The human brain uses the data flows from the sensorimotor system to construct information resonances around 3D objects and behaviors perceived in the real world. These sensorimotor data fields are the underlying 'alphabet' of internal 3D object formation and animation. There is no known mathematical relationship between the sensorimotor data fields and the subsequent memory resonances of consciousness. But it may be possible to link the two domains through dynamic transform functions that can maintain binding and coherence between the perceived reality and the virtual representations.

Significant processes in the real world can be identified as positions, relationships or motions in euclidean space, and expressed with regard to a simple cartesian coordinate system x,y,z and time involving only 'real' numbers.

Tensor mathematics is used to manipulate these classes of data and is ideally suited to computer modeling through the notation of matrices in handling vector arithmetic.

An example translation might be the non trivial vector transformation between the alignment of the eye based on visual cues and the muscular control of the same. For instance visual input following a snow flake and the muscular outputs controlling the eye muscles. (A snowflake object, a 2D camera view port a short distance away and a mechanical tracking system). There is clearly a relationship or transformation between these domains while still sharing the same euclidean coordinate system.

Although this is interesting research in its own right, it actually speaks more about the grace of cybernetic processes and how dedicated biological neural structures can both learn and apply these transformations outside of higher cognitive oversight.

Our problem is of a similar class, in that there is a transformation between the real and the mental or 'imaginary' euclidean spaces. Although there is arguably some plasticity in the coordinate system, it is only to the extent they are congruent that cognitive processes have any relevance or potency in the real world. Traditional mathematics does not really speak to the kinds of processes found in either nature or cognition. A new kind of mathematics is required, perhaps something more akin to Stephen Wolfram research where the creation of complexity arises from simple rules.

The speculation being that if the representation of complex models and behaviors could be generated from simple rules, and the brain could manipulate such rules using its parallel nature rather than the sequential nature described in Stephens discoveries, then the possibility arises of the brain being a general rule manipulation workspace, and memories not really existing anywhere until they are manifested by the rules. Thus the brain becomes a general purpose processor for generating rules from complex modality patterns and subsequently processing those rules. But there are so many unanswered questions. How are the rules recognized and constructed from say vision. How are the rules manipulated, morphed and aligned to the modalities, or to creative processes. Unless there are special, as yet unknown properties these rules possess, it does not necessarily move us closer to the actual raw mechanism of consciousness processing if we have no math or knowledge for dealing with such phenomina.

If a rule exists for say a candle structure, and another rule for a liquid flow behavior, are there methods by which such rules can engage, a burn rule, a flow rule, a bounce rule etc. The modality flow does not explicitely define form or behavior, they are hidden within complexity. Even 3D representations do not imply function or behavior (from collisions etc.) These things can be processed using complex algebra or relation to precedents, but if the generation rules themselves could be merged and aligned to direct the resulting representation, the way small changes in genetic code can cause significant effects in the phenotype. Then a mechanism might be in place to direct consciousness. In nature, the rules process matter to define its form and behavior (plants, animals). In consciousness, the rules process representations of matter as the fluid virtual forms and behaviors of thoughts.

Part of this research involves attempting to find any mathematical relationships between the parallel data flows originating from the outside world and the data structures held within memory. But more important even than this, is the method by which subsequent coherent structured animations can take place.

The initial problem of testing an incoming parallel 2D data stream like vision, to matching render planes of internal 3D objects could be conceptualized as a brute force calculation of comparing each incoming data set to trillions of potential simulation 'fits'. Conventional computing may remain substantially underpowered for this class of problem for some time, but quantum computation may be more suited.

The fundamental barrier to AI is image instantiation. For example, what challenge would exist in order to instantiate a well with a bucket being manually wound up and down the shaft. How would a visual system go about resolving what was happening mechanically, to the rope etc. You have the environment, the actors and the animation. Each of these needs to be resolved. In Q1 2003, Nvidia will release a cinematic quality real time hardware render engine in a single chip. This will be capable of over 100 high definition, fully rendered frames per second including atmospheric, lighting and camera effects. Assuming a sufficiently deep library of models and environments, Could this tool be used to track real time vision in order to build and match a 3D simulation by shear brute force trial rendering. If initially seeded with a human generated reference world as close to the vision as possible? Would traditional software engineering be up to the challenge of following subsequent motion through the positive feedback using 100 render frames per second. Adjusting the model world to minimize the render plane differences? The 3D models that generate the matching render planes to the input vision, become legitimised and learned as credible object precedents within the growing internal universe.

The current research is centered around the neural morph tunnel metaphor. See below

An initial 8 by 8 pixel morph tunnel is to be modeled in computer simulation to test the basic principles of operation - Initially, using very simple image data sets of un-shaded object primitive outlines (coin, cube, cone etc.) to investigate the boundaries and weaknesses of the architecture. Methods for incorporating surface detail; for pre-scaling of data; layer organization; environmental constraints; object norms; object degrees of freedom; match probability thresholds; noisy data sources; trained error consequences and correction mechanisms. Techniques for prioritizing search pathways and encoding object confidence weightings.