The Turing Test: Affirming the Consequent?

[inocente]

That was an example. I’m not bothered by a single comment, just by the fact that we can’t have a debate without bringing every retort as close to an insult as is possible without breaking the forum rules.

If you think that’s my aim then I’m out of here.

 

[inocente]

That was an example. I’m not bothered by a single comment, just by the fact that we can’t have a debate without bringing every retort as close to an insult as is possible without breaking the forum rules.

I think Quine would agree that there is one objective reality. It does not seem like the “One True Model” characterization of his view does it justice, though. His point seems to be that whenever we interpret data we interpret it with respect to background assumptions. So people have different models; Quine’s relativism, I think, would be likely to impose limitations on our ability to single out a true model.

I’m saying there can be no one true model, since it’s a model.

I am using underdetermination as an example of what is “logically impossible” and so not under the domain of God’s omnipotence. A state of affairs cannot determine one interpretation to the exclusion of others. One could not say, “Well maybe God can make the state of affairs determine a single model,” but that is not possible. I am suggesting the same is true of mind. Could God make a machine intelligent? Perhaps. But what God could not do is make intelligence emerge out of the physical facts of a machine.

You now need to prove that you logically and necessarily possess some essence or factor that no machine can ever possess.

 

[inocente]

I’m guessing you’ve adapted Daniel Dennett’s argument in your previous demand that I show that you have the same qualia as me or that my qualia are the same as yesterday. But of course, the issue with Dennett’s argument is that it misses the point, since no one has claimed that qualia are the same for all people over the course of time, or that comparative questions about qualia need to have discoverable answers.

I claim qualia don’t exist. You claimed “qualia are perfectly falsifiable”. Come on then, where’s the experiment?

*Signal and avoidance is not the same as the subjective feeling. Creatures of all orders respond to and avoid stimuli reflexively. Even plants do it. To say they all experience a subjective feeling of pain seems to reduce the proposal to absurdity. But then the response to “something wrong” and the avoidance of further damage is not at all contiguous with the experience of pain.

(A further point is that such responses occur even on a cellular level; cells respond to non-optimal stimuli, modify their environment, signal for responses to invaders/damage, etc. What in principle is different? But it doesn’t make sense to say that pain is felt on a cellular level, unless we give up materialism and settle for panpsychism.)*

I think it’s been shown that the sensation of pain is created within the brain itself. I described pain as something which “keeps hammering for our attention so we’ll do something to make it to go away”. That description could only apply to a machine/being/system capable of modifying its behavior, there would be no point otherwise. But equally it could apply to any machine/being/system, there is nothing in the description which requires anything inexplicable or mystical.

The issue, though, is whether they are thoughts at all. The presence of representative data in the nervous system does not seem like a sufficient condition. Take the camcorder again. I could be checking the stream of visual data at any given point. I could even reproduce an image according to it (for, in the thought experiment, I designed the camera, and that’s exactly what it does). But it doesn’t follow that the data present in the camera’s system constitute “thought.”

Well no, but gadgets apart, I am conscious of some thoughts and therefore they occur and therefore they or some correlate could be logged.

Not the point I was making. I’m just saying that science probably has epistemic limitations even given full-blown materialism. It seems implausible that humans - which, by materialistic hypothesis, evolved messily and suboptimally - should have some intrinsic limitless capacity to discover all truths about the natural world. One could, of course, define reality by what is discoverable by science, but such a notion is pretty ad hoc, since science is, again, the activity of principally limited human beings.

Sure, but explaining thinking isn’t a difficult problem. It’s just that philosophy is the wrong tool, hasn’t got anywhere, and so has made it look like a hard problem. Early days as far as the science is concerned, and progress is rapid.

You know, circularity is not an example of wrongness. It does not prove anything, but it does not prove that a claim is, for example, inconsistent. So reconstructing my statement to be circular in a way in which I did not state it proves nothing.

I am pretty content to say that I have qualitative experiences. If you are a philosophical zombie, then I guess your denial of qualia suddenly makes sense.

What I mean by circular is calling experiences qualia and qualia experiences is a content-free exercise, it says nothing new, they are just different words for the exact same thing.

 

[inocente]

This view is highly implausible. “Pain,” “red,” and “bitter” cannot simply be “labels we attach to the neural configurations” because we have knowledge of them completely independent of our knowledge of neural configurations, while the neural configurations add a coincidence of physical conditions. But we already have other coincidences: for instance, I feel pain someone pokes me with a pencil, and I see red when a red object is in front of me. Those physical conditions are not the same as “pain” or “red.” So the burden is on the materialist to construct a model on which internal physical conditions simply are equivalent to sensations of pain and color. But no such effort is here made: it is asserted that they “just are” and that’s the end of it.

