Shulou(Shulou.com)06/03 Report--

Introduction: intelligent and warm ~

A few days ago, after Lei Feng.com reported the news of the establishment of an artificial intelligence institute at Wuhan University, some readers saw "Renmin University of China-Hillhouse Institute of artificial Intelligence" in the list of established artificial intelligence colleges in China. He expressed curiosity about what the National people's Congress did when it set up an artificial intelligence institute. In this article, Lei Feng explains in detail the "thinking" and "doing" of the "Hillhouse Institute of artificial Intelligence".

If there are a thousand readers, there are a thousand Hamlets.

A thousand universities will also have a thousand artificial intelligence colleges / research institutes with different characteristics.

Renmin University of China, together with Qing Dynasty and North China, is known as the three top universities, Tsinghua Institute of Technology, Peking University of Arts, Science and Medicine, and Social Sciences of the National people's Congress. People have always been under the impression that the National people's Congress is a giant of social science, but science and technology has little sense of existence.

However, at the beginning of this year (January 19), Renmin University announced the establishment of the Hillhouse Institute of artificial Intelligence with the following positioning:

"Hillhouse College of artificial Intelligence is a secondary college under the school, which is responsible for the planning and construction of artificial intelligence-related disciplines, and carries out personnel training and scientific research in this field.... give full play to the existing advantages of the school in artificial intelligence-related disciplines, build a world-class artificial intelligence discipline, and enhance the school's international influence and competitiveness."

Professor Zhang Guofu, associate dean of the Hillhouse Institute of artificial Intelligence, has said many times: "We want to do a different AI from the two schools in the north."

How is it different? The National people's Congress has its own considerations, and we can start with the change of the paradigm of scientific methods.

(the following part draws lessons from the speech delivered by Professor Wen Jirong, dean of the Hillhouse Institute of artificial Intelligence, on November 19th, which has not been confirmed by President Wen Jirong himself and represents only his personal opinion. )

Y = f (x): the change of scientific methods 1. Traditional methods

In the past hundreds of years, thinking about the core nature of their research methods, whether in natural or social sciences, can always be summed up as follows:

All the studies are just to find out the objective law hidden behind the appearance from the complicated world and the changeable samples, an immutable and stable law. Especially in the field of natural science, we always hope to express the objective law as a model or equation.

This method is essentially a scientism tradition or rationalism tradition, we hope to be able to draw such a model, function or equation from intuition or a small number of samples through induction, deduction and other methods. Once we have mastered such a model / function / equation, we can use it to explain all kinds of seemingly changeable phenomena and to analyze the stable things in the impermanent world.

In the history of human development, especially in the process of science, we have been looking for models that are as simple and beautiful as possible in various fields. The simpler the model is, the more beautiful it is, and the more universal it is, the better it will be.

This method has achieved great success in the field of natural science, and has become the core driving force in modern society. It can be said that every aspect of today's life is the result of science and technology under the guidance of this thought.

We have found some stable, immutable and objective laws in nature.

But we should also note that in the past many years, when we applied this scientific method to social sciences, we found a lot of difficulties. For example, when we use formulas to describe economic laws, predict the stock market, and so on, there are often unpredicted results.

This shows that this method is not applicable in social sciences.

The reason is that social science is a complex, nonlinear, (super) multivariable system, and it is often difficult to reveal the real invariant law behind such a system through small data / intuition.

More importantly, natural sciences can collect a large amount of data through a large number of experiments, while social sciences are difficult to obtain data through repeated experiments, so there is a scarcity of sample data. This also gives people the impression that "social science is not scientific".

2. Big data method / empirical method

The emergence of big data era provides us with an unprecedented opportunity. We suddenly have the opportunity to collect a lot of data, especially in situations where it was difficult to do experiments before. We find that as we collect more and more data, we don't even have to look for the hidden models behind the data.

Big data provides such a method, which can directly map from input to output, which is equivalent to a purely empirical method. We know that if we have enough experience, we can solve the problem directly with experience instead of looking for the model. This is the essence of using big data's solution.

This approach has been successful in many areas, but there is a problem with this approach, that is, there is often not enough data. In many cases, you will find that your data cannot cover all situations.

3. New paradigm: big data + intelligence

Artificial intelligence is a new method that emerged under the background of big data: although the problem is very complicated, I can still find the model behind the data and grasp the invariance and regularity of things. Its method is different from the traditional method, it is to seek "complex model" from "massive sample data".

In the past, the method of natural science was to find a simple model from a small amount of data and a small number of experimental samples, which can be used in natural science, but not in social science, because the problem is too complex and there are too many variables.

But now that we have the artificial intelligence method based on big data, we will find that we can find complex models from huge amounts of data.

A system may have tens of millions or billions of variables, and there may be very complex nonlinear problems behind it. It doesn't matter, we can still build such a model.

Deep learning is the representative of this method, which can learn complex models very efficiently from massive data. In fact, deep learning is not just about face recognition and autopilot, it is also effective for social sciences, and its application to social sciences will bring about a subversive revolution. Because it provides us with an effective new method to study complex problems.

This new scientific paradigm is: big data + artificial intelligence research paradigm.

Second, a preliminary study of the new paradigm

Deep neural network can be understood as a very complex function f. In computer vision, we use it to represent what happens when the modeler sees an object, and now we have achieved very good results in this area, which has been able to surpass the performance of human beings in specific areas.

