lecture plan: 3 points, after Churchland & Churchland: a) reformulating the argument as clear as possible b) syntax vs. semantics c) question begging premise #3 --- Hello, today we are going to look at an approach to understanding Searle's "chinese room" thought experiment. This approach is due to the philosophers Paul and Patricia Churchland. We shall first state one explicit argument that Searle has said is the argument extractable from the "chinese room" thought experiment, then look at two key terms it uses to make its point, and then using those terms as the key issues, show that the argument begs the question against traditional AI. I start by looking at the argument Searle purports to have used: Searle's argument - reformulated ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1. Computer programs are formal (syntactic) 2. Human minds have mental contents (semantics) 3. Syntax by itself is neither constitutive of, nor sufficent for semantics. Therefore, says Searle: Programs are neither constitutive of, nor sufficient for minds. Let's see how this maps on to the "chinese room" A computer program (for purposes of this discussion) can be looked as a system having three parts: input, internal states, and output. In the "chinese room", inputs come into the system as written symbols. These are processed by the internal states of the system which in this case are the states of the "lookup in the book process" that the person in the room does. The output is then created by the internal states of this process by a matching procedure. Here is a point where it appears the argument is rather odd - how is this matching to procede. Searle isn't clear on the point. But we shall ignore this - we can imagine any sort of matching algorithm or procedure we like, to be as fair as possible to Searle. This is an instance of what is sometimes called the principle of charity. With the outputs "determined", the person in the room then releases them into the outside world. We can imagine (Searle has us suppose, this being *his* invocation of the principle of charity) this being done in such a way so as that native speakers of the language the person in the room is manipulating are convinced that they are dealing with another person and another speaker of their language. Hence they've passed the turing test. But this was all purely a syntactic prosess - no *meaning* was involved. So no *meaning* can be attributed on the base of a computer program. Let's look at syntax and semantics in greater detail to understand what Searle is claiming. [pause for questions] Syntax & Semantics ~~~~~~~~~~~~~~~~~~ Syntax is the sentence structure of a language. We have spent some time in this course looking at how to represent the logical aspects of this structure. Remember then, how we abstracted away from the meanings of things when we found that we could model some inferences merely on the form of sentences and their combinations. The logic we have looked at is syntactical. A computer program is similar; in fact there are programming languages modelled after certain aspects of logic and related fields (Prolog, LISP). A compiler or interpreter looks at a computer program's *form*, much like we did when we did translations into logical notation. This is to be contrasted with the semantics of a language. When you hear me say something like "computer program", you are aware not only of the sounds - if you are aware of them at all - but of the meanings of the words, because you are a speaker of English. If I said "programme d'ordinateur", I have uttered some sounds that are perhaps unfamiliar to some of you - those of you who do not speak French. But they have the same meaning. The study of how utterances and inscriptions like the *written marks* "computer program" get meaning is what is (in part) studied by semantics; often we use the term loosely just to mean "meanings". Searle, then, thinks that the input to a computer program has no *meaning* to it, because the program is formal only. Some of you may see now why I (and the Churchlands) think that Searle's argument is question begging. But - apart from that - are there any questions? Begging the question in #3 ~~~~~~~~~~~~~~~~~~~~~~~~~~ You may remember from past lectures our discussion of arguing in a circle, or begging the question. We have just seen how Searle's argument turns on an understanding of semantics and computer programs. He is claiming that input to computer programs is processed by them purely in virtue of form, hence purely syntactically. But by claiming that computer programs can only deal with things syntactically is in opposition to AI! The argument is circular on that ground. The whole point of AI is to precisely figure out how to "squeeze" meaning out of form. As the philosopher Daniel Dennett puts it, there is no *intrinsic* meaning to any symbols we take as meaningful. Hence, there is some process that *does* get meaning out of form, in us at least AI is just trying to recreate that process. I shall quote directly from the Churchlands on this: "Moreover, to assume its truth [Premiss 3 - KD] is tantamount to begging the question against the research program of classical AI, for that program is predicated on the very interesting assumption that, if we can just set in motion an appropriately structured internal dance of syntactic elements, appropriately connected to inputs and outputs, we *can* produce the samge cognitive states and achievements found in humans" I said that I would focus on premiss 3 of the reconstructed argument from Searle. But, those of you who have studied computer science may also have noticed a possible equivocation in Searle's argument about the nature of computer programs, as well. This is an aside; once one finds one flaw in an argument it is sufficient in many cases. However, should Searle respond to the objects offered above by the Churchlands, he still must deal with the possible slip in premiss 1. "Computer programs are purely formal." Are they? Is this [show slide of source code] a computer program? Some source code to a program is formal, much like our logic. But that doesn't entail an "embodied" or running program based on that source code is purely formal. One way of seeing this is that the formal is static, unchanging. But a running computer program is a *process*. Does this affect Searle's argument? Perhaps, as it changes the way in which we think of the inputs to a program. Rather than again being formal, we can think of them as states of the world getting involved in a process. This is much more alike our own brain processes which are our cognition; so perhaps this premiss also is also, if I may speak losely, weakly question begging as well.