byline chip

Chinese AI Players Face Blacklist Roadblocks

In widening its punishment of China in the ongoing US-China trade war, the US has extended its blacklist to directly handicap its biggest competition in a much higher stakes race, for world domination in AI.

Recently, almost ten more AI companies – including providers of video surveillance,  facial and speech recognition and data recovery; were added to US trade black list. The reasons cited were related to the  violation of human rights by the supposed usage of AI technology in China’s repression the Muslim ethnic minority groups of the Uygur region.

Here is an interesting digression; that one of the most notable Chinese AI companies in this most recent US blacklist is SenseTime Group (known for its facial recognition AI tech),  whose founder Tang Xiao’ou was appointed as the foreign national to Malaysia’s sovereign wealth fund, Khazanah Nasional. SenseTime is the top AI ‘unicorns’ startup  from China with a valuation of over USD7 billion.

The most significant impact of the  US blacklist is that it makes it harder for is Chinese AI players to source AI chips and computing hardware to run their offerings on. These hardware include GPUs for ‘deep learning’ training servers, sensors and FPGAs for signal processing, bioinformatics, device controllers, etc.

As it is, a lot of superior AI hardware  and patents are  from US suppliers such as Nvidia, Intel, AMD, Western Digital,  Seagate, Ambarella and more.

Nonetheless, this supply cut of AI hardware has been anticipated by the far-sighted Chinese as early as  2017. Leading AI players had already started to develop the necessary hardware and recent announcements show the results.

In Shanghai two months ago, Huawei launched Ascend as the world’s fastest AI (server) cluster (claiming to beat Nvidia’s Tesla v100).

The company also announced shared cloud-based AI training hubs with state government to provide 3rd party hardware support for ecosystem partners to development AI applications.

Earlier last year Alibaba’s subsidiary T-Head produce its own AI inference chips; and also the announcement of the Xuantie 910 chipset for autonomous driving using 5G networks, AI and autonomous driving.

In July, search engine giant Baidu launched its Honghu AI chip voice control in intelligent car (partnership with Geely automobile), home and city scenarios.

In fact, many Chinese AI champions already have made some sort of strategic investments in AI hardware – whether independently, with other local partners, and even some still in patent collaboration with US-based tech suppliers.

For now, the Chinese need to play catch up for their AI software to be effective, by quickly growing sufficient supply of powerful AI hardware that are based on neural networks for the necessary deep learning training and also more data ‘deep lakes’ – (such as that provided by Japanese company Hitachi Vantara’s LUMADA suite of data services are crucial for), for data modelling and scenario inferencing.

IT Bytes Back says: In the longer term, there is no doubt that this US blacklist can deter some AI competition; but short term wise, China definitely faces a setback unless its AI players can significantly ramp up its AI hardware supply i. order to successfully deploy all the ambitious and innovative intelligent solutions that it is building for the world.