1. I think Searle really needed to pin down what it was that he felt machines could not do. In fact, he ignores certain schools of pyschological thought (skinner?) where human behavior is only the product of reflex reactions and thought is merely the process by which these reactions take place. I also seem to recall that there are philosophical schools that take a similar viewpoint where humans have no independence of thought. In addition, I don't think Searle's viewpoint reconarch on human thought because there does seem to be a certain mechanistic aspect to human thought that he seems unable to deal with. Searle's viewpoints seem very shaped by whatever his own underlying religious and/or philosophical beliefs are.
2. I found this paper particularly annoying. I must admit that I cannot dispute Searle's arguements based on the example that he has given. However, the example he gives makes it impossible for the Chinese translator to have any understanding of the Chinese. In one part of the article, he states that one of the proponants of strong AI argues that it is possible for computers to understand exactly as humans do. I found it remarkable that he would state this and then give an example that has no correlation to human understanding. If he wants to dispute that computers can understand, he must set up an example that captures how humans understand, and then prove that the computer does not understand. This would be quite a feat.
3. If we look at the Chinese experiment, in many ways, it essentially models the way that we learn a new language or a piece of music by rote. We mimic what we see or hear, and learn to make responses, given some rules at the time. Had Searle taken the Chinese experiment further, the next step most likely would have been learning to associate the rules that he followed with some comparable event or memory, thereby increasing understanding. However, he stopped at a point before that happens, preventing additional learning. In the hamburger stories, I think that he conveniently ignores the amount of learning and cultural associations that went on until that point, that allows the reader to predicate a behavior. In essence, the stories and experiment really were taken out of context of their surrounding learning and culture, all of which allows a human to "understand" and to later make inferences and draw conclusions.
4. Although Turing's paper is weaker, Searle's is not strong. While Searle's Chinese translation example is intriguing, I would not agree with him that the distinction between understanding and not can be made by whether the manipulated formal symbols are interpreted or uninterpreted. Humans likely store in their minds some uninterpreted formal symbols that they use to understand the world around them. Furthermore, I would venture to guess that if a person (note that I am not implying computer) is in the translation room (or environment) long enough, (s)he would eventually understand the Chinese characters as they relate to each other, for which (s)he has correlation instructions. Had Searle known about neural networks, how would his argument have changed?
5. Still latter he must resort to saying that anything that merely executes a program could not possibly posses intentionality or understanding. The hardware as the brain and the program as the mind is obviously a weak and refutable analogy. I would argue that all living things result from the execution of programs. Scientists have very good reason to believe this. We were not all been created independently but somehow to the same design, without parents and with no developmental histories. We are all the results of a wonderfully complex program and we all think, believe, feel and understand. Clearly the execution of a program can result in unexpected things. We just can't limit our definition of what a program is.
1. Another reason I don't buy the imitation game theory is this example. Suppose for that this assignment I could use the Web search mechanism on my Internet browser to find a copy of someone else's homework from last year. I could turn that in and it would be very likely that you woul have no way of knowing that I did not do the homework. However, as far as you are concerned I did the homework. So you might wrongly conclude that I did my work and learned something about whether machines can think
2. This paper provides an interesting historical perspective on computers and society. Turing starts out by having to show how computers store instructions and how they are able to calculate. Turing spends much time discussing how electronics are a valid medium for such machines. Turing writes: "Readers must accept the fact that digital computers can be constructed ..." Today, we take these things for granted in much of society. Computers are ubiquitous throughout the western world, and the fact that a machine can do calculations does not have to be proven. Or the description of what a machine is and how computers fall into this realm is no longer necessary.
3. The most convincing, yet still not flawless, argument is that from Informality of Behavior. To give an alternative counter to the argument I will also rework it a little, as Turing did. It assumes that a human's actions can not be predicted with certainty, but a computer's actions can. This is due to the fact that a human uses complex and sensitive components, e.g. chemicals, whose levels can not be perfectly measured even if we did completely understand how the brain works. Anyone familiar with the concept in the study of Chaos know as sensitivity of inputs would probably agree with this. Furthermore the argument assumes that this immeasurable randomness inherent in humans is what, or at least one thing that, sets us apart from machines. All of a machine's data can be precisely measured because it must be using a finite amount of storage for any given value or else any one calculation that a computer might do would take forever. Hence computers must be inherently different then humans. This argument falls apart when you consider a theoretical hydraulic computer. Imagine a computer that stores its values as water levels in a series of vials. Suddenly, like with humans, every attempt to find the value of any of the computer's data can be nothing more than an estimation because no matter how close our value is to the real value in the vile, it still must be finite while it does not have to be finite in nature. This means that any attempt to predict what the computer will do will eventually be erroneous as the minute errors in the estimation of the values compound into larger and eventually substantial errors.
4. I thought that it was interesting that Turing's predictions were not too far off mark. He statest that within 50 years or so it will be possible to program computers with a storage capacity of about 10^9 to make them play the imitation game so well that an average interrogator will not have more than 70 percent chance of making the right identificat that there are several Turing test applications out there that could probably live up to this standard.
Winograd and Flores, in their book Understanding Computers and Cognition, characterize both minds and computers as being symbol manipulation systems. It is the ability to manipulate symbols which is characterized as intelligence and both artificial minds and biological minds exhibit this property. The arguments of Searle seem to fall into one of the falicious categories outlined by Turing as "the argument from consciousness", i.e., "not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols, could we agree that machine equals brain". The argument has a certain appeal. However, perhaps a machine can be built that comprises not only input, processing and output of sufficient refinement, but also feedback mechanisms of various sorts such that the "thought" of a hamburger is associated with, or triggers, other "thoughts" such as how a hamburger is cooked or what condiments typically are served with hamburgers, and further that each subsequent "thought" or association may trigger further "thoughts" such as what are the various types of mustard which are available. Perhaps such a cascade of associations might be considered to be a form of "understanding".