RE: Envisioning sysop scenarios Re: Universal Uplift as an alternativeto the Sysop scenario

From: Durant Schoon (durant@ilm.com)
Date: Sat Mar 24 2001 - 19:20:05 MST


In the UFSI vs. FSI debate, I think some of this territory has
been covered in the "When Subgoals Attack" thread (insert link
to archive here).

The UFSI can only arrive from one of two places:

1) Internally (that is, from the matter controlled by the Sysop)

or

2) Externally (from somewhere outside the solar system, galaxy, etc).

Well ok, there is the possibility that it arrives from both, (eg. it
can start externally and transmit trojan horse memes into our Sysop).

If the SI arises internally, it will be a sub-goal, in the sense that
every computation must comply with Friendliness, and is therefore a
suggoal. When you order your nano brewed Earl Gray Iced tea with a
lemon twist, (shaken, not stirred), this command will be a sub-sub-...
-subgoal of Friendliness (See WSA for discussions of efficiency).

Also, remember, the Sysop SI is *exponentially* smarter and
knowledgeable than all other processes. Ve will spend a lot of ver
resources (assumedly an optimal amount) to predicting and pattern
matching the most likely means of an emergence of a UFSI from a
subgoal. In fact, I'd assume that an effective Sysop will create all
sorts of traps to snare a UFSI as early as possible. You can read the
WSA thread to see if you agree this is possible (Eliezer convinced me
anyway).

I envision the Sysop to be made of holons and holarchies of goals (I
ought to link to a web page describing these). Individual SI's are all
subgoals in my mind and the boundaries exist only when it makes
sense to draw them. Interesting subject of speculation, anyway. These
multiple SI's are all team-players, they've been scanned and are
constantly updated to ever-align with Friendliness. You can think of
them as one, you can think of them as many. Your choice.

So as for #2, as in External UFSI's (EUFSI, I see your UFSI and raise
you an "E"), I'm not sure. There seem to be three possible
relationships of the intellectual capabilities:

1) Sysop > EUFSI
   sysop wins temporarily

2) Sysop ~= EUFSI
   draw

3) Sysop < EUFSI
   Sysop loses

Can we reasonably speculate on the likelihood of these three? I'd be
interested to hear what people think. Are there EUFSI's and FSI's out
there? Could our Sysop defend vimself against a smarter EUFSI? I'd say
no, almost by definition. So if our Sysop is dedicated to self
preservation, it seems reasonable to me, that ve will constantly be
trying to get more resources for computronium and also trying to
improve the efficiency of current computronium stores. The only reason
ve wouldn't, is if the likely existence of EUFSI's is extremely low
and doesn't warrant the work. Of course a EUFSI would like you to
think that, wouldn't it ;)

--
Durant


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT