Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > alt.folklore.computers > #159025 > unrolled thread

How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network?

Back to article view | Back to alt.folklore.computers

Contents

  How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Neo Fortune <serenity3043@gmail.com> - 2016-02-08 02:28 -0800
    Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Ahem A Rivet's Shot <steveo@eircom.net> - 2016-02-08 11:33 +0000
      Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? The New Other Guy <Newsgnus@gmail.com> - 2016-02-08 04:43 -0800
        Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Morten Reistad <first@last.name.invalid> - 2016-02-08 18:54 +0100
          Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Anne & Lynn Wheeler <lynn@garlic.com> - 2016-02-08 13:01 -0800
        Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? "Charles Richmond" <numerist@aquaporin4.com> - 2016-02-08 14:23 -0600
          Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? jmfbahciv <See.above@aol.com> - 2016-02-09 14:21 +0000
            Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Ahem A Rivet's Shot <steveo@eircom.net> - 2016-02-09 14:54 +0000
              Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Neo Fortune <serenity3043@gmail.com> - 2016-02-10 01:21 -0800
            Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? "Hilo Black" <hbk897@gmail.com> - 2016-02-10 02:07 +1100
            Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Walter Bushell <proto@panix.com> - 2016-02-14 09:55 -0500
              Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Morten Reistad <first@last.name.invalid> - 2016-02-14 21:38 +0100
                Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? jmfbahciv <See.above@aol.com> - 2016-02-15 14:14 +0000
                  Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? "Rod Speed" <rod.speed.aaa@gmail.com> - 2016-02-16 05:14 +1100
              Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? jmfbahciv <See.above@aol.com> - 2016-02-15 14:14 +0000
      Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? "Charles Richmond" <numerist@aquaporin4.com> - 2016-02-08 14:22 -0600
        Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Ahem A Rivet's Shot <steveo@eircom.net> - 2016-02-08 21:39 +0000
          Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? "Osmium" <r124c4u102@comcast.net> - 2016-02-08 16:43 -0600
            Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? jmfbahciv <See.above@aol.com> - 2016-02-09 14:21 +0000
              Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? "J. Clarke" <j.clarke.873638@gmail.com> - 2016-02-09 19:12 -0500
                Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? jmfbahciv <See.above@aol.com> - 2016-02-10 12:46 +0000
                Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Roberto Waltman <usenet@rwaltman.com> - 2016-02-15 00:49 -0500
    Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Stephen Wolstenholme <steve@easynn.com> - 2016-02-08 12:54 +0000
      Re: How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network? Mike Spencer <mds@bogus.nodomain.nowhere> - 2016-02-08 15:31 -0400

Page 1 of 2  [1] 2  Next page →

#159025 — How Many Raspberry PI's Would It Take To Emulate The Human Brain's Neural Network?

https://www.raspberrypi.org/

[toc] | [next] | [standalone]

#159026

On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
Neo Fortune <serenity3043@gmail.com> wrote:

> https://www.raspberrypi.org/

	I recently read that the human brain holds around a petabyte of
data in some form. Given that the latest and greatest raspberry-pi has a
gigabyte of memory I would not expect it to be possible with less than a
million of them. Interconnections would be troublesome.

-- 
Steve O'Hara-Smith                          |   Directable Mirror Arrays
C:>WIN                                      | A better way to focus the sun
The computer obeys and wins.                |    licences available see
You lose and Bill collects.                 |    http://www.sohara.org/

[toc] | [prev] | [next] | [standalone]

#159027

On Mon, 8 Feb 2016 11:33:56 +0000, Ahem A Rivet's Shot <steveo@eircom.net>
wrote:

>On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
>Neo Fortune <serenity3043@gmail.com> wrote:
>
>> https://www.raspberrypi.org/
>
>	I recently read that the human brain holds around a petabyte of
>data in some form. Given that the latest and greatest raspberry-pi has a
>gigabyte of memory I would not expect it to be possible with less than a
>million of them. Interconnections would be troublesome.

Hell, you could do it with THREE, if you wanted to emulate a politician.

[toc] | [prev] | [next] | [standalone]

#159052

In article <fa3hbbhh2qld1tcf2v3c4f278s4b7kbmdd@4ax.com>,
The New Other Guy  <Newsgnus@gmail.com> wrote:
>On Mon, 8 Feb 2016 11:33:56 +0000, Ahem A Rivet's Shot <steveo@eircom.net>
>wrote:
>
>>On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
>>Neo Fortune <serenity3043@gmail.com> wrote:
>>
>>> https://www.raspberrypi.org/
>>
>>	I recently read that the human brain holds around a petabyte of
>>data in some form. Given that the latest and greatest raspberry-pi has a
>>gigabyte of memory I would not expect it to be possible with less than a
>>million of them. Interconnections would be troublesome.
>
>Hell, you could do it with THREE, if you wanted to emulate a politician.

To be serious; the connect speed is the one where we do have
a chance. The very max interconnect speed is around 10kbit; usually
significantly less. There are also surprisingly large delays; neuron
signals take ~30 ms from your fingers to your brain, ~70 from your
toes.

With 10 billion neurons, times 1000 interconnects (which seems large)
times 10 kilobits (also large) we get 10E16 bits/second, or around
10000 trillion; or 10 petabits. This is an upper bound, I would guess
the real figure is in the thousands lower; some tens of terabits. This
is around what we can push, today, through a standard optical (G.652)
fiber. 

Not that we can acheive this bitrate all in one place. Nor can the
brain. But with around 100 100Gbit swithces and WDM, we can.

The endeavour would cost us on the order of 100 kilowatts.

The individual neurons seem to be around as capable as an apple ][; 
or somewhat less. There are 10 billion of them. Most are mostly idle most
of the time. If we take todays technology and implement 6502-style
computers we should be able to make them use ~10mA each, down in the
uA range when idle. 

10 billion times 5V times 10 mA should give us a few hundred megawatts.

That is currently way out of scope. Even idling at 100 uA they would
draw several megawatts.

But the comms is barely doable in ~4 19" racks.

-- mrr

[toc] | [prev] | [next] | [standalone]

#159067

Emergent Chip Vastly Accelerates Deep Neural Networks
http://www.nextplatform.com/2015/12/08/emergent-chip-vastly-accelerates-deep-neural-networks/

-- 
virtualization experience starting Jan1968, online at home since Mar1970

[toc] | [prev] | [next] | [standalone]

#159064

"The New Other Guy" <Newsgnus@gmail.com> wrote in message 
news:fa3hbbhh2qld1tcf2v3c4f278s4b7kbmdd@4ax.com...
> On Mon, 8 Feb 2016 11:33:56 +0000, Ahem A Rivet's Shot <steveo@eircom.net>
> wrote:
>
>>On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
>>Neo Fortune <serenity3043@gmail.com> wrote:
>>
>>> https://www.raspberrypi.org/
>>
>> I recently read that the human brain holds around a petabyte of
>>data in some form. Given that the latest and greatest raspberry-pi has a
>>gigabyte of memory I would not expect it to be possible with less than a
>>million of them. Interconnections would be troublesome.
>
> Hell, you could do it with THREE, if you wanted to emulate a politician.
>

NOG, we want to emulate a person who actually *has* a brain... *not* a 
politician!!! :-)

-- 

numerist at aquaporin4 dot com

[toc] | [prev] | [next] | [standalone]

#159098

Charles Richmond wrote:
> "The New Other Guy" <Newsgnus@gmail.com> wrote in message
> news:fa3hbbhh2qld1tcf2v3c4f278s4b7kbmdd@4ax.com...
>> On Mon, 8 Feb 2016 11:33:56 +0000, Ahem A Rivet's Shot <steveo@eircom.net>
>> wrote:
>>
>>>On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
>>>Neo Fortune <serenity3043@gmail.com> wrote:
>>>
>>>> https://www.raspberrypi.org/
>>>
>>> I recently read that the human brain holds around a petabyte of
>>>data in some form. Given that the latest and greatest raspberry-pi has a
>>>gigabyte of memory I would not expect it to be possible with less than a
>>>million of them. Interconnections would be troublesome.
>>
>> Hell, you could do it with THREE, if you wanted to emulate a politician.
>>
>
> NOG, we want to emulate a person who actually *has* a brain... *not* a
> politician!!! :-)
>
I don't understand why anyone would want to emulate a human.

/BAH

[toc] | [prev] | [next] | [standalone]

#159101

On 9 Feb 2016 14:21:10 GMT
jmfbahciv <See.above@aol.com> wrote:

> Charles Richmond wrote:

> > NOG, we want to emulate a person who actually *has* a brain... *not* a
> > politician!!! :-)
> >
> I don't understand why anyone would want to emulate a human.

