Shulou(Shulou.com)06/02 Report--

Big Data Digest

Authors: Wei Zimin, Jiang Baoshang

In 2017, the annual report of AI Index, led by Stanford University and composed of Wu Enda, Kai-fu Lee and other experts, was released for the first time. It reviews the global development of artificial intelligence throughout the year from academic, industry development, government strategies, etc., and is called the strongest report on artificial intelligence throughout the year.

On December 12, 2018, Harvard University, Massachusetts Institute of Technology, Stanford University and OpenAI jointly released the second annual AI Index report.

The pace of progress in the field of artificial intelligence is measured not only by the actual production of products and research results, but also by the predictions and concerns of economists and policymakers. The goal of this report is to use hard data to measure developments in the field of artificial intelligence.

The first report, published in December 2017, highlighted the unprecedented pace of investment and job growth in AI, particularly in gaming and computer vision.

This 18-year report analyzes developments in the field of artificial intelligence with finer data and a broader perspective.

The report has repeatedly mentioned the development of artificial intelligence in China and Tsinghua University:

The United States accounts for only 17% of the world's published papers, Europe is the most productive country, with 28% of the papers published in 18 years, followed by China at 25%.

The registration rate of artificial intelligence and machine learning-related courses in universities has increased significantly all over the world, among which Tsinghua University is the most notable. The registration rate of related courses in 2017 is 16 times higher than that in 2010 and nearly 3 times higher than that in 2016.

Different countries attach different importance to the application direction of artificial intelligence. China attaches great importance to the application of agricultural science, engineering and technology. Compared with 2000, China has paid more attention to agriculture in 2017.

Wu Enda also recommended the report heavily on Twitter today, saying that "there is too much data," and highlighted two highlights of the report: artificial intelligence is developing rapidly in both industry and academia; and the development of artificial intelligence still needs to be more diverse and inclusive.

Let's take a look at the essence of the report!

Globalization of artificial intelligence research

Without a global perspective, there is no AI story. The 2017 report is heavily biased towards North American activities, and the 18-year report narrows the global gap, using a "global perspective" to analyze research and investment in artificial intelligence.

This year's AI Index points out that AI commercial implementation and research results are exploding around the world, and this trend is more pronounced in Europe and Asia. China, Japan and South Korea are ahead of other eastern countries in publishing AI papers and books. In fact, as shown in the chart above, the United States accounts for only 17% of the world's published papers, Europe is the most prolific country, with 28% of the world's papers published in 18 years, followed by China at 25%.

In machine learning, the majority of papers published are on "man-machine wars" in games.

Artificial Intelligence Course Registration Volume Tsinghua Attention

Tsinghua University, China's most artificial intelligence university, has also been mentioned many times in this report: the registration rate of university artificial intelligence and machine learning-related courses has increased significantly worldwide, the most striking is the growth rate of Tsinghua University-related courses, from 2017 registration rate is 16 times higher than 2010, nearly three times higher than 2016.

Artificial intelligence application direction by country

It can be seen from the report that countries attach different importance to the application direction of artificial intelligence. China attaches great importance to the application of agricultural science, engineering and technology. Compared with 2000, China has paid more attention to agriculture in 2017. While Europe and North America place great emphasis on humanities and medical applications, Europe's research approach is more comprehensive.

AI skills and jobs

Job vacancies are serious

The graph above shows the number of job openings per year for the required Al skills, and of course, the growth rate of job openings for AI skills. The data of each position is correlated.

While machine learning is the most important skill, deep learning (DL) is growing at the fastest rate, with the number of job openings for deep learning increasing 35-fold between 2015 and 2017.

Gender distribution: imbalance still exists

The gender gap in AI-related jobs remains significant. The diagram above shows that skills are classified according to skills, and there are correlations between skills. The number of applicants does not represent the number of people employed, but overall, men account for 71% of applications for AI-related jobs in the United States. Among the subdisciplines of artificial intelligence, deep learning and robotics have a relatively balanced gender ratio compared to other disciplines.

International conferences and framework developments

The graph above shows the popularity ratio of various Al and ML packages on GitHub, and the number of stars can be used to approximate which AI framework is most popular. From 2015 to 2018, Tensoflow held the first place and scikit-learn held the second place.

Among large conferences, NeurIPS (formerly known as NIPS), CVPR and ICML are the three largest AI conferences with the largest number of participants. Since 2012, the growth rate of the number of participants, these three are also ahead of other conferences. It is worth mentioning that ICML nearly doubled in 2018.

The latest government policy

On the Chinese side, in July 2017, the government released a new generation of AI development plan, which formulated an AI development strategy until 2030. It is planned to bring China's AI technology to the most advanced level in the world by 2020. By 2025, China will make a major breakthrough in AI theory, and artificial intelligence will become the driving force for industrial upgrading and economic restructuring. By 2030, China will become the world's major new AI hub.

In September 2018, DARPA announced an investment of more than $2 billion to overcome the limitations of AI technology. Private companies play an important role in AI technology development/investment in the United States, and in 2017, technology companies such as Amazon and Alphabet provided a large amount of funding to the AI field.

In the EU region, France has recently published three reports: (1) the French Al Plan, which proposes a strategy for AI research, education and innovation;(2) a report focusing on social and regulatory issues; and (3) the Mission Villani report, which focuses on policy and reducing brain drain and addresses the importance of diversity in AI research. The UK has released a series of series since 2016.

April 2018: The government invests £ 950 million ($1.2 billion) to support research, education and enhance the UK's data infrastructure.

two more thing

Computer vision is a subdiscipline of artificial intelligence that has wide applications in autonomous vehicle development, augmented reality, and object recognition capabilities, and just like machine learning, training algorithms improve over time.

In terms of performance, AI computing power continues to soar, especially in areas such as computer vision. By measuring ImageNet's benchmark performance, the report found that within 18 months, the training time for the most accurate image classification model dropped from "around hours" to "around four minutes." Training is about 16 times faster. Other areas, such as object segmentation, improved accuracy by 72 percent in just three years.

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.