Analysis: Artificial Intelligence, where is the outlet?

(News free download) machine learning, artificial intelligence into public view

In recent years, with the development of large amounts of data and algorithms of support, the machine can be reality scenes crawl and capture, and real-world scenarios by the algorithm of the data so that it can be machine readable, to the perception of reality . After obtaining the data of the reality of scene data, with a huge number of sample data to compare training to achieve recognition of the scene, called machine learning. Because of these two factors related to computing power, algorithms and large data support limitations in the experimental stage. AlphaGo win at chess and makes the machine learning increase the heat, and eventually into public view with LiShiShi.


In the Chinese Academy of Engineering (Shenzhen) Chinese University of Hong Kong president Xu Yangsheng opinion, perception, cognition and action, these three elements constitute three artificial intelligence. Examples of the computer vision point of view, to make a computer image recognition, feature extraction is first, the importance of the pixels in the difference extraction, this is perceived. Then important elements of labeling, marking to become a computer can recognize by the sign, so that the computer can understand the content of the picture, this is the perception. Finally, the computer-generated image of a paragraph to describe, this is the last step in the operation.

Beijing University of Aeronautics and Astronautics Wang Tian Miao believes that 50 years of history between the time after the development of artificial intelligence, researchers will focus on most aspects of the operation, because there is no ability to complete the first two steps. Generally speaking, human-computer interaction is the three factors of the action, is now generally can be achieved, the most common is the factory of the mechanical arm, by the programmer of the machine incorporated into fixed program code to achieve the mechanical arm repetitive movements. Why would repeat the action, rather than arbitrary step? The reason is that the machine is set artificially limit the scope and generate action, not in response to the machine according to their own understanding after made.

That is, before machine learning, artificial intelligence and robotics research stay in the main action, the lack of perception and cognition research, and the current robot learning only a perception and artificial intelligence in the early stages of development of cognitive levels and at this stage of the research it is still experimental.

Artificial intelligence is easy to form an oligopoly situation?

AI fire up after this year, regardless of the technology giant or start-up companies, want to seize the opportunity in the bonus period. Regardless of domestic and international technology giant through its own research and development or through acquisitions, in stepping up the layout of the field of artificial intelligence, to stay ahead in this area. So, from the current market, which the company has a chain of artificial intelligence on it?

The first is the computing and information storage chip giants like Intel, NVIDIA and other companies, they are the most upstream in this area, and the ability to provide computing solutions for midstream and downstream industry chain, they determine the depth of the development of artificial intelligence. Followed by the Big Data industry chain of raw data acquirer, including operators, BAT, Microsoft, Google and other Internet portal of holding companies, which have the necessary data resources, machine learning, artificial intelligence determines the breadth of development.

In addition, artificial intelligence technology research and development group, where autopilot, depth of learning, speech recognition and image recognition and other fields have achieved leading companies and their respective teams. Work directly led by the tech giants, such as Google autopilot, IBM Waston, Baidu autopilot and voice recognition; there are outstanding representatives of laboratories and start-up companies, such as DeepMind roots deep learning; in addition, there is inherently possess strength characteristics Tesla autopilot, iFLYTEK voice recognition.

However, the development of artificial intelligence requires constant investment and continuous data accumulation. In the field of artificial intelligence research has such a statement, AI requires a lot of data to support, and machine learning for feedback data will increase the quantity and quality of data acquisition, a huge high-quality data will further accelerate the efficiency and effectiveness of machine learning , forming a virtuous circle. Consequence of this development is to continue, this area started leading companies will be more leading companies at a disadvantage and will gradually be eliminated, and ultimately the formation of a handful of oligopoly.

At the same time, some companies are in the heart of the industrial chain of AI, by virtue of its technical and financial resources, through acquisitions and strategic stake, etc., to achieve monopoly control of the industrial chain. For example, Google acquired startup DeepMind AI in 2014, two years before we see its R & D AlphaGo defeat Li Shishi scene.

B-end market, the next outlet artificial intelligence will be medical and financial?

On the summit of this artificial intelligence and robotics, academia and industry movers who have expressed their views on the AI a vent.

Computer Science Department University of Oxford Michael Wooldridge believes that the next AI application should be in the medical field. At present, we can through smart wristbands and other intelligent hardware detects the user's heart rate, blood pressure, blood sugar, while counting steps. If we deal with these data to AI, we can achieve some health applications. This application is to let the doctor with you at any time, 24 hours a day monitoring. It knows how much you sleep, how much to eat, you know the level of blood sugar, you know the situation of the movement and so on, while I suggest you when to exercise, what time can not eat too much or drink too much wine.

Michael Wooldridge introduced in the UK national health system, including the cases of all British case data, medication records. If AI can be applied to the medical field, it will bring new discovery, artificial intelligence but also the next outlet should be that our entire medical profession in the medical. According to foreign media reports, in early June this year, DeepMind start planning their algorithm is applied to the healthcare industry, and plans to use machine learning process British National Health Service within five years.

Hong Kong University of Science Professor Yang Qiang believes that artificial intelligence can not do without data, we now have a judgment if the next outlet, it is necessary to see which industries have large data resources complete closed system. In the financial sector, a lot of people all throughout all business processes documented here with a digital approach. In the future prediction information and processing aspects, if a closed system in which, under the premise of large data, there are funds to promote the influence of the financial industry is the most likely to succeed in a field.

C terminal market, with voice as the starting point of the consumer revolution or will soon

From the development trend, if stuck in the artificial intelligence laboratory, the Institute of the stage, do not enter the commercial phase of the C-terminal face can not be in contact with the general public, but high above the black and Technology. Even as AlphaGo win as famous players, creating a sensation, just the talk of gossip.

But never let AI promote capital only laboratory product, AI want to commercialization, the first thing to consider is the technical difficulty. From the current point of view several areas of artificial intelligence, automatic driving many security risks can not be resolved in the near future, although the depth of learning made significant achievements, but still in the exploratory stage, robotic automation can not achieve automatic programming. But the field of voice control, a children's chat robot market driving outside, so intelligent voice in the mass consumer market applications fast.

Whether Microsoft, Google, IBM and Apple, or domestic Baidu, Sogou and IFLYTEK abroad in the field of speech recognition and semantic analysis has made some progress, and even subvert the industrial structure the way a number of areas. Moreover, the intelligent voice market penetration must be specific people, specific scenarios to reflect. In specific scenarios, there is the potential usage scenarios and customer service capabilities, assistant class scene.

In short, the development of artificial intelligence and robotics, and in response to an aging population, rising wage costs, industrial restructuring and other issues, the country's economic development has an important role. It should be noted here is that people do not have to worry about the robots will replace the labor force caused the unemployment problem, the application of new technology will bring the development of productivity, production efficiency, but also creates new ideas manner wisdom of life, intelligent life, etc. change, and stimulate the release of new demand and create new growth points. The arrival of artificial intelligence, is more on the structure of human society presents new challenges. (Source: ZDNet text / Sherwood compile: China Electronic Commerce Research Center)

Internet Research Papers