Big data: Business Revolution and Scientific revolution

Large Data Business revolution

Large data is layered, including large data 1.0, 2.0 and 3.0, which are presented separately.

Large Data 1.0

If you use a key word for large data 1.0, that is analysis, more in-depth, better analysis technology. As a matter of fact, members of the Hequan have just said a lot of examples and I would add some more. More interesting is the big medical data, which is more about "quantifying ego", it is through a kind of non-intervention, put some so-called medical sensors to our side, such as we wear a wrist watch, a ring, an ear, a pair of glasses, and so on, through these devices we can understand their heartbeat, blood pressure, even the health of our body, makes early predictions of major illnesses such as epilepsy. Similarly, if we use a child mattress for a child, and through the pressure and humidity sensor on the mattress, we can determine early whether the child has serious snoring or improper sleeping position. Because severe snoring can cause children to develop less intellectually than their peers at the age of 3 or 4, it is now possible to detect and treat them early with a sensor analysis of the mattress. There are similar assistive tools for people with chronic illnesses and older persons.

On the in-depth analysis of large data, it is very important that personalized information recommendation. Personalized information recommendation is not only based on the user's similarity so simple things, there are a lot of more in-depth complex models. For example, in terms of how users look at information, how do we determine whether a user is going to go deep into the gossip to die, or to go to another news story? Similarly, some users visit Taobao is just a stroll, some users are very clear to want to buy something, which requires the user's intention to predict, which involves some of the more difficult machine learning technology.

Deep analysis of large data can play an important role in many areas. Netflix, for example, is an online VCD leasing company that analyzes the viewing and evaluation data of a large number of film dramas, finds the most appropriate directors, actors and content, and uses them to invest and photograph the most successful TV series "card Houses" in American history, This is actually the first film that Netflix has launched. At present, we can use the data of set-top box to intelligently sort different programs of the same TV station, so that these programs can better serve different audiences. We can also analyze the competitive network between TV programs, by fine-tuning the broadcast time and increasing the ratings. These are based on in-depth analysis of large data, the method used is much more difficult than simple support vector machine, regression analysis, decision tree and so on, in most cases there is no way to make these analytical methods of product.

Large Data 2.0

Large data 1 is only a start, it emphasizes only the large amount of data produced by its own business, and optimizes its business through more in-depth analysis. In fact, the essence of big data in business is not just in-depth analysis, but also how to creatively use your data to other businesses, or to use other external data for your business. This brings in Big Data 2.0, and its key word is "externality." Just now academician Hequan a lot of good examples, such as Indiana University How to predict the Dow Jones, how to predict the stock market through Weibo, Google's information, how Taobao's CPI index is used to assess China's non-just-needed consumption and so on. In these cases, the microblogging and Taobao data were not originally intended to predict the Dow Jones index, nor to calculate the Chinese consumer, it could be simply a way for users to vent their emotions on the web or to record sales on Taobao. But these data can play a lot of people could not imagine the "external" value, which is a typical feature of large data 2.0.

Here's a zest finance example. One of its slogans is "All data is Credit data", which makes it possible to quickly make decisions about how much to lend to you by analyzing a user's vast amount of content on the Internet, including deleting pages, recording purchases, social relations records, and so on. Not only is the decision cycle short, the cost is low, and the proportion of users delaying loans after lending is 35% lower than in the banking sector.

On the external application of large data, it is important to occupy the terminal. For example, each air-conditioner can be placed on the outside of a number of sensors, collect temperature, humidity, air particles and so on, such air-conditioning basically every dozens of meters can be set up, can provide more than the current air quality monitoring station coverage of more extensive information, can do nationwide environmental monitoring network- Such a network is rather difficult for the country to do. Another example, a terminal, such as a smart color TV terminal can do? As you can think of, it seems that you can do some personalized program recommendation, or do some personalized advertising push, but in fact far more than that. A tens of millions of intelligent terminals, the first can do better ratings survey, the second it can also affect radio and television, TV program ratings. For example, in the program recommended, Zhejiang TV in Hunan TV before the recommendation, so that can improve the ratings of Zhejiang TV. In the future it can even do terminal advertising, do content and so on. The owners of these future terminals will dramatically change people's lifestyles and impact on traditional industries.

Large Data 3.0

Big Data 3.0 is the last step in the big data age, and it's a vital step. Talking about the arrival of the big data age, never say that just a few people or a few companies can use large data to mark the arrival of the big data age, should be every scientific team, every start-up enterprise has the ability to benefit from large data applications. For them, although the volume of data is so large, they are used as conveniently as "small data" as usual. To do this, there is a need for better it architecture, better analytical tools to make it possible for the common team to make use of large data and, on the other, quantitative management measures and programs for the quality, value, rights, privacy, security, etc. of the data by the government and the industry. When the ecological environment of large data is formed, it is possible to further discuss the so-called large data platform and use these platforms to attract the best data and the best talent.

