Scenarios for the Next Century
Forecasting is hard. Scenario planning — disciplined exploration of plausible futures — is more useful than point prediction for AI-shaped civilization.
Key facts
- Scenario planning is widely used in policy, military, and corporate strategy (Royal Dutch Shell pioneered the modern form).
- No serious forecaster gives any single AI scenario above ~50% probability.
- Ord (2020) estimated total existential risk this century at ~1-in-6, with AI as the largest single contributor.
- Open Philanthropy, RAND, and CSER publish public scenario analyses.
Soft Takeoff, Broad Prosperity
AI capability rises gradually; institutions adapt; productivity gains are broadly shared. Requires strong governance, distribution, education, and global coordination.
Closest historical analogs: electrification (1880–1930), the postwar productivity boom, and the post-1990 internet diffusion.
Hard Takeoff, Concentration
Rapid capability gains concentrate power with whoever controls compute, data, and frontier models. Outcomes depend on the values, accountability, and oversight of those actors.
Concerns explored in detail by Bostrom (Superintelligence), Christiano (Without Specific Countermeasures), and the AI Now Institute.
Plateau and Stagnation
Scaling and algorithmic returns diminish. AI becomes important but ordinary, like electricity — transformative over decades, not years.
Hardware constraints, data limits, or fundamental algorithmic ceilings could each produce this outcome.
Catastrophe
Misalignment, deliberate misuse (bio, cyber, autonomous weapons), or geopolitical breakdown produce severe global harm. Avoidable in principle, but requires deliberate effort and international cooperation.
The Ord / Bostrom existential-risk taxonomy distinguishes recoverable from unrecoverable catastrophes.
Multipolar AI World
Multiple powerful AI systems and actors — US labs, Chinese labs, open-source ecosystems, sovereign-AI programs — coexist with no single dominant winner. Coordination, standards, and competition policy become central.
Probably the most empirically consistent scenario with 2024–2026 conditions.
How to Use Scenarios
Scenarios are planning artifacts, not predictions. Robust strategies hedge across scenarios; brittle strategies optimize for one.
Major militaries, oil companies, and policy think tanks all run formal scenario exercises; AI governance has begun adopting the practice.
Frequently asked
Which scenario is most likely?
+
Honest answer: no one knows. The distribution itself is the planning artifact, not a single forecast.
Should I plan for catastrophe?
+
At the policy and institutional level, yes — even low-probability tail risks justify hedging. Individually, the highest-leverage actions are usually civic and professional rather than personal preparedness.
Who does this work?
+
Open Philanthropy, RAND, CSER (Cambridge), CSET (Georgetown), GovAI, and frontier AI labs' policy teams all run formal scenario work.
Sources & further reading
Continue in this series
Macroeconomics
The Post-AGI Economy
Biology
AI, Biotech, and Longevity
Philosophy
Transhumanism and Human Enhancement
Long-termism
AI, Space, and the Long-Term Future
Human Condition
Meaning, Purpose, and Identity in an AI Era