Comment by mitxela

They're not fundamentally unsolvable - even bigger networks with even more training can simply be trained to give the correct answers to all of these questions.