Home' Technology Review : November December 2007 Contents 64 FICTION
TECHNOLOGY REVIEW /
A month or two. A part of Lincoln recoiled, but
another part thought: that wouldn t be so bad. A break
from the farm, doing something di erent. Meeting new
people, learning new skills, working with animals.
Rats, most likely.
Steve Hasluck had been part of a team of scien-
tists developing a new kind of medical nanomachine,
re ning the tiny surgical instruments so they could
make decisions of their own, on the spot. Steve s team
had developed an e cient way of sharing computing
power across a whole swar m, allowing them to run
large, complex programs known as "expert systems"
that codi ed decades of biological and clinical knowl-
edge into pragmatic lists of rules. The nanomachines
didn t really "know" anything, but they could churn
through a very long list of "If A and B, there s an 80
percent chance of C" at blistering speed, and a good
list gave them a good chance of cutting a lot of dis-
eases o short.
Then Steve found out that he had cancer, and that
his particular kind wasn t covered by anyone s list
of r ules.
He took a batch of the nanomachines and injected
them into a roomful of caged rats, along with samples
of his tumor. The nanomachines could swarm all over
the tumor cells, monitoring their actions constantly.
The polymer radio antennas they built beneath the rats
skin let them share their observations and hunches
from host to host, like their own high-speed wireless
Internet, and report their ndings back to Steve him-
self. With that much information being gathered, how
hard could it be to understand the problem and x it?
But Steve and his colleagues couldn t make sense of
the data. Steve got sicker, and all the gigabytes pour-
ing out of the rats remained as useless as ever.
Steve tried putting new software into the swarms.
If nobody knew how to cure his disease, why not
let the swarms work it out? He gave them access to
vast clinical databases and told them to extract their
own r ules. When the cure still failed to appear, he
bolted on more software, including expert systems
seeded with basic knowledge of chemistry and phys-
ics. From this starting point, the swarms worked
out things about cell membranes and protein fold-
ing that no one had ever realized before, but none
of it helped Steve.
Steve decided that the swarms still had too narrow
a view. He gave them a general-purpose knowledge
acquisition engine and let them drink at will from the
entire Web. To guide their browsing and their self-
re nement, he gave them two clear goals. The rst
was to do no har m to their hosts. The second was to
nd a way to save his life or, failing that, to bring him
back from the dead.
That last rider might not have been entirely crazy,
because Steve had arranged to have his body pre-
ser ved in liquid nitrogen. If that had happened, maybe
the Stevelets would have spent the next 30 years ferry-
ing memories out of his frozen brain. Unfortunately,
Steve s car hit a tree at high speed just outside of Aus-
tin, TX, and his brain ended up as ambé.
This made the news, and the Stevelets were watch-
ing. Between their lessons from the Web and whatever
instincts their creator had given them, they gured
out that they were now likely to be incinerated them-
selves. That wouldn t have mattered to them if not
for the fact that they d decided the game wasn t over.
There d been nothing about resurrecting charred
esh in the online medical journals, but the Web
embraced a wider range of opinions. The swarms
had read the sites of various groups convinced that
self-modifying software could nd ways to make itself
smarter, and then smarter again, until nothing was
beyond its reach. Resur recting the dead was right
there on every bullet-pointed menu of miracles.
The Stevelets knew that they couldn t achieve any-
thing as a plume of smoke wafting out of a rat cremato-
rium, so the rst thing they engineered was a breakout.
From the cages, from the building, from the city. The
original nanomachines couldn t replicate themselves,
and could be destroyed in an instant by a simple chemi-
cal trigger, but somewhere in the sewers or the elds or
the silos, they had inspected and dissected each other
to the point that they were able to reproduce. They took
the opportunity to alter some old traits: the new gen-
eration of Stevelets lacked the suicide switch, and they
resisted external meddling with their software.
They might have vanished into the woods to build
scarecrow Steves out of sticks and leaves, but their
software roots gave their task rigor, of a kind. From
the Net they had taken ten thousand crazy ideas about
the world, and though they lacked the sense to see that
they were crazy, they couldn t simply take anything on
faith, either. They had to test these claims, one by one,
as they groped their way toward Stevescence. And
while the Web had suggested that with their power
to self-modify they could achieve anything, they found
that in reality there were countless crucial tasks that
remained beyond their abilities. Even with the aid of
dexterous mutant rats, Steveware Version 2 was never
going to reëngineer the fabric of space-time, or res-
urrect Steve in a virtual world.
Within months of their escape, it must have become
clear to them that some hurdles could be jumped only
Links Archive September October 2007 January February 2008 Navigation Previous Page Next Page