Back to feed
News Story
Import AI
1 sources

RAND Publishes US Superintelligence Strategy Report

RAND has published a lengthy paper outlining what the US should do to secure geopolitical advantage on an uncertain path to superintelligence. The main takeaway is that because so much of how the next phase of the AI takeoff will occur is unknown, the best thing is for the US to maintain a 'Freedom of Action' strategy. The strategy includes four key ingredients: building a human-AI ecosystem, building an AI-security architecture, adapting legacy national security enterprises, and investing in the capacity of citizens, firms, and governments to respond.

SynthePulse Insight · AI deep readingMembers

RAND's "Freedom of Action" Strategy: Why America Chooses to Keep All Options on the Uncertain Path to Superintelligence

Version 1 · 1 source

RAND releases a long paper arguing that the United States should adopt a "freedom of action" strategy on the unknown path to superintelligence, the core of which is to spend money to preserve options rather than betting on a single direction.

  • Core argument of the RAND paper: because the development path to superintelligence is highly uncertain, the best strategy for the United States is to maintain "freedom of action," that is, to spend money to preserve future options.
  • The "freedom of action" strategy contains four major elements: building a human-machine ecosystem, building an AI safety architecture, transforming the traditional national security system, and investing in the response capacity of citizens and institutions.
  • The paper summarizes three major categories and seven archetypal strategies: coexistence (dominance, joint development, preparation), denial (pause, deterrence, social continuity), and acceleration.
Open section navigationA Strategy Paper That Does Not Bet on a Single Direction

A Strategy Paper That Does Not Bet on a Single Direction

RAND has released a long paper exploring how the United States should "secure geopolitical advantage on the uncertain path to superintelligence." The paper's core conclusion is that because there are too many unknowns about exactly how the next stage of AI takeoff will unfold, the best approach for the United States is to maintain a "freedom of action" strategy. The purpose of this strategy is to "build and maintain capabilities to secure U.S. geopolitical advantage during the transition to superintelligence, and to ensure humanity survives with agency."

The logic behind this judgment is: since it is impossible to determine which endgame is most likely, spending money now to preserve options is more reasonable than betting on a single path. The paper does not assert that any one future is bound to arrive, but instead takes uncertainty itself as the starting point for strategic design.

Four Pillars: From the Human-Machine Ecosystem to Citizen Response Capacity

The paper proposes four key components of the "freedom of action" strategy. First, build a human-machine ecosystem: invest in AI safety, develop tools to protect human agency, invest in sharing the benefits of technology, prepare society for AI-driven disruption, and shape the incentive structure of the AI ecosystem.

Second, build an AI safety architecture: develop tools to understand the frontier of AI systems at home and abroad, including visibility into the underlying compute; build technology to verify potential agreements; and create regulatory expertise to govern the technology.

Third, transform the traditional national security system for the AI era: comprehensively reform the country's security institutions so that they are prepared to respond to the changes brought by AI and can leverage AI capabilities.

Fourth, invest in the response capacity of citizens, businesses, and government: provide information to decision-makers to help them think about AI; develop and diffuse positive uses of AI to respond to the changes it brings; establish "break glass" plans for responding to crises; and improve AI literacy.

Free for now

Read the full analysis

3 more sections of analysis, plus the full takeaway

Loading

Credibility boundary

This article is based on Import AI issue 473's reporting on the RAND paper. The specific content of the RAND paper, the strategy categories, and the list of uncertainties all come from that secondhand report, and the original text could not be independently verified. Import AI author Jack Clark's comments are personal views.

Sources for this version

  1. Import AI 473: The US's superintelligence strategy; human brain in a mouse skull; and machine hermeneuticsImport AI

Primary report

Import AI

Primary source