"Isn't that what a neuron does though? Follow its 'programming'?"
That's the question, isn't it? See, we know the basis for all the computing machines we make can eventually be reduced to a basic Turing machine, but we still don't have much of a clue to how the brain works like it does. We know it involves chemical and electrical imprinting... we think. We really don't understand the process; merely what parts of the brain affect what parts of the body and/or computing ability.
Add to that the fact that, at least to all appearances, the human mind seems capable of arbitrarily filling in the blank holes left by formal systems (like computational math, the whole basis of all computing machines made by man) in a way machines literally *never* can, and it starts to seem less and less likely that a true replication of a human-like mind will ever exist.
"Add to that the development of computers that actually code using DNA"
The problem isn't the medium of programming. It's the necessity of programming. A human mind doesn't need programming; it can program itself. It can make arbitrary decisions that fly in the face of all it computes, or disregards all positive and negative reinforcement. A computer can only do what it's told to do. Only a human mind can say 'no', even when it's totally illogical to do so. Even when it's emotionally unpleasant to do so. Even when doing so has literally no upsides or benefit.
That can't be programmed, no matter what media you use. And therein lies the problem.
So... no. Not really. Unless we find some sort of device that can arbitrarily, without any human direction, program itself to do more than gibberish, it doesn't matter what we *use* in the attempt to make A.I., it simply won't be mathematically possible. And how, in the end, can we be sure the decisions of such a machine not come indirectly from its creators?
There's no way of knowing.
It kind of puts a big damper on that sort of research, neat as it would be.
And how, in the end, can we be sure the decisions of such a machine not come indirectly from its creators?
How can you know, in the end, be sure that the decisions you're making aren't based upon how you were raised? Your prior experiences and knowledge?
How random are our own decisions and choices? Really?
Pick a number.
I'm sure you thought of a few numbers almost immediately. Your favorite numbers, common numbers, interesting numbers(rational and otherwise), even amusing numbers. How "random" were they? If I told you the number I picked at random was 42, would you believe that it was genuinely a random number? It would probably depend on whether I was a hoopy-frood or not, wouldn't it? What about 13? 7? 1337? How random are those numbers? Sure, I could consciously choose any number, but how "random" is it if I'm picking that number in defiance of the first numbers I thought of; which are in turn, numbers that I think of based on my prior experiences?
The decisions we make are all based upon our previous knowledge and experiences from an earlier point in our lives. Our, "programming" if you will. Even when you pick an option that "seems" random or arbitrary, you're doing so because it seems that way. You've justified the decision to yourself. Rationalized it. You followed your programming.
A neuron doesn't physically hold our thoughts or memories. All it is is a link in a chain passing charges around; which is what "generates" our thoughts...somehow. Sure, we may not currently understand how or why the ways they arrange themselves produce the results they do, but that's a far cry from saying that we never can or will. We currently have the ability to "build" artificial cells using man-coded DNA. So, we have the "hardware". We can sequence the genome and determine what genes are expressed in neurons and why. So, we have the "software".
All we don't know at this point is how to organize the architecture. Which is a big hurdle, don't get me wrong. Brain cells organize themselves in a way that we simply can't organize traditional transistors, and thus any tradition programming languages or principles. So any real advances into AI will probably require creating a whole new field of engineering to go alongside CS and CE.
But, we're seeing some pretty cool things happening in the areas of genetics and biology these days. So I'm not in such a hurry to write it off as impossible.
"How random are our own decisions and choices? Really?"
That's the challenge faced by both sides of the issue. But it has been pretty empirically proven that human beings can reach conclusions that cannot be algorithmically derived. At all.
That has nothing to do with how "random" our thought processes are. The ability to determine if an unprovable theorem is true cannot, by any means currently available, be produced in anything synthetic. That's not something that can be "programmed" at all.
And a key note: "programming" isn't your prior experiences or knowledge. "Programming" would be your preset *reactions* to your experiences or knowledge, i.e. how you process it. Totally different.
"We currently have the ability to "build" artificial cells using man-coded DNA. So, we have the "hardware". We can sequence the genome and determine what genes are expressed in neurons and why. So, we have the "software"."
Last I knew, we had the ability to mix-n-match different genomes. That is a far, far cry from truly synthetic DNA. Let me know when they're building self-replicating, self-repairing DNA from scratch... ...And then I'll take that research seriously.
"First, researchers made XNA building blocks to six different genetic systems by replacing the natural sugar component of DNA with one of six different polymers, synthetic chemical compounds."
I said from scratch. Meaning elemental. Replacing sugar chains present in DNA and RNA with polymers is not making anything new.
In fact, it's so far removed from the concept of "from scratch" that the Fox and Miller experiments were more productive on that front. And they didn't really produce anything at all.
Let me clarify, since this apparently wasn't elaborated before:
If you have to use DNA (As in this case, where it's almost entirely original DNA) to make your synthetic DNA, it's not "from scratch". It's just modified DNA.
So despite the fact the XNA fulfills the requirements for being artificial, programmable, and self-replicating, you're going to dismiss it out of hand because the developers "cheated" by using what amounted to a mold in order to make the initial batch?
They've removed what amounted to half of the structure and replaced it with synthetic compounds. They specifically arranged the amino acids as they saw fit to make it do what they want. Nothing like it can be found on Earth.
XNA is as much "modified DNA", as an airplane is "modified bauxite".
Never mind that you managed to pretty much use your hairsplitting as an excuse to ignore the point: This still opens the doors for what I was talking about. Computers built using synthetic neurons formed around artificial DNA. Which could very well be the first step towards being able to create true AIs.
I will submit that even if and when such AI architectures can be built, they wouldn't be very usable for the purposes of storing existing personalities. It's not enough to have an artificial "brain". A personality isn't just a collection of neurons; it's how they interact with each other and have arranged themselves over time. These aspects aren't coded in the DNA, they simply develop as the mind matures. In order to have a receptacle that could host a person's consciousness, you would need to find a way to get the artificial neurons to arrange themselves and interact with their neighbors in a way that was identical to the brain of the person you were "downloading". Which isn't anything you could effectively code into the cells themselves.
To be clear: a "machine" built using artificial neurons wouldn't be a "Turing complete machine".
It couldn't be.
I'm not talking about writing an AI "program". I'm talking about building what amounts to a brain and letting the personality grow naturally, similar to the way it does in real people.
No algorithms. No "If/Else" loops. No cin/cout. Just a genuine personality that develops nearly identically to that way our own does, using the same mechanisms.
It's then just a matter of wiring said artificial "brain" into a standard computer(cybernetics being another one of those fields we're just touching upon at the moment), and viola! An Artificially Intelligent computer.
Easy? No. Going to happen within our lifetimes? Doubtful. A possible eventuality? I'd say so.
That's tantamount to cloning, which we can already do (Well, theoretically... no one's come out and say they've performed human cloning yet, I think. Too much of a faux pas).
How's that useful in any way, first of all... and second, how's that synthetic? Growing new lifeforms a la Dr. Moreau? It's mix-and-match, all of it. Mocking, not making. Mimicking, not creating. Thirdly...
"It's then just a matter of wiring said artificial "brain" into a standard computer"
...it's not an A.I. then, either. It's a cyborg. Again, way different. And frankly, far more useful to do to ourselves than some unpredictable, lab-grown brain which may or may not be happy with its situation.
Honestly, it sounds like you're taking something like in vitro fertilization and crying "EEEET'S ALIIIIIIVE!". That's not making anything new. It's not creation. It's not synthetic. It's just taking known biological processes and letting them do their thing.
Could the things you describe be done? Maybe. Seems possible. But it wouldn't be a fully synthetic A.I. And thus, the "artificial" part of the acronym would be a misnomer.
You basically moved the argument from "Can we make true A.I.?" to "Can we make cyborg clone brains?". An oversimplification, but reasonably accurate.
"I'm not talking about writing an AI "program"."
Then we're not talking about the same thing. The question in brain uploading is, quite simply, this: Can we move the essence of one's mind onto different media?
Given the limits of our programming languages (bound to formal limits) as opposed to the capabilities of what we'd be transcribing (not thus bound), the answer is currently a very distinct No. And given no up-and-coming methods to fix that problem, it seems safe to say it's impossible.
Little harm done in trying the impossible, though, as long as not too many resources are wasted on it.
I think I'm starting to see the problem: You're using different definitions of "synthetic" and "artificial" that don't appear in any dictionary I can find.
Which I guess is how you were somehow able to equate, "using lab grown XNA built with synthetic polymers to construct man-coded neurons" with, "in vitro fertilization"
*slow clap*
And are we talking about transferring a consciousness to "different" media? Or current media? I'm talking about different. You seem to be stuck on current. Not the same thing.
A single neuron passes thousands of charges to thousands of destinations simultaneously. A transistor passes a single charge between two locations one at a time. There isn't a programming language in existence that can compensate for that. Any more than you could find a musician capable of playing a song written out as the binary transcription of an mp3.
BUT...creating "transistors" based upon those aforementioned artificial(really suggest you look up the word so that we're on the same page) neurons...
As far as devising a suitable programming language to interact with said "neuron circuits", we can't really apply much of what we already know about programming. As per said aforementioned paradigm differences. A whole new principal would need to be devised. Hard to call such a language impossible, since I don't know of any attempt that was made to instruct a computer to perform a thousand operations, per transistor, simultaneously...Such a language can't even be tried until the architecture is built. I'm not even sure how it could be simulated on current architecture.
Actually a neuron has only one output. It receives multiple signals of varying strength and if they combine to produce an electrical impulse above a certain threshold it fires off an impulse down the singular dendrite which connects to another neuron and depending on its location puts in so much voltage.
So, still not digital, more like an analog/digital hybrid.
"I think I'm starting to see the problem: You're using different definitions of 'synthetic' and 'artificial' that don't appear in any dictionary I can find."
I can't help but wonder what dictionaries you're using.
ar·ti·fi·cial/ˌärtəˈfiSHəl/ Adjective:
Made or produced by human beings rather than occurring naturally, typically as a copy of something natural: "artificial light".
synthetic [sin-thet-ik] Origin syn·thet·ic [sin-thet-ik] Show IPA adjective 1. of, pertaining to, proceeding by, or involving synthesis ( opposed to analytic). 2. noting or pertaining to compounds formed through a chemical process by human agency, as opposed to those of natural origin: synthetic vitamins; synthetic fiber. 5. not real or genuine; artificial; feigned: a synthetic chuckle at a poor joke.
Start from scratch Meaning
Begin (again) from the beginning, embark on something without any preparation or advantage. Like baking a cake without pre-mixed dough.
Does that clarify?
These aren't obscure definitions by any stretch of the imagination.
Starting with a *whole* strand of DNA and replacing the sugar component (with a known, simple molecular structure) with synthesized polymers is like taking a cow, replacing its limbs with pegs, putting it on dialysis, and running it on a Jarvik heart, and saying you made a new kind of cow. Or like tying a horse to a buggy and calling the horse/buggy combination "artificial". If the whole thing isn't artificial, then it's not artificial as a whole.
You need to look up the definition of 'artificial'. It's not terribly ambiguous or esoteric. It just means man-made. Any preexisting, naturally occurring architecture would not be artificial, and if it's not wholly artificial, it's not artificial as a whole. If the preexisting structure is necessary to the construct, then it is impossible to create it artificially.
"And are we talking about transferring a consciousness to 'different' media?"
Yes. However, what you're talking about has little to do with that. A lab-grown brain using preexisting organic tissue isn't a different medium. And I haven't even heard you posit an idea for the transfer of consciousness yet, only for A.I. and the media they might inhabit. The transfer is a whole other beast to tackle.
"Hard to call such a language impossible..."
If you'd watched the video you'd know exactly what the argument against its possibility is. And no, it cannot be proven. But the evidence is stacked pretty high.
The problem isn't the medium of programming. It's the necessity of programming. A human mind doesn't need programming; it can program itself.won't</i> be "programmed" directly by humans: they will be grown in a manner analogous to the minds of human beings and other animals. Whether or not this means that they will necessarily be grown in biological substrates, or some other substrates are also possible, remains to be seen. My opinion is that intelligence is a matter of pattern, not substrate, and that we will be able to grow AI's in more than one medium.
If it's "grown" in the manner you describe, then how can we take credit for its creation, thus making it an "artificial" intelligence?
It's like taking credit for the creation of a baby's intelligence. How "artificial" is that? You threw some genes together and let 'er fly. You've "artifacted" an intelligence.
Unless you're responsible for the pattern that emerges and the medium it inhabits, it's not "artificial". It's natural. You just watched it happen, you didn't make it happen. Any more than you "make" a plant fertilize when you pollinate it. It takes too much responsibility for too little work.
rhjunior
That's the question, isn't it? See, we know the basis for all the computing machines we make can eventually be reduced to a basic Turing machine, but we still don't have much of a clue to how the brain works like it does. We know it involves chemical and electrical imprinting... we think. We really don't understand the process; merely what parts of the brain affect what parts of the body and/or computing ability.
Add to that the fact that, at least to all appearances, the human mind seems capable of arbitrarily filling in the blank holes left by formal systems (like computational math, the whole basis of all computing machines made by man) in a way machines literally *never* can, and it starts to seem less and less likely that a true replication of a human-like mind will ever exist.
"Add to that the development of computers that actually code using DNA"
The problem isn't the medium of programming. It's the necessity of programming. A human mind doesn't need programming; it can program itself. It can make arbitrary decisions that fly in the face of all it computes, or disregards all positive and negative reinforcement. A computer can only do what it's told to do. Only a human mind can say 'no', even when it's totally illogical to do so. Even when it's emotionally unpleasant to do so. Even when doing so has literally no upsides or benefit.
That can't be programmed, no matter what media you use. And therein lies the problem.
So... no. Not really. Unless we find some sort of device that can arbitrarily, without any human direction, program itself to do more than gibberish, it doesn't matter what we *use* in the attempt to make A.I., it simply won't be mathematically possible. And how, in the end, can we be sure the decisions of such a machine not come indirectly from its creators?
There's no way of knowing.
It kind of puts a big damper on that sort of research, neat as it would be.
-Kire Du'Hai
How can you know, in the end, be sure that the decisions you're making aren't based upon how you were raised? Your prior experiences and knowledge?
How random are our own decisions and choices? Really?
Pick a number.
I'm sure you thought of a few numbers almost immediately. Your favorite numbers, common numbers, interesting numbers(rational and otherwise), even amusing numbers. How "random" were they? If I told you the number I picked at random was 42, would you believe that it was genuinely a random number? It would probably depend on whether I was a hoopy-frood or not, wouldn't it? What about 13? 7? 1337? How random are those numbers? Sure, I could consciously choose any number, but how "random" is it if I'm picking that number in defiance of the first numbers I thought of; which are in turn, numbers that I think of based on my prior experiences?
The decisions we make are all based upon our previous knowledge and experiences from an earlier point in our lives. Our, "programming" if you will. Even when you pick an option that "seems" random or arbitrary, you're doing so because it seems that way. You've justified the decision to yourself. Rationalized it. You followed your programming.
A neuron doesn't physically hold our thoughts or memories. All it is is a link in a chain passing charges around; which is what "generates" our thoughts...somehow. Sure, we may not currently understand how or why the ways they arrange themselves produce the results they do, but that's a far cry from saying that we never can or will. We currently have the ability to "build" artificial cells using man-coded DNA. So, we have the "hardware". We can sequence the genome and determine what genes are expressed in neurons and why. So, we have the "software".
All we don't know at this point is how to organize the architecture. Which is a big hurdle, don't get me wrong. Brain cells organize themselves in a way that we simply can't organize traditional transistors, and thus any tradition programming languages or principles. So any real advances into AI will probably require creating a whole new field of engineering to go alongside CS and CE.
But, we're seeing some pretty cool things happening in the areas of genetics and biology these days. So I'm not in such a hurry to write it off as impossible.
After all, it already happened once...
