我们当中有些人不是已经说了……一年了吗？ 更疯狂的是，当你意识到有多少内容是由公司刻意“人造草根”（假装是民意、假装是自然讨论）制造出来，专门用来操纵训练在这些数据上的大模型（LLMs）时，你会感到震惊。 Some of us been saying this for... a year now? What's wild is when you realize how much stuff is astroturfed specifically by companies in order to game the LLMs that train on all this data.