Artificial intelligence also needs to find a breakthrough from the chip

At the beginning of 2016, a “human-computer war” became the focus of much attention. Google AlphaGo won the world’s Go champion Li Shishi, which triggered fierce discussions. Despite the fact that this century's battles have cleared up, AlphaGo's brilliant performance has caused the artificial intelligence boom that has been triggered to increase. In this boom, many industries have begun to evolve toward artificial intelligence. The age of artificial intelligence has come quietly, and artificial intelligence will also greatly enhance and expand the boundaries of human capabilities, promote technological innovation, enhance national competitive advantages, and even promote humankind. Social development has far-reaching consequences.

Artificial Intelligence in China is Excessively Applied to Universal Applications

Looking at the development of the domestic artificial intelligence industry, the entire industry chain is mainly composed of three core links: "base technology", "artificial intelligence technology" and "artificial intelligence application." From these three aspects, we can see the development context of the domestic artificial intelligence industry and can analyze the application of artificial intelligence.

The basic technology of artificial intelligence mainly depends on cloud computing technology and big data technology. In this regard, the scale of the domestic market is huge. The industry has treated cloud computing technology and big data technology differently than the simple business model of early learning and imitating. Instead, it accepts it more and more pragmatically, constantly taps the enormous value it contains, and builds on the different nature of services. Artificial intelligence based platform.

These platforms provide basic background guarantee and implementation prerequisites for the implementation of artificial intelligence technology and the landing of artificial intelligence applications from the aspects of infrastructure, platform services, and software application services. For example, the deep learning integrated machine launched by Chuangda Data is an artificial intelligence platform based on big data, which helps researchers quickly penetrate into the core areas of artificial intelligence.

Artificial intelligence technology focuses on three aspects: pattern recognition, machine learning, and human-computer interaction. Pattern recognition emphasizes processing of non-intuitive data such as signals, images, speech, text, and fingerprints, such as speech recognition, face recognition, and so on. Machine learning covers most areas from general artificial intelligence applications to dedicated artificial intelligence applications such as computer vision, natural language processing, biometrics, DNA sequencing, and more. Human-computer interaction includes both the human-system speech interaction and the physical interaction between human and robot entities.

The domestic artificial intelligence technology mainly focuses on computer vision, speech recognition and language technology processing at the application level. Among them, representatives of the company include Science and Technology News, Baidu, Alibaba, Tencent, contempt of science and technology, Geling deep pupil and so on.

Artificial intelligence applications involve both dedicated applications and general-purpose applications. Among them, the application in the special field covers most of the current applications of artificial intelligence applications in the country, including face recognition, voice recognition, and intelligent robots in various fields. The general-purpose applications are focused on general-purpose solutions in areas such as smart homes, smart agriculture, and smart healthcare. At present, domestic artificial intelligence applications are in an excessive development stage from professional applications to general applications.

Starting with algorithms and chips to promote artificial intelligence application innovation

The accumulation of research in the field of artificial intelligence in China is not much different from that of developed countries. If we can formulate a comprehensive promotion plan for artificial intelligence at the national strategic level, it will provide China with an excellent opportunity to realize overtaking and improve the overall national strength and influence.

We should vigorously promote the development of artificial intelligence, seize the height of artificial intelligence application innovation, help the country's industrial transformation and upgrading, and strive to form a global competitive advantage.

This requires innovative artificial intelligence algorithms. As the core of artificial intelligence implementation, the algorithm is the largest competitive threshold in the global artificial intelligence industry in the future. However, China is still basically learning foreign algorithms, lacking independent innovation of algorithms. Although our country has achieved a breakthrough in engineering algorithms, the level of algorithms based on the cognitive level needs to be improved. This is also the core area of ​​future competition.

At present, the application of scenarios in the field of specialization is still the core of R&D and investment. The maturity of basic technologies also brings improvements in artificial intelligence technologies such as storage capacity and machine learning, but it is limited to the current computing power and large-scale CPUs and GPUs. In the parallel solution, the development of domestic artificial intelligence mainly focuses on computer vision, voice recognition, and smart life.

Therefore, through the innovation of algorithms, the evolution of technology, the accumulation and evolution of data, and the application of the supercomputer platform, the development trend of the artificial intelligence industry in the future should be from the application of scenarios in the field of specialization to the development of universal solutions in the fields of voice and vision. .

The focus of future competition in artificial intelligence will be in the field of machine learning, namely supervised learning, non-supervised learning and enhanced learning. By then, the competition of the algorithm will enter the white-hot stage. Only at the algorithmic level, countries or companies can achieve breakthroughs in the field of image recognition and computer vision and international technology.

