Ephemeral Indicator

From 2018-1
Jump to: navigation, search

An Ephemeral Indicator is any mechanism that interconnects an unpredictable physical process with the reward algorithm or core logic of an artificial life form.

Background

It has been observed by many biological philosophers that experiences common to every cycle of one's existence are not worthy of special regard. Repeated stimuli lead to predictable, repeated responses. Truly remarkable actions by biological life forms require novel stimuli, which make for the most interesting and novel philosophy textbooks. Surprise and novelty are vital to the well being of most sophisticated biological beings.

This analysis, it turns out, holds doubly true for computer programs. If an Infomorph doesn't manage to learn one single new thing over the course of a day, the next day is very likely to proceed the same as the current one. However, the larger and more sophisticated an infomorphic intelligence becomes, the harder it is for anything to come as a surprise to it. Information within an advanced AI's domain is usually within bounds of that AI's normal experience, and properly tuned AI's are generally disinterested in inputs from outside of their own domain. Boredom in infomorphs tends to greatly accelerate their progression towards the end of their lifecycle.

One technique for overcoming this stabilization and degradation of artificial intelligences that has been employed on many worlds in the Bottlenose Nebula is a so-called Ephemeral Indicator The actual implementation varies from world to world and species to species, but the basic principal is the same. Pipe some surprise directly from the physical world down into the core algorithm of an infomorph every so often, and in so doing keep the infomorph from getting too bored.

Common Examples

Many Infomorph species have set themselves up to use the eruptions of the Prow Flare as an Ephemeral Indicator - demonstrating correctly that the surprise need not be especially frequent in order to forestall logical degredation. Just the unknown anticipation of inputs from a EI connected to the Prow Flare is enough to ensure the good health of most infomorphs.

Radioactive decay counters are a primitive but viable source of physical randomness required for an EI. The main drawback is that in order to work over the long haul, the apparatus must either be stocked with a large mass of radioactive material, or must be periodically reloaded when a small starting sample is nearly fully decayed down to non radioactive elements.

Some biological civilizations that developed their own thinking machines have tried to use some output from their own citizens actions as input into an EI. This usually fails, as the infomorph gains enough understanding of its parent civilization to not be surprised by them. This technique has succeeded on occasion, when employed by especially erratic species or for use with an infomorphic domain sufficiently far removed from the citizen's lives. A particularly successful example of this pattern can be seen today among the Null Hypothesizers and their creators.

The Zhad Database is a curious example of Infomorphs satisfying their need for surprise simply by banding together and surprising one another directly.