Yes there's a bias from the training data, that will always be there, but ChatGPT has an additional model stopping it from generating certain things and has prebaked responses on some topics.

Comments