[ExI] SETI for Post Singularity Civ

Anders Sandberg anders at aleph.se
Sun Jan 18 10:50:26 UTC 2015


Keith Henson <hkeithhenson at gmail.com> , 18/1/2015 7:35 AM:

Re the whole topic related to M brains, Jupiter brains and related,
nobody has ever answered my interrelated concerns of waste heat
dissipation and speed of light delays.  The laws of the universe, as I
understand them, don’t allow for physically large objects to be fast
thinkers.


Fast thinkers in what sense? A large object can have very fast local computations but long lags between parts. It doesn't have to be a computational unity.


Consider a spherical M brain of radius R and heat production P per volume. If it just cools by blackbody radiation the max radius for a given max temperature T is R = 3 sigma T^4 / P. If it does C computations per volume per second, the total computations are 4 pi R^3 C / 3. The delays will be of order R/c. If we demand that a computational module is roughly the size it takes to transmit a signal across it per computational cycle, and the relevant smallest components inside have linear size r and hence cycle time 1/(Cr^3), then modules need to be smaller than c/(Cr^3). So you get  (Cr^3R/c)^3 modules inside the M brain. 


If the figure of merit is just total computations then it scales as T^12 P^-3 C - it really pays off running it hot! If the figure of merit is the computations per unitary module then it scales as C^-2 r^-9 - running really tiny components is the key thing. If on the other hand you value modules, then they scale as C^3 r^9 T^12 P^-3 - back to the large hot brain. The overall ratio between M brain extent and cycle time is CRr^3; keeping it around unity makes the system about as integrated as a human brain.






Anders Sandberg, Future of Humanity Institute Philosophy Faculty of Oxford University
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20150118/3a7fccbd/attachment.html>


More information about the extropy-chat mailing list