M E T A D R E A M S

                        With an Investigation
                             into their 
                    Potential and Ultimate Future

                        by Benjamin D. Cooley





"If a tree falls in an empty forest, does it make a sound?"  

We've all heard the question, and inevitably, most of us come up with the
same answer.  Since the universe is consistent as far as each of us has
ever experienced, we assume that its laws apply everywhere, and that the
laws continue to apply at all times.  So when the tree falls in the empty 
forest we must beleive, based on our own experiences, that the answer
is yes, the tree does make a sound.

But suppose the universe isn't as consistent as we beleive it to be.
Suppose it only extends to the limits of our perceptions, or the limits
of the perceptions of all of the intelligent beings on the planet.  Then
the tree, if no one is there to see it fall, doesn't actually fall, and
therefore makes no sound at all.  

As an answer to the tree question, it's an interesting perspective, and
certainly one that's always considered in any discussions, but in what
way is it related to real life?  It obviously doesn't affect us in our
day to day lives where cause and effect seems pretty consistent, nor
has it ever come up in research where, so far, every incongruous result
has had a definite cause.  So, since there is no way to objectively test 
this hypothesis one way or the other - at least as far as we know, it
seems that we might as well stick with the conventional answer to the
tree question and forget this one.  So why did I bring it up at all?

I brought it up because this way of looking at the tree question, or
reality in general, provides a solution to a very difficult problem
encountered in implementing virtual reality simulations that are meant
to be indistinguishable from ordinary reality.

Maybe I should start with the problem so that the solution, when
explained, makes sense.  The problem is that, as the size or resolution
of a virtual reality simulation increases, the storage and processing power required
to perform the virtual reality simulation also increases.  And since the
increase in space simulated is exponential (i.e. from cubic foot to
cubic miles to cubic lightyears, etc.) the storage required to simulate
the space also increases exponentially.

Now, I'm talking about simulations that the participants in are supposed
to be unable to distinguish from reality, not the simple simulations
that they can easily discover are fake, though even artificial looking
simulations suffer from the increase of storage to a lesser degree.
The storage and processing requirements of the realistic simulations
are enormous since they must simulate everything to such a high degree
of accuracy, and therefore the need to reduce their storage space and
processing power requirements is greater.
 
An easy way to decrease the amount of storage space needed is to remove
redundant information like empty space, straight walls, etc.  But as
the redundant information is removed, resolution of the simulation
inevitably suffers.  For instance, suppose we were to attempt to reduce
the resources needed by the simulation by saying that empty space was
completely uniformly empty.  The only problem is that real space isn't
like that, so the user when he investigates finds that space,
which should be filled with the occational atom in differing 
distributions, is inexplicably empty, or is simply filled with an
unbeleivably uniform amount of atoms.  So we're stuck, either simulate
everything, or be satisfied with the occational inconsistencies.

But suppose we divided the data model used in generating the simulation
into two distinct catagories: absolute modeling and perception
modeling.  We'll define absolute modeling as the modeling of reality
by storing all information needed to generate the reality to the desired
resolution and size.  And we'll define perception modeling ass storing
only those components of the reality that can be perceived by the users
in the reality. Obviously the storage requirements of the Absolute method
would tend to be greater than the perception method, and so the
perception method seems like the better route.  By the way, notice that
the two methods are not mutually exclusive.  The perception method is
actually a superset of the absract method.  In other words, the
perception method uses the abstract method to simulate reality, but only
the components of the reality which the user will actually perceive.

Althought the perception method seems much better at first glance, 
there are obvious problems.  Some components of real reality are easy
to eliminate using the perception modeling method.  For instance, if
none of the participants have electron microscopes or radios, the
simulation can dispense with simulating the electromagnetic spectrum.  
Also, since some solid objects may not ever be broken by the
participants, their internal structure may not need to be simulated.  

However, beyond these obvious examples, picking the parts of reality
that the participants won't notice have been fudged becomes a lot
harder.  For instance, suppose we wanted to take a newly plastered
virtual wall and represent it as random distribution of lumpy blobs on
a straight wall.  If the wall was originally plastered by someone in
the reality, he would come back and notice that the wall now looked
different from how it looked when he plastered it.  So the sytem would
have to remember the strokes of the plasterer if it wanted to make it
look realistic, or perhaps even the motions of his body as he plastered
it, then map the random distibution of the plaster blobs into his
motions to make it look right, and even then it might not convinve the
plasterer.  Even then if the plasterer remembered a particular blob on
the wall the system is screwed.  So eventually, just to make sure the
system hadn't missed anything, it would have to record the entire wall
down to a faily detailed resolution in an absolute model.

As it turns out, based on these kinds of situations, perception based
modeling really doesn't provide much savings in the storage and
processing department.  The increase in the size of storage required
for increasingly large simulated areas is still pretty high.

Of course there is another solution.  You can abandon creating a
realistic simulation and create a cardboard mock-up simulation.  This
type of simulation uses pre-defined cut out algorithms, surfaces, and
models, to decrease the amount of storage required.  All plastered wall
surfaces in this type of reality are either random distributions of
blobs, or based on the same surface map, or combinations of the two.  
What you get is a cookie cutter world where every surface looks exactly
like every other, every object works the same.  The simulation justs
pays lip service to working like the real world, and does not attempt to
be consistent or detailed.  In our universe everything, even those things
that seem the same like mass produced goods, are subtly different.  In
the cardboard virtual reality everything would look different, but in
actuality would all be identical.

However, there is another alternative.  We can use perception modeling to
decrease the amount of storage and processing power needed if we add
one more needed component to the virtual reality simulation computer: 
the minds of those that are to use the simulation.  If the computer
knows exactly what the participants are going to notice and not
notice, it can specifically generate the important things to the correct
level of detail and fudge the rest.  The participant will never know the
difference.

