Forums > Social Discussion > Artificial Intelligence, the Internet and Myspace/Facebook/etc
Login/Join to ParticipateStargazer
6,650 posts
6,650 posts
Posted:
Just thinking in the 'hype' of myspace, facebook etc:
If I were to generate Artificial Intelligence, I would need so much input of various kinds about humans, that I would generate a web (like the internet) where people (like you and me) will key in anything that concerns them (thoughts, ideas, reactions, pictures, movies, etc)
It would subsequently generate some profile/ cross section of human intelligence, more than any study could ever produce: User Generated Content...
Not saying that there is a conspiracy or anything like it, only noticing and sharing. No doomsday or fear of an Alien project, nothing in my mind about that.
The internet becomes the great archive of (at least) mankind ... Not sure where it will lead to...
Just thinking in the 'hype' of myspace, facebook etc:
If I were to generate Artificial Intelligence, I would need so much input of various kinds about humans, that I would generate a web (like the internet) where people (like you and me) will key in anything that concerns them (thoughts, ideas, reactions, pictures, movies, etc)
It would subsequently generate some profile/ cross section of human intelligence, more than any study could ever produce: User Generated Content...
Not saying that there is a conspiracy or anything like it, only noticing and sharing. No doomsday or fear of an Alien project, nothing in my mind about that.
The internet becomes the great archive of (at least) mankind ... Not sure where it will lead to...
the best smiles are the ones you lead to 
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
its not quite that simple ...
you require "intelligence" to exist to be able to autonomously integrate the new content entered by people, hence a chicken and egg problem. To evolve a system takes substantial feedback as you need a success criteria to guide its development (not to mention the immense complexity of self modifying evolutionary code)
Interpreting unrestricted images is hugely complex this is why you dont see robots outside of highly restricted environments.
Personally I would be very disturbed if AI got its "knowledge" about the world from the internet masses ... no doubt it would end up with a great understanding of p orn, flaming people and 1337n355
its not quite that simple ...
you require "intelligence" to exist to be able to autonomously integrate the new content entered by people, hence a chicken and egg problem. To evolve a system takes substantial feedback as you need a success criteria to guide its development (not to mention the immense complexity of self modifying evolutionary code)
Interpreting unrestricted images is hugely complex this is why you dont see robots outside of highly restricted environments.
Personally I would be very disturbed if AI got its "knowledge" about the world from the internet masses ... no doubt it would end up with a great understanding of p orn, flaming people and 1337n355
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
BRONZE Member
Insert Champagne Here
13,215 posts
Location: without class distinction, Australia
Insert Champagne Here
13,215 posts
Location: without class distinction, Australia
Posted:
When I think of artificial intelligence and the internet I think of Wikipedia.
How many people use wikipedia as their main source of into on a topic?
When I think of artificial intelligence and the internet I think of Wikipedia.
How many people use wikipedia as their main source of into on a topic?
i would have changed ***** to phallus, and claire to petey Petey
Rougie: but that's what I'm doing here
Arnwyn: what letting me adjust myself in your room?..don't you dare quote that on HoP...
GOLD Member
Introverted
833 posts
Location: Norf London, United Kingdom
Introverted
833 posts
Location: Norf London, United Kingdom
Posted:
More discussion of a similar idea over on the [Old link] thread from a few years back...
More discussion of a similar idea over on the [Old link] thread from a few years back...
My Mind is a Ship
Emotions become the Waves
Soul is the Ocean
If a quizz is quizzical, what is a test?
BRONZE Member
poiromaniac
162 posts
Location: liverpool, uk
poiromaniac
162 posts
Location: liverpool, uk
Posted:
well theres always thr turing test that needs to be passed if you want to call it intellegence (thats where somebody has a conversation with a human an a machine, both of which he cannot see, but can read their text replies and if the person would not be able to distinguish between the human and the computer then you have yourself some AI)
well theres always thr turing test that needs to be passed if you want to call it intellegence (thats where somebody has a conversation with a human an a machine, both of which he cannot see, but can read their text replies and if the person would not be able to distinguish between the human and the computer then you have yourself some AI)
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
Re Turing Test - read the Chinese room argument, all the Turing test could prove is that the programmers had "intelligence" .... whatever that is
Re Turing Test - read the Chinese room argument, all the Turing test could prove is that the programmers had "intelligence" .... whatever that is
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
BRONZE Member
Carpal \'Tunnel
5,814 posts
Location: Perth, Australia
Carpal \'Tunnel
5,814 posts
Location: Perth, Australia
Posted:
Must go play WoW
Must go play WoW
Is it the Truth?
Is it Fair to all concerned?
Will it build Goodwill and Better Friendships?
