In Artificial Intelligence studies, your studies are not complete until you understand the Evolution Heuristic.
A Heuristic is a strategy for approaching a problem. Unlike a logical statement which must be either true or false, a Heuristic is neither true nor false but instead an approach that works or doesn't work in a specific situation. What works in one situation may not work in another with different parameters.
The Parameters for Evolution are:
1. It must operate on a population rather than an individual
2. There must be a source of change repopulation and recombination
3. There must be a test of fitness applied at the individual level
4. unfit individuals must be eliminated from the population in some manner.
3 and 4 when applied together form the "Test to extinction" hypotheis of the meaning of survival.
It is obvious that most people do not understand evolution so they often talk as if individuals were evolved.
It is also obvious that people mistake the guidance of the heuristic to come from the source of change, rather than from the test of fitness.
To consider why this might be the wrong approach, imagine a statistical scatter graph of a random population based on any parametric property. Now mark a location as the top of the graph, and turn the graph so that location is at the top of the page. Now draw a line through the graph from top to bottom on the right hand side. This is the population that has been selected as unfit.
If you recombine the remaining population some of the new population will fall over the line an be considered unfit, over time, the new populations will gradually lose the ability to recombine to form variations that fall over the line. If we look at the new resulting scatter graph, we will see that the graph has moved to the left. It is however just as scattered, it just no longer contains the variations that used to fall over the line.
Randomness is direction neutral, but fitness is not.
If you understand that, then you will see that Evolution is very similar to a hill climbing heuristic, or a controlled descent heuristic the primary difference being that it is based on survival.
Survival however is not easily understood as a fitness measure, unless you understand Entropy.
The Second Law of Thermodynamics states that nothing happens unless it makes the Universe a little more entropic. A corrollary is that the more entropy is promoted by an event the more likely that event is to happen.
Organisms exist because they increase the entropy of the world significantly while reducing the entropy internal to themselves. One way of looking at it, is organisms export entropy. So do societies, if you want to see an example of this all you have to do is look at E-waste. People here in civilization all want a fast computer, as a result computers are constantly being replaced with faster ones, causing the old computers to become obsolete. Here we see it as buying a new computer and recycling our old one.
In china however, what is seen is large shipments of E-Waste that are stripped in primitive conditions in order to recover the valuable materials. This creates pollution, which makes china's industries unsuitable for making toys for our populations. In other words our tidy little management of E-waste here in first world countries, promotes pollution in second world countries and third world countries. This is the entropic cost of being a first world country.
One of the problems I have with the Neural Networks Discipline is that sometime in the 70's or 80's people started to hype it all out of proportion. It almost died, except for two things one group that doggedly tried to model the neuroscience, and another group that tried to use the principles of distributed processing without hewing too closely to the neurology of the brain. Unfortunately it is not obvious in reading books from that era, which group was which. As a result, many researchers were warned against doing exactly what we need to do now to progress beyond the basic neural network, they were told not to try to model heterogeneous groups of neurons.
Hmm... Well, while the majority of the population is not threatened, there is still a trend for some portion of the population to die before they're forty years old. Further, it is interesting to note a higher level of infant and child mortality in the poorer sections of the population. I think that this is an indication that Evolution still works. But when you ask where it is heading, you may be surprised to find that intellect is not a factor it finds important. Nor is race. As I have said elsewhere "Survival of the Fittest" is not how evolution works or has ever worked, it is a misreading of the nature of natural selection. Think more along the lines of Survival of the Sufficiently fit, and you will find that as long as most of the population is not threatened by the environment, Evolution does not have to go anywhere at all.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence