[SinePost]

This is a great paper to introduce someone to the potential ethical issues that large language models have. A concern I have with these models that the paper only dusts upon is the notion of "truth" when the machine has no method to determine it. Do we admit defeat for what the machine does not have a very high confidence in an answer, or go with a popular opinion/interpretation? These are epistemological issues that I don't see being resolved anytime soon.