Will it be Beneficial to all concerned?
Im in a lonely battle with the world with a fish to match the chip on my shoulder. Gnu in Binnu in a cnu
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
your such a junky, set away from the computer you have a big weekend of spunoutness to prepare for, just remember even though your pretending its WoW dont attack people they wont like it ... jed might but the rest wont
your such a junky, set away from the computer you have a big weekend of spunoutness to prepare for, just remember even though your pretending its WoW dont attack people they wont like it ... jed might but the rest wont
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
SILVER Member
veteran
1,228 posts
Location: on the wrong planet, United Kingdom
veteran
1,228 posts
Location: on the wrong planet, United Kingdom
Posted:
If an AI was to base it's opinion of humanity on what it could find on the internet, more than likely it would just go
OMG? WTF? NUKE!
With any luck in 3 years I'll be doing a Masters in AI programming (for computer games, granted, but they're getting very close to reality simulators, and who knows what another 3 years of technological advancement may bring).
If an AI was to base it's opinion of humanity on what it could find on the internet, more than likely it would just go
OMG? WTF? NUKE!
With any luck in 3 years I'll be doing a Masters in AI programming (for computer games, granted, but they're getting very close to reality simulators, and who knows what another 3 years of technological advancement may bring).
You aren't thinking or really existing unless you're willing to risk even your own sanity in the judgment of your existence.
Green peppers, lime pickle and whole-grain mustard = best sandwich filling.
SILVER Member
veteran
1,228 posts
Location: on the wrong planet, United Kingdom
veteran
1,228 posts
Location: on the wrong planet, United Kingdom
Posted:
If you were to write www in Aramaic it wouldn't make any sense (them not having computers back then), so they'd just translate it as a number, seeing as they used the letter characters to represent numbers too.
Guess what number the 'w' character is
If you were to write www in Aramaic it wouldn't make any sense (them not having computers back then), so they'd just translate it as a number, seeing as they used the letter characters to represent numbers too.
Guess what number the 'w' character is
You aren't thinking or really existing unless you're willing to risk even your own sanity in the judgment of your existence.
Green peppers, lime pickle and whole-grain mustard = best sandwich filling.
Stargazer
6,650 posts
6,650 posts
Posted:
IMO it all depends on the programmers, we also would have MSN and other components available, to simulate human emotional reactions... The thing is much more complex than just a few typosites only.
What actually IS "intelligence"? Interesting enough you can get some pretty tough AI when it comes to gameplay.
I reckon the thing is just behind the next corner.
IMO it all depends on the programmers, we also would have MSN and other components available, to simulate human emotional reactions... The thing is much more complex than just a few typosites only.
What actually IS "intelligence"? Interesting enough you can get some pretty tough AI when it comes to gameplay.
I reckon the thing is just behind the next corner.
the best smiles are the ones you lead to
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
To put things in perspective, the average human brain has 100 billion neurons, each neuron acts a computer, it on average is connected to around 10,000 other neurons, the input from each neuron influences the firing threshold of the neurons it is connected to, in essence each neuron acts like a micro computer.
So if you look at the internet as a comparison, there are a possible maximum 255 x 255 x 255 x 255 computers on the internet at any one time ~ 4 billion computers.
Look at the complexity of human vision, we do it without thinking, for vision for AI you need to have rules to determine if a edge exists for each of the 8 million (ish) colours with regards to the other 8 million ish colours, but to make matters more complex vision is a function of space so its at least 8 mill x mill x 8 mill combinations. This is a requirement if u want it to be able to process images or video.
.... real AI is a VERY LONG way off
Game AI is "simple" because you create the rules of the virtual universe, real AI is not because you cant just make up the rules for the real world.
Written by :FireTom
What actually IS "intelligence"? Interesting enough you can get some pretty tough AI when it comes to gameplay.
I reckon the thing is just behind the next corner.
To put things in perspective, the average human brain has 100 billion neurons, each neuron acts a computer, it on average is connected to around 10,000 other neurons, the input from each neuron influences the firing threshold of the neurons it is connected to, in essence each neuron acts like a micro computer.
So if you look at the internet as a comparison, there are a possible maximum 255 x 255 x 255 x 255 computers on the internet at any one time ~ 4 billion computers.
Look at the complexity of human vision, we do it without thinking, for vision for AI you need to have rules to determine if a edge exists for each of the 8 million (ish) colours with regards to the other 8 million ish colours, but to make matters more complex vision is a function of space so its at least 8 mill x mill x 8 mill combinations. This is a requirement if u want it to be able to process images or video.
.... real AI is a VERY LONG way off
Game AI is "simple" because you create the rules of the virtual universe, real AI is not because you cant just make up the rules for the real world.
