The Future of Artificial Intelligence Governance
0 likes212 views
The document discusses the current state of AI governance, contrasting approaches in Europe, China, and the U.S. It highlights Europe's regulatory framework, China's authoritarian oversight, and contrasts Biden's AI executive order with Trump's revocation of it to enhance U.S. AI leadership. It raises concerns about the implications of AI technology advancement on safety, regulation, and international competition, especially with China's growing capabilities in AI.
The Future of Artificial Intelligence Governance
- 1. The Future of AI Governance Charles Mok Research Scholar, Cyber Policy Center, Stanford University @charlesmok www.linkedin.com/in/charlesmok/
- 2. Let’s begin with the present, or, is it the past? ● The techno-optimists and the pessimists – Risks and benefits of genAI ● Governments wants economic and social benefits, but also safety ○ Scientific and technology leadership ○ Data intensive industries, from financial/manufacturing to robotics and creative industries ○ Concerns for job displacements ○ Safety against rogue AI and crimes – disinformation, synthetic imagery, cyber scams, etc. ○ Military and national security considerations? ● Legal issues ○ Training on copyrighted materials ■ If that is ok, then what about distillation? ○ Libel and defamation
- 3. The Three Digital Empires: First, Europe ● From Prof Anu Bradford’s “Digital Empires” on digital regulations to AI governance ● Europe: The AI Act ○ Building on GDPR, DSA, DMA, Copyright Directives etc. ○ Four risk categories ■ Unacceptable risk – e.g. social scoring, person-based crime prediction -> PROHIBITED ■ High Risk – e.g. critical infra, edu, healthcare, employment, law enforcement, justice -> risk management, data quality & governance, record-keeping, transparency, cybersecurity etc. ■ Limited Risk – e.g. chatbots, GAI -> require transparency to inform users ■ Minimal or no risk – e.g. video games, spam filters -> no regulations ○ Also, the Cyber Resilience Act (cybersecurity rules) and Product Liability Directive (liability for “defective AI systems” ○ Also required: supply chain and training data transparency, energy consumption, etc. ○ Stakeholders are involved in enforcement by developing codes of practice
- 4. Then, China ● Authoritarian state control ● Multiple related regulations on digital spheres and also specific to AI and aspects of AI: ○ Previously: CSL, DSL, PIPL, etc. ○ Also: Algorithm Recommendation Provisions, Deep Synthesis Regulations, Interim Measures for Generative AI Services, Basic Safety Requirements for Generative AI Services, etc. ○ Adherence to “socialist principles” and XJP thoughts: Close monitoring, surveillance and censorship ● The reality is a cat-and-mouse game ○ Delicate balance between control and technology advancement goals ■ DeepSeek example reveals how “easy” it is to overcome censorship, or to produce malicious codes and disinformation despite all monitoring efforts
- 5. Finally, the US (before) ● First, before Jan 20, 2025 – Joe Biden’s EO14110 “Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” ○ 8 guiding principles and policy priorities for federal agencies and authorities, and improved AI oversight in federal govt ○ Establishing non-binding standards on AI risks and cybersecurity principles with industry and stakeholders ○ Support for research and development, e.g. National Artificial Intelligence Research Resource (NAIRR) Pilot under NSF ● State legislations ○ E.g. California’s “Safe and Secure Innovation for Frontier Artificial Intelligence Models Act,” vetoed by Gov Newsome.
- 6. The US (after, now) ● Now, after Jan 20, 2025 – on Jan 23, Donald Trump revoked Biden’s EO14110 ○ “The Biden AI Executive Order established unnecessarily burdensome requirements for companies developing and deploying AI that would stifle private sector innovation and threaten American technological leadership.” ○ Instead, Trump “calls for departments and agencies to revise or rescind all policies, directives, regulations, orders, and other actions taken under the Biden AI order that are inconsistent with enhancing America’s leadership in AI.” ○ “American development of AI systems must be free from ideological bias or engineered social agendas. With the right government policies, the United States can solidify its position as the leader in AI and secure a brighter future for all Americans. The order directs the development of an AI Action Plan to sustain and enhance America’s AI dominance, led by the Assistant to the President for Science & Technology, the White House AI & Crypto Czar, and the National Security Advisor. It further directs the White House to revise and reissue OMB AI memoranda to departments and agencies on the Federal Government’s acquisition and governance of AI to ensure that harmful barriers to America’s AI leadership are eliminated.” ○ Then, he repeats his previous (unfulfilled) AI orders on spending on R&D etc. from 2019-2020.
- 7. The Big Picture ● It’s unclear to begin with, plus the uncertainty from the US ○ If US Big Tech indeed will take over the US direction, what direction is that really? ○ Where will the AI Safety Summit, G7 Hiroshima AI Process, G20 AI Principles, etc. go from here? ○ Will the “Brussel Effect” on digital platform services still be relevant for AI? ○ The DeepSeek shock ■ Yes, China can (to some extent), and that’s already giving hope to other countries, from Europe to the Global South ■ The Trump paradox ● DeepSeek vs TikTok re: data transfer ● How to respond to safety (or national safety) concerns while pledging to “ensure that harmful barriers to America’s AI leadership are eliminated” ● Scaling vs. other innovation and optimization ● Export restrictions – more, or less? ● Tariffs – 100% on Taiwan/TSMC? Really? ● To H1-B or not to H1-B, that is the question
- 8. Paper: Taking Stock of the DeepSeek Shock ● 6 questions ○ How did DeepSeek happen? ○ $6M only, really? ○ Scaling vs. optimization? ○ Do export restrictions work? ○ Data transfer to China? ○ Did they cheat? ● 6 inferences ○ The U.S. is still ahead in AI but China is hot on its heels. ○ The genie is out of the bottle. No longer can U.S. or any AI companies overly rely on brute-force scaling. ○ The U.S. government and industry must realize that China has a thick playbook to disrupt the U.S.-led AI ecosystem. ○ The U.S. must redouble its effort in supporting fundamental research and talent development. ○ DeepSeek is also disrupting its Chinese AI competitors and may contribute to restructure the future AI ecosystem of China and the world. ○ The U.S. will need to develop a response to China’s open source strategy on AI and other technology areas.