Wednesday, August 19, 2009

Analog Machines

For many years I used to imagine building an analog computer. Unlike the binary system we are so committed to nowadays I considered the issues of ranges of values other than 1 and 0. Now, those numbers can just be arbitrary smooth fractions between 1 and 0, or just indefinite numbers of unsigned integers, either of which have however many digits they need. The computer would be made "somehow", even if only of "unobtainium." A value could be a voltage, or a frequency, or a shape of a waveform. Some kind of signal.

In an analog computer the instructions would be made of signals. The signals are entirely analog. There are only as many "pieces" of the signal as is deemed necessary by the receiver of the signal. For instance an "ear" would be sensitive to the low end of the dynamic range. A radio would be sensitive to a large part of the middle ranges. There are also digital inputs that come from the outside and come from the inside. However they are really just analog signals shaped in such a way to approximate the square waves sufficiently to be interpreted as a digital signal. These could be as high or low of frequency needed, up to a certain practical limit (super gamma X ray frequency would not be safe for use in any kind of computing device.)

This computer must be capable of storing and retrieving signals. One such storage would be a light beam reflecting inside one of many tubes of mirrors. Electronically it could be voltage amplitudes bouncing between two identical feedback amplifiers. They only increase the output amplitude a fractional amount in order to make up for propagation entropy or attenuation. Time lapse would occur, however, such that the time cost equals one echo per connection. Such propagation effects already exist in current machines. The goal is to minimize time delay, however, so that reality can be "modeled" or interacted with in as realistic a manner as possible.

However the action is done, whether it be an optical means, using light beams and mirrors, etc. or if using cascades of transistors, capacitors and inductors, the sources of electrical signal would not be so much high speed all the time, just very accurately handled, regardless of frequency, whether many very very slow signals are used (allowing interim multiprocessing), or very tightly synchronized ultra high speed signals that test the fastest limits of the hardware.

For computational instance, if I wanted to get the square root of a number, I would put 25 volts in one hole of "the square rooter circuit" and the other hole would instantly output 5 volts. Only the reaction time of some transistor or other component is important, and they are generally quite fast, almost to the point of taking near 0 time. If a component was bypassed "to save time," there would still be some time used by electrical conductivity in the wire, however nearly 0 it was. Other mapping functions could be that perhaps X volts input equals Y(f(X)) volts output, or whatever. There are far more possible instances of specific electronic mathematics that can be integrated within the flexible fabric of electronic connectivity. There are too many to enumerate since they cover almost everything in reality and everything that has never happened yet.

The single input and single output is only a simplification, and there may be many inputs and many outputs for any given "thing which does something" in the machine.

You don't think of radio as being a "batch job" that takes all day to process, like baking pastries. If pastry baking was like radio they would take very little time or energy to bake. It takes little time for the input to a radio transmitter to be sent and received simultaneously by multiple radio receivers at long distances and at a vast number of locations. It would be as if as soon as you shoveled eggs, flour, sugar and milk into the input pipe, fully cooked cupcakes would instantly flutter from the output pipe in somebody else's living room. Computers made in such a way would be far faster than todays linear step-by-step digital computers. (I don't think cupcakes will ever be quite so convenient this way...)

As time goes on, various electronic component designs are made at in ever increasing numbers. There will always be far many more ideas than actual implementations because of the need for real wires and real components. Talk is cheap but pickles cost nickels. Yet, over time, these accumulated parts have been segmented, cataloged and stored in a vast, human maintained memory, such as in all the engineering file cabinets and databases in all the companies in the world that make such things.

There are fields within fields of connections possible between devices, and some are like ponds and rivers of undulating liquids with millions of separate sources and sinks for the voltages.

A huge amount of things would not even have names, such as nobody has a name for every grain of sand. A Z-transistor with a K-capacitor and an I-coil might have a name in a database, such as Part Number ZKI-123. A computer which manipulated or used such devices might even create them on the fly, only momentarily hooking up parts X, M and P for a microsecond, never to be used again, because the conditions necessary to use that combination never occurs again.

Flexible machines of this sort, which could synthesize parts on a need-be basis, could also evolve solutions to problems never encountered before. Much like human immune cells which can recognize things that are NOT correct in the body, a machine could recognize when nothing in its current configuration is suitable for whatever current problem occurs. It would counter the problem with a shaped electrical field and collection of molecules made ideally for the situation, just as water in gravity field will fill every nook and cranny of a cup it is poured into. The water does not "think" about the shape, it just automatically assumes that shape. Auto-pilot in airplanes has some similarity to this effect.

An analog computer could be made from mirrors, light emitters and light sensors, hopefully of very high reaction speed. A device made within spheres within spheres, with mirrored surfaces, can emit light signals on one side of the sphere and sensors on the far sides could pick up the reflections, with no wiring necessary. The "individuality" of signals would not be in discrete wires, but in discrete colors of the spectrum, as many as are needed all the way down to Planck's wavelength (theoretically). Also, signals could be serial and monochromatic and use only encoded "from-to" addressing to separate one signal from another, similar to how Ethernet works -- certainly not as fast as discrete, private spectral subdivisions could be. The parts which synthesize other physical stuff, such as unique molecules, are more chemical and nano-mechanical in nature, however the brains for such things would be in the analog computer itself.

Powering such devices might occur at the "heat" level. The heat radiating from the planet passes through everything on its way out -- sort of a photon wind. Those photons and the resulting molecular motion within crystals could then produce controlled oscillations which act as tiny "generators" and "motors" in our little machines. Of course there may be limits, and the entropy may be such that it actually damages machines instead of powering them! The implementation details are left to more qualified minds -- I am only thinking out loud. Perhaps if such things were possible, living systems would have already stumbled upon them.

Nevertheless, I can imagine, on some far distant world, after billions of years of biological evolution had run its course, that what we now consider "machines" would evolve in this way. What we consider nano-tech in our vernacular would be the DNA of their evolution. I can't imagine what the end result might be, nor whether it would be good or bad. I suppose there are many bad science fiction stories that could be made about such things. But, regardless of whether God created our own evolution, there is nothing, so far, to indicate that any other kind of evolution would thereby be prohibited.

No comments: