Super Artificial Intelligence, a naive approach

[Aridas]

They had already achieved 10^14...

Of something else. You're trying to say that an order of magnitude between notably different things should be treated as if it's just an order of magnitude, by that logic.

By that kind of logic, 20000 ameobas can be treated as if they weigh more than 2 elephants. There's a 10^4 difference in number, after all! Nevermind anything else that might be of relevance to the calculations.

That's not to say that number is everything, again, though. How something is used tends to be much more important, and it's distinctly possible that even 10^14 could potentially outdo the 10^17 which would be just as much in the range of your assumptions.

My 3 year old nephew understands that the range was

(1) not my conjuring.

You've yet to show it. That your nephew understands it doesn't mean that it's actually true or that others should just accept it blindly.

(2) the precise range is pointed out in sources linked.

Where? So far, you've substantiated the 10^15 number that you popped out later with something quite irrelevant to the point you were trying to make and have repeatedly tried to misuse it. You've given no indication where you got the 10^16 number at all. The 10^18 number is the only one that you've backed up at all.

So no, your statement here is entirely unacceptable.

 

[MikeG]

See the sources provided.

I have done. There is no mention of this figure as far as I can see. Could you explicitly quote chapter and verse, with a link, so that we can verify this number.

 

[MikeG]

Please don't consider a career in any profession where missing by a factor of 10 is considered important. Heck, I'm scratching my head trying to think of any career in which 1 of something isn't "notably different" to 10 of something.

I think even suggesting this is an indication that what you're doing isn't working.

I'm thinking seismologist. The Richter scale is logarithmic, with a force 6 earthquake being one tenth as powerful as a force 7. Our buddy could announce to the world that the quake which just flattened, say, San Francisco, was "roughly 7 or 8" on the Richter scale. No-one would notice........

 

[Aridas]

Alright, though. I'll accept that you're intent on staying your course of failing to back up your points.

With that said, then... I'll actually respond to your OP.

(i)
Life's meaning probably occurs on the horizon of optimization:

(source: mit physicist, Jeremy England proposes new meaning of life)

Looking at the link, it doesn't support the claim that you look like you're making here. The link just addresses a proposal to redefine life, yet again, by the look of it. There's already a bunch of definitions for life floating around, though, and being used for various purposes in science and otherwise. There's not much to say of note here, then, beyond that "Life's meaning" is pretty much exclusively used in a completely different way. Namely, one directly in line with the age old question of "What is the meaning/purpose of life?"

(ii)
Today, artificial intelligence exceeds mankind in many human, cognitive tasks:

(source: can we build ai without losing control over it?)

Yup. The moment that self-optimizing and learning functions were figured out, it was pretty much only a matter of time and resources before such became possible. This is hardly new news, though.

(iii)
The creation of general artificial intelligence is so far, mankind's largely pertinent task, and this involves (i), i.e. optimization.

This is an empty assertion based solely on opinion and little more. As I'm sure other posters noted, though, pertinent is being misused here.

The human brain computes roughly 10^16 to 10^18 synaptic operations per second.

Links are fixed in this version, by the way. Still, this is pretty much a "So what?" bit of information.

(iv)
Mankind has already created brain based models that achieve 10^14 of the above total in (iii).

And? Raw numbers, even if impressive, don't answer the more important questions about the ability of an AI.

If mankind isn't erased (via some catastrophe), on the horizon of Moore's Law, mankind will probably create machines, with human-level brain power (and relevantly, human-like efficiency), by at least 2020.

Economic forces have been projected to interfere with Moore's Law, regardless. When it comes to human level brain power, though, there's still the very real question of what that actually means. By at least a few measures, computers overtook that long ago, for that matter, on a quick look.

(v)
Using clues from from quantum mechanics, and modern machine learning, I have composed (am composing) a naive fabric in aims of absorbing some non-trivial intelligence's basis.

This does not lend itself to meaning something in English. The words, individually, do have meanings, but combined, it does not lead to any meaningful interpretation.

(vi)
Criticism is welcome/needed.

My first direct question, then, is "What is the point of this thread?" It is not obvious at all. Going from the very general gist of (v), was this just an unnecessarily complicated attempt to show off a bit of the work that you're proud of and hopefully get some helpful pointers for how you could improve it? If so, you communicated such extremely poorly and threw in lots of distractions, and have compounded on your errors since then, by the look of it.