Look at the image for thirty seconds, keeping your eyes on the bottom right star, then look at a blank surface. You should see an after image of a true color flag (rapidly blink your eyes if necessary).

See en.wikipedia.org/wiki/Afterimage

http://upload.wikimedia.org/wikipedia/en/thumb/f/f3/US_flag(inverted).svg/800px-US_flag(inverted).svg.png

 

[Peter_Plato]

This last half sentence: “when it, in fact, doesn’t” is where we depart.

It all boils down to approximation. If we would build a cart with triangular wheels, it would be next to impossible to let that cart “roll”, it could only be “dragged”. We can then improve on the design and use “square” wheels. Not much of a change, the ride is still very bumpy. None of these approximations can be asserted to be a good “replica” of a circular wheel. As we keep on improving and use a multi-sided polygon, at some point the ride starts to become smooth. There is no way to decide if the wheel is a million-sided polygon, or an actual circle. Especially, since the “real” circle is just a mathematical abstraction, which can only be approximated in real life.

You may say that “ontologically” speaking, we “only” have a million-to the power of-trillion-sided polygon and not a “true” circular shaped wheel, but this “difference” is impossible to detect, even in principle. And this is where our difference lies. You cannot say that the wheel is not “circular”, because you have no epistemological method to find the “difference”. I say that if the difference is impossible to detect (in principle), then it makes no sense to even talk about a “difference”. **Ontology without epistemology is just empty speculation. **

This fundamental difference (and it can be detected, all right ) is why we can only have a partial agreement. But partial agreement is still good to have.

I would see it just the opposite. Epistemology without ontology is merely a delusion of certainty. Certainty without correspondence to actuality is empty. Period.

Where your cart wheel analogy breaks down is in the very presumption that human thinking is reducible to behaviour, such that as we more closely approximate human-like behaviour we actually do create human-like thinking. The choice of your analogy merely reinforces your presumption, it doesn’t present any case to support that we ought to accept its principle tenet that intensional thinking of the human kind is, indeed, reducible to behaviour, except as a presumption. You just beg THAT question, completely.

To be clear, your analogy begins with a presupposition that behaviour (iterations of polygons) is the same kind of thing (2D figure) as thinking (circle) is. That, however, is precisely the issue at stake. Why should we believe that thinking is merely complex behaviour when it is plausibly true that complex behaviour could be created via sophisticated AI technology without subjective consciousness being required, i.e., a philosophical zombie?

It is like arguing from a premise similar to “A man is merely a woman without a womb,” to a conclusion that if technology ever gets to the point of perfecting an artificial womb which can successfully deliver a new human baby via implantation, we have overcome all gender differences. The problem here, as with your analogy, is that we have no reason to accept that a woman (circle) is merely a man (polygon) with a womb (curves instead of straight lines), nor do you provide any reason to accept that human subjective consciousness is reducible to exhibited behaviour no matter how strongly you (merely) insist or presume it to be so, or beg the question entirely.

 

[Peter_Plato]

Woman with twins in womb?

 

[Peter_Plato]

I am pretty content to say that I have qualitative experiences. If you are a philosophical zombie, then I guess your denial of qualia suddenly makes sense.

Is it possible that those who deny subjective self awareness deny it because they specifically lack it or lack a qualitatively unique awareness of it?

The same could be true regarding free will or free moral agency. Perhaps it is not so much that deniers do not apprehend the idea of “free will” as a conceptual possibility, but simply lack a frame of reference (free will itself) from which to fully realize it beyond a mere concept.

Perhaps a similar problem even exists regarding a denial of God.

Similar, for example, to how heavier than air flight deniers could not conceive of the possibility of flight until it was realized. Most people don’t understand how flight is possible, even today, and couldn’t explain its principles, but merely accept it because they can experience it, though not because they can understand it. Yet understanding the possibility (and theory) of flight had to precede actualizing the possibility, which would not have taken place if the possibility was consistently denied. (A general problem with some forms of skepticism, perhaps?)

Similarly, “thought” deniers may not accept the reality of qualia because of their own metaphysical presumptions regarding the question. Until the “idea” of thought itself is fully flushed out, the deniers will continue to insist on the “lack of evidence” for the existence of qualitative experiences, even if it means denial of their own qualitative experiences (assuming they have them) because of metaphysical presumptions regarding what counts as evidence - a position that was the entrenched position (one of ignorance parading as lack of evidence) held by heavier than air flight deniers prior to the reality of flight.

 

[polytropos]

Ah, but I do not say that that “intelligence” is a particularly human trait.

