There are a number of theories as to what autism is. These range from Cohen's extreme male brain theory to neuroanatomical theories such as those involving the mirror neuron system and cognitive theories such as weak central coherence theory.
About the only thing we can really agree on is that no one knows for sure what the cause or causes of autism is or are.
I have a number of ideas of my own. I hope to one day be able to investigate them. These range from one that's connected to a meditation technique properly called mushin no shin to another, more basic, hypothesis.
My actual reasoning for this is connected to something called neural networking theory. As the human brain can be described as a neural network, this has considerable relevance. Unfortunately, my reasoning also involves fractal algorythms, which are something that I've found that the English language is fairly poor at expressing.
No, explaining the concepts is actually fairly easy. The problem is that my original reasoning was framed in terms of them. English kinda sucks at describing fractal patterns, especially when it comes to following a chain of reasoning that's framed in terms of them. While it's not that difficult with simpler fractals like the Sierpinski Triangle or the Koch snowflake, I actually think in a form of what can be called semi-fractal, recursive pseudocode. Were I to try to explain my actual thoughts on the matter in English, I'd quickly get bogged down in the details.
This is hardly a new experience for me.
That said, the hypothesis itself is actually fairly elegant. While it's possible to view mental retardation as a defecit in the ability to learn (i.e. acquire new information, new mental circuitry, and/or new schemata), I suspect that autism may be related to a defecit in the ability to unlearn (i.e. the ability to forget existing information, revamp existing mental circuitry, and/or engage in accomodation for matters in which there are existing cognitive schemata).
On the flip side, this could also be phrased as a statement that autism is simply an overabundance of the ability to retain existing skills and remember things, which would help explain autism's suspected connection to genius.
There is, of course, a great deal more to this theory, but I'm saving it for a potential future research paper.