Generative design describes a process in which a designer defines technical constraints, like weight, strength, and stipulations for manufacturing, through a computer program; then an algorithm comes up with designs that fit all of the designers’ specifications. The software company Autodesk has been working on generative design for years, and when Starck approached the company with the idea of doing a project, the group decided to use Autodesk’s experimental generative design software platform to create a chair using as little material as possible. That meant inputting Starck’s creative vision and the technical constraints of the injection molding process to Autodesk’s software, which dreamed up hundreds of different chairs before Starck settled on one design–soon to be mass-manufactured this year.
The final design, which looks almost organic, with small tendrils acting as supports in unexpected places, is called “A.I.,” named so because the chair is a collaboration between a human and machine.
Over the past few years, Autodesk has worked with technical experts on generative design projects, but A.I. is unique to those conceptual proposals. None of Autodesk’s previous generative projects–like creating a space lander for NASA, car parts for GM, and a proof-of-concept super-light airplane cabin seat–have made it to market. Instead, they act as experiments to show off the company’s technology and help it design for more futuristic scenarios, unlike A.I., which is being produced within just a few months. Likewise, one of Starck’s concerns–that, ultimately, the chair was beautiful–doesn’t usually come up in more engineering-focused applications for generative design. “Those are very different requirements versus the performance-driven engineering requirements that we’re used to talking about, whether it’s high-performance motor sports or aerospace,” says Mark Davis, senior director of design futures at Autodesk.
Using software creatively comes with its own challenges. Davis says that Starck had high expectations of the software; he imagined being able to “just say what he wanted or describe what he wanted and out it would pop on the other end,” as Davis puts it. But generative design, and software that enables it, is still in its early stages, and the design process required much more human input than something made purely by a computer. As a result, much of the design work was done by people, who piggybacked on the software’s organic formulations–similar to the way machine learning algorithms are used in other creative situations. An animation of some of the different iterations shows just how mangled some of the tool’s ideas were before humans went in and refined them with clean lines, symmetry, and balance. Similarly, the software can’t do things like design chairs to be stackable, so people had to manually ensure that the finished chair would be able to stack.
The software that Starck and Autodesk’s designers worked with isn’t commercially available, and while it’s still limited in terms of the constraints it can work with–primarily, technical ones–a platform that could control aesthetic principles as well has huge possibilities for Autodesk. The company is already working on an experimental algorithm called Deep Style that can mimic a particular company’s industrial design ethos. The goal is for the computer to generate designs that are aesthetically pleasing or closer to a company’s brand identity, even if they’re less functional. “They might as well have the aesthetics of their brand embedded into the form and shape and substance of their product,” Davis says. “I think it’s going to be a differentiator if they can express design language in terms of form.”
While Davis’s idea for what generative design could eventually do sounds relatively similar to what industrial designers do today, that technology is still a long way off. Humans are much more creative than robots, and may always be. Still, one day, star designers like Starck may only need an algorithm to do their grunt work.