[extropy-chat] D-Wave premiere of 16 qubit processor

[Eugen Leitl]

On Wed, Feb 14, 2007 at 06:33:24AM -0500, Robert Bradbury wrote:
> 
>    Can anyone show me how quantum computers *or* AGI apply to the
>    following problems:
>    1) Protein folding (presumably done through classical molecular
>    dynamics simulations) [1]

If QC parallelism is suitable for concurrent sampling of conformation
space, or to QM treatment of subsystems or entire system it is
very applicable, assuming QC is ever practical.

>    2) Nano-part design

Parallel sampling of design space, of conformation space, full
QM treatment of parts or entire assemblies, all also with above
caveat.

>    3) Nano-system, esp. nanorobot, design

Basically, the same thing as 2).

>    These in my book are the *hard* problems.  Whether we can crack
>    encryption keys means almost *nothing* IMO relative to a singularity

QC, if indeed practical, is a good match for some problems (RSA,
via number factoring) but a bad one for others (IDEA, for instance).
I doubt QC is useful for brute-forcing most block cyphers, for instance.

>    I personally was much more excited by the Intel 80 core 1.# TF
>    processor [3] because the implications are that with 1000 processors

I am also quite excited by it (the Cell done right, at last), but
it's vaporware. They voiced intent to push this into their mainstream
product, but making cores do x86 will make them much fatter. 

>    (@ 62,000W) [4,5] I have human brain level computational capacity

The human brain computational capacity is chronically difficult to
estimate, and it also involves total amount of bits held, not just
bit manipulation rate (a waferful of ring oscillators or adders
in 65 nm could be really hard to beat, then).

>    1. One can see the growth in known structures in the PDB database
>    [2].  Though we have not reached that point yet, I suspect we are
>    within a few years of the protein folding problem largely going away.

Protein folding is one thing (and it's not just bottlenecked by the
performance, but also by the forcefield accuracy), but the inverse
protein folding problem is harder (at the very least it a large resource
multiplicator, if ran e.g. by GA, easily a factor of a million).

>    There is not much in nature that is "novel".  Once a sufficient number
>    of protein structures has been determined experimentally (I'd guess
>    from 100-200,000) there will be very few structures in the phase space
>    which do not have a close relative whose structure has already been

The catalytic space/antibody diversity seems to be about 100000 IIRC
(have to look it up, not sure).

>    determined.  At that time one will care much less about protein
>    folding because it is the final structure which is of much more

The folding pathway is also important, at least in vivo. Also,
some structures in vivo are toxic, so it's not the shape, but the
surface feature presented that must be kosher.

>    5. As the chip also appears to be designed so one can stack memory
>    chips on top of the processor it is nice to see Intel is following up
>    on my 10 year old suggestions for 3D chip architectures, e.g. [6].
>    6.
>    [3]http://www.aeiveos.com:8080/~bradbury/Conferences/Extro3/cpu4.jpg

I did a very good simile (some things done better, arguably) of Intel's 
design in about 1996-96, with the ULIW design. Ideas alone are unfortunately
cheap, prototyping in a foundry unfortunately far less so.


-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820            http://www.ativel.com
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 191 bytes
Desc: Digital signature
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20070214/e28780cc/attachment.bin>