[extropy-chat] Re: Limitpushers and Limitexplorers (Was: Altered genes)

[Anders Sandberg]

söndagen den 28 mars 2004 20.10 wrote Robert J. Bradbury:
> So I don't think one can explore much of the phase
> space(s) without pushing on the limits that the universe
> seems to be presenting to us.  So one has to strike
> some kind of balance between pushing those limits (if
> that is possible) and exploring what they allow. 

Hmm, the glass is (1/sqrt(2))(|full> + |empty>) :-)

You are right in that the space of possibilities is extremely vast, and that 
exploring it would require *extreme* pushes beyond current limits (probably 
into the vicinity of Omega Point-like methods). But are we interested in 
exploring *all* 1 nm nanostructures? Even if we could, the vast majority 
would be rather uninteresting and given the tech needed to do it it would not 
be very necessary to do so. There would be far richer domains to explore. 

This is in many ways the evolutionary problem. Life has not had the time (and 
never will) to explore genotype space fully. It just randomwalks through it, 
producing a range of unique and highly contingent species. In some cases 
these come up with further adaptations that enlarge the state space further 
(more genome, alternate splicing, nervous systems), making it even harder to 
explore it all. And what need is there for evolution to explore all <1kb 
genomes? 

The unmanageable vastness of things to explore is in my opionion wonderful, 
because it makes choices matter. If one could just brute-force through all 
designs and ideas, one would in the end likely end up with just a few 
"optimal" results - the best nanoassembler, the best social system, the most 
appealing artwork etc. The future becomes convergent. But if we cannot 
explore all possibilities, then the future is divergent. Depending on what we 
explore (and how) we will find different solutions, making the world 
contingent. It would likely never repeat itself if it was re-run. This ties 
in with my views that humans (and species) gain much of their individual 
value just by being irreplaceable; no process can recreate the uniqueness 
that emerges from individual development (I leave the ethics of copying 
uploads for later :-). 

Still, in practice the choice of what to explore and to what ends matters much 
and involves plenty of trade-offs. The exploration-exploitation dilemma in 
learning theory is a good example: spend time using what we have learned 
(safe) or try to learn more (risky, but could have far greater rewards)? The 
answer (at least in some cases) is to both explore and exploit, with 
intensities depending on past experience. Often the best breakthroughs seem 
to occur due to unrelated exploration; deliberate and concerted pushes work 
better when one knows there is a crack in one of the limits and want to 
exploit it. 

Right now the originally basic science of genetics is becoming a very powerful 
tool to push back biological limits. Here random exploration of interesting 
things seems to work very well with finding powerful tools that we can apply 
to many problems, and the mutual feedback is strong. In mathematics on the 
other hand many breakthroughs happen out in the esoteric spaces far from 
reality with little connection to it, but sometimes have a very strong impact 
(like in cryptography and computer science). It remains to see if the 
borderlands of applied math and complexity will have the same kind of 
feedforward or feedback effects on our limits. 

-- 
Anders Sandberg
http://www.nada.kth.se/~asa
http://www.aleph.se/andart/

The sum of human knowledge sounds nice. But I want more.