In the realm of artificial intelligence, the year 2028 looms as a pivotal moment, offering two contrasting scenarios for the global AI leadership race between the United States and China. This article delves into the critical factors that will shape this future, emphasizing the need for democracies to maintain their edge in AI development and deployment. With AI's rapid advancement, the stakes are higher than ever, and the decisions made today will have far-reaching implications for the world's technological and geopolitical landscape.
The Compute Advantage
At the heart of AI development lies access to powerful computer chips, known as 'compute'. The US, with its innovative ecosystem of companies, currently holds a substantial lead in this domain. However, China, under the jurisdiction of the Chinese Communist Party (CCP), is not far behind, thanks to its world-class talent and strategic exploitation of loopholes in US export controls. The CCP's AI labs have been leveraging distillation attacks to illicitly extract innovations from American companies, narrowing the gap in model intelligence.
The upcoming years will witness a period of great acceleration in AI capabilities, with the 'country of geniuses in a data center' potentially becoming a reality. This acceleration makes policy action even more urgent, as the US and its allies must act now to secure a lead in frontier capabilities, estimated to be 12-24 months by 2028. A strong lead in this period would be invaluable, but the window of opportunity is fleeting.
Two Scenarios for 2028
Scenario One: Democratic Leadership
In this optimistic scenario, America and its allies successfully defend their compute advantage. Policymakers tighten export controls, disrupt China's distillation attacks, and accelerate AI adoption in democracies. As a result, democracies set the rules and norms for AI, ensuring a safe and beneficial transition to transformative AI. This scenario also presents an opportunity for productive engagement with China on AI safety, which is supported by the US.
Scenario Two: Authoritarian Triumph
Conversely, if the US chooses not to act, China's AI labs will quickly catch up and even surpass America. In this scenario, AI norms and rules are shaped by authoritarian regimes, and the best models enable automated repression at scale. The US, by not tightening loopholes on China's access to compute, inadvertently contributes to this authoritarian triumph, which occurs on the back of American compute.
The Imperatives of Staying Ahead
The US and its allies approach AI competition from a position of strength, with the tools for AI dominance built by an innovative ecosystem in democratic nations. However, the task at hand is to avoid squandering this advantage and not make it easier for the CCP to catch up. The competition is not just about intelligence but also about domestic adoption, global distribution, and resilience.
AI as a Dual-Use Technology
Frontier AI will significantly impact the future military balance, and the CCP is already building its military for an AI-enabled battlefield. The PLA strategists view AI as a means to catch up and surpass the US military. The risk compounds as frontier AI will accelerate other critical technologies, widening the lead in the national security technology stack.
Neck-and-Neck Competition and Responsible AI
A neck-and-neck race between American and Chinese AI labs could hinder responsible AI development and governance. PRC labs, despite their concerns with AI safety risks, have not translated these concerns into safety practices comparable to US labs. The pattern of releasing dual-use capable models as open-weight further exacerbates the problem, making models accessible to malicious actors.
Policy Objective: Creating and Maintaining a Lead for Democracies
The US supports policies to build and maintain a safe, near-term lead over the CCP in intelligence, domestic adoption, and global distribution. This lead is crucial to avoiding authoritarian AI leadership and protecting national security interests. Engaging with AI experts in China on safety is essential, but the US must maintain a strong capabilities advantage to influence AI safety effectively.
The Mythos Preview Wake-Up Call
The release of Mythos Preview, a model by Anthropic, signals an acceleration period that demands urgent policy action. With access to this model, Firefox was able to fix security bugs more efficiently, highlighting the potential of AI in enhancing cybersecurity. Frontier AI capabilities will soon approach the 'country of geniuses in a data center' portrayal, making policy decisions even more critical.
Four Fronts of the Competition
The US-China competition for strategic advantage in frontier technologies is multifaceted, playing out on four fronts: intelligence, domestic adoption, global distribution, and resilience. Intelligence is the most crucial, driving geopolitical competition and market adoption. However, intelligence alone is insufficient; the integration of near-frontier AI systems into China's economy and security apparatus could shift the balance of power.
The State of the Competition
Compute, the advanced semiconductors needed for AI, is a critical input on all four fronts. Democracies currently hold the lead in compute leadership, but the gap is narrowing. China's chipmakers, despite state-backed investment, remain behind due to US and allied export controls. The US-PRC compute gap is widening, and tightening controls and enforcement can help close loopholes propping up the CCP's AI ecosystem.
How Democracies Built the Lead
The compute lead is attributed to the innovation of companies like NVIDIA, AMD, and others, along with forward-looking policy actions by the last three presidential administrations. Bipartisan efforts have protected US and allied innovation by restricting access to the US AI stack by PRC firms. However, some worry that this may force PRC labs to innovate, potentially reducing the American lead.
How the CCP Stays Competitive
Despite export controls, China's AI labs have been able to stay close on intelligence through illicit and evasive compute access and distillation attacks. The US has taken steps to address these threats, but the competition remains intense. The decisions made today will determine the future of transformative AI and the global order.
Ensuring Democracies Lead
To secure a democratic lead in 2028, the US supports policy actions to close loopholes, restrict model access, and promote the export of American AI. These measures aim to sustain and expand the AI lead, ensuring a safe and beneficial transition to transformative AI. The world's future hinges on these decisions, and the US is committed to leading the way.
