The Ultra Intelligent Machine

[wayfriend]

Machine Learning is only one component of AI. It's not intelligence, it's gaining knowledge. I refer here to the difference between intelligence and knowledge. Programmers have discovered how to gain knowledge - in limited circumstances - without understanding anything. A powerful trick, and useful in the way that algebra is useful, as a tool. One which, as it happens, everyone wants to exploit. Who doesn't want knowledge without the pesky cost of learning anything? Only the wise.

[Skyweir]

good points Wayfriend ..

And wow V .. youre working on a paper or a book for publishing .. thats pretty damn cool. All power to you ..

I do agree with Wayfriend that its not intelligence per se. What is interesting is that when an AI is programmed .. is there any indication that an AI can move beyond its program?

[wayfriend]

is there any indication that an AI can move beyond its program?

A program cannot move beyond it's program ... there is only the variation in the freedom that the program has been given. In other words, how specific or how general is it's abilities.

This is why non-deterministic impulses always become part of the AI discussion when considering AI as self-programming. As if true free will derives from RNG. Mankind will need to break out of the false dichotomy of determinism vs randomness before we'll solve that one.

[Fist and Faith]

What's RNG?

[Avatar]

Random Number Generation

--A

[Fist and Faith]

Ah. Thanks.

[Esmer]

Q) what is molybdenum?

A) 42

#LedZeppelin

[Fist and Faith]

The way I see it, wf, the physical brain is made of all those neurons that release ions into synapses under certain conditions. Along with various other processes that are materially reducible. Yet consciousness is not materially reducible. How is it that consciousness is dependent on, yet free of, the physical? And since it is, how can we create a counterpart in the electronic realm?

I don't know about randomness and free will. If the decision I make between two very attractive choices is not the result of any system of determinism, and I don't randomly pick one, then how do I make it? I suppose it could be randomly chosen for me, via a chaotic process. That wouldn't be determinism or my choosing randomly. But it wouldn't be free will, either.

[wayfriend]

Without rehashing old arguments .... it's clear that, until we figure out what the source of consciousness is, whether it reside in the physical or not, we're not going to replicate it in a machine or in software. Not unless there is a happy accident. In the mean time, people strive (in their ideas, anyway) to create an illusion of free will by using randomness, which is then filtered by feedback and machine learning until it produces something which is both intelligently directed and yet unpredictable. (Like a chess program which calculates the best three moves to make and then randomly picks one.) However, that to me is not free will, just the illusion of it. You can be a slave to random numbers as much as you can be a slave to deterministic causality.

To rehash old arguments ... I summarized mine quite well here.

[Zarathustra]

How is it that consciousness is dependent on, yet free of, the physical? And since it is, how can we create a counterpart in the electronic realm?

I recently used a metaphor of a language, as a system of rules, vs a novel, as an instance of applying those rules. Any meaningful speech act, or written work, is going to depend upon a set of rules for its mere existence. Without those grammar rules, that syntax, it's impossible for meaningful utterances to exist (just as it's impossible for minds to exist without the physical laws that produce brains). And yet, novels are not determined by that grammar, merely because they manifest through it and depend upon it for their very existence.

Likewise, just because physics and chemistry are the rules of this universe doesn't mean that those rules are deterministic of minds. It's true that anything that exists will be dependent upon those rules for its existence, and its "shape" will conform to those rules, but that's not the same as being determined by the rules . . . no more than a book being determined by grammar.

Now, here's the catch: a book is written by a conscious mind. So we already have within this metaphor the phenomenon we're trying to model or explain by virtue of comparison. The reason that novels cannot be predicted by studying grammar rules is because minds use these rules to introduce novelty (specifically, meaning; see below). So, if the metaphor can be applied to reality, we must ask what introduces novelty into the universe while simultaneously making use of rules?

I'm tempted to say it's the very same phenomenon: mind, consciousness. But that feels circular, since the rules in this instance "create" the mind. But a mind is not merely consciousness, but always/necessarily consciousness OF something. That means it has content, not merely bare existence. In fact, it cannot exist without content. Thus its "bare existence" requires that it has a meaning, or meaningfulness (i.e. content). That content is ideal, not actual.

And here's where I hang my argument: the content of consciousness is not strictly caused from the bottom-up by physical processes, but often (if not always) caused "sideways" from the content of previous states of consciousness. So we are like a novel writing itself. Indeed, when thinking verbally, our thoughts are following grammar rules more than they are following the rules of physics. The content of consciousness in this case is pure linguistic meaning, not mathematical formulas that trace neural firing. But grammar rules are not in physics--nowhere to be found in them! And yet our thoughts are shaped into these patterns that have nothing to do with the laws of nature.

