Baidu unveils new cloud chip, still lags behind rivals like Google
China internet giant Baidu Inc unveiled on wenesday its first self-based artificial intelligence (AI) chipset, but it still lags behind rivals including Google, according to industry insiders. The search engine firm released its first cloud-to-edge AI chip called Kunlun which can be used in a number of scenarios, such as public clouds and autonomous vehicles, Robin Li Yanhong, founder of Baidu, said at the company’s developer forum in Beijing. “The chipset was born from a massive AI algorithm, which is also based on Baidu’s eight years of research on CPU [central processing unit], GPU [graphics processing unit] and FPGA [ field-programmable gate array] technology,” he said.
CPU, GPU and FPGA are used in high-performance computing.
“Over the past 40 years, Chinese companies have been heavily dependent on imports of high-performance chipsets, which has been a pain for us,” Li said, noting that Kunlun chip can support various scenarios such as voice recognition, search ranking, natural language processing and autonomous driving.
However, industry insiders were cautious about Baidu’s chipset. “It still lags behind Cambricon in terms of specifications,” Xiang Yang, an industry expert at Beijing-based CCID consulting, told the Global Times.
Compared to its power consumption capacity, Cambricon has better performance, he noted.
Chinese Academy of Sciences released the country’s first cloud AI chip, developed by Cambricon Technology, in May, according to the Xinhua News Agency.
The chipset will have accurate and fast big data processing ability, especially in image and voice search methods.
Baidu, considered China’s Google as it mainly focuses on search engine business, does not own core chipset technologies.
Although Baidu’s breakthrough on chipsets reflected progress by Chinese companies in this domain in recent years, the company had to cooperate with US tech firm Intel in chip design,” Yang Jing, CEO of AI Era, told the Global Times.
Compared to the tensor processing unit AI accelerator chipset developed by US tech giant Google, Baidu still faces a narrow application system, she noted.
Baidu also unveiled its autonomous bus, which has entered mass production. Apolong, a fully autonomous L4 mini bus developed by Baidu in collaboration with vehicle manufacturer King Long, has reached 100 units in volume in its production facility in Xiamen in East China’s Fujian Province.
The bus will be put into operation in cities and regions such as Beijing, Shenzhen, Xiongan New Area and also overseas markets like Tokyo.
“Baidu needs to focus on not only the performance of its chipset but also a build-up of the AI system, which means how many companies will use its chipsets,” Xiang said.
Still, it has advantages in its algorithm structure, he added. Based on Baidu’s open source platform for deep learning, it must connect the chipset with different algorithms to become more competitive.