Benhard E. Boser 1992
Early in the 20th century, in 1991, an article was released for peer review on IEEE Journal of Solid State Circuits. In the article, Bernhard E. Boser, described neural network capabilities in pattern classification, speech recognition and control.
The article and the formula has been part of a reference for several studies on neural network parallel processing.
Hadi Esmaeilzadeh, Adrian Sampson, Michael Ringenburg, Luis Ceze, Dan Grossman, and Doug Burger* University of Washington / Microsoft Research
In June 2012, a team from University of Washington posted a presentation on the challenges to implement neural networks. This is what their system looked like.
At IFA 2017, Huawei Consumer Business Group, unveiled Kirin 970 platform. Kirin 970 is powered by an 8-core CPU, a 12-core GPU and also featured a dedicated neural processing unit. It is built on a 10nm advanced process, packing 5.5 billion transistors in one cm². Huawei has opened the chipset to developers and partners who can find new and innovative uses for its processing capabilities.
“As we look to the future of smartphones, we’re at the threshold of an exciting new era,” Yu said. “Mobile AI = On-Device AI + Cloud AI. Huawei is committed to turning smart devices into intelligent devices by building end-to-end capabilities that support coordinated development of chips, devices, and the cloud. The ultimate goal is to provide a significantly better user experience. The Kirin 970 is the first in a series of new advances that will bring powerful AI features to our devices and take them beyond the competition.”
After years of development, Cloud AI has seen broad application, but user experience still has room for improvement in areas such as latency, stability, and privacy. The goal is for Cloud AI and On-Device AI to complement each other. On-Device AI offers strong sensing capabilities, that are the foundation of understanding and assisting people. Sensors produce a large amount of real-time, scenario-specific, and personalized data. Supported by strong chip processing capabilities, devices will become more cognitive of user needs, providing truly personalized and readily accessible services.
The Kirin 970 is powered by an 8-core CPU and a new generation 12-core GPU. Built using a 10nm advanced process, the chipset packs 5.5 billion transistors into an area of only one cm². Huawei’s new flagship Kirin 970 is Huawei’s first mobile AI computing platform featuring a dedicated Neural Processing Unit (NPU). Compared to a quad-core Cortex-A73 CPU cluster, the Kirin 970’s new heterogeneous computing architecture delivers up to 25x the performance with 50x greater efficiency. Simply put, the Kirin 970 can perform the same AI computing tasks faster and with far less power. In a benchmark image recognition test, the Kirin 970 processed 2,000 images per minute, which was faster than other chips on the market.
New developments in AI require joint effort across the entire value chain, involving tens of millions of developers, and the experience and feedback of hundreds of millions of users. Huawei is positioning the Kirin 970 as an open platform for mobile AI, opening up the chipset to developers and partners who can find new and innovative uses for its processing capabilities.