A multi-institution research team has developed an optical chip that can train machine learning hardware.
An international team of researchers has designed and built a chip that runs computations directly in memory and can run a wide variety of AI applications–all at a fraction of the energy consumed by computing platforms for general-purpose AI computing. The NeuRRAM neuromorphic chip brings AI a step closer to running on a broad range of edge devices, disconnected from the cloud, where they can perform sophisticated cognitive tasks anywhere and anytime without relying on a network connection to a centralized server.
Some analysts say that the end of Moore’s Law is near, but Patrick Naulleau, the director of Berkeley Lab’s Center for X-Ray Optics (CXRO), says that it could be decades before the modern chip runs out of room for improvement, thanks to advances in materials and instrumentation enabled by the CXRO.
During the COVID-19 pandemic, Johns Hopkins Medicine Media Relations is focused on disseminating current, accurate and useful information to the public via the media. As part of that effort, we are distributing our “COVID-19 Tip Sheet: Story Ideas from Johns Hopkins” every other Tuesday.
Columbia engineers are the first to build a high-performance non-reciprocal device on a compact chip with a performance 25 times better than previous work. The new chip, which can handle several watts of power (enough for cellphone transmitters that put out a watt or so of power), was the leading performer in a DARPA SPAR program to miniaturize these devices and improve performance metrics.