Lightmatter, a startup using light to accelerate artificial intelligence, wins the 2017 MIT $100K Entrepreneurship Competition
Lightmatter is a Massachusetts based startup created at the MIT Quantum Photonics Laboratory. The startup’s technology uses light to calculate faster and more efficiently, than existing artificial intelligence hardware.
“The team’s technology replaces electricity in processors with light, mimicking the computational operation of the brain to process larger and more complex algorithms with less power. ” MIT Management Sloan School said.
“Today’s computers do operations one by one,” says Yichen Shen, who earned his PhD in physics from MIT in 2016 and remains at the instutite as a post-doctoral associate. “We do operations all together, when light is passing through, it’s all simultaneous. We change the architecture completely. We use light instead of electricity, so that it’s even faster and saves power.”
“There’s a lot of data crunching that has to be done. In the brain, all this information happens in an analog fashion. The thing that makes the brain such a powerful processer is the connectivity of neurons,” said Dirk Englund, assistant professor at the Quantum Photonics Laboratory. “Every neuron in your brain is connected to in the order of 7,000 other neurons. In an electrical circuit, you have lots of wires crossing over one another. Light flies through to make these highly connected networks, creating a huge advantage.”
Lightmatter took home the $100,000 Robert P. Goldberg grand prize at the event Wednesday night. They had to beat out seven other teams. A startup called “Shrink Wrap For Crap” took home the audience choice prize.
Lightmatter believes that this technology will create a new and faster kind of computing hardware.
“While big datasets and the steady advancement of electronic computing hardware have enabled the artificial intelligence revolution, progress in electronics is slowing as the transistor—the workhorse of modern computers—approaches its fundamental size limit,” says Nicholas Harris, co-inventor of the startup’s optical neural network technology. “At Lightmatter, we’re using light to accelerate artificial intelligence. By computing with light, rather than electrical signals, Lightmatter’s systems are able to calculate faster and more efficiently than existing artificial intelligence hardware.”
The startup and technology were created by; Nick Harris: An EECS PhD candidate and National Science Foundation Fellow studying optical quantum computing and artificial intelligence at MIT; Yichen Shen: A postdoctoral associate at MIT, he joined the MIT Physics department to study nanophotonics and artificial intelligence. He earned his Ph.D. from MIT in 2016 and is co-inventor of Lightmatter’s optical neural network technology; Darius Bunandar: Developed software that visualizes night skies that have been distorted by black holes. The blockbuster movie Interstellar recently used this software to create scientifically accurate images. After enrolling in the Physics PhD program at MIT, he joined the quantum photonics lab to investigate artificial intelligence in the context of quantum computation and communication, and Thomas Graham: Currently pursuing his MBA at MIT Sloan School of Management. He previously held business strategy and business development roles at Google and Google X.