MIT researchers and other collaborators developed photonic chip concept, optical computing systems to do AI calculations faster and more efficiently than traditional processors

Briefing

MIT researchers and other collaborators developed photonic chip concept, optical computing systems to do AI calculations faster and more efficiently than traditional processors

August 18, 2017

Briefing

  • Programmable Nanophotonic Processor – MIT researchers and other collaborators created new computing chip that uses light instead of electricity to perform complex AI algorithm calculations faster and more efficiently than conventional computing systems
  • Optical Approach – Directs multiple light beams through modifiable, interconnected waveguides programmed to perform specific computations, allowing them to interact with each other to process result of calculation
  • Initial Demonstration – Used processor to create simple neural network that detects four basic vowel sounds, already achieving 77% accuracy, compared to traditional systems with 90%
  • Potential Applications – Can be used in data centers, security systems, drones and self-driving cars once fully-functioning system becomes commercially available

Accelerator

Sector

Information Technology

Organization

Massachusetts Institute of Technology

Source

Original Publication Date

July 12, 2017

Leave a comment