Artificial intelligence accelerated penetration is not far away

Source：Updated：2017-10-13 10:07:17

As "a new generation of artificial intelligence development planning", the state council began to deployment of artificial intelligence in our country from the overall development plan, and about 2030 years ago, the overall plan of the new development of artificial intelligence in our country, strategic objectives and major tasks and safeguard measures for planning and deployment of the system.
 
From the perspective of national strategy, the combination of artificial intelligence and manufacturing will become the main means to promote intelligent manufacturing in the future. From the point of industry development, industry structure of artificial intelligence have begun to take shape, mainly including foundation support layer, technology driven and scenario application layer, the author will mainly from the three levels for arranging and analyzing ai industry chain.
Artificial intelligence accelerated penetration is not far away
Figure artificial intelligence industry chain three layers structure
 
Basic support: the international IT giant has long been entrenched, and Chinese start-ups are hard to enter
 
In the field of artificial intelligence, the traditional chip computing architectures have been unable to support the deep learning and so on large-scale parallel computing needs, which will require new the underlying hardware to better reserve data, speed up the calculation process. The basic layer is mainly hardware, which includes hardware, neural network chips, sensors and middleware for performance acceleration such as GPU/FPGA. These are the prerequisites for supporting artificial intelligence applications. These hardware are the computing power of the entire artificial intelligence, which is mainly dominated by international IT giants.
 
Currently in the GPU area, nvidia has a large scale network acceleration of industrial scale, and has launched a new processor based on Volta, the first speed to exceed 100TFlops. Intel mainly builds its industry around the FPGA, introducing artificial intelligence chips that mimic the human brain. Google also launched a second generation TPU chip to provide chip support for its own open source TensorFlow framework. In addition to these when it comes to the industry's giants, there are numerous in this area, start-up companies, such as transit the science and technology, the Cambrian and west well, but in the industry layout and research and development power also cannot match these giants.
 
Technical driving layer: algorithm and computing power are the main driving force, and open source is the trend
 
Technology layer is the core of the development of artificial intelligence, intelligent degree play a decisive role in the application layer products, in the process of this development, algorithm and computation power play a main role to the development of AI. Technology mainly rely on the operation platform of base layer and data layer resource modeling mass recognition training and machine learning, and development for different areas of application technology, consists of two phases of intelligence and cognitive intelligence. Among them, the intelligent perception stage by sensors, such as search engines and human-computer interaction realizes the person and the connection of information, receive the data modeling, such as speech recognition, image recognition, natural voice processing and biological recognition, etc.; The cognitive intelligence stage models the acquired data and USES the thinking function of the human brain such as deep learning to obtain the results, such as machine learning, predictive class API and artificial intelligence platform. On this basis, artificial intelligence can master the basic information input and processing ability of "see" and "listen", so as to evolve more application-oriented products at the user level.
 
At present, the artificial intelligence technology platform in the application of domestic level mainly focus on computer vision, speech recognition and language processing technology, domestic technology co., LTD. Has made a rapid development momentum is, one of the representative of the enterprise including hkust xunfei, spirit deep pupil, and access business group hua sheng electronic cloud (spirit), the horizon, SenseTime, yong hong science and technology, kuang depending on technology, sound cloud know, etc.
 
Scenario application layer: AI and scene depth fusion, domain application more extensive
 
The application layer is mainly based on the integration of the base layer and the technology layer to the traditional industry, realizing the application of different scenarios. With the technological breakthrough of artificial intelligence in the fields of speech, semantic and computer vision, the application of artificial intelligence will be accelerated to various industries.
 
The application layer can be divided into two parts: consumption level terminal application and industry scene.
 
The consumption level terminal includes intelligent robot, intelligent uav and intelligent hardware. The application of scene is mainly to the AI application scene of various external industries. In recent years, domestic enterprises launch the application layer of the products and services, such as small robots such as intelligence, wisdom teeth, customer service, customer service, I ask "go out", "the secret", such as virtual assistant, industrial robots and also emerge in endlessly, humanoid robot application layer products and services is gradually fall to the ground.
 
Among them, IBM first laid out artificial intelligence, "universal Watson" to promote multi-industry transformation; Baidu's "baidu brain" plans to focus on driverless cars; But Google's artificial intelligence business is more complex, and many fields are blossoming, including AlphaGo, driverless cars, intelligent surgical robots and so on. Microsoft is leading in areas such as language semantic recognition and computer vision. In addition, the home appliance industry has also set up a craze for artificial intelligence. Many home appliance enterprises have targeted AI technology to develop AI technology and apply it to household appliances. So far this year, changhong, midea, gree and galanz are all making the transition to intelligent manufacturing, trying to get a foothold in "Smart Home", combining artificial intelligence and intelligent family.
 
Conclusion: weak artificial intelligence has accelerated infiltration, strong ai is not far away
 
In the author's opinion, the artificial intelligence field has already ushered in the new industrial explosion period, but in general, artificial intelligence is still the weak artificial intelligence in the specific application field. The vast majority of systems, including deep learning neural networks and machine intelligence, are still not up to the point of being able to think. To continue to climb into the strong ai, the computer's cognitive ability is necessary, which is also the point that artificial intelligence is trying to break through. AI originates from technology, and finally, in the scene, the intelligence of the industry needs to be fully integrated with the industrial scene to realize the real landing.