2018年-数据局_大数据文摘：2018年度斯坦福AI指数_94页_3mb

AI Index 2018 Annual Report Summary

Core Content

The AI Index 2018 Annual Report is a comprehensive resource that tracks and analyzes the growth and development of artificial intelligence (AI) globally. It aims to ground discussions about AI in data by providing metrics on the volume of activity, technical performance, and other relevant factors. The report is led by a Steering Committee and managed by a Project Manager, with contributions from various institutions and organizations.

Main Objectives

1. Refresh metrics from the previous year.
2. Provide global context for AI activities and research.

Key Contributors

• Steering Committee: Yoav Shoham (Chair), Raymond Perrault, Erik Brynjolfsson, Jack Clark, James Manyika, Juan Carlos Niebles, Terah Lyons, John Etchemendy, Barbara Grosz
• Project Manager: Zoe Bauer

Key Metrics and Findings

Volume of Activity

• Academic Publications:
  • AI papers on Scopus increased 8x since 1996.
  • CS papers increased 6x during the same period.
  • In 2017, 28% of AI papers originated in Europe, 25% in China, and 17% in the U.S.
  • Machine Learning and Probabilistic Reasoning accounted for 56% of AI papers in 2017, up from 28% in 2010.
  • Neural Networks saw a 37% CAGR from 2014 to 2017.
• Corporate and Government Affiliation:
  • In 2017, Chinese government-affiliated AI papers increased by 400% since 2007, while corporate papers only increased by 73%.
  • U.S. corporate AI papers were 6.6x more than those in China.
  • U.S. AI authors were cited 83% more than the global average.
• Conference Attendance:
  • Large AI conferences (with over 2000 attendees in 2017) like NeurIPS and ICML have seen the most growth in attendance since 2012.
  • NeurIPS attendance grew 4.8x, and ICML grew 6.8x.
  • Small AI conferences, such as ICLR, have also experienced significant growth, with 20x increase in attendance from 2012 to 2018.

Technical Performance

• AI performance continues to improve across various domains, including:
  • Vision: Enhanced object detection capabilities.
  • Natural Language Understanding: Considerable improvement in this area.
• The GLUE metric and COCO leaderboard are new areas of study added to the report.

Diversity and Inclusion

• Gender Diversity: 80% of AI professors at leading universities are male.
• Community Initiatives:
  • AI4ALL and Women in Machine Learning (WiML) are increasing in participation.
  • AI4ALL alumni grew by 900% from 2015 to 2018.
  • WiML workshop registrations increased by 600% since 2014.
• Nonprofit Involvement: Over 50% of the Partnership on AI's members are nonprofits.

Government Initiatives

• The report highlights recent government investments in AI, particularly from the U.S., China, and Europe.
• The AI Vibrancy Index is introduced to measure the overall activity and impact of AI research across academia and industry.

Open Source and Robotics

• ROS (Robot Operating System):
  • Total downloads increased by 352% since 2014.
  • Unique downloads increased by 567%, suggesting more distinct users.
  • China saw the most growth in ROS pageviews, with an 18x increase from 2012 to 2017.

What's Missing?

The report includes suggestions from experts on how to make the AI Index more comprehensive and representative. It emphasizes the need for further collaboration and data transparency to better reflect the global AI landscape.

Additional Highlights

• Course Enrollment:
  • AI and ML course enrollments are increasing globally, with Tsinghua University in China showing the most significant growth.
  • In 2017, AI + ML course enrollment at Tsinghua was 16x higher than in 2010.
  • U.S. introductory ML course enrollment grew 5x from 2012 to 2017.
• Gender and Racial Diversity:
  • There is a notable lack of gender diversity in AI classrooms, with a significant proportion of AI professors being male.
  • Increased efforts are being made to promote diversity through initiatives like AI4ALL and WiML.

Conclusion

The AI Index 2018 Annual Report demonstrates that AI is a rapidly growing field with increasing academic and industry participation. While the U.S. and Europe continue to dominate in terms of citations and corporate involvement, China is emerging as a major player with significant growth in both publications and government initiatives. The report also highlights the importance of diversity and inclusion in AI research and development, urging greater transparency and collaboration to ensure a more representative and comprehensive understanding of AI's global impact.