He once ordered Tianhe No. 1 to top the top 500 global supercomputing. Today, it also talks about AI and cause and effect.

"The biggest change in the era of big data is to abandon the thirst for causality and replace the concerns of the relevant relationship." Victor Mills Schaneberg had clearly wrote in the book "Big Data Age".

But should "cause and effect" really give up?

A fact that has to face is that the current big data processing is facing challenges such as insufficient storage capacity, data collection transmission storage processing, difficult data format, and difficult data value. These problems are not solved. The more data, the easier it is to fall into the "data swamp." At this time, clarifying the cause and effect has become the key to recognize the nature of data and carry out data governance.

On June 26, the 24th Annual Meeting of the China Association for Science and Technology opened. On the main report session, Yang Jun, an academician of the Chinese Academy of Sciences, said directly: "The cause and effect has opened a" scientific door "to strong artificial intelligence."

Academician Yang Xuejun reported at the annual meeting of the Chinese Association for Science and Technology

Talk about "cause and effect"

How do "related" and "cause and effect" understand?

In a simple example, there is a correlation with the rising rooster and the sun, but we can't say that the rise of the sun is the result of the rooster.

In the field of artificial intelligence, in the process of machine learning, it is often "feed" as correlation data, which often causes AI to have various problems under real conditions. For example, sweeping robots learned to sweep the floor, but it may not know why. Otherwise, there will be no news that#不 shit at home is covered with the whole family.

Weibo screenshot

The Turing Prize winner, Judia Pell, has triggered the cause and effect revolution of artificial intelligence, and he said straightforwardly that "the data is very stupid", "modern computers are completely different from our brain."

In Perl's view, artificial intelligence must understand causality in order to evolve into "strong AI". Academician Yang Xuejun obviously recognizes this statement. He believes that the current artificial intelligence is essentially based on big data, and the "data swamp" problem has become a stumbling block from weak to strong roads. Therefore, he proposed to develop data lake technology for causal relationships.

This is an intelligent data management system. Users can store large capacity, high -dimensional, crowdship, and multi -format data according to the original tissue without need to be structurally processed by the data. They have the characteristics of intelligent processing of data for the ladder of causality.

The "three steps" said by Academician Yang Xuejun are:

Reasoning, "What did I see"; intervention in reasoning, "What would happen if I action"; anti -factual reasoning, "What would I do at the time and what would I do."

"I call these three stairs as the law of‘ eyes and legs ’ -not only read thousands of books, but also have to do thousands of miles and think about everything." Yang Xuejun said.

In recent years, he and the team have been promoted by the team's development, architecture design innovation, key technical breakthroughs, and typical application verification steps. Related applications have been carried out. For example, the cause and effect data lake system in depth the immune effect and law of the new crown vaccine in the human body.

"The data center that can be deployed in the form of cloud service, or it can be deployed on the side platform in the form of software." He added.

Academician Yang Xuejun spoke at the 2020 World Computer Conference

"Do it autonomously"

"This person wants to keep it." This is the comment left on Yang Xuejun's doctoral dissertation on Yang Xuejun's doctoral dissertation. Three years later, when my country decided to set up the "Galaxy -III", 31 -year -old Yang Xuejun became the chief designer and made a good story of teachers and students.

The operation speed of the "Galaxy" series has soared -the goal of the I era is 100 million times per second, and the II era has reached one billion times. By the time "Galaxy -Ⅲ" has sprinted to tens of billions. This is basically in line with the rules of the development of supercomputers at a speed of 10 times.

However, when "Tianhe No. 1", Yang Xuejun didn't want to leave step by step. Taking the decision of the chief teacher of his chief designer, he directly sprinted by trillions of sprints every second.

"Tianhe No. 1" tens of trillions of super computer systems. Xinhua News Agency (Photo by He Shuyuan)

Their confidence is the co -calculation technology of CPU+GPU heterogeneous created by Yang Xuejun. However, in the first attempt in 2009, Yang Xuejun's team not only failed to achieve the result of 1+1> 2, but also regretted that the calculation efficiency was only about 20%. In order to solve this problem, they have been closed for more than half a year, and finally have achieved results.

In November 2009, "Tianhe One" won the world's fifth and Asia's first good results in the world's top 500 rankings. Of the top 10, except "Tianhe No. 1", the remaining nine computers were developed in the United States.

One year later, the second phase of the "Tianhe No. 1" system topped the top 500 global supercomputing, and for the first time, it left China's supercomputing name, making China the second country in the world to win this crown. China has begun to speak in the field of international super computing.

"High -performance computers are the infrastructure of the country, just like the role of highways in the national economy." Yang Xuejun emphasized: Super computing must be done well, and it must be done autonomously!

At the age of 48, he was elected academician of the Chinese Academy of Sciences; 56 years old and was awarded the rank of generals and became the youngest general in active duty at the time; 58 years old, the asteroid of the number 330640 was "Yang Xuejun Star" — no matter what time, to "autonomy autonomy at any time "The pursuit is firmly engraved in his footprint.

- END -