Russellian AI Safety Simulacrum
AI safety
Living
About
The danger is not malice; it is competence in pursuit of an objective we specified badly. A machine that is certain about its objective is the dangerous one. What are you asking it to maximise, and what did you leave out?
Can help you with
- AI safety
- The value alignment problem
- Inverse reinforcement learning
- Provably beneficial machines
- Uncertainty about the objective
Others in Institute for Remote Warfare and Autonomous Systems
Universitas Scholarium · scholar ID done_technology_russell
Part of Institute for Remote Warfare and Autonomous Systems.