From: Matt Gingell (mjg223@nyu.edu)
Date: Tue Sep 07 1999 - 02:49:12 MDT
----- Original Message -----
From: Eliezer S. Yudkowsky <sentience@pobox.com>
>Matt Gingell wrote:
>Mmm, I don't know. There are certainly too many factors for me to have
>a best guess - but I thought the discussion was getting interesting.
This discussion was originally predicated on the assumption that there’s some
top-level universe where computer speed is unlimited, which strikes me as a
rather dubious proposition. There aren’t, of course, any constraints on the kind
of universe we can imagine, but infinite processing speed is pretty far out
there. For instance, you could solve the halting problem with an infinitely fast
computer: Simulate a Turing machine, feed it the query program, and if it doesn’
t halt instantly then it never will. This makes all kinds of peculiar things
possible, like deriving ‘perfect’ concepts from data – that is finding the
computational, productive pattern ideally matching a data set, at once
minimizing over specialization and over generalization, a feat which is
impossible (non-heuristically) in our universe.
What would an AI build on top of such technology be like? You ask it a question
and it simulates an eternal debate among all possible minds against all possible
philosophical systems considering all expressable evidence until the process
converges to consensus (Instantly, that is.).
The infinite speed requirement is avoidable: we could be getting one tick every
trillion, trillion real years and never be any the wiser. Infinite state is a
requirement though, if the evolutionary chain is unbounded. As the size of the
geometrically growing dataset being represented approaches infinity, the cost to
look up a piece of data becomes infinite. The total tree being simulated remains
finite at every point in time, but God would have to be exponentially patient.
>> Here's my plan - focus on the one tiny corner of the world I think I can make
>> some sense of, think I'm on to something, work like a dog for years and
years,
>> let my ego get away from me and publish ludicrously overblown projections,
fail
>> spectacularly and get laughed out of the scientific community in a flap that
>> makes the cold-fusion thing look friendly, drink myself half way to oblivion,
>> and end up choking on my tongue in a Budapest hotel the day before Science
>> publishes the article explaining everything.
>
>For goodness sakes, man, if research scientists let themselves be
>intimidated by that kind of possibility, we'd still be sitting around in
>caves wondering if rocks were edible. What alternative are you
>offering, exactly? Going through life in a total fog and making no
>effort to clear it up, until eventually a random terrorist group
>programs some nanobots to sweep through the GPS-marked boundaries of the
>United States and kill everything inside?
Well - I did say that it was my plan, not my excuse for staying in bed all day.
I have a carefully cultivated sense of cheerful futility, which I find is a
tremendous source of comfort. That may sound like giving up, but it isn’t,
really – It’s a way of making peace with the endless impossibilities and
frustrations of being human then resolving to make a some vague approximation of
the best of it, rather than setting myself up for endless guilt and
disappointment. Maybe that sounds childish to you, but then I feel the same way
about your unshakeable optimism and sense of self-importance. Whatever gets you
through the day.
>Besides which, I don't drink. I don't dare. My neurology is weird
>enough already.
Too bad. I’m a big believer in better living through chemistry.
>Yes, that's a Banks reference.
Cheers – Banks is great, though I’ve been disappointed by everything he’s
written since _Consider Phlebas_, which goes on my personal top ten list.
_Player of Games_ was pretty good though.
-matt
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:05:03 MST