JuxaLOTmoreChocolate, that's a bloomin' good question, which is why many of the experts have given up even trying to classify people fully at the moment. There's an international 'think group' working towards a new set of diagnostic criteria for DSM V in the USA, but SO much more research needs to happen before they can report. I think the reporting date is being pushed back further and further at the moment.
A child who is classically autistic, highly intelligent, non-verbal at 8yo, no eye contact, no pretend play, lots of repetition? To their embarrassment, many of the researchers are finding that supposedly "low IQ" individuals with autism have actually got very high specialist IQs (rather than general common-sense IQs), but since our behaviour is SO different to that of other people and our communication with people is often irrelevant to us, no-one knew it. They thought all the repetition was just some sort of motor problem or random brain activity.
When they started to really analyse it, they realised there are subtle patterns to it - often too subtle for anyone else to see unless they apply computer technology etc.
An example for you: As a child, I spent probably 300 hours picking grass apart to look at the insides of the seed bit. Pick up a piece of grass, open it, look at contents. Pick up a piece of grass, open it, look at contents. Repeat that thousand upon thousand of times. You'd think I was doing nothing apart from obsessing, yes? Nope, I was analysing the different genetic possibilities for that particular type of grass. How did it grow, how often did it develop differently? I just had no reason to talk to anyone else about it. It was my grass, my research, my interest. It's not my fault if other people thought it was just stupid random repetition.
Simon Baron-Cohen talks in recent papers about profoundly autistic individuals who, for example, shake a piece of string. He's realised that in some cases they're studying the physics of it - how fast do you have to shake it, what happens when you do, how do different lighting conditions affect the visual results. It's caused him to rethink what he thought he knew about autism. Someone looked at a child who had drawn hundred upon hundred of the same picture over and over and over again. They didn't see the pattern changes. A computer did. It wasn't the same picture - there were subtle predictable variations happening. They were apparently testing how far they could alter the picture for it to be the same picture.
Is this true of every autistic individual? No, it isn't. Some really are profoundly learning-disabled. But many are highly intelligent but have absolutely no interest in interacting with you or playing with you. None at all. We're extreme specialists. So extreme that most people have NO idea what we're doing, or why.
As for the whole "mild" and "severe" thing, don't get me started. I'm in agreement with many of the experts who say that we have no idea how to define it right now, and we really shouldn't try until we've got more research done. I'm minded of a chap who told me that no way could I have any idea at all what REAL autism was like as Asperger syndrome was so mild that it was irrelevant. Scarily, he was running a major parent's help board for autism and Asperger syndrome. He's a wiser man, now. You can have mild autism and mild Asperger syndrome. You can have severe autism and severe Asperger syndrome. There is no clear mild-severe boundary in either case.