General Motors will research and develop advanced driver assistance systems (ADAS) technologies with the aid of a U.S. national lab’s artificial intelligence system, the company announced April 27.
Artificial intelligence (AI) systems known as neural networks mimic the way the brain works to carry out tasks such as detecting patterns in text, images or sounds—for example, recognizing faces in photos. Although neural networks can operate far faster and at much greater scale than humans, designing effective neural networks can take even the most expert programmers up to a year or more.
An AI system from the U.S. Department of Energy’s Oak Ridge National Laboratory can dramatically speed up the process of developing a useful neural network. The AI, known as MENNDL (Multinode Evolutionary Neural Networks for Deep Learning), uses an evolutionary algorithm. This automatically combines and tests millions of neural networks to see which ones are good or bad at solving a given problem, essentially breeding high-performing neural networks optimized to perform a desired task.
“In a matter of hours instead of months or years, you have a full set of network designs for a particular application,” Robert Patton, head of Oak Ridge National Laboratory’s Learning Systems Group and leader of the MENNDL development team, said in a statement.
Since its inception in 2014, MENNDL has found use in applications ranging from identifying neutrino collisions to analyzing data from scanning transmission electron microscopes. In 2020, MENNDL was used on Oak Ridge National Laboratory’s Summit supercomputer to create neural networks that can detect cancer markers in biopsy images much faster than doctors are able to do, in a project with the Stony Brook Cancer Center at Stony Brook University in New York.
General Motors will use MENNDL to confront a major problem in ADAS research and development—how can cars quickly and accurately perceive their surroundings to navigate safely through them? The hope is that MENNDL can develop advanced neural networks that can instantly analyze on-board camera feeds and correctly label each object in the car’s field of view.
MENNDL is designed to run on a variety of systems equipped with graphics processing units (GPUs), from desktops to supercomputers. This is the first commercial license for MENNDL, as well as the first AI technology to be commercially licensed from Oak Ridge National Laboratory.