From: Michael Roy Ames (michaelroyames@yahoo.com)
Date: Sat Feb 08 2003 - 11:54:58 MST
Dear SL4,
Last month in the #sl4 chat channel, there was a discussion about
competitive games and whether we should teach a seed AI, or Friendly
Super Intelligence (FSI), about playing them. One group was arguing
that we should not teach zero-sum, or competitive, games at all.
Another group thought it would be useful to do so. I found myself
sympathizing with both viewpoints during the discussion, but have later
come down firmly on the 'openness' side of the argument. Here are some
thoughts.
Competitive Games
-----------------
It is not the competitiveness, epitomized in some human games that we
should avoid passing on to an FSI. It is instantiating an
enjoyment/goal of zero sumness (or negative sumness) that should be
avoided. In competitive games where the plainly written rules explain a
microworld of win-loose zero sumness, there is (almost always) an
unmentioned missing element. This missing element is the enjoyment, the
diversion, the positive-sum *funness* that the humans experience when
playing. This is much less culturally evident in U.S. society than it
is in European society... and therefore may not be immediately obvious
to thinkers with a U.S. based upbringing. This human viewpoint should
be explained to an FSI. The purpose of competitive games is co-equal
with that of non-competitive games, i.e. 'to have fun' and maybe learn
something. That we can simultaneously want to have winners and losers
is an artifact of our human makeup, and an FSI must be able to
understand this. After all, *we* understand it... therefore it cannot
be that difficult to grasp.
Us Versus Them
--------------
It is not the 'us' versus 'them' meme that is a problem, it is the
sanctifying and demonizing that often follows that leads to zero
sumness. Taking sides and competing is a good way for independent
actors to determine what works best, and have everyone profit by that.
It is convenient for humans to adopt this method because with a free
market system (of goods *and* ideas) individual selfishness works to
benefit society as a whole. It is when the over-arching idea of 'common
goals' is ignored (or rejected) that different POVs ossify into
destructive memes.
FSI Exposure
------------
We should not avoid exposing an FSI to competitive ideas in an attempt
to somehow channel the intellect into cooperative-only pathways. Why
not? Firstly, it would not work. Even the most cursory examination of
human knowledge would reveal competition at every point. Even the most
casual conversation with a normal human being will lead an FSI to
formulate the ideas of competitiveness and opposing
ideas/groups/opinions. This information is not something that we should
avoid teaching to an FSI, but rather something we should concentrate on
explaining well. "Look at this," we should say. "This is how we have
made progress, and these are the reasons." We should not shy away from
providing important information about humans, about what we have found
to work and not work. We should show our successes and failures, and
our analysis of both, to an FSI. We should not be saying, "Don't look
there!". Rather we should be saying, "Check out this mess we made.
Don't make that mistake! Here are three better paths, and we think
there are many more better paths. Got any suggestions?".
Michael Roy Ames
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:41 MDT