Alibaba Group today announced its first AI inference chip, a neural processing unit called the Hanguang 800, that says it can perform machine learning tasks much faster and more energy efficient. The chip, announced today at Alibaba Cloud's annual Apsara Computing conference in Hangzhou, is already being used for features on Alibaba's e-commerce sites, including product search and personalized referrals. It will be provided to customers of Alibaba Cloud later.
As an example of how the chip works, Alibaba stated that Taobao normally takes an hour to categorize the one billion product images that are uploaded to e-commerce platforms each day by merchants, and prepare them for search and personalized Recommendations. With Hanguang 800, Taobao completed the task in just five minutes.
Alibaba already uses Hanguang 800 in many of its business processes requiring machining. In addition to product search and recommendations, this also includes automatic translation on e-commerce websites as well as advertising and news services.
Although Alibaba has not announced when the chip will be available to its cloud customers, the chip may be able to help Chinese companies reduce their dependence on US technology, as the trade war complicates business partnerships between Chinese and American technology companies. It can also help Alibaba Cloud grow in markets outside of China. According to the Synergy Research Group, Alibaba Cloud is the market leader in China, but behind Amazon, Microsoft and Google in the Asia-Pacific region.
Hanguang 800 was developed by T-Head, the company that leads the development of chips for cloud and edge computing within the Alibaba DAMO Academy, the global research and development initiative that brings Alibaba more than $ 1
In a statement, Alibaba Group CTO and Alibaba Cloud Intelligence President Jeff Zhang said, "The introduction of Hanguang 800 is an important step in our quest for next generation technologies to improve the performance of the company Increase Computer Both our current and our emerging business while improving energy efficiency. "
He added," In the near future, we plan to strengthen our customers by providing access through o your cloud business for the advanced computing that enables the chip anytime, anywhere.
T-Head's other product launches included the XuanTie 910, an IoT processor based on RISC-V, the open-source hardware command set that began life as a project at UC Berkeley. XuanTie 910 is designed for IoT high-performance applications such as edge servers, networks, gateways, and autonomous vehicles.
The Alibaba DAMO Academy works with universities around the world, including U.C. Berkeley and Tel Aviv University. The program's researchers focus on machine learning, network security, visual computing and natural language processing, with the goal of serving two billion customers by 2035 and creating 100 million jobs.