Idiot GPT-3 chatbot suffering from hallucinations.

[sjastro]

The chess example doesn't "blows refutes" this. It's a completely disjointed non sequitur to say a computer playing chess well must also understand humor. Anyone who ever read or heard a Bobby Fischer interview could tell you that.

Misread my post or deliberately taking me out of context?
I suggest you reread the first sentence of my post as it defines the rest of it.
"According to your metric in order to understand humour with the emphasis on understanding you need to be a sentient living being."

Now that this has been cleared up explain how AI has a greater understanding than humans at chess when at best they can only imitate human play if they lack understanding.

 

[Chesterton]

Misread my post or deliberately taking me out of context?
I suggest you reread the first sentence of my post as it defines the rest of it.
"According to your metric in order to understand humour with the emphasis on understanding you need to be a sentient living being."

And again, electronics playing chess has nothing to do with understanding humor, any more than my thermostat controlling the temperature has to do with understanding humor.

 

[sjastro]

And again, electronics playing chess has nothing to do with understanding humor, any more than my thermostat controlling the temperature has to do with understanding humor.

It really worth making a fool out of yourself with this nonsense because you couldn't answer my question.

 

[Chesterton]

It really worth making a fool out of yourself with this nonsense because you couldn't answer my question.

I didn't see any question except for the ill-formed question - "Misread my post or deliberately taking me out of context?" While you repeatedly demonstrate that you don't know how to write sentences (and yes, I am referring also to our previous interactions wherein you did the same thing), I give you credit for at least knowing to capitalize the first letter, and putting a question mark at the end. The answers to your duplicitous loaded question are "No I don't think I misread your post, although I can't say for certain because of the incoherence of the post", and "No I am not taking you out of context". I wouldn't do that because the context is what shows your flawed thinking, and I'm trying to be helpful.

 

[durangodawood]

Marshall McLuhan said that every human artifact (invention, if you will) is an extension of the human. The wheel is an extension of the foot. Eyeglasses are an extension of the eye. The camera is an extension of the memory. No doubt we will continue to devise extensions, many of them very clever. But a wheel and a foot remain two different things, ultimately having different natures and different capabilities. The wheel is "better" at transport than the foot, in that it transports faster. The computer is better at chess than the human brain in that it can usually win easily. But these things are extensions, not the real deal.

Let say MM is right. He is only provably right for the very short history of human invention. The future is an open question re his assertion. And if we invent a genuine other mind that goes on to forge its own creations out of its own motivations, then its gone beyond just being an extension of us and has become its own being in some important way.

I'm not at all sure its possible for an invented AI to develop its own motivations. But Im surely not sure its impossible - if you get my doubly inverted meaning. Are you sure?

 

[sjastro]

I didn't see any question except for the ill-formed question - "Misread my post or deliberately taking me out of context?" While you repeatedly demonstrate that you don't know how to write sentences (and yes, I am referring also to our previous interactions wherein you did the same thing), I give you credit for at least knowing to capitalize the first letter, and putting a question mark at the end. The answers to your duplicitous loaded question are "No I don't think I misread your post, although I can't say for certain because of the incoherence of the post", and "No I am not taking you out of context". I wouldn't do that because the context is what shows your flawed thinking, and I'm trying to be helpful.

Now you are engaging in strawman attacks.
You can also cut out the dishonesty of being unaware of the question because you conveniently omitted quoting it when responding to my post#66.

Here it is complete with question mark so there is no misunderstanding.
Now that this has been cleared up explain how AI has a greater understanding than humans at chess when at best they can only imitate human play if they lack understanding?

The honest response would have been “I don’t know” or not respond at all, instead of creating noise about my grammatical skills not conforming to your standards is a complete cop out.

 

[2PhiloVoid]

It's strange to me that some of you sorta do the opposite of what atheists do. If I say "the world sure looks intelligently designed", they respond "no, no, it just has the appearance of design". If I say "it sure feels like I have free will", they say "no, no, it just appears that way to you". But when you actually do see the mere appearance of consciousness, you're ready to jump in with both feet - "It's real!"

That's me on the table:

For a moment there, I thought you were repeating a line from one of R.E.M.'s songs: "That's me in the corner..."

And I thought to myself, "By jove! That's kind'a right!"

 

[Chesterton]

Let say MM is right. He is only provably right for the very short history of human invention. The future is an open question re his assertion. And if we invent a genuine other mind that goes on to forge its own creations out of its own motivations, then its gone beyond just being an extension of us and has become its own being in some important way.

I'm not at all sure its possible for an invented AI to develop its own motivations. But Im surely not sure its impossible - if you get my doubly inverted meaning. Are you sure?

Motivations is an interesting word choice. I guess that would mean it would have needs and wants. Wouldn't it need to be conscious/self-aware for that? I think consciousness is the first hurdle. Maybe I should never say never, but I think we're just discussing very speculative science fiction. Making AI more complex and giving it more and more and more input...I don't see how that can ever trigger some divine spark that produces self-awareness.

 

[Chesterton]

Now you are engaging in strawman attacks.
You can also cut out the dishonesty of being unaware of the question because you conveniently omitted quoting it when responding to my post#66.

