i n c a - research

- R5 -

If you take a VCR, you could generate hundreds of thousands of words and images to describe the intricate details of its operation, down to the tiniest transistor and coil. But the essence of a VCR can really be described quite accurately in just a few words - 'A machine to store incoming video data onto removable magnetic tape cartridges, which can subsequently be replayed.'

The human brain is similar, complex in detail, but simple in principle - that principle being -

'The ability to store parallel data streams from reality exposure, that can subsequently be recalled internally to simulate that reality.'

It doesn't matter which sensory modality we consider (sight, sound, touch etc.), they all exist as parallel time varying signals encoding data objects from the real world.

Reality is presented to the senses, and thus to the brain, as a jumble of mixed and partial object data sets (chair, violin, mouse, river etc.) The brain recognizes patterns to build a library of 'instances' and subsequently of 'archetypal forms' - data sets and abstractions that represent individual objects and features of reality. As these forms become tagged with language patterns, they invariably become universals - that 'particular' tree becoming a 'universal' tree archetype.

This library of forms is used to decipher subsequent sensory flow to form models and animations based on precedent, thus understanding and predicting events in reality.

The brain is programmed to emotionally grade (judge) those simulations and act (animate) accordingly.

So we have:

The ability to extract, identify and link patterns from slow parallel analog data streams.
The ability to then manipulate those patterns independently.
The ability to arrange those patterns into a universal 'stage' map.
The ability to link those patterns to other data sets known as language patterns (both oral and written).
The ability to construct and interpret animated scenes from these language pattern tags.
The ability to use those serial language tags to receive and transmit scene compositions socially and with machinery.
The ability for language patterns to also attach themselves to emotional grading schemas.

The human brain is able to build 3D form from 2D modality exposure, particularly vision. The parallel data from the eyes (or indeed any sense modality) enters the neural network structure of the brain and is broken down into individual correlation sets, such as:

regions of pattern surface similarity
regions of relative motion (over time)
pattern (form) edges

Those branches are then instantiated in the brain to identify associated/connected object surfaces. Those surface patterns are then connected to form an object whole. Then those whole forms are joined to form animation (motion or shape) behavior precedents.

These data sets can then later be recalled through pattern recognition, language addressing tags, or object association links (cat/basket/fur/Felix, cup/tea/cake/kettle, auto/road/gas etc.). The object sets are used as the foundation of scene recognition, then following object behavior into settings of pasts and futures. Finally, these data are available for creative and imaginative thinking through morph or substitution translations.

The data exists in an analog form of parallel signals varying in intensity over time. The brain can only hold a finite number of data frames (time) in buffer memory. It needs to be translated (decoded) down to the underlying 3D objects and animation narrative in order to be either understood (in relation to the internal known universe and wider time frames) or used to guide any cognitive or real life physical action.

This research looks at the kinds of data pathways and structures found in the biological brain, and which have the potential to so organize the parallel data such that.

Data patterns can be stored (either as the raw parallel data wavefronts- or more likely as the more efficient extracted feature sets from the incoming patterns.
data can be associated with pre-stored records
data can be recalled (e.g. through language attachments)
data can be used for high speed pattern matching (sub second)
Research into the volume of data stored and the encoding format efficiencies to yield the overall useable memory capacity and any memory limits of the biological brain. The effect of overload, of data fragmentation (of both objects and linkages). The data pathways, width and velocity. The effect of physical or chemical disruptions etc.

Research Challenge:

Search for any correspondence between biological memory systems and silicon. Whether biological memory formation can be synthesized from current memory technologies, or whether new classes of memory are needed. The kinds of speeds and size limits to be expected. The resilience to data brittleness, how is the brain able to use deteriorated data sets. Where are the limits to instantiation success. How are the probabilistic object matches assessed. How are emotional tags associated with the 3D objects or narrative and used to grade and guide the scenes?