This summit brought together players such as voice, chips and other areas of artificial intelligence, ARM China Vice President Jin Yongbin, Intel Vice President Li Wei, HKUST Executive CEO Hu Yu, Si Bei CEO Gao Sixing and other industrial experts and the Chinese Academy of Sciences. Institute of Computing Technology researcher Han Yinhe and other academic representatives. At the same time, He Guili, deputy director of China Institute of Information and Communications, also released the "Global Artificial Intelligence Industry Map."
First, the field of chips: leading industrial giants, academics have dedicated themselves
'Future algorithm will be as common as APP'. The first speaker to give a speech is Jin Yongbin, vice president of ARM China Market. He mentioned that ARM is building an open and collaborative innovation ecosystem in the artificial intelligence ecosystem. Last month, ARM launched the ultra-mini GPU: A brand new Mali Multimedia IP Suite for mobile phones, tablets, TVs, and other terminal chips. The new Mali Multimedia IP Suite includes the Mali-G52 GPU, Mali-G31 GPU, Mali-D51 Display Processor, Mali -V52 video processor, Mali-G52 GPU improves machine learning performance by 3.6 times, Mali-G31 is ARM's smallest processor, can bring artificial intelligence capability to low-end smart devices. At the same time, ARM also Nvidia creates a dedicated AI chip for IOT equipment and integrates it into Arm's Project Trillium platform for machine learning.
In addition to the chip giant ARM, this summit also invited Li Wei, vice president of Intel, another giant in the chip industry. He mentioned the outbreak of data, computing infrastructure growth and the current artificial intelligence in health, finance, education, and medical care. The growth of such application directions will become an advantage in the development of China's artificial intelligence industry. Intel has also accumulated its assets in AI, including hardware, software libraries, frameworks, tools to platforms. Intel has a general-purpose CPU in hardware, in terminals and The cloud is also doing chips. At the same time, he also said that the development of artificial intelligence algorithm framework is very important, according to Intel's experience this optimization makes CPU performance increased by 100 times in two years.
In recent years, China’s research institutes are also exploring the field of chips. The world’s first “neural network” processor chip “Cambrian” capable of 'deep learning' was published by the Institute of Computing Technology of the Chinese Academy of Sciences. Today’s Chinese Academy of Sciences’ computing technology Han Yinhe, director of the Institute's Control Computing Laboratory, also came to the summit. This time they are building machine-specific chips. Han said that the evolution of computing platforms has changed from PCs to smart phones. Robots will become new computing platforms in the future. However, existing chips in the robotics field are directly borrowed from computers or mobile phones, and do not take into account the characteristics of robots. The processor used for robots should draw on the human brain and use partitioned heterogeneous multi-core computing. Institute of Computing Technology, Chinese Academy of Sciences The development of the Dadu robotic chip platform is currently under way, using a time-division multiplexing structure algorithm to implement small-scale embedded designs, and using approximate algorithms for ultra-low-power design. Currently, the development of this processor will be divided into three tasks: Robot's neural network processor, motion control capability design, robot visual perception construction, most The formation of an open eco-robot platform hardware and software combination.
Second, the robot field: humanoid robot is the next generation human-computer interaction center
As a representative of the domestic intelligent robot community, Ubuntu has always been exploring humanoid robots and established an artificial intelligence laboratory. At the summit, CTO Xiong Youjun, an excellent candidate for CTO, elaborated the development trend of service robot technology from the technical point of view. First of all, he mentioned the key technologies in terms of hardware. The activities of the joints of the robot body require high-performance, high-precision servo motors. In the hands, joints, etc., where flexibility requirements are different, the requirements for servo motors are also different. In these core hardware, there is still a big gap between the domestic enterprises and the world advanced level. The intelligence of the robot is from the intelligent of the whole machine, which also includes the support of hardware such as sensors, motors, motors, and operation control.
In addition to the hardware capabilities, humanoid intelligent robot gaits also need corresponding technical support to achieve high energy efficiency. This includes hand-eye coordination for grasping and operating capabilities, gait walking techniques, 3D visual navigation techniques, and AI human behavior understanding and recognition. Technology, identification of body language, analysis and understanding of the environment and many other technological developments.
Regarding the development trend of future service machines, he believes that humanoid intelligent robots will be the development trend of service robots. Only when they are closer to humans in form, can they be better accepted by humans and better perform human-computer interaction. In the future, humanoid intelligent robots In motion control and artificial intelligence, it will approach people infinitely and even beyond humans. But Xiong Youjun also said that although robots are currently superior to humans in a single scene such as playing chess, robots are still in the early stages of development, both in terms of form and artificial intelligence capabilities. However, in the future, humanoid robots will become the next generation of consumer-level human-computer interaction centers, which will gradually win the trust of family members and eventually become family members. At the same time, new interaction methods will redefine the value of data, and the existing industry ecology will be overturned. Scene-based intelligent robot data ecology will be the future big data entry. In the next 5-10 years, intelligent service robots will usher in large-scale industrial applications.
Third, the "Global Artificial Intelligence Industry Map (V1.0)" was released
At the 2018 China Artificial Intelligence Industry Entrepreneurs Summit, He Guili, deputy director of the China Institute of Information and Communications, also released the “Global Artificial Intelligence Industry Map (V1.0).” He mentioned that current global research institutions and companies are conducting artificial intelligence. Research and innovation, artificial intelligence development is generally favored, especially in voice, vision and other technologies have gradually entered the practical and commercial phase, driving the industry scale growth.
In terms of computer infrastructure, innovations in GPUs, DSPs, FPGAs, ASICs, and brain-like artificial intelligence chips are very active. They not only realize the cloud, but also meet the needs of terminal AI computing. Chinese chip companies are also promoting Research and innovation of various types of AI chips, such as the Cambrian, Shenjian Technology.
In terms of software algorithms and platform construction, the global data traffic increases rapidly, with a compound annual growth rate of 24%. This provides data traffic for deep learning, and at the same time it has different public data in the field of machine vision, NLP, and speech recognition. Set is also providing resources for the creation of enterprises.
From the perspective of global development, China and the United States have become leaders in the global artificial intelligence industry. The United States has more strength in core technologies, but it is not comprehensive in laying out applications; China is headed by Beijing, Shanghai. 5. Guangdong and Jiangsu and Zhejiang provinces are gradually accelerating their development.
Finally, Mr. He pointed out that capital, talents and supervision are the guarantees for speeding up the construction of the artificial intelligence basic environment. Capital should increase the frontier research of artificial intelligence, key common technical problems, achievements transformation, base platform construction, innovative application demonstration support, At the same time, it optimizes the financing environment for artificial intelligence creation. In terms of human resources, it is necessary to cultivate or introduce top talents, to make breakthroughs in the research of original technologies, and to train engineers who meet industry requirements. In terms of supervision, laws related to artificial intelligence should be strengthened. Research on ethical and social issues, especially in smart driving, smart medical and other fields.
In addition, the summit also invited Synopsys chairman and co-CEO Dr. Aart de Geus, Yun Tian Li Fei co-founder and CEO Chen Ning, Dahua Technology vice president Zhang Xingming, Tencent smart platform product deputy general manager Chen Qian, Yun Cong Technology. Sun Qingkai, Partner and Senior Vice President, Sun Yuanhao, Founder of Singhuan Technology, Yao Song, CEO of Shenzhen S&T, Professor of Tsinghua University, Supervisor of Doctoral Candidates, Zhang Lin, Dean of Tsinghua Berkeley Shenzhen Institute, and other industry and academic representatives on the artificial intelligence industry Development has carried on the wonderful round table forum.