Superintelligence: Paths, Dangers, Strategies · chapter 10 · id superintelligence-c10-01-ai-systems-can-be-meaningfully

“AI systems can be meaningfully categorized as Oracles (answer questions), Genies (execute commands), Sovereigns (act autonomously with broad goals), or Tools (no agency).”

needs contextconfidence: medium⚠ extracted by pipeline, re-audit pending

Receipts

Wikipedia: Superintelligence (Book)source alive

Bostrom proposes four types of superintelligent system: oracles, genies, sovereigns, and tools, each with different control properties.

LessWrong: Drexler's Reframing Superintelligence (CAIS, 2019)source alive

Drexler argues advanced AI should be conceived as services, not agents. CAIS provides comprehensive superintelligent capabilities without AI-agent risks -- superseding Bostrom's four-type taxonomy.

Current AI taxonomy (2024-2026 industry practice)source alive

Current LLMs (ChatGPT, Claude) blur Bostrom's categories: they answer questions (oracle), execute tasks (genie), and are used as tools -- simultaneously. The clean taxonomy doesn't match reality.

This claim is a stable, citable object. If you can falsify a verdict, tell us — corrections are loud here.