A robot will uncritically do what you tell it.
An experienced welder will tell you you're an idiot when you're an idiot.
A robot will uncritically do what you tell it.
An experienced welder will tell you you're an idiot when you're an idiot.
You could imagine a situation where the automated QA is trusted and does its job in sets, with humans interspersed to backup the QA programs. But one of the QA checks is actually buggy and OK’s the missing weld, but the full batch of work covers it up before reaching a human for the double-check
In which case, both issues are at fault: too much reliance on the QA program, and allowing too much work to be done before it reached a human (bad double-check setup)
Especially in case the robots share code and e.g. have the same parse error when loading your list of welding points.
And might well muck it up one time out of 50.