斯坦福-AI（人工智能）指数大牛解读行业发展-2017.11_101页-7mb

AI Index 2017 Annual Report Summary

Core Content

The AI Index 2017 Annual Report is an open, not-for-profit project launched by the One Hundred Year Study on AI at Stanford University. It aims to provide data-driven insights into AI activity and progress across multiple domains, including academia, industry, open-source software, and public interest. The report highlights the rapid growth of AI-related fields and calls for broader community involvement in tracking and analyzing AI trends.

Main Sections

Volume of Activity

• Academia

  • The number of AI-related papers published in the Scopus database has increased more than 9x since 1996.
  • Stanford's introductory AI and ML course enrollments have grown by 11x since 1996.
  • AI paper publishing growth outpaces general Computer Science paper growth, indicating a stronger focus on AI.

• Industry

  • The number of active US AI startups has increased 14x since 2000.
  • VC investment in AI startups has grown 6x since 2000.
  • AI job openings have increased 4.5x since 2013, with the US still dominating the market.

• Robot Imports

  • Industrial robot shipments into North America and globally have seen significant growth.

• Open Source Software

  • GitHub data shows increased interest in AI and ML software, with TensorFlow and Scikit-Learn being particularly popular.

Public Interest

• Media Sentiment
  • The percentage of media articles referencing AI has grown, with a mix of positive and negative sentiment.

Technical Performance

• Vision

  • Object detection error rates in the ImageNet competition have dropped from 28.5% in 2010 to below 2.5%.
  • Visual Question Answering (VQA) systems have made progress, though the dataset is evolving.

• Natural Language Understanding

  • Parsing performance has improved significantly.
  • Machine translation between English and German has reached high accuracy.
  • Question Answering systems, such as those tested on SQuAD v1.1, have improved to near human levels.

• Speech Recognition

  • Speech recognition systems have achieved performance close to human-level in controlled environments.

• Theorem Proving & SAT Solving

  • Theorem proving tractability has improved, showing better performance by AI systems.
  • SAT solvers have also shown strong performance on industry-applicable problems.

Key Insights

• AI Vibrancy Index

  • A composite metric that combines academic activity (paper publishing and course enrollment) and industry activity (VC investment) to measure the overall vitality of the AI field.
  • The index shows that academic activity initially drove AI progress, but industry investment became the main driver around 2013, with academia catching up since.

• Human-Level Performance

  • AI systems have made notable progress in areas such as game playing (Othello, Checkers, Chess, Jeopardy!), object detection, and skin cancer classification.
  • However, generalization and real-world application remain challenges, as AI often excels in narrow, controlled tasks.

Limitations and Future Directions

• Data Bias

  • The report is heavily US-centric and lacks international data.
  • It does not include demographic breakdowns or government and corporate investment data.

• Missing Areas

  • The report does not cover all technical areas, such as dialogue systems, planning, and commonsense reasoning.
  • Areas like recommender systems and standardized testing are yet to be adequately measured.

• Call to Action

  • The AI Index invites contributions from the broader community to enhance data collection, analysis, and the development of new metrics.
  • It emphasizes the need for diverse and inclusive data to understand the societal impact of AI.

Conclusion

The AI Index 2017 Annual Report provides a comprehensive overview of AI's growth and technical progress, but acknowledges its limitations. It serves as a foundation for future reporting and encourages collaboration to ensure a more accurate and holistic understanding of AI's development and impact.