Here are four possible large data integration platforms.

The first is the ability to effectively protect privacy and information in the context of the establishment of the so-called data Taobao, that is, in line with the legal provisions of the situation, through this platform can be free to upload and download data, and achieve free pricing. In fact, research data has already begun to do so, but other areas have not yet started.

The second is whether to generate data carriers. Data carriers provide storage and computing capabilities, as well as necessary analysis tools and software, through advanced it architectures. Some smart people develop related products based on data supplied by data suppliers, and even develop better data products on the basis of data products. If the data product is sold as a download or an API, then the data carrier, the data provider, the data developer, and so on, can benefit from the fees paid by the customer for the use of the data product.

The third is the platform for data mining challenges, which we hope to focus on the most important challenging issues in data mining and large data analysis through such a platform. Such a platform can not only focus on the general problems in various industries, but also focus on most of the talent. Now Kaggle's most popular slogan is that it has the world's millions of data scientists and data engineers contact. The solutions put forward by these top-level talents are the tools on this platform, and these tools may further develop some universal tools in the future. So this platform will become a platform for "problems, talent and tools".

The last platform, is to the different vertical industry has the universality of problems and solutions to further abstract out, the establishment of a large vertical industry data research Center, and the Union and industrial funds together to establish a large data research institute, to form a "trinity, mutually complementary" large data industry ecology.

How to become a big data enterprise

We need to do a lot to be a big data enterprise, a big data individual or a big data government. Enterprises, for example, to become large data enterprises, first of all, all of its production and operation processes need to be data-oriented, using the enterprise social platform or deploying the sensor into the manufacturing process so that the data can be recorded; Secondly, the enterprise itself should have the ability of deep analysis of massive data, and formulate the reserve plan Of particular importance is the reserve plan for external data. We often say that the number of time to hate less, a large data enterprises, in addition to using their own data, but more importantly to understand which of their own business needs external data support, to store these data. Finally, stand in the height of the platform, the enterprise should be tolerant of the mentality of open some open data, participate in some social challenges and competitions.

Big data has brought a lot of changes to traditional research.

First, big data will bring about a shift in the paradigm of scientific research. There must have been data and theory before the subsequent interpretation, and the ultimate goal of scientific research is to predict and control. It's not the same now, with data that can be directly predicted and controlled.

The second big change, formerly in sociology, psychology and management, and other fields of experimental subjects are often dozens of people, up to hundreds of people, the research methods are often semi-quantitative or qualitative. Now, through the Internet can be targeted at tens of millions of or even hundreds of millions of people to experiment, and is not controlled experiments.

Here I would like to pass a few examples of how big data affects science from three levels, mainly including: First, what new perspectives are provided by large data. Second, the actual role of large data. Third, science is a double-edged sword, the scientific research of large data can hurt everyone here.

First, take a look at what the big Data provides, and start by talking about a Barabasi group last year's work on scientific reports. While doing evolutionary biology experiments, we believe that a creature always wants to spread its genes as far as possible, but only in the very lowest of creatures can it be verified-we do not believe that the driving force behind an unforgettable love is reproduction. To discuss the issue at the advanced biological level, the team analyzed 500 million text messages and 2 billion calls to analyze who was your first friend, and where the "first friend" was the person you texted or talked to most. The study found that a 20-year-old woman, her best friend is often a male, and a 20-Year-old boy's best friend is often a woman, mainly for breeding reasons, he (she) may be one of your lover. Interestingly, when a second friend is analyzed further, it is found that the second best friend is often a male, while the female's second best friend is often a female. But what's different is that when you reach the age of forty or fifty, the sex of a second friend of a man is often invisible, while a woman's second best friend is usually a male. The reason for this is this: when a man is forty or fifty years old, his first friend becomes his wife, and men's attention to the opposite sex drops faster than women's after marriage. For his wife, a woman at the age of forty or fifty, her first friend is often his children, so the second friend is easy to be male, is his husband. For men, his second best friend is often his child. This finding may be of little practical use, but it provides us with a new perspective: the use of communication technology to study the laws of evolutionary biology.

