[extropy-chat] XYZBrain History [was: sjbrain calcs]

[Robert J. Bradbury]

On Mon, 22 Dec 2003, Alan Eliasen wrote:

> I'm earnestly following this discussions of Brains the Size of a
> Planet and all that, and going a bit insane in the membrane trying to
> keep the alphabetical procession of different XYZbrains in order.  Neat
> stuff, though.

I suppose a bit of a history lesson is in order for people who haven't
been on the list for decades.

Sometime in the early '90s Perry Metzger proposed the concept of a
Jupiter Brain (JBrain).  The exact date is lost in the private Extropian
archives.  See:
  http://www.aeiveos.com/~bradbury/JupiterBrains/
Anders Sandberg did some work in the mid-'90s documenting Dyson Spheres
(really Shells):
  http://www.student.nada.kth.se/~nv91-asa/dysonFAQ.html
he also started work writing some detailed theoretical calculations on
what the limits might be which was published in 1999 in the Journal
of Evolution and Technology:
  http://www.jetpress.org/volume5/Brains2.pdf
The appendix

>From 1997 to 2002 I gave talks at several conferences (including Extro3)
and published some papers about Dyson Shells and an extended concept
known as Matrioshka Brains (MBrain):
  http://www.aeiveos.com/~bradbury/MatrioshkaBrains/

During various visits to Spike's house he and I have played around with
various extensions to some of my ideas.

My efforts have at least planted these ideas in the heads of
a few astronomers and SETI researchers (though the mainstream
folks have generally rejected them).  In addition, Damien Broderick
has picked up some of the ideas in his book "The Spike" so they
are slowly filtering out into public awareness.

The quick summary regarding MBrains is that each "node" provides
the approximate computing capacity of 100,000 to a 1,000,000 human
brains and and entire MBrain contains a sufficient number of nodes
that you have the computational capacity of ~10^15 human *species*
(at our current population level) per star.

Spike's sjbrain is an interesting new variant.


> Can I ask what may be a silly question?  Why would one build so many
> little far-flung nodes instead of a more monolithic structure?  So many
> things are simpler, faster, and more efficient with a single
> structure--communications, energy expenditure, time lag, that I guess I
> don't see why the structure wouldn't be made much more compact.

The compact structure is in the MBrain (and to a lesser extent the JBrain).
The problem is that you hit the heat removal limits.  That's in the
MBrain architecture using 1 cm^3 Drexlerian nanocomputers that have
to radiate 100,000 W.  See for example:
  http://www.aeiveos.com/~bradbury/MatrioshkaBrains/MatrioshkaBrains.html#Figure1

The radiator size ultimately is much bigger than the computer size.

>    Communicating with someone with a one-minute lag would probably be
> pretty inefficient and painful.

We are going to have to get used to it.  A really advanced civilization
would probably consist of a number of Matrioshka Brains organized
the same way globular clusters are (a million or so stars within
a few light-years of each other).  But after they get through
star-lifting their lifetimes are in the trillion year range so
they probably don't sweat the inter-brain delay times.

> So, my question is, why intentionally make it so hard and painful to
> communicate?  I've always thought that future supercomputers would
> cluster as close as possible together so that communications lags would
> be minimized.  Or, why not?  Robustness?

They are clustered close together.  That's why the Cray-1 and Cray-2
had circular designs.  See:
  http://futuretech.mirror.vuurwerk.net/cray1.html

IBM was probably as good as Cray at figuring out how to cool things
(you may be too young to remember that some of the early 360's
and 370's were based on ECL logic and had to be water cooled!).
To avoid the problems that heat causes in accelerating failure
rates IBM is producing its high (up to 65K?) processor count
machines like Blue Gene with sharply constrained clock speeds
on the CPUs.  Even the PowerPC's in Univ. of Virgina supercomputer
(currently 2nd fastest in the world) run at clock speeds significantly
below current Intel CPUs.

>    Frink documentation:
> http://futureboy.homeip.net/frinkdocs/

Interesting Alan, I'll have to take a look at it.

Hope the above brings everyone who hasn't been around the last 6
years up to speed.

Robert