The most interesting "dare" thus far was presented by anonymous commenter "A":
Find the flaws in the Chinese Room argument against AI, and explain them to normal folk.The Chinese Room is a thought experiment first proposed by John Searle in 1980 that attempts to illustrate that Strong AI will never be accomplished by mere symbol manipulation. Strong AI claims that a computer can be programmed in such a way that it will be a mind, capable of conscious thought and comprehension. Weak AI says that, although computers can be useful tools for understanding minds, no computer (as currently conceived) will ever achieve real intelligence.
The thought experiment is as follows:
In the Chinese room thought experiment, a person who understands no Chinese sits in a room into which written Chinese characters are passed. In the room there is also a book containing a complex set of rules (established ahead of time) to manipulate these characters, and pass other characters out of the room. This would be done on a rote basis, eg. "When you see character X, write character Y". The idea is that a Chinese-speaking interviewer would pass questions written in Chinese into the room, and the corresponding answers would come out of the room appearing from the outside as if there were a native Chinese speaker in the room.The man in the room represents a computer following a set of rules (it's "program"). Even though he can generate outputs that appear to reflect comprehension, all he's doing is following rules and manipulating symbols.It is Searle's belief that such a system could indeed pass a Turing Test, yet the person who manipulated the symbols would obviously not understand Chinese any better than he did before entering the room.
I subscribe to the weak AI position, but as the Wikipedia article notes near the end, the actual truth may not be important. If a computer appears intelligent to an observer, does it matter if it really is or not? Perhaps only when considering moral and ethical questions, such as whether or not an apparently-intelligent computer can be "murdered". The problem becomes philisophical rather than scientific; I could argue that there's no way for me to know that you are actually intelligent, other than by observing you.
This "other minds" objection certainly doesn't refute the argument (it doesn't matter how I know you're really intelligent or not), but it may make the argument scientifically unimportant. We have no way to make decisions other than by making observations, and if we can't observationally distinguish between real and artificial intelligence then there's no science to be done. (Of course, if we can visually inspect the "brain" of the being in question we could decide easily.)
Searle refutes many other common objections in his own writing, and I suggest that you don't send additional objections to me until you understand his arguments -- it's likely that he's covered your objections already. The point is this: weak AI means that computer scientists will always know which intelligences are genuine and which are not, even if the general public can't tell the difference.
Miss "A" asks me to "find the flaws" in the argument, and I have only one to offer that hasn't been covered extensively elsewhere (to my knowledge, and it makes the argument stronger, not weaker). Searle assumes the existence of something I don't think is possible: a set of rules that the man in the room could follow that would allow him to appear to understand Chinese to outside observers.
Human language can be used to construct questions that aren't recursively enumerable, which means it may take an infinitely long time to find an answer, if one exists. For example:
Given a program and input parameters, will that program run forever?That question is undecidable, because for the input programs that will run forever the respondent will have to wait that long before knowing the answer. Somehow humans identify such situations and behave appropriately, approximating and guessing and making-do with rough estimates. A computer could "guess" also, but how would it know when to do so? That problem itself is undecidable.
It's certain that, even with an infinite rulebook, the outside observers would eventually come up with a question that would take the man in the room an infinite amount of time to answer. If the man understood Chinese he would be able to respond with an appropriate guess, but since he must rely solely on symbol manipulation he can only "guess" by replying with a random symbol, thereby revealing his lack of comprehension.
This whole discussion is vastly oversimplified, but "A" wanted me to try to make it accessible to "normal folk" so I did the best I could. For more information you can read the Wikipedia entries I linked to above -- they're mostly accurate.
hm...a neat post. but you could have demolished Searle even if you don't believe in strong AI.
i really liked your last point.
but, look at this load a malarkey searle says:
-----"All the same," Searle maintains, "he understands nothing of the Chinese, and . . . neither does the system, because there isn't anything in the system that isn't in him. If he doesn't understand then there is no way the system could understand because the system is just part of him"
-----Besides, Searle contends, it's just ridiculous to say "that while [the] person doesn't understand Chinese, somehow the conjunction of that person and bits of paper might" (1980a, p. 420).
Searle's argument is absurd! First, he says "collapse the boundary! make the person the system. THEN, resort to false syllogism! Claim the person still didn't understand simply because I asserted before that the person didn't understand!--but NOW, I've conflated that person to a SYSTEM!"