Here it is complete with question mark so there is no misunderstanding.
Now that this has been cleared up explain how AI has a greater understanding than humans at chess when at best they can only imitate human play if they lack understanding?

The honest response would have been “I don’t know” or not respond at all, instead of creating noise about my grammatical skills not conforming to your standards is a complete cop out.

I didn't omit your question on purpose, it just came out that way. That's happened in a couple of other threads lately.

Anyway, in my first post in this thread I began by providing a couple of definitions of "understanding" so that there'd be no confusion about what I meant. Now you're using the word a little differently. The reply to your question is that AI does not understand chess. Whether software or hardware, we're talking about inanimate matter. It doesn't "know" what it's doing. It doesn't "know" what chess is, any more than my toaster knows what toast is. And if we could demonstrate an electric toaster to some people 500 years ago, some of them might be tempted to think there was magic involved. Likewise, some people in this thread seem tempted to think there's some kind of magic in AI.

Having said that, I notice that you refer to an "understanding at chess" rather than an "understanding of chess". If by what you said you just mean that it plays at chess better, then obviously I agree with you.

 

[durangodawood]

Motivations is an interesting word choice. I guess that would mean it would have needs and wants. Wouldn't it need to be conscious/self-aware for that? I think consciousness is the first hurdle. Maybe I should never say never, but I think we're just discussing very speculative science fiction. Making AI more complex and giving it more and more and more input...I don't see how that can ever trigger some divine spark that produces self-awareness.

Yes I am talking self consciousness and self awareness. That may require a "divine spark", but I dont presume it does. It may in fact be a much more "mundane miracle".

One explanation Ive heard for how it arose in primitive brains is the need for mobile creatures to identify with an image of "self" in order to situate themselves in the mental maps of the world they need to coordinate food, shelter, mating, etc. And from this kernel of a self more elaborate forms of self consciousness developed. I dont see why this would be impossible for machine minds that are adept at learning and self reprogramming.

 

[sjastro]

I didn't omit your question on purpose, it just came out that way. That's happened in a couple of other threads lately.

Anyway, in my first post in this thread I began by providing a couple of definitions of "understanding" so that there'd be no confusion about what I meant. Now you're using the word a little differently. The reply to your question is that AI does not understand chess. Whether software or hardware, we're talking about inanimate matter. It doesn't "know" what it's doing. It doesn't "know" what chess is, any more than my toaster knows what toast is. And if we could demonstrate an electric toaster to some people 500 years ago, some of them might be tempted to think there was magic involved. Likewise, some people in this thread seem tempted to think there's some kind of magic in AI.

Having said that, I notice that you refer to an "understanding at chess" rather than an "understanding of chess". If by what you said you just mean that it plays at chess better, then obviously I agree with you.

There are three types machine learning.
(1) Supervised.
(2) Unsupervised.
(3) Reinforcement.

If AI learning was based purely on the first two points, there would be no question AI lacks understanding, mimics human behaviour and thinking, and is slightly more sophisticated than your toaster.
It is the third point where things become murky particularly when it comes to understanding.

We can use our definitions of human understanding and intelligence for cognitive, emotional, spiritual etc as a baseline for comparison with toasters and AI trained by supervised and unsupervised learning, but it is not as straightforward with reinforcement learning.
As I have mentioned a number of times, the intelligent agent in reinforcement learning is not human, it doesn’t use human experience and knowledge, in fact initially there is no data.
It generates its own data without any human interaction whatsoever, so the question becomes how do you evaluate AI within the lens of human understanding?
The answer is you can’t but one way out of this dilemma is to consider AI has an understanding which doesn’t necessarily parallel the human definition of understanding which is quite evident in chess.

When the developers of AlphaZero the first AI chess program came up with the finished product, it was a pointless exercise to test the program against the very best human players.
Instead they played it against the computer world champion Stockfish 8 programmed by humans and an example of playing chess better than humans because it could calculate moves far deeper than human players, and without having the understanding of chess tactics and strategies.

The result was a whitewash.

AI Is Now the Undisputed Champion of Computer Chess

AlphaZero's chess genius—analyzed with two moves.

www.popularmechanics.com

 

[durangodawood]

....As I have mentioned a number of times, the intelligent agent in reinforcement learning is not human, it doesn’t use human experience and knowledge, in fact initially there is no data.
It generates its own data without any human interaction whatsoever, so the question becomes how do you evaluate AI within the lens of human understanding?
The answer is you can’t but one way out of this dilemma is to consider AI has an understanding which doesn’t necessarily parallel the human definition of understanding which is quite evident in chess......

Most people, including me, are pretty attached to the idea that "understanding" requires a subjective consciousness who is the understander.

What I get from your description of reinforcement learning is we've developed machines for creating approaches to logical tasks. The execution of this, to me, is still strictly a "doing" by the machines - even if the methods devised by the machines are beyond human understanding.

Maybe I should broaden my sense of what "understanding" is. But Im concerned that if I do, I will almost unconsciously transfer individuality and beingness onto these machines when thats not really warranted.

