Study Guide: Chapter 5 — Decisive Strategic Advantage

Core Idea

A "decisive strategic advantage" (lead big enough for world domination) can arise even without a fast takeoff, because the crossover dynamic (Ch. 4) can turn a modest head start into total separation. Unified AI systems face fewer of the restraints (agency problems, risk aversion, internal coordination failure) that historically kept human powers from converting temporary advantage into permanent control.

Key Terms

Decisive strategic advantage · singleton · crossover · agency problem · Baruch Plan

Case Summary

US nuclear monopoly, 1945-49: could have threatened first strike (Russell, von Neumann) or negotiated a strengthened veto-less UN (1946 Baruch Plan — rejected by Stalin, fearing being outvoted). Neither hardline nor benign path was taken to its full conclusion; Cold War arms race followed instead. Closest real precedent for a power holding — and not fully using — a decisive strategic advantage.

Application Checklist

• [ ] Distinguish "who reaches parity first" from "who ends up with an unbridgeable lead" — a moderate takeoff can still produce the latter
• [ ] Check whether a leading actor's advantage is compounding (crossover-like) before assuming small leads stay small
• [ ] When assessing monitoring/detection strategy, weigh physical-capital-heavy paths (easy to monitor) against pure-software paths (hard to monitor)
• [ ] Ask what specifically restrains a human organization from exploiting an advantage — then check if that restraint would hold for a non-human agent

Self-Test

1. Why can a moderate (not fast) takeoff still produce a decisive strategic advantage for the frontrunner?
2. What structural advantage does a unified AI project have over a human organization in maintaining secrecy and avoiding internal betrayal?
3. Why does Bostrom think the historical restraints that stopped the US from using its nuclear monopoly to force a singleton might not apply to a superintelligent AI?