The Chinese technology company launched the latest version of Baidu’s open AI platform, Baidu Brain 7.0. They also announced that Kunlun II AI chips are now in mass production. Baidu annouced these news at Baidu’s annual World 2021 technology conference, which was live-streamed from Beijing.
According to the company, Baidu Brain 7.0 is one of the world’s largest AI open platforms and is used by Baidu to support industrial applications of AI and Baidu Cloud. The latest version integrates a broader range of information sources and deep learning, including language comprehension and reasoning. It uses various technologies to provide output in various formats, including language, audio, and visual.
Baidu Brain first launched as a platform in 2016. According to the firm, the previous version, Baidu Brain 6.0, has built over 270 core artificial intelligence capabilities and created over 310,000 models for developers, making it a key driver of intelligent transformation in various industries.
According to a statement from Baidu, customers will be able to use the latest chips in combination with Baidu Brain 7.0 software to allow a new generation of AI applications.
The Kunlun II chips, independently developed by Baidu, promise two to three times more processing power than the first generation of Kunlun chips. The Kunlun II has a 7nm technology and is equipped with Baidu’s second-generation XPU neural processing architecture for edge, cloud, and AI, according to the company.
The Kunlun II processors, according to Baidu, offer up to 256 TeraOPS (INT8) and 128 TFLOPS (FP16) with a maximum power consumption of 120 watts. Arm CPUs and high-speed interconnects, security and virtualization, as well as an updated compiler engine and developer kit, are all included in the chips. The Kunlun II chips have yet to get any new specifications or performance data from the firm.
The first-generation Kunlun chips were released in 2018, and commercial production started in late 2019. According to Baidu, more than 20,000 first-generation Kunlun chips have been produced and are being used for search engines, smart assistants, and cloud business purposes.
The original first-gen Kunlun chips used a 14-nanometer process technology and included 512 Gbps (gigabytes per second) of memory bandwidth. They provided a performance of up to 260 TOPS at 150 watts. According to Baidu, they allowed Ernie pre-training models for natural language processing to infer three times faster than the conventional GPU/FPGA-accelerating model.
The new Kunlun II chips, according to Baidu, could be utilized in a variety of cloud, terminal, and edge computing scenarios for a number of applications, including high-performance computer clusters, biocomputing, intelligent transportation, and autonomous driving. The Kunlun II chips enable deep learning frameworks such as Baidu’s open-source deep learning platform, PaddlePaddle. They are geared for AI technologies such as voice, natural language processing, and pictures.
According to Reuters, Baidu’s Kunlun chip division got a new round of funding in March for around $2 billion. Baidu spun off the Kunlun chipmaking section as its own company in June.