Baidu, a notable Chinese search giant recently unveiled a new AI-based processor known as Kunlun during its yearly Baidu Create event. According to the company, the processor has the power to handle edge and datacenter workloads. What’s more, Kunlun will comprise both the 818-100 and 818-300 for inference.
This move by Baidu propels it to the selection of other Chinese companies that are designing hardware specifically for machine-learning.
During the announcement, Baidu revealed that it embarked on the development of a field-programmable gate array artificial intelligence (AI) accelerator back in 2011.
It also said that Kunlun is approximately 30 times faster compared the initial FPGA-based processor. Even so, the chips boast 512GBps memory bandwidth can handle 260 tera operations a second at 100 watts and are created with Samsung’s 14nm process.
Baidu’s recent announcement also revealed that Kunlun uses the company’s artificial intelligence (AI) ecosystem that comprises of AI scenarios including deep learning and search ranking frameworks such as PaddlePaddle.
The company also said that its years of experience in improving the performance of such AI frameworks and services gave it the necessary expertise required to build a world-class artificial intelligence (AI) chip.
Aside from supporting the normal open-source deep learning algorithms, the Kunlun chip can support a broad variety of AI applications. They include large-scale recommendations, autonomous driving, natural language processing (NLP), search ranking, and voice recognition.
China is seeking to boost its domestic production of processors amid the recent threat to the current existence of ZTE due to a seven-year ban imposed by the US on the importation of American components. Fortunately, Donald Trump ultimately withdrew the ban.
Pony Ma, Tencent’s founder and chief executive officer, said at the close of May that the ZTE issue showed that China lacked vital technologies including operating systems and processors. He went on to state that Tencent was fully devoted to supporting the growth of the chip industry in China.
Furthermore, the Chinese Government is looking forward to achieving its Made in China 2025 strategy that will see domestic chips meeting about 40 percent of the country’s requirement.
The AI capabilities of Baidu are growing at a fast rate. In fact, recently the company joined forces with King Long, a Chinese commercial vehicle maker, and SB Drive, a Japanese SoftBank subsidiary to develop an Apolong self-driving minibus to serve the Japanese market.
Yuki Saji, the chief executive officer of SB Drive, pointed out that public transportation in Japan is currently facing numerous challenges such as driver aging, driver shortage, and traffic network reduction. As such, self-driving technology is expected to be the solution to such issues.
Baidu is also known for creating EasyDL, deep-learning tool intended for individuals with minimal technical training. The tool is currently allowing physicians to DIY diagnostic aids including more efficient parasite screening.
Having customized hardware will enable Baidu to boost its AI potential at a reduced cost. Additionally, China’s has a weak semiconductor industry, which is also becoming a key focus for the nation amid the ZTE debacle. The recent trade issues with the United States appear to be stirring up the country’s interest in key technologies.
