The hottest 2019 global AI application innovation

  • Detail

The 2019 global AI application Innovation Summit held yesterday

ai accelerated the transformation of traditional industries

in recent years, the tide of artificial intelligence has swept the world. AI is no longer the exclusive term of technology enterprises. It is more and more used in various industries. In the future AI world, there will be autonomous vehicle on the road. Factories will be intelligent and personalized production. Even hospitals will use AI to quickly diagnose the disease. The prospect is very good, but the reality is also bone like - the difficulty of idling and landing computing power still exists. Yesterday, the 2019 global AI application innovation summit was held in Wuzhou Hotel. This year is the second. 500 academicians, guests, experts, scholars and industry masters from the global artificial intelligence field gathered here. How to break through the limit of computing power of AI chips in the future and what are the development trends of AI? Let's listen to AI masters


general AI chips will be more intelligent in the future

now, we have seen that artificial intelligence is slowly infiltrating into ordinary social life. Face recognition technology facilitates cashless payment and protects street safety; In a word, you can adjust the brightness of the lights and the temperature of the air conditioner, and speech recognition opens up a different kind of home life; Autopilot disengages the steering wheel from manual operation

"algorithm, data and computing power are the three driving forces of AI." At yesterday's summit, Professor weishaojun, Dean of the Department of micro nano electronics of Tsinghua University, summarized in his speech. With the continuous maturity and development of 5g communication technology, IOT, big data and other technologies, AI is undergoing profound changes, but there are also many obstacles behind the development of AI

5g with low delay and high bandwidth will bring geometric progression data. The continuous evolution of the algorithm will enable AI to eat a large amount of data and spit out accurate decisions, which correspondingly increases the requirements for computing power and challenges the development of AI chips

weishaojun believes that traditional CPU, GPU and other chip architectures can not perfectly meet the requirements of artificial intelligence. To make face recognition more accurate and autonomous vehicle respond faster, it requires a special AI chip, but AI chips face many difficulties

"an AI algorithm often corresponds to a chip." Weishaojun said that the R & D of a chip has to go through a cycle of "definition - Design - Manufacturing - sales". When a new chip is launched oneortwo years later, the AI algorithm may evolve continuously, and the algorithm has changed since the chip was launched

with the deepening of AI application, the processing of massive data has brought about a sharp rise in energy consumption. Industrial and agricultural video monitoring needs 2 watts, and automatic driving needs 100 watts. Low power consumption has also become one of the requirements of future AI chips

in Wei Shao's assumption that there are also scientific research institutions and enterprises to contact Guoliang copper on cooperation matters, AI chip will be a common platform in the future and more intelligent. "The architecture of the chip needs dynamic variability to adapt to different algorithms; it needs programmability to adapt to the evolution of algorithms."

current situation:

Shenzhen AI chips are emerging

during the two sessions in Shenzhen this year, some CPPCC members proposed that Shenzhen could take a differentiated development path in the chip industry and find a new path in the fields of AI, sensors and so on. At yesterday's summit, I saw some breakthroughs made by Shenzhen in the field of AI chips

Dr. Niuxinyu, founder and CEO of Kunyun technology, shared Kunyun's new attempt on AI chip with the guests present. He introduced that the high-performance AI inference chip developed by Kunyun technology can enable face recognition that previously only supported 10 cameras to reach 100 in the future; An AI car that reacts 50 times a second reacts 500 times. This means safer automatic driving, more efficient face recognition and faster industrial automation production lines

in addition, the new chip architecture has versatility and high scalability under different algorithms, and has been applied in different industries. For example, UAV and robot patrol inspection in the power industry; Security and people flow analysis of smart cities; There are many fields such as sensor anomaly detection and analysis of C919 large aircraft in China


AI transformation in traditional industries accelerated

in addition to AI chips, another hot topic of the summit focused on innovation and industrial development in the AI field. The popularity of AI in recent years has made various industries expect to achieve transformation and upgrading through AI. AI technology, computing power and industry implementation are all making great strides forward