	To find out if it can be done (we'll never be sure it can't).

	For the fun of watching theologians arguing about whether or not
the emulation has a soul. 

	For the fun of watching lawyers arguing about whether or not it has
rights.

	As a stepping stone to making something more intelligent.

-- 
Steve O'Hara-Smith                          |   Directable Mirror Arrays
C:>WIN                                      | A better way to focus the sun
The computer obeys and wins.                |    licences available see
You lose and Bill collects.                 |    http://www.sohara.org/

[toc] | [prev] | [next] | [standalone]

#159129

On Tuesday, February 9, 2016 at 3:00:02 PM UTC, Ahem A Rivet's Shot wrote:
> On 9 Feb 2016 14:21:10 GMT
> jmfbahciv <See.above@aol.com> wrote:
> 
> > Charles Richmond wrote:
> 
> > > NOG, we want to emulate a person who actually *has* a brain... *not* a
> > > politician!!! :-)
> > >
> > I don't understand why anyone would want to emulate a human.
> 
> 	To find out if it can be done (we'll never be sure it can't).
> 
> 	For the fun of watching theologians arguing about whether or not
> the emulation has a soul. 
> 
> 	For the fun of watching lawyers arguing about whether or not it has
> rights.
> 
> 	As a stepping stone to making something more intelligent.
> 

Hal 9000 VS Dave - Ontological scene in 2001: A Space Odyssey
https://www.youtube.com/watch?v=HwBmPiOmEGQ

[toc] | [prev] | [next] | [standalone]

#159105

"jmfbahciv" <See.above@aol.com> wrote in message 
news:PM00052B56F0BC16B9@aca405cc.ipt.aol.com...
> Charles Richmond wrote:
>> "The New Other Guy" <Newsgnus@gmail.com> wrote in message
>> news:fa3hbbhh2qld1tcf2v3c4f278s4b7kbmdd@4ax.com...
>>> On Mon, 8 Feb 2016 11:33:56 +0000, Ahem A Rivet's Shot 
>>> <steveo@eircom.net>
>>> wrote:
>>>
>>>>On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
>>>>Neo Fortune <serenity3043@gmail.com> wrote:
>>>>
>>>>> https://www.raspberrypi.org/
>>>>
>>>> I recently read that the human brain holds around a petabyte of
>>>>data in some form. Given that the latest and greatest raspberry-pi has a
>>>>gigabyte of memory I would not expect it to be possible with less than a
>>>>million of them. Interconnections would be troublesome.
>>>
>>> Hell, you could do it with THREE, if you wanted to emulate a politician.
>>>
>>
>> NOG, we want to emulate a person who actually *has* a brain... *not* a
>> politician!!! :-)
>>
> I don't understand why anyone would want to emulate a human.

Because they can still do some stuff that computers can't do. 

[toc] | [prev] | [next] | [standalone]

#159343

In article <PM00052B56F0BC16B9@aca405cc.ipt.aol.com>,
 jmfbahciv <See.above@aol.com> wrote:

> I don't understand why anyone would want to emulate a human.
> 
> /BAH

For some (alleged) people, it would be an improvement if we mean a
functional person. For decades we are able to emulate catatonics and
non directive shrinks.

[toc] | [prev] | [next] | [standalone]

#159386

In article <proto-683C05.09555614022016@news.panix.com>,
Walter Bushell  <proto@panix.com> wrote:
>In article <PM00052B56F0BC16B9@aca405cc.ipt.aol.com>,
> jmfbahciv <See.above@aol.com> wrote:
>
>> I don't understand why anyone would want to emulate a human.
>> 
>> /BAH
>
>For some (alleged) people, it would be an improvement if we mean a
>functional person. For decades we are able to emulate catatonics and
>non directive shrinks.

Perhaps some idealised humaoid. 

<red scare> So we can start implementing communism.</red scare>

-- mrr

[toc] | [prev] | [next] | [standalone]

#159432

Morten Reistad wrote:
> In article <proto-683C05.09555614022016@news.panix.com>,
> Walter Bushell  <proto@panix.com> wrote:
>>In article <PM00052B56F0BC16B9@aca405cc.ipt.aol.com>,
>> jmfbahciv <See.above@aol.com> wrote:
>>
>>> I don't understand why anyone would want to emulate a human.
>>>
>>> /BAH
>>
>>For some (alleged) people, it would be an improvement if we mean a
>>functional person. For decades we are able to emulate catatonics and
>>non directive shrinks.
>
> Perhaps some idealised humaoid.

