judgment.
- Every technological leap questions whether past training was worthwhile.
- What will be replaced is the hand, what will not be replaced is the choice.