Deng: Wireless city with large data

From August 21, 2012 to August 22, China Mobile held "the sixth International Symposium on Mobile Internet" in Beijing International Conference Center, focusing on cloud, tube, end, and "Taiwan", to explore the construction of a new industrial ecosystem of cooperation and mutual benefit, and to create a new life of mobile interconnection. The following is a speech by Dr. Deng, Carnegie Mellon computer robotics specialist.

The following are lecture shorthand:

Deng: I was embarrassed because I had just been hit a few hours ago when I met an acquaintance on the way to the meeting, and he asked you what to do? I said I go to China Mobile's conference, he said what topic? I said wireless city with big data, he said you go again? How do I say this? He said you used to talk about cloud computing is a trick, and then talk about the Internet, fog is a flicker, that thing has not finished, you start to talk about Big data, will be a cheat? I said your question is very good, I make two guarantee, the first guarantee is that I today's speech guarantee is dry goods, second absolutely is to tell the truth, burst aniseed.

First answer a question, big data is not a bluff? The so-called talk is simple, straightforward point, that is, can you earn money? If you can earn money is not a lie, if everyone busy a white busy is a cheat. So I'm going to give you a typical big data scenario and see if the big numbers make money. This example is a Canadian company example, in 1999, the company called Goldcorp, is a mining company, until 99, the entire company's financial situation deteriorated sharply, we feel that the last resort. This time, they changed a new CEO, the CEO of Younger, more radical thinking, just came back from MIT class, this person is not a computer, but heard the Kaiyuan system such a fresh thing, he felt that since our engineers in the underground to dig out the gold, Is it possible for me to put all the geological data on the Internet and perhaps some of the gods of God can find it? This is the experience of learning Kaiyuan. And then he did it, he put all of his company's geological data on the area from 1948 to the Internet, and the natural good came, and soon received a variety of feedback, the company set up 110 exploration sites, more than 80 locations to find gold, so the share price returned. So many reporters came to interview him for his experience.

But you go to look at such famous case analysis and feedback, we seem to be very, say you open source, open the data, the situation happened. Is this actually the case? With a careful interview with their CEO, you will find the first thing, is it that I opened the data in the world to find 110 occurrences of people? Take a closer look at the CEO's interview. There's actually a lot of intermediate, the CEO said one thing, it is said that after the data was launched, this is the Japanese Mitsubishi or which company is based in Massachusetts Institute, those who study the CT signal, but also from the CT signal to produce human organs of the 3D map, medical imaging research, and geology is not related. But one of the researchers heard that there was a geological data, so they took the medical imaging data to the geological aspects of the modification, the result was a great success. As the CEO described it, he said that when the institute's people came to their company to use medical 3D technology to show the geological structure, all the executives present were almost jumping. But remember a problem, this is not the last occurrence, but because you have such a 3D geological model, it is easy to find the direction of the mine accident, so that he found 110 occurrences after the creation of a very good condition. So, this is the first aspect that, after opening the data, produces a lot of beneficial intermediate results, which contributed to the eventual success.

The second is that opening up the data, especially for the mining companies, which are very traditional companies, is inconceivable. A lot of people think this is the company's food, like Coca-Cola formula, is not it? Many interviewers have said that it doesn't matter, you can open it, so the so-called company secrets may be just the company's internal engineers do not want to let themselves embarrass excuses. Is this actually the case? No, because after that, they succeeded, and many voices said that you were opening up all the mineral geology data and maybe finding more gold, but they didn't, and he was open in desperation. So, this place has a big contradiction, if you open up the data, brainstorming, you can create value. However, you open the same time, a lot of confidential data also went out, how to solve the secret and you open the middle of contradictions? This is a big challenge.

