When you increase "software intelligence", you exponentially increase incorrect specification risks, code complexity, testing costs... If you want a cheap, testable, reliable software, better make it as simple as feasible. You can go for more complex softwares only if you have the extra-money... and are willing to take the extra-risks that come with them...
Naturally.
There'll have to be a demonstrated need, i.e., accident reports, pilot concerns, claiming there becomes too much for the crew to handle under a certain set of conditions. If true, then that'll become reflected in requirements to fix it. You can want a cheap airplane, but the requirements will tell you how cheap it can be.