why is it that implementations of computer vision using machine learning algorithms have not yet been achieved human like perceptual abilities

AI will only every truly understand human communication when the following is true:

AI is developed not as code in a box, but as code in an AI virtual world - that provides experiences, context, simulated physical laws, and the passing of time. AI is coded with purpose - which we would call emotions, so as to allow it to 'wish' to communicate, to 'want' to acquire knowledge, to 'enjoy' contact. AI is not taught human language, but allowed to experience it in the context of the above two models. A virtual world for the AI to virtually inhabit, and virtual emotions to drive the AI to do more than simply sit in the box.