 

[ProgrammingGodJordan]

Of something else. You're trying to say that an order of magnitude between notably different things should be treated as if it's just an order of magnitude, by that logic.

By that kind of logic, 20000 ameobas can be treated as if they weigh more than 2 elephants. There's a 10^4 difference in number, after all! Nevermind anything else that might be of relevance to the calculations.

That's not to say that number is everything, again, though. How something is used tends to be much more important, and it's distinctly possible that even 10^14 could potentially outdo the 10^17 which would be just as much in the range of your assumptions.



You've yet to show it. That your nephew understands it doesn't mean that it's actually true or that others should just accept it blindly.



Where? So far, you've substantiated the 10^15 number that you popped out later with something quite irrelevant to the point you were trying to make and have repeatedly tried to misuse it. You've given no indication where you got the 10^16 number at all. The 10^18 number is the only one that you've backed up at all.

So no, your statement here is entirely unacceptable.

No.

(A)

Roughly 10^15 synapses is given in the original source. (WikiPedia link 1, Koch Christof, biophysics of computation 2004)

The 10^15 corresponds to the 2020 year computation, and it also corresponds with 10^16 sops.

So, roughly 10^15 synapses = 10^16 sops, all in the original post.


(B)

If you look back from the original post to now, my expressions have aligned with the above.

I was repeating that it (10^15) was included in source, and beings (including yourself) did not want to accept that roughly 10^15 synapses = 10^16 synaptic operations per second.



(C)

So, you guys did not want to accept that roughly 10^15 synapses = 10^16 sops, (in original post), regardless of the instance that I had indicated this in reply #115:

(A)
The artificial synapses correspond with the speed.

(D)

As for your response in reply #224 above, the criticism is redundant and or garbage.
Read the material carefully, please.



FOOTNOTE:
Try to not refer to other posters' opinions to justify your responses.
As is observed, the other posters were wrong.

 

[ProgrammingGodJordan]

So you derived 10¹⁶ by starting from 10¹⁵ and multiplying by 10 operations per second, and you got 10 operations per second because you started with 10¹⁶ and needed to derive 10¹⁵? This doesn't strike you as at all circular?

No.

Please read carefully.

The expressions differ, in terms of rate/size


Thus, roughly 10^15 synapses = 10^16 synaptic operations per second.

....where the LHS is the size, while the RHS represents the number of operations the LHS size can perform.
(So, lhs corresponds with rhs, i.e. speed)

 

[ProgrammingGodJordan]

Please don't consider a career in any profession where missing by a factor of 10 is considered important. Heck, I'm scratching my head trying to think of any career in which 1 of something isn't "notably different" to 10 of something.

I think even suggesting this is an indication that what you're doing isn't working.

This comment of mine applies:

No.

Please read carefully.

The expressions differ, in terms of rate/size


Thus, roughly 10^15 synapses = 10^16 synaptic operations per second.

....where the LHS is the size, while the RHS represents the number of operations the LHS size can perform.
(So, lhs corresponds with rhs, i.e. speed)

 

[ProgrammingGodJordan]

I had a look but they 404. The other repositories I looked at are equally trivial apart from a potentially interesting javascript learning library you forked from elsewhere. Yeah, some of us know git, java, etc. Nothing to see here folks, move along.

The coding of neural nets and new programming language(s) (like the language and neural nets I programmed, from reply 189) may be hard to some.

I observe where highly intellectual beings tended to be intrigued by neural nets, from basic ones such as (item iv from reply 189) to slightly more complicated ones, such as (item ii from reply 189)

I am curious. One of my repositories (the one from the original post) naively compounded supermanifolds & reinforcement learning, using clues from quantum computing basis. Since you find manifolds and deep learning 'trivial', where manifolds are observed to be potential ways to solve SEVERE ISSUES within deep learning, do you have any tips for me?

 

[MikeG]

Constantly restating that you multiplied one figure by ten to get to another figure doesn't even begin to explain why you used the figure 10. Unless you address this, the entire thread will grind to a halt, as no-one is going to let you get away with plucking an oh-so-convenient figure like this right out of thin air. Where does your figure of 10 "synaptic operations per second" come from? No-one will take your word for it, so please support anything you say with evidence, including direct quotes and links. Your sloppy thinking is bogging this thread down in the mire.

 

[fagin]

I am curious.

http://www.imdb.com/title/tt0061834/

 

[Aridas]

No.

(A)

Roughly 10^15 synapses is given in the original source. (WikiPedia link 1, Koch Christof, biophysics of computation 2004)

The 10^15 corresponds to the 2020 year computation, and it also corresponds with 10^16 sops.