Thinking well without emotions and chemistry solves problems.
Emulating human thinking will create problems.  Very few people
are able to set biases aside and look at situations objectively.
That's what AI could do well.

>
> <red scare> So we can start implementing communism.</red scare>

Every human equally poor.

/BAH

[toc] | [prev] | [next] | [standalone]

#159438

"jmfbahciv" <See.above@aol.com> wrote in message 
news:PM00052BCFA433F634@aca40706.ipt.aol.com...
> Morten Reistad wrote:
>> In article <proto-683C05.09555614022016@news.panix.com>,
>> Walter Bushell  <proto@panix.com> wrote:
>>>In article <PM00052B56F0BC16B9@aca405cc.ipt.aol.com>,
>>> jmfbahciv <See.above@aol.com> wrote:
>>>
>>>> I don't understand why anyone would want to emulate a human.
>>>>
>>>> /BAH
>>>
>>>For some (alleged) people, it would be an improvement if we mean a
>>>functional person. For decades we are able to emulate catatonics and
>>>non directive shrinks.
>>
>> Perhaps some idealised humaoid.
>
> Thinking well without emotions and chemistry solves problems.
> Emulating human thinking will create problems.  Very few people
> are able to set biases aside and look at situations objectively.
> That's what AI could do well.
>
>>
>> <red scare> So we can start implementing communism.</red scare>
>
> Every human equally poor.

Nope, the apparatchiks aren't. 

[toc] | [prev] | [next] | [standalone]

#159431

Walter Bushell wrote:
> In article <PM00052B56F0BC16B9@aca405cc.ipt.aol.com>,
>  jmfbahciv <See.above@aol.com> wrote:
>
>> I don't understand why anyone would want to emulate a human.
>>
>> /BAH
>
> For some (alleged) people, it would be an improvement if we mean a
> functional person. For decades we are able to emulate catatonics and
> non directive shrinks.

I would rather develop a thinking entity which can think clearly
without all the problems humans have.

/BAH

[toc] | [prev] | [next] | [standalone]

#159063

"Ahem A Rivet's Shot" <steveo@eircom.net> wrote in message 
news:20160208113356.2bcf169d61f23ea5842e2696@eircom.net...
> On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
> Neo Fortune <serenity3043@gmail.com> wrote:
>
>> https://www.raspberrypi.org/
>
> I recently read that the human brain holds around a petabyte of
> data in some form. Given that the latest and greatest raspberry-pi has a
> gigabyte of memory I would not expect it to be possible with less than a
> million of them. Interconnections would be troublesome.
>

ISTM that there's a video on YouTube by Steven Furber (an ARM guy) about 
attempting to create an artificial human brain using 1,000,000 Raspberry 
Pi's.

-- 

numerist at aquaporin4 dot com

[toc] | [prev] | [next] | [standalone]

#159071

On Mon, 8 Feb 2016 14:22:05 -0600
"Charles Richmond" <numerist@aquaporin4.com> wrote:

> "Ahem A Rivet's Shot" <steveo@eircom.net> wrote in message 
> news:20160208113356.2bcf169d61f23ea5842e2696@eircom.net...
> > On Mon, 8 Feb 2016 02:28:20 -0800 (PST)
> > Neo Fortune <serenity3043@gmail.com> wrote:
> >
> >> https://www.raspberrypi.org/
> >
> > I recently read that the human brain holds around a petabyte of
> > data in some form. Given that the latest and greatest raspberry-pi has a
> > gigabyte of memory I would not expect it to be possible with less than a
> > million of them. Interconnections would be troublesome.
> >
> 
> ISTM that there's a video on YouTube by Steven Furber (an ARM guy) about 

	Heh I remember him as an Acorn guy making the tube interprocessor
interface for the BBC micro (we couldn't wait at Torch so I used back to
back parallel ports).

> attempting to create an artificial human brain using 1,000,000 Raspberry 
> Pi's.

