According to my research, in order to create machine intelligence of a similar type to our own, we would need to convert the 2 dimensional source materials in the outside world, into an inner 3 dimensional simulation of time and space using the resources of 4th dimensional memories of unpolarized form and behavior. That is, form with no prior scaling or perspective; behavior through time; or observer / observed collapse.

The physical objects and behaviors referred to in the 2d source materials need to be recognized (instantiated from a 4th dimensional library of forms and their potentials), and animated together within a 'mental' scene simulating 3d space and time.

The subsequent animations need to be resolved, or evaluated according to a value system (emotion). This is a 4th dimensional information process. One which resolves, or collapses, the non-polarized (unconscious) memories of form, into a polarized scene of object forms and behaviors. That is, to an observer perspective (the observed, or the conscious state). And remaining linked back to the 4th dimensional potentials (the observer side of the equation, or the unconscious state).

An enabling technology might be digital memory organized into a translation matrix. Where a common notional 3d space is mapped onto a 'vocabulary' of touch and muscle scripts (necessary to navigate the space), and a library of form and behaviors made up from that core sensorimotor vocabulary.