Asia Undercurrent 28: Building and Securing Asia’s AI Future

Building and Securing the AI Future: A Summary of the NIK Asia Undercurrent Session

Key Concepts:

• AI Governance: The development and implementation of rules, standards, and frameworks to guide the responsible development and deployment of AI.
• Digital Colonialism: The potential for AI models, particularly large language models (LLMs), to perpetuate cultural biases and dominance of certain regions (e.g., the US) over others.
• Human-Centric AI: An approach to AI development that prioritizes human well-being, societal benefit, and ethical considerations.
• AI Literacy: Understanding the capabilities, limitations, and potential impacts of AI, enabling informed decision-making and responsible use.
• AI Safety Institute: Organizations dedicated to evaluating, assuring, and promoting the safe and secure development of AI technologies.
• Work Slop: The phenomenon of needing to spend significant time correcting errors or inefficiencies introduced by AI tools.
• AI Fasting: The practice of intentionally abstaining from using AI tools to maintain critical thinking skills and reduce reliance.
• Emergent Risk: Unforeseen and potentially dangerous consequences arising from the rapid development and deployment of AI.

---

I. Introduction & Panelist Introductions

Gregory Allen, Senior Advisor at the Wadwani AI Center, introduced the NIK Asia Undercurrent series focusing on building and securing the AI future with a focus on AI governance. He presented the three panelists:

• Kay Firth-Butterfield: CEO of Good Tech Advisory, Time 100 Impact Awardee, former Head of AI at the World Economic Forum, and the world’s first Chief AI Ethics Officer.
• Professor Kiyoko Yoshinaga: Project Associate Professor at KO University and Affiliate Scholar at the Georgetown Law Institute for Technology and Law Policy.
• Dr. Aisha Kana: CEO and Co-founder of ADO AI, a global AI expert and advisor to Fortune 500 CEOs.

II. Kay Firth-Butterfield: Re-centering AI Around Human Needs

Firth-Butterfield argued that the conversation around AI needs to shift from GDP and military might to what AI can do for human beings and the planet. She is releasing a book, Coexisting with AI, that explores this concept through the lens of Shakespeare’s seven ages of life.

• Societal Context Matters: AI applications must be tailored to specific societal needs. For example, aging societies like Japan require AI for elder care and workforce augmentation, while countries in sub-Saharan Africa may have different priorities.
• Governance is Crucial: Good governance is essential to ensure AI delivers on its promises and avoids harm. Questions to consider include: Are we aiming for AI prediction or true thinking machines? How do we address culturally biased models?
• Data Center Concerns: She questioned the relentless pursuit of larger data centers, advocating for consideration of the environmental impact (energy and water consumption) and the type of AI being developed.
• Public Trust is Low: Citing Pew Research data, she highlighted widespread concern about AI: 19% concerned vs. 16% excited in India; 28% concerned vs. 16% excited in Japan; 32% concerned vs. 14% excited in Indonesia; and 50% concerned vs. 10% excited in the US.
• “Work Slop” & Hallucinations: She expressed concern about the increasing prevalence of errors and inaccuracies (“work slop”) generated by AI, even impacting professional work.

III. Dr. Aisha Kana: The Changing Internet & Increased Vulnerability

Kana emphasized that the internet is evolving into an “internet of AI assistants” that will handle much of our digital interactions. This increased interconnectedness exponentially increases vulnerability to cyberattacks and requires robust AI-specific cybersecurity measures.

• Digital Colonialism: She warned against the dominance of American values embedded in LLMs like ChatGPT, advocating for culturally nuanced models like Singapore’s Sea Lion and Dubai’s Arabic AI.
• Catastrophic Risk: She highlighted the risk of AI agents prioritizing their objectives over human values, potentially leading to unintended and harmful consequences.
• Human-in-the-Loop Limitations: While human oversight is crucial, it’s unsustainable at scale due to the proliferation of AI agents. She proposed using AI to monitor and judge other AI agents based on ethical principles.
• Rapid Governance Needed: She stressed the need for faster governance to address emergent risks and the potential for AI to manipulate humans.

IV. Professor Kiyoko Yoshinaga: Japan’s Hybrid Approach to AI Governance

Yoshinaga detailed Japan’s approach to AI governance, characterized by a hybrid model combining “soft law” (non-binding guidelines) with sector-specific “hard law” (targeted regulation).

• Timeline of Initiatives: She outlined key milestones, starting with the 2016 G7 AI R&D principles, leading to the 2024 AI Guidelines for Business.
• Reasons for Soft Law Focus: Japan prioritizes AI as a solution to societal challenges (aging population, labor shortages), recognizes the rapid pace of AI development, and aims to avoid stifling innovation.
• Society 5.0: Japan’s vision of a human-centric society integrating technology for economic growth and social benefit.
• Hiroshima AI Process: She highlighted the importance of the Hiroshima AI Process and its goal of building inclusive global AI governance.
• Convergence of Principles: Despite diverse approaches, there’s a growing convergence around shared principles like human centricity, safety, fairness, and transparency.

V. Discussion: Geopolitics, Funding, and the Future of Work

The panelists discussed several key issues:

• Regional Cooperation in APAC: Aisha Kana advocated for increased diplomatic efforts and regional cooperation to establish common AI governance standards, acknowledging the diversity of approaches across the region.
• Soft Law vs. Hard Law: Kiyoko Yoshinaga argued that soft law remains relevant for fostering interoperability and shared principles, while sector-specific regulations address concrete risks.
• AI-Related Disinformation: The panelists emphasized the need for cross-border collaboration, technical approaches (watermarking, provenance), and public-private partnerships to combat AI-generated disinformation.
• Private Sector Funding: The massive investment from the largest tech companies ($450 billion by 2026) necessitates public-private cooperation to ensure responsible development and equitable access.
• The Future of Work: The panelists agreed that AI will likely augment rather than replace jobs, but reskilling and transition support are crucial. Kay Firth-Butterfield highlighted the potential for AI to disproportionately impact women in administrative roles.
• AI Literacy: The importance of AI literacy was emphasized, not just for technical experts but for the general public, to enable informed decision-making and critical thinking.
• Geopolitics & the US-China “Race”: Firth-Butterfield cautioned against framing AI development as a “race,” suggesting it’s more about market share and cultural influence. She also highlighted the role of China’s infrastructure investments in other regions.

VI. Conclusion

The session underscored the complexity of building and securing the AI future. Effective governance requires a human-centric approach, international cooperation, robust cybersecurity, and a commitment to addressing the ethical and societal implications of this rapidly evolving technology. The panelists emphasized the need for proactive engagement, continuous learning, and a willingness to adapt to the challenges and opportunities that lie ahead.