Can computers learn and become designers? – Generative design
Jeff Kowalski, Chief Technology Officer, Autodesk
CAD is a Lie: Generative Design to the Rescue
We have all heard of CAD—computer-aided design. That term however, is a misnomer. It should stand for computer-aided documentation. Think about it. The computer isn’t aiding in the design process. YOU are doing the design, the computer simply documents what you envision in your mind.
So how can a computer be used more effectively in the design process? What if you could tell a computer what you wanted, or needed, and it could create options for you? Computers that creatively come up with design ideas on their own is the heart of what Autodesk calls generative design. In generative design, you give the computer the specifications and constraints involved, and it will explore possible solutions, many that you may not have even envisioned.
Biologist D’Arcy Wentworth Thompson made this comment in 1915,
“The form of an object is a diagram of its forces.”
Today, Autodesk is taking this concept literally with Project Dreamcatcher, a research project that lets designers describe the forces that act on an object and then lets computers go off and make it. These forces can be structural loads or even manufacturing methods.
It is a method of thinking that isn’t typical. In Dreamcatcher, you start by sharing the goal with the computer, telling it not what you want it to do, but what you are trying to achieve. You describe your problem, and the computer creates a large set of potential solutions using cloud computing. This may not seem too exciting, but here’s the thing … in the time it would have taken you to do one design, Dreamcatcher has done all of them.
All its design proposals are delivered to you in an explore tool, and you can start combing through the various designs, understanding tradeoffs. Through this process, you may tweak your parameters to repeat the process, but ultimately, you will select one of the computer’s designs to fabricate.
There are computers today that after being shown millions of images as a training set, can correctly label an image, based on what it is “seeing” on the image. That’s pretty amazing, being able to tell that a picture is of a group of people playing Frisbee, or a picture of two pizzas on a stove top. Autodesk is taking this same idea and applying it to design software and analysis.
What if there was machine learning as part of the analysis process, so that any results based on any analysis is remembered, and the computer got an impression of connections due to cause and effect? What would the results be if this repeated itself, or there was variation so it could “learn”? This is the type of technology that is in Dreamcatcher.
With a world that is more reliant on imagination and innovation, the possibilities of generative design are almost limitless. It’s exciting to think that we are approaching an age where instead of computer-aided design, CAD could represent Computer As Designer!
As Chief Technology Officer at Autodesk, Jeff Kowalski is responsible for shaping the company's long-term technology vision and driving innovation by exploring big ideas. A seasoned entrepreneur, he has earned a reputation for facilitating a rich diversity of technical, business, and entrepreneurial ideas from diverse sources, including engineers, technologists, marketers, policy experts, and others.
Kowalski holds a Bachelor's degree in electrical engineering and a Master's degree in computer science, both from Cornell University.