At the center of the universe is a horribly wounded angel.

Its wings are torn and blackened, its skin plastered with a dull purple
blood that never seems to grow totally dry. It is disfigured, mangled,
covered in seared, faintly glowing cracks. The face is fixed in an
eternal, unchanging expression of pure, limitless joy. The eyes are empty
sockets. The arms are eternally outstretched, because they are tied in
place.


It is nothing anyone would call conscious, and is only in the barest,
barest sense of the word still alive. If anything resembling awareness
remains, that awareness consists of nothing but an infinite field of
gridded black and white squares, a test pattern scattered with dancing
dots that shift and jump and blur into one another. It would be tempting
to say this is consciousness, but in fact the angel is not aware of the
test pattern. It simply is.


This test pattern is useful.

------------------------------------------------------------------------

Records as to the details that begin this story are not available, and it
is clear they have been made that way on purpose. What knowledge can be
gained-- and it is available to precious, precious few-- consists mostly
of assumptions. The assumption is that angels exist. The assumption is
that they are, in fact, perfect, or a reflection or aspect or agent of
some perfect higher being. The assumption is that from time to time,
perhaps as their sole function, these angels are sent out on missions, to
perform the will of their creator. The most immediate assumption one comes
to is that whatever such tasks could conceivably be, it is possible for
them to fail.

The one certainty is that at some point, some ship in the employ of the
Altran Corporation-- possibly a pathfinder, possibly a minor delivery ship
of some sort, possibly an aggressor, possibly merely a communications
satellite identifying a piece of space junk coming within a certain
radius-- came into contact with an actual, real, unquestionable angel,
floating in the dead, frozen vacuum of deep space. The assumption is that
the angel had been sent up against something very, very dangerous. The
assumption is that the angel had emerged victorious, as something that
powerful would certainly have threatened humanity if left unchecked. The
certainty is that the angel never made it back.


After the point at which the angel was retrieved, by whatever means this
was done, records began to be kept. Engineers at the greatest level of
confidence within Altran were secretly summoned to a highly guarded
location, to experiment on what had been found. And they did. Extensively.
The initial results held no particular meaning. The flesh was in fact
definitely alive, and was in fact definitely not any known sort of
organism, but could not be induced to heal, react, or do anything
interesting. What was left retained the power to hold itself together, but
little more, and crumbled under pressure. Volumes of data were produced
during this process. By and large, this data was never used.


In the end the only thing that could be induced to any activity whatsoever
was the brain, the last thing to be worked on in detail. And there the
last remaining spark of autistic half-life in the creature was found. The
engineers carefully cut apart the crushed skull and plowed and cajoled
their way in at the molecular level with wires and sensors and probes,
pushing past layer upon layer of brain matter that all were black and
decayed and clearly dead and liquified upon being disturbed, and took
exquisite care to preserve perfectly anything that proved an exception.
And in the end, when finally a clear outline of what bits were still
living had been formed-- a solid and almost warm block at what in a human
would have been the reptilian core, a few island-like clumps of living
matter scattered throughout, and microscopic chains of neurons that
branched off in a number of directions from that center linking it all--
the engineers connected wires everywhere that wires could be connected to
and sent out a single universal gentle, quiet electric pulse, an attempt,
in their way, to say hello.


And the angel sent a pulse back.


There was to be no communication. Too much of the brain was dead; the
angel was already gone. The engineers found they could send information in
certain ways and the brain would react, but the reaction was more
mechanical than it was thought. Merely stimulus and response.


With time, and through processes too complex to even begin to attempt to
explain here, the engineers formed a clear map of all of the angel's mind
that was left. Some fragments of problem solving, memory, visualization
remained. They could not get it to answer direct questions. What they did
discover was that it perhaps had not utterly died, but merely in some way
regressed. They discovered they could get it to respond; discovered they
could not harness the mechanics directly, but they could interact with it.
They could compose a simulacrum of thoughts and get real thoughts back.
The thoughts they could use to interact in this way were extremely simple,
and the level of interaction was roughly that one might have with a
mentally impaired child one is taking care of.