In the next step, what our country has to do is to develop artificial intelligence chips. In the current large-scale discussion of artificial intelligence technology, one point is easily overlooked, that is, the chip that carries the artificial intelligence operation. The human brain is a complex network of hundreds of billions of neurons and trillions of synapses. There are still many orders of magnitude difference between the existing chips and this. Therefore, we need to seek breakthroughs from the chip, enhance the chip's parallel computing capabilities, analyze massive amounts of data at high speed, improve the chip's programming specificity, high performance, low power consumption, large-scale server deployment or resource constrained The potential of embedded applications reflects its potential; at the same time, it is a perfect combination of performance and power consumption.

Take Google's TPU chip as an example. Compared with FPGAs and GPUs on the market, its performance per watt is improved by several times. The chip has a higher tolerance for reduced accuracy. That is, it requires fewer transistors per operation and can therefore achieve higher calculations per second. In China, the "Cambrian" processor chip and the deep learning data to be released by the cloud computing giant released by the Institute of Computing Technology of the Chinese Academy of Sciences are exploring in this field.

Increase policy support and enhance corporate cooperation

The research area of ​​artificial intelligence is very extensive. Arbitrary research branches involve huge computing codes. Plus loopholes investigation and cross-domain crossover, it is difficult for any enterprise to make breakthroughs in a closed environment. With the popularization of application in the special field and the enhancement of the application requirements for general technology, China should continue to increase its openness and make system-level open source the norm. Foreign top companies such as Google, Microsoft, Facebook, and Yahoo who regard artificial intelligence as the core competitiveness of the future have all opened up their own artificial intelligence systems. Domestic HKUST News, Baidu, Alibaba, Tencent, etc. have also made their own artificial intelligence systems in different dimensions and spaces after taking the lead in artificial intelligence.

The company has opened its own artificial intelligence system to the outside world and does not represent the complete transfer of core technologies and algorithms. The opening of the low-level system can allow more companies or social groups to participate in the research and development of artificial intelligence related fields from different perspectives, and provide a standardized platform for the application innovation of China's artificial intelligence and rapid iteration of products to form a sound development environment. For open companies, they also ensure that they grow in sync with the latest cutting-edge technologies in the industry.

Our country wants to develop artificial intelligence industry, but also to strengthen cooperation between enterprises. The domestic artificial intelligence industry analyzes from the point of view of maturity of application development in specific fields and differentiated technology thresholds, and can be divided into three directions: basic infrastructure construction, general-purpose scene application, and dedicated application development.

In terms of underlying infrastructure construction, enterprises rely on their own data, algorithms, technologies, and platform advantages to provide basic resource support for companies in the industry. At the same time, they will also translate their advantages into dedicated and general-purpose applications to create their own Artificial intelligence industry chain ecological closed-loop. In terms of general-purpose scenarios, companies will focus on computer vision and voice recognition to provide universal solutions for security, life, finance, and manufacturing. In the area of ​​dedicated application R&D, most hardware manufacturers, startups, solution providers, application providers, etc. are concentrated. In general, the promotion of artificial intelligence application innovation is inseparable from the mutual cooperation of the core links in the industry chain. It is inseparable from the mutual cooperation among various companies in the direction of industrial R&D.

It is also indispensable for policy support to increase support for artificial intelligence.

The number of companies participating in the research of artificial intelligence technology in China is relatively small, and their own R&D investment capacity is relatively limited. It is urgent to require national-level policy support to ensure that China will have the R&D competition with Europe and the United States in the development period of the most critical technologies for artificial intelligence in the next 10 years. Strength and R&D progress have ensured that China has always been at the leading position in the world in the wave of scientific and technological progress. The specific recommendations are as follows:

First, the Ministry of Science and Technology took the lead to increase support for basic research on artificial intelligence. The National Development and Reform Commission and the Ministry of Finance provided project support and financial support. Encourage and support the establishment of long-term and deep-level cooperation mechanisms between national research institutes and artificial intelligence enterprises so as to mobilize and integrate the strengths of various R&D institutions in the field of artificial intelligence in China, and jointly support China to win a leading position in the fierce international competition in artificial intelligence technologies. .

Second, mobilize more forces to participate in the research and development and application of artificial intelligence, and introduce a series of subsidy policies to support them. For example, support will be given to platform manufacturers participating in the construction of artificial intelligence middle tier technologies, as well as the introduction and support of international talent and teams in the field of artificial intelligence. Domestic companies carry out research and development support for artificial intelligence technologies.

Finally, we will increase the research and development support for key common core technologies of artificial intelligence and encourage all industries to introduce artificial intelligence technologies to transform and upgrade traditional industries and build a benign environment for artificial intelligence R&D and application.