HELLFIRE If they’re training these models on users’ social media posts, maybe they should consider a more appropriate, honest name: artificial stupidity. How can these models be credible or trustworthy if those are the training data used? What proportion of social media content is even coherent, much less useful?