2022斯坦福AI指数报告-230页_29mb

AI Index Report 2022 Summary

Core Content

The AI Index Report 2022 provides a comprehensive overview of the global state of artificial intelligence (AI), including trends in research and development, technical performance, AI ethics, economic impact, and policy developments. It is a collaborative effort by the Stanford Institute for Human-Centered AI, drawing on data from various academic, private, and nonprofit organizations.

Main Points and Key Information

Private Investment and AI Adoption

• Private investment in AI surged in 2021, reaching approximately $93.5 billion, more than double the 2020 total.
• The number of newly funded AI companies decreased from 1051 in 2019 to 746 in 2021.
• The number of funding rounds worth $500 million or more increased from 4 in 2020 to 15 in 2021.
• The "data management, processing, and cloud" sector received the highest private AI investment in 2021, followed by "medical and healthcare" and "fintech."
• The United States led in both total private investment and the number of newly funded AI companies, with figures 3 and 2 times higher, respectively, than China.

Global AI Research and Collaboration

• The United States and China had the most cross-country collaborations in AI publications from 2010 to 2021, with a fivefold increase since 2010.
• The U.S.-China collaboration produced 2.7 times more publications than the U.K.-China collaboration.
• China led the world in the number of AI publications (journals, conferences, and repositories) in 2021, with a 63.2% increase compared to the U.S.
• The U.S. had a dominant lead in AI conference and repository citations among major AI powers.
• Educational and nonprofit organizations were the most productive in cross-sector collaborations, followed by private companies and educational institutions, and then educational and government institutions.

AI Technical Performance

• AI systems have become more affordable and higher performing, with the cost to train an image classification system decreasing by 63.6% since 2018 and training times improving by 94.4%.
• Large language models have set new records on technical benchmarks but also show increased bias, with a 280 billion parameter model in 2021 showing a 29% increase in elicited toxicity compared to a 117 million parameter model from 2018.
• Computer vision subtasks like medical image segmentation and masked-face identification have seen rising interest.
• General reinforcement learning has improved significantly, with AI systems achieving 129% improvement on more general tasks compared to narrow tasks like chess.

AI Ethics and Bias

• Research on fairness and transparency in AI has increased fivefold since 2014.
• Algorithmic fairness and bias has transitioned from an academic focus to a mainstream research topic.
• Multimodal models reflect societal stereotypes and biases, with experiments showing that images of Black people were misclassified as nonhuman at over twice the rate of any other race.
• Researchers with industry affiliations contributed 71% more publications at ethics-focused conferences compared to the previous year.

AI in the Economy and Education

• New Zealand, Hong Kong, Ireland, Luxembourg, and Sweden had the highest growth in AI hiring from 2016 to 2021.
• In the U.S., California, Texas, New York, and Virginia had the most AI job postings, with California having 2.35 times more than Texas.
• Washington, D.C. had the highest rate of AI job postings relative to its total job postings.
• AI patents increased 30 times from 2015 to 2021, with a 76.9% compound annual growth rate.

AI Policy and Governance

• The number of AI-related bills passed into law in 25 countries grew from 1 in 2016 to 18 in 2021.
• Spain, the United Kingdom, and the United States passed the most AI-related bills in 2021, each adopting 3.
• In the U.S., only 2% of proposed AI-related bills became law from 2015 to 2021.
• State legislators passed 1 out of every 50 proposed AI-related bills in 2021, while the number of such bills proposed grew from 2 in 2012 to 131 in 2021.
• The 117th U.S. Congress had the highest number of AI-related mentions since 2001, with 295 mentions by mid-2021, compared to 506 in the previous session.

Global AI Trends

• Robotic arms have become significantly more affordable, with the median price dropping fourfold from $50,000 in 2016 to $12,845 in 2021.
• AI ethics research is becoming more widespread and is increasingly being driven by industry researchers.

Structure of the Report

Contributors and Steering Committee

• Co-Directors: Jack Clark and Ray Perrault
• Members: Erik Brynjolfsson, James Manyika, John Etchemendy, Yoav Shoham, and others
• Staff and Researchers: Daniel Zhang, Nestor Maslej, and others
• Affiliated Researchers: From institutions like The World Bank, Microsoft, and others
• Contributors: A wide range of individuals and organizations contributed to different sections of the report.

Public Data and Tools

• The report includes raw data and an interactive tool for comparing up to 29 countries across 23 indicators.
• Public data is available on Google Drive and through the Global AI Vibrancy Tool.

Supporting Partners

• The report is supported by organizations such as Google, OpenAI, Open Philanthropy, Bloomberg Government, and others.

Citation

• The report can be cited as:
  Daniel Zhang, Nestor Maslej, Erik Brynjolfsson, John Etchemendy, Terah Lyons, James Manyika, Helen Ngo, Juan Carlos Niebles, Michael Sellitto, Ellie Sakhaee, Yoav Shoham, Jack Clark, and Raymond Perrault, "The AI Index 2022 Annual Report," AI Index Steering Committee, Stanford Institute for Human-Centered AI, Stanford University, March 2022.

Conclusion

The AI Index Report 2022 highlights the rapid growth and increasing complexity of AI, emphasizing its global reach and the growing emphasis on ethical considerations. It underscores the dominance of the U.S. and China in AI research and collaboration, the affordability and performance improvements in AI systems, and the rising importance of AI ethics in both academic and industry settings.