can computers think?
In Computing Machinery and Intelligence, Turing argues that if a computer can pass the test that he devised, then it could be stated that a computer is thinking though, not in the way a human thinks. In this paper, I am going to explain the Turing Test and I am going to explain what Turing thought his test proved. I will also discuss two objections that Turing talks in his paper and the Chinese Room argument. Finally, I will argue against the idea that computers think because they can not make mistakes.
In Turings paper Computing Machinery and Intelligence, Turing explains a test that he devised, and that if a computer passes this test, then Turing states that it could be stated that a computer is thinking.
The test is as follows: There is a person in a room (the interrogator) and something else outside of the room. The point of the test is to see if the something outside of the room is a human or not - the interrogator is allowed to ask questions to find out the answer. If the interrogator is convinced that the something is a human, then one could say that the something is thinking. But what if the something is a computer? If a computer can trick a person into thinking that the computer is a human being, does that mean that the computer thinks?
Turing thought his test was a sufficient condition for thought, but not necessarily a necessary condition for thought. By a sufficient condition of thought, Turing means that what computers do is a certain kind of thinking. By a necessary condition of thought, Turing means that not everyone who thinks need to do what computers do. Thus, computers think, but not all thinking things do what computer do.
Turing, knowing that many people would disagree with him, stated nine objections to his conclusions, and tried to refute these objections. Though, I think that delineating and categorizing these concepts into separate entities is simplifying the ordeal (I think that many of the objections go together and when combined together could be powerful arguments that Turing would have a hard time fighting against), I also think that Turing did a good job arguing against them.
One objection is The Heads in the Sand Objection. This objection states, The consequences of machines thinking would be too dreadful. Let us hope and believe that they cannot do so. (Hofstader & Dennet, pg 58). I really liked Turings response to this objection, which had a mocking feel to it, I do not think that this argument is sufficiently substantial to require refutation. Consolation would be more appropriate: perhaps this should be sought in the transmigration of souls. (Hofstader & Dennet, pg 58). I think people use this objection, not only for computers thinking, but for, practically, anything that they disagree with.
Another objection, which I think is more reasonable, and I think has more ground, is the Lady Lovelaces Objection. The Lady Lovelaces Objection states, The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform. (Hofstader & Dennet, g 63). That is, a computer cannot invent anything up it it has to be programmed to do so thus, a computer does not think. Turing argues against this by stating that computers surprise him all the time with what they come up with: Machines take me by surprise with great frequency. This is largely because I do not do sufficient calculation to decide what to expect them to do, or rather because, although I do a calculation, I do it in a hurried slipshod fashion, taking risks.
In Minds, Brain, and Programs, Searle argues against computers thinking, or to say more precisely, Searle argues against people who support strong AI. Searle states, according to strong AI, the computer is not merely a tool in the study of the mind; rather, the appropriately programmed computer really is a mind, in the sense that computers given the right programs can be literally said to understand and have other cognitive states. (Hofstader & Dennet pg. 353) Instead of using the Turing test, Searle comes up with the Chinese Room argument. In the Chinese Room Argument, one imagines themselves in a room, they are given Chinese symbols, from some outside source, that they do not understand (given that they are not a native speaker of Chinese), and they are told to look the symbols up in an exhaustive book of Chinese symbols, and write the symbol that the book states to write. They would then give the Chinese symbols that they wrote back to the outside source. Now, Searle argues that he would be translating Chinese, but he still wouldnt understand Chinese, and he states this is how computer works; thus, a computer doesnt think. To Searle, symbol manipulation is not sufficient for thinking because they dont have intentionality - that is the symbols mean nothing to a computer. Searle states, In the linguistic jargon, they have only a syntax but no semantics. (Hofstader & Dennet pg. 368) I agree with Searle and his assessment. I did an improvised Chinese Room experiment with my friend Greg, and found that he knew what to put down and when to put things down (the syntax), but he didnt exactly know what he was stating (the semantics). Attached to this report is the experiment (including key to symbols), the results of the experiment, and some questions I had Greg answer about the experiment before telling him what he was doing.
I personally dont think that passing the Turing test is sufficient to state that a computer is thinking. I think that passing the Turing test in conjunction with the ability to make a mistake is sufficient to state that a computer is thinking that is, that is if a computer can pass the Turing test and be able to make a cognitive mistake, then it is sufficient to state that it is thinking. But that is because, in my opinion, the only sufficient definition of thinking is the ability to make a cognitive mistake. I did not just come up with this litmus test because I feel like it that is, I am not just saying this to disagree with Turing. I pondered on this subject. I started to think about the meaning of to think and when do people use it and when people dont use it; and then finally, the definition Turing seems to give it. Turing seems to give a sufficient definition, that boils down to, of thinking as the ability to manipulate data. That is if a computer can manipulate date to fool a person into thinking that the computer is a person, then the computer is thinking.
But, I started to think about things we consider as non-thinking things, such as plants. And, I think according to the definition the ability to manipulate data you could state that plants think. What is data? What constitutes it? One can construe that sunlight could be data, and that plants manipulate it to create energy in the process known as photosynthesis (which, in my opinion, is closer to thinking than a computer can get to). But, the way that people use the term, thinking, does not include that plants think. Then something is missing.
I think what is missing is the ability to make a mistake. One thinks, because one can think wrongly, that is, to think is possibly to err. When I think of all things that I think are thinking, the thinking things all have the ability to make a mistake. A raccoon can misjudge the distance of car in a street, after crossing the same street over a hundred times, and get hit by a car; a person can take an extra step down a flight of steps and their foot can hit the bottom of the floor with a heavy plop, although having walked down those stairs thousands of times before; etc. As the ability to make a mistake lessens, the ability to think seems to lessen.
Computers do not think because they cant make mistakes. One can argue that the effects of a computer virus are mistakes, such as the computer crashing or the computer sending out your information to hackers, etc. But this is the difference between a hardware mistake and a software mistake the hardware was not meant to do what the computer is doing, but the software is doing what was expected to do. Viruses are made, particularly, to make your life hell. But computers were not made to have viruses. Its analogous to saying that a human who has the flu and is sleeping more often is mistaken. Lets entertain the idea that computers make mistakes, that is software mistakes, that is mistakes that they are programmed to do. If computers are programmed to make mistakes and they are successful in completing these mistakes, then they are not mistakes, because the by the very definition of mistake it is implied that mistakes are not done on purpose. So if a computer is making a mistake on purpose, then it is not making a mistake, but doing something that it is programmed to do. If it cant make a mistake, then it cant think.
In conclusion, I think Turings article is wonderful, but I like Searles outcome better. I think that thinking is not just the mere manipulation of data, but something else - I agree with Searle that thinking is instantiated by organic substances.
Douglas R. Hofstadter and Daniel C. Dennet; The Minds I: Fantasies and Reflections on Self and Soul; Basic Books, inc., Publishers; New York, New York; 2000.