The brain is an amazing machine, capable of many different types of processing, learning, thinking and creativity. Computers are also very flexible - let's not be discouraged.
Computers Simulating Neurons
At the moment, in order for a computer to simulate neural processes, transistors are combined into gates that build processing cores that manipulate floating point numbers that model equations that act on arrays held in memory to map their current states to the next states. Simulating learning requires a whole set of other mathematical processes layered on top.
This is a 'scenic route' to getting where we want to go (implementing intelligence inside a machine). Unfortunately, we don't really have a map that would tell us about the direct route.
Neurons Simulating Computers
After a human brain has learned to process images in front of its eyes, and recognize written words, further education can lead to the ability to handle mathematical concepts within the realm of "the imagination". The processing and manipulation of these concepts takes place using the same underlying infrastructure as is utilized in the nervous system of (say) a worm.
Here, the brain seems to be taking a 'scenic route' to processing mathematical concepts. And it seems like its somewhat pot-luck whether individual humans have any aptitude for it.
Better Adaptation through Design
There hasn't been enough time (on an evolutionary scale) for the brain's circuits to have become well-adapted at processing (say) calculus efficiency.
Just because the brain's method of solving image-recognition tasks seems leaps-and-bounds better than 'plain' computer science approaches doesn't mean that we have to rethink every aspect of how to go about designing machine minds.
In terms of computer algebra, for instance, we have already figured out a better route from A to B than the scenic route that the brain takes. There are likely to be more optimizations available - particularly in the higher-order functions that have been around less long in evolutionary terms.