What would be the number one requirement?

[Jewel34]

inocente:

Turing called it the Imitation Game rather than a test. The machine’s objective is to fool the interrogator, so yes deception is key to playing the game.

That is one way to look at it. But it is not the only way. Suppose that in the future we can build actual “machines” which look like and act like human beings, true humaniform “robots”. Asimov, who was the best to explore these questions presented a story in “I Robot”, the title is “Evidence”. Its plot revolves around the “unmasking” of a possibly robot, possibly human “entity”, who scrupulously protects his identity. And the reason is to spare the humans the embarrassment lest they would find out that the best leader is “only” a positronic robot, who embodies the best traits a human might have.

inocente:

The machine is isolated and communicates by texting, so it only needs to pretend to have human traits, it doesn’t actually need to possess them to win.

This brings up the question: “where does the emulation end and reality begin?”. If the “emulation” is so perfect, that is it cannot be detected, then it makes no sense to talk about emulation any more.

inocente:

We are learning that many of our traits are also displayed by other species of social mammals, so I wouldn’t have thought they’re anything to be ashamed about. Wouldn’t it be something to celebrate if and when human ingenuity builds a machine that can win the game?

I would love it.

---

Bahman:

Here is another version of argument:

1. To lie you have to construct a subject (lie) which oppose the real subject (truth)
2. This requires semantic (understand the meaning of subject)
3. Computer work is based on syntax
4. Syntax does not suffice semantic (following chines room thought experiment)
5. Hence computer cannot lie

What meaning is “meaning” is a little tricky but you can find a clear explanation here.
And the main Chinese room outcome is “Syntax does not suffice semantic”. You can read the details on the provided link.

Sorry, I presented my questions, and I am interested in your answers, not some links. What is “understanding”, what is “meaning”, and how do we know that our conversation partner understood what we say? I am genuinely interested in what you have to say, and totally uninterested in what other people might say.

The phrase “mental image” simply stands for the internal structure of the entity involved.

We deal with information exchange between two partners. The exchange of information takes place in a communication channel. The sender of the information has his own mental “image” of the information he is sending. The receiver of the information interprets the information, and if his mental “image” is sufficiently close to that of the sender’s then we speak of understanding.

The meaning of the information is simply the mental image that the information conjures up in the receiver. None of this presupposes a biological environment. The “syntactical rules” only make the understanding process easier, but the language is so redundant with extra information that even with somewhat distorted syntax one can (usually) decipher the “meaning” of the information. (An interesting example is this: “take a piece of white paper and cover up the bottom part of each line in the paper. You will be able to read it even with such distorted “syntax”.)

Finally, we know that the receiver “understood” the information, if the continued exchange will reveal this. A very simple analogy from the IBM mainframes: In a communication channel there are two instructions “RUOK” and “IMOK” which stand for “are you ok?” and “I am ok”. When both computers send and receive these simple messages, they both “know” that the communication channel is open and operational.

 

[PseuTonym]

That is simply not true. Ordinary humans will not pass a “Mensa-type genius test”, but that is not what the Turing test is all about. The Turing test is not about extraordinary intelligence.

The numerals were a clue that I was trying to refer to a different test. Perhaps I should have called it “the HAL9000 test”, named after the machine HAL9000. (Note the numerals in that name). I foresee that HAL9000 will propose that test. We distinguish it from “the Turing test”, which was proposed by a man named “Turing.”

We cannot pass the HAL9000 test because our interrogator can simply challenge us to a game of chess. Being unable to play as well as a machine, I cannot fool the machines into thinking “PseuTonym is a machine.”

 

[Jewel34]

The numerals were a clue that I was trying to refer to a different test. Perhaps I should have called it “the HAL9000 test”, named after the machine HAL9000. (Note the numerals in that name). I foresee that HAL9000 will propose that test. We distinguish it from “the Turing test”, which was proposed by a man named “Turing.”

We cannot pass the HAL9000 test because our interrogator can simply challenge us to a game of chess. Being unable to play as well as a machine, I cannot fool the machines into thinking “PseuTonym is a machine.”

This is obvious. A genius can pretend to be a moron, but a moron cannot pretend to be a genius.

 

[Bahman]

Sorry, I presented my questions, and I am interested in your answers, not some links. What is “understanding”, what is “meaning”, and how do we know that our conversation partner understood what we say? I am genuinely interested in what you have to say, and totally uninterested in what other people might say.

Understanding is the ability to convert information (which differs from knowledge), a phrase for example, to a conscious mental state which correlate the current information to understood concepts which are already part of knowledge.

Meaning in regard to language is one to one relation between a piece of information with other understandable form of information. Giving meaning requires a conscious mental process which breaks knowledge related to a piece of information to simpler and understandable form of information. It works in opposite direction of understanding.

 

[PseuTonym]

a moron cannot pretend to be a genius.

If it cannot be done, then how did I do it? Never underestimate the power of human stupidity.

 

[tonyrey]

I am open to accept more if and only if it is necessary to explain what mind is.

The mind is the aspect of a person which has an unmechanistic, creative power of insight, free will, a conscience and a capacity for unselfish love!

 

[polytropos]

I write a program as follows: It prints, “Integer multiplier: enter two numbers.” The user enters two numbers. If the numbers are both 9, then it outputs 82. For any other pair of numbers, it outputs the product.

Is the machine lying to the user when it outputs 82 after the user enters 9, 9? To me that seems doubtful.

One could say that it’s different if the machine is capable of passing the Turing test. But I don’t see why that’s so, for a couple reasons. In this case, it is not merely the returning of what the user didn’t asked for that becomes the condition (“dishonesty” or “stupidity”) for human-likeness, since the conversational ability of the machine is playing a pivotal role in distinguishing it from other “deceptive” programs (contrary to the OP’s thesis that it is the lying which is the fundamental character of a machine which passes the test). The other reason is that this could be exactly how multiplication is defined for the machine which passes the Turing test.

Also, it’s a computer. Suppose that in designing our machine we do something to simulate the effects of consciousness in humans; ie. the (name removed by moderator)uts are in some way processed centrally. When we’re designing its software, we, the software engineers, are at complete liberty to stipulatively divide which operations are accessible to this “conscious” component. And we can make it so that multiplication is not performed by the computer’s central processing, but is simply an accessory feature which overrides the computer’s usual functions when it is given that particular (name removed by moderator)ut.

But then the computer is not deceiving at all. It’s essentially just bad at multiplying. (Actually, a closer analogy would be that it’s a person with a faulty calculator attached to their arm. When you ask them most questions, they answer like any human would. When you ask them another question, somehow the question doesn’t reach their consciousness, but the calculator instead answers for them, often inaccurately. The person here isn’t even stupid or dishonest. Multiplication just ceases to be the sort of questions he answers, although the interrogator gets an answer sometimes, and it turns out to be wrong every once in a while. The details of this analogy could be spelled out in more details, but I think you get the picture.) Even if it “knew” anything, it certainly doesn’t know that it’s providing with the user with an incorrect answer, because as programmers, we are absolutely able to segment whatever particular operation we choose. And the computer would function exactly the same as though none of these stipulative adjustments were made.

One might extend this to a larger refutation of the Turing test as a whole. There seems to be no hard limit on which of its operations we can segment from its “awareness” without changing its outputs. In fact, since the architecture of a modern computer is nothing like a human brain, even assuming that materialism is true, it seems that such divisions and segmentations in processing might be present in any attempt at passing the Turing test. To take the above example of a human with a calculator attached to its arm, a modern computer could behave just the same if you “outsourced” each function to a computational role analogous to the calculator, such that every type question is handled in this manner (accurately, or inaccurately, if you want), and none of the questions reach that conscious center. There need not be a conscious entertainment for the (name removed by moderator)uts and outputs to line up coherently - and given the structure of computers and the state of computer science, this is even likely how computer scientists would design such a machine to pass the Turing test.

 

[polytropos]

Q: what would be the number one requirement for a “machine” to pass the Turing test?
A: to be able to lie!

After all one of the first questions would be a direct approach, asking the tested one: “are you a human?”. And if the other party is not human, and admits it, then the game is over, it failed to be able to pretend that it is a human. Therefore the number one necessity to be able to emulate a human is dishonesty! Is this not sad?
…
Getting real depressing, doesn’t it? To be “human”, one needs to be dishonest, dumb and cruel? And we are the crown of creation?

What are your thoughts?

I don’t find the “dishonesty” observation very compelling.

