When one sees the term artificial intelligence (AI), the idea that immediately comes to mind is machine learning and neural networks.
A young startup company, Syntiant, is trying to expand the power of AI to low-power power semiconductors for battery-powered devices. The company has developed what they term as Neural Decision Processors (NDPs) that utilize custom analog neural networks to optimize deep learning functions at the transistor level.
“Our position to marry both AI and semiconductor disciplines,” said Kurt Busch, CEO OF Syntiant, in a recent conference call with ECN.
According to Busch, traditional semiconductor solutions are unable to support advanced machine learning applications at the device level, due to the high power costs of moving and processing data. Syntiant expects to leverage the power of artificial intelligence to improve the performance of an emerging generation of analog semiconductor used in mobile phones, wearable devices, drones, and other devices, by optimizing data and memory to make efficient use of power. Applying analog neural networks and deep learning algorithms can help analog devices function more efficiently, Busch notes.
Syntiant is targeting its NDPs at always-on applications for battery powered devices. These include keyword spotting, speaker identification, wake word, event detection, image recognition, and sensor synthesis.
According to Busch, Syntiant is working with several companies and expects the company to design-ins by the end of 2018. The company has already announced a development agreement with Infineon Technologies to complement its NDP with the company’s high-performance microphone technology.
Headquartered in Irvine, CA, Syntiant has assembled management and engineering teams that combines expertise in analog semiconductors and AI software. The company has completed an initial round of funding led by Intel Capital.