Looking Glass

by Andrew Mayne

  Today I can sit at my computer and play with a program that mixes and matches genes like LEGO blocks, click “Send,” and have a laboratory custom make a bacteria with that gene sequence.

  This technology has already saved lives and theoretically will save far more than it kills with its bad applications—at least we hope so. The genie was let out of the bottle when a nineteenth-century monk started playing with pea pods; trying to limit the tools or the flow of information is only going to make the good guys less capable and informed.

  Probably.

  But along with all this interest in the military applications for engineered bacteria has come a number of nonlethal ways to use germs. One of them is in a project I was tangentially involved with as a consultant.

  Suppose you think Terrorist A is part of a terror network, but you have no idea if he knows Terrorist C in another country. If they’re smart enough to not use electronic communication that links them and always use an intermediary—Terrorist B—how do you connect A to C without waterboarding Terrorist B or even letting him know you’re onto him?

  You could follow everyone who meets with A, then follow everyone who meets with them and hopefully find a short connection to Terrorist C, but the problem is you quickly run out of agents to follow all the possible intermediaries. As the human-connection tree branches out again and again, you soon find that there aren’t enough people on the planet to manage the job.

  This was one of the problems the intelligence community was dealing with: the sheer cost, and ultimately the impossibility, of employing enough people to follow other people.

  One solution, and the project I was brought in to consult on, was to use not people but bacteria.

  A wealthy superpower with a first-rate science program could modify a benign bacteria—the kind that’s already in your glass of water—with a special tag that would allow you to differentiate it from its brothers and sisters.

  You could then spray this on Terrorist A as he passes you in the street. As the bacteria multiply in his mouth and nasal passages, he’ll spread it for several days before his immune system kills it off.

  It sounds like a potential nightmare scenario in the making.

  Well, it isn’t and it is.

  The problem with the San Francisco experiment—beyond the ethical ones—was that they didn’t have access to designer bacteria. They had to use strains that were presumed harmless but still rare enough that they could attribute their spread to their testing.

  Using a common benign bacteria wouldn’t have worked, because it would already be crawling all over everyone in San Francisco.

  Whereas now we can take a friendly strain like Neisseria lactamica—and not a turncoat germ like Streptococcus pneumoniae, which can too readily mutate into something harmful—add some markers to the genome, and wait to see if our special strain shows up on Terrorist C by stealing his Starbucks coffee cups and swabbing anything else he touches.

  Through the miracle of adaptation, we can even get a rough estimate of how many people our bacterial spy had to go through to reach Terrorist C, telling us the “distance” between them. And then using some software I helped create, we can estimate the size of the network by looking for the number of mutations we find on Terrorists D and E.

  This sounds scary to some people. To me, it’s not. The statistical likelihood of a mutated dangerous strain of Neisseria lactamica being unleashed in a project like this is about one-trillionth the likelihood that you’ll launch a civilization-destroying superflu the next time you sneeze Streptococcus pneumoniae in a crowded McDonald’s.

  What is scary to me is that a friend of mine told me that, while the hunt for Osama bin Laden was still ongoing, there were serious talks about engineering a particular strain of flu virus that would target only him. Researchers were about 50 percent certain they could do this with a several-billion-dollar budget. But they had to explain to officials the other 50 percent probability, which was that a slight mutation could wipe out the entire bin Laden family—including the benevolent ones. And an even bigger mutation, actually just a defect in the part that looked for his specific gene markers, would stand a high probability of wiping out 90 percent of all primates on the planet—including Homo sapiens.

  The project was abandoned. Or so I’m told.

  Fast-forward to today: the intelligence community now has access to specifically engineered bacteria and viruses that can perform a number of nonlethal tasks, if you know what to ask for. They don’t name these germs for their true intent. Instead they’re given boring number sequences and categorized for uses that are supposed to be limited to the petri dish.

  I special ordered five separate strains of Neisseria lactamica with a couple of modifications not in the books. Since I don’t have access to high-end laboratory equipment or even a nifty scanner with custom assaying chips, I have to be able to spot my strains in the field. That’s why they have a special gene that will make them glow pink under an ultraviolet light when sprayed with a protein compound.

  So after I go place my spy cams at each botanica, I’m going to enter each store and douse every bottle of Snake Bite I can find, covering them with my microbial bloodhounds.

