Re: Delicate complexity? (was: RE: computronium prime-oxide_

Hal Finney (
Sun, 22 Nov 1998 19:11:47 -0800

Hal Finney wrote:
> >
> >> they ran into balancing problems, where trivial
> >> strategies would dominate (often a problem with alife simulations).
> >> At last report they were introducing various ad hoc rules and limitations
> >> to try to get robust evolutionary behavior.

Nick Bostrom wrote:
> >
> >At this most basic level, complexity or growth in compexity does
> >certainly not seem to be the most likely or natural situation. But
> >since it's boring (for the human psychology) we change the parameters
> >until we get something interesting. I makes me wonder if we may not
> >be prone to overestimate the likelihood that the far future will be
> >complex.

Curt Adams wrote:
> I expect that it will be complex. Complexity is thermodynamically
> favorable - it's simplicity that's hard.
> In the wild speculation department, I wonder if part of the reason
> that Alife has a problem with trivial strategies arises because
> alife doesn't generally try to deal with thermodynamics. Hence
> they're not modelling one of the biggest drives to complexity.

Stuart Kauffman is known for his belief that life evolves "to the edge of chaos". Something about evolution has kept it from tipping over into maximally simple strategies. It may be that much of the early history of life on earth was a matter of "learning to evolve", that is, the creation of the genetic code, sexual recombination, and the other basic parameters which govern reproduction. This system had to be such that it not only rewarded innovation, but that it was robust enough to maintain the characteristics of rewarding innovation.

You could imagine many alternatives being tried, some of which worked well initially, but which drifted away from the properties which allowed evolution to work. Only when a system was developed which somehow carried its own incentives to maintain itself, to maintain the properties which allow evolution to work, was it truly stable. At that point evolution took off and organisms descended from that system took over the world. This is then what Kauffman observes as a tendency to stay with near-chaotic behavior; only a genetic system which had such a tendency would allow for long-term evolution.

This is all speculation, of course.

Extending it, we may have had the same thing happen culturally. We developed cultural systems which not only encourage and reward innovation (for surely such have arisen at random many times), but which maintain themselves. Such a cultural system holds itself at the edge of chaos and allows cultural evolution to occur.

When we initially enter a new realm in which evolution can occur, as with early alife, there is no a priori reason to expect complex behavior. Most of the time, a simple strategy dominates. We have to play with the systems and try out many possibilities before we find some which evolve complexity. If these various alife systems were actually competing in some forum (which they are, in a way - they are competing for our attention and computer cycles), then only systems which can preserve their complexity-enhancing characteristics will win. That's not a big problem for today's alife programs because usually the way inheritance and reproduction works are fixed by the creator, but in the future it may become more of an issue as the simulations get more realistic.

Similarly, as we move into the Singularity, if the nature of evolution is fundamentally different from the cultural evolution of today, there is no reason to expect chaotic behavior. It is more likely that a simple strategy will dominate. Only over a longer period of time, when many forms of Singularity entities can form and reproduce and grow, will there be time to evolve a mode of reproduction which encourages chaos and evolution as we know it. It's not clear whether we get that far, though, whether the post-Singularity entities will be such that reproduction, variation and selection can occur.