Jump to content
Objectivism Online Forum

The Singularity

Rate this topic


WI_Rifleman

Recommended Posts

As for the math function, when the x range is time, progress is limited, because one does not have control over the progression of time. I don't think you want to get into a mathematics rigorous definition of function and variables because I have a second bachelor's in applied math and I am pretty sure I am right there.

Of course I want to get into that math, but just as far as this topic, nothing too deep. If you were a construction worker, then I would want to stay away from the subject completely, and ask you to do the same...

Fists off, sure, you did not say what you mean by unlimited, so I was wrong to assume you define a limited function as one that has all its values from a limited [a,b] interval, for all x's. (like sin(x) ) I shall now change my ways, and ask for clarifications instead.

You said this:

I am not saying it is impossible but I believe man can do better than exponential or linear. He is unlimited in his capacity to reason and find ways of explaining and replicating his environment. Transhumanists should have a higher respect for what is human.

You know math, so you are aware that linear and exponential are math terms, they refer to specific types of functions. So there must be a type of function of time (Kurzweil is talking about function of time, not something else) that's "better", and somehow unlimited (I can't really wrap my mind around your kind of unlimited yet, do you mean something like f(x)=1/(1-x)?) , which has been shown to describe the human capacity to reason.

Please, most importantly, explain to me in math terms what you mean by better and unlimited. Then, how would that unlimited and better function you referred to, look like?

Then, I will have understood why Kurzweil has less than the appropriate amount of respect for the human capacity to reason. But if you say that you cannot measure the human capacity to reason, then what possible meaning could the word unlimited have? If you know it's unlimited, you must've measured it somehow, and ran out of measuring tape.

Edited by Jake_Ellison
Link to comment
Share on other sites

  • Replies 60
  • Created
  • Last Reply

Top Posters In This Topic

Of course I want to get into that math, but just as far as this topic, nothing too deep. If you were a construction worker, then I would want to stay away from the subject completely, and ask you to do the same...

GREAT, it doesn't say anything about your background on your profile so I don't know where you are at, please don't take any offense, I just don't know what level of math capability you are at but we'll try.

Fists off, sure, you did not say what you mean by unlimited, so I was wrong to assume you define a limited function as one that has all its values from a limited [a,b] interval, for all x's. (like sin(x) ) I shall now change my ways, and ask for clarifications instead.

I mean unlimited in the conventional definition of unlimited i.e. he has no limits to his understanding or no one can set limits to his rational faculty. He has free will which means he is free and therefore has no limits that can be placed logically on him. (except within time limits like the crow epistemology stuff: the context here is all long term anyway)

You know math, so you are aware that linear and exponential are math terms, they refer to specific types of functions. So there must be a type of function of time (Kurzweil is talking about function of time, not something else) that's "better", and somehow unlimited (I can't really wrap my mind around your kind of unlimited yet, do you mean something like f(x)=1/(1-x)?) , which has been shown to describe the human capacity to reason.

Please, most importantly, explain to me in math terms what you mean by better and unlimited. Then, how would that unlimited and better function you referred to, look like?

Lets say that for arguments sake that Kurzweil is right that he has a function that he has put through several data points and the curve is defined by those points as an exponential function i.e. let's take Kurzweil at his word. It is not linear but exponential. There is a subset of applied mathematics that searches for r^2 deviation value. This is common in statistics (Ki square value) and data analysis fields. If you have used a trend-line in excel you know of a similar function. You can even give excel a form of an equation and it will solve for the constants for an r^2 value. The R^2 value is the standard deviation of the points to the function and a value of 1 means that the point exactly match a r^2. Almost never do you get an R^2 value of 1 so lets say Kurzweil has one of .99, which is common for good data. The function is then set. Or is it? Does this now mean that there will not be a nuclear war, an economic crisis that wipes out civilization in a conventional war or anything that would stop Kurzweil's dream of a singularity for 500 years. There has been times in history when man has stepped backward in technological progress and this is why Kurweil's functions do add up.

He is claiming to predict the future very accurately. He might not have the exact curve but he has picked the family of curves and this limits him to an always improving function. The derivative of an exponential function is always positive. and this is even on his wiki: http://en.wikipedia.org/wiki/Raymond_Kurzweil

Daniel Lyons, writing in Newsweek, criticized Kurzweil for some of his predictions which turned out to be wrong; such as the economy continuing to boom from the 1998 dot-com through 2009, a US company having a market capitalization of more than $1 trillion, a supercomputer achieving 20 petaflops, speech recognition being in widespread use and cars that would drive themselves using sensors installed in highways; all by 2009[63].

Then, I will have understood why Kurzweil has less than the appropriate amount of respect for the human capacity to reason. But if you say that you cannot measure the human capacity to reason, then what possible meaning could the word unlimited have? If you know it's unlimited, you must've measured it somehow, and ran out of measuring tape.

Unlimited as in not limited to an output that is always positive or unlimited as in not limited to a function on a graph. You don't have to measure "unlimited" it is unlimited by definition. just like infinity can not be measured( except in terms of other infinities by deduction in set theory) See above. This is also the problem with determinism as things get more and more predictable the capability of free will drops. You have no free will because things are predictable. They are at cross purposes deductively.

All of this is not to mention that inductively none of his prediction capabilities make sense. Who has a tap on what the economy and technology is going to do? Man the individual certainly but it is not Mr. Kurzweil.

Edited by Paul Lemke
Link to comment
Share on other sites

Edited because somehow I missed a whole page of this thread, haha:

Paul Lemke:

Kurzweil has stated that there maybe slow-downs or even downward movement in his curve, depending on conditions. He actually has identified some slow-downs (great depression for example) where technology advanced a bit slower than predicted. Massive wars might make it go negative. His point is that his curve is an average, it is the best fit ignoring all the slow downs (or more properly, accounting for them by decreasing the overall rate of growth somewhat).

Separately, on the main topic of the singularity and the possibility of the creation of an actual consciousness equivalent in power to that of a biological human in a computer:

To say that we will never be able to create a consciousness in any form of a computational substrate is to ignore the fact that we do so all the time when we have babies. The real question is if we can create one without going through any remotely standard reproductive channels, and even more specifically, to do so without using a biological substrate. Well, in order to deny this possibility you must either 1) assert that the brain is so complex that we will never have enough computational power to emulate it, meaning that there is a cap at the maximum theoretical amount of computational power that man can pack into a computer of arbitrary size or 2) to assert that somehow human consciousness has to have a biological substrate, which is to assert the existence of a non-corporeal soul totally outside the realm of physics (since given enough computing power, we can emulate any physical system of any size, at least theoretically).

Since 2 is just ridiculous and might as well come right out of the mouth of Jesus, I'll only address 1.

The brain is about one kilogram of matter. Given access to any amount of matter, it seems absolutely certain that we could emulate its behavior down to the level of atoms. More important though, we shouldn't have to do that. While the behavior of synapses is likely affected by the dynamics of cell membranes and ion transport and concentration, any effect below the level of individual proteins is likely just noise, not actually computationally significant. Noise can be replicated in a computer by a randomness generator (if true randomness is needed, we could use a pile of radioactive isotope). Computation must be emulated. Neuroscientists estimate the needed computational power would be in the range of no more than about 10^28cps. That would be reached by 2045 at the present rate of increase in computer power. It may be far sooner than that since that is the high range of estimates.

There is no reason to think that we couldn't create something equal in size to the brain that could emulate its computational activities perfectly. The brain has a lot of junk it doesn't need for computing, so if we strip that all out, we don't even have to be as efficient as the brain is in order to match its power. If you emulate all of the activity of the brain, then you've created the same platform for consciousness as the human brain, and so why shouldn't it be conscious? If I can emulate its behavior, it will do the same things, and the emulated person (for example) will behave exactly as the person it was based on (allowing for natural differences resulting from different random influences by biology for the person and the noise generator for the computer). The emulated person would recognizably be the same as the biological one, and no one would be able to tell the difference. As a result, there is no reason to think anything has changed, or that somehow the computer isn't "conscious" like a human is conscious, since it behaves the same.

Edited by nanite1018
Link to comment
Share on other sites

That's outrageous.. Would you please write in an appropriate way!! He is already participating in the discussion.

Sorry, I edited it, I missed a whole page. Also, sorry for calling that person (not sure who) an idiot, though asserting that making a prediction is an assertion of determinism makes no sense at all, especially based on the passage you quoted from Kurzweil. I tend to be more loose with terms like that in conversation, since its pretty obvious that I never actually think the person is truly stupid when in person (got to get to know me I guess), but it is inappropriate on a forum, since tone of voice, facial expression, and experience with the person doesn't really exist.

Link to comment
Share on other sites

I can imagine reason just fine, in an entity that isn't trying to survive, but is instead just trying to solve some other problem.

But why will this hypothetical entity want to solve this "other problem"?

You are wrong, it is not axiomatically impossible to imagine reason used by a machine.

Reason needs volition to be exercised. And volition needs desires. And desires are possible only for a living being.

Fine. Artificial intelligence means man made intelligence, not machine intelligence.

To create an intelligent living organism all you need is to find a woman and to conceive a baby with her. That would be a man-made intelligence. There is nothing tricky in it.

