Pardon my ignorance but what exactly was the issue with what LeCun said in that response? I feel like it's a pretty agreed upon paradigm that feeding an AI/ML algo with biased data for training will make the AI/ML algo biased. That seems unsurprising and like a perfectly valid statement so there must be something I'm missing and, at least from the article, I couldn't glean what the issue was since the response was just "I'm sick of this framing" without any kind of additional information. If I'm reading it correctly, the issue maybe was that he didn't go far enough and put too much emphasis on data set bias?