	It is about the right scale - 10,000 neurons per pi gives you 100K
bytes per neuron (less a bit for code) which should be plenty for state and
Morten's bandwidth calculations would seem to indicate that a million
100Mbit connections could handle the traffic easily (just don't ask me to
design the switch array). It might run a bit slow but that's probably not a
bad thing. Now as for the software ... well it's just a small matter of
programming.

-- 
Steve O'Hara-Smith                          |   Directable Mirror Arrays
C:>WIN                                      | A better way to focus the sun
The computer obeys and wins.                |    licences available see
You lose and Bill collects.                 |    http://www.sohara.org/

[toc] | [prev] | [next] | [standalone]

#159075

"Ahem A Rivet's Shot" wrote:

> It is about the right scale - 10,000 neurons per pi gives you 100K
> bytes per neuron (less a bit for code) which should be plenty for state 
> and
> Morten's bandwidth calculations would seem to indicate that a million
> 100Mbit connections could handle the traffic easily (just don't ask me to
> design the switch array). It might run a bit slow but that's probably not 
> a
> bad thing. Now as for the software ... well it's just a small matter of
> programming.

I think a great amount of progress is made based of faulty inputs. One hears 
or reads a word wrong, sees something and the brain renders it wrong, sounds 
the wrong note on a musical instrument. I guess that doesn't make the 
programming harder, I just think it's interesting.

Also, If I try to solve a problem, my sub-conscious overwhelms my conscious. 
Fix a problem firmly in my mind and a few days later an answer pops up out 
of nowhere.  If I'm lucky, that is.

I  suspect there are people who can just bull ahead and force a solution, 
too. I am not thinking of "school type" problems here. 

[toc] | [prev] | [next] | [standalone]

#159096

Osmium wrote:
> "Ahem A Rivet's Shot" wrote:
>
>> It is about the right scale - 10,000 neurons per pi gives you 100K
>> bytes per neuron (less a bit for code) which should be plenty for state
>> and
>> Morten's bandwidth calculations would seem to indicate that a million
>> 100Mbit connections could handle the traffic easily (just don't ask me to
>> design the switch array). It might run a bit slow but that's probably not
>> a
>> bad thing. Now as for the software ... well it's just a small matter of
>> programming.
>
> I think a great amount of progress is made based of faulty inputs. One hears
> or reads a word wrong, sees something and the brain renders it wrong, sounds
> the wrong note on a musical instrument. I guess that doesn't make the
> programming harder, I just think it's interesting.
>
> Also, If I try to solve a problem, my sub-conscious overwhelms my conscious.
> Fix a problem firmly in my mind and a few days later an answer pops up out
> of nowhere.  If I'm lucky, that is.

I would always dream my solutions.

>
> I  suspect there are people who can just bull ahead and force a solution,
> too. I am not thinking of "school type" problems here.
>

/BAH

[toc] | [prev] | [next] | [standalone]

#159122

In article <PM00052B5701ECA104@aca405cc.ipt.aol.com>, See.above@aol.com 
says...
> 
> Osmium wrote:
> > "Ahem A Rivet's Shot" wrote:
> >
> >> It is about the right scale - 10,000 neurons per pi gives you 100K
> >> bytes per neuron (less a bit for code) which should be plenty for state
> >> and
> >> Morten's bandwidth calculations would seem to indicate that a million
> >> 100Mbit connections could handle the traffic easily (just don't ask me to
> >> design the switch array). It might run a bit slow but that's probably not
> >> a
> >> bad thing. Now as for the software ... well it's just a small matter of
> >> programming.
> >
> > I think a great amount of progress is made based of faulty inputs. One hears
> > or reads a word wrong, sees something and the brain renders it wrong, sounds
> > the wrong note on a musical instrument. I guess that doesn't make the
> > programming harder, I just think it's interesting.
> >
> > Also, If I try to solve a problem, my sub-conscious overwhelms my conscious.
> > Fix a problem firmly in my mind and a few days later an answer pops up out
> > of nowhere.  If I'm lucky, that is.
> 
> I would always dream my solutions.

FWIW, my current employer made an 8-week meditation course available 
last year to any interested employees (through UMASS Medical Center--not 
some pop-culture guru), during work hours and at no charge.  I took it 
and have been using the techniques.  Sometimes I'll get stumped on a 
problem, decide to give it up for the day and go to something else, do a 
10 minute meditation to change mental gears, and the solution pops up.  
Kind of surprised me the first time it happened.

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | alt.folklore.computers

csiph-web