Neural chip could aid voice recognition, man-machine learning
- By George Leopold
- Jun 09, 2016
A neural chip technology startup focused on Siri-like human-machine interactions emerged from stealth mode this week to unveil "military-grade voice recognition and authentication technology" along with a proprietary neural processor approach.
San Diego-based KnuEdge was founded a decade ago by former NASA Administrator Dan Goldin and has so far raised about $100 million in private funding while developing its neural computing platform over the last five years. The startup is positioning its neural chip and voice authentication technology as advancing the state of the art for human-machine interactions and voice biometrics.
As part of its unveiling, KnuEdge also announced a voice recognition and authentication platform that the startup claims addresses security and background noise issues. It also rolled out neural processor technology based on an architecture it claims is "entirely different" from graphics processors, CPUs or FPGAs, or field-programmable gate arrays.
KnuEdge said the processor scales up to 512,000 devices and is based on "neurobiological principles" that promises to "reset" chip and system-level computing in datacenters and emerging applications such as those that make up the Internet of Things. KnuEdge's mission, Goldin declared in a statement, was nothing less than to create "technologies that will in essence alter how humans interact with machines, and enable next-generation computing capabilities ranging from signal processing to machine learning."
While emphasizing that its neural processor and voice recognition technologies would meet military specifications, KnuEdge also said it is offering the technology to businesses, claiming "significant interest" from Fortune 500 companies in the banking and health care sectors.
Before heading NASA, Goldin spent 25 years at the aerospace giant TRW Corp., heading its satellite TV development efforts. He launched KnuEdge in 2005, organizing a group of private investors the company said were willing to provide patient capital in lieu of incremental technology development. That, the company added, gave Goldin's team the opportunity to focus on longer development times to advance the state of the art for voice recognition and neural processing.
Goldin also said he stayed away from building large development teams and the resulting "high cash burn rates."