[D] Why are ChatGPT's initial responses so unrepresentative of the distribution of possibilities that its training data surely offers? Submitted by Osemwaro t3_ziwuna on December 11, 2022 at 3:41 PM in MachineLearning 8 comments 2