20211101-世界经济论坛-The_AI_Governance_Journey_Development_and_Opportunities_31页_3mb

The AI Governance Journey: Development and Opportunities

Overview

Since its inception in 2019, the Global AI Council has worked to shape international governance for artificial intelligence (AI) through multistakeholder collaboration. Key goals include ensuring responsible AI applications to build public trust, bridge the “Pacing Problem” (governance lagging behind technological progress), and accelerate innovation. Current challenges include translating principles into practice and addressing gaps in global regulation.

Current Trends and Key Highlights

✱ Development of Governance Frameworks

• 4 Eras of AI Governance (2009-Present):

  • Pre-2010: AI寒冬与复苏，需技术评估防止滥用
  • 2010–2016: 技术加速带来挑战
  • 2016–2019: 原则与指南的诞生
  • 2019–Present: 技术加速与治理创新，风险层级管理体系

• Key Standards Addressed:
  Privacy, fairness, transparency, explainability, human oversight, cybersecurity, algorithmic bias, human-centred design, child-centric AI.

✱ AI Governance by Practice

• From Principles to Action:
  Mechanisms include standardized labelling, third-party auditing (e.g., ethical certification programs), risk-based regulations, and AI procurement guidelines.
• Public Awareness:
  Proposed tools include Denmark’s Data Ethics Seal and Malta’s first-of-its-kind AI certification programs.

✱ Multistakeholder & Agile Governance

• Strategy: Collaboration among industry, government, academia, and civil society bridges knowledge gaps and ensures broader buy-in.
• Examples:
  • Gaia AI Accelerator launched in 2021
  • Agile testing of FRT systems via public assessments.
  • Regulatory sandboxes in India, Malta, and the UK.

Key Challenges and Gaps

✱ Stifling Innovation?

• Risk: Lengthy regulation could hinder startups.
• Solution: Context-sensitive regulation and voluntary frameworks.

✱ Governance Shortcomings:

• Economic Inequality: AI may displace 400M–800M jobs by 2030.
• Ethical Limits: Bias in training data and unintended outputs.
• Geopolitical Fragmentation: Differing national AI policies (e.g., China, the EU).

✱ Climate Footprint:**AI Training,"Greenwashing," and Sustainability Directives.

Future Directions

1. Global Harmonization:
   Interoperable governance mechanisms via alliances like GPAI and GAIA.

2. Frontier Use Cases:
   Monitoring bias, transparency, fairness, ethical computing, disinformation, quantum governance.

3. Positive Economic Policies:
   Investing in reskilling, profit-sharing from AI (via “Windfall Clause”), and equitable benefit distribution.

4. Proactive Interventions:
   ⚖️ Policymaking to integrate fairness, safety, and diversity in AI products early on.

Concluding Thoughts

Multistakeholder governance is essential for responsible AI development. By fostering cross-sector collaboration, we can ensure AI benefits society while leveraging cutting-edge innovation.