BRUCE – Barnes' Rather Unique Cognitive Entity
A (virtual) cubic array of flexible, unassigned neural nodes, with
multiple fixed inputs and outputs spaced strategically around the
The node logic is flexible and self-deterministic. Internal
feedback features provide self-excitation, self-control, opportunity
for self-training. External support systems provide persistent and deep
memory and a few dedicated functions in advance of NNET learned
replacement. (Example: text/speech in/out.) Training can be external or
internal data match/optimization, self-acquired patterns, internal or
external positive/negative feedback.
Goal: training through
conversation. ??? 4 matched CPU/board/mem sets for increased nnet
processing throughput and larger nnet space. (120g ram) / (24byte nnet
size) = 5g nnet nodes
My thought: Humans don't (very often) produce even ONE horsepower.
Every attempt at general AI is using literally tens of thousands of horse power - and still without success.
I *think* I will need less than 50 four-core systems to do what I want. More than one horsepower, but less than
a dedicated nuclear power plant. Just a LITTLE more than my solar panels will produce, maybe.