It keeps making the same mistakes over and over again. If used you need to probe further. Always have it lay out the assumptions and references that are used. Verify those references!!! I have sought several references that I ultimately found did not exist.
I always get the sneaky suspicion that I am training it with my questions. My company has oked its use, and it is definitely "book smart". It does fall apart at the same point that I do (really need to test and analyze this).
If we plan to continue with retail education (where the professor just gets online graded assignment, many multiple choice or fill in the blank), then it will be a problem. And of course essay writing even though I think I could smell out an essay created by AI (that isn't the say that what the hard work of gathering references and interpreting references hasn't been short circuited by AI). I would recommend profs go back to pen and pencil in a proctored environment (alternatively dumb terminals to complete essays).
My prof for my engineering class didn't have an issue with us using AI (It was part of what he was trying to evaluate). Again it was idiotic in attempting to write code for a specialized finite element code (ADPL). Still it is great for getting blocks of output written (always laborious since ADPL seems to be on a foundation of FORTRAN).
My prof has only one class a semester with 20 or so Jr. to graduate student level engineering students so he can get away with traditional engineering tests. Not sure what you would do if you had to grade 300 tests now days. Even though my FIL did that for years carrying 2-3 classes (2 when he was head). He also taught the Mandatory Elective Math course that most students not in math or sciences had to take (I would cut my throat before I had to do something like that).
= new reply since forum marked as read