Re: singularity and exi-GIMPS team

From: hal@finney.org
Date: Wed Aug 11 1999 - 16:14:52 MDT


Spike Jones, <spike.jones@lmco.com>, writes:
> Someone please forward this to Eliezer. thanks!
>
> Question Eliezer and other singulatarians: would it enhance
> progress toward the singularity if a large number of personal
> computers were to be set up to run 24-7-52 and were inter-
> connected? It would seem so, and this gives me an idea.
>
> We have been talking about an extropians GIMPS team, and
> we already have a seti@home team. But, we dont *have* to
> run either. We can combine the teams into a parallel computing
> infrastructure, that would then be usable as a resource for other
> projects, perhaps ones that are more pertinent to extropians,
> such as Shane Brubaker's sim of a leech brain, or some wild
> scheme by Eliezer, for instance.

This is an interesting idea, but historically it has been difficult
to adapt software to run effectivelly on a loosely connected parallel
processing system like this. Only certain kinds of projects are suitable,
and the technical hurdles can be difficult to overcome.

A cautionary tale is Thomas Ray's "Network Tierra" project. Ray got a lot
of publicity in the mid 90s for his Tierra alife simulation, which seemed
to evolve various forms of artificial life forms, including parasites,
predators, and other complex ecological entities. He got the idea to
extend it to run on a network of personal computers similar to what
Spike is suggesting. The computers would have different properties
(some would have more run time or memory than others, for example) and
life forms could migrate from system to system. The idea was that the
variation in environments would be an additional engine for evolution.

However, the project got bogged down in technical problems (i.e. bugs).
http://www.hip.atr.co.jp/~ray/tierra/tierra.html describes the current
status, but it hasn't been updated for a year.

Ray is now stuck teaching a "computer programming for non programmers"
class, http://faculty-staff.ou.edu/R/Thomas.S.Ray-1.Jr/Zoo5970/zoo.html,
at the University of Oklahoma. Had Network Tierra worked he'd probably
be a household name (again).

>
> The important thing is to *start* somewhere, which is to pick
> a project, such as seti or GIMPS, and get extropians comfortable
> with leaving their computers on, and with the notion that something
> is running behind the screensaver scenes. GIMPS is a good
> technology demonstration platform for this, since I know from
> experience that it does not crash your computer and it gets
> outta your way when you are doing something else.

Can anyone suggest a more Extropian project? Finding primes is pointless,
and the SETI thing seems badly managed and inherently unlikely to succeed.
Many of us see nanotech or AI as the paths to the singularity, and perhaps
some bottlenecks in these fields are due to lack of computing resources.
However it is necessary to identify a problem which can be solved on a
loosely coupled network where communication costs are very high.

> A few months ago, I would have been loathe to do *anything*
> which might hasten the singularity, however, as I think it over
> the idea is growing on me. Of course, I still have some dread
> that the singularity might want to kill us all. That would be "bad".

As in, "don't cross the beams" bad?

> But maybe not.
>
> Eliezer, please take me outta your kill file temporarily for the
> sake of this discussion, which is up your alley, but delete
> anything from me which is not labelled singularity. thanks! spike

-- Hal



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:04:43 MST