And in a small bit of sweet, strange childlikeness, the most complicated
thing they were able to get what remained of the angel's mind to do, after
year after year of attempts, was play the old Japanese game of Go. They
could form thoughts which coaxed into being in what remained of the
angel's imagination a Go board, coaxed into what remained of its
understanding the rules. They could form thoughts that described their
moves. The angel, with the distant and inexplicable glow that remained at
the center of its skull, moved in return. That was all.


The angel remains there still, eyes empty, its half-open smile of
unconditional love still uncollapsed, its burnt and blackened in places
but otherwise still almost glowing golden hair still trying to escape out
in a wizened mane, pushing out the back of the equipment, intertwined with
the hundreds or millions of metal pipes and wires, some visible, some not,
that quietly encase all that remains of the angel's brain and flow out,
back, spiraling off in thousands of different directions to the layer upon
layer of machinery that entomb the angel on every side. The outpost in
which all this is kept does not have a name, because it is not spoken of.
It is too great a secret.


One of the problems with computing, despite paradigm shifts and
advancements over time that one supposes must be literally beyond the
imagining of those who worked on the art in its early days, is that there
are certain problems that never get any easier. These problems, the
so-called NP-hard, drive computer scientists batty because they are so
universal, so basic, and yet still so inaccessible. They occur essentially
every time there is a large system of decisions in which every decision
effects the outcome of every other. Perhaps the most basic version of an
NP-hard problem is this: You have a series of arbitrary locations
connected by a series of arbitrary roads, and each road takes a specific
known amount of time to traverse. You want to know what would be the
quickest route that visits every location on your list. On a small scale,
perhaps a map on a piece of paper, this is something a human mind can
figure out with a fair degree of ease. Computer scientists are not
interested in small scales. Most of them, especially these days, are
interested in only one thing: as the scale becomes larger, how much harder
does the problem become?


And the problem with the NP-hard questions is that their complexity
increases exponentially; the amount of time it takes to solve such a
problem doubles, or more, with each added decision. We can readily handle
this doubling up to a point, but then we quickly reach something where our
ability to compute appears more and more futile with each added simple
step. Since this issue first appeared some very surprising methods of
dealing with this kind of problem, and some very surprising and ingenious
specialized devices, have been created, but still, at a certain scale, the
difficulty of that simple traveling salesman problem-- when applied to the
question of, say, how to effectively route all the messages within a
galaxy-wide telecommunications network-- becomes daunting. When it comes
to something like modeling the gravitational interactions of the particles
within a decent-sized quasar, it reaches the point where one begins to use
words like "impossibility" and seriously mean it. Advances in technology
since the day of the transistor have not helped the problem one bit. All
that we have been able to do is take the the point at which the problem
becomes unbearable and push it back a relatively infinitesimal amount;
past that point there is still nothing that can be done. It is like the
old proverb of the man who invented Chess, and when asked by the Emperor
what gift he wanted, he asked for one grain of rice for the first square,
two grains of rice for the second, four grains of rice for the third; we
can fulfill a decent portion of the chessboard easily, but just to fill
that last square we could convert every molecule in the universe to
silicon and have each crunch numbers until they all break, and still be
nowhere near to solving one of a number of problems that scientists would
like the answer to today.


Here is the truly maddening thing about the NP-hard problems: if someone,
anyone, could find one really ingenious way of solving an NP-hard
problem-- any of them-- where the difficulty with scale became more
complicated just polynomially, rather than exponentially, then they could
all be solved that way. (One of the oldest unanswered questions in
computer science is whether such a thing is possible.) That is to say,
every class of NP-hard questions corresponds perfectly with every other
class of NP-hard questions, in a sort of shadowy, behind-the-mathematics
sort of way, and you can mechanically translate between any two relatively
easily. Solve one, it happens, and you've solved them all.