The key is meaning. The laws of nature can build things that are capable of discerning an entirely different, "higher" level of reality. That level is the meaning of being, not mere being itself. Ideal reality. That doesn't mean fictional, fake, or fanciful. The ideal is real. And in us, it can be connected to physical actuality. We are the connecting point. And in that connection lies the paradox, the "miracle" of free will and all the rest. Causation stops being purely bottom-up at this point, once matter touches "heaven."

Now, the question is, how do we build a machine that discerns meaning? How do we write a program--which is always pure syntax--that can also understand semantics? I don't think it's possible. Consciousness does not come from syntax. It arises from experiencing the world, being embodied, being alive.

[Esmer]

a book filled with blank pages is still a book. it isn't content that determines it's actuality but it's capacity for the potential that exists within it.

an empty mind is still a conscious one. you don't have to think to be alive, and being alive isn't dependent upon whether or not you believe you are, or not.

reality is that which, when you stop believing in it, doesn't go away.

the difference lies between knowing and knowing what you know

what is the difference between consciousness and thought?

what's the difference between a cock and a rooster? they both have big mouths in the morning

gnosis.org/naghamm/thunder.html

[Zarathustra]

Esmer, those are some interesting quotes. However, while a book with blank pages might technically still be a book, it is not a novel. And though a mind that is blank of thoughts may still be conscious, consciousness is always necessarily consciousness of something. This does not have to be consciousness of a thought, it is merely the structure of consciousness, the phenomenological structure. Consciousness is intentional, characterized by intentionality, meaning that it is always directed towards an object, whether that object is a thought, a feeling, a sensation, or being itself.

[Fist and Faith]

How is it that consciousness is dependent on, yet free of, the physical? And since it is, how can we create a counterpart in the electronic realm?

I recently used a metaphor of a language, as a system of rules, vs a novel, as an instance of applying those rules. Any meaningful speech act, or written work, is going to depend upon a set of rules for its mere existence. Without those grammar rules, that syntax, it's impossible for meaningful utterances to exist (just as it's impossible for minds to exist without the physical laws that produce brains). And yet, novels are not determined by that grammar, merely because they manifest through it and depend upon it for their very existence.

Likewise, just because physics and chemistry are the rules of this universe doesn't mean that those rules are deterministic of minds. It's true that anything that exists will be dependent upon those rules for its existence, and its "shape" will conform to those rules, but that's not the same as being determined by the rules . . . no more than a book being determined by grammar.

Now, here's the catch: a book is written by a conscious mind. So we already have within this metaphor the phenomenon we're trying to model or explain by virtue of comparison. The reason that novels cannot be predicted by studying grammar rules is because minds use these rules to introduce novelty (specifically, meaning; see below). So, if the metaphor can be applied to reality, we must ask what introduces novelty into the universe while simultaneously making use of rules?

I'm tempted to say it's the very same phenomenon: mind, consciousness. But that feels circular, since the rules in this instance "create" the mind. But a mind is not merely consciousness, but always/necessarily consciousness OF something. That means it has content, not merely bare existence. In fact, it cannot exist without content. Thus its "bare existence" requires that it has a meaning, or meaningfulness (i.e. content). That content is ideal, not actual.

And here's where I hang my argument: the content of consciousness is not strictly caused from the bottom-up by physical processes, but often (if not always) caused "sideways" from the content of previous states of consciousness. So we are like a novel writing itself. Indeed, when thinking verbally, our thoughts are following grammar rules more than they are following the rules of physics. The content of consciousness in this case is pure linguistic meaning, not mathematical formulas that trace neural firing. But grammar rules are not in physics--nowhere to be found in them! And yet our thoughts are shaped into these patterns that have nothing to do with the laws of nature.

The key is meaning. The laws of nature can build things that are capable of discerning an entirely different, "higher" level of reality. That level is the meaning of being, not mere being itself. Ideal reality. That doesn't mean fictional, fake, or fanciful. The ideal is real. And in us, it can be connected to physical actuality. We are the connecting point. And in that connection lies the paradox, the "miracle" of free will and all the rest. Causation stops being purely bottom-up at this point, once matter touches "heaven."

