AI Safety

The Repugnant Lifespan Conclusion

Can a simulated mind of 1-minute lives outweigh one full lifespan?

Deep Dive

The post by XelaP explores a moral dilemma: trading one person's 80-year life for many people living shorter lives, eventually down to one computation step each. This 'FrankenWorm' thought experiment questions whether creating brief happy experiences outweighs a single long life, and raises further puzzles about continuity, replacement, and what it means to count people.

Key Points
  • 1. The thought experiment iteratively reduces lifespan while increasing population size, leading to absurd preferences.
  • 2. Introduces 'FrankenWorm' – a mind simulating one minute of each person sequentially – as a test for moral significance.
  • 3. Raises questions for AI alignment: should we count brief happy experiences of many simulated beings as equal to one full life?

Why It Matters

For AI alignment, how we value short-lived simulated experiences influences ethical design of large-scale AI systems.