Form follows function: The most famous slogan in design. First formulated by Louis Sullivan in the late 19th century, it became a basic principle for rationalist design in the sixties of the last century. With machine learning, ‘form follows function’ turns to ‘form learns function’.
How does it work? In the coded example below, simplified bike shapes are generated, then tested for performance in the form of a race on always changing roads. Bike designs which won a race are kept to run against newly generated bike designs in the next race. Over the course of several generations, performance becomes markedly better, and designs become more similar.
The code here is made to run on home computers. With more complex and processor-intensive apps, more complex designs can be tested in more complex environments, thus getting closer to simulate real world conditions.
Possible applications span from transportation design (cars, motorbikes, yachts) and product design -where product features can be optimized for scenarios involving users and environments – to architecture, where building surfaces can be optimized for environmental conditions. With the addition of crowd simulation algorithms, interior designs and layouts can be optimized for human perception and wayfinding; applications here include the optimization of walking distances, the improved arrangements of shops in shopping malls, or the improvement of layouts for emergency escape routes.
This post was originally published in April 2016.