Or you can grow a human baby or even some other intelligent creature in a vat. I'm fine with it. But the question is wether it is possible to give reason to a heap of metal.

Edit: grammar

Edited by Gavagai
Link to comment
Share on other sites

Sorry, I edited it, I missed a whole page. Also, sorry for calling that person (not sure who) an idiot, though asserting that making a prediction is an assertion of determinism makes no sense at all, especially based on the passage you quoted from Kurzweil. I tend to be more loose with terms like that in conversation, since its pretty obvious that I never actually think the person is truly stupid when in person (got to get to know me I guess), but it is inappropriate on a forum, since tone of voice, facial expression, and experience with the person doesn't really exist.

If you don't believe me as an eventual robot engineer. How about Rodney Brooks, one of the founders of iRobot, he started one of the only successful American robotics companies manufactures robots for a living.

"I'm not worried about the singularity anytime soon.(time frame 32:27)"

http://fora.tv/2009/05/30/Rodney_Brooks_Re...g_With_Robotics

But maybe you have to be a neophyte in technology order to believe in things like the singularity and not a person who designs and actually builds something on an everyday basis. Is anyone else here a robotics engineer or going to become one?

Link to comment
Share on other sites

Two years later, things are not so certain: http://hplusmagazine.com/articles/multimed...-asimo-learning

Not so fast... This is pattern recognition, with the impressive addition of pattern generation.

A new object is observed, the robot captures views of it to create a new pattern. It stores several views of the object, perhaps even generates and stores a 3-D model of the object. When it is shown the same object later, it compares views of that object to stored views using cross-correlation algorithms. When it registers a spike - meaning, when it finds a view in its memory that correlates closely with the current view - it "recognizes" that object. On a very basic level, this is done by normalizing the histogram of the objects being compared - scaling them and centering their pixel values at zero. Then each pixel of the seen object is multiplied by the pixel in the stored object and all pixel products are summed. If a pixel correlates in both objects, it will have either a positive value in both or a negative value. Multiplying them together gives a positive value if they're correlated, a negative value if they're not. If a large number of pixels are correlated, then the output of the algorithm produces a positive sum "spike." This can be done iteratively starting at a low resolution to choose the views the compare, but you get the idea. If the robot can produce distance vectors using stereoscopic vision algorithms, then the distances, rather than pixel values, can be correlated, but the idea is the same.

The chair trick is completely subjective and obviously rigged to produce the desired result. The robot sees a wooden chair and later associates it with a backless stool. Would you have made the same "conceptual" connection between the two, given that no one ever sat in either object? If he had brought out a small end table, would the robot have dismissed it, or connected it based on size and height of the horizontal plane? How about a short table lamp with a picture frame balanced on top?

I've seen enough demos to know that they are invariably rigged ... optimized, to the specific objects or scenarios given. Had the observer been invited to bring his own object, would the robot have succeeded? Had he held the toy upside down the second time would the robot have succeeded? If he had held it up in front of a mirror? etc.

What we are seeing here is not the recognition of an entity, an identification of that entity or an integration of two different entities based on common characteristics. We are seeing simple pattern correlation algorithms based on hard-coded rules, possibly in 3-D, definitely adjusted for size. It's a step, possibly, in the right direction. Very impressive, but not necessarily a quantum leap over target recognition capabilities that have been around for at least a couple of decades.

Link to comment
Share on other sites

Not so fast... This is pattern recognition, with the impressive addition of pattern generation.

Very impressive, but not necessarily a quantum leap over target recognition capabilities that have been around for at least a couple of decades.

I have to agree with almost everything you have just said. The only thing I might not agree with is the 10 years, by what standard i.e. how'd jou come up with dat?

What would have been more impressive in the video is some capability of "close" pattern recognition e.g. they show a slightly more dark red car after the original red car, that is shaped different and the robot is capable at grasping the "abstraction" of Car. Without this robots will undoubtedly have to have massive amounts of memory storing jpgs or 3d data and processing it is going to be a resource hog.

I often wonder why they don't do "slate"-"super" computing with these robots. (Slate computing or cloud computing with super computers(http://en.wikipedia.org/wiki/Tablet_PC#Slates topic heading Slates, cloud computing http://en.wikipedia.org/wiki/Cloud_computing) I think most of the time these robots have the computer on board. Having the computer on board is to me insane and so limiting to the computation algorithms. I also see little point to the human robot form any more. I think robots might be biped but arms legs and 5 fingers. Asimo's current design came out 10 years ago and they still have not up the ante where Sony has gone with their little QRIO.