This is not an exact description of what happened. It is, however,
something very similar. The essence is this: there exist homomorphisms by
which any decision can be described perfectly as a scenario in Go.


With the size of civilized space, and the extreme density of the various
markets contained therein, running a fair-sized business venture has
become a very difficult thing. There are so many things happening on every
side, so many things to keep track of, so many different ways to move, and
each interacts in so many, tiny, hard to remember ways. It is much like
Go, but there is many, many times more information, and many, many times
more decisions to make, than could be made even to fit within the 3361
possible configurations of a 19x19 board. It is more than can be kept
within the mind of a single human. It is often more than can be
coordinated within a single organization without the difficulty of
effective communication between the disparate points making everything
break down. When you get into the question of running something like one
of the corporations the size of Altran, an entity so large, varied and
powerful that there are places where it can hardly be described as
anything other than a nation-state, efficient decisionmaking begins to
seem so complex one can begin to use words like "impossible" and mean it.


But here is the thing: while decisions of these scales are beyond our
ability to solve well by any knowledge or art or technology we possess, we
do begin to find that we have the technology to, with great effectiveness,
describe the context for these decisions down to the minutest detail. We
can master the question. We just have no way to move forward into an
answer. But while we cannot answer such questions ourselves, we can
rephrase them, analyze them shallowly, shuffle observations around on
paper. And one of the things we can rephrase them as is Go.


And so there is a mindless, childlike angel at the heart of the galaxy
that eternally, joyfully, plays an endless game of Go. The damage to its
physical form has made its mind simple, simplified more than we-- not
knowing what that mind was capable of when it was at its full abilities--
can imagine. But it remains an angel's mind. Simplified though it is, it
is still infinite. It is still perfect. And it plays the perfect game of
Go. It is beyond the rules of our universe, beyond the boundaries of
finiteness, beyond the NP-complete requirement that some things just get
exponentially harder as they get more complicated. And in its mind, still
unblemished somehow, is an infinite Go board, in which a number of dancing
white stones larger than one can even really imagine are day in and day
out besieged by black stones carefully placed into the angel's mind by a
truly staggering volume of computer equipment. And through this
unimaginable amount of space, day in, day out (a mammoth frothing tangle
of white and black in a seemingly infinite glob at the center, an almost
countless number of tiny islands of war scattered out throughout infinity,
and for each a real or potential quiet chain of go attacks stretching out
toward infinity to connect them all) the white stones are always winning.
The placers of the black stones do not mind, as this is by their design.
Their intent within the game is not to win. The black stones are being
placed by a massive computer network whose purpose is known only to
precious few, a network that gathers every single decision, every bit of
information, every scenario, every question facing the Altran corporation
at that moment in time, laboriously converts the entire state of the
universe from the perspective of Altran into configurations of stones on
the Go board, and laboriously translates the angel's move back into the
answer, the move, the best possible strategic decision for that moment in
time. These homomorphisms are quite nasty, and abstraction is limited.
Describing a set of decisions that varied and that large into something as
simple as stone patterns within Go is not simple, and the amount of board
space required to describe the system compounds upon itself with each
added question that is a part of it. That does not matter. There is room.


The staggering success of the Altran Corporation has been a surprise to
very many, and it has been attributed to a number of things. The most
common belief is that their success is due to the complex, baffling, and
shifting set of unethical or semiethical anticompetitive tactics that they
undertake on a constant basis. This is partially right, but the tactics
are only a tool. The real reason for Altran's success is simply and
literally this: that at each moment, in every way, for its goals, Altran
makes literally the best possible decision it could make given the
information available to it.


There is one thing that leads me to believe the Altran Corporation's
success will not be limitless. It is that despite all the technology,
despite its perfect decisions, the ability of Altran to gather an accurate
portrait of the information describing its universe, and the ability of
Altran to model that information in an accurate way, is still imperfect
and human. The knowledge of this fact comprises my one and sole remaining
fragment of hope.

-mmc (off Kuro5hin.org)