So what's the difference in applying this method to the social sciences?

Professor Wen Jirong gave several examples:

1. + Economics

Wen Jirong said that they extracted the data of industrial transformation from 300000 pieces of news. Among them, x is the 300000 of the news data, and by establishing a model, we draw the industrial transfer result y = industrial transfer number as shown in the figure (time, province 1, province 2). After removing the low-frequency data, you get the picture on the right.

From this, we can easily see that China's industrial transfer is centered on the north, Shanghai and Guangzhou, while the transfer situation in the three places is different. Beijing turns to the periphery, Shanghai turns to the central and eastern regions, while Guangzhou still turns locally. In this way, the original data can be visualized and visualized, thus revealing the economic laws that were previously impossible to discover.

2. + Sociology

Wen Jirong continued to cite an achievement of cooperation with the Chinese Academy of Social Sciences to study ideological issues. They found a lot of interesting phenomena based on 170 million users and 2.7 billion Weibo data (including millions of big Vs). For example, they found from Weibo data that in fact China has not had serious populism in recent years, and that populism is diversified and that people are more concerned about domestic affairs and policies.

3. + Politics

The third example is Wen Jirong's work at Microsoft. In 2012, Wen Jirong used public online data to analyze the public opinion base of Obama and Romney, so as to predict the US presidential election, the results are quite accurate. Wen Jirong joked that the prediction of this model is no longer accurate, "because the US navy has grown up."

4. + History

Note: the disaster distribution map on the right comes from the network, here is only an example

History can also be done. We can digitize historical data and model them in various ways, such as how disasters (floods, droughts, locusts) occurred in successive dynasties, what kind of impact they have, and what they have to do with birth, GDP, war, plague and so on. We can turn these problems into problems of data analysis, which can be done by current technology, but it is still a long way from the real history of data, which involves how to build a huge technology platform.

5. + jurisprudence

In terms of law, Wen Jirong has made a lot of achievements. For example, in the picture above, they use tens of millions of legal text judgments to do an analysis system and type in the "triad" to get the distribution map on the right side of the picture above, in which the darker the color is, the more cases there are. From this we can see that Jiangxi is very special, but the three eastern provinces seem to be different from what people think.

The above is only the combination of some social disciplines and the new paradigm scientific methods, and is only preliminary. In fact, there are many people in the author of similar things, a lot of social science research is also benefiting from big data + artificial intelligence.

However, at present, there are still a series of problems in the process of combining this method with various disciplines. For example, there is often no direct data, or the data exists in unstructured text.

In addition, computer experts who know how to use this technology do not know the corresponding social sciences, while experts in other fields are often not good at the use of big data and artificial intelligence. What data is extracted for analysis? Analyze what problems? How to analyze it? Experts in the field should cooperate deeply with computer experts.

What the National people's Congress lacks most is experts in the field of social science.

This is the advantage of the National people's Congress, and it is also the advantage of the Hillhouse artificial Intelligence Institute of the National people's Congress.

III. Ten Frontier issues of Intelligent Society Governance

As mentioned earlier, if Hillhouse AI is to "give full play to its existing strengths in AI-related disciplines", it must work closely with other colleges.

On the 19th of this month, with the support of the China Association for Science and Technology, Renmin University of China hosted the first Intelligent Social Governance Forum.

Participants in this forum include China Association for Science and Technology-Research Center for Intelligent Social Governance of Renmin University of China, National Institute of Development and Strategy of Renmin University of China, Culture and Science Park of Renmin University of China, Committee of Renmin University of China, Hillhouse artificial Intelligence College, School of Economics, School of Law, School of Society and population, School of Journalism, School of Labor and personnel, Institute of the Future Rule of Law, Journalism and Social Development Research Center.

In fact, this meeting can be seen as a symbol of the formal "alliance" between the National people's Congress Hillhouse Institute of artificial Intelligence and various fraternal colleges.

Wen Jirong, dean of the Hillhouse Institute of artificial Intelligence, announced at the conference ten "frontier issues of intelligent social governance" as a target for their cooperation with brother colleges in the future. They are:

The first topic: the construction of big data platform for intelligent social governance (artificial intelligence college).

The second topic: the algorithm and mechanism design of intelligent social governance (artificial intelligence college).

The third topic is the legal regulation of intelligent social algorithms and data (law school).

The fourth topic: the legal liability of the intelligent social Internet platform (law school).

The fifth topic: economic regulation and competition policy in intelligent society (School of Economics).

The sixth topic: intelligent society digital economy and China's economic transformation (School of Economics).

The seventh topic: public rationality and public opinion governance in intelligent society (School of Journalism).

The eighth topic: the construction and standardization of public ethics in intelligent society (School of Journalism).

The ninth topic, intelligent social Internet and interpersonal relationship reshaping (School of Social and demographic Studies).

The tenth topic: intelligent social order and intelligent governance (College of Society and population).

From this list, we can see that the position of the National people's Congress in the development of artificial intelligence is completely different from that of the Qing Dynasty and the North. The National people's Congress puts more emphasis on the combination of artificial intelligence technology and various social sciences, and changes the traditional social sciences through new technologies and new research paradigms. In this process, the artificial intelligence institute is both the center and the edge.

Perhaps we can think that the National people's Congress regards AI technology as the "new mathematics" of social sciences.

Https://www.leiphone.com/news/201911/5enGTF6LCAB1oL4V.html

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.