THE POLYTHENE AND pins memory-surface processed the information that was put onto it. All this processing involved was that the water did not stay where it fell but flowed over the contours of the surface to find its own level. Apart from flowing, it also altered the contours of that surface. This shift in the position of the water means a shift in the memory pattern; a shift in the memory pattern means a change in the memory pattern; a change in the memory pattern means a processing of the information. It is as if one tried to take a photograph of a camel, but each time the pattern on the photographic plate shifted to give a picture of the pyramids. The plate would be processing the camel pattern to change it into the pyramids pattern.
The fundamental activity involved is flow. Flow means that the water moves from one area to another area. Water tends to flow downhill. This is not so easy to describe in functional terms, apart from saying that one area is preferable to another. There is a natural direction of flow, and this is determined by the contours of the surface which make some areas preferable to others.
One can now try and translate this flow principle and the general behaviour of the polythene and pins memory-surface to a new model.
There are advertising displays that consist of a surface which is made up of hundreds of separate bulbs. Different pictures can be shown on this surface by lighting up different patterns of bulbs. Pictures can also be made to flow across the surface by lighting up new bulbs and turning off lighted ones.
The new model memory-surface is made up in a similar manner. Instead of the surface being a homogenous sheet like the polythene sheet it is made up from a huge number of separate points, and each point is a bulb. Each bulb can either be on or off. As with the advertising display, the pattern on the memory-surface is given by the arrangement of lighted bulbs.
A bright pattern falling on this memory-surface would light up those bulbs on which the pattern fell. If a square pattern were put onto the surface then a square pattern of bulbs would light up. This would be arranged by having each bulb operated by a very special switch.
There are street lights which automatically turn on at dusk because the level of daylight has fallen below a certain level. The special switches in the model would work the other way round. When the light falling on the switch exceeded a certain level then the bulb would go on. This would apply to light from any source except the bulb itself (the switch would be shielded from this).
The distribution of light in this model would be equivalent to the distribution of water in the polythene and pins model. But what about the all-important flow behaviour? To understand this, one has to go back to the tilting board on which were placed the columns of children’s blocks. As the board was tilted, the taller columns toppled over. The ones that toppled over were made taller still and that made them still easier to topple over. The degree of tilt required to topple a column over could be called the toppling point. Nothing would happen until the degree of tilt reached the toppling point, but when it had then the column would take off on its own and fall over. The interesting point here is that the column would not just gradually lean over until it was lying flat but that at a certain point it would take off on its own and suddenly topple right over. In other words it needed prodding up to a point but then it would go over on its own. This toppling point could be called a threshold for action. Below this threshold nothing would happen. Once it was reached things would take off on their own.
As the columns got taller and taller the toppling point would be reached sooner and sooner. The threshold for action would be reached sooner and sooner. The threshold could be said to get lower and lower.
We can now go back to the bulbs with their special switches. These switches would behave like the block-columns. Once their threshold was reached they would switch on by themselves. Each time they switched on, the threshold for switching would get lower (just as the columns became easier to topple each time they were toppled).
This difference in ease of switching depending on what had happened in the past would be equivalent to the contours in the polythene and pins model. Those switches which were hard to activate (high thresholds) would correspond to the peaks on the polythene sheet, while those which were easy to activate (low thresholds) would correspond to the valleys.
Whenever there was a pattern of light on the memory-surface there would be an edge to it. At this edge there would be lighted bulbs next to bulbs which were unlit. The light-sensitive switches on the unlit bulbs would respond to the already lit bulbs and would tend to light up their own bulbs. Thus the pattern would tend to spread at the edges just as the water tended to flow in the polythene and pins model. Those switches with low thresholds would light up more easily than those with high thresholds, and so the pattern would tend to spread towards areas which had been activated before, just as the water in the other model tended to flow downhill towards areas which had been used before.
Whether an individual light bulb was lighted or not would depend on two things. First, whether the pattern actually fell across it. Second, whether the pattern spread to it. And the ease with which the pattern spread would depend on how often that particular bulb had been lighted previously.
At first sight such a model would seem to give a fairly good imitation of the polythene and pins surface. The reason for having this additional model is that its behaviour is much more like the organization of the nerves in the brain. The nerve network consists of a collection of individual switches which may be either on or off. Each switch has a threshold for activation and once this is reached the switch flips over by itself.
Unfortunately there is one huge difference between the polythene and pins model and the light bulb model. In the polythene and pins model there is a limited quantity of water. This water cannot be in two places at once. So if the water flows towards one area it must flow away from another area. This gives genuine pattern-shifting from one area to another according to the contours of the surface. This genuine shift of pattern gives true information-processing. Things are different in the light bulb model.
In the light bulb model the edges of the pattern will spread. New bulbs will light up and the pattern will be extended. The vital difference, however, is that though new bulbs light up, the bulbs already alight remain lighted. Thus there is no genuine shift of pattern from one area to another. The pattern just gets bigger and more blurred around the edges, but does not actually shift. It is true that the extension of the pattern will be towards the light bulbs with low thresholds, but this does not improve things if the already-lit bulbs with high thresholds remain lighted. It is as if in the polythene model the water filled the valleys but remained on the peaks as well.
As things stand, what one would get from a simple pattern put onto the light bulb surface would be a much enlarged and blurred pattern. Perhaps the whole surface would light up eventually. Unless there is genuine shift, and not just spread, then there is no useful information-processing but just a mess.
The difference between the two models is that in the polythene model it is the water itself which moves across the surface, and there is a limited amount of water. In the light bulb model the input just triggers off the light switching and there is no limit to this.
If the number of lights that were lit at any one moment were strictly limited, then the light bulb model would better resemble the polythene one and would be a useful system. In fact this can be arranged quite neatly by making use of a circular systems effect.