Biology 103

2005 Final Paper

On Serendip

In order to answer the first two questions, it is necessary to explain a bit more information about the inner workings of CCortex™. First of all, though AD claims that CCortex™ is very similar in process and "organization" to the human cortex, it is completely dissimilar in terms of size. It is composed of 1,000 processors contained in 500 separate network elements. (2) Thus, even were the CCortex™ system to accurately emulate the human cortex, its application would be limited because it is extremely expensive and space-consuming to reproduce.



In terms of actual function, the CCortex™ system runs a personality emulation program that can "learn". The computer network is trained using a text-based chat-like interface, through which the network can respond to a number of questions. Based on the network's answers, the builders give the computer a reward based on the answer, and so the computer is taught using the classical conditioning method. (3)



Does all of this add up to a reasonable, or even desirable, reproduction of the workings of the human cortex? I argue that it does not, but also that it can not. Granted, this constitutes what, in all likelihood, could be the closest computer approximation of human thought and interaction. But what is 'close' in a case like this is very far away from an actual human brain. Even discounting the size issue, alluded to earlier, there seem to be a number of differences in terms of complexity that question the ultimate similarity between CCortex™ and the human brain.



The first of these is CCortex's™ manner of "learning" by this fairly basic Pavlovian response method. Classical conditioning is just one of many ways to learn for humans.(4) In addition, classical conditioning normally only works under certain circumstances and in certain ways, such as for unconscious processes. (5) CCortex™ has been programmed to only learn in a certain way, and, consequently, cannot learn in any way outside of its specifications. This is not the same as humans, who can learn in many different ways depending on what it is they're learning; classical conditioning cuts down what CCortex™ can possibly learn. Also, different humans learn better through varying methods. Some learn better by reinforced visual stimuli, some learn better by reinforced audio stimuli, etc. If another CCortex™ was built, unless it was specifically designed otherwise, it would learn by exactly the same method as the original. This is true no matter how many CCortex™ machines are built. Granted, the things that each machine "knows" will be different on a case-by-case basis, but their methods of learning will be identical. These are some things that ultimately separate CCortex™ from the human brain.



Even if CCortex™ could properly emulate human thinking and learning, I argue that this is not a desirable outcome. To start of, the field of psychology has been devoted to examining human intelligence and instinct for over a hundred years, so to a fairly large extent, this work is already being done, and CCortex™ may be mostly redundant for these purposes. One use for CCortex™ that is not inherent to the human brain already is the fact that it was built by humans. This could help us understand more exactly the nature of neural connections and the neural system in the brain. Even given this, however, our understanding of the brain through this fact is limited. AD is working with a simulation of neurons, and a simulation of neural connections, and so the building of CCortex™ can, in this vein, only help us to understand how to properly simulate the human neural system in a very specific way – through a computer program. There are most likely other possible simulations of the neural network that might also give us a firmer understanding of the brain, but these are not being explored here.



This entire discussion calls into question the issue of human learning and creativity. I have assumed that human thought and creativity is somehow individualized, and not merely a long and complicated response algorithm within the human brain. Could this be the case though? Humans all act somehow differently from one another, despite our general similarities. This difference can potentially be explained away, though, by noting the fact that every human deals with slightly different circumstances than all other humans. Given these different circumstances, is it possible that difference can be explained entirely by a person reacting to their unique environmental condition? I am tempted to disagree with this notion, if only because of the entire field of genetics which makes the very good point that environment interacts with potential differences (genes) within all of us to create the person we are. The question is, at its core, a good one to consider, though, and a worthwhile topic for further discussion.



One important aspect of CCortex™ that I have purposefully left out is AD's professed applications for this network. AD is not merely acting on a whim to try to absolutely emulate the human brain. Rather, they are viewing CCortex™ as a tool primarily for international business purposes, with other applications in the field of, obviously, artificial intelligence, as well as other psychological applications. They make no pretense of being able to duplicate a human's thought process, and it does not seem as if they are even interested in this at all. They are trying, rather, to create a more accurate computer simulation of the human brain in order to learn more about both artificial intelligence and humans simultaneously. (1) I used CCortex™ primarily as a jumping-off point for a discussion of artificial intelligence in general, and so many of my points are applicable to a wider range of cases than just this one.



1) The AD Information Page on CCortex™.

2) Roland Piquepaille's Blog, contains some pictures of CCortex™.

3) Ian Yorston's Blog, links to an AD press release.

4) Wikipedia's Article on "Learning".

5) Lynda Abbot, Ph.D.'s, Page on Behavioral Conditioning.



| Course Home | Serendip Home |