5 ms·

> I had been taught that Searle’s 1980 theory of the Chinese Room closed the book on the possibility of thinking machines. In Searle’s view, minds must be made

> I had been taught that Searle’s 1980 theory of the Chinese Room closed the book on the possibility of thinking machines. In Searle’s view, minds must be made of biological dynamics. Computers, which Searle must have envisioned as non-networked computers of the Commodore era, can only simulate these dynamics. > In hindsight, to reprise this theory in 2016 was laughable, especially given some, let’s call them, changes in computing in the past four decades. The internet long ago turned a computer from a four-walled “room” to an infinite set of portals which lead to billions of biological and language-using minds, with which our computers work in continuous and limitless collaboration. Searle's Chinese Room argument has never made any sense to me, but this makes even less sense. How did the internet change anything regarding that argument?

jfengel 4y ago
As far as I can tell, Searle is a bit like Noam Chomsky: deservedly famous for work on language and also for his anti-war work. He then wrote an absolutely abysmal paper on artificial intelligence, and everybody got wrapped up in it. The Chinese room argument has no merit whatsoever, and should be forgotten. Millions of better things have been said on the subject. But Searle, being famous, gets the mindshare.
- boole1854 4y ago
  > The Chinese room argument has no merit whatsoever, and should be forgotten. I have a different take on the argument. I believe it was widely misunderstood to be an argument against the possibility of artificial intelligence, when it was actually merely an argument that there is a logical difference between symbol manipulation and sentient understanding. Searle contributed to the misunderstanding by terming the position he was arguing against "Strong AI". He then argued that "Strong AI" was not possible on the basis of his argument. However, "Strong AI" has a different meaning to many people so that the "Strong AI" which Searle was trying to refute was not the same as the "Strong AI" that most people have in mind when they hear the term. For example, most people would probably say that a computer which could converse, write poems, compose novels, author novel scientific papers, etc. in perfect Mandarin, indistinguishable from a highly educated, human, fluent speaker of Mandarin, would be by definition a Strong AI. But Searle never refuted, or tried to refute, the idea that a computer could accomplish all of that, so he was not arguing against the position that most people would take to be "Strong AI". His point was to argue that having the qualia of conscious understanding is logically different than being able to accomplish the symbol manipulation required to achieve the tasks above.
  - MichaelCollins 4y ago
    Isn't Searle's argument equivalent to arguing that 'Strong [natural] Intelligence', the qualia of consciousness or whatever you want to call it, doesn't exist because individual neurons in a brain have no high level understanding of what the brain is doing?
    - boole1854 4y ago
      It isn't precisely equivalent so long as it is granted that the brain does something other than symbol manipulation. ARGUMENT 1. Symbol manipulation is not the same as conscious understanding. 2. The brain can perform the equivalent of symbol manipulation via neuronal interactions. 3. ? Conclusion: The brain does not have conscious understanding. Depending on how you fill in (3), you can get two possible conclusions: 3a. The brain does nothing other than the equivalent of symbol manipulation described in (2). With (3a), the conclusion above does follow. But, as an alternative: 3b. In addition to being able to perform the equivalent of symbol manipulation via neuronal interactions as described in (2), the brain has the ability to instantiate conscious understanding (perhaps this is done by some special kind of neuronal interactions or perhaps some other way). With (3b), the conclusion above does not follow. And since we know that the conclusion is false, we should therefore also consider (3a) to be false. It is true that the mysterious "ability" that the brain has is not explained by (3b). Searle's view is thus more palatable for people who are okay with the position that we do not currently have a satisfactory explanation for qualia. However, it should be noted that (3b) is not necessarily an endorsement of the existence of an immaterial "soul substance" or anything of the sort as it is compatible with physicalist theories such as integrated information theory (in which what neurons are doing is more than just symbol manipulation).
      - jfengel 4y ago
        Point 1 appears to be begging the question. Why shouldn't sufficiently-sophisticated symbol manipulation, producing results the same as "conscious understanding" (which seems to be the same as "semantics" as Searle was using it), be identical to "conscious understanding"? I don't know that it is, but Searle doesn't know that it isn't. I would think it was the question we were trying to prove. The question of whether the brain does symbol manipulation or not seems irrelevant. We know it doesn't have the same mechanism as either expert systems or neural networks. That seems trivial and not worth thinking about.
- grantcas 4y ago
  [dead]

