Yes, I know. It's just kind of interesting how it can make inferences about complicated things but not get multiplications correct that would almost definitely

Yes, I know. It's just kind of interesting how it can make inferences about complicated things but not get multiplications correct that would almost definitely have been in its training set many times (two digit by two digit)