Q: what would be the number one requirement for polytropos to convince Jewel that he’s inocente?
A: to be able to lie!

No need for depression. That doesn’t mean that to be inocente is to be a liar. It just means that to pretend to be something that you’re not is to be a liar.

Furthermore, I’d hope I wouldn’t have to pretend to be stupid or cruel to fool you into thinking that I’m inocente. Or are inocente and I, as humans, just equally stupid and cruel?

 

[inocente]

That is one way to look at it. But it is not the only way. Suppose that in the future we can build actual “machines” which look like and act like human beings, true humaniform “robots”. Asimov, who was the best to explore these questions presented a story in “I Robot”, the title is “Evidence”. Its plot revolves around the “unmasking” of a possibly robot, possibly human “entity”, who scrupulously protects his identity. And the reason is to spare the humans the embarrassment lest they would find out that the best leader is “only” a positronic robot, who embodies the best traits a human might have.

Well no, it isn’t one way to look at it, it’s the only way to look at it. I linked Turing’s paper, it’s got nothing to do with sci-fi robots. Turing’s game only needs to be played by the very first machine which its designers claim can think.

Once the idea that machines can think is accepted, there’s no point in testing other machines. We don’t test every child to see if they can think and there’s no reason why we’d test every machine either.

This brings up the question: “where does the emulation end and reality begin?”. If the “emulation” is so perfect, that is it cannot be detected, then it makes no sense to talk about emulation any more.

No one was talking about emulation. Have you read Turing’s paper? He was attempting to answer the question of how could we tell if a machine can think, but without getting caught up in the meaning of words. He writes:

I propose to consider the question, “Can machines think?” This should begin with definitions of the meaning of the terms “machine” and “think.” The definitions might be framed so as to reflect so far as possible the normal use of the words, but this attitude is dangerous, If the meaning of the words “machine” and “think” are to be found by examining how they are commonly used it is difficult to escape the conclusion that the meaning and the answer to the question, “Can machines think?” is to be sought in a statistical survey such as a Gallup poll. But this is absurd. Instead of attempting such a definition I shall replace the question by another, which is closely related to it and is expressed in relatively unambiguous words. - loebner.net/Prizef/TuringArticle.html

I would love it.

Me too.

 

[inocente]

I don’t find the “dishonesty” observation very compelling.

Q: what would be the number one requirement for polytropos to convince Jewel that he’s inocente?
A: to be able to lie!

No need for depression. That doesn’t mean that to be inocente is to be a liar. It just means that to pretend to be something that you’re not is to be a liar.

Furthermore, I’d hope I wouldn’t have to pretend to be stupid or cruel to fool you into thinking that I’m inocente. Or are inocente and I, as humans, just equally stupid and cruel?

It’s interesting (or at least to me) that these days the internet gives us all the opportunity to take on different personas. No one is necessarily who they claim to be, everyone is to some extent playing Turing’s game. We do not actually know even that we are talking to other humans.



en.wikipedia.org/wiki/On_the_Internet,_nobody_knows_you%27re_a_dog

 

[Paddy_Walker]

Well, I don’t get it. If you raise two animals together of different species as pets they will probably both think they belong to the other species. If you raise a computer intelligence in a human environment it too will probably think its human (and the human may believe its part of a computer program). So the computer won’t be lying it will just be honestly behaving human-ly. And only a human with, y’know, free will, morality and sin etc., can actually lie. A machine would just be mimicing whatever is the behaviour of humans.

 

[Bahman]

That is one way to look at it. But it is not the only way. Suppose that in the future we can build actual “machines” which look like and act like human beings, true humaniform “robots”. Asimov, who was the best to explore these questions presented a story in “I Robot”, the title is “Evidence”. Its plot revolves around the “unmasking” of a possibly robot, possibly human “entity”, who scrupulously protects his identity. And the reason is to spare the humans the embarrassment lest they would find out that the best leader is “only” a positronic robot, who embodies the best traits a human might have.

What do you mean with identity? That is consciousness which grants identity. To best my knowledge there is problem in definition of consciousness, its function, how it is produced by brain or brain+soul.

This brings up the question: “where does the emulation end and reality begin?”. If the “emulation” is so perfect, that is it cannot be detected, then it makes no sense to talk about emulation any more.

What do you mean by real? Do you mean that machine become conscious?