That’s exactly what I was saying. You do not find intelligence particularly human. But acting like a human is the basis of the Turing test, which you use to determine whether something else is intelligent. If intelligence is not particularly human, then why is human language processing the criteria? Suppose you have a computer which can pass the Turing test. Remove all of the language processing software; just let it solve riddles and equations like an intelligent human would. There seems to be no reason to hold that it’s not intelligent just because it doesn’t act like us (because, of course, we aren’t human). But then the test loses its “rhetorical force,” because we don’t really mean to say that computers that just do complex operations are intelligent beings.

Animals also exhibit a certain level of intelligence. Apes can conduct a conversation in sign language. Apes can use and manufacture tools.

Ape sign language is rather limited. For instance, apes who are taught sign language tend not to initiate conversation and have hard limits on the number of signs they can learn. To quote one author (who quotes a linguist):

Whatever the sincerely held hopes and the promises of success, however, the attempt to teach apes sign language can only be viewed objectively as a failure. no ape has ever been taught genuine American Sign Language. As the linguist Steven Pinker explains: ‘This preposterous claim is based on the myth that ASL is a crude system of pantomimes and gestures rather than a full language with complex phonology, morphology and syntax. In fact the apes had not learned any true ASL signs.’ He goes on to quote the one deaf native signer on the Washoe team [Washoe was one of the famous chimps who was engaged to learn sign language], who complained that while he could not see Washoe producing even one ASL sign, the other experimenters were gleefully noting down hundreds of ‘signs,’ and showing implicit displeasure that the deaf observer’s logbook was not full enough. … Unlike children, who also produce gibberish in their early years of language learning, the chimps do nnot refine their behaviour to the point where they are regularly producing correct, grammatical sentences.

Tool use is likewise dubious. For example, Jane Goodall thought that apes made tools because she observed them stripping leaves from twigs to get insects out of their nests. But certain birds do the same thing. So tool use and communication in apes are quite limited, but have never really been regarded as paradigms of intelligence anyway.

The reason we wish to use the human intelligence as a “measuring rod” is simple. Of all the beings we encountered so far, humans seem to have the highest degree of this nebulous intelligence.

Quite true. But that does not really answer the argument I’ve made, which is that it would not be human language processing that determines intelligence, even if human intelligence were a rough scale. The issue is that if you take two computers, one that can “fool” a human, and the other identical except unable to fool a human, it’s not clear what has substantially changed in the activity of the computer. The complexity of its internal operations and the way it manipulates data could very well be the same. The fact that it can fool a human does not seem to add anything. It just seems to fool people.

Even the most advanced super-computer is incredibly simple compared to the neural structure of an ant (not to mention a rat!). I am talking about those “computers” which have billions of CPU’s arranged in a parallel processing structure, which have at least visual and auditory I/O ports (can have others, too). Which can remember their past actions and anticipate future ones. Which can generalize and act in a purposeful manner. (We are all aware that the linear accumulation of parts can switch into a qualitative change. A simple example is the critical mass of fissionable material. A more complicated one would be a growing crowd, which turns into a mob.)

What you will need to show is that we’re not just finding a way to deceive ourselves. Properties emerge; they emerge for reasons (ie. we realize that it must be a critical mass of fissionable material, not just any material, which explains the volatility of the critical mass; if it were another material, it would not be a critical mass, so the observation of emergence as a general phenomenon does not absolve one of the need to substantiate claims with theoretical models).

Another question would be: what do the visual and auditory ports add? Certainly a human does not lose his intelligence by losing the capacities of his eyes and nose (or even the ability to process other human language). It seems like latent computational ability should be sufficient for intelligence; if you have a computer that you deem intelligent, it would have to remain intelligent if you remove whatever “sensory” (name removed by moderator)uts it is getting. The further concern is as to why we should regard its activity is different. The camcorder example seems to bother people, but it does make it clear that we can’t regard the presence of sensory data as a sufficient condition for thought or qualia. What does the complexity of the processing do to change this fact? At what point do the 0s and 1s that a camcorder uses to represent an image to its human users turn into an image which is represented to your supercomputer as an individual subject, by virtue of the data being “within the system”?

 

[polytropos]

You misunderstand me. I do not assert that the concept of pain is identical to the “actual” pain. When someone pokes you with a sharp object, it is processed by the nerve endings, transmitted by the nerves, and processed by the pain center. You still talk about “coincidences” here, which is not warranted. It implies the lack of physical causative factor. And we both know that it is not simply a “coincidence” (which may or may not happen) that we feel pain when poked with a sharp object.

Then feel free to clarify. You said that they were labels standing for neural configurations. When I’m poked, I feel pain and certain nerve endings are active. No one is disputing that. But the claim (if you’re making it) that the feeling of pain and the activity of certain nerve endings are the same thing seems to just beg the question without any explanation.

No one asserts that each individual “cell” is supposed to “feel” pain. This would be a kind of reductionism which is falsely attributed to the physicalist’s approach.

No one speaks here of “generally”. I do not say that “any or all” stimulus-and-response mechanisms will experience “pain”.

I did not attribute it to your approach. We can agree that cells don’t feel pain. The question I am posing is what distinguishes the computer’s complex stimulus-and-response mechanisms from those of a cell? The activity of a cell is very complex. It responds to stimuli, manipulates “data” encoded in that stimuli, and responds accordingly. Cells also communicate with each other (and humans would be hard pressed to learn to communicate with a cell!). But that in itself is not a sufficient condition for thought. So when we take a computer that can respond to stimuli and converse with humans, what makes it different from the immense number of complex stimulus and response mechanisms active in the biological world? You and I agree that, on view of the proponent of the Turing test, intelligence and communication are not uniquely human; why, then, cannot alternative manifestations of response to stimuli and communication in the biological world be examples of intelligence?

So it can’t be stimulus and response mechanisms in general. Then: what is it? Why is it particular types of communication and stimuli response? What criteria do you use to sort out the intelligent communication from the non-intelligent communication in a way that is neither ad hoc nor question begging?

Why does not the lack of generality go the other way? ie. you can say that not all stimulus and response mechanisms are intelligent, but then can we say that all human-like behavior is intelligent?

Suppose that you have a sadistic person, who wishes to live out his fantasy. Yet, he does not want to torture another human being, so he constructs a crude dummy, and beats that. This does not give the satisfaction he craves, so he starts to “improve” on the design. First he puts a device into the dummy, which emits groans, shrieks and begs the torturer to stop. Then he “improves” again. The next dummy will exhibit more pain responses, attempts to escape. Then comes some physical response; the “dummy” which bleeds, which exhibits all the obvious signs of suffering. But there is still more room for “improvement” (in a very sick sense!). The next dummy will be a homeostatic being, which is aware of its fate, which is scared of death, and yet craves death since it is the only escape of the pain.

Where in this line is the dividing step where “emulation” stops and becomes “reality”?

I find the situation vacuous, for arguments given throughout the thread; the creation would never be intelligent. The idea of creating a lifelike dummy in order to punish it would be immoral on its face, but because of the intention of the person making the dummy, not because the dummy has rights.

And this is the problem you face. If you say that all this is just an “emulation” and the “visible suffering” of the dummy is of no relevance, then you need to give an epistemological method which can tell apart the emulation from the “real” thing.

I do not, since I have never claimed that my judgments are not defeasible. If I’m fooled, then I’m fooled; I imagine I’d call it a human and that would be that. If it were convincing enough, as your scenario posits, then I would probably not believe someone who claimed that it was in fact artificial. None of this has any bearing on the ontology of whether the being is intelligent, thinks, or feels; I would just be wrong in thinking that it did (like I might be fooled if a good actor pretends to be in pain).

You may say that “ontologically” speaking, we “only” have a million-to the power of-trillion-sided polygon and not a “true” circular shaped wheel, but this “difference” is impossible to detect, even in principle. And this is where our difference lies. You cannot say that the wheel is not “circular”, because you have no epistemological method to find the “difference”. I say that if the difference is impossible to detect (in principle), then it makes no sense to even talk about a “difference”. Ontology without epistemology is just empty speculation.

But I’m not saying that the machines “approximate intelligence.” I’m saying that they approximate human-like behavior, and if you want to call that “artificial intelligence,” then that’s fine and good. The practical conclusion doesn’t take you further than that. I’m not the one in need of epistemological method, since I’m not making positive claims; so I agree that ontology without epistemology is empty speculation. The issue is that the claim that a machine is intelligent is the one that lacks an epistemology.

 

[polytropos]

I’m giving up on this point, since your insistence that “I know I think; I am human; therefore only humans can think” is an impenetrable fortress.

That is, however, not my conclusion, nor my argument. I am saying that intelligence, by any credible definition, is fundamentally subjective, and the lack of a necessary connection between intelligence and human-like behavior imposes limits in principle on what the Turing test could ever show about the subjective state of those it observes. You are trying to make my argument into a positive claim that only humans can think; but it is actually the Turing-test proponent who needs to make the positive claim, and who is up the creek without a paddle if his argument just doesn’t follow.

It’s very frustrating that you keep on about camcorders. I asked you to think about designing a machine which can survive in a complex environment like we can, by itself for year after year, and to see that hierarchical pattern matching and prioritization mechanisms are required for it to sift out the important information it needs to survive from the reams of data hitting its sensors every second. Why do you keep on about some nineties gadget? I don’t know whether it’s me or you but we’re just failing to communicate so I’m giving up on this point as well.

