OECD+2023年G7广岛会议：生成式人工智能进程-英-37页_2mb

G7 Hiroshima Process on Generative AI Summary

Report Overview

This report analyzes the G7 Hiroshima Process on Generative AI, focusing on opportunities, risks, and international priorities. It is based on a questionnaire administered in Q3 2023, involving G7 members to stocktake generative AI developments and guide policy discussions. The report highlights trends, incidents, and policy gaps, emphasizing the need for international cooperation.

Key Findings on Generative AI Growth

• Public and Economic Impact: Generative AI has rapidly entered public discourse, with an eightfold increase in news articles and tweets between late 2022 and early 2023. Productivity gains, research, and venture capital investments surged, reaching USD 12 billion globally in 2023's first half, driven by advancements like large language models.
• Technological Trends: Open-source developments show gradual growth, with key risks including disinformation, intellectual property rights violations, and privacy breaches. Early data indicates over 1,000 incidents related to generative AI based on news reports.

Major Opportunities

• Identified by G7 members include productivity improvements, innovation in healthcare and climate solutions, promoting fair recruitment, and enhancing democratic processes. Examples like AI in education and the workplace show potential benefits.

Primary Risks

• G7 members unanimously listed disinformation and manipulation as major threats. Other key risks include intellectual property infringement, privacy breaches, cybersecurity threats, and bias in AI outputs. These challenges affect sectors such as media, education, and healthcare.

G7 Priorities

• Most Urgent Actions: Responsible use, addressing disinformation, and governing AI systems are top policy priorities. Other considerations include human rights protection, data governance, and robust AI quality control.
• Notable Trends: G7 countries are leveraging existing frameworks and developing guidelines; for instance, Japan is creating unified guidelines, while the EU proposes the AI Act. Gaps vary by country, with Canada and Italy addressing risks through regulatory tools and codes of practice.

International Cooperation

• G7 members stress the need for global alignment on standards, interoperability of frameworks, and dialogue frameworks like the OECD AI Principles. Recommended actions include voluntary codes, research on risk tools, and international initiatives to mitigate global threats.

Conclusion

Generative AI holds vast potential but requires swift, coordinated action to manage risks and harness opportunities. The report calls for continued international dialogue and policy development to ensure safe and equitable AI deployment.