Home >> History, Ideology & Science >> Globalization

     Email   Print 

Intelligence Emulation

Bart Lasater - 9/9/2007

Since the advent of the programmable machine, there have been philosophical debates regarding the concept of machine intelligence as it relates to human intelligence. Among the questions that arise are the following: Can a machine understand a language as we do? Can it formulate decisions as we do? Does the presence of a program written by an intelligent person mean that the machine possesses intelligence?

Unfortunately, such debates frequently deteriorate into arguments about the definitions of words such as understand, intelligence, and what it actually means to think. Like most students of computer science, I’ve developed some strong opinions regarding the intelligence capability of programmable machines, and I realize that there are different schools of thought on the subject. While reading the text of an author of one school remarking about the ideas of another school, I realized how sensitive this subject can be. The following are simply my own conclusions and opinions - comparatively explained - based on my studies.

The Chinese Room Argument (a good place to begin)

One of the most widely debated (and hotly divided) arguments regarding machine thought and intelligence is a concept put forth by American philosopher John Searle regarding the difference between the syntactical processing (symbol manipulation) of a computer program and the semantic understanding of the human mind. This concept has come to be known as the Chinese Room Argument (CRA from this point onward). The CRA is a refute of (among other things) the idea that a program can understand words and symbols much as our minds understand them, and that human thought and intelligence is not simply computation. In summary:

A person who does not understand the Chinese language is placed in a room with many Chinese symbols (a database) and a rule book (program) for how to process any incoming questions. These incoming questions are in the form of other Chinese symbols that are fed into the room through a slot in the door. When this happens, the person looks in the rule book for the steps on how to process the incoming symbols. After following the rules correctly, the person passes back the answer to the questions in the form of Chinese symbols. What the person is doing is acting as the processor implementing a program to answer questions in Chinese (Searle 11).

Though Dr. Searle offers several axioms related to his argument which aren’t discussed in this article, the essence of the CRA is this: A program follows a syntactical (not semantic) structure for processing a task. These syntactical rules guide the programmable machine in its tasks, and the machine need not have a semantic understanding of the symbols with which it is working. Much like the man who did not understand Chinese, it simply worked with symbols and followed the rules of processing, while a person who actually could understand Chinese need not refer to such rules, and could likely make due with grammatical errors. You may remember reading something that was written by a terrible speller, but were still able to make it out since the semantics of the language were enough to help you understand what the poor speller was trying to convey.

The CRA is an excellent place to begin a debate about the intelligence capabilities of programmable machines. When I first read about the CRA, the argument seemed well-formed and logical, but like all arguments it has its share of detractors who also have well-formed arguments against some of the CRA’s finer points.

For the sake of argument, let’s assume that the program (instruction book) can effectively pass the Turing Test. The Turing Test is a test of human imitation named for computer scientist Alan Turing (1912 – 1954). In summary:

There is a person, a machine, and an interrogator. The interrogator is in a room separated from the other person and the machine. The object is for the interrogator to determine which of the other two is the person, and which is the machine. The interrogator knows the other person and the machine by the labels ‘X’ and ‘Y’ (but doesn’t know which is the machine). The interrogator is allowed to put questions to the person and the machine of the following kind: “Will X please tell me whether X plays chess?” Whichever of the machine and the other person is X must answer questions that are addressed to X. The object of the machine is to try to cause the interrogator to mistakenly conclude that the machine is the other person; the object of the other person is to try to help the interrogator to correctly identify the machine (Oppy, Dowe, 2005). If the machine succeeds in convincing the interrogator that it is the person, it is said to pass the Turing Test.

A Detraction (or two)

The following is from Sam Vaknin, an author and doctor of philosophy, who refers to the book of instructions that the man in the Chinese Room uses to answer the questions:

“Whomever composed the book of instructions must have been conscious, possessed of mental states and of cognitive processes. Moreover, he must also have had a perfect understanding of Chinese to have authored it. It must have been an entity capable of thinking, analyzing, reasoning, theorizing and predicting in the deepest senses of the words. In other words: it must have been intelligent. So, intelligence (we will use it hitherto as a catchphrase for the gamut of mental states) was present in the Chinese Room (Vaknin, 2006). ”

I believe that understanding denotes intelligence, and vice-versa. Since the book of instructions symbolizes the program, I agree with the fact that a level of intelligence went into the creation of the instructions that are utilized by the machine and such programs have always been present in our programmable machines. However, does this denote that the machine understands the Chinese (or any other) language in the manner that the programmer understood it?

