Artificial intelligence opportunities and challenges experts say

"Currently, many people regard artificial intelligence as strong artificial intelligence (AI) in science fiction, that is, the ability to reach the human brain. However, in the next decade or so, the biggest opportunity must be the so-called weak artificial intelligence, that is, not at all. Need to achieve all the functions of the human brain." In the "2018 Beijing Artificial Intelligence Industry Summit Forum and Beijing Frontier International Artificial Intelligence Research Institute inaugural meeting", Li Kaifu said.

At present, artificial intelligence has become the focus of global pursuit. How should China develop, opportunities and challenges?

Even weak AI has great changes.

It is very difficult for artificial intelligence to fully realize all the functions of the human brain. Li Kaifu said that the artificial intelligence mentioned today, in short, is to use a large amount of data to make more accurate judgments than people in a specific field. To create value, create efficiency, and liberate mankind without having to do repetitive work, but to do more creative and loving work.

"Today's artificial intelligence is just a tool, but only weak AI will bring great changes and opportunities." Kai-Fu Lee said that there are institutional conservative forecasts that China's GDP in 2030 is about 38 trillion US dollars, of which 7 trillion. The dollar is brought by AI.

In Kai-Fu Lee's view, artificial intelligence will permeate every industry in the world. Its development will probably experience four waves. The first wave is the Internet wave. The second wave is to train existing data to improve the artificial intelligence model. Traditional industry, in the past, talking about Internet+, the future must be AI+, because not every enterprise can use Internet+, but with AI+ many industries, 1+1 will be equal to or greater than 3 or 4. The third wave is to create new data, and the fourth wave is automated AI.

"These four waves are not happening one after another, but at the same time. The first and second waves activate the existing data and replace the blue-collar repetitive work. The third and fourth waves create new data. White-collar repetitive work is replaced," Kai-Fu Lee said.

China needs to develop AI patents and talents need to improve

Talking about the advantages of China's development of AI, Li Kaifu believes that the main performance is in three aspects: First, the academic level is improved, the proportion of high-quality papers is 42.8%; the second is the huge market and data volume; and then the support of national policies. However, it is undeniable that the gap between China and the US artificial intelligence industry is still large, mainly in terms of talents, patents, and capital.

“In this case, the amount of investment and the number of AI companies can be achieved through corporate efforts. Personal concern is that the number of patents is still far behind the United States, and behind the patents implies top research, all AI papers such as Nature. science "of these top papers, scientific experts, the United States is still living on. also, the number of China's AI engineers, scientists also inadequate in this regard should strive for." Kai-fu Lee said.

Based on this, Beijing Frontier International Artificial Intelligence Research Institute came into being. The institute is organized in accordance with the three-tiered open organizational structure of “1 Institute + N Innovation Centers + N Research Conditions Platform”. The first batch of three innovation centers were established, namely Beijing Artificial Intelligence Basic Research Innovation Center, Beijing Smart Social Innovation Center and Beijing Artificial Intelligence Patent Innovation Center, aiming to attract more international talents and cooperate with domestic and foreign universities to create first-class Open research and technology transformation center.

It is well known that to develop artificial intelligence, computing power and algorithms are key. "We all know that deep learning in artificial intelligence has high demands on computing and data. Take the example of "human-machine wars" in which AlphaGo participates. Model training often requires a large amount of GPU or TPU to participate." Chinese Academy of Sciences Computer Network Information Deputy Director of the Center Chi Xuebin said.

To this end, the Chinese Academy of Sciences has specially set up an artificial intelligence computing and data application service platform for the institute. At present, it has gathered tens of thousands of GPU cards, which will soon increase to tens of thousands. “The platform is the infrastructure, just like the highway, providing a convenient working platform, so that researchers and engineers can focus more on the debugging of the model, thus promoting the rapid transformation of scientific and technological achievements and applications.” Xue Bin said.