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
SILVER Member
member - tee hee "member"
279 posts
Location: Bethnal Green, London, England (UK)
member - tee hee "member"
279 posts
Location: Bethnal Green, London, England (UK)
Posted:
Wheres Rob Nunchux? He's doing an AI Masters at Brum!
Wheres Rob Nunchux? He's doing an AI Masters at Brum!
"Whats wrong with the cat?" - Mrs Schrödinger
GOLD Member
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
Posted:
I didn't know the 10,000 average connections per neuron figure (got a source for that? I'm interested). But here's the thing: a given neuron has one output: either it fires or it doesn't. On a more complex level, the firing rate is the ultimate determinant of the output of the neuron.
Let me put this in another perspective: A newborn infant demonstrates a stereotyped behavior repertoire that is smaller than that of your average insect. In essense, every little cute thing an infant two hours old does is simply reflex. I can elicit almost any available behavior from a newborn with a maneuver specific to that behavior. I can make a newborn wiggle his butt, grab my hand, burp, suck and swallow, and even poop. And it will work >98% of the time and the remaining ~2% of infants probably have some neurological pathology.
By the time that infant is two weeks old, he is fixing on peoples' faces.
By the time he is six months old he is cooing, laughing, interacting, and you can begin to see a little human being coming to consciousness.
At about a year he's walking (+/- 2 months).
These milestones are set in stone. A baby who is not walking by 14 months of age is developmentally delayed. The milestones will all come in pretty much the same order for every baby. In that respect, a good deal of human intelligence is inborn, pre-programmed, and hard-wired.
But over the next 17 years, that child will be exposed to a world in which there are rules, expectations, patterns, and probabilities. By age 5 he will speak a language fluently and as a native. If he hasn't learned a given language by age 10-12 he will *never* speak it fluently and as a native, no matter how good at it he gets.
Without those experiences and inputs, however, that stereotyped pattern of development becomes disordered and delayed. Until a certain age, children can recover those milestones, but after a certain degree of developmental deprivation, they are destined to be mentally retarded, in spite of having been born with a perfectly good brain.
And that is why it is so hard to build from scratch a computer that can behave like an adult human being. It takes 16-20 years to make an adult.
Written by :ben-ja-men
To put things in perspective, the average human brain has 100 billion neurons, each neuron acts a computer, it on average is connected to around 10,000 other neurons, the input from each neuron influences the firing threshold of the neurons it is connected to, in essence each neuron acts like a micro computer.
So if you look at the internet as a comparison, there are a possible maximum 255 x 255 x 255 x 255 computers on the internet at any one time ~ 4 billion computers.
I didn't know the 10,000 average connections per neuron figure (got a source for that? I'm interested). But here's the thing: a given neuron has one output: either it fires or it doesn't. On a more complex level, the firing rate is the ultimate determinant of the output of the neuron.
Let me put this in another perspective: A newborn infant demonstrates a stereotyped behavior repertoire that is smaller than that of your average insect. In essense, every little cute thing an infant two hours old does is simply reflex. I can elicit almost any available behavior from a newborn with a maneuver specific to that behavior. I can make a newborn wiggle his butt, grab my hand, burp, suck and swallow, and even poop. And it will work >98% of the time and the remaining ~2% of infants probably have some neurological pathology.
By the time that infant is two weeks old, he is fixing on peoples' faces.
By the time he is six months old he is cooing, laughing, interacting, and you can begin to see a little human being coming to consciousness.
At about a year he's walking (+/- 2 months).
These milestones are set in stone. A baby who is not walking by 14 months of age is developmentally delayed. The milestones will all come in pretty much the same order for every baby. In that respect, a good deal of human intelligence is inborn, pre-programmed, and hard-wired.
But over the next 17 years, that child will be exposed to a world in which there are rules, expectations, patterns, and probabilities. By age 5 he will speak a language fluently and as a native. If he hasn't learned a given language by age 10-12 he will *never* speak it fluently and as a native, no matter how good at it he gets.
Without those experiences and inputs, however, that stereotyped pattern of development becomes disordered and delayed. Until a certain age, children can recover those milestones, but after a certain degree of developmental deprivation, they are destined to be mentally retarded, in spite of having been born with a perfectly good brain.
And that is why it is so hard to build from scratch a computer that can behave like an adult human being. It takes 16-20 years to make an adult.
