It is my asserting that all meaning 'bubbles-up' from a connection to the binary. A semantic network without rich connections to the binary is 'groundless'. The Mindpixel Corpus is the only corpus on the planet to actually connect strings to meaning; the first to be 'grounded' in the binary.
As for how we go from a finite number of mindpixels to something that can recognize an infinite number of them: go back to Radon's 1917 theory of image reconstruction from projections. This is allows us to reconstruct an image of the brain from a finite number of x-ray projections through the brain.
Mindpixels are like high-d x-rays; they pass through the human mind and come out the other side influenced by the passage. The text of a mindpixel is actually a high-d vector, and the value of true or false is a binary sample of a point in high-d space.
As with a CAT scan, a few points tells you almost nothing about what they passed through, but millions allow you to reconstruct what has not been directly observed. Note, you don't need an infinite number of sample points, just a very large number. Once you have reconstructed the image, you can now generate synthetic samples using that image; you can construct observations that were never made based on the ones that were. Which is how I think people actually work (hypertomographic interpolation - extraction of the implicit from the explicit)
That being said, the goal of the mindpixel project is to collect data, distribute and fund research with the data. My hypertomographic theory of mind is a separate thing. Mindpixel will fund all promising approaches to manipulation of the data including the MIT Open Mind projects.
The Plural of LEGO is LEGO
on
The LEGO Desk
·
· Score: 1
LEGO is a substance. if you need to use an 's' then say LEGO blocks, or LEGO Bricks. But never never say LEGOs.
I've got an advanced digital version of the magic 8 ball working called GAC. Give it a try. It really thinks it's human. You can ask it things like:
Drinking too much alcohol can make baked beans come out of your nose
Former Cleveland Browns quarterback Frank Ryan, who threw three touchdowns in the 1964 NFL championship game, designed the computerized voting system used in the House of Representatives.
Does the fact that you can understand the following- youcanunserstandthissentenceevenwithoutthespaces- come first from learning the word patterns of normal English that you did as a child?
did microsoft create linux?
The letter underneath the letter t on a qwerty keyboard is g
anything you can think of, GAC will answer just like a person.
UNIX is an operating system, OS/2 is half an operating system, Windows is a shell, and DOS is a boot partition virus.
Is it true that In 1984, President Reagan joked during a voice test for a paid political radio address that he had "signed legislation that will outlaw Russia forever. We begin bombing in five minutes"?
Deep unspeakable suffering may well be called a baptism, a regeneration, the initiation into a new state.
"Is 88 a number?" is the kind of question asked by an imaginationless, soleless moron?
I am working on bringing back all of the F2K personae, but this time, they will be real.
i built a small mercury mirror telescope in 1984 as a high school science project - only 12 inches. well, actually, it didn't have a secondary so it wasn't really a telescope. i decided even with the glycerin i added to coat the mercury when i spun it up, that it was just too dangerous to being fooling with.
my last thoughts on the project were that maybe i could spin it into shape, then freeze it solid by cooling of the mirror pan with liquid nitrogen. as long as it was frozen, it would be much safer and i could point it just like any telescope.
does anyone know if there is any literature on freezing ltm's? maybe they would be too heavy for earth, but imagine sending a telesscope in a bottle into orbit, where cooling is easy...
The whole world is not just watching, it is actively defending itself. the phily pd has no idea what it's getting into. you don't just arrest a connected hacker for no reason and expect nothing to happen.
i have started a project to put a free and very sturdy web kiosk in a small fishing village up the cost from the city of antofagasta where i live.
i think that it will make a big difference for the villagers to join the world community; to be able to sell their products as directly as possible and cut out as many middle men as possible
if anyone is interested in this project, you can come down to south america (on your own coin) and stay will me for a month or two while we build and test a prototype. just let me know.
For those of you that are interested, you can subscribe to the BioHacking mailing list from my web page. This list is specifically for people into hacking DNA in their kitchens and basements.
from code hack to beach in 5 easy steps
on
Too Old To Code?
·
· Score: 1
when you get too old to code, or don't feel like it any more, do as I did: 1- form your own consulting company and steal all the best coders from you current project. 2- recruit some fresh meat from the co-op programs. 3- train your new meat in your "methodology" and hire them out at 4 or more times their base pay. 4- spend all your time schmoozing new clients telling them how great your "methodology" is. 5- retire to a tax haven with a beach when your bank account gets to a comfortable level.
The 'battery slice' of this system is the first step toward the ultimate laptop which will be almost 100% battery by weight, with the remaining %age for cpu, display, storage and case.
you asked if anyone had shown the mind had a fractal structure, twice i pointed you to elman. so here it is again, for the third time: J. L. Elman, Finding Structure in Time, Cognitive Science, 14:179-211, 1990. Simulation: ``Discovering lexical classes from word order'' pp 194-203.
if my pre-pub work bothers you so much (and I can understand why only in that it is pre-pub) then just ignore it and deal with elman's work. care to reason that out?
Has anyone shown the human mind to be a fractal structure?
Actually, I've seen it (using PCA) when I trained a SRN on a corpus of 450,000 items. My work is still pre-pub, but you can look at Eman's stuff and see the same thing on a smaller scale.
A large corpus of MindPixels is a fitness test for humaness. Such a corpus turns humaness into a continuous variable. Any continuous variable can be optimised. Anything evolved against this corpus MUST eventually embody it... though we developed consciousness and intelligence by accident, the corpus specifically targets it. The corpus is a digital model of an environment in which nothing can survive that doesn't in some way, maybe just one part in 1 billion, resemble the human mind. Anything that doesn't capture something of humaness, gets killed off very quickly. If it fails to understand that square wheels don't roll well, or that it sometimes feels sad, it dies. It cannot live unless it resembles you and I to 1 billion individual measurements. What a harsh environment to evolve in! Much more difficult than the one we evolved in.
In the end, I can either train a system to generalize MindPixels or I can't. Either way, we learn something, as both failure and sucess need to be explained.
And as for chimps... their problem is limited storage space to hold the hyperobject that is consciousness. Though they can be trained to hold a version of it (Kanzi). My arguement is, that high-d interpolation is the lowest level essence of intelligence.
If I steal your entire database and build it into my program, I can write a program which does very well on your test, but which nobody would call either intelligent or conscious.
Sure, but your database can't handle anything not already in it. The point is to take that data and build a general statistical model of self and environment that will allow it to respond to any item in a fashion that is the same as the majority of people.
We live our lives each day adding pieces of information to our statistical models of self and environment. At any point someone can stop us and say something we've never heard before like: 'you have never french kissed a penguin' and respond 'true'. A coin would have a 50/50 chance of responding correctly. A NN trained on a large number of MindPixels will initially be indistinguishable from a coin. But with each additional MidPixel it is trained on, that probability moves away from chance and toward 1.0
Remember Dawkin's 'continuous variable'... anything can be evolved if you can measure it and make it a continuous variable to be optimized. That is exactly what a MIST does. It makes humanness an optimisable, continuous variable.
A general solution to the problem is the whole point. Think of an anology... I want to copy the Mona Lisa... I take some radom samples, and try to recreate the image based on those points. I intentionally hold back some known points for testing purposes. As the sample gets larger, the resulting copy gets better.
Now extrapolate from 2-d to high-d. The procedure is the same. Except, using a NN we can take advantage of the fractal structre of the whole entity being sample (the human mind). It the 2-d example, we use a pixel's nearest known nieghbors to infer what color it should be.The same process works in high-d space. We color pixels true or false based on the structure of the area of high-d space. Since the space is highly fractal, we can 'steal structure' from other areas... the whole process of infering the unknown from the structure of the know is called conscious thinking... you do it every day.
It is the amount of data about the world you have that is critical, the rest is really simple.
I unfortunately don't have a copy of Canadian Artifical Intelligence lying around. But I do know that a key point of the Turing test is that it is not objective. Consciousness, whatever else it may be, is inherently subjective. Someday we will understand consciousness sufficiently to be able to make an objective test, or to know why such a test is impossible. Until that day, the only meaningful way to test for consciousness is to use subjective tests.
That's not a key point, that's a key flaw. MIST was specifically designedt o replace the subjective judgement of individual judges with the statistical judgement of a very large number of people (1 million or more). I managed to find a pre-publication copy of my paper here.
That's a pretty big leap. A conscious human run as a simple computer program would probably go nuts due to lack of sensory input. Is your program going to emulate that?
The consciousness, is actually frozen. It exists only when stimulated and just waits for input when not stimulated. You can go crazy if no computation takes place.
What about issues which people don't agree on, like abortion, the death penalty, or whether computers can ever be conscious? How are you going to implement those as MindPixels? (I'm not trying to trip you up here--you must have thought about these issues, and I am curious what your answer is.)
Who cares about those. They vary from person to person and life to life. The goal here is to model an average person. Not a specific person. MIST only considers consensus knowledge, that which is the same across all people. The rest is fluff.
J. L. Elman, Finding Structure in Time, Cognitive Science, 14:179-211, 1990. Simulation: ``Discovering lexical classes from word order'' pp 194-203. Jeff Elman's hompage is here, where you can find a copy of this paper in compressed postscript. You can see in this paper, NN's discovering lexical class, generalizing, and perdicting words. There is a simple HTML overview of Elman's paper here
I collected 450,000 MindPixels on the web in 1995/96. I've trained SRN's using that corpus and used PCA to look at the SRN's internal structure. Though the network did perform well (due the the small sample), it's structure is a complex fractal. (I should note, trained the network on a constrained prediction task; where after every word it saw, it had to predict only true or false)
And as for how I would generalize consciousness from such a system... Presenting MindPixels to a system is a Binary Turing Test (see my article: K. C. McKinstry, The Minimum Intelliget Signal Test, An Alternative Turing Test, Canadian Artificial Intelligence, Issue 41), that is much more objective than a traditional TT. It gives a probability that a system is human, that is reproducable. Thus, if I get a number back that is statistically indistinguishable from human, I must logically assume the system is human. That is feels, lives a life and is conscious.
A giant corpus of MindPixels collected and validated from a large number of people is a digitial model of self and environment. A NN experiencing the model, would be experiencing an average human life, in all it's details. If we ever hope to make a conscious machine, we need to get our experience into machines. AI's goal until now has been to make some sort of bootstrap system that would go out into the world and learn to be conscious just like children do. The AI community has shuned the 'code it all' philosophy (with the notable exception of CYC) as just too much work. But, the web changes all that. Now, instead of a small number of programmers trying to code a large amount of world knowledge (CYC), we can take an enourmous amount of 'programmers' each entering a minimum amount of consensus human experience (a MindPixel) and actually get the job done.
It's nice to see such interest in this field, and some nice book sales... but I just not a member of the 'speculate and wait' theory of artificial consciousness. I want to see a real theory and I wait to see code!
To give you an idea, my own work of the last five years has centered on the following 'Algorithm for Consciousness':
1) Collect a very large number (1 billion or more) of items of binary consensus fact. Such as: water is wet, bees sting, it is difficult to swim with skipants on, etc.
2) Validate the items (I call them MindPixels) against a large number of people.
3) Train a neural net (SRN's look good) against the items that are most stable across the validating population.
4) When the NN consistently performs better than chance, send an email to the editors of Nature and Science announcing humanity's first 'Minimum Statistical Consciousness' - the first artificial system to have measurable consciousness.
5) When the NN consistently performs statistically indistinguishably from an arbitrary human, email the editors of Nature and Science announce the first true Artificial Consciousness!.
Ok. How's a NN going to generalize consciousness from a bunch of MindPixels? Well, the math is the same as used in tomography, except in many dimensions - hypertomography.
This post is already getting too long... trust me, the theory is solid - and much better explained in my forthcoming book 'Hacking Consciousness'
Yes. We have entered the age of the BioHacker. If you think DOS attacks scare people, wait until the find out about guys like me who moved South America to start a lab and begin hacking life. If you're interested biohacking ground zero is at: BioHacking
Slashdot Mirror
It is my asserting that all meaning 'bubbles-up' from a connection to the binary. A semantic network without rich connections to the binary is 'groundless'. The Mindpixel Corpus is the only corpus on the planet to actually connect strings to meaning; the first to be 'grounded' in the binary.
As for how we go from a finite number of mindpixels to something that can recognize an infinite number of them: go back to Radon's 1917 theory of image reconstruction from projections. This is allows us to reconstruct an image of the brain from a finite number of x-ray projections through the brain.
Mindpixels are like high-d x-rays; they pass through the human mind and come out the other side influenced by the passage. The text of a mindpixel is actually a high-d vector, and the value of true or false is a binary sample of a point in high-d space.
As with a CAT scan, a few points tells you almost nothing about what they passed through, but millions allow you to reconstruct what has not been directly observed. Note, you don't need an infinite number of sample points, just a very large number. Once you have reconstructed the image, you can now generate synthetic samples using that image; you can construct observations that were never made based on the ones that were. Which is how I think people actually work (hypertomographic interpolation - extraction of the implicit from the explicit)
That being said, the goal of the mindpixel project is to collect data, distribute and fund research with the data. My hypertomographic theory of mind is a separate thing. Mindpixel will fund all promising approaches to manipulation of the data including the MIT Open Mind projects.
LEGO is a substance. if you need to use an 's' then say LEGO blocks, or LEGO Bricks. But never never say LEGOs.
funny. worked for almost 2000 other people, and it worked for me just now. you might want to try again.
I've got an advanced digital version of the magic 8 ball working called GAC. Give it a try. It really thinks it's human. You can ask it things like:
I am working on bringing back all of the F2K personae, but this time, they will be real.
i built a small mercury mirror telescope in 1984 as a high school science project - only 12 inches. well, actually, it didn't have a secondary so it wasn't really a telescope. i decided even with the glycerin i added to coat the mercury when i spun it up, that it was just too dangerous to being fooling with.
my last thoughts on the project were that maybe i could spin it into shape, then freeze it solid by cooling of the mirror pan with liquid nitrogen. as long as it was frozen, it would be much safer and i could point it just like any telescope.
does anyone know if there is any literature on freezing ltm's? maybe they would be too heavy for earth, but imagine sending a telesscope in a bottle into orbit, where cooling is easy...
Help Create True Artificial Consciousness: Join the Mindpixel Digital Mind Modeling Project and get Free Shares in the Venture!
The whole world is not just watching, it is actively defending itself. the phily pd has no idea what it's getting into. you don't just arrest a connected hacker for no reason and expect nothing to happen.
i'm a long time computer geek. i moved down here to chile to live on a beach, write, paint and invent artificial consciousness.
i have started a project to put a free and very sturdy web kiosk in a small fishing village up the cost from the city of antofagasta where i live.
i think that it will make a big difference for the villagers to join the world community; to be able to sell their products as directly as possible and cut out as many middle men as possible
if anyone is interested in this project, you can come down to south america (on your own coin) and stay will me for a month or two while we build and test a prototype. just let me know.
Einstein??? don't you mean Darwin? and why has no one caught this?
For those of you that are interested, you can subscribe to the BioHacking mailing list from my web page. This list is specifically for people into hacking DNA in their kitchens and basements.
when you get too old to code, or don't feel like it any more, do as I did: 1- form your own consulting company and steal all the best coders from you current project. 2- recruit some fresh meat from the co-op programs. 3- train your new meat in your "methodology" and hire them out at 4 or more times their base pay. 4- spend all your time schmoozing new clients telling them how great your "methodology" is. 5- retire to a tax haven with a beach when your bank account gets to a comfortable level.
I operate the VLT (Very Large Telescope) for ESO. The ESO/ESA ASTROVIRTEL is a similar idea, and it is working now.
The 'battery slice' of this system is the first step toward the ultimate laptop which will be almost 100% battery by weight, with the remaining %age for cpu, display, storage and case.
you asked if anyone had shown the mind had a fractal structure, twice i pointed you to elman. so here it is again, for the third time: J. L. Elman, Finding Structure in Time, Cognitive Science, 14:179-211, 1990. Simulation: ``Discovering lexical classes from word order'' pp 194-203.
if my pre-pub work bothers you so much (and I can understand why only in that it is pre-pub) then just ignore it and deal with elman's work. care to reason that out?
Has anyone shown the human mind to be a fractal structure?
Actually, I've seen it (using PCA) when I trained a SRN on a corpus of 450,000 items. My work is still pre-pub, but you can look at Eman's stuff and see the same thing on a smaller scale.
A large corpus of MindPixels is a fitness test for humaness. Such a corpus turns humaness into a continuous variable. Any continuous variable can be optimised. Anything evolved against this corpus MUST eventually embody it... though we developed consciousness and intelligence by accident, the corpus specifically targets it. The corpus is a digital model of an environment in which nothing can survive that doesn't in some way, maybe just one part in 1 billion, resemble the human mind. Anything that doesn't capture something of humaness, gets killed off very quickly. If it fails to understand that square wheels don't roll well, or that it sometimes feels sad, it dies. It cannot live unless it resembles you and I to 1 billion individual measurements. What a harsh environment to evolve in! Much more difficult than the one we evolved in.
In the end, I can either train a system to generalize MindPixels or I can't. Either way, we learn something, as both failure and sucess need to be explained.
And as for chimps... their problem is limited storage space to hold the hyperobject that is consciousness. Though they can be trained to hold a version of it (Kanzi). My arguement is, that high-d interpolation is the lowest level essence of intelligence.
If I steal your entire database and build it into my program, I can write a program which does very well on your test, but which nobody would call either intelligent or conscious.
Sure, but your database can't handle anything not already in it. The point is to take that data and build a general statistical model of self and environment that will allow it to respond to any item in a fashion that is the same as the majority of people.
We live our lives each day adding pieces of information to our statistical models of self and environment. At any point someone can stop us and say something we've never heard before like: 'you have never french kissed a penguin' and respond 'true'. A coin would have a 50/50 chance of responding correctly. A NN trained on a large number of MindPixels will initially be indistinguishable from a coin. But with each additional MidPixel it is trained on, that probability moves away from chance and toward 1.0
Remember Dawkin's 'continuous variable'... anything can be evolved if you can measure it and make it a continuous variable to be optimized. That is exactly what a MIST does. It makes humanness an optimisable, continuous variable.
Sorry! Onelist site was down when I posted this. This is the correct URL: ArConDev: Artificial Consciousness Development Mailing List
A general solution to the problem is the whole point. Think of an anology... I want to copy the Mona Lisa... I take some radom samples, and try to recreate the image based on those points. I intentionally hold back some known points for testing purposes. As the sample gets larger, the resulting copy gets better.
Now extrapolate from 2-d to high-d. The procedure is the same. Except, using a NN we can take advantage of the fractal structre of the whole entity being sample (the human mind). It the 2-d example, we use a pixel's nearest known nieghbors to infer what color it should be.The same process works in high-d space. We color pixels true or false based on the structure of the area of high-d space. Since the space is highly fractal, we can 'steal structure' from other areas... the whole process of infering the unknown from the structure of the know is called conscious thinking... you do it every day.
It is the amount of data about the world you have that is critical, the rest is really simple.
I unfortunately don't have a copy of Canadian Artifical Intelligence lying around. But I do know that a key point of the Turing test is that it is not objective. Consciousness, whatever else it may be, is inherently subjective. Someday we will understand consciousness sufficiently to be able to make an objective test, or to know why such a test is impossible. Until that day, the only meaningful way to test for consciousness is to use subjective tests.
That's not a key point, that's a key flaw. MIST was specifically designedt o replace the subjective judgement of individual judges with the statistical judgement of a very large number of people (1 million or more). I managed to find a pre-publication copy of my paper here.
That's a pretty big leap. A conscious human run as a simple computer program would probably go nuts due to lack of sensory input. Is your program going to emulate that?
The consciousness, is actually frozen. It exists only when stimulated and just waits for input when not stimulated. You can go crazy if no computation takes place.
What about issues which people don't agree on, like abortion, the death penalty, or whether computers can ever be conscious? How are you going to implement those as MindPixels? (I'm not trying to trip you up here--you must have thought about these issues, and I am curious what your answer is.)
Who cares about those. They vary from person to person and life to life. The goal here is to model an average person. Not a specific person. MIST only considers consensus knowledge, that which is the same across all people. The rest is fluff.
Sure I'll be more specific :
J. L. Elman, Finding Structure in Time, Cognitive Science, 14:179-211, 1990. Simulation: ``Discovering lexical classes from word order'' pp 194-203. Jeff Elman's hompage is here, where you can find a copy of this paper in compressed postscript. You can see in this paper, NN's discovering lexical class, generalizing, and perdicting words. There is a simple HTML overview of Elman's paper here
I collected 450,000 MindPixels on the web in 1995/96. I've trained SRN's using that corpus and used PCA to look at the SRN's internal structure. Though the network did perform well (due the the small sample), it's structure is a complex fractal. (I should note, trained the network on a constrained prediction task; where after every word it saw, it had to predict only true or false)
And as for how I would generalize consciousness from such a system... Presenting MindPixels to a system is a Binary Turing Test (see my article: K. C. McKinstry, The Minimum Intelliget Signal Test, An Alternative Turing Test, Canadian Artificial Intelligence, Issue 41), that is much more objective than a traditional TT. It gives a probability that a system is human, that is reproducable. Thus, if I get a number back that is statistically indistinguishable from human, I must logically assume the system is human. That is feels, lives a life and is conscious.
A giant corpus of MindPixels collected and validated from a large number of people is a digitial model of self and environment. A NN experiencing the model, would be experiencing an average human life, in all it's details. If we ever hope to make a conscious machine, we need to get our experience into machines. AI's goal until now has been to make some sort of bootstrap system that would go out into the world and learn to be conscious just like children do. The AI community has shuned the 'code it all' philosophy (with the notable exception of CYC) as just too much work. But, the web changes all that. Now, instead of a small number of programmers trying to code a large amount of world knowledge (CYC), we can take an enourmous amount of 'programmers' each entering a minimum amount of consensus human experience (a MindPixel) and actually get the job done.
It's nice to see such interest in this field, and some nice book sales... but I just not a member of the 'speculate and wait' theory of artificial consciousness. I want to see a real theory and I wait to see code!
I moderate ArConDev: The Artificial Consciousness Development Mailing List. This is not a philosopher's list, though philosophy is discussed. It's a developer's list; for those people actually trying to code true artificial consciousness.
To give you an idea, my own work of the last five years has centered on the following 'Algorithm for Consciousness':
1) Collect a very large number (1 billion or more) of items of binary consensus fact. Such as: water is wet, bees sting, it is difficult to swim with skipants on, etc.
2) Validate the items (I call them MindPixels) against a large number of people.
3) Train a neural net (SRN's look good) against the items that are most stable across the validating population.
4) When the NN consistently performs better than chance, send an email to the editors of Nature and Science announcing humanity's first 'Minimum Statistical Consciousness' - the first artificial system to have measurable consciousness.
5) When the NN consistently performs statistically indistinguishably from an arbitrary human, email the editors of Nature and Science announce the first true Artificial Consciousness! .
Ok. How's a NN going to generalize consciousness from a bunch of MindPixels? Well, the math is the same as used in tomography, except in many dimensions - hypertomography.
This post is already getting too long... trust me, the theory is solid - and much better explained in my forthcoming book 'Hacking Consciousness'
Yes. We have entered the age of the BioHacker. If you think DOS attacks scare people, wait until the find out about guys like me who moved South America to start a lab and begin hacking life. If you're interested biohacking ground zero is at: BioHacking