I bring up camcorders because I think we can all agree that visual data is present in their system, but they do not have qualia. If you have a theoretical model for why the addition of complexity and the handling of other sensory (name removed by moderator)uts makes a difference between the more complicated computer and the camcorder, then you may provide it. It is not clear why hierarchical pattern matching and prioritization mechanisms should do this. I agree that they might be necessary*, but there is just no argument as to why they should be regarded as sufficient. (There are certainly complex nonintelligent prioritization systems. The question is, what distinguishes them?)

I asked you to prove qualia exist, and instead you went circular on me: p-zombie is defined as having no qualia, so absence of qualia is defined as p-zombie.

The point I was making was that denying that qualia exist is tantamount to denying that you are a personal subject who has qualitative experience. I find denying them to be incoherent (unless, of course, you are not a personal subject with qualitative experiences). It’s like someone saying, “Prove to me that you think” or “Prove to me that you don’t have an inverted spectrum.” I don’t think I can do that for you, but if you want to deny it, I am willing to say, “Go ahead.”

 

[polytropos]

I’m saying there can be no one true model, since it’s a model.

So basically… Quine’s position.

I claim qualia don’t exist. You claimed “qualia are perfectly falsifiable”. Come on then, where’s the experiment?

Are you a subject who experiences things like the color red?

Well no, but gadgets apart, I am conscious of some thoughts and therefore they occur and therefore they or some correlate could be logged.

It seems like “some correlate” is much too weak a claim, since I don’t dispute that correlates of thoughts can be detected (that’s indeed what the studies you cited did show). Since whether thoughts in themselves can be logged is what is at issue, we won’t be able to take that on bare assertion.

Sure, but explaining thinking isn’t a difficult problem. It’s just that philosophy is the wrong tool, hasn’t got anywhere, and so has made it look like a hard problem. Early days as far as the science is concerned, and progress is rapid.

It seems doubtful that someone could even construct an experiment to show that thoughts exist. As I’ve said before, on its face, it seems like a category mistake. So the logical positivist is left with redefining thinking in weaker terms more generous to his theory or becoming an eliminativist.

What I mean by circular is calling experiences qualia and qualia experiences is a content-free exercise, it says nothing new, they are just different words for the exact same thing.

Well, I’ve been saying “qualitative experiences” since that is the definition of qualia. I don’t think I’ve been saying that experiences are qualia, full stop (doing a calculation in my head, for instance, wouldn’t be qualitative). If I did, it was probably stated loosely, for which I apologize.

 

[polytropos]

Is it possible that those who deny subjective self awareness deny it because they specifically lack it or lack a qualitatively unique awareness of it?

Similarly, “thought” deniers may not accept the reality of qualia because of their own metaphysical presumptions regarding the question. Until the “idea” of thought itself is fully flushed out, the deniers will continue to insist on the “lack of evidence” for the existence of qualitative experiences, even if it means denial of their own qualitative experiences (assuming they have them) because of metaphysical presumptions regarding what counts as evidence - a position that was the entrenched position (one of ignorance parading as lack of evidence) held by heavier than air flight deniers prior to the reality of flight.

It seems like someone could only deny qualia if they don’t understand the concept. They seem to me to be a basic aspect of being a human - like thinking itself. Their intransigence to science seems to be what leads materialists like Daniel Dennett and the Churchlands to deny them, whereas the inability of neuroscience to discover them should simply make one skeptical of materialism.

 

[Jewel34]

polytropos:

I find the situation vacuous, for arguments given throughout the thread; the creation would never be intelligent.

So many misunderstandings! Very frustrating. Maybe I am too dumb to communicate my points clearly. The example has nothing to do with the specifics of “intelligence”. It was offered as a much broader conceptual problem of “emulation” versus “reality” – whether we talk about intelligence or pain or whatever. That is a generic problem, it could be applied to any situation where there is something called “original” and another thing we call “emulation”. In a sense it points beyond the scope of this thread, but it is an obvious result of examining the “details”.

And the problem has nothing to do with “morality” of creating such a dummy. By the way, if that dummy would be a philosophical zombie, then there would be no “moral” problem, just like there is nothing morally wrong to take a hammer, and “torture” a piece of rock. Of course the concept of a “philosophical zombie” is exactly as ridiculous as a hypothetical copper-wire-look-alike, which is identical to the “real” copper wire, but which cannot transmit electricity. It is amazing how many incorrect thought experiments can some philosophers come up with.

polytropos:

If intelligence is not particularly human, then why is human language processing the criteria?

