Huawei focuses on the development of artificial intelligence systems for
Hardware

Huawei focuses on the development of artificial intelligence systems for corporate customers

Chinese giant Huawei Technologies has been under severe sanctions from the US and its foreign policy allies for several years. The company is forced to pay more attention to the domestic market and prefers to rely on solutions for corporate customers based on the hardware base of its own design when developing artificial intelligence systems.

    Image source: Huawei Technologies

Image source: Huawei Technologies

At a conference for cloud solution developers, as reported by South China tomorrow postHuawei introduced the third version of the Pangu AI model, which uses the resources of Ascend processors of its own design and the MindSpore software environment. The solution is aimed at corporate customers, including companies specializing in the operation of railway transport or mining.

Meanwhile, according to representatives of the specialized department of Huawei Cloud, the computer infrastructure in China can no longer meet the demand for specialized resources. Many local companies are forced to pay big bucks for specialized NVIDIA accelerators and wait months to get them. In this situation, Huawei tries to offer an alternative based on processors of its own design. According to company officials, its Ascend processor-based cluster can be 10% faster than “mainstream GPU” when training models. Huawei employees refuse to disclose details of the comparison conditions.

As Huawei Cloud management points out, Pangu’s language model is not used to write poetry, but focuses on meeting the needs of the company. Huawei’s mission is to break into the business environment and provide businesses with valuable technology. So far, Huawei’s AI solutions have been applied to more than a thousand projects across industries, including eight coal mines.

About the author

Dylan Harris

Dylan Harris is fascinated by tests and reviews of computer hardware.

Add Comment

Click here to post a comment