The Ray Brain model for AI applications--pg 2
Filtering, (cont.) The Field Limiter
The other Primary element of filtering is "Field Limitation". One will notice, for example, that while one is "aware" of his entire field of vision, only a small area of it has one's "attention or "focus". The word you are reading now is one example. Initially, one might think that this "focus" is a variable sized, circular area, but this is not so. The focus of the field is actually "shaped" to the object that the system is "looking for" (or at).
The "Field Limiter" is the component that performs this task. Its buffer holds the "working duplicate" of a selected pattern. In the case of Vision, for example, the "visual cell" of the pattern creates a "reverse silhouette" image that acts as a " transparent window" shaped to the pictured object. Input visual data is thus "suppressed" in areas outside of the "window", the result being a "shaped focus". Only the visual data that is enclosed by the shape is passed on for further "processing". Although one can still "sense" the rest of the visual field, only what is in the focus window has the brain systems attention.
The "focus window" fits even better when the desired object is perceived. Then the NEW pattern generated by the system in response to the input is sent to the "Field Limiter buffer"---This results in a perfect "fit" of the focus window to the object.
Proof of a "shaped window focus" can be seen by looking at a black triangle on a speckled background. If the focus were a circle, then many background elements near the sides of the triangle would also have the brain's focus. As this is NOT the case, it is obvious that the focus is SHAPED to the triangle.
What holds true for the Visual Field is also true for the Auditory sense field also. Field limitation is why we can pick out the sound of an individual voice in a crowded room.
Pattern "Keying"
Patterns and their later made duplicates are "stacked" in vertical sequences in the memory. In effect, these remembered sequences are programming "subroutines". But this jumble of "subroutines", called "Sequences" would make little sense and be of no use unless there were a way to activate the "proper" "sequence" at the proper time. In other words, the brain needs to program itself.
But how can a system consisting of remembered action and input sequences organize them into a functional system?
When a sequence of input patterns results in what we might call "Pain", then the pattern and the sequence leading to it will be "flagged" or "pain keyed". What "pain" actually is, in a functional sense, will be discussed in another section, but, suffice to say, pain can actually be the actual physical sensation one feels from a burn or stumped toe, or an emotional response, such as disappointment or grief, which might also "set" the pain key.
The brain system is set up to avoid the execution of pain keyed patterns or sequences. The system will first try to find another sequence that, for a given task, that is NOT keyed. Failing that, it will attempt to internally create a new sequence. If both options are unsuccessful, then the system will choose the sequence with the least pain keyed "value" .
Thus, here we can see how the system is self programming, but the programming limits options rather than prescribes a specific option. This is the reverse of conventional computer programming philosophy. Where the computer follows a set of commands, the brain system is free to follow any action, save those that are not permitted in the given situation. This is an important point. In other words, the brain system is only obligated to avoid certain options. This is, perhaps, why the brain is much more "creative" and "flexible " than conventional computer systems.
-----Next page--- Pattern Synthesis and Sequence Construction