So, roughly 10^15 synapses = 10^16 sops, all in the original post.


(B)

If you look back from the original post to now, my expressions have aligned with the above.

I was repeating that it (10^15) was included in source, and beings (including yourself) did not want to accept that roughly 10^15 synapses = 10^16 synaptic operations per second.

Yet, there's no reason to accept the conversion when the only thing that you've presented to back it up is, so far, empty assertion. Assertion that would seem to be very much in doubt given that the 10^18 number is the only one that you gave any evidence for in the first place, which points to a conversion rate more like 1000x rather than 10x. More like 10000x, if we're going by the numbers in the one Modha paper, for that matter.

(C)

So, you guys did not want to accept that roughly 10^15 synapses = 10^16 sops, (in original post), regardless of the instance that I had indicated this in reply #115:

Which you've based on what?

(D)

As for your response in reply #224 above, the criticism is redundant and or garbage.
Read the material carefully, please.

So, no answer for the only truly important question there? Again, what was the intended purpose of this thread? There wasn't any clear purpose in the first place, which has made it fairly inevitable that it get bogged down on the distractions you provided.

FOOTNOTE:
Try to not refer to other posters' opinions to justify your responses.
As is observed, the other posters were wrong.

You simply asserting that someone is wrong isn't especially convincing, quite frankly.

 

[abaddon]

Please don't consider a career in any profession where missing by a factor of 10 is considered important. Heck, I'm scratching my head trying to think of any career in which 1 of something isn't "notably different" to 10 of something.

I think even suggesting this is an indication that what you're doing isn't working.

This comment of mine applies:

No.

Please read carefully.

The expressions differ, in terms of rate/size


Thus, roughly 10^15 synapses = 10^16 synaptic operations per second.

....where the LHS is the size, while the RHS represents the number of operations the LHS size can perform.
(So, lhs corresponds with rhs, i.e. speed)

Nope. All that tells us is that you trod in a basic math error and tried to insert random numbers just to escape from that error.

It isn't working.

 

[ProgrammingGodJordan]

Nope. All that tells us is that you trod in a basic math error and tried to insert random numbers just to escape from that error.

It isn't working.

Here is a publicly available data, showing 10 impulses per second:

http://www.merkle.com/brainLimits.html

 

[ProgrammingGodJordan]

That paper also says that human brains have 10¹⁴ synapses. Any particular reason you're ignoring that part of it?

Your comment revealed AN ERROR OF MINE. This event has lead to the correction of a figure on wikipedia, by myself (where this would be my first wikipedia public data edit)..



///////(A) MY ERROR

(1)
I had initially encountered this source, where I observed the 10^15 estimation synapses value for human brain.

(2)
Source is (1) seemed potentially outdated, so I did brief WikiPedia, where I found orders of magnitude link, that contained 10^15 synapses. (which I took to be probably updated)


(3)
I briefly read IBM source, and did not detect that koch source was referenced as 10^14. In my quick reading, I related the 10^14 to ibm's chip.




///////(B) SMALL WIKIPEDIA UPDATE BY ME

(1)
Based on (A), I had recently came to recognize that 10^15 was potentially a bad listing on wikipedia's orders of magnitude link. (the link from original post)


(2)
This particular WikiPedia link presents HEADINGS in short/long scale.
This essentially means that after HEADING 10¹², for example, we won't see 10¹³ as a HEADING, but we may see 10¹³ under the heading 10¹².


(3)
From (3), I concluded that 10^15 was invalid on the wikipedia's orders of magnitude link, both by approximation, and naturally, by position on page.

So, now, on wikipedia before and after my edit looks like:

You can view the edit history of the orders of magnitude edit HERE, or go to orders of magnitude link, and click view history, to see my edit record.




///////(C) IMPLICATIONS

Essentially, I had done my computations based on this source, but now I see that it is outdated, based on the above.

This means that IBM has already achieved roughly similar number of artificial human level synapse, but those artificial synapses are still rough, although they are of unprecedented efficiency.

 

[The Don]

Here is a publicly available data, showing 10 impulses per second:

http://www.merkle.com/brainLimits.html

From your link:

We might count the number of synapses, guess their speed of operation, and determine synapse operations per second

The 10 operations per second was just a guess for illustrative purposes to demonstrate how the calculation could be made not an accurate measure based on any kind of research

 

[The Don]

[qimg]http://i.imgur.com/SXRniwT.jpg[/qimg]

Your comment revealed AN ERROR OF MINE. This event has lead to the correction of a figure on wikipedia, by myself (where this would be my first wikipedia public data edit)..



