Perhaps cognition is a collection of problems with exponential complexity. 

In that case we would see some success, but not be able to scale to much really useful.

Also, the brain would have to not be a Turing machine, and the strong physical Church-Turing thesis would be false. 

Seems simpler maybe?