And then, even better, he resorts to reductionism! It's "absurd" that a systems' explanation would work because why? because a thing plus another thing is NEVER ANY MORE than the two things. Well... how do cars work? or bodies? Or computers? or vacuum cleaners?
Not to mention that he's got no way of explaining HUMANS KNOWING ANYTHING. He's resorts to reductionism to saw that a computer can pass the Turing test but not know anything--so I guess, us humans do too? Is he simply a nihilist?
But your last point is cool. Because in fact, it argues that searle's whole thought experiment is just plain UNREASONABLE. Why? because building the room is impossible! So, hello, he can't make the infinite sets of responses to an infinite set of questions so that the machine always halts. So, guess what? you could tell the diff. between a human and the computer. tsk tsk. bad searle!
I still liked my other question, though, the one about explanining how brilliant people can be SO wrong. Since obviously Searle is, too.
A: I don't think you're giving Searle nearly enough credit, and it's not at all obvious that he's wrong.
What you say is a load of "malarkey" isn't at all. As he says, there's no reason the human couldn't memorize the symbols and the rules and then execute the behaviors without anything other than himself as the system. Even still he wouldn't understand Chinese.
Further, he doesn't need to explain how humans know anything. Aside from the fact that no one knows, it's irrelevant to the discussion.
The main reason computer scientists stopped wrestling with the problem is that it's unsolvable and irrefutable. Searle is probably right, but it's not clear yet how it matters, other than in hypothetical moral dilemmas.
it IS malarkey, because his chinese room argument NEVER GETS TO THE ISSUE of "consciousness" because, by his own admission, humans could behave that way and not be conscious. Gee, what a straw man to argue that therefore, that way wouldn't mean a computer "thinks!" "this is all the machine is doing; THAT doesn't count as consciousness! (er, hm, that might be all a person does, too, but a) if that's conscious, I can't explain it, and b) if it isn't, then strong AI is right! c) if that isn't all a person's doing, I've got no way of saying why not without resorting to a non-reductionist perspective, and therefore, I shouldn't for the computer...D'oh!)
yes, it's unsolveable, and irrelevant. And therefore again, Searle is demolished, since he thinks that his argument ever addressed the problem. So Searle's argument is wrong--whether or not strong AI ever could be.
brilliant people make big mistakes. Searle's one of them. Marx was too. So's Penrose, Ryle, Einstein (re: quantum mechanics) etc. I see no problem with giving Searle credit but not for this poor argument.
I disagree with you about the "moral" argument, though. Everything's a moral/philosophical argument, right up until we make it a scientific one. That's the funny thing about philosophy questions. As soon as they are made concrete enough to answer, they start being answered in science, and fall out of philosophy. So this one will get answered too, and when it does, it won't be a philosophical isse.
we used to not understand the philosophy of sight. we thought about the philosophy of perceiving light or pictures, or having copies of pictures in our minds, or etc..then we discovered how photons work, how the retina works, how our optic nerve does fourier transforms, and low and behold, the question of "Sight" isn't philosophical anymore. now we just ask "but how do we KNOW that pattern that we see is a ...
the same will happen here. computers will think, and we'll know how we made that happen, and then we'll be asking "yes, but how do we KNOW we know that the computer is thinking?" etc. we'll still not know if it's moral to turn the darn thing off. :)
Very little in this world is provable, and it's particuarly hard to prove a negative (halting problem, again). Searle's Chinese Room is just a thought experiment, and it's a powerful indication that Strong AI isn't possible. But it's not a proof.
As it stands, either position is based largely on faith.
The problem with the Chinese room problem is that it is poorly conceived. At the cellular level, all a brain is doing is following instructions. What we think of as intelligence is probably an emergent property of the totality of the equipment and the rule sets, and there is no real reason I'm aware of at the moment that mind "cannot" arise if the equipment is silicon and copper instead of protein and neural transmitters.
If you don't know what it is or where it comes from in the first place, making up rules about what it can and cannot come from can be a tad silly.
In other words, it's only a "problem" in the first place if you have a definite position on a point not in evidence -- that is, if you deny the possibility that consciousness is or can be in an instruction set.
BP: The entire point of the illustration is that intelligence can't be merely an emergent property of a set of rules. If you're not aware of reasons why, then you didn't read the earlier discussion of non recursively enumerable languages.
But yes, the illustration was designed to defend a specific position. I don't think that has any impact on its effectiveness.