Huawei Connect 2018: Decentralised approach to powering AI applications
Huawei has observed that the centralised computing capability provided by data cetres, is unable to satisfy the needs of high-bandwidth and low-latency scenarios. One pertinent example of this is autonomous driving.
The Chinese company finds that hierarchical computing capability through terminal, edge and cloud, is the suitable way for artificial intelligence use cases to pervade more aspects of human society and economy.
To that end, they had launched the Atlas Intelligent computing platform last year, an integrated portfolio of products that includes key technologies like intelligent heterogenous computing and edge-cloud collaboration.
During Huawei Connect 2018, this Atlas portfolio was further complemented with the launch of the Atlas 200 AI acceleration module that is based on Huawei’s ASCEND AI processor and can support real-time analytics of sixteen channels of HD video, the Atlas 300 AI acceleration card that will also support requirements of high-density video surveillance, the Atlas Edge station that enables cloud-edge collaboration, and the Atlas 800 AI appliance that provides an optimised AI environment to improve performance as well as monitor performance.
The Atlas 800 AI appliance does integrate software solutions like Huawei cluster management and task scheduling. Beyond these and when it comes to applications that the Atlas portfolio of hardware can enable, President of Intelligent Computing Business, Qiu Long explained that Huawei would provide reference algorithms for applications like facial recognition, for customers to select to integrate.
(This journalist is a guest of Huawei’s to their 3rd annual conference in Shanghai).