Stanford University releases the artificial intelligence index 2019 annual report
Stanford University, in conjunction with MIT, Harvard, OpenAI and other institutions, released the “artificial intelligence index 2019 annual”. The 2019 edition has three times as many data sets as the 2018 edition. This report includes nine aspects of global AI dynamism that analyze the development of AI.
The 2019 report also released two tools: a tool for searching AI research papers, AI Index arXiv Monitor, which ” is another tool that enables search of the full text of papers published to this pre-print repository, providing the most up-to-date snapshot of technical progress in AI.” The other is the Global AI Vibrancy Tool, which is “an interactive tool that compares countries across 34 indicators, including both a cross-country perspective and an intra-country drill down.”
AI Index 2019 Report Highlights
1. Research and Development
• Between 1998 and 2018, the volume of peer-reviewed AI papers has grown by more than 300%, accounting for 3% of peer-reviewed journal publications and 9% of published conference papers.
• China now publishes as many AI journal and conference papers per year as Europe, having passed the US in 2006. The Field-Weighted Citation Impact of US publications is still about 50% higher than China’s.
• Singapore, Switzerland, Australia, Israel, Netherlands, and Luxembourg have relatively high numbers of Deep Learning papers published on arXiv in per capita terms.
• Over 32% of world AI journal citations are attributed to East Asia. Over 40% of world AI conference paper citations are attributed to North America.
• North America accounts for over 60% of global AI patent citation activity between 2014-18.
• Many Western European countries, especially the Netherlands and Denmark, as well as Argentina, Canada, and Iran show relatively high presence of women in AI research.
2. Conferences
• Attendance at AI conferences continues to increase significantly. In 2019, the largest, NeurIPS, expects 13,500 attendees, up 41% over 2018 and over 800% relative to 2012. Even conferences such as AAAI and CVPR are seeing annual attendance growth around 30%.
• The WiML workshop has eight times more participants than it had in 2014 and AI4ALL has 20 times more alumni than it had in 2015. These increases reflect a continued effort to include women and underrepresented groups in the AI field.
3. Technical Performance
• In a year and a half, the time required to train a large image classification system on cloud infrastructure has fallen from about three hours in October 2017 to about 88 seconds in July, 2019. During the same period, the cost to train such a system has fallen similarly.
• Progress on some broad sets of natural-language processing classification tasks, as captured in the SuperGLUE and SQuAD2.0 benchmarks, has been remarkably rapid; performance is still lower on some NLP tasks requiring reasoning, such as the AI2 Reasoning Challenge, or human-level concept learning task, such as the Omniglot Challenge.
• Prior to 2012, AI results closely tracked Moore’s Law, with compute doubling every two years. Post-2012, compute has been doubling every 3.4 months.More…