I have to say no. Programs are written by intelligent people, but a computer system utilizes a program as the steps to follow under certain conditions of input. On a less sophisticated scale, this is much like an engraving machine following a stencil in order to carve a certain pattern into a surface. Computing is far more sophisticated in its tasks, but the program is still consulted with input and it gives output derived from following the program as a guide.

However, Dr. Vaknin continues with another argument in contrast to the statement I just made:

“Surely, the programmer is the source of any intelligence that a computer possesses. But is this relevant? If the computer were to effectively make use of the intelligence bestowed upon it by the programmer – wouldn't we say that it is intelligent? If tomorrow we will discover that our mental states are induced in us by a supreme intelligence (known to many as God) – should we then say that we are devoid of mental states? (Vaknin, 2006)”

Though that last statement opens the door for the greatest theological discussion of them all, this goes back to my point about the engraving machine and the stencil. I don’t believe that because a machine can mechanically follow the stencil that was created by a person skilled at drawing makes that machine skilled at drawing. As in the CRA, if someone were to write a program that could pass the Turing test (in any language), a programmable machine would still be using it as a guide in order to complete a task. Letters or symbols would be read in one at a time, formed into words, those words would be compared to surrounding words (if any) to check for patterns, answers would be derived according to the program, and output given. When bestowed upon the machine, the program is no more than a sophisticated guide, or an emulation of human understanding.

Some Anticipated Sidebar Questions

Yes, but if I give an instruction book to a person who follows it to perform a task, does that just mean that they are only emulating the intelligence that went into the instruction book?

Typically when a person repeatedly follows instructions, eventually they will memorize them and no longer need to view them anymore to perform a task. Though they needed a guide at first, eventually the intelligence needed to perform the task becomes their own through repetition and memorization. A person may even figure out ways to make the process more efficient on their own. I feel that this is in contrast to what a machine does when following a program to complete a task.

Though a clever programmer can include dynamic routines to make a process perform more efficiently under certain conditions (such as page-replacement algorithms in virtual memory), a machine is relegated to following the steps of the program (its instruction book) each and every time that it needs that program to perform a task. Any measures of increased efficiency are credited to the programmer, not the machine.

A computer is given a program which it stores on the hard disk and loads into memory when needed. Isn’t this much like the person memorizing the instructions and utilizing them when needed?

There is a critical difference in the human method. The person is capable of finding ways of performing the task in a more efficient manner on their own outside of the instructions. For example, the instructions given for assembling a chair may state that the armrests should be attached to the seat first, then the legs attached afterward. However, a person may find that placing the armrests on the chair first makes it difficult to turn the chair over on its seat (either on top of a workbench or on the floor) in order to attach the legs to the bottom of the seat. The armrests may affect the stability of the seat, and the person may have to find some additional way to brace the seat and armrests while trying to attach the legs. This could hinder the speed at which the legs are attached. On the other hand, if the armrests were not attached yet, the flat seat bottom would easily rest on a work bench or floor without the need for additional bracing while the legs are attached in a quicker manner than if bracing were needed. Perhaps the person could write the manufacturer of the chair and suggest the instructions be updated?

This demonstrates a quality that humans possess but machines lack: Independent development. If a machine were given a program to follow, it would be required to follow it step by step unless the programmer included exceptions that would allow the machine to deviate from the order of the steps in a case of recognized efficiency (and even then, the machine would only be able to recognize cases of efficiency if the programmer had cleverly included that capability). On the other hand, the person could recognize such cases without outside intervention and choose to deviate from the instructions in order to complete the task in a more optimal manner.

Programmable machines can also determine optimal paths and steps and inform a user of such findings and offer suggestions. Is this unlike what was just described?

Once again, any such actions are credited to the programmer that has already put them into place. In the previous example, the person demonstrated another human quality: Choice. The person can choose to search for more efficient ways of completing a task despite the steps in the instructions. The machine would only do so if the programmer had desired to give it such capability. Any appearance of choice is only an emulation made possible by a human-written program.

