People that don't code think that something can do code 98% correct is surely better, without seeing the hard enforcement of a 2% error without manual intervention. I think all of the jokes about stupid computers and weird behavior of languages when you use them out of spec (famously labeled wat) gave people the wrong impressions about why those problems exist or how they can or cannot be fixed. You'd think they'd be able to transfer that cynicism quicker to models since they are also dumb computer things but apparently they say "hello" and that is tricking them out of that?