Yeah I haven't looked into the specifics of what the Moral Turing Test are.
I imagine philosophers must have some (at least somewhat) reasonable scales and methods to quantify what moral answers are, at the very least.
Whether the LLM actually is doing better moral reasoning or happened to be able to trial-and-error its way to what the researchers wanted to see, I definitely don't have the expertise to say!
Start here. 3Blue1Brown is one of the best math communicators out there. It’s 30 min well spent and will erase any idea of “reasoning”. The article on LLM’s and cons completes the circuit, by circling back to how human reaction to the stuff.
Add comment