AI chips move toward edge computing
Views : 9
Update time : 2022-09-17 14:49:43
With all kinds of smart devices such as smart cars, smart watches, smart home appliances, VR/AR, etc., more and more into people's lives, artificial intelligence is rapidly moving from the "cloud" to the "edge". Under this trend, from the recently held "World Artificial Intelligence Conference 2022" can be found, more and more chip companies began to strengthen their layout in the field of edge intelligence.
At the beginning of this round of artificial intelligence boom, many enterprises believe that the most suitable place for artificial intelligence is in the cloud, which is also the main reason why a large number of enterprises will migrate data to the cloud, and the corresponding processing operations are carried by the cloud server. However, with the further development of digitalization, the amount of data collected by the edge and terminal is increasing exponentially, and there are higher requirements for real-time response and low latency, so the extension of artificial intelligence to edge computing becomes unstoppable.
Qualcomm President and CEO Ammon pointed out the importance of edge innovation in his keynote speech for WAIC2022. "AI processing power needs to be extended to the edge side, where context-rich data is generated, and distributed intelligence is born. This will significantly facilitate the large-scale deployment of richer AI applications, while improving cloud intelligence as a whole." Ammon also highlighted a number of key advantages of terminal AI. "Terminal AI improves security, protects privacy, and allows sensitive data to be stored on the terminal rather than sent to the cloud. It is able to detect malware and suspicious behavior, which is critical for large-scale shared environments. "End side AI will enable new disruptive technologies, and it will help make more efficient use of limited network resources and bandwidth." Lars Reger, Global Executive Vice President and CTO of NXP Semiconductors, also highlighted the need to rely on edge computing architectures when real-time response and low latency are critical factors in his talk entitled "Enabling Edge End Intelligence". Chen Tianshi, chairman of Cambrian, said that with the popularity of AI, the demand for AI computing power is growing exponentially. In the face of this trend, what Cambrian did was to provide product support for the intelligent development of all industries.
IDC predicts that 64% of data will be generated outside of traditional data centers in the next few years, which means that more intelligent processing of data will be done at the end and edge sides. With the access of massive data, the center of gravity of AI processing is moving to the edge. As the focus of artificial intelligence continues to migrate to the edge side, more and more chip manufacturers begin to strengthen the development layout of edge side products. At this conference, Qualcomm actively promoted the popularization and ecological cooperation of AI on the terminal side. Ziad Asghar, vice president of Product Management at Qualcomm, said Qualcomm is committed to continuously driving performance power ratio improvements by providing efficient hardware, algorithms, and software tools.
Aixin Yuanzhi, an artificial intelligence visual perception chip company, also showed a full range of end-side edge side AI chips and application solutions in smart city, smart transportation and other fields. In recent years, the introduction of edge AI technology has brought significant improvement to image and video quality processing. For example, in the dark light scene at night, the intelligent detection of obstacles such as people, cars, cyclists, road edges, railings and so on in the blind area of vehicles, and the early warning and alarm are carried out, showing the application prospect of edge AI.
In the process of the migration of the center of gravity of artificial intelligence to the edge side, autonomous driving has become an important direction for the development and application of edge AI. Chen Tianshi delivered a keynote speech titled "Vehicle-cloud collaboration, intelligent vehicle upgrade with core power" at the conference, which deeply introduced the Cambrian intelligent vehicle ecological development layout under the concept of vehicle-cloud collaboration. In the field of autonomous driving, computing power has become the core driver, according to Chen Tianshi. Under this trend, Cambrian Xingge, a subsidiary of Cambrian Holding Company, actively lays out the vehicle end product line and promotes the development of vehicle cloud coordination system.
Data closed loop and continuous optimization of AI model can be realized through vehicle-cloud cooperation. In the cloud, Cambrian has training chips and clustering products. At the end of the vehicle, Cambrian Hangge can provide on-board intelligent chips and computing platform, which can quickly send back the data of the vehicle through the cooperation of the vehicle cloud, and realize the rapid iteration and upgrade of the AI model.
In the field of vehicle-mounted intelligent driving AI chips, domestic companies such as Black Sesame and Horizon also performed well. Black Sesame's Huashan series vehicular intelligent driving chips can meet the computing power needs of L2+ level and above autonomous driving functions, and form industrial chain partners with FAW, NIO, SAIC, BYD, Dongfeng, etc. Horizon's Journey series AI chip has been installed in the Ideal One, Zhiji L7, Chang 'an UNI and other production vehicles.
How many chips does a car need?
Position and Function of Main Automotive Sensors
Chip: The increasingly intelligent electronic brain
LDA100 Optocoupler: Outstanding Performance, Wide Applications