Simple. Because it seems to be the best criterion. Language usage is fraught with difficulties, homonyms, synonyms, misleading and multi-use connotations, jokes (!), and especially puns (!!). To weed out all the pitfalls seems like an excellent way to separate the chaff from the wheat. Can you offer a better way? Interpreting language is much more difficult than solving mathematical problems. It is rather amazing to see Watson to be able to decipher the intentionally fuzzy clues given on Jeopardy. It is not surprising to see “him” beat the human contenders. The speed is not relevant, the accuracy is.

polytropos:

Why does not the lack of generality go the other way? ie. you can say that not all stimulus and response mechanisms are intelligent, but then can we say that all human-like behavior is intelligent?

You bet we cannot. There are many people who are not “intelligent”, even outside the obvious groups of newborns and the mentally retarded.

I attempted to offer a definition for intelligence: “the ability of solving complex problems, of being able to conceptualize and to generalize, of having the ability to separate the relevant factors from the incidental ones…” and so on. I am very much aware that these are also fuzzy criteria. What is “complex” for one person is very simple for another one. That is why the Turing test offers the (also fuzzy) “long enough sequence” of conversation. Contrary to the opening post’s assertion, if we keep on asking questions and we consistently receive correct answers; that is a very good indication that the respondent has actual knowledge and does not rely on “coincidental” answers. Indeed we all rely on this variant of the “formal” fallacy of “affirming the consequent”. The whole building of science uses this method of substantiating a hypothesis (while being aware that the process will never prove that the hypothesis is correct).

 

[polytropos]

So many misunderstandings! Very frustrating. Maybe I am too dumb to communicate my points clearly.

I hope we can have a civil conversation.

The example has nothing to do with the specifics of “intelligence”. It was offered as a much broader conceptual problem of “emulation” versus “reality” – whether we talk about intelligence or pain or whatever. That is a generic problem, it could be applied to any situation where there is something called “original” and another thing we call “emulation”. In a sense it points beyond the scope of this thread, but it is an obvious result of examining the “details”.

And the problem has nothing to do with “morality” of creating such a dummy. By the way, if that dummy would be a philosophical zombie, then there would be no “moral” problem, just like there is nothing morally wrong to take a hammer, and “torture” a piece of rock. Of course the concept of a “philosophical zombie” is exactly as ridiculous as a hypothetical copper-wire-look-alike, which is identical to the “real” copper wire, but which cannot transmit electricity. It is amazing how many incorrect thought experiments can some philosophers come up with.

You say it has nothing to do with morality… but the thought experiment posed was a case of torturing a human-like entity. I’m not sure why the scenario is supposed to appear morally egregious if morality is not relevant to it.

The presence of intelligence and pain are to the point because they answer one of your questions; in actuality, it does not cease to be emulation and become reality. At no point, no matter how convincing the dummy is made, is one torturing a real human being.

Not to mention, I’m not sure what all the hysterics are for, because if we look back to my response in #88, I did answer the other considerations (ie. the epistemological question) of your thought experiment (which it seems you attach greater important to, than to my apparently off-topic preface about intelligence and morality). It’s not like I tossed out intelligence/moral concerns as a half-response and then ignored the rest. (Perhaps you are still formulating the rest of your reply, but then I can’t understand the order of your reply, in which you started with my response to your thought experiment, which was at the end of my series of posts, and went on to comment on unrelated chains of the debate without saying anything about my response to the epistemological question.)

Simple. Because it seems to be the best criterion. Language usage is fraught with difficulties, homonyms, synonyms, misleading and multi-use connotations, jokes (!), and especially puns (!!). To weed out all the pitfalls seems like an excellent way to separate the chaff from the wheat. Can you offer a better way? Interpreting language is much more difficult than solving mathematical problems. It is rather amazing to see Watson to be able to decipher the intentionally fuzzy clues given on Jeopardy. It is not surprising to see “him” beat the human contenders. The speed is not relevant, the accuracy is.

Human language processing is complex. The question though, is: why is it different in principle from other complex operations that a computer does, from the standpoint of a computer? I’m not sure if you’ve ever used a program as powerful as Wolfram Mathematica, but there is virtually no upper bound on the level of complexity of the math problems I could throw at a computer. It seems like eventually, the complexity has to surpass that of human language processing. Now, it’s tougher for us to program a computer that processes human language - but merely asserting that it’s a complicated process (with “homonyms, synonyms, misleading and multi-use connotations, jokes, and puns”) does not seem to answer the question at all.

You bet we cannot. There are many people who are not “intelligent”, even outside the obvious groups of newborns and the mentally retarded.

My use of the word “all” was probably misleading. By “human-like behavior” I mean “behavior sufficient for a machine to be pronounced intelligent by the Turing test,” not any behavior performed by humans (which need not be paradigmatically intelligent). The problem posed was this: You have claimed that you are not committed to regarding all manners of communication in the biological world as intelligent, ie. immensely complicated communication (like that between cells or occurring within the human body) does not entail intelligence between participants. But why, then, can we regard any entity capable of engaging in human language as intelligent, just by virtue of that capability? We exclude some complex manners of communication, but not others; this discrepancy needs to be accounted for in a non-ad hoc manner.

