Alibaba is a name that is synonymous with the e-commerce and technology industry, with Taobao and Lazada among the companies owned by the Chinese giant. As part of their ongoing initiative to “lead technology development”, they’ve announced a new AI inference chip that helps accelerate machine learning: the Hanguang 800.
At the Alibaba Cloud’s Apsara Computing Conference 2019 in Hangzhou, China, Alibaba explained that the NPU is currently being used internally for Alibaba’s business operations. The AI, according to them, helps with product search and automatic translation on e-commerce sites, as well as personalised recommendations, advertising, and intelligent customer services.
Given the level of computing power required to complete these tasks, the time taken to complete these tasks is often relatively lengthy. For example, one billion images are uploaded to Taobao by merchants on a typical day, which would take 1 hour to process. The Hanguang 800, on the other hand, only takes 5 minutes to complete the same task—categorise, tailor search, and personalise recommendations. This is further supported by performance tests results of a single-chip computing performance of 78,563 IPS at peak moment, while the computation efficiency was 500 IPS/W during the Resnet-50 Inference test.
“In the near future, we plan to empower our clients by providing access through our cloud business to the advanced computing that is made possible by the chip, anytime and anywhere.”
– Jeff Zhang, Alibaba Group CTO and President of Alibaba Cloud Intelligence
Interestingly, access to the Hanguang 800’s computing power is something that Alibaba wants to provide for their clients—via Alibaba Cloud. This is part of an effort to narrow gaps in various industries by providing access to technology, with an end-goal of making the world “more inclusive”.