Attractors and Values

From: N.BOSTROM@lse.ac.uk
Date: Sat Feb 01 1997 - 19:09:54 MST


          Not all interesting statements about the future need to be
          specific. Suppose, for example, that we want to claim that
          all advanced civilisations tend to approach some common
          ideal state, but that we don't want to commit ourselves to
          exactly what this state is. Well, why not define a
          convergence thesis, saying that the possible civilisations
          trajectories through configuration space tend to converge in
          the positive time direction. This expresses an interesting
          form of sociological/technological determinism (one that
          doesn't have anything to do with physical determinism on the
          microlevel).
                As it stand, the convergence hypothesis is quite
          unsatisfactory, however. It is instructive to think about
          how we could begin to refine and sharpen it.
                We could begin by clarifying what we mean by
          "possible" civilisation. We could mean every civilisation
          that is consistent with physical laws, excluding boundary
          conditions; but something that is more restrictive might be
          more fruitful. So we could say that the "possible"
          civilisations are all possible civilisations that are
          compatible with what we know about our civilisation; the
          idea being that we are interested in what might happen to
          the human civilisation and that we say something about that
          by saying that all possible civilisations, which have all
          properties we know that the human civilisation has, will all
          share the same long term fate. We might then want to soften
          this a bit by modifying it to "almost all of the reasonably
          probable specifications of human civilisations (modulo our
          knowledge) will share a similar long term fate". -Still very
          much too vague, but a step in the right direction. We might
          go on to decide how long the "long term" is supposed to be,
          and who "we" shall be taken to refer to (you and me? the
          intellectual elite? all living humans?), and how similar the
          shared fates are supposed to be, etc. Needless to say, we
          are not aiming at mathematical precision here, that would be
          to shoot far above the goal.
                An interesting variant is to extend the denotation of
          the "possible civilisations" to include not only possible
          civilisations that could turn out to be our but also other
          possible civilisations that are sufficiently advanced. We
          might want to say something like "Almost all civilisations,
          once they have become sufficiently advanced, will become
          even more advanced, and as they advance they will become
          more and more similar in most important aspects.". Add a
          little precision, and you would have formulated an
          interesting proposition.
                There are other flavours of the convergence thesis. We
          might be interested in a thesis that says that all possible
          civilisations into which we could transform our civilisation
          will share a similar fate. (If that were true, we would be
          powerless to change the world in the long run.) Here it is
          very important to specify what we mean by "we". For example,
          if "we" were all living humans, then we could easily
          transform our society into one in which no crimes were
          committed -and that might be a good idea-, but if "we"
          refers to you and me, then we can't do that. (I find that
          discussions about politics often suffer from a lack of
          relativisation of policy to agents: what should you do? what
          should your interest group do? what should your country do?
          what should civilised educated people do? - it is hopeless
          to try to work out a good policy in general; one can only
          make a good policy for such and such agents in such and such
          situations, (given such and such aims).)
                One rival hypothesis would be the divergent track
          hypothesis, according to which the future trajectories will
          divide up into a small number (3 2) of diverging clusters,
          the trajectories within each cluster tending to converge. It
          is slightly misleading here to speak of converging
          trajectories; what is meant is rather "routes of development
          of civilisations tending toward the same goal-state". As an
          illustration, take the following somewhat ludicrous story.
          Some deep investigation reveals that in each possible
          civilisation similar to ours in certain specified ways,
          there will emerge either one or the other of two religions,
          A and B, with roughly equal probability. These religions
          will be such as to inspirit their adherents with such zeal,
          cohesion and adaptability that they will eventually come to
          dominate the culture in which they arise. Having obtained
          local power, they will employ new technologies (drugs,
          electrodes etc. etc. ) to cement their old strongholds and
          to win converts from other groups as well. The stronger
          these religions become, the better they are able to optimise
          their strategy. Thus a positive feedback loop sets in and
          soon leads to total domination on earth. Then the religions
          embark on the project of transforming as much of cosmos as
          they can into the structures on which they place most
          values; perhaps they generate the cosmic equivalent of the
          Tibetean prayer wheels, giant ultra-centrifuges rotating
          trillions of inscriptions of "Gloria in excelsis Deo A" or
          "Deo B" as the case might be. All civilisations in which one
          of these religions emerges, will converge in some sense:
          they will all lead to the rapid transformation of earth and
          the gradual transformation of cosmos into the specific value
          structures of the religion in question, although the timing
          and precise execution may vary somewhat between different
          possible civilisations.
                In this case, one could say that the artefactual
          configuration space of the universe (i.e. its configuration
          with respect to its content of artefacts; two universes are
          in the same artefactual state iff they contain identical
          artefacts) will have two attractors: world dominion of
          religion A or of religion B. Moreover, we could say that the
          paths toward the attractor centre are quite uniform over all
          realistic directions of approach. When this is the case, we
          say that the artefactual configuration space contains
          tracks, courses of development such that once a civilisation
          has begun to travel along them, it is unlikely that it will
          diverge from them barring major external event
          interposition.
                 We are now in a position to formulate and argue for
          an interesting hypothesis about the future's topology: the
          track hypothesis, saying that the artefactual configuration
          space for all civilisations roughly comparable to present
          human civilisation contains trenchant tracks in the future
          direction, either one track or a small number of them.
                The outlines of some fragments of the argument for
          this claim (a full exposition would presumably require
          book-length treatment) could begin to be drawn as follows.
          As progress is made in science, technology, infrastructure,
          economic structure etc., this will have the effect of making
          us more effective. New technologies will increase our power;
          augmented cognitive capacities (whether through >AI or
          through mere extension of present systems such as science,
          education, information technology etc.), will increase our
          understanding of the consequences of using this power in
          various ways. The result of this is that we will have
          increased ability to make reality conform to our desires.
          There is no reason why we shouldn't also be able to mould
          our desires according to our higher-order desires. Thus, if
          are only a few highest-level desires that are genuinely held
          by large number of influential agents, then, it might be
          argued, there are only a few attractors into which our
          civilisation could sink, and if it could be established that
          the approach to any of these attractors would tend to be
          rather uniform over all realistic directions of approach,
          then we would have found that our future-topology contains
          so many tracks, and we would have made a case for the track
          hypothesis.
                One could then go on to list some prima facie
          plausible basic goals or values. (Basic ones, not values
          such as playing golf, for those who value that activity
          presumably does so because they think it is fun; but if,
          e.g., they could have much more fun by having their reward
          centres directly stimulated chemically or electronically,
          without any ill side effects, then there is no reason to
          suppose that they would insist on continuing the golf.) Here
          are some out of the hat: (1) maximal total pleasure
          (hedonism); (2) average of present human meta-desires
          ("humanism"); (3) maximal consciousness, pure consciousness,
          religious experiences, wonderful deep experiences
          ("spiritualism"); (4) maximal reproduction ("Darwinism",
          could this be argued for on Darwinistic grounds if several
          competing value systems are present? -Hanson (1994)); (5)
          maximal practical utility, such as safety, computational
          power etc. ("pragmatism"); (6) annihilation, voluntary or
          involuntary (nihilism). Involuntary annihilation is not a
          value, but a very real possibility anyway; it's a plausible
          candidate for being one of the tracks in our
          future-topology.
          
          Nicholas Bostrom n.bostrom@lse.ac.uk
          



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:44:09 MST