What does the

say about this story? Big data is not a trick is whether to find value, if not found in the data value, it is a flicker; found value, that is to create a huge profit point. Does this matter have anything to do with our wireless city and our China Mobile? I think there is, this picture is actually our 3G general architecture, 3G in the middle of a lot of network links, each link will produce a lot of data, we are recorded where the data? In the log, in the database and so on, this information is not used? Some people say that there is, someone said no, for example, someone told me, after this data, you can see any place can see the city where there are more people, where there is less people. Suppose you know that A and B are very close, what do you want to do? They can sue you for privacy, so these big data, especially what the network operators accumulate in the data, what is valuable, and how to find this information is actually an art, an experience, not just a technical problem. So, as I said just now, I said I would never cheat and never lie, and one of them is that I am adamant in telling the truth.

Back to the question, if China Mobile opened all the blogs, what information in these large data is valuable? My answer is no, but maybe you can open up the data and brainstorm and let people around the world help you find that information. The next question is that you want to open a data, you have to do a platform, how do you do the platform? I know that. This is a hypothetical picture of us, but do not think that I am an armchair, I sell first. The right side of the picture is very simple, is a number of distributed cloud storage, said that we use a lot of cheap servers to China Mobile, Unicom and other large data into the inside, the storage is not enough? is not enough, because we have just said that the big data is not a bluff? The key question is, can you find the value and find out what it is worth? By the middle of the place, algorithm, data mining algorithm, so, you in the middle of this place, you have to have data mining algorithms plus cloud computing parallel computing. Why parallel Computing? Because the amount of large data is too large, you simply calculate, a few months are not down, time is part of the value, so you have to use parallel computing to speed up the processing of data. What's the front? It is an app Stroe. Those who are open to outsiders may not be the final answer, but they can provide a lot of valuable intermediary tools, in tandem, it is most likely to lead to the discovery of the final value information. So, among us, we emphasize the middle tool, which is the meaning of our app Stroe. Cloud computing platform first is your storage to be cheap, you buy a lot of garbage servers, strung together have a very beautiful name is called cloud storage.

What I said just now is the technical framework, in which there are actually a lot of people and things involved in this technical framework. First, the left is that we have a lot of data sources. We have just said that China Mobile has a blog, we have a public opinion analysis and so on, we call it the data source. When the data comes in, you can't find the value. That's rubbish. I have just said that it is art, that is experience, rely on a lot of people, the most important one is the professor, scholar. Those people looked at the copy all the way, and then he found a new algorithm, application developers to apply, to develop a very easy to use the application, which means that I have a functional things, how to let users like it? We need a product designer, you have the product, also can push to the market, but need money, who will help you do it? Investors come in and you have a product for whom? Three kinds of people, end consumers, enterprises, government. So if you want to contribute to the prosperity of this big data industry, it's really going to unite a lot of people to form an ecosystem of common prosperity, the most important of which is actually two. The first is to expand the source of data, the second is to find the value of large data as far as possible, seize these two, the big data industry can prosper.

This is actually an industry of an open platform called KDD, this is a Los Angeles branch, in all data applications in the middle, this place is the highest cited. Large data samples A lot of people are donating, and have already contacted China Mobile, China Unicom and several banks, as well as Chinese customs, Chinese government departments, they will provide some samples of data, what is called sample? The sample is said to be a part of the region, not the whole country, and is processed, to remove some real information and protect privacy. So, the data has, then, we just said that the data platform, it is for money, and now the money has. This picture is my Friday just took, in the West two flags Zhongguancun Software Park, right flagpole this place's building is the Beijing government designated to cloud base, in the cloud base to make a what thing? Call the Republic of Geekcafe, first of all, it's a coffee shop. The idea of communication is a collision. The second is called Geekshow, which is actually a showroom where people make a product model that can be sold. The third is Geeklab, you can sit down and work together. What the hell are we doing with this thing? We actually want to do an experiment, the first is to discover the value of large data, the second is to unite the gang, this is a very good engineer, these people represent the capacity of the third is an immature product prototype. With these three, we will be able to demand, research, investment, product, market several elements of the series. So, these people who are involved in this collection of republic, there are scholars, engineers, media, investors, you develop a thing, we sell to large companies, he developed products, we have a 10%,15%geeklab to feed.

(Responsible editor: The good of the Legacy)