Further, computers communicate in their own languages over the internet all of the time. Like the mathematical equation, it seems like there’s no upper bound on the level of complexity of such an interaction. It’s communication. It’s complicated. It’s even between two entities which you propose are potentially intelligent, if in other circumstances (in conversation with a human interlocutor). But then the question is: why can’t we regard these beings as intelligent? Perhaps by communicating with 0s and 1s their communication is even more efficient than ours. In this sense the Turing test commits the same anthropocentric bias of which it accuses its opponents.

 

[polytropos]

Contrary to the opening post’s assertion, if we keep on asking questions and we consistently receive correct answers; that is a very good indication that the respondent has actual knowledge and does not rely on “coincidental” answers. Indeed we all rely on this variant of the “formal” fallacy of “affirming the consequent”. The whole building of science uses this method of substantiating a hypothesis (while being aware that the process will never prove that the hypothesis is correct).

If one argues that the reasoning is meant to be inductive, then the fallacy is actually still committed. Inductive reasoning could allow one to move from “effects” to a likely “cause.” Affirming the consequent, though, is committed in the step of deciding what constitutes evidence for thinking. Human-like behavior simply is not a sufficient condition for thinking (under any sufficiently rigorous definition with some basis in the terms common usage, as a subjective process), so the Turing test does not provide evidence that a machine likely thinks.

It is kind of like starting with a hypothesis about what quantity of water would constitute a critical mass. One posits some amount of water, and then keeps adding it to a pool, then a lake, then an ocean… in order to test the hypothesis. But of course, there never is a sufficiently large quantity of water to constitute a critical mass. It is a case where the evidence being gathered does not even suggest the desired conclusion.

The analogy breaks down in that eventually we might start to suspect that our hypothesis was bad. In the case of artificial intelligence, it’s even worse: we are searching for a phenomenon that is fundamentally subjective by any coherent account, with a test that can’t even test for subjective phenomena. It’s a fool’s errand. And it won’t do to say to the skeptics, “That’s not fair: we don’t know of any tests for subjective phenomena. Why don’t you find an epistemological method to prove that you think?” because, of course, the skeptics of the Turing test aren’t the ones making a positive claim, and its the fault of the epistemology behind the Turing test that it could not in principle find what it has set out for.

 

[Jewel34]

polytropos:

You say it has nothing to do with morality… but the thought experiment posed was a case of torturing a human-like entity. I’m not sure why the scenario is supposed to appear morally egregious if morality is not relevant to it.

Ok. Valid point. I am sure we can agree that the “first approximation” of a carved, wooden dummy can be beaten, or chopped to pieces, and there is nothing morally objectionable about it. According to your remark in post #88:

“I do not, since I have never claimed that my judgments are not defeasible. If I’m fooled, then I’m fooled; I imagine I’d call it a human and that would be that. If it were convincing enough, as your scenario posits, then I would probably not believe someone who claimed that it was in fact artificial. None of this has any bearing on the ontology of whether the being is intelligent, thinks, or feels; I would just be wrong in thinking that it did (like I might be fooled if a good actor pretends to be in pain).”

It seems to me that you might object to the practice of torturing the dummy as long as you thought that it was a human being, but would cease to object if you had convincing evidence that the weeping, the begging, the bleeding (etc…) is only an emulation. Am I right in this assumption?

This seems to be supported by your current observation:

The presence of intelligence and pain are to the point because they answer one of your questions; in actuality, it does not cease to be emulation and become reality. At no point, no matter how convincing the dummy is made, is one torturing a real human being.

Apart from the fact that the definition of a “realsup[/sup]” human being is missing, there is a moral imperative of not to torture any being with a sufficiently developed nervous system which registers “pain” (to gain some “sadistic pleasure” from gratuitously inflicting such pain). There is no problem with “torturing” water, or a rock, or a wooden human-like dummy.

Obviously that ever-enhanced dummy will navigate to be an ever-more-precise human-look-alike (and not a dog… for example). The way I read your response is that if this process would take place in front of your eyes, and so you would be aware that the final “approximation” is still “just” an emulation, you would not object to the torturing process. As you said in post #88: “None of this has any bearing on the ontology of whether the being is intelligent, thinks, or feels; I would just be wrong in thinking that it did”. Clearly here you speak of a “philosophical zombie” (which has a nervous system, nerve endings, a pain center in the brain, but does not “feel” any pain, only acts as if it did). The problem is that this “zombie” is exactly as impossible as the oft-repeated copper wire “look-alike”, which cannot be distinguished from the real copper, but which would not conduct electricity. As such, your position is grounded on a physical impossibility.

