For quite a while now, I have been intrigued with the problems surrounding Artificial Intelligence. I think what initially fed my interest in this field was probably reading Snow Crash and Cryptonomicon by Neil Stephenson. But either way, over the last fifty or sixty years, a lot of time, effort, research, and development has gone into trying to create Artificial Intelligence. We have gotten to the point, where we have structures like Neural Networks and various data structures (trees, hashes, linked lists, etc) which have all found some sort of use in this area of theoretical computing. We also have tests like the Turing Test to determine whether and A.I. is truly intelligent.

This is where, I think, we're flawed in our research. Several weeks ago, I was talking with Andrew in IRC, and was mentioning my thoughts in pursuing adaptive A.I.. Apparently, I spurred some thoughts in his head, which then proceeded to consume him. And then, a few days ago, we caught up again and had another lengthy discussion on actually building, not an artificial intelligence, but rather a synthetic intelligence. Since then, I've done some poking around online, and I'm not seeing a whole lot in this direction. Wikipedia has a lot of information on Articificial Intelligence, which is all fine and dandy, and has a lot of important ideas and theory that we can't overlook in developing intelligence. However, although there is an article on Synthetic Intelligence, there isn't much there, and it's contents are nearly bare. So, let me try and explain this all (and I warn you, I may fail horribly, but I need to try).

Artificial Intelligence is somewhat of a misnomer. Lets start by breaking it down into bare definitions. (From Answers.com, extra definitions not applicable to the topic have been removed to preserve space)

artificial

1. Made by humans; produced rather than natural.
2. Made in imitation of something natural; simulated: artificial teeth.
3. Not genuine or natural: an artificial smile.

intelligence

1. The capacity to acquire and apply knowledge.
2. The faculty of thought and reason.
3. Superior powers of mind. See synonyms at mind.

artificial intelligence

1. The ability of a computer or other machine to perform those activities that are normally thought to require intelligence.
2. The branch of computer science concerned with the development of machines having this ability.

In other words, artificial intelligence is faked. This may not seem like much of a problem at first. But how is a faked intelligence going to develop a conscience if there is no real intellect. Now, synthetic and artificial are synonyms, but that doesn't necessarily they are identical terms. In the English language, when we think of artificial, we think fake, or man-made. Synthetic, typically produces similar thoughts, however there is a key difference. Where an artificial item is percieved to be identical on the outside to the real object, its construct can be wildly different. (Like, say, artificial sugar, it may taste similar, but the chemical compositions are no where near identical). Now, we look at a synthetic item, and thought it may not be natural, the chemical composition will be nearly identical. Obviously, there are differences many times, but the variance is negligible.

In other words, artificial intelligence as we think of it, is fundamentally flawed. Much of what I've seen so far seems to use structures like neural networks, markov chains, and other structures and methods that simply manipulate data. Not much as far as self-referencing and modifying logic. I'm sure work has been doing in self-morphing and self-referencing code, but I haven't seen much stuff that's been released and actually worthwhile.

There is the Cyc and OpenCyc, which, I haven't looked too deeply at, but after looking over it briefly now, shares some similar ideas to what I'm thinking of now.

Basically, in order to create an artificial, or rather, a synthetic intelligence, we need to analyze what exactly defines intelligence, and then try and break that down into a finite number of features. At that point, it's simple a matter of connecting the pieces together and feed it input.

Andrew has come up with, what we think, are four basic areas, and have a design that we're slowly getting put together. I'll try and post some about that later, but I still need more time to wrap my head around it all.

Over fifty years of research and development has gone on in the area of artificial intelligence, and we have made a lot of head way in that amount of time. But what many have that to be improbable, and what many people fear to happen, may soon be within our reach. Computing power and storage has become more and more powerful and cheap that creating intelligent and conscience machines are nearly within our grasp.

The future may not be here yet, but it is inching ever closer, and an exceedingly rapid rate.

--nullpuppy(out)