Artificial intelligence and two approaches to it
Artificial intelligence is the term for making computers become smart or less dumb would probably be closer to the point.
In the most abstract way you can probably use the Turing test to define what artificial intelligence is. The Turing test is a simple conversation where you can not decide if you are talking to a computer or to another human being. But that is on a higher level than we are at right now. What artificial intelligence is today is a way to program a computer to learn by itself instead of being told what to do in every single step.
In basic we have two approaches to do this at present day, and both approaches are influence by nature. One is based on the brain neurons and the other is based on genetics, go natural evolution.
So if we start with the neuron approach that is called Artificial neural network. We simulate a neuron by doing something simple. We create a simple function that accept all incoming values, we put a threshold value that either fires or does not fire. Fire in this case is either send something to the next neuron or not. We then connect a number of neurons in one layer, connect all those neurons to the next layer and finally have some neurons that gives out a result.
We then take old cases of the problem we want to teach the network, run it through the network. If you get the wrong result you correct the network a small way towards the right answer and run the examples again. By doing this we can, if we are lucky, teach the system to predict outcomes. For example we teach it to the multiplication system, but we don’t teach it all of the results. With a well balanced system it will learn the result for untrained examples. I have tried this myself and successfully trained a network to learn multiplication and even predict (somewhat) winning horses in a horse race.
Genetic algorithm are another approach to solve similar problems. With genetic algorithms we encode the problem as a genetic code. We then create a a population of genes, totally random ones, and then we do some sexy genetic things. We mate the genes, take half the genes from one, and half the genes from another and combine them. We do some mutations, switch a 0 to a 1 and vice verse. We then check what genetics works best for our problem and say in a very natural selection kind of way, you can mate on. We then keep running the problem through the population and in the end we will have a super human, well super algorithm to solve our problem.
Both can work, but the problem is there is no way to actually see what will or will not work. How many neurons should you use, how many layers. In genetic algorithms, how big population do we need, what values are needed to be encoded into gene to solve the problem. Here we are at best guesses with both approaches. They are like black boxes, they work but we can not really mathematical prove what each neurons should do to solve the problem. However they do work, in the best cases, and that in itself is a very cool thing.
What was blitzkrieg during world war 2
First lets get the terms right, blitzkrieg was actually a term, as far as I know, used in some newspapers, among them a British newspaper (1939) regarding the tactics of the German army. The German term is “Bewegungskrieg“ or war of movement combined with a “Kesselschlacht“, the battle of encirclement.
One can go very far back in history to find the operational reason for this kind of war, but for this article i will only go back until world war 1.
But even before that, what is a war of movement in a world war 2 setting.
The first tactical decision is to concentrate your forces in one point to swamp the defending forces, create a breakthrough at one or a few single points. The second part is to feed quick moving forces, tank divisions or motorized division, through this breakout point and harass and disrupt the enemies rear. They would cut supply lines and communications to the front units and so impair their fighting ability. The last step, the “Kesselschlacht“, was to encircle and eliminate the surrounded forces, usually with infantry.
So what spawned these tactics and why were they developed in the first place?
Germany could not fight a prolonged war, specially against France/England and Russia and win a war of attrition. They did not have the manpower or the industrial capacity to win such a war. That had been proved in world war 1 where the imperial German army tried offensive on mass and the losses were severe with little or no gain. The one tactic that came out of world war 1, and gave a foundation for the “Bewegungskrieg“, was the stormtroopers. They were heavily armed soldiers that bypassed the enemy strong points and by doing so managed to create a deep penetration of the front. Being infantry on foot they still had great limitations in the depth of penetration that they could manage, but it was a tactic that proved highly successful, even if they failed in the end.
In world war 2 they created the panzer division to do the part of the stormtroopers. Combining this with air force cooperation they managed to do some very deep penetrations, managing the “Kesselschlacht“, especially in the early stages of the war. This doctrine does have some sever limitations. In Russia, the logistic worked against it. You could preform big encirclements but as the Soviet union managed to create new divisions, trade space for time and so the big encirclements that worked on a tactical level, actually did not work on the strategic level.
The other way to combat “Bewegungskrieg“ is defense in depth. The most striking example of this must be the battle of Kursk. The Germans wanted to encircle a Russian enclave, but as they were forewarned about this operation, they created defense in depth.
Defense in depth is to slow down the offensive spearheads until they loose momentum, creating high causalities and hindering the “Kesselschlacht“, in fact creating a war of attrition, one that Germany could not hope to win.
This was a brief introduction to blitzkrieg during world war 2, and its limitations.