The Vision and Parsing Subsytems
- Vision is described elsewhere, but there is a retina of
on-off and off-on detectors, a primary visual cortex for
edges and angles, and a secondary visual cortex which
recognizes stalactites and pyramids.
- The retina and V1 are reasonably biologically accurate.
V2 is not.
- Parsing is done with a stack, and produces semantic frames.
- Variable binding by short-term potentiation supports the
stack and frames.
- It's a bit slow to reset for the next part because the
system has to wait several hundred cycles for the
bindings to be erased.
- The symbolic hook here is to read the next word of the
sentence.
- Otherwise, it is all neural.
