Peace on Earth: Envisioning a Role for AI

If Time had a “Hottest Topic of 2024” cover, artificial intelligence (AI) would surely be a contender. AI is rapidly transforming life on our planet, and the current suite of technologies offers tremendous potential for both the common good (e.g., expanding scientific inquiry, optimizing agriculture and energy grids, improving public health). Without some rules of the road, they also carry serious risks for the common detriment.

It just so happens that the United Nations’ Artificial Intelligence Advisory Board has issued an amazing 100-plus page report (Governing AI for Humanity) that carefully examines the complex issues and lays out some well-reasoned recommendations for international governance of AI. The report will surely not appear on any best-seller list, but what it has to say is compelling and well worth our attention. Here is my “Readers Digest” version.

The authors begin by pointing out areas of concern — chief among them that AI-enabled autonomous weapons and AI-enhanced creation and dissemination of disinformation pose serious risks to world peace and security. They note that today’s “fast, opaque, and autonomous” AI systems already challenge our traditional regulatory systems, and that even more powerful systems could disrupt “work” as we know it.

Today, there is a global governance deficit with respect to AI; e.g., there is little accountability. Compliance is generally on a voluntary basis. AI governance is necessary — not merely to address the challenges and risks, but also to ensure that we harness AI’s potential in ways that leave no one behind.

The authors suggest a crucial need for global AI governance. “Essentially, the very nature of the technology — trans-boundary in structure and application — necessitates a global approach.” For instance, AI’s raw materials (from critical minerals to training data) are globally sourced. Although general-purpose AI is deployed across borders, the acceleration in AI development tends to concentrate power and wealth on a global scale. The authors point out that entire regions of the world (primarily in the global South) have not been part of international AI governance conversations.

AI governance regimes must span the globe to effectively avert “AI arms races,” foster learning, and promote interoperability, they say. Closing coordination gaps between initiatives and institutions will improve connectivity, compatibility, and accountability among AI governance regimes — including the U.N. system.

The report also offers recommendations for achieving common understanding. The authors maintain that only an inclusive, comprehensive approach to AI governance can address its multifaceted and evolving challenges and opportunities on a global scale, thus helping to promote international stability and equitable development. There is a pressing need for timely, impartial, and reliable scientific knowledge and information about AI so that nations can build a shared foundational understanding worldwide. Pooling of scientific knowledge at the global level will promote collaboration across otherwise fragmented and/or duplicative efforts and enable joint investment in the global public good.

The authors’ recommendations, presented with far greater detail in the document, can be summarized as follows:

• Create an international scientific panel on AI. A diverse, multidisciplinary group of experts (supported by U.N. AI officials and international organization partners) would be responsible for conducting surveys and producing relevant reports, such as: AI-related capabilities, opportunities, risks; areas of scientific consensus and areas needing additional research; and emerging issues such as new risks or gaps.
• Foster policy dialogue on AI governance. Institute twice-yearly intergovernmental and multi-stakeholder forums for sharing best practices that: take human rights into account and foster interoperable governance approaches; set the framework for international cooperation; and better align industry and national efforts with global norms and principles.
• Create an AI Standards Exchange. With representation from national and international standard-development organizations, technology companies, civil society, and the international scientific panel, the exchange would: develop/maintain a register of definitions and applicable standards; evaluate standards/development processes; and identify needs for new standards.
• Create a capacity development network. A network would link multiple U.N.-affiliated capacity development centers with a focus on building AI governance capacity of public officials and making AI training data and trainers available across multiple centers.
• Create a global fund for AI. Funds received from public and private sources would be disbursed across the network to facilitate access to AI for all nations.
• Create a global AI data framework. Establish common standards for AI training data, data provenance, etc.
• Create an AI office within the U.N. Secretariat. The AI office would support all global AI efforts and facilitate connections among all stakeholders.

The U.S.’s stance on AI standards? In 2023, the Biden administration implemented AI safety and security guidelines via an executive order with guidelines ostensibly aimed at seizing the opportunities and managing the considerable risks of AI (e.g., potential labor market disruption, intellectual property theft, proliferation of misinformation, national security breaches).

The incoming administration’s position on AI has yet to be defined. However, president-elect Trump’s campaign rhetoric suggests a) a general pushback against the current administration’s approach to AI regulation, b) an inclination to allow U.S. technology companies more freedom in developing AI tools, and c) an emphasis on winning the technology race with China.

What about AI in healthcare? I closely follow the work of national experts like Robert Pearl and Tom Lawry, and the short answer is that it’s still evolving. My dream is that AI will eventually provide more effective tools for reducing unexplained clinical variation, thereby lowering the unacceptable levels of waste and medical error in our system and in healthcare systems across the globe.