Super Artificial Intelligence, a naive approach

[ProgrammingGodJordan]

Apologies. A quick look shows that is indeed a standard neural network

Yes, I learnt how to write basic neural net from scratch in roughly 3 hours (perhaps typical), then I wrote the first item.

That first item above is nothing more than basic, as indicated in Readme.

However, the second item, the residual neural network (for heart irregularity detection) though standard, has an underlying that is non-trivial, while the top layers are from an easy to manipulate machine learning library, called mxnet.

 

[Squeegee Beckenheim]

I have not scrutinized other users here, despite the mountains of errors they have made.

So what? Do you believe that that means that your claims shouldn't be scrutinised? Leaving aside the fact that you specifically asked for your ideas to be examined in the OP, if you don't want your claims to be scrutinised, then you shouldn't post them on a sceptic's message board.

(1)
See exascale computing on wikipedia for 10^18 estimation. (as provided in original post)

Then why are you using a figure of 10¹⁵ in this thread, rather than 10¹⁸?

(2)
You may need to believe redundantly in science or other paradigms, but such is not necessary for others, like Neil deGrasse tyson, or myself.

Quote by Neil deGrasse Tyson:

Quote source video: https://www.youtube.com/watch?v=yRxx8pen6JY

Non-sequitur. This has nothing to do with you only skimming a paper and then misrepresenting the contents because you didn't understand what you were reading. It certainly doesn't explain why such behaviour is consistent with your ideas being well-founded or your reasoning being something that anybody should take even remotely seriously.

 

[ProgrammingGodJordan]

So what? Do you believe that that means that your claims shouldn't be scrutinised? Leaving aside the fact that you specifically asked for your ideas to be examined in the OP, if you don't want your claims to be scrutinised, then you shouldn't post them on a sceptic's message board.



Then why are you using a figure of 10¹⁵ in this thread, rather than 10¹⁸?



Non-sequitur. This has nothing to do with you only skimming a paper and then misrepresenting the contents because you didn't understand what you were reading. It certainly doesn't explain why such behaviour is consistent with your ideas being well-founded or your reasoning being something that anybody should take even remotely seriously.

(1)
There is hardly any data from that paper that I cannot understand.
After-all, I do have an understanding of the mathematics behind quantum computing's basis:

https://www.quora.com/How-does-quantum-computing-work/answer/Jordan-Bennett-9




(2)
I had long expressed that 10^15 synapses = 10^16 sops was used to compute a minimum year, using the minimum value 10^15.
One can easily compute a year, for the maximum value 10^18.



(3)
Do you know any quantum computing?

 

[The Don]

I already mentioned that I could not access page 15 using Porpoise of Life's link.

As seen in the screen shot provided (top right), that is page 15, not 12, from my copy.

Furthermore, advancing 3 pages on my copy, still does not yield any thing.

Page 15 is not accessible, to me.

Did you access it using Porpoise of Life's link ?

All I did was then click on the blue highlighted link on that page (reproduced below)

1 The Membrane Equation

...and scroll down

 

[Aridas]

Did you access it using Porpoise of Life's link ?

All I did was then click on the blue highlighted link on that page (reproduced below)

1 The Membrane Equation

...and scroll down

Worked for me. To quote the most relevant sentence, "At about 1 billion chemical synapses per cubic millimeter of cortical grey matter, there are lots of synapses in the nervous system (on the order of 10^15 for a human brain)."

 

[ProgrammingGodJordan]

Did you access it using Porpoise of Life's link ?

All I did was then click on the blue highlighted link on that page (reproduced below)

1 The Membrane Equation

...and scroll down

I did but strangely nothing showed up for page 15.

Anyway, data to keep in mind:

(1) The prior wiki data expressed 10^15. (a 2004 estimation)
(2) Other wiki data expresses 10^14, (2012 estimation)

What do you think is the next optimal move?


PS: My initial 2020 calculation is correct after-all, at 10^15 syapses = 10^16 sops, with respect the sources linked the.PS(2): Thanks for link "Porpoise of Life" this confirms the original post data, with respect the sources linked then.

 

[MikeG]

.......One can easily compute a year, for the maximum value 10^18.........

I did:

Moore's Law, hey. That one where we get a doubling in performance of silicon chips every two years? The right law, I take it?

OK, well, let's have a look. 2020 is 3 years away. At the moment you're own figures for computer performance has the best of them at 10^14 (somethings), which is one ten-thousandth of the level of humans, again, according to your own figures. Applying Moore's Law to those figures suggest computers will be at around approximately 3 ten-thousandths of that of humans by 2020. In fact, by applying Moore's Law alone, it looks as though it will take computers approx. 15 years to get to that same level, or approx. 2032.
As I probably said before, you just make **** up.

 

[ProgrammingGodJordan]

I did:

The minimum though, remains at 2020, if we use 10^15 synapses, koch christof, as I had long stated (See original post)

 

[MikeG]

....... My initial 2020 calculation is correct after-all, at 10^15 syapses = 10^16 sops, with respect the sources linked the.........

Like I said, you make **** up. This figure is not justified.

 

[Porpoise of Life]

I did but strangely nothing showed up for page 15.

Anyway, data to keep in mind:

(1) The prior wiki data expressed 10^15. (a 2004 estimation)
(2) Other wiki data expresses 10^14, (2012 estimation)

What do you think is the next move?

The next move is to find a citation that backs up your number or undo your edit.
I've checked the sources in the other wiki article and could not find anything referring to number of synapses. As far as I can tell all three articles only mention number of neurons.

Also, if 2004 is too old to be reliable, then could you explain how you can base your estimation of 10 signals per second per synapse on a 1985 article (and without citing the article itself, only another article that cites it as a source)?

 

[MikeG]

The minimum though, remains at 2020, if we use 10^15 synapses, koch christof, as I had long stated (See original post)

It's not a minimum, FFS. The word you want there is earliest.

 

[ProgrammingGodJordan]

The next move is to find a citation that backs up your number or undo your edit.
I've checked the sources in the other wiki article and could not find anything referring to number of synapses. As far as I can tell all three articles only mention number of neurons.

Also, if 2004 is too old to be reliable, then could you explain how you can base your estimation of 10 signals per second per synapse on a 1985 article (and without citing the article itself, only another article that cites it as a source)?

As long mentioned, I could not find anything later for impulses per second.

As for the wikipedia articles, that is not a thing for a blog post, and so later estimations are listed, as observed on Wiki/neuron, and Wiki/List_of_animals_by_number_of_neurons.

 

[The Don]

PS: My initial 2020 calculation is correct after-all, at 10^15 syapses = 10^16 sops, with respect the sources linked the.PS(2): Thanks for link "Porpoise of Life" this confirms the original post data, with respect the sources linked then.

Now you need to show how 10¹⁵synapses is equivalent to 10¹⁶ SOPS all you have is a 30 year old guess

Then you have the issue that the current state of play isn't 10¹⁴ SOPS (IBM have rolled that one back) - but even if you accept that it is, IBM claim to be "only" running 1542 times slower than real time so you have.....

A factor of 100 in capacity (10¹⁴ vs 10¹⁶)

A factor of 1542 in speed

giving 154200 in total

Even if Moore's law holds (and remember according to informed sources, due to physical limitations it no longer does), it'll take quite a while to overcome that shortfall....

 

[ProgrammingGodJordan]

Now you need to show how 10¹⁵synapses is equivalent to 10¹⁶ SOPS all you have is a 30 year old guess

Then you have the issue that the current state of play isn't 10¹⁴ SOPS (IBM have rolled that one back) - but even if you accept that it is, IBM claim to be "only" running 1542 times slower than real time so you have.....

A factor of 100 in capacity (10¹⁴ vs 10¹⁶)

A factor of 1542 in speed

giving 154200 in total

Even if Moore's law holds (and remember according to informed sources, due to physical limitations it no longer does), it'll take quite a while to overcome that shortfall....

Some-else here mentioned that IBM "rolled back", expressing 10¹⁰, instead of 10¹⁴. (I had been referring to 10^14 synapses, but the being appeared to be unaware of the distinction between neuron/synapse number)

I trivially showed that that IBM was referring to the number of simulated neurons (10¹⁰, which is a smaller number than the number of simulated synapses (10¹⁴), so no such roll back occurred, it had always been 10¹⁴, synapses.

Just to clarify, what did you mean by IBM rolled back?

 

[The Don]

Just to clarify, what did you mean by IBM rolled back?

The fact that it's running 1542 times slower than realtime

 

[MikeG]

As long mentioned, I could not find anything later for impulses per second..........

And as "long mentioned" (couldn't we speak English?) that source describes this figure as a guess.

 

[ProgrammingGodJordan]

And as "long mentioned" (couldn't we speak English?) that source describes this figure as a guess.

Anyway, that value (by neurosientists) did not influence my edit of wikipedia.

With regards to the original, post, based on the 2020 forecast, as far as I could find, "10 impulses per second" exists.

 

[Squeegee Beckenheim]

(1)
There is hardly any data from that paper that I cannot understand.

Whether or not you can potentially understand it is moot. You didn't understand it because you didn't pay attention while skim-reading it. How can anybody be expected to take you at all seriously if you formulate theories and come to conclusions based on on papers that you admit you haven't even read properly?

 

[Porpoise of Life]

Anyway, that value (by neurosientists) did not influence my edit of wikipedia.

With regards to the original, post, based on the 2020 forecast, as far as I could find, I found 10 impulses per second.

So you didn't find it.
You made it up because that would make your estimations correct. That's not evidence, that's circular reasoning.

 

[ProgrammingGodJordan]

Like I said, you make **** up. This figure is not justified..

I already linked to those values, as expressed by neuroscientists.