Liu Jun, general manager of artificial intelligence and high performance products Department of Inspur Group, shared his observations on AI trends with the participants. He believes that the current AI application scenarios are also concentrated in traditional industries, such as face recognition, automated customer service, security, fraud analysis, or vehicle recognition, intelligent street lights, and intelligent traffic police in the smart city category

with the further development of technology, after 2020, intelligent factories and intelligent robots will usher in an explosion period; After 2025, autonomous driving and intelligent medical diagnosis will also usher in large-scale applications

"if you want to be rich, build roads first." Liu Jun described that at present, China's AI industry is still in the "paving" stage. The higher the computing power means the faster the highway, and the faster the development of AI. He said that this year, China's AI turnover will reach 2.5 billion US dollars, two thirds of which are in the computing field, which includes AI chips, algorithms, applications, cloud computing, etc

Liu Jun believes that relying on the high computing power developed in AI chips and other fields, China's AI commercial landing process will be greatly accelerated in the future, and the AI transformation of traditional industries will also be accelerated. At present, the AI security screening machine in high-speed rail stations has increased the passenger traffic efficiency by five times, and the AI diagnosis error rate of tuberculosis is less than 1%

facing the huge prospect, there are also great challenges. In Liu Jun's view, the AI industry is still an underwater iceberg that has not surfaced, and will become a trillion level market in the future. More and more traditional industries place their hopes on AI to achieve transformation and upgrading, but the challenges remain

"many people said that they had completed a comprehensive investigation on the licensed enterprises of hot-rolled steel bars for construction before June 30, 2017, and found AI leading enterprises such as bat (China's three Internet giants Baidu, Alibaba and Tencent) in the industry, hoping to use cooperation to do industry AI." Liu Jun said that after several years of work, AI enterprises found that they could not continue because they could not meet the needs of thousands of industries, resulting in a gap in the transformation of AI in traditional industries


to bridge the gap, we need to build an AI ecosystem

in the face of the AI needs of all walks of life, even giants such as bat will inevitably be unable to meet their needs. Liu Jun believes that we need to build an AI industry ecosystem, and the industry AI is done by many ecological partners

the same gap also spans between the implementation of AI chips and AI technologies. In an interview with Guangzhou all media, Niu Xinyu said that in order to make artificial intelligence technology a commodity available to everyone and change our lifestyle, we need to reduce the cost of AI chips. However, the high computing power brought about by the poor toughness of thermosetting resin due to its cross-linked molecular structure will promote more users and scenarios to use artificial intelligence, and AI will eventually become "available to everyone"

we must face it squarely that compared with foreign countries, many core chip fields in China started relatively late at the time of distance cutting. Niu Xinyu believes that the establishment of an industry ecology in the field of AI chips is the greatest opportunity and challenge in the field of artificial intelligence - the larger the shipment of chips, the lower the production cost of chips. At the same time, there should be not only chips, but also a very wide range of applications based on chips. The combination of the two can form a good "ecological closed loop". This requires extensive cooperation between chip and application development partners, and gradually expand to a broader field

at the same time, it should be noted that chips have always been a globally competitive industry. Everyone must stand on the international stage and compete positively, and then occupy a place in various indicators of performance, application and power consumption, so that the chip performance level has an "international voice". At present, there is no unified standard in the artificial intelligence chip industry, and it is difficult for people to have an authoritative way to know what level their chips are in the world. The most direct way to know the performance of a chip is to evaluate it. Therefore, Niu Xinyu especially called for the establishment of "artificial intelligence chip evaluation standards" and "let the unified chip standards reduce the communication costs of chip manufacturers and artificial intelligence algorithm manufacturers, and establish an orderly competitive environment"

article/Guangzhou Daily all media cuiningning, Liu Chang

picture/Guangzhou Daily all media xuanhui

[: Zhong Shangling]

Copyright © 2011 JIN SHI