Ghosts and Mind Copies - The Identity Question

[jeff_wilson]

Originally Posted by whswhs

That sounds like the bit in Job where God lets Satan kill all of Job's children—but after Job affirms his ultimate faith in God, God gives him another set of children, so it's all right. I always thought that was really creepy.

From the viewpoint of Hebrews in antiquity, having your entire family die by an act or acts of God is not an unusual occurrence. The sudden change of fortunes for the better following such an event probably was, though.

[sir_pudding]

Originally Posted by tantric

Frankly, I don't buy AI at all. They might be *useful* and even smart, but not aware.

I can't think of any thing that can support this, that doesn't also "prove" that humans aren't actually aware either. I'm fairly convinced that human consciousness, identity, whatever you want to call is an emergent property of a very complex system. I see no reason why a similar system composed of computer hardware and software can't generate an analogous emergent property, theoretically anyway.

[Flyndaran]

Originally Posted by sir_pudding

I can't think of any thing that can support this, that doesn't also "prove" that humans aren't actually aware either. I'm fairly convinced that human consciousness, identity, whatever you want to call is an emergent property of a very complex system. I see no reason why a similar system composed of computer hardware and software can't generate an analogous emergent property, theoretically anyway.

I'd sooner believe that we aren't truly aware as it's often meant than consciousness is somehow intrinsically a feature of "wet" biology.

[sir_pudding]

Originally Posted by Flyndaran

I'd sooner believe that we aren't truly aware as it's often meant than consciousness is somehow intrinsically a feature of "wet" biology.

Yeah, as far as I can tell the Chinese Room means that nobody "actually" speaks Chinese; which tells me that either the premise of the experiment is flawed, or that I don't actually understand it.

[jeff_wilson]

Originally Posted by sir_pudding

Yeah, as far as I can tell the Chinese Room means that nobody "actually" speaks Chinese; which tells me that either the premise of the experiment is flawed, or that I don't actually understand it.

eh? The people outside the room speak it, and the person who wrote the instructions for the room operator spoke it.

[sir_pudding]

Originally Posted by jeff_wilson

eh? The people outside the room speak it, and the person who wrote the instructions for the room operator spoke it.

Or do they have a sophisticated system for parsing the language and an internal lexicon of Chinese words? Don't they just have a set of instructions for speaking Chinese, just in their memories and speech centers rather than a filing cabinet. If the expirement proves that an AI doesn't actually speak Chinese because it is just using an internal set of instructions to do so, how is a Chinese speaker any different?

[jeff_wilson]

Originally Posted by sir_pudding

Or do they have a sophisticated system for parsing the language and an internal lexicon of Chinese words? Don't they just have a set of instructions for speaking Chinese, just in their memories and speech centers rather than a filing cabinet. If the expirement proves that an AI doesn't actually speak Chinese because it is just using an internal set of instructions to do so, how is a Chinese speaker any different?

A Chinese speaker encompasses all parts of the Chinese language speaking process, even if in the case of the room the process is done by parts (the operator, the reference books, and the author) that individually don't have full comprehension or expression:

I. The author knows Chinese and the language of the operator but does not know what conversation will take place.
II. The operator can see what characters are coming and going and knows how to look them up and see what characters to respond with, but does not know what they mean, and
III. The books contain the rules for how to generate outputs based on inputs, but have no capacity to act.

[sir_pudding]

As I said, either it's flawed or I don't understand it.

Originally Posted by jeff_wilson

A Chinese speaker encompasses all parts of the Chinese language speaking process,

Why can't you say the same thing about the Chinese speaking AI?

even if in the case of the room the process is done by parts (the operator, the reference books, and the author) that individually don't have full comprehension or expression:

Can't you break the Chinese speakers brain out into parts too and equate those processes with the parts of the Chinese room, just as the original experiment attempts to equate the room with the parts of a Chinese speaking AI? Say that the speech centers are like the instructions and the hippocampus is like the filing cabinet and the lexical nodes are like the dictionaries, and the prefrontal cortex is like the operator and so on. So the Chinese speaker is just another kind of Chinese room (like the AI) and nobody "actually" speaks Chinese.

II. The operator can see what characters are coming and going and knows how to look them up and see what characters to respond with, but does not know what they mean, and

Your visual cortex alone doesn't know what written words mean, this is pretty clear from brain damage studies.

III. The books contain the rules for how to generate outputs based on inputs, but have no capacity to act.

Your speech centers alone appear to have the ability to generate outputs based on inputs but aren't capable of generating intention by themselves, as far as I understand the neurobiology. This is why so much human speech is able to be done with minimal conscious thought and people are able to "speak before they think", right?

[jeff_wilson]

Originally Posted by sir_pudding

Why can't you say the same thing about the Chinese speaking AI?

You can. When I said, "in the case of the Chinese room," I was treating it as a case of a Chinese speaker.

Originally Posted by sir_pudding

Can't you break the Chinese speakers brain out into parts too and equate those processes with the parts of the Chinese room, just as the original experiment attempts to equate the room with the parts of a Chinese speaking AI?

Mostly you can, but the comparison is not exact.

Originally Posted by sir_pudding

So the Chinese speaker is just another kind of Chinese room (like the AI) and nobody "actually" speaks Chinese.

I think the point of the exercise is that the reference materials are supposed to be the product of an AI research effort. This would require a very adept Chinese speaker who also is skilled at articulating each step of a hypothetical process for interpreting and responding in Chinese if a person were to do it. If a machine can reproduce it, and it produces typical conversation responses, it does not seem supportable to say the machine doesn't really understand Chinese.-0

Originally Posted by sir_pudding

Your visual cortex alone doesn't know what written words mean, this is pretty clear from brain damage studies.
Your speech centers alone appear to have the ability to generate outputs based on inputs but aren't capable of generating intention by themselves, as far as I understand the neurobiology. This is why so much human speech is able to be done with minimal conscious thought and people are able to "speak before they think", right?

I think so.

[sir_pudding]

Originally Posted by jeff_wilson

You can. When I said, "in the case of the Chinese room," I was treating it as a case of a Chinese speaker.

Isn't the point of the thought experiment to demonstrate that an AI that can pass a Turing test isn't really sapient? Because the procedure carried out by the AI and the one carried out by the room are supposed to be equivalent?

Mostly you can, but the comparison is not exact.

Is the comparison of the Chinese Room and the Chinese AI exact? If not then why isn't this a flaw in the Chinese Room's conclusion?

I think the point of the exercise is that the reference materials are supposed to be the product of an AI research effort.

Well, yes.

This would require a very adept Chinese speaker who also is skilled at articulating each step of a hypothetical process for interpreting and responding in Chinese if a person were to do it.

I'm not sure how this bit is relevant. Isn't that speaker's brain also reducible to a set of interconnected functions? Like the room? Like the AI?

If a machine can reproduce it, and it produces typical conversation responses, it does not seem supportable to say the machine doesn't really understand Chinese.-0

Isn't that the entire point of the thought experiment? It purports to show that the machine, like the room, doesn't really understand Chinese (or by extension any language).

I think so.

Hence "Nobody 'actually' speaks Chinese" is the only conclusion I can derive. Which is absurd. Which means either I don't understand it (most likely) or it's possibly seriously and obviously flawed.