 

[inocente]

That is, however, not my conclusion, nor my argument. I am saying that intelligence, by any credible definition, is fundamentally subjective, and the lack of a necessary connection between intelligence and human-like behavior imposes limits in principle on what the Turing test could ever show about the subjective state of those it observes. You are trying to make my argument into a positive claim that only humans can think; but it is actually the Turing-test proponent who needs to make the positive claim, and who is up the creek without a paddle if his argument just doesn’t follow.

That can’t be. If intelligence was fundamentally subjective then any of us could feel as intelligent as Einstein and win a bevvy of Nobels.

I bring up camcorders because I think we can all agree that visual data is present in their system, but they do not have qualia. If you have a theoretical model for why the addition of complexity and the handling of other sensory (name removed by moderator)uts makes a difference between the more complicated computer and the camcorder, then you may provide it. It is not clear why hierarchical pattern matching and prioritization mechanisms should do this. I agree that they might be necessary, but there is just no argument as to why they should be regarded as sufficient. (There are certainly complex nonintelligent prioritization systems. The question is, what distinguishes them?)*

My argument is simply that qualia are not magic, or to generalize, that there is no such thing as magic.

The point I was making was that denying that qualia exist is tantamount to denying that you are a personal subject who has qualitative experience. I find denying them to be incoherent (unless, of course, you are not a personal subject with qualitative experiences). It’s like someone saying, “Prove to me that you think” or “Prove to me that you don’t have an inverted spectrum.” I don’t think I can do that for you, but if you want to deny it, I am willing to say, “Go ahead.”

I don’t think anyone would deny we have subjective experiences, what is problematic is the invention of qualia. It is a self-fulfilling prophecy that inventing things which can’t be observed, quantified or explained means, by definition, they can’t be observed, quantified or explained.

 

[inocente]

So basically… Quine’s position.

No, I’m saying there’s no point in his navel gazing.

Are you a subject who experiences things like the color red?

The stickies say you can’t answer a question with a question. Sounds like I’m on a winner here so I’ll ask again - You claimed “qualia are perfectly falsifiable”. Come on then, where’s the experiment?

It seems like “some correlate” is much too weak a claim, since I don’t dispute that correlates of thoughts can be detected (that’s indeed what the studies you cited did show). Since whether thoughts in themselves can be logged is what is at issue, we won’t be able to take that on bare assertion.

By correlate I mean that presumably there are various grammars and syntaxes to thoughts - it’s unlikely that when playing tennis or composing a tune, your thoughts of playing the ball or adding a note are English sentences.

It seems doubtful that someone could even construct an experiment to show that thoughts exist. As I’ve said before, on its face, it seems like a category mistake. So the logical positivist is left with redefining thinking in weaker terms more generous to his theory or becoming an eliminativist.

The trouble with all the philosophical words used (mind, consciousness, thought, qualia, etc.) is that they are either general and bland, or else terms of art with multiple definitions. A scientific explanation will either define them far more precisely or drop them in favor of a precise language.

No one cares what logical positivists thought, as that’s yet another failed philosophy to add to the pile.

Well, I’ve been saying “qualitative experiences” since that is the definition of qualia. I don’t think I’ve been saying that experiences are qualia, full stop (doing a calculation in my head, for instance, wouldn’t be qualitative). If I did, it was probably stated loosely, for which I apologize.

No, it wasn’t you, I mean that concepts like qualia don’t add anything to the party.

 

[inocente]

It seems like someone could only deny qualia if they don’t understand the concept. They seem to me to be a basic aspect of being a human - like thinking itself. Their intransigence to science seems to be what leads materialists like Daniel Dennett and the Churchlands to deny them, whereas the inability of neuroscience to discover them should simply make one skeptical of materialism.

Substituting, your argument is that those who don’t believe in invisible pink unicorns have failed to understand the concept, and if they can’t discover any evidence for invisible pink unicorns either then there’s no hope for them.

This is perhaps not your best argument ever.

 

[Peter_Plato]

Pilate’s skepticism:

If you don’t have a reason find someone innocent for a lack of evidence, proclaim them guilty with an ill-formed appeal to “What is truth?” and a washing of hands.

Judas’ skepticism::

When you don’t have the personal integrity necessary to side with the truth of Being sell your integrity for hard evidence or cash (whichever is more compelling.)

Either way, if truth takes tangible form in front of your eyes, cite “lack of evidence” and go for the hard cash, gold or silver, which is eminently “objective” and convertible to any currency.

Interesting that Jesus’ words were not, “You must lose yourself. Period.” They were better translated as: "You must be willing lose your self when faced with Truth because then you will trade a false notion of self to find your true self. Simply denying “self” completely for the sake of “evidential proof” is unconscionable.