Uniquify, a Silicon Valley neural network technology and AI edge computing company, is introducing a transformative neural network fabric technology at CES 2022. Today’s neural network technology utilizes multiply-accumulate (MAC)-based operations to realize visual, audio, data, and natural language processing AI models. Uniquify’s Neural Network 2.0 technology shrinks the neuron in the neural network by replacing MAC operations with proprietary AI processing element (AIPE) technology. The area and cost reduction from AIPE technology enables AI edge computing anywhere and everywhere, including consumer markets, data centers and the cloud, enterprises, manufacturing, health and medicine, and government applications.
“Neural network technology is a new paradigm and complements the conventional software technology that has permeated all aspects of our lives for the last several decades. The expensive MAC hardware used to implement advanced but bulky neural network models often inhibits many edge computing application markets from realizing their full potential,” says Josh Lee, CEO of Uniquify. “Our Neural Network 2.0 technology shrinks the neuron in neural networks using AIPE technology to implement the most complex and advanced AI visual, audio, and natural language processing models using a fraction of real estate required by MAC-based neural networks. It’s a game changer in the AI edge computing arena.”
Recommended ITech News:Â Semrush Offers Free Traffic Tools Subscription to Amazon Alexa Internet Users
Most neural network operations–including CNN, RNN, FNN, AE, Batch Norm, and activation functions–consist of multiplication and addition operations running in inference (prediction) mode. These operations can be implemented in hardware as a MAC block. Uniquify’s AIPE block is 20 times smaller in size, power, and cost compared to the conventional MAC block and can be aggregated to achieve 70 tera operations per second (TOPS) per watt on a 7nm technology node. With such economic use of silicon real estate, Neural Network 2.0 technology has a wide variety of applications, including consumer appliances, artificial intelligence of things (AIoT), surveillance, vision inspection, diagnosis, and autonomous driving.
“We are excited to announce the rollout of our Bethel product family of neural network processing SoCs based on our Neural Network 2.0 technology,” Sam Kim, COO of Uniquify, comments. “Bethel Lite targets low power applications such as AIoT and consumer appliances. Bethel Pro is most suitable for visual and audio processing applications while Bethel Max is designed for high bandwidth and the best performance applications. These three products cover many vertical segments of the current neural network edge computing market. We are engaging with a handful of tier-1 customers to deploy our technology in the near future.”
Recommended ITech News:Â NTT DATA to Acquire Chainalytics, Strengthening Its Supply Chain Consulting and Analytics Capabilities
Uniquify is applying the Neural Network 2.0 technology to all major neural network visual, audio, data, and natural language processing models and committed to bring the most advanced AI technologies and models from cloud computing to edge computing so that consumers can benefit from these AI technologies and protect their privacy by keeping their data off cloud servers. Experts anticipate an explosion of size and complexity in today’s AI models and that makes Neural Network 2.0 technology the clear winner in addressing the virtually limitless market segments and industries that are ready to adopt neural network and AI technologies.
Uniquify’s core business includes neural network platform and modeling, AI edge computing, consumer application SoC development, and semiconductor IP licensing. After years of research and development, Uniquify is ready to unveil Neural Network 2.0 technology, Digital Content AI models, and high bandwidth and performance AI edge computing SoC through an invitation-only demo at 2022 CES.
Recommended ITech News:Â Carlyle to Acquire, Expand Data Center Company Involta
[To share your insights with us, please write to sghosh@martechseries.com]