  Call it revenge for the headache I still have a day later.

  CHAPTER FORTY-SIX

  BIAS

  The hard part wasn’t putting the little spy cams near the aluminum door frames of the botanicas or even spraying their inventory of Snake Bite as I followed the delivery truck while it made its rounds from shop to shop. The challenging part was sorting through the thousands of images gathered by the cameras and narrowing them down to the 323 people that all walked through their doors over a two-day period.

  My camera placement on the second day at Blessed Angel Spiritual Wonders was a little off, and the only usable images were the ones captured in the reflection of the door as people entered and left.

  Out of all of them, I have a feeling one could be the Toy Man, which is encouraging and frustrating because the image is so poor. Compounding that is the awareness of my own bias. I’ve already cast someone in the role of the Toy Man in my head, and this man meets that description. Is it because I think he matches what Artice told me? Or because he’s my idea of a scary black man?

  I could second-guess my innate prejudices all day long, but in the end, it’s never a substitute for an expert opinion. In this case, the opinion of Artice himself.

  I’d scheduled to have Artice use the jail’s video-calling service to speak with me in case he remembered anything else that might be important—also to let him know I was actively pursuing this and I hadn’t forgotten him.

  As I wait for the conference software to tell me he’s calling, the waning moon is visible through my hotel window, reminding me there are only four days until the Toy Man kills again, if he’s keeping with his schedule.

  While another killer might hold off while there’s an investigation under way, I suspect that the Toy Man won’t … for several reasons. The first is that police are looking for him on the wrong side of the country and searching a house he abandoned years ago. They haven’t even found his other Los Angeles killing field. The second is that he’s arrogant. Like other highly intelligent killers, he wants to believe that his actions are invisible and that he can’t get caught. Were he to stop, that would be an admission to himself that the cops could be at least as smart as him. The last reason I think he’s going to kill again on the new moon is because he believes in magic. This is the most powerful time of the lunar cycle for him to conduct his blood ritual.

  My computer chimes, and Artice’s somber face appears. He’s not the unflappable young man I saw before.

  “Artice?”

  “Yo, Theo.” He glances off to the side. Usually there’s a guard in the room when they do these sessions, so he’s probably just reminding me.

  “Are you okay?” I ask.

  “Yeah … um … I had a visit earlier from a Detective Chang …”

  “Chen?�
��

  “Yeah, her and some dude prosecutor I hadn’t spoke to before. They showed me a photo and asked if it was the Toy Man.”

  “Was it?”

  Artice lowers his head and cups the phone. “Between you and me? No. I never saw that guy before. But they really, really wanted me to say yes.”

  “Did they say a name?”

  “Yeah, something Sims.”

  Clearly they’re putting pressure on their only witness to name the late Brazilian hit man Ordavo Sims. But how far are they willing to go?

  “They pressured me hard, man. All I had to do was go before a grand jury and say that was the guy. No trial or anything.”

  That’s so wrong, dangling something like that in front of Artice. I try to keep my anger to myself.

  “But I said I wasn’t sure. Then they told me that I was just a kid back then and everything is different when you’re young and memories can get all messed up.”

  “How did that go over?”

  “I told them to go fuck themselves. I already went through that shit before with people telling me I made up the Toy Man. Now they think he’s real, but they’re going to tell me who he is and who he isn’t? Fuck them. I don’t care what they put in front of me or what promise they make. If that ain’t the Toy Man, then it ain’t him, you know? I’m not gonna let them go easy on me knowing he’s out there hurting some other kid. How can a decent person live with that?”

  Indeed. Even a guy like Artice, who’s caused his share of mayhem, sees that. “So now what?”

  “They said they were going to let me think about it and come back tomorrow.”

  “Did they offer you anything specific as far as a deal?”

  “No. But I got the feeling the prosecutor guy was willing to talk to the judge on my case and maybe do something. Who knows. I don’t trust them. Anyway, what you got?”

  “I have some photos, too. Unfortunately, I can’t do much to help your case.”

  Artice nods his head. “I know that. But if it goes down like the last time you went after one of these motherfuckers, then that’s cool with me.”

  I cough. “Uh, I’m hoping for a less physical resolution.”

