Re: No AI for Nano/No Nano for copyloads

From: Emlyn (onetel) (
Date: Sun Jul 16 2000 - 04:03:17 MDT

I'll say if the destructive scan procedure is 10% likely to work, I'm in
(given that I'm a corpsicle). I'd like to say a higher number, but I think
I'm likely to be outcompeted. I'm not quite willing to go lower, although I
suspect I might still be too conservative.

When you pick a particular chance for success, you are betting that all the
prior experiments, done before that level of sophistication is reached, will
fail. At the 90% chance level, I think you are almost a dead certainty (ooh)
to miss the first wave. Big critter cloning is like this currently, is it
not; many failures, very few successes? Dolly is a really lucky gambler.

A ramble, which will lead back to the main point:
I've been playing an online game, Utopia, for the last few months. It's a
massively mutliplayer strategy games (press says 60,000 players, and I think
that's actually about right).

In this game, you play a fantasy province; Orcs, elves dwarves, halflings
(yawn, yep same old stuff). It has a interestingly simplistic geographic
model - there's no map, you just exist in this abstract world, with a
certain amount of acres of land, all kinds of production statistics and
settings which lead ultimately to an expansion rate of population
(particularly limited by land) and income level (production rate of gold).
You use those things to make an army, and you use your army to get more

The bread and butter of the game is land. Your amount of land drives &
bounds all other aspects of your province. When you attack someone and
succeed, you get an amount of their land, and they lose that amount. Land is
totally abstract, just a single integer value.

The interesting thing is that there are various inputs to the game which
cause the mean level of land to increase over time (the actual mechanism is
that new players who don't understand the game too well turn up constantly,
are allotted an initial amount of land, fail to cope, and drop out, losing
all that initial allotment to other players in the process; it becomes
subsumed into the greater land economy).

Also (and even without that growth factor), larger provinces dominate
smaller ones; they have larger armies. So, amongst really clued in players,
there is exponential growth, limited by the number of other players who are
at a similar size to you (because you cannot grow effectively by hitting
players far smaller than you, due to various game mechanics).

The game is reset every 2 months or so; everyone starts anew. This is
because by about that stage, there is no hope for new players to make any
kind of decent go at getting ahead; the game becomes strongly stratified.

When you start from the reset point, and have really sophisticated tactics
(people have written incredibly elaborate software, some of which is freely
available, for generating and testing strategies offline), the game has a
very interesting character. Specifically, as time goes on, the size of the
largest provinces in the game increases (exponentially). If you are one of
those largest provinces, initially you must play a very tight game indeed to
stay out the front of the pack. This has enormous advantages; later, when
you are one of the biggest, you almost never get attacked (no one else is
big enough). If you are slightly below the size of the largest provinses,
you get hit occassionally (by the biggest, who are few), which slows your
growth further. If you are nowhere near the front of the pack, you get hit
quite often, which costs you heaps economically even if the attacks fail;
thus your size, over time and relatively speaking, falls far below that of
the leaders.

It's a really fantastic example of the founder effect. Even though it is a
highy dynamic and highly competive situation, if you lead, your are more
than likely to continue to lead. If you are in front from the start, the
game is a lot easier than if you are trying to come up from below; you can
achieve the same effect with far less risk. This becomes more true as the
game stratifies further, and there are less and less people who can truly be
considered to be at the top (countered only by the fact that once there are
only a few really big ones, they can identify each other, and can tend to
attack each other out of vanity, to become the truly top province; provinces
that get involved in such spats tend to find that neither is a contender in
fairly short order, as they waste each other's resources.)

A countering effect also is that once you are clearly out amongst the front
few, there is a tendency to rest on your laurels; falling from the front
creeps up on you gradually, and by the time you notice it is often too late
(happened to me this round).

Anyway, what's my point? Experience of the founder effect in this kind of
environment is incredibly instructive if you want to really understand what
it means. To be first, you have to go to extraordinary lengths. Also, you
have to take some quite extraordinary risks. Why is this?

In a game like Utopia, I would estimate that as the round starts, there is
something in the order of 5,000 truly competent players on the starting
blocks; these are players who have played before, and/or have researched the
net resources (extensive!) to find out how to have an optimum start. They
all have a roughly equivalent chance of getting to the top 50 (the winners
group), as there is not that much randomness involved in the game (except,
of course, that of the combined decisions made by 60,000 players, hmm). By
the end, there is quite a difference between the top and bottom of the top
50, and they are, in aggregate, incredibly far ahead of the average player.

The truly clued up players, who are going for glory, know that you cannot
play a conservative game and get to top spot. Say you have an oportunity to
take a risk which you estimate is 50% likely to put you significantly ahead,
and 50% likely to stuff up and be quite detrimental; knock you right out of
the top league (a very common situation in Utopia). I would normally, as a
relatively conservative wargamer, not take such a risk. However, in Utopia,
there are so many players, that the story is quite different to your usual
wargame including a handful of players. If you have choices to take risks
like this, so do other players. Some of them will blow out, and some will
succeed. In a short amount of time, the top leagues will be made up only of
these who took the risk and succeeded. When you add a bunch of such
decisions on top of each other, most risk takers tend to get weeded out. But
the top provinces, the truly powerful ones, are made up of those players who
beat the odds. They are no better than the players who took the risk and
failed. Just luckier.

(I'm getting the wind up signal from the floor manager) Situations like the
proposed upload race, where the first really successful uploads will be
those that dominate the social environment, whatever that may be, remind me
a lot of the Utopia style situation. I think we can assume that there will
be quite a lot of people who understand the potential of being amongst the
first uploads, and judge a given risk of failure to be preferable to their
more mundane alternatives (deep freeze, or else lives which they value
less). Some of those people will choose a very high risk level, in order to
be first. If there are enough of those people, success will occur at that
extremely high risk level (because at least a couple of them will succeed).
And we are only talking about needing to get a handful of uploads done, if
more than one, is that correct? I'm betting (based on nothing) that at
around 10% success probability, you are still in the front wave.

But as usual, I'm probably being overly conservative.


(By the way, I think there's huge potential for testing the base concepts of
parts of dynamic social systems in the online gaming environment. You don't
need to simulate what people will do in a given situation, you can get real
people to make the decisions. Has anyone tried using the online massive
multiplayer game environment for serious research?)

> Emlyn (onetel) wrote:
> > If I was in a position to be cryonically suspended (which I don't think
I am
> > in Oz), I think I would make it very, very clear that I *volunteer* for
> > first uploading experiments. ... Exploit me, I don't care; becoming one
of the
> > first few Posthumans has got to come with a price tag. And ye gods, the
> > payoff...
> I am a cryonics customer, and I also want to make it clear that I
volunteer to
> be one of the first. Specifically, if a destructive scan procedure is up
to an
> 90% chance of getting all the relevant info (ideally as judged by an idea
> futures market), then scan me! And run my scan info in lots of
> trying to get the simulation rules right. Oh, if I seem to be in
excessive pain
> you might halt the simulation and try something else. But yes, the payoff
> be huge.

This archive was generated by hypermail 2b29 : Mon Oct 02 2000 - 17:34:39 MDT