Interview: "ChatGPT isn't made to speak the truth". "It's designed to tell you the most plausible thing given its training data. Sometimes that will coincide with the truth, but very often it won't."
"Not only does it get things wrong, it gets things wrong in very plausible ways."
These models compress large amounts of data into the neural network. But that compression process can't be perfect.
~Mike Wooldridge
https://www.youtube.com/playlist?list=PL65IchQPZKwnAIaxht94BtOADVOQOvJH1