///////(A) MY ERROR

(1)
I had initially encountered this source, where I observed the 10^15 estimation synapses value for human brain.

(2)
Source is (1) seemed potentially outdated, so I did brief WikiPedia, where I found orders of magnitude link, that contained 10^15 synapses. (which I took to be probably updated)


(3)
I briefly read IBM source, and did not detect that koch source was referenced as 10^14. In my quick reading, I related the 10^14 to ibm's chip.




///////(B) SMALL WIKIPEDIA UPDATE BY ME

(1)
Based on (A), I had recently came to recognize that 10^15 was potentially a bad listing on wikipedia's orders of magnitude link. (the link from original post)


(2)
This particular WikiPedia link presents HEADINGS in short/long scale.
This essentially means that after HEADING 10¹², for example, we won't see 10¹³ as a HEADING, but we may see 10¹³ under the heading 10¹².


(3)
From (3), I concuded that 10^15 was invalid on the wikipedia's orders of magnitude link.

So, now, on wikipedia before and after my edit looks like:


You can view the edit history of the orders of magnitude edit HERE, or go to orders of magnitude link, and click view history, to see my edit record.




///////(C) IMPLICATIONS

Essentially, I had done my computations based on this source, but now I see that is outdated, based on the above.

This means that IBM has already achieved roughly similar number of artificial human level synapse, but those artificial synapses are still rough, although they are of unprecedented efficiency.

A lot of your recent links come back to Ralph C Merkle's site. I don't know about his background or his qualifications but I would note that he is very keen on the singularity and may not be the most impartial source of information.

 

[ProgrammingGodJordan]

A lot of your recent links come back to Ralph C Merkle's site. I don't know about his background or his qualifications but I would note that he is very keen on the singularity and may not be the most impartial source of information.

(a)
Read carefully.

10 impulses per second comes from:

Principles of Neural Science, by Eric R. Kandel and James H. Schwartz, 2nd edition, Elsevier, 1985.


(b)
Please edit out or update my prior quote, because:
(1) It lengthens the post more than it needs to. (merely 2 of 7 links relates to Ralph C Merkle's site, and those two links are the same link.)
(2) I removed the red headings.
(3) You were invalid to say that relied on the site only, as seen in (a).

 

[The Don]

(a)
Read carefully.

10 impulses per second comes from:

Principles of Neural Science, by Eric R. Kandel and James H. Schwartz, 2nd edition, Elsevier, 1985.

Don't have access to that text but the use of the words "guess" and "roughly" seem to indicate that it is neither a precise nor accurate figure.

 

[The Don]

(b)
Please edit out or update my prior quote, because:
(1) It lengthens the post more than it needs to. (merely 2 of 7 links relates to Ralph C Merkle's site, and those two links are the same link.)
(2) I removed the red headings.
(3) You were invalid to say that relied on the site only, as seen in (a).

I never said the highlighted - stop creating strawmen

 

[ProgrammingGodJordan]

Don't have access to that text but the use of the words "guess" and "roughly" seem to indicate that it is neither a precise nor accurate figure.

I never said the highlighted - stop creating strawmen

You based your reference merely on the instance that I took from Ralph, which was invalid (reply 236).

You didn't mention the source that Ralph actually referred to, so you only mentioned that I used ralph. No 'strawman' neccessary.

Anyway, you don't need to quote the entire sequence, as (merely 2 of 7 links relates to Ralph C Merkle's site, and those two links are the same link.)

So please edit your previous post, by removing my long quote, and keeping the link (although your criticism was invalid), as you were wrong based on your initial expression; both about supposedly "most of the links" reaching back to the Ralph link, and the instance explained atop this response.



FOOTNOTE(1):
I edited the last link in the quote you quoted (in 236), to fix such a link.


FOOTNOTE(2):
BOTH the authors (Eric R. Kandel and the late James H. Schwart) from Ralph's website reference in question are NEUROSCIENTISTS, so unlike Ralph, now you know the background of the 10 impulses per second reference.


FOOTNOTE(3):
Observing footnote (2), you may probably see why your quote is invalid:

The 10 operations per second was just a guess for illustrative purposes to demonstrate how the calculation could be made not an accurate measure based on any kind of research.