That's the challenge faced by both sides of the issue. But it has been pretty empirically proven that human beings can reach conclusions that cannot be algorithmically derived. At all.
That has nothing to do with how "random" our thought processes are. The ability to determine if an unprovable theorem is true cannot, by any means currently available, be produced in anything synthetic. That's not something that can be "programmed" at all.
And a key note: "programming" isn't your prior experiences or knowledge. "Programming" would be your preset *reactions* to your experiences or knowledge, i.e. how you process it. Totally different.
"We currently have the ability to "build" artificial cells using man-coded DNA. So, we have the "hardware". We can sequence the genome and determine what genes are expressed in neurons and why. So, we have the "software"."
Last I knew, we had the ability to mix-n-match different genomes. That is a far, far cry from truly synthetic DNA.
Let me know when they're building self-replicating, self-repairing DNA from scratch...
...And then I'll take that research seriously.
"After all, it already happened once..."
...When? Who did it?
-Kire Du'Hai
I said from scratch. Meaning elemental. Replacing sugar chains present in DNA and RNA with polymers is not making anything new.
In fact, it's so far removed from the concept of "from scratch" that the Fox and Miller experiments were more productive on that front. And they didn't really produce anything at all.
Let me clarify, since this apparently wasn't elaborated before:
If you have to use DNA (As in this case, where it's almost entirely original DNA) to make your synthetic DNA, it's not "from scratch". It's just modified DNA.
-Kire Du'Hai
They've removed what amounted to half of the structure and replaced it with synthetic compounds. They specifically arranged the amino acids as they saw fit to make it do what they want. Nothing like it can be found on Earth.
XNA is as much "modified DNA", as an airplane is "modified bauxite".
Never mind that you managed to pretty much use your hairsplitting as an excuse to ignore the point: This still opens the doors for what I was talking about. Computers built using synthetic neurons formed around artificial DNA. Which could very well be the first step towards being able to create true AIs.
I will submit that even if and when such AI architectures can be built, they wouldn't be very usable for the purposes of storing existing personalities. It's not enough to have an artificial "brain". A personality isn't just a collection of neurons; it's how they interact with each other and have arranged themselves over time. These aspects aren't coded in the DNA, they simply develop as the mind matures. In order to have a receptacle that could host a person's consciousness, you would need to find a way to get the artificial neurons to arrange themselves and interact with their neighbors in a way that was identical to the brain of the person you were "downloading". Which isn't anything you could effectively code into the cells themselves.
It couldn't be.
I'm not talking about writing an AI "program". I'm talking about building what amounts to a brain and letting the personality grow naturally, similar to the way it does in real people.
No algorithms. No "If/Else" loops. No cin/cout. Just a genuine personality that develops nearly identically to that way our own does, using the same mechanisms.
It's then just a matter of wiring said artificial "brain" into a standard computer(cybernetics being another one of those fields we're just touching upon at the moment), and viola! An Artificially Intelligent computer.
Easy? No. Going to happen within our lifetimes? Doubtful. A possible eventuality? I'd say so.
How's that useful in any way, first of all... and second, how's that synthetic? Growing new lifeforms a la Dr. Moreau? It's mix-and-match, all of it. Mocking, not making. Mimicking, not creating.
Thirdly...
"It's then just a matter of wiring said artificial "brain" into a standard computer"
...it's not an A.I. then, either. It's a cyborg. Again, way different. And frankly, far more useful to do to ourselves than some unpredictable, lab-grown brain which may or may not be happy with its situation.
Honestly, it sounds like you're taking something like in vitro fertilization and crying "EEEET'S ALIIIIIIVE!". That's not making anything new. It's not creation. It's not synthetic. It's just taking known biological processes and letting them do their thing.
Could the things you describe be done? Maybe. Seems possible. But it wouldn't be a fully synthetic A.I. And thus, the "artificial" part of the acronym would be a misnomer.
You basically moved the argument from "Can we make true A.I.?" to "Can we make cyborg clone brains?". An oversimplification, but reasonably accurate.
"I'm not talking about writing an AI "program"."
Then we're not talking about the same thing.
