In fact, artificial intelligence in China is still at a very early stage.
Yesterday, Netease released the "2016 Global Artificial Intelligence Development Report", showing that the United States, the United Kingdom and China accounted for 65.73% of the total number of artificial intelligence companies in the world, while China ranked second with 15745 patents in the number of global artificial intelligence patents.
From the Google Alpha Dog War Li Shishi to the excellent must-select robot spring evening group dance, the word "artificial intelligence" is gradually becoming familiar to the public. Various robot-named companies, different looks of drones, and posing in the mall give the children a shake. The VR product of the car is a utopia of Hollywood science fiction.
From the double-creation week activity started on the 12th of this month, the entrepreneurial culture of Shenzhen has been ignited again. Cook has also made a special trip. Your Shenzhen has also screened in the circle of friends: “Shenzhen once again attracts the eyes of the world†and “World The smartest minds are rushing to Shenzhen. It seems to tell the world that this is the center of the universe.
In the past two days, I took the opportunity to go to the main venue for two days. Unsurprisingly, the most exhibited products on the spot were robots, drones and AR/VR glasses. From the appearance, it seems that these products are very cool, in order to find out the level of industry development and find out a few products that can make people scream, I asked people about the performance and industrialization direction of each product.
The conclusion is that it seems that you are not as arrogant as you said. . .
What I saw was that the three-and-five-year-old children couldn’t put down the "Little Man" who sang "Little Apple" - it was much more fun than the toy excavator that couldn’t sing; also saw the hot-selling robot at the time, I don’t know how she In the small restaurants that are nowadays, I also saw that the patrol robot "Xiaoming" was surrounded by audiences and could not avoid obstacles. When I asked the technicians next to me, how did the patrol robot identify the thief and the owner, he smiled at me.
As we all know, the core of artificial intelligence is machine learning. Of course, the more mainstream now is deep learning, training the model through data, and using the model to issue predictions. This is the basic "educational" principle of the machine. With the help of massive data to train the neural network, so that a cold metal or plastic has wisdom, it is quite cool, then the problem comes, which involves three levels of problems.
First of all, what is big data? What big data has value in use? How to sort these big data? What do we need to stuff into the robot brain? This is the premise of artificial intelligence. If there is no large database supply of various dialects such as Sichuan dialect and Fran dialect, how can the speech recognition of Keda Xunfei be recognized? Only the robot "Xiaomeng" who can understand Mandarin can understand the owner of Henan dialect.
Second, how does the machine get instructions? This involves a cliché such as human-computer interaction. Commonly used are speech recognition, image recognition and text recognition. The machine sends calculation commands to the computing hub by recognizing speech, images or text. Some people will say that this is very simple, that is to say, let it understand! In fact, it is not so simple. In terms of Chinese alone, there are no more than one hundred dialects in China. There is no difference in the pronunciation standard of dialects. Everyone’s words are not the same. It’s hard to let him understand. Not ordinary difficulty!
Of course, in the end, I have to talk about the most core model building or network neural training. Neurosurgery is the process of getting the machine to get instructions, calculating the results, and outputting instructions. The high school math teacher teaches us that if A=1, then...; if A=2, then..., this model that derives different results based on variables is the most basic algorithm.
For a chestnut, I watched the double exhibition last night, and the apprentice Xiao He, who is working in soft work, asked me to eat. This time she was late again. I relived my past experience with Xiao He’s meal and looked at how much it was late for the number of meals I had with her. I use this to predict the likelihood of her being late. If this value is beyond a certain limit in my heart, then I will choose to wait for a while before starting. Suppose I have been with Xiaohe about 5 times. The number of times she is late is 1 time. Then she is 80% on time. 80% is the standard line for me to make a meal with me. If I am 5 times. The number of late arrivals was 4, which is 20% of her arrival on time. Since this value is lower than my red line, I chose to postpone the time. This model is the soul of the machine, and the algorithm is of course the key, so the search engine company has a natural algorithm advantage, whether it is the driverless on Tesla or the news push on the mobile phone interface.
Through various human-computer interaction recognition technologies, instructions are issued to the machine, and the machine calculates a result based on the existing network neural (computation model), and then issues an execution instruction to the machine hardware, which is the basic principle of the robot work.
After reviewing the basic principles of machine learning, look at the robots you exhibited, shouting "small flowers" - wake up the machine, sing a song - the recorded "Little Apple" began to sing, the patrol robot used The camera avoids obstacles, there is no large database support for the face image, and the captured data will not match. It becomes a mobile camera. How can the “small flower†patrol you in the three-step and one-step community? !
Nowadays, the endless stream of robots can be roughly divided into industrial robots, entertainment robots, companion robots and nanny robots. The industrial robot arm can repeat the simple mechanical work, and the robots that need to interact with people are difficult to achieve alone. Function, of course, is important. On the one hand, it is difficult for people to maintain a long-lasting freshness of something that is not completely intelligent. For example, only singing, dancing and simple dialogue robots can be performed. You may not be interested in her for three days.
Of course, the accumulation of artificial intelligence technology has reached a window, regardless of the current state of development, the prospects are bright. Privately, in terms of robots alone, it is still difficult to industrialize a large area under the current technology accumulation. The only hope is that when the combination of artificial artificial intelligence technology and intelligent hardware is not fully intelligent, we will The combination of the prior art and the existing hardware to improve the performance of various aspects of the existing hardware is currently the best choice for the marketization of artificial intelligence.
In driving, if the computer vision technology is combined with the car, the driver on the road may be alerted when the danger is detected in front of the car. Or voice recognition is implanted in the car control center. On a high-speed car, it is safer to navigate the car through voice. Of course, many car manufacturers have already tested the water in this area, and some have matured. It is also an example of the current industrialization of artificial intelligence.
The scene in "Star Wars" will eventually appear, but now that the "nanny robot", I can only say: Hehe.
Logic Comparators,Comparator Circuit Logic,Comparator Gate Logic,Comparator Circuit Digital Logic
Shenzhen Kaixuanye Technology Co., Ltd. , https://www.iconlinekxys.com