Panel, Conference Presentation
The Geopolitics of AI | Global Conference 2024
Milken InstituteErik Schatzker, Karen Kornbluh, Eva Maydell, Anne Neuberger, Karen Pierce, Kevin Rudd
- The U.S.-China relationship is defined as a bipartisan "strategic competition" where China views AI as a geopolitical "game changer" with concentrated policy efforts dating to 2017–2018.
- Chinese leadership anticipates deep anxiety regarding domestic political risks of deploying generative AI and remains concerned about the dynamism of the American ecosystem, though they are currently ahead of democracies in facial recognition due to fewer privacy constraints.
- American companies are projected to lead the development of the largest and most powerful models, with the U.S. and allies expected to innovate and maintain the leading edge for an indefinite period despite short-term uncomfortable moments.
- Competitive races are expected to persist in image identification, economic competitiveness, and cyber offense, with defensive applications seen as a potential area where the U.S. and allies may achieve greater success than in offensive domains.
- U.S. export controls are anticipated to slow but not prevent China's progress in computer chip technology, while different nations are expected to mobilize top talent and government programs to compete in the AI sector.
- A new European Commission, anticipated within a couple of months, will prioritize renewing China policy and balancing alignment with U.S. partners against maintaining a distinct European approach to AI.
- There is a projection that Europe's regulatory norms risk stifling innovation if they become overly precautionary, specifically regarding the EU AI Act, which may hinder competitiveness.
- Future leadership in AI is expected to depend on establishing adequate red-teaming, testing, and "human-in-the-loop" controls rather than solely on technical capabilities.
- Plurilateral cooperation among democratic nations and the Global South is viewed as feasible, whereas multilateralism involving all states faces difficulties due to authoritarian value sets limiting negotiation scope.
- U.S. export controls and private sector engagement are expected to be complemented by proposals for an entity similar to the International Energy Agency focused on critical minerals and chips.
- Public-private partnerships are deemed essential for AI governance, with private firms bearing responsibility for publishing model cards to ensure transparency regarding training data and bias.
- Societal literacy regarding AI's tailored messaging is currently low, creating urgent risks of election weaponization that will require heightened vigilance during the UK and U.S. elections within the next year.
- Regulators are projected to struggle to keep pace with the slow unfolding of AI technologies, necessitating immediate negotiation of basic rules of the road before the technology becomes more sophisticated.
- Lessons from prior military technologies like missiles and submarine detection are expected to inform the development of large-scale AI models, including potential national programs for drug discovery.
- Efforts to bridge understanding gaps include proposals for a group mirroring the 1950s Pugwash movement to bring together scientists, policymakers, and private sector members.
- There is an expectation that the U.S. and allies will need to impose a price on hostile foreign actors engaged in the deliberate subversion of democracies, while also increasing government fluency by recruiting tech sector personnel.
- Confidence in AI models is expected to grow as companies improve citizen competence regarding model trustworthiness, though current governance models require innovation due to the private sector's primary role in driving advancement.