RAND maps seven archetypal US strategies for superintelligence and picks "Freedom of Action"

Import AI·medium signalXBlueskyLinkedInCopy link

Import AI 473 covers a RAND analysis that sorts US approaches to superintelligence into seven archetypes across three families — coexistence, denial and acceleration — and recommends a "Freedom of Action" posture built on four pillars: a human-AI ecosystem, an AI-security architecture, adapted national security institutions, and more capacity for citizens, firms and government. RAND names five critical uncertainties, including how close the danger is and whether restraint is feasible. The same issue covers a Nature paper on xenocortical mice carrying human cortical organoids that matched controls on memory maze tasks, and a Toronto/Harvard research agenda on pacing rather than halting AI, which identifies compute, algorithms, model weights and deployment as the intervention points and flags weights as the hard case because they copy freely.