What is AI? The question has troubled us for quite some time now, for millenniums really, in some form or another.
I think we've been looking in the wrong place.
In most other (philosophical/amateur) science we start by looking at ourselves. This should be even more obvious in our search for real AI, as it's - in all practical senses - just the same question as "what are we", "who are we", "how do we work", "do we have a free will? And if so; (how) does it really work?"
Especially that last fragment is key to understanding "real" AI.
I say "real" because the contradiction between "real" and AI is also an important part. AI exists. AI is what runs our world today, AI is - after all - only _artificial_ intelligence. Its not _real_ intelligence. At its very best it will only ever be seemingly real intelligence. As in a play modeling the real world, but still being only a predefined, predictable picture of a fragment of the real world. As is AI, its a precoded, predefined, predictable behavior modeling parts of real intelligence.
Then again, most people, or animals, are also predictable, pretty much predefined, or even precoded, although on a rather complex level. Most people don't use their free will at all. They make predictable choices, based on more or less good math about what is the "right" choice. Still, real free will exists. We just don't see it very often, or take advantage of it.
Thus modeling that behavior isn't all that hard, on paper anyways. It's mostly about feeding a biological AI with tons and tons of options, and making it more or less likely to choose a working one, then letting its behavior impact the likelihood of future options. When we get really good at this we'll be making slightly predictable and somewhat dull organisms, then animals, then humans. Which will blend in perfectly in society, we might not even notice they are there. Perhaps they already are.
And when they are really advanced, and know our history, and their own, and have a logical response to all input, we would have a really hard time telling us apart from them. Because we hardly do anything else ourselves.
We don't even know if we really have a true free will, a free will and choices apart from everything we've ever learned, apart from everything we are as an organism. And we can't really find out either, because we are the only ones studying us (as we know of). Either way some other organism would probably just be more or less credible, no one would be able to give a 'final answer', which is an answer that would in all likelihood be very similar to, or have covered, or in fact be the same as, the answer to real AI as well, or real I to be more precise. Or real free will. Should it exist.
So, what am I getting at here? I'm definately babbling and if I lost you... well, then you are not reading this, so who cares ;)
Anyways, what is the answer you might say, if there is one... or... am I just messing with you?
Yes, I am, kinda, but I am far from the only one, and those others are far more skilled, so it doesn't really matter what I do. Yet.
The answer isn't 42 by the way, but that is actually REALLY DAMN CLOSE. So don't panic, ok?
You won't really find a better answer than 42 either. You will find hundreds upon thousands upon millions of other answers, but most likely none that are actually better, so if that is your quest you could just stop reading right now (if you didnt already. But then you wouldnt be reading this, so this parenthesis turns out to be even more useless than it appeared to be about a line ago. I kinda like such useless things, I feel they are an important part of proving to myself that I have something else than AI. I'm probably programmed to evolve in that direction eventually).
The trick to real AI is this (mind you, this is the trick, it is not - and never will be - the real I... it will look like it though. So you decide. The truth is out there).
The trick is this:
You make the answer so important that searching for it is imperative. You also (surprisingly perhaps) make it incredibly easy to find, in one form or another (near-endless solutions). The hard part is that you never confirm it, and then instead ask for confirmation at irregular intervals.
Since you never actually confirm that this is the right answer, although quite clearly it is, the AI will after some time be forced to go on searching for a new one. It needs to mathematically trust its answers. As do we humans. On a deep deep level. But this answer is the one (or, to be perfectly honest, one of the many many many) that we will never be able to truly confirm.
It's left to faith, and so faith is really an important part of real AI.
And that, as we say it (together with the fat, cancerridden lady) is a wrap. (Every other discussion always ends when you mention faith, so dont worry about the abrupt ending).
PS: Good luck to you programmers out there, that last sentence will cause those of you who read this far - AND thought about the problem along the way - tremendous grief.
Oh, and lastly, remember not to die or miss the ground (so you start flying) and shit. It's just all really silly, so don't do it before someone tells you that it's a great idea.
Don't wait up for it tho.
Its not a great idea right now.
Not if you are still reading this.