In the mid-1960s, Soviet computer scientist Mikhail Moiseevich Bongard created sets of visual puzzles where the objective was to spot an easily justifiable difference between two sides of a single image (for instance, white shapes vs black shapes, etc...). The idea was that these puzzles could be used to teach computers the general faculty of abstraction: perhaps by learning to spot the differences between these sorts of images, a computational agent could learn about inference in general. Considered a global (...) expert on Bongard problems, cognitive scientist Harry Foundalis developed the Phaeaco cognitive architecture for his PhD thesis - based on emulating cognition by solving the problems, creating a kind of artificial intelligence. In this paper, the extent to which Foundalis' approach allows for artificial general intelligence (the ability to reproduce a wide range of human abilities, or the goal of cognitive models) will be evaluated - with reference to Daniel Dennett’s reductive theory of mind and Immanuel Kant’s concept of the phenomenon and the noumenon. The point of view presented is that Phaeaco is missing several characteristics of general artificial intelligence. (shrink)