jfengel 4y ago

As far as I can tell, Searle is a bit like Noam Chomsky: deservedly famous for work on language and also for his anti-war work. He then wrote an absolutely abysmal paper on artificial intelligence, and everybody got wrapped up in it. The Chinese room argument has no merit whatsoever, and should be forgotten. Millions of better things have been said on the subject. But Searle, being famous, gets the mindshare.

boole1854 4y ago
> The Chinese room argument has no merit whatsoever, and should be forgotten. I have a different take on the argument. I believe it was widely misunderstood to be an argument against the possibility of artificial intelligence, when it was actually merely an argument that there is a logical difference between symbol manipulation and sentient understanding. Searle contributed to the misunderstanding by terming the position he was arguing against "Strong AI". He then argued that "Strong AI" was not possible on the basis of his argument. However, "Strong AI" has a different meaning to many people so that the "Strong AI" which Searle was trying to refute was not the same as the "Strong AI" that most people have in mind when they hear the term. For example, most people would probably say that a computer which could converse, write poems, compose novels, author novel scientific papers, etc. in perfect Mandarin, indistinguishable from a highly educated, human, fluent speaker of Mandarin, would be by definition a Strong AI. But Searle never refuted, or tried to refute, the idea that a computer could accomplish all of that, so he was not arguing against the position that most people would take to be "Strong AI". His point was to argue that having the qualia of conscious understanding is logically different than being able to accomplish the symbol manipulation required to achieve the tasks above.
- MichaelCollins 4y ago
  Isn't Searle's argument equivalent to arguing that 'Strong [natural] Intelligence', the qualia of consciousness or whatever you want to call it, doesn't exist because individual neurons in a brain have no high level understanding of what the brain is doing?
  - boole1854 4y ago
    It isn't precisely equivalent so long as it is granted that the brain does something other than symbol manipulation. ARGUMENT 1. Symbol manipulation is not the same as conscious understanding. 2. The brain can perform the equivalent of symbol manipulation via neuronal interactions. 3. ? Conclusion: The brain does not have conscious understanding. Depending on how you fill in (3), you can get two possible conclusions: 3a. The brain does nothing other than the equivalent of symbol manipulation described in (2). With (3a), the conclusion above does follow. But, as an alternative: 3b. In addition to being able to perform the equivalent of symbol manipulation via neuronal interactions as described in (2), the brain has the ability to instantiate conscious understanding (perhaps this is done by some special kind of neuronal interactions or perhaps some other way). With (3b), the conclusion above does not follow. And since we know that the conclusion is false, we should therefore also consider (3a) to be false. It is true that the mysterious "ability" that the brain has is not explained by (3b). Searle's view is thus more palatable for people who are okay with the position that we do not currently have a satisfactory explanation for qualia. However, it should be noted that (3b) is not necessarily an endorsement of the existence of an immaterial "soul substance" or anything of the sort as it is compatible with physicalist theories such as integrated information theory (in which what neurons are doing is more than just symbol manipulation).
    - jfengel 4y ago
      Point 1 appears to be begging the question. Why shouldn't sufficiently-sophisticated symbol manipulation, producing results the same as "conscious understanding" (which seems to be the same as "semantics" as Searle was using it), be identical to "conscious understanding"? I don't know that it is, but Searle doesn't know that it isn't. I would think it was the question we were trying to prove. The question of whether the brain does symbol manipulation or not seems irrelevant. We know it doesn't have the same mechanism as either expert systems or neural networks. That seems trivial and not worth thinking about.
grantcas 4y ago
[dead]