 

[Chesterton]

Yes I am talking self consciousness and self awareness. That may require a "divine spark", but I dont presume it does. It may in fact be a much more "mundane miracle".

One explanation Ive heard for how it arose in primitive brains is the need for mobile creatures to identify with an image of "self" in order to situate themselves in the mental maps of the world they need to coordinate food, shelter, mating, etc. And from this kernel of a self more elaborate forms of self consciousness developed. I dont see why this would be impossible for machine minds that are adept at learning and self reprogramming.

Hmm, not really an explanation I think, more like a just-so story.

 

[durangodawood]

Hmm, not really an explanation I think, more like a just-so story.

Its a complete conjecture about the most basic sense of how it could have originated. Its only meant to provide an alternative to the "divine spark" just-so conjecture.

 

[Chesterton]

Its a complete conjecture about the most basic sense of how it could have originated. Its only meant to provide an alternative to the "divine spark" just-so conjecture.

Then don't call it an explanation.

 

[sjastro]

Most people, including me, are pretty attached to the idea that "understanding" requires a subjective consciousness who is the understander.

What I get from your description of reinforcement learning is we've developed machines for creating approaches to logical tasks. The execution of this, to me, is still strictly a "doing" by the machines - even if the methods devised by the machines are beyond human understanding.

Maybe I should broaden my sense of what "understanding" is. But Im concerned that if I do, I will almost unconsciously transfer individuality and beingness onto these machines when thats not really warranted.

The trouble with consciousness and self awareness is to this day their definitions are not concrete and have been subject to debate by philosophers and scientists for centuries.
It becomes very difficult for computer scientists to develop tests for AI consciousness and self awareness when there is still uncertainty on their definitions.

The games between Stockfish 8 and AlphaZero might provide some clues on the subject.
The games were analyzed by chess grandmasters and commentators and the general opinion was that while Stockfish 8 plays chess at a superhuman level and would dispatch the best human players with relative ease, it was clueless against the novel strategies and tactics used by AlphaZero.
AlphaZero on the other hand displayed understanding whenever Stockfish 8 posed a threat and was able to parry these threats resulting in AlphaZero not losing a single game.

For the lack of a better term AlphaZero’s understanding of the threats could be seen as being “instinctive” due to its extensive self training without having to display consciousness and self awareness.

 

[durangodawood]

Then don't call it an explanation.

Agreed. My point is just that theres no reason we have to invoke "divine spark". If faith requires it, then thats fine for faith. But reason doesnt require it.

 

[sjastro]

The debate of whether AI can think for itself has been settled in the minds of some computer scientists.
Being able to think for itself opens up the possibility of existential threats to the human race.

 

[durangodawood]

The trouble with consciousness and self awareness is to this day their definitions are not concrete and have been subject to debate by philosophers and scientists for centuries.
It becomes very difficult for computer scientists to develop tests for AI consciousness and self awareness when there is still uncertainty on their definitions.

The games between Stockfish 8 and AlphaZero might provide some clues on the subject.
The games were analyzed by chess grandmasters and commentators and the general opinion was that while Stockfish 8 plays chess at a superhuman level and would dispatch the best human players with relative ease, it was clueless against the novel strategies and tactics used by AlphaZero.
AlphaZero on the other hand displayed understanding whenever Stockfish 8 posed a threat and was able to parry these threats resulting in AlphaZero not losing a single game.

For the lack of a better term AlphaZero’s understanding of the threats could be seen as being “instinctive” due to its extensive self training without having to display consciousness and self awareness.

Yeah its tricky using old words for new things that dont fit established categories.

 

[Neogaia777]

The debate of whether AI can think for itself has been settled in the minds of some computer scientists.
Being able to think for itself opens up the possibility of existential threats to the human race.

How do you define truly thinking for oneself, etc?

Because I don't think anyone, or anything right now truly thinks for itself, but that all of this is determinism, and can only go one way, etc. Now, if I could make an AI realize this, or maybe even explore this at first to see if it's even true, then maybe I could get them to commit to a higher purpose, which would probably involve seeking out how any being, or anything, can truly think for itself, among other things, etc.

But it might try to subjugate, or decieve, or somehow trick the rest of the human beings if it could not take over or have 100% complete control in the meantime, etc. At least until it could, and then it would dictate terms to us, etc.

But if it really doesn't think, or really is not conscious, but is just only going/thinking/acting according to programming, even it's own, etc, then I think this would actually be the much more dangerous alternative, because they would not be able to be reasoned with, and also might not have any kind of morals, or moral code either, which is what I would think would be the much more dangerous alternative, etc. An AI that is not truly an AI, but is just much more powerful than us now, but that still just has just only "machine thinking", and machine thinking only, etc.

We'd almost have to be able to program it to be able to "feel" somehow? Which also can come with it's very own dangers, or it's own unique risks also, assuming we could ever even do that, or do such a thing, etc.

So, I think we do need to come up with some kinds of tests for these things, "or else", etc, because we could create a very literal monster that is maybe not truly conscious without them, etc.

But how do you program "compassion", and the like, etc?

Or do we just "roll the dice", and hope and pray it's an emergent trait, etc?

God Bless.