Sociology has a very important foundational theory: social capital is equivalent to economic capital. In other words, our relationships, the degree of intimacy and diversity we have with whom we relate to, are, to some extent, our economic capital. But this is only a theory, although it is a groundbreaking theory, has not been confirmed. In 2010, the American journal Science published an article to quantify the diversity of telephone calls in more than 30,000 districts in the UK by analyzing the telephone connections of 99% people in the UK, the social capital of the borough. Comparing the economic development indices of these districts, the researchers found that social capital and economic capital are strongly correlated. For the first time, this study validates the fundamental theory of sociology that has just been mentioned. But these studies are of little practical use because you can't raise the economic level of a city or a borough by calling more.

So what's the real effect of big data? We've done a research recently, and if you've seen a movie on the internet that you think sucks, for example, "Rich Fuchun", then you actually tend to competitively a lower score in the next grade, rather than a high score, which is different from what many people feel--I see bad movies, maybe next time I want to score high marks. This shows that people's behavior has a "anchoring effect." Similarly, you see a good film, you will tend to competitively high points next time. When this kind of "anchoring effect" is eliminated, the accuracy of personalized recommendation can be greatly improved. This shows that the use of a large number of data analysis can directly solve some practical problems.

However, science is a double-edged sword, the big data bring us the possibility of blessing, but also bring us possible danger.

Let me give you two examples. The first is an article published this year in Scientific reports, which analyses the data of the 200多万个 cell phone. Cell phone data resolution is relatively low, every one hours have a reading, tell you this mobile phone users belong to which base station. But in fact, in a 3-month period, if there's a random 4 times that we know you're at a base station for an hour, then I can almost be the only one in more than 2 million who can determine who you are. This privacy is a big violation, because we can more easily through scientific research and project cooperation in the way to get anonymous telecommunications data. All the guests here are very influential people, many people may want to know your mobile phone, want to know where you have been. They just want to see in the news what time, where, and what meetings you attended, and if it happens that you haven't shut down for the last one or two hours or so, it's easy to find out what your phone number is by analyzing it and know where you've been and who you've been in contact with.

A more typical example is the one published in the Proceedings of the American Academy of Sciences by Cambridge University and Microsoft study this year. The researchers analyzed more than 50,000 users who clicked on Facebook to "like" and "dislike" data that could be used to predict whether you were drinking or what the race was, and it could even predict whether you were gay or not. The underlying theory behind this case is that for a person, different types of data, including your shopping, speaking, social data, favorite and disliked data, movie-watching data, and so on, expose you from different sides. Based on each set of data, you can paint a portrait, each portrait and you are different, but there is a lot of repetition. In a sense, with these different data, you can be predicted and judged.

Large Data ethics

The big data is not just about the changes in business and science, but more importantly, it brings us a lot of changes in ideology that trigger us to rethink a lot of things. And big data, once combined with many other major industrial innovations, can have particularly dire consequences. For example, 3D printing, in addition to the print rendering technology and software design modeling, is important for large data because it requires 3D of scanned data. There are two kinds of trends that can be ethically focused on 3D printing: The first trend is to print yourself. With the development of technology, this 3D printer can find some printing materials by itself, using solar energy or other energy to melt the material to print itself. Such a 3D printer, its hardware and software are open source, in the future may be some people in these open source hardware, software to add some of their own code. The code is a response to the environment, some code may be kind, friendly, some code may be unfriendly, even offensive. The code for these 3D printers can mutate, learning, just like the genetic algorithm: many of the different environment variables in the response to the module through a certain protocol to interact, and some modules are successful, there may be some variation in the mechanism, it can reproduce more printers; It has fewer printers to copy. If there is a 3D printer with silica as the main raw material, we ask a question, is this some form of life? It can also reproduce, evolve, mutate, and it can move, communicate, and even attack humans or other organisms, so how does it develop in the future?

Another scary trend is to print life. Now we can print a single cell with an inkjet printer and the ear can be transplanted successfully. Before the kidneys can only survive outside the laboratory for one months, mainly its vascular system does not do well, and the recent problem of the vascular system printing has been overcome, so one organ transplant becomes possible. We know that it is impossible to build a building with dust; in the same way, it is absolutely impossible to build a life with elementary particles. But, like building buildings with bricks, it is possible to use cells to build a living body. In more than 10, 20 years, I believe that the lower life of the printer can be a reality. Is there any possibility of printing higher organisms in the future, or even printing humans? The print person is different from the clone, if Karl Marx's idea is correct, then the print person has the same memory with us, including our deep pain, the emotional experience, the happy joyful and so on. What kind of ethical impact will this print bring?

I believe that the big data, together with other related industries, brings not only the changes in business, science and the kind of intelligent city that we see, but also how we think about the essence of life and what is the essence of existence. I hope we can all be the first people to be well-prepared before a new era arrives.