When the read next word rule (from the Rule One subnet) comes
on, a symbolic action turns the last word off and turns the
next word on.
The bar one subnet has two CAs, and the first is word active.
That's turned on too.
That's the only symbolic action. I have medium term hopes (2 years)
of hooking this up to a vision system for reading words, then the
system will be entirely subsymbolic.
The girl saw the dangerous pyramid with the stalactite.
The combination of word active and the The input
ignites the the CA in the other subnet.
Later, girl becomes the input. In combination with
word active, this is enough to ignite the girl CA
in the Noun Access subnet.
This is enough to ignite the girl CA in the Noun
Semantics net.
Two caveats on neural reality here. First most of the CAs
are set up so that half the neurons are on at all times
flipping back and forth.
Second, almost all CAs are orthogonal. The semantics are the
exception so with similar words (via word net semantics) having
overlapping CAs.