That leads to another human quality that machines can only emulate if it were included in a program: Desire. The person assembling the chair would want to find a way to make it easier (especially if they had to do it many times). It was such a desire that lead them to choose to think of ways to deviate from the instructions, which then lead them to find a more efficient method and thus independently develop as an assembler.

The statement by Dr. Vaknin raises another interesting point. If our intelligence and mental states were bestowed on us by a higher power, wouldn’t these gifts make us emulators of the actions that were just described, and make us devoid of intelligence or mental states?

Even if our mental states were bestowed upon us by a higher power such as God, unless God is constantly guiding us along through each step of our lives like puppets, our desires and ability to choose outside of intervention give us independence, or free will. People have demonstrated their ability to choose to do many things, rational and irrational, throughout history. If I so chose, I could stop talking mid-sentence during an important presentation and start doing pushups for no apparent reason. When complete, I could pick up right where I left off with no explanation. This choice can be entirely unplanned and be done at that moment to prove a point about being controlled like puppets by a higher power. However, the integrity of the presentation is important to me and I would likely refrain from such silly actions. Even though people have done far more irrational things than what was just described, it is my choice not to interrupt the presentation since I desire not to confuse those in attendance.

This leads me to several conclusions about the three human qualities that were described: Our desires prompt choice. The chair assembler desired to find more efficient ways to perform the task, then chose to deviate from the instructions when they recognized a way to do so. Also, during presentations I desire not to confuse the audience so I choose to stay on subject and not do anything out of the ordinary.

Our choices can lead to independent development. Since choosing to deviate from the instructions led to the discovery of a more optimal process, the person developed into a more efficient assembler of chairs. Since I choose to stay on subject during presentations, I can spend more time gauging their reactions (which are unaffected by any unexplained acts on my part) to my presentation methods in order to help determine which things I am doing right and which I am doing wrong (Does making eye contact help keep people’s attention, or does it cause them to shy away? When I raise my voice, do people seem more interested or put off?). Spending more time with these concerns help people become more effective speakers.

I feel that desire, choice, and independent development are the aspects of human intelligence that have allowed us to make continual advances in the technological world. These aspects could not exist if our decisions and movements were controlled by an outside power. If I were being controlled, that would mean that I have no desire and choice of my own. Any development along the way would not be independent of the power controlling me.

However, none of these aspects could emerge without the ability to have an understanding of our surroundings. The understanding of a situation in which we confront difficulties prompts our desires to make the situation more tolerable. We then have the choice as to whether or not we wish to take steps to improve the situation. Whether or not we are successful in improving the situation, we can develop independently through the experience. Even if we fail to improve a situation, we likely gain some incidental yet beneficial knowledge along the way.

Human understanding is the base of any and all human-written programs. The programmer implies their level of understanding to a situation which may be solved by some type of automation or machine processing. They desire to create instructions that may be utilized by a machine to solve (or help with) the problem, and choose to write and test the program itself, which becomes a guide for the machine.

Summary

The following are a summary of my conclusions on this subject:

1. Humans have the ability to understand their surroundings and recognize areas of improvement which can prompt their desire to create methods of improving a situation.
2. Desire prompts choice which can lead to independent development; unless a higher power is constantly forcing our decisions and controlling our movements, these three things are undeniably part of human intelligence.
3. Human understanding and intelligence are present in programmed instructions.
4. Except in the face of some unexpected mechanical, electrical, or coding error, machines will follow a program’s steps (including any decision structures or methods of added efficiency under certain conditions) without deviation.
5. Following a program’s steps without deviation denotes the lack of desire and therefore choice and independent development.
6. On a programmable machine, human intelligence can only be emulated and a machine cannot possess any intelligence that it can call its own. Input processing is an emulation of understanding, decision structures are an emulation of desire and choice, and efficiency methods are an emulation of development, but this development is prescribed by and belongs to the programmer(s), not the machine. The program is no more than a sophisticated guide that the machine must follow. The fact that the machine must follow a program represents a boundary that prevents the machine from possessing its own intelligence.

In closing, the opinions in this article represent only one set of ideas when it comes to the intelligence capability of machines, and certainly do not cover all possible aspects and arguments on this vast topic. I must also note that my intentions are not to belittle the sciences of automated computation and their achievements. I will always remain amazed by the advent of the programmable machine and the surprising advancements that are continually made in these fields.