The mission of the AI Index is to ground the conversation about AI in data. Talking about their efforts, the team in the report says, “The AI Index is an effort to track, collate, distill, and visualize data relating to artificial intelligence. It aspires to be a comprehensive resource of data and analysis for policymakers, researchers, executives, journalists, and the general public to develop intuitions about the complex field of AI.” Its is interesting to note the growing importance of China in the world of AI. China was mentioned 69 times in 94 page report and India was mentioned barely 5 times in the report.
The AI Index is prepared by Yoav Shoham who chairs the committee and works at Stanford University. The rest of the team includes, Raymond Perrault from SRI International, Erik Brynjolfsson from MIT (Massachusetts Institute Of Technology), Jack Clark from OpenAI, James Manyika from McKinsey Global Institute, Juan Carlos Niebles from Stanford University, Terah Lyons from Partnership On AI and John Etchemendy from Stanford University and Barbara Grosz from Harvard University.
AI Index puts out major insights into the world of Artificial Intelligence and the report has four sections:
- Data: Volume of Activity and Technical Performance
- Other measures: Recent Government Initiatives, Derivative measures, and Human-Level Performance
- Discussion: What’s Missing?
Here are the key takeaways of the report:
1.AI Research Papers Publishing Increased
The report says that Europe has consistently been the largest publisher of AI papers. The share of Europe’s research paper is 28% of all AI papers on Scopus in 2017. The report says that the number of papers published in China increased 150% between 2007 and 2017. The number of Chinese papers increased despite the spike and drop in Chinese papers around 2008.
There is also a suggestion, that AI publishing is growing because of heightened interest in computer science. More findings in the report suggest that 56 percent of papers came from the the Machine Learning and Probabilistic Reasoning category. The same number was only 28% in 2010. The most remarkable observation was that Neural Networks research paper had a compound annual growth rate of 3% from 2010—2014, followed by a CAGR of 37% from 2014—2017.
2. Investments And Gender In AI
The report says that from January 2015 to January 2018, the amount of AI startups increased 2.1x and the number of AI startups has seen exponential growth. Interestingly, from 2013 to 2017, the report also observed that VC funding in AI increased 4.5x, which is huge. Naturally, ML is the largest skill cited as a requirement in job openings. Lately, deep learning (DL) is growing at the fastest rate which increased 35x.
The report also shed light on the gender disparity in artificial intelligence. The report observed men make up 71% of the applicant pool in the U.S. In 2014 30% of all AI patents came out of the U.S followed by South Korea.
3. Open Source And Public Perception Of AI
The popularity of various AI-programming frameworks can be measured by the stars on its Github repo. The recent trend observed is that frameworks backed by major companies are really popular. Some examples are Tensorflow (Google), Pytorch (Facebook), mxnet (Amazon).
The team also showed that AI articles in the media have become less neutral and more positive. Articles have been 12% positive in January 2016 and went to 30% positive in July 2016. The percentage of positive articles has sticked to nearly 30% since then, the report observed.
The mention of the term “Artificial Intelligence” is used more in comparison of “Machine Learning”. The report said that in the U.S. words like Machine Learning or Artificial Intelligence was said at least once in a given event or a discussion. In United Kingdom’s data indicates that Machine learning or Artificial intelligence was said at least once in a given comment online.
4. Technical Performance
The report says that ImageNet accuracy has seen large improvements in training time and accuracy. Machine translation has also improved consistently, BLEU scores from English to German are 3.5x greater today than they were in 2008, the report said. The performance in question answering has improved also. The report said that the performance in 2018 has gone from 63% to 69% on the Easy Set for question answering and from 27% to 42% on the Challenge Set for question answering.
The technical performances and training times have constantly improved pointing out the obvious fact that artificial intelligence is revolutionising technology and is essentially the core of computer science.