Re: Let's form an RC5 cracking team

Bartley R. Troyan (bt26+@andrew.cmu.edu)
Wed, 17 Sep 1997 18:23:59 -0400


Eliezer S. Yudkowsky wrote:
> I'm more impressed by Bovine getting 38% of a 2^56 space. That's enough to
> do, what, almost a quarter of a second of human thought? And all this by
> offering $1K? In a few years, anybody with a Power-program to debug will be
> able to get human-level processing power, albeit slowed down by a couple of
> thousand to one, by offering a mere $10K.

It's exciting to imagine the possibilities that arise even today from
such a large mass of distributed compute servers. It looks to me like
Bovine's group has implemented a distributed computing architecture of
high quality and resilience (although I have yet to see their code).
They've made it almost effortless to donate cycles--I logged into a
sparc ultra today and in 60 seconds with no prior experience had its cpu
load up to 100% running their client. Not bad...

Although they have made clients available for many popular platforms,
I'd like to see a Java applet client, allowing anyone with a web browser
to just load up a page to start donating cycles. Of course having a jvm
doing the compute-intensive operations is pretty slow but it's possible
to write native methods that would speed things up quite a bit. And
until you get around to writing them for all the different architectures
it will at least work slowly but correctly. At least the GUI/networking
code can be written once and reused everywhere.

But forget about the RC5 project and the paltry "reward" they're
offering--what else can be done with a million pentium-class machines
all offering lots of free cycles to the world? Abstract away the
rc5-cracking-specific code from their client and allow it to run any
module of compute-intensive code that applies to a problem at hand.
What kind of distributed neural network simulations can be built on top
of it? Or does the bandwidth become a limiting factor that would make
it impossible/pointless to run in real time?

Maybe compute cycles will become a unit of currency...

> In fact, how do we *know* these are key-testers we're running and not a
> distributed, sentient mathematician?

Yeah, I think that's just running in the background as another thread.
It's 4am...do you know what your CPU is doing?

Bart Troyan
Carnegie Mellon University

> --
> sentience@pobox.com Eliezer S. Yudkowsky
> http://tezcat.com/~eliezer/singularity.html
> http://tezcat.com/~eliezer/algernon.html
> Disclaimer: Unless otherwise specified, I'm not telling you
> everything I think I know.