The system has not been able to define a solid test for "non-obviousness" which is one of the criteria for patentability. As a result, in my experience, approvals are based most notably on "novelty" -- If no one has seemingly done in the past, it tends to get approved.
In my opinion, that's what has gone wrong with the system, i.e., not the theoretical criteria for patentability but its execution.
In my opinion, that's what has gone wrong with the system, i.e., not the theoretical criteria for patentability but its execution.