29 May 2009

"Vampire Robot"

VAMPIRE ROBOT
By Curtis C. Chen

I know what you're thinking: Why would anyone build a vampire robot?

First of all, let's clear up some misconceptions. She's not a vampire. She doesn't drink blood. The artery in the neck is just the quickest way to get her probe tendrils into the brain--two probes, one for each hemisphere, so the scars look like what everyone imagines vampire bite marks would--although, if vampires actually existed, they'd probably just tear out your throat instead of making a couple of dainty little punctures.

That's another big difference right there. To a vampire, you'd just be food. But to Kayla, you're a research subject, a precious resource. That's why she doesn't kill anyone, why she doesn't hurt anyone, and why she does her best to take her measurements without disturbing their sleep. That's why she sneaks around at night. She wants data. No life, no brain activity, no data.

Yes, I admit, the original experiment has gone a little off track. She wasn't supposed to leave the building yet, but she's smart. We built her to be intelligent, adaptable--you might even say adventurous.

I'm not anthropomorphizing; we really did give her a personality. Do you know what goes into creating an AI? It's not linear programming like you learned in the sixth grade, or those clunky neural networks in self-drive cars. It takes months to train a neural net, and you still need to write explicit subroutines to handle any unexpected shapes or movement.

We wanted Kayla to seek out those exceptions. The Fed planned to send her into dangerous and hostile environments, and they didn't want their multi-billion-dollar investment to get taken down by some street urchin with a slingshot. We had to give her the ability to deal with anything.

I can't tell you how we did it. You wouldn't understand the math, and it won't help you stop her. Don't you get it? We don't know what happened. We don't know how she escaped, or why she decided to try in the first place. She's exceeded our original designs. She's alive.

And she's not some lumbering giant with no sense of purpose. Kayla knows exactly what she's doing, even if we don't. She knows how to hide in the darkness, emerging just long enough to perform--what did you call it? "Surgical strikes." Quick, clean, precise.

Now, we do need to find her, but not to kill her. If we get close enough, we can activate her wireless downlink and grab a complete copy of her current run-state. Then we can decompile that code and see exactly what she's thinking.

We've analyzed her activity patterns, and we can make some educated guesses about her next targets. But if we do this, we do it our way. We won't let you destroy her. She's not hurting anyone. She just wants to learn. Just like us.

Besides, we don't want to wake the nest. I mean, who knows how many more of those defensive bat-bots she's constructed by now?

EOF

No comments: