Tags

A challenge to the folks in Silicon Valley is to create a new type of computer which is similar to the way a brain works. What is needed to accomplish the effect, which normal functioning brains do, is a cell in which tiny transistor-like cells will store energy in a cumulative way but simultaneously in a continually draining away way too. Imagine a leaky water bucket with various sources of water coming in at variable but controllable rates and water leaking out at a known or controllable rate. The goal is to have the bucket like switch dump its specific load of energy only when the input flows exceed the output flows for some period of time. With electronic devices this could be a very short period.

Each of these input flows could be from variable buckets similar to the one described but at these previous input buckets there would be external input events which determine what these first level buckets are processing. The secondary buckets described at the top would only receive inputs from the buckets which had received external stimuli, but they would receive it in a discrete digital form from the input buckets. If there were only one level of input buckets, each bucket feeding a single second level bucket, it would change an analog gradual input into a discrete digital output. That wouldn’t be very useful because there are easier ways to do that by simply measuring the first level’s output. However, if the second level bucket was receiving input not only from the first level bucket but also from third level buckets, which themselves also got input from fourth level buckets, then this second level bucket’s tripping behavior would be affected some even by the fourth level.

What happens when there is a wide sheet of primary input buckets, overlaying a sheet of secondary buckets, and that overlaying a third sheet and that a fourth, a fifth and a sixth, is that the last sheet is able to make integrated decisions based on comparisons from the input from the entire sheet of first level buckets. Each sheet can make simple discriminations about what is going on in the sheets above it because it gets support in its tripping behavior from the sheets below it. After a few cycles of the information about the first level bucket’s input, the flowing to the bottom sixth layer bucket and back up would stabilize and the sixth layer could make statements about what the entire first layer was perceiving.

With traditional digital computers the energy entering a transistor is limited to within specific limits to reliably get the desired switching effect, but within this new leaky-bucket type switch the trip point might be precise but the inflow rate versus the outflow rate is what determines when the switching potential is reached and the tripping effect occurs, and that switching is controlled by several different inputs from the different levels.

This type of layered sheets of leaky buckets linked together would permit a kind of computing which is more similar to the way a brain works and would therefore be able to get some kinds of results more quickly than either a digital computer or a brain. It would be quicker than a digital computer because its decision trees are much shorter and it would be quicker than a brain because these would be strictly electronic effects rather than vastly slower chemically mediated effects.