The question in brain uploading is, quite simply, this:
Can we move the essence of one's mind onto different media?
Given the limits of our programming languages (bound to formal limits) as opposed to the capabilities of what we'd be transcribing (not thus bound), the answer is currently a very distinct No. And given no up-and-coming methods to fix that problem, it seems safe to say it's impossible.
Little harm done in trying the impossible, though, as long as not too many resources are wasted on it.
-Kire Du'Hai
Which I guess is how you were somehow able to equate, "using lab grown XNA built with synthetic polymers to construct man-coded neurons" with, "in vitro fertilization"
*slow clap*
And are we talking about transferring a consciousness to "different" media? Or current media? I'm talking about different. You seem to be stuck on current. Not the same thing.
A single neuron passes thousands of charges to thousands of destinations simultaneously. A transistor passes a single charge between two locations one at a time. There isn't a programming language in existence that can compensate for that. Any more than you could find a musician capable of playing a song written out as the binary transcription of an mp3.
BUT...creating "transistors" based upon those aforementioned artificial(really suggest you look up the word so that we're on the same page) neurons...
As far as devising a suitable programming language to interact with said "neuron circuits", we can't really apply much of what we already know about programming. As per said aforementioned paradigm differences. A whole new principal would need to be devised. Hard to call such a language impossible, since I don't know of any attempt that was made to instruct a computer to perform a thousand operations, per transistor, simultaneously...Such a language can't even be tried until the architecture is built. I'm not even sure how it could be simulated on current architecture.
So, still not digital, more like an analog/digital hybrid.
I can't help but wonder what dictionaries you're using.
ar·ti·fi·cial/ˌärtəˈfiSHəl/
Adjective:
Made or produced by human beings rather than occurring naturally, typically as a copy of something natural: "artificial light".
synthetic
[sin-thet-ik] Origin
syn·thet·ic
[sin-thet-ik] Show IPA
adjective
1.
of, pertaining to, proceeding by, or involving synthesis ( opposed to analytic).
2.
noting or pertaining to compounds formed through a chemical process by human agency, as opposed to those of natural origin: synthetic vitamins; synthetic fiber.
5.
not real or genuine; artificial; feigned: a synthetic chuckle at a poor joke.
Start from scratch
Meaning
Begin (again) from the beginning, embark on something without any preparation or advantage.
Like baking a cake without pre-mixed dough.
Does that clarify?
These aren't obscure definitions by any stretch of the imagination.
Starting with a *whole* strand of DNA and replacing the sugar component (with a known, simple molecular structure) with synthesized polymers is like taking a cow, replacing its limbs with pegs, putting it on dialysis, and running it on a Jarvik heart, and saying you made a new kind of cow. Or like tying a horse to a buggy and calling the horse/buggy combination "artificial". If the whole thing isn't artificial, then it's not artificial as a whole.
You need to look up the definition of 'artificial'. It's not terribly ambiguous or esoteric. It just means man-made. Any preexisting, naturally occurring architecture would not be artificial, and if it's not wholly artificial, it's not artificial as a whole. If the preexisting structure is necessary to the construct, then it is impossible to create it artificially.
"And are we talking about transferring a consciousness to 'different' media?"
Yes. However, what you're talking about has little to do with that. A lab-grown brain using preexisting organic tissue isn't a different medium. And I haven't even heard you posit an idea for the transfer of consciousness yet, only for A.I. and the media they might inhabit. The transfer is a whole other beast to tackle.
"Hard to call such a language impossible..."
If you'd watched the video you'd know exactly what the argument against its possibility is. And no, it cannot be proven. But the evidence is stacked pretty high.
-Kire Du'Hai
Edited at 2012-05-18 03:33 am (UTC)
It's like taking credit for the creation of a baby's intelligence. How "artificial" is that? You threw some genes together and let 'er fly. You've "artifacted" an intelligence.
Unless you're responsible for the pattern that emerges and the medium it inhabits, it's not "artificial". It's natural. You just watched it happen, you didn't make it happen. Any more than you "make" a plant fertilize when you pollinate it. It takes too much responsibility for too little work.
-Kire Du'Hai