AI can already dream up imaginary celebrities, so perhaps it can help the Army imagine revolutionary new engine parts or aircraft, too.
That’s the goal of a new project from the Defense Advanced Research Projects Agency (DARPA), the research wing of the US Defense Department. DARPA wants entrants to rethink the way complex components are designed by combining recent advances in machine learning with fundamental tenets of math and engineering.
AI is increasingly being used to imagine new things, from celebrity faces to clothing (see “The GANfather: The man who’s given machines the gift of imagination”). The systems being used to conjure up new ideas are still in their early stages, but they show a path forward.
Machine learning is also already used in some areas of design and engineering, but the DARPA project aims to apply it more broadly, and to the crucial task of determining function and form. “We are using very few computational tools,” says Jan Vandenbrande, the DARPA program manager in charge. “It’s very artisan.”
One project selected for funding by DARPA is D-FOCUS, from researchers at the University of Wisconsin-Madison and PARC, the research company spun out of Xerox.
D-FOCUS doesn’t come up with new designs from scratch but offers up alternatives to existing designs. If the early phase of the design process is automated, a human designer can explore more design options and compare trade-offs with each option before committing to a potentially very expensive plan, says Johan de Kleer, the PARC lead on the project.
Under the DARPA challenge, software has to come up with designs for machines that can solve classic engineering questions, like how to transport water uphill.
Using hard-coded laws of physics along with functional requirements provided by a human designer, D-FOCUS can explore potential design concepts. For the moving-water-uphill problem, for instance, the system suggested using the Leidenfrost effect—a phenomenon where water droplets on a very hot surface create a thin layer of vapor beneath themselves, causing a repulsive force that makes the water hover above the surface. The researchers admit that this concept is largely impractical, but it is the type of out-there thinking that can push designers to come up with innovative designs.
DARPA has a long history of backing early technologies. The DARPA Grand Challenge was the first long-distance competition for driverless cars, back in 2004, and it kicked off the current boom in self-driving technology. More recently, DARPA funded an Explainable AI (XAI) program to develop new AI systems that were easier for humans to understand.
Mike Haley, Autodesk’s senior director of machine intelligence, says AI could expand design beyond boundaries imposed by the bias and groupthink that humans can succumb to. “We are going to think beyond our brains and come up with ideas that we would have never come up with before,” Haley says. “It’s like having the world’s most wonderful mentor.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
Unpacking the hype around OpenAI’s rumored new Q* model
If OpenAI's new model can solve grade-school math, it could pave the way for more powerful systems.
Generative AI deployment: Strategies for smooth scaling
Our global poll examines key decision points for putting AI to use in the enterprise.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.