I am ok with clearly-labeled posts that are in dedicated comms and occassionally enjoy examples of the shitshow that the models produce when they hallucinate. However, if the model is trained on works without authorization from and compensation to the creators of its training data, I find their use beyond ethically-questionable.