This is fascinating reading…

The purpose of the experiment…was to explore ChatGPT’s conception of humanist ethics. Doing so using a sensitive topic such as religion underscores the malleability of ChatGPT’s responses. Especially considering the number of soft logical fallacies we got away with, it should give pause to the reader’s confidence in AI. ChatGPT already has a reputation for generating fake references while at the same time giving convincingly human-like responses. In addition to this empirical unreliability, we believe we have demonstrated ethical unreliability in the program.

Read more.