Re: SI: Singleton and programming

Eugene Leitl (
Sun, 22 Nov 1998 01:09:20 +0100

Anders Sandberg writes:
> Do you really think posthumans can ignore legacy systems and design
> everything from scratch whenever they need it? That sounds very
> uneconomical.

I think it would depend on how much baggage one has to lug around. If we assume an advanced posthumanist civilization, where all available matterenergy is turned into computational substrate (unless you have this ultimative in anal-retentiveness dispelled by a writhing wriggle of hosts and parasites of all scales doing their evolution dance also in physical space -- meshes and hiearchies of fleas ad infinitum), you are obviously bound for a virtual ecology with enormous scope of complexity as artefacted by coevolution. You of course have to represent skills evolved in the past which you might need in the future. However, there is a tradeoff. The resources needed for representation (and be it molecular tape in realspace) cannot be timeshared for computation, and hence for evolving adequate solutions from scratch. If the baggage becomes too large, you are outperformed by individua who had weighted their resource allocation differently. Of course a tabula rasa person constantly reinventing itself is also an equal absurdity as the walking Library of Babel locked in the loop of constant retrieval.

> > Or in other words, what makes you think that redundancy is the best way to
> > deal with the kind of flaws (if any) that exist in Jupiter brains?
> I am not saying it is the best way of dealing with flaws of JBs, I'm
> saying that redundancy makes a system more robust against local
> damage.

Assuming that all known systems, whether natural or men-made (I know this is an arbitrary classification) are hierarchical, why should virtual ecologies of giant computing clusters be different? Maybe I'm stupid, but I don't see any plausible trajectory how one could arrive at an even a flat-hierarchy JB starting with where we are now. (Of course talking about post-Singularity events is meaningless, but it's so fun...)

> > Perhaps
> > there will be redundancy, but without duplication - multiple, separate
> > algorithms.
> Not unlikely. Isn't this why we want to have individuals?

Another reason would seem that signalling costs to obtain some distant copy might be prohibitive, necessity to integrate an off-shelf module into a distinct persona (mutation) set aside. Remember Kauffman's lettuce which we break down do common building blocks instead of cosmically fusing with it.