By Brian Hill, One-Man Think-Tank and Resident Pseudo-Scientist

Let me begin this little rambling with a disclaimer: I am not a scientist; I don’t even play one on TV. This entire article is based on my knowledge of AI research approaches, including reading of mainstream scientific publications (Discover, Scientific American), web articles on the subject, and scientific television programs (Discovery Channel, TLC, PBS). It’s a topic that interests me, so I have done some research, but since I am not a scientist, I am not privy to some of the inner workings of the latest research. If someone more learned than me wishes to comment on this article, by all means, post it.

Now, the basic premise of this rambling is that all current AI approaches are doomed to failure. This is due to one primary problem with the approaches: They are too damn complex, and focused on the wrong process. All current AI approaches (with the possible exception of Neural Network approaches) are focused on writing code to emulate the cognitive functions of the human brain. This is idiotic, and doomed to failure.

First, if you are going to try to reverse engineer something, you have to know the thing you are attempting to emulate pretty damn well. At present, we know very little about the functionality of the mind. The basics, as always, are easy. We understand impulses, and chemical signaling, and all of the other basic functions. Our problem is, we don’t understand how everything fits as a cohesive whole. There isn’t a scientist in the world that can tell you what all of the exact, intricate details occur inside the brain to cause the formulation of a musical inspiration, for example. If you don’t understand something fully, you can’t possibly emulate it.

Second, trying to emulate the human thought process is a down-right idiotic way of going about AI design. Computers are not people, and never will be. The “hardware” is completely different. Our minds are analog devices, able to sense an infinite number of very slight changes and respond to those changes. Computers are digital devices. They have a finite number of possible combinations. While that number may be very high, it is still finite. This means that our method of formulating thoughts will never perfectly translate into a computer, at least not until we design a true analog computing platform.

Third, the approach is fatally flawed. True AI will be achieved not by trying to figure out how thinking “works”, but by creating a proper framework which ALLOWS thinking to occur. Think about it like this: There are no hard and fast rules for how you think. You think in a different manner than every other human on this planet. The pathways in your brain which are responsible for your thinking process were formulated by you over years of experience. They simply formed. There wasn’t an “if/then” statement that caused them to form, they did so on their own. For another example, imagine a newborn child. They emerge from the womb with a VERY limited amount of knowledge. Their instincts are all they have to guide them, and their instincts urge them to explore and interact with everything. Over time, this interaction begins to build a data bank of information related to their environment. Connections between this information form on their own inside the brain. These connections are what are responsible for the ability to form complex thoughts.

But the point is, trying to emulate the thinking process of a grown human being is the wrong approach, and needlessly complex. We should instead be trying to create an environment within the machine which allows the machine to formulate it’s own relationships between data, rather than attempting to create a framework that defines the relationships between data. Personally, I think the answer to making this happen is something simple and probably not directly related. I think part of the answer lies in building an analog computing platform, as I don’t believe a digital platform will ever “think” in a manner that we would recognize. The rest of the answer lies in discovering the prerequisite formula that allows connections to be formed in the mind. With those two things, I think we will see true AI, and without them, I believe that AI research is forever doomed.