Take our plastered wall for instance.  The plasterer certainly remembers,
perhaps subconsciously, some of the details of the wall, and he also
knows generally how the wall should look.  So the computer takes that
information from the plasterers mind, combines it with its own absolute
information as the the position and size of the wall, etc. and recreates
the wall to a level sufficient to convince the plasterer that the wall is
real and consistent with what he rememebers.  With access to the
participants minds, the computer knows exactly what it can fudge, and
what it can't. And not only that. From the information in the user's
minds, the computer can actuall generate most of the reality without
really needing much more additional information!

What this means is that the storage space and processing power required
to generate realistic virtual reality is greatly reduced.  In fact, the
amount of processing power and storage is not even really primarily
related to the amount of space simulated, but to the number of people
that are part of the reality, and the proportions of this relationship
are linear not geometric.  Add one person, add the storage space for
that person plus the overhead that his presence adds.  The size of the
universe, and more importantly, its complexity still play a part, but
the primary components of this type of reality are the people themselves.

This type of simulation I will call a METADREAM since it really can be
defined as a computer assisted group dream that's shared by all the
participants in the simulation.  The metadream model of virtual reality
has several important components: the participants, the reality
generator, the reality creator, and the consistency auditor.

The participants are the largest component of the system, providing the
primary storate for the reality and also the rules and expectations by
which the reality is generated.  The reality generator extracts the
necessary information from the minds of the participants to generate
the information that the participants will perceive through their
senses, the reality.  The reality creator also accesses the reality
generator when there are gaps in the information it retrieves from the
participants (for example, the distribution of blobs on the plaster
wall).  The reality creator would take the relevant information from
the reality generator (which came originally from the participants), 
combine it with its own internal models and sets of physical laws, and
return sensory information to fill in the gaps.  The consistency
auditor would catch when the other two systems had made a mistake and
possibly roll back the metadream to a previous state and hopefully
avoid the mistake when it came up again, or unobtrusively manipulate
the metadream to remove the mistake, or evidence of it, from the current
point. Together these three systems make up an AI level program which I
like to call "The Mother of all Dungeon Masters".

Now you might be able to understand why I used the "Tree" question to
try and explain my metadream concept.  In a metadream, everything is
entirely perception based.  Only those things that are currently
experienced are actually real.  Objects such as trees, houses, and
other inanimates simply don't exist but for memories or abstract models
until someone is around to actually experience them.  Therefore the
tree falling in the empty forest in a metadream doesn't make a sound, 
nor does it even fall until the reality creator needs to make a forest
and decides that a certain number of trees, based on the climate of the
area, and the rainfall, etc. have fallen.  

The metadream is a pretty powerful concept, but it's not the limit of
the potential of virtual reality.  For the be all end all of virtual
reality I'll coin another term.. Virtual Universe.  A virtual universe is
a metadream on a grand scale.  The virtual universe is the simulation
of an entire population of people to the limits of their perception.  
In our case, our metadream would focus on our own planet, with minor
detail only on the other stuff in the universe to make everything look
consistent.  The metadream would consist of six (or seven?) billion
intelligent minds, numerous semi-intelligent minds (who, by the way, 
can be statistically simulated as their intelligence and their direct
interaction with the intelligent population decreases).  Of course this
particular metadream would require a massive control system to keep
track of the different reality models that would be needed for the
planet and apply them when needed.  However, since most of the models
required are localized, the control system can remain relatively
decentralized, working perhaps with each continent, each city, and each
house, with different levels of detail and differnt models.  Once the 
planet is successfully simulated, the rest of the universe is easy, 
since the perceptions by the inhabitants on the planet are very
limited.

So basically, using the metadream form of virtual reality, you can
conceivably simulate an ENTIRE UNIVERSE (more or less, i.e. don't look
behind those curtains Dorthy!).  Maybe not a universe as we understand
the concept, a huge infinitely complex, infinitely detailed, expansive
cosmos, but still a universe that could fool us if we didn't know the
truth.  Perhaps if the Grand Dungeon Master of the virtual universe is
smart enough, we could possibly be fooled indefinitely.  At the very
least, we can esacape the cardboard virtual reality simulations and get
real, imperfection rich, detailed simulations with metadreams.

The metadream and virtual universes will probably be an inevitability in
the evolution of our species.  They won't mean instant utopia, but they
may aid us in reaching our full human potential.  Besides, they would be
a blast.  Think of it.  There can be as many types of universes as can be
imagined and built.  Pern, Middle Earth, Dune, all these places can
exist, or at least places that are like these can exist, modified by the
fact that there are real people inhabiting them and not fictional
characters.  In many universes, magic will probably be a reality.
Perhaps all those heroic fantasy stories of Tolkein, Feist, Moorecock,
and others could actually take place some day.

I beleive that over time the real universe will fill up with theses
things, each planet containing millions.  Imagine the planet earth with
a million subuniverses dotted across its surface.  An incredible density
of life!  Perhaps trillions of intelligent beings sharing the same
planet, with maybe no more than a billion actually on its surface as
living beings at any given time.  It would be overcrowding but without
any ecological disaster, and inside there would be elbow room for
everyone.  The systems wouldn't be unlimited, the planet would still
only support a certain number of people, but the number would be
greatly increased, perhaps a thousand, perhaps even a million times.

So, combine the arguments for digital people: imortality, increased
intelligence, body independence, and faster and cheaper travel, and
combine it with the added possibilities raised by the metadream, and
you have a pretty persuasive argument that humanity's future will be a
digital one.

END.