Now, the question is, how do we build a machine that discerns meaning? How do we write a program--which is always pure syntax--that can also understand semantics? I don't think it's possible. Consciousness does not come from syntax. It arises from experiencing the world, being embodied, being alive.

I'm never on firm ground in these conversations. But I don't see how this addresses how the mind can be free of material reductionism. Which it is. The brain is syntax. It's all particles and groups of particles interacting in the ways they must. Yet it is somehow also the mind, free to pursue thoughts other than those that are thrust upon it, and also able to initiate a series of interactions of the particles to achieve what it wants. It's like the words in the book turning into the physical objects they are describing.

And if it can happen in one medium, why would it not be possible in another?

[wayfriend]

And if it can happen in one medium, why would it not be possible in another?

I have no doubt that it could. Perhaps we will create it. Perhaps we will find it. Perhaps it will find us!

I have a strong feeling, though, that it's not possible with software. Because software isn't a 'medium', it just calculates what an imaginary medium would do.

[Zarathustra]

I don't see how this addresses how the mind can be free of material reductionism. Which it is. The brain is syntax.

The brain is physical and actual. That's not syntax, which is merely the rules of sentence formation. While there may be "rules" of how particles interact (I'd say patterns, not rules), the interaction of particles is nothing at all like the formation of sentences. Sentences convey meaning through symbols and words. Particles just exist and interact.

It's all particles and groups of particles interacting in the ways they must. Yet it is somehow also the mind, free to pursue thoughts other than those that are thrust upon it, and also able to initiate a series of interactions of the particles to achieve what it wants. It's like the words in the book turning into the physical objects they are describing.

Not sure about that. A book is made up of particles, too. Where it transcends its physical incarnation is in its meaning. That's where mind transcends its physical incarnation, too.

And if it can happen in one medium, why would it not be possible in another?

It is possible in another. Computers contain information that transcends their physical incarnations, just like novels. But information doesn't have meaning without a mind to understand that information, just like a novel requires someone to read it. Can we write novels that can read and understand themselves? No. Nor can we write computer programs that can read and understand themselves.

The medium is only part of the problem. Computers and books aren't alive. That's a huge hurdle. But they also rely solely on syntax, whereas minds do not. Minds can be aware of the ineffable, that which cannot be expressed in objective language. For instance, qualia. Someone who can't see has no meaning for the word, "red." We are only able to convey this to each other in as much as we both have the same subjective experience. Computers could only manipulate symbols without ever understanding what those symbols mean. Meaning isn't the manipulation of symbols.

[Fist and Faith]

And if it can happen in one medium, why would it not be possible in another?

I have no doubt that it could. Perhaps we will create it. Perhaps we will find it. Perhaps it will find us!

I have a strong feeling, though, that it's not possible with software. Because software isn't a 'medium', it just calculates what an imaginary medium would do.

The medium is the electronic hardware that the software is running on. The hardware is subject to the same properties of the material world that our medium, our brains, are. The software is the equivalent of the properties/behaviors that are specific to the groups of particles known as brains.

[wayfriend]

The medium is the electronic hardware that the software is running on.

It's *a* medium, I will grant you that.

I know it's hard to explain. Let me try this way: You have a brain, you can think. With your thinking brain, you can imagine a ball thrown into the air and landing. You can calculate where the ball would be in any given instant. You can make a CGI video showing a ball flying through the air.

But there is no ball. It's just some calculations about a theoretical ball.

When a computer simulates intelligence, there is no intelligence. It's just some calculations about a theoretical intelligence. What would it answer if we asked it this and it worked like that.

The "brain" doing the calculating is very, very dumb, and very, very deterministic. But the "brain" doing the calculating isn't the intelligence, just as your brain is not the ball.

There is no intelligence. There is only a simulation of how a theoretical intelligence might process input and produce output. It's a calculation of where it would be in any given instant. You can even calculate what sounds it could make and what pictures it would create.

But there is no ball.

My gut tells me that the lack of actual existence is a barrier to developing self-awareness. But that's just my gut.

Now, if you could CHANGE computers so that they didn't calculate what intelligence would do, bit instead DO it ... if you could make the transistors hold memories instead of just holding instructions ... if you could connect them in complex ways that do more than just ADD, MOVE, TEST, and JUMP .... then it might develop self-awareness. Make the chip like an actual brain rather than a primitive instruction-follower.

[Fist and Faith]

I don't see how this addresses how the mind can be free of material reductionism. Which it is. The brain is syntax.

The brain is physical and actual. That's not syntax, which is merely the rules of sentence formation. While there may be "rules" of how particles interact (I'd say patterns, not rules), the interaction of particles is nothing at all like the formation of sentences. Sentences convey meaning through symbols and words. Particles just exist and interact.

Yeah, I was lazy there. I just said what I really meant to wf.

It's all particles and groups of particles interacting in the ways they must. Yet it is somehow also the mind, free to pursue thoughts other than those that are thrust upon it, and also able to initiate a series of interactions of the particles to achieve what it wants. It's like the words in the book turning into the physical objects they are describing.

Not sure about that. A book is made up of particles, too. Where it transcends its physical incarnation is in its meaning. That's where mind transcends its physical incarnation, too.

Yeah, I wasn't too clear this morning. I just meant it as comparing two seemingly impossible things.

And if it can happen in one medium, why would it not be possible in another?

It is possible in another. Computers contain information that transcends their physical incarnations, just like novels. But information doesn't have meaning without a mind to understand that information, just like a novel requires someone to read it. Can we write novels that can read and understand themselves? No. Nor can we write computer programs that can read and understand themselves.

Well, that's the question, eh? And I have strong doubts, too. But maybe we can help them achieve it.

The medium is only part of the problem. Computers and books aren't alive. That's a huge hurdle. But they also rely solely on syntax, whereas minds do not. Minds can be aware of the ineffable, that which cannot be expressed in objective language.

But how? Built from rules as strict as programming, the mind goes beyond it. How? Why do we assume that what happened with particles and the rules that govern their interactions can't happen with electronics and programs? Especially considering nature did it using a hit-and-miss approach over billions of years, whereas we can try different things as often and frequently as we want.

Is experience the key? We can try to set it up in what we think should be the final stage, but we're certainly not going to get it right. But we started as a little connection of input connected with something that gave the ability to react. So make our prototype a simple thing like that. And simulate the input a million times an hour, so it gets the equivalent of what took nature a very long time to do.

Of course, the experience is useless without the ability to change. No change means no mutation and natural selection. I suspect the natural selection would take care of itself. But how do we give it the ability to change? Give it the ability to replicate itself. Can we make a self-replicating program that replicates itself PERFECTLY? I'll bet not. I suspect a program copying itself a billion times will make mistakes. So we give each copy the input a million times an hour, and see what happens.

I'm sure all this has been done. I know there are SciFi stories along these lines. Anybody know any specifics about what's been tried? And guesses about why it didn't amount to anything?

[Zarathustra]

FF, did you (or anyone else) check out the link I gave above? It's an article in the Stanford Encyclopedia of Philosophy on the Chinese Room thought experiment. I believe we have discussed it before. But the rebuttals section is very interesting. It might help you with your conceptual struggle. It actually argues pretty forcefully against my position. Perhaps I was wrong to say that computers only manipulate symbols. They aren't like a book. They are causal engines. They have processes and actions. Inputs can cause outputs. And they don't really manipulate symbols (syntax), they manipulate electricity and then we interpret this in terms of symbols. Now, there is a lot in there that I disagree with, such as the idea that a processor understands what it is doing in following the rules of its programming. But I clearly need to do more thinking about where semantics comes from, and what understanding is.


I think the key is consciousness. We can be conscious of semantics and syntax, but semantics and syntax don't produce our consciousness. So even if there is some sense in which computers transcend syntax (i.e. their programming) to achieve meaning*, this doesn't mean they are conscious.


We are trying to simulate intelligence. And we're making progress in this regard. But we never tried simulating consciousness, even the most rudimentary consciousness we see (or assume!) in simple organisms. May that's the problem. Not only do we not know what consciousness is in ourselves, we can't say for sure that other organisms even have it. Maybe they just have complex behaviors and instincts. So if we can't say this for sure in living creatures, how could we ever say it for a computer _ much less build it?


Whether or not Searle is correct to say that you can't derive semantics from symbol manipulation, I think he is still correct to say that you can have symbol manipulation without understanding or consciousness of that semantics. So maybe computers will one day be intelligent, but they won't know it.

*["Achieving meaning" could be understood (by us) in terms of causal relations i.e. the programming producing meaningful effects.]

[Fist and Faith]

I had seen the Chinese Room before, so didn't check out your link, thinking I knew the point you were making. I'll check it out, and your latest post, tonight.


EDIT: Wow, that link is much more than simply presenting the Chinese Room!