-Mike
Certified Mad Doctor and HoP High Priest of Nutella
A buckuht n a hooze! -Valura
BRONZE Member
wandering thru the woods of WI
3,556 posts
Location: Wisconsin, USA
wandering thru the woods of WI
3,556 posts
Location: Wisconsin, USA
Posted:
I think you are underestimating, Doc
I think you are underestimating, Doc
Faith
Nay, whatever comes one hour was sunlit and the most high gods may not make boast of any better thing than to have watched that hour as it passed
SILVER Member
veteran
1,228 posts
Location: on the wrong planet, United Kingdom
veteran
1,228 posts
Location: on the wrong planet, United Kingdom
Posted:
And computers are seriously limited in their ability to learn. One of the biggest limitations is that they can't interact with their environment, to make use of what they've learned.
Also what abilities they do have are mostly preprogrammed by us. Unfortunately learning isn't among them, except in the most rudimentary forms.
And computers are seriously limited in their ability to learn. One of the biggest limitations is that they can't interact with their environment, to make use of what they've learned.
Also what abilities they do have are mostly preprogrammed by us. Unfortunately learning isn't among them, except in the most rudimentary forms.
You aren't thinking or really existing unless you're willing to risk even your own sanity in the judgment of your existence.
Green peppers, lime pickle and whole-grain mustard = best sandwich filling.
GOLD Member
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
Posted:
underestimating what?
Written by :faithinfire
I think you are underestimating, Doc
underestimating what?
-Mike
Certified Mad Doctor and HoP High Priest of Nutella
A buckuht n a hooze! -Valura
GOLD Member
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
Posted:
The other thing to remember is that human intelligence evolved as an evolutionary solution to a series of shifting pressures on a particular branch of life over a period of some 2.5 billion years.
The other thing is that the human brain is shockingly similar to that of a chimpanzee or a bonobo. And our genomes are 97% concordant. What the hell is in the other 3%?
There is something very complex about intelligence and brain architecture that makes intelligence an emergent phenomenon, rather than an inherent one (and I use those terms mathematically and not colloquially).
And personality itself is at least as much inborn as it is nurture. Ask any parent of two or more children; they all spring from the womb with a certain inherent temprament that follows them through life.
The other thing to remember is that human intelligence evolved as an evolutionary solution to a series of shifting pressures on a particular branch of life over a period of some 2.5 billion years.
The other thing is that the human brain is shockingly similar to that of a chimpanzee or a bonobo. And our genomes are 97% concordant. What the hell is in the other 3%?
There is something very complex about intelligence and brain architecture that makes intelligence an emergent phenomenon, rather than an inherent one (and I use those terms mathematically and not colloquially).
And personality itself is at least as much inborn as it is nurture. Ask any parent of two or more children; they all spring from the womb with a certain inherent temprament that follows them through life.
-Mike
Certified Mad Doctor and HoP High Priest of Nutella
A buckuht n a hooze! -Valura
BRONZE Member
old hand
929 posts
Location: Oregon, USA
old hand
929 posts
Location: Oregon, USA
Posted:
But the real question is does intelligence really require emotion, or just comprehension and abstract thinking?
Look at the way translating software works, Google's for example doesn't actually know what any of the words you put in mean. It doesn't look for definitions. It's comparing your words, with words of documents it has that have already been translated into dozens of languages (like UN documents for example). It doesn't *understand* what you're writing.
But, if you truly want to look at the potential future, I would say Google's cluster network is quickly approaching the if not already at the processing power of the human brain, it's just busy running search queries not being used to develop AI.
But the real question is does intelligence really require emotion, or just comprehension and abstract thinking?
Look at the way translating software works, Google's for example doesn't actually know what any of the words you put in mean. It doesn't look for definitions. It's comparing your words, with words of documents it has that have already been translated into dozens of languages (like UN documents for example). It doesn't *understand* what you're writing.
But, if you truly want to look at the potential future, I would say Google's cluster network is quickly approaching the if not already at the processing power of the human brain, it's just busy running search queries not being used to develop AI.
#homeofpoi -- irc.newnet.net Come talk to us we're bored 
Warning: Please Do Not Jump On The Seals
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
R. Miikkulain, J.A. Bednar, C. Yoonsuck & J. Sirosh, “Computational Maps in the Visual Cortex”, p4
the bits on a computer also have one output, they are either ones or zeros, its the way in which those ones and zeros are connected together and interpreted that gives meaning.
Written by :Doc Lightning
I didn't know the 10,000 average connections per neuron figure (got a source for that? I'm interested).
R. Miikkulain, J.A. Bednar, C. Yoonsuck & J. Sirosh, “Computational Maps in the Visual Cortex”, p4
Written by :Doc Lightning
But here's the thing: a given neuron has one output: either it fires or it doesn't. On a more complex level, the firing rate is the ultimate determinant of the output of the neuron.
the bits on a computer also have one output, they are either ones or zeros, its the way in which those ones and zeros are connected together and interpreted that gives meaning.
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
GOLD Member
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
Posted:
Thankye. I'll look it up.
The difference is that a neuron produces a firing rate. That firing rate determines the concentration of a neurotransmitter in the synapse at the end of that axon. That influences the next neuron's likelihood of firing. There's a lot of fuzziness in there that bits and bytes can't replicate.
Written by :ben-ja-men
R. Miikkulain, J.A. Bednar, C. Yoonsuck & J. Sirosh, “Computational Maps in the Visual Cortex”, p4
Thankye. I'll look it up.
Written by :ben-ja-men
]
the bits on a computer also have one output, they are either ones or zeros, its the way in which those ones and zeros are connected together and interpreted that gives meaning.
The difference is that a neuron produces a firing rate. That firing rate determines the concentration of a neurotransmitter in the synapse at the end of that axon. That influences the next neuron's likelihood of firing. There's a lot of fuzziness in there that bits and bytes can't replicate.
-Mike
Certified Mad Doctor and HoP High Priest of Nutella
A buckuht n a hooze! -Valura
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
The firing rate is measured by the rate at which the synapse fires, observing the rate at which a register value changes over time would have the same effect, code looking something like
bool b;
bool c;
bool d;
for (int a=0;a<=10000;a++)
{
if (int(a/2)-a/2==0)
{ b=true; }
else
{ b=false; }
d=XOR(b,c)
}
b changes with time and affects the states of d over time with relation to other variable c.
I agree that there is alot more complexity in the neurons as the firing threshold inside of each one changes with relation to the states of the microtubules inside it, which act as a cellular automata to determine the threshold and are influenced by the incoming signals, but in essence I don't see a difference in the principles behind how they process information.
Written by :Doc Lightning
The difference is that a neuron produces a firing rate. That firing rate determines the concentration of a neurotransmitter in the synapse at the end of that axon. That influences the next neuron's likelihood of firing. There's a lot of fuzziness in there that bits and bytes can't replicate.
The firing rate is measured by the rate at which the synapse fires, observing the rate at which a register value changes over time would have the same effect, code looking something like
bool b;
bool c;
bool d;
for (int a=0;a<=10000;a++)
{
if (int(a/2)-a/2==0)
{ b=true; }
else
{ b=false; }
d=XOR(b,c)
}
b changes with time and affects the states of d over time with relation to other variable c.
I agree that there is alot more complexity in the neurons as the firing threshold inside of each one changes with relation to the states of the microtubules inside it, which act as a cellular automata to determine the threshold and are influenced by the incoming signals, but in essence I don't see a difference in the principles behind how they process information.
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
Scientist of Fortune
1,189 posts
Location: Edinburgh
1,189 posts
Location: Edinburgh
Posted:
The gene are pretty much the same, but the regulatory elements are fairly different. There was a conference on gene regulation recently in the US. Pharygula has a pretty decent write-up if you are interested. Article link
You're pretty much hitting the nail on the head with regards to why our rather enthusiastic expectations of ai aren't coming true. Learning and producing an object capable of learning are two (related) problems. Human brains have a couple of billion years behind them, so computers still have some catch-up to do. Mind you, drosophila brains only have about 30000 neurons, and 2/3 of them are given over to vision analysis. Besting them is probably a realistic goal in the short term.
Of course, so little is understood that the field is rife with superstition and pseudoscience at the moment. You don't have to go far to find some kook peddling their pet theory about how "consiousness" works, for example. A better understanding of how brains work will certainly give some good clues to programmers about how to go about making an ai, but then there's the issue of whether the best approach really is to mimic brains. The first really impressive ai's could well be produced by completely novel approaches.
All in all there is a lot of hard graft to go before we get a decent ai, both in terms of neuroscience and computing.
Written by :Doc Lightning
The other thing is that the human brain is shockingly similar to that of a chimpanzee or a bonobo. And our genomes are 97% concordant. What the hell is in the other 3%?
The gene are pretty much the same, but the regulatory elements are fairly different. There was a conference on gene regulation recently in the US. Pharygula has a pretty decent write-up if you are interested. Article link
You're pretty much hitting the nail on the head with regards to why our rather enthusiastic expectations of ai aren't coming true. Learning and producing an object capable of learning are two (related) problems. Human brains have a couple of billion years behind them, so computers still have some catch-up to do. Mind you, drosophila brains only have about 30000 neurons, and 2/3 of them are given over to vision analysis. Besting them is probably a realistic goal in the short term.
Of course, so little is understood that the field is rife with superstition and pseudoscience at the moment. You don't have to go far to find some kook peddling their pet theory about how "consiousness" works, for example. A better understanding of how brains work will certainly give some good clues to programmers about how to go about making an ai, but then there's the issue of whether the best approach really is to mimic brains. The first really impressive ai's could well be produced by completely novel approaches.
All in all there is a lot of hard graft to go before we get a decent ai, both in terms of neuroscience and computing.
According to Heisenberg's Uncertainty Principle of Quantum Dynamics, we may already be making love right now...
Stargazer
6,650 posts
6,650 posts
Posted:
Wow - that's a lot of useful and specific input. Thanks for that guys, you rock.
From what you are writing I understand that "real" AI might not present but future. How close is this future? - now that is the better question...
There is a LOT of information already available, lying idle on maybe a billion PC's worldwide? I'm not certain of the computing powers of a single neuron - so
how many computers would it really take to resemble a human brain?
What I do know is that in an enclosed environment (like for say a game of chess) you can't beat a PC. Sure, a computer is only as good as the people who programmed it - but the information stored on a computer seems[ to be far more reliable and precise that the information stored in a human brain. There is little guesswork on a hard drive, no?
Now when it comes to the "catching up" that computers would have to do.... how about CERN or - more specific - how about "the Grid"? It seems that we're not as far from generating a (web based) supercomputer as we might think.
Now the entire history of mankind and this planet is stored on hard drives - by the time an AI (as in a "critical mass") is reached, the learning process could be similar to a big bang.
True - computers do have a limited environment to act - so do humans. We're very limited creatures - compared to others we're weak, slow and dependent to a certain environment. What makes us the 'superior' race on this planet is our "intelligence"... The same would apply to AI.
Wow - that's a lot of useful and specific input. Thanks for that guys, you rock.
From what you are writing I understand that "real" AI might not present but future. How close is this future? - now that is the better question...
There is a LOT of information already available, lying idle on maybe a billion PC's worldwide? I'm not certain of the computing powers of a single neuron - so
how many computers would it really take to resemble a human brain?
What I do know is that in an enclosed environment (like for say a game of chess) you can't beat a PC. Sure, a computer is only as good as the people who programmed it - but the information stored on a computer seems[ to be far more reliable and precise that the information stored in a human brain. There is little guesswork on a hard drive, no?
Now when it comes to the "catching up" that computers would have to do.... how about CERN or - more specific - how about "the Grid"? It seems that we're not as far from generating a (web based) supercomputer as we might think.
Now the entire history of mankind and this planet is stored on hard drives - by the time an AI (as in a "critical mass") is reached, the learning process could be similar to a big bang.
True - computers do have a limited environment to act - so do humans. We're very limited creatures - compared to others we're weak, slow and dependent to a certain environment. What makes us the 'superior' race on this planet is our "intelligence"... The same would apply to AI.
the best smiles are the ones you lead to
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
how long is a piece of string
a computers information is embodied in the real world through its human users, so remove the human and the information is meaningless (in its current form without real AI to do stuff) to give an example, the electrons in the wall behind you could be vibrating in such a way that they are emulating the windows operating system, but without the correct filter to view the wall its just a wall, not an operating system.
with the source code and a photographic memory you could beat it, read this to understand why processing information does not equal intelligence if you used deep blues program to play checkers it would lose as it was designed specifically to play chess and has no ability to learn anything outside of the world of chess
it helps but its the rules for processing the data that is lacking. Without the rules on how to process sensory input (video pictures etc) its about as sentient as your toaster
Written by :FireTom
How close is this future? - now that is the better question...
how long is a piece of string
Written by :FireTom
There is a LOT of information already available, lying idle on maybe a billion PC's worldwide? I'm not certain of the computing powers of a single neuron - so
how many computers would it really take to resemble a human brain?
a computers information is embodied in the real world through its human users, so remove the human and the information is meaningless (in its current form without real AI to do stuff) to give an example, the electrons in the wall behind you could be vibrating in such a way that they are emulating the windows operating system, but without the correct filter to view the wall its just a wall, not an operating system.
Written by :FireTom
What I do know is that in an enclosed environment (like for say a game of chess) you can't beat a PC. Sure, a computer is only as good as the people who programmed it - but the information stored on a computer seems[ to be far more reliable and precise that the information stored in a human brain. There is little guesswork on a hard drive, no?
with the source code and a photographic memory you could beat it, read this to understand why processing information does not equal intelligence if you used deep blues program to play checkers it would lose as it was designed specifically to play chess and has no ability to learn anything outside of the world of chess
Written by :FireTom
Now when it comes to the "catching up" that computers would have to do.... how about CERN or - more specific - how about "the Grid"? It seems that we're not as far from generating a (web based) supercomputer as we might think.
it helps but its the rules for processing the data that is lacking. Without the rules on how to process sensory input (video pictures etc) its about as sentient as your toaster
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
BRONZE Member
old hand
929 posts
Location: Oregon, USA
old hand
929 posts
Location: Oregon, USA
Posted:
Just thought this might interest you guys...
https://www.ted.com/talks/view/id/125
Just thought this might interest you guys...
https://www.ted.com/talks/view/id/125
#homeofpoi -- irc.newnet.net Come talk to us we're bored
Warning: Please Do Not Jump On The Seals
GOLD Member
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
HOP Mad Doctor
13,925 posts
Location: San Francisco, CA, USA
Posted:
Well, microtubules play a strange, and ill-defined role in neuronal processing. Then there is post-transcriptional modification of all sorts of neuronal proteins ranging from proteins that control synaptic vesicle release to receptors to switching of the voltage threshold for voltage-gated sodium channels.
It's hideously complicated.
Written by :ben-ja-men
I agree that there is alot more complexity in the neurons as the firing threshold inside of each one changes with relation to the states of the microtubules inside it, which act as a cellular automata to determine the threshold and are influenced by the incoming signals, but in essence I don't see a difference in the principles behind how they process information.
Well, microtubules play a strange, and ill-defined role in neuronal processing. Then there is post-transcriptional modification of all sorts of neuronal proteins ranging from proteins that control synaptic vesicle release to receptors to switching of the voltage threshold for voltage-gated sodium channels.
It's hideously complicated.
-Mike
Certified Mad Doctor and HoP High Priest of Nutella
A buckuht n a hooze! -Valura
Stargazer
6,650 posts
6,650 posts
Posted:
Ben: Bottom line is you say AI is not to be achieved?
"Intelligence" and "the ability to act (intelligent)" is not necessary the same - to me. More so I think intelligence has nothing to do with a body.
It's true - put a program for chess into a (for say) automobile production and the outcome would be disastrous. Then again: let the area in your brain, responsible for movement, govern your taste buds... maybe you'd even start to enjoy Marmite
Hmm, thinking of it - with the source code and a photographic memory I would be King in Vegas... who can beat the right computer in a game of chess? I side you that "computing power" does not equal "intelligence" - at least not "higher intelligence" or "consciousness". Yet: is the ability to process informations (fast) not the foundation to intelligence?
Sure, if it would be "easy" - hey it would be present already. We are facing the amazing fact that humans are the only one on the planet who developed this specific kind of intelligence, even though monkeys are genetically so similar to our selves.
What is that 3% genetic difference? Why does it make this much of a difference and what is the "landslide effect" or "critical mass"?
So with higher processing capability and all the information stored, all the programs available it is possible to generate AI... I'd say. There are determinations of for say facial expression and their translation (anger, stress, comfortable, happy...) same is possible for to determine the human voice, etc. already in place. Maybe it's only the "missing link" that is .... well, missing.
A program might be able to only correct he spelling of a sentence and not get the meaning. Same applies to a secretary, who is transcribing a recorded tape into the computer. S/he would afterwards not be able to tell you the meaning of what s/he has transcribed.
There is also a difference between "intelligence" and "awareness" as I see it.
Mayeb I'm just sending out a garbled message, but as I se it - with all the information about human "dis/likes" stored in the www it is possible to recreate human patterns. I might have to meditate more upon it and dig deeper into respective literature to get across what I mean. At the same time, there are ppl out there who studied this topic for a lifetime and still have not accomplished a satisfactory result...
Ben: Bottom line is you say AI is not to be achieved?
"Intelligence" and "the ability to act (intelligent)" is not necessary the same - to me. More so I think intelligence has nothing to do with a body.
It's true - put a program for chess into a (for say) automobile production and the outcome would be disastrous. Then again: let the area in your brain, responsible for movement, govern your taste buds... maybe you'd even start to enjoy Marmite
Hmm, thinking of it - with the source code and a photographic memory I would be King in Vegas... who can beat the right computer in a game of chess? I side you that "computing power" does not equal "intelligence" - at least not "higher intelligence" or "consciousness". Yet: is the ability to process informations (fast) not the foundation to intelligence?
Sure, if it would be "easy" - hey it would be present already. We are facing the amazing fact that humans are the only one on the planet who developed this specific kind of intelligence, even though monkeys are genetically so similar to our selves.
What is that 3% genetic difference? Why does it make this much of a difference and what is the "landslide effect" or "critical mass"?
So with higher processing capability and all the information stored, all the programs available it is possible to generate AI... I'd say. There are determinations of for say facial expression and their translation (anger, stress, comfortable, happy...) same is possible for to determine the human voice, etc. already in place. Maybe it's only the "missing link" that is .... well, missing.
A program might be able to only correct he spelling of a sentence and not get the meaning. Same applies to a secretary, who is transcribing a recorded tape into the computer. S/he would afterwards not be able to tell you the meaning of what s/he has transcribed.
There is also a difference between "intelligence" and "awareness" as I see it.
Mayeb I'm just sending out a garbled message, but as I se it - with all the information about human "dis/likes" stored in the www it is possible to recreate human patterns. I might have to meditate more upon it and dig deeper into respective literature to get across what I mean. At the same time, there are ppl out there who studied this topic for a lifetime and still have not accomplished a satisfactory result...
the best smiles are the ones you lead to
GOLD Member
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
just lost .... evil init
2,474 posts
Location: Adelaide, Australia
Posted:
sure but at the end of the day regardless of how complex the system its just processing streams of information.
No im saying that AI is not to be achieved by the methods put forward in this thread
yes this is the embodiment issue previously mentioned
its a requirement and it creates a medium in which "intelligence" (whatever that is) can emerge, however without the mechanism to process information in a meaningful way its still just as "intelligent" as your toaster
No they are "cheap tricks" that apply to very very specific situations, facial recognition basically searches for specific patterns, if you tried to locate a snakes face with the same software it would not work, the "intelligence" is that of the programmer who has recognised a specific pattern and written a program for it, the program has no embodiment in the real world it is simply performing symbol manipulation (see Chinese Room Argument mentioned earlier). If you want "intelligence" you need a robust system that can operate autonomously and has meaning that is embodied in the real world and even then its questionable as to whether its just a good trick or not (see Eliza https://www-ai.ijs.si/eliza/eliza.html experiments have been done where ppl have thought they where talking to a real person for therapy and bonded very strongly to eliza, this is the joys of anthropomorphization, eliza just uses simple rules to rearrange statements as leading questions, clever programming yes, intelligence no)
If there where AI it could use the net to build a good profile of humans BUT the net cant be used as a source to create AI.
Written by :Doc Lightning
Well, microtubules play a strange, and ill-defined role in neuronal processing. Then there is post-transcriptional modification of all sorts of neuronal proteins ranging from proteins that control synaptic vesicle release to receptors to switching of the voltage threshold for voltage-gated sodium channels.
It's hideously complicated.
sure but at the end of the day regardless of how complex the system its just processing streams of information.
Written by :FireTom
Ben: Bottom line is you say AI is not to be achieved?
No im saying that AI is not to be achieved by the methods put forward in this thread
Written by :FireTom
More so I think intelligence has nothing to do with a body.
yes this is the embodiment issue previously mentioned
Written by :FireTom
Yet: is the ability to process informations (fast) not the foundation to intelligence?
its a requirement and it creates a medium in which "intelligence" (whatever that is) can emerge, however without the mechanism to process information in a meaningful way its still just as "intelligent" as your toaster
Written by :FireTom
So with higher processing capability and all the information stored, all the programs available it is possible to generate AI... I'd say. There are determinations of for say facial expression and their translation (anger, stress, comfortable, happy...) same is possible for to determine the human voice, etc. already in place. Maybe it's only the "missing link" that is .... well, missing.
No they are "cheap tricks" that apply to very very specific situations, facial recognition basically searches for specific patterns, if you tried to locate a snakes face with the same software it would not work, the "intelligence" is that of the programmer who has recognised a specific pattern and written a program for it, the program has no embodiment in the real world it is simply performing symbol manipulation (see Chinese Room Argument mentioned earlier). If you want "intelligence" you need a robust system that can operate autonomously and has meaning that is embodied in the real world and even then its questionable as to whether its just a good trick or not (see Eliza https://www-ai.ijs.si/eliza/eliza.html experiments have been done where ppl have thought they where talking to a real person for therapy and bonded very strongly to eliza, this is the joys of anthropomorphization, eliza just uses simple rules to rearrange statements as leading questions, clever programming yes, intelligence no)
Written by :FireTom
I might have to meditate more upon it and dig deeper into respective literature to get across what I mean. At the same time, there are ppl out there who studied this topic for a lifetime and still have not accomplished a satisfactory result...
If there where AI it could use the net to build a good profile of humans BUT the net cant be used as a source to create AI.
Our deepest fear is not that we are inadequate. Our deepest fear is that we are powerful beyond measure. It is our light, not our darkness that most frightens us. We ask ourself, who am I to be brilliant, gorgeous and talented? Who are you NOT to be?
© 1998 - 2024 Home of Poi
All prices converted to United States dollars using XE.com