Christopher Goodman
April 11, 1997

Can machines think? Or rather, can we develop true artificial intelligence in the sense of machines that think and understand as we humans do? This is an interesting problem that is becoming more and more relevant in our lives as computers become more complex and integral to our lives. Two articles in our textbook, John Searle's "Minds, Brains, and Programs" and William Lycan's "Robots and Minds", present two different answers to this question and also raise several new questions. John Searle takes the position that on one level computers do think - they manipulate symbols - yet on another level they do not think - computers do not understand the symbols they are manipulating to mean anything in the sense that we humans do. Lycan takes the position that yes, computers do think, and that it is quite possibly only a matter of time before a machine can be created that not only looks and behaves like a person, but also thinks like a person. Therefore, Lycan claims, the suitably programmed machine of this complexity is a person as much as you and I are. I fall more on Lycan's side of the argument.

Words such as "intelligence" and "understanding" have variations in their definitions depending on whom you ask. It is often hard to come up with even a simple definition once one delves into the problems at hand. But, since we as humans (in particular Searle) often try to separate ourselves from computers by saying that we understand the meaning of the symbols we manipulate, it is necessary that I give a useful and accurate meaning to the words (or symbols) I will be using.

I define "thinking" as processing information, with any level of complexity. I include in the thinking category a thermostat making the decision to turn on the heat as a result of data stating that the temperature is too low. I include the human contemplating the enormous amount of information contained in his Philosophy book while trying to write a paper for class. Central to this definition is the idea of taking in data and making a decision (even if the decision is to do nothing). I realize that this definition of thinking is a bit broader than common usage, but I do not think that it is outside of acceptable limits. The act of decision making, and limiting the decision making to data, is what separates my definition from Searle's amusing idea of a thinking stomach. The stomach takes in matter and physically processes it in a certain set way. However, any changes in how the stomach treats this matter are the result of the brain sending out decisions it has made by processing the data the stomach has sent to it.

I define "understanding" as the more complex process of making associations between a symbol and other symbols, as well as between a symbol and the actual thing it represents. Therefore, for something to understand a symbol, it must have stored memory about the symbol, and be able to make associations between the symbol and other related things. Depth of understanding relates to the quantity and quality of stored references (as well as to quality of perception), and the complexity of the associations made with them. For example, the first time I heard the term "web page", I had no understanding of the term (term being synonymous with symbol). After seeing one on a computer on a TV news story, I had an idea of some type of document similar to a word processing document - not inaccurate, but still showing a lack of understanding. However, after coming to school, surfing the web myself, and listening to a few lectures briefly covering the topic, I developed a reasonably firm grasp as to what a "web page" was, though still not as deep of an understanding as my roommate who took a class on designing web pages. I had significant first-hand perception of web pages. I had been told by teachers and friends what one was. And most of this was stored in my memory for future reference, so I can easily recall different characteristics of web pages, as well as associate them with other related data.

Now, by taking these two definitions and applying them to normal, healthy humans, we see that humans are thinking, understanding beings. We also can apply these two definitions to even our current computers. The difference being only that computers are currently far behind humans in complexity of understanding and in depth of perception. We cannot deny that computers perceive - computers have keyboards, cameras, microphones, scanners, and many other ways of taking in data. One can say that a lot of this data is forced into computers, but this does not change the fact that computers perceive - humans are often forced to perceive things, too. Computers think - they manipulate the data they have taken in and make decisions based on it. Computers, although still on a simple level, understand. For an example, I can say to my computer, "Open Control Panels." The computer then takes these spoken words and translates them to the file names, directories, and commands that are part of its language. It translates "Open" to mean the action of taking something from a memory location and displaying it on the screen. It translates "Control Panels" to mean the name of a particular folder - the one that needs to be open. Thus, the computer understands the command, and the symbols that make up the command. Now, admittedly this is not a very deep understanding, but according to the given definition, it is understanding. The computer then takes this data, plus the data coming from the necessary association it has had to make to understand the command (knowing what the control panels folder is, where it is located, and whether or not it can be opened), and makes the decision to open the control panels folder. This shows that the computer perceives, understands, and thinks.

I doubt that anyone would say that my computer approaches anywhere near human intelligence, despite the fact that it understands more words than many dogs I know. But it does exhibit a lot of similarities with human intelligence. So, what is the difference? Memory capacity, quality of perception, complexity or depth of understanding, and degree of freedom in decision making are the areas where the computer is behind. In the first two of these four areas, simple improvements in hardware are all that is needed, though big enough improvements are still probably far in the future. The last two areas involve more complex programming, which is probably much further in the future. With these developments, though, I believe a true artificial intelligence, on par with humans, will eventually develop.

This raises several questions, and objections. One popular question that Lycan refers to is "Could a mere blood-less runner of programs have states that feel to it in any of the various dramatic ways in which our mental states feel to us?" More and more often people are not giving the resounding "NO!", and neither does Lyman. I feel that emotions are just a highly developed central part to our programming that lets us feel particular states. For instance, we feel fear because it is our own programmed way of identifying a state of danger. We feel pain because it is our own programmed way of identifying injury, emotional as well as physical. These states can be caused in many different ways (even sometimes without reason) because we are complex. We feel a wide variation of feelings because this system is highly developed. Once again, given time and effort, computers can conceivably develop a complex system of emotions, too.

The other questions of particular interest to me are: 1.Do these computers have rights? and 2.Where do we draw the line? Lycan spends several paragraphs on this topic, and I think it is one we need to consider. In answer to the first question, he says yes, for who are we to decide whether or not Harry (his example of a robot that is remarkably human in nature) and Henrietta (his example of a human who's body is almost entirely mechanical) are people? If we cannot distinguish their minds from regular humans', then we have no basis for denying that they have the same basic rights. As for the second question, that is much tougher, and he makes the analogy to animal rights, for even now computers exhibit intelligence that is roughly equal to many animals. However, since I have proved my main point, and run over my page limit, I will now bow out. The area of animal rights is still hotly debated, and we still often disagree on many aspects of human rights. These areas are the subject of many papers, articles, news stories, even organizations. Maybe, if only to prevent a repeat of our confusion morally over our sudden ability to clone complex biological organisms, we should start looking now at the issue of "computer rights". It still sounds strange to me.