Future Projections
Overview
Section titled “Overview”This section explores different scenarios for how AI development might unfold. Scenario analysis helps prepare for multiple futures rather than betting on a single prediction.
Scenarios
Section titled “Scenarios”AI development succeeds with safety:
- Alignment research proves sufficient
- Safety measures are implemented effectively
- Humanity retains meaningful control
- Probability estimate: 15-35%
Misaligned AI causes major harm:
- Alignment fails despite effort
- Deceptive or power-seeking AIRiskPower-Seeking AIFormal proofs demonstrate optimal policies seek power in MDPs (Turner et al. 2021), now empirically validated: OpenAI o3 sabotaged shutdown in 79% of tests (Palisade 2025), and Claude 3 Opus showed...Quality: 67/100 emerges
- Ranges from setback to existential catastropheAi Transition Model ScenarioExistential CatastropheThis page contains only a React component placeholder with no actual content visible for evaluation. The component would need to render content dynamically for assessment.
- Probability estimate: 5-25%
Development is deliberately slowed:
- Government intervention or lab coordination
- More time bought for safety research
- Possible through regulation or crisis response
- Probability estimate: 10-25%
Gradual change without clear transition:
- No single “AGI moment”
- Ongoing adaptation to incremental progress
- Neither utopia nor catastrophe
- Probability estimate: 30-50%
Multiple powerful AI systems compete:
- No single winner-take-all
- International competition dynamics
- Complex coordination challenges
- Probability estimate: 25-45%
Scenario Analysis Uses
Section titled “Scenario Analysis Uses”Scenarios help:
- Test robustness - Which interventions help across scenarios?
- Identify early signs - What would indicate we’re heading toward each?
- Prepare contingencies - What should we do if each materializes?
- Communicate - Make abstract risks concrete
Key Uncertainties
Section titled “Key Uncertainties”The scenario we end up in depends heavily on:
- AGI timelines (sooner favors speed, later favors preparation)
- Alignment difficulty (harder favors pessimistic scenarios)
- Coordination success (better coordination enables pause/redirect)
- First-mover dynamics (monopoly vs. multipolar outcomes)