I think the reason that American engineers haven't gone after a humanoid bot is that most of this tech is pointless without a power supply that is at least a 100 times more dense. spherical actuators eat up a lot of juice.

Link to comment
Share on other sites

... is likely just noise, not actually computationally significant. Noise can be replicated in a computer by a randomness generator (if true randomness is needed, we could use a pile of radioactive isotope). Computation must be emulated. Neuroscientists estimate the needed computational power would be in the range of no more than about 10^28cps. That would be reached by 2045 at the present rate of increase in computer power. It may be far sooner than that since that is the high range of estimates.

Noise is probably necessary for the functionality. I know of signal processing algorithms that require the introduction of noise back into a filtered circuit in order to work. Bart Kosko has a new nook out on Noise and its uses.

Also, piling up computational power isn't going to be enough to simulate a brain. There has to be knowledge of how it works, how it is structured, how it is interconnected. No electronic interconnect approaches the fanout of 1000 to 10,000 achieved by neurons in the brain, so there is another huge problem to be solved.

Link to comment
Share on other sites

  • 2 weeks later...
Of course I want to get into that math, but just as far as this topic, nothing too deep. If you were a construction worker, then I would want to stay away from the subject completely, and ask you to do the same...

GREAT, it doesn't say anything about your background on your profile so I don't know where you are at, please don't take any offense, I just don't know what level of math capability you are at but we'll try.

Fists off, sure, you did not say what you mean by unlimited, so I was wrong to assume you define a limited function as one that has all its values from a limited [a,b] interval, for all x's. (like sin(x) ) I shall now change my ways, and ask for clarifications instead.

I mean unlimited in the conventional definition of unlimited i.e. he has no limits to his understanding or no one can set limits to his rational faculty. He has free will which means he is free and therefore has no limits that can be placed logically on him. (except within time limits like the crow epistemology stuff: the context here is all long term anyway)

You know math, so you are aware that linear and exponential are math terms, they refer to specific types of functions. So there must be a type of function of time (Kurzweil is talking about function of time, not something else) that's "better", and somehow unlimited (I can't really wrap my mind around your kind of unlimited yet, do you mean something like f(x)=1/(1-x)?) , which has been shown to describe the human capacity to reason.

Please, most importantly, explain to me in math terms what you mean by better and unlimited. Then, how would that unlimited and better function you referred to, look like?

Lets say that for arguments sake that Kurzweil is right that he has a function that he has put through several data points and the curve is defined by those points as an exponential function i.e. let's take Kurzweil at his word. It is not linear but exponential. There is a subset of applied mathematics that searches for r^2 deviation value. This is common in statistics (Ki square value) and data analysis fields. If you have used a trend-line in excel you know of a similar function. You can even give excel a form of an equation and it will solve for the constants for an r^2 value. The R^2 value is the standard deviation of the points to the function and a value of 1 means that the point exactly match a r^2. Almost never do you get an R^2 value of 1 so lets say Kurzweil has one of .99, which is common for good data. The function is then set. Or is it? Does this now mean that there will not be a nuclear war, an economic crisis that wipes out civilization in a conventional war or anything that would stop Kurzweil's dream of a singularity for 500 years. There has been times in history when man has stepped backward in technological progress and this is why Kurweil's functions do add up.

He is claiming to predict the future very accurately. He might not have the exact curve but he has picked the family of curves and this limits him to an always improving function. The derivative of an exponential function is always positive. and this is even on his wiki: http://en.wikipedia.org/wiki/Raymond_Kurzweil

Daniel Lyons, writing in Newsweek, criticized Kurzweil for some of his predictions which turned out to be wrong; such as the economy continuing to boom from the 1998 dot-com through 2009, a US company having a market capitalization of more than $1 trillion, a supercomputer achieving 20 petaflops, speech recognition being in widespread use and cars that would drive themselves using sensors installed in highways; all by 2009[63].

Then, I will have understood why Kurzweil has less than the appropriate amount of respect for the human capacity to reason. But if you say that you cannot measure the human capacity to reason, then what possible meaning could the word unlimited have? If you know it's unlimited, you must've measured it somehow, and ran out of measuring tape.

Unlimited as in not limited to an output that is always positive or unlimited as in not limited to a function on a graph. You don't have to measure "unlimited" it is unlimited by definition. just like infinity can not be measured( except in terms of other infinities by deduction in set theory) See above. This is also the problem with determinism as things get more and more predictable the capability of free will drops. You have no free will because things are predictable. They are at cross purposes deductively.

All of this is not to mention that inductively none of his prediction capabilities make sense. Who has a tap on what the economy and technology is going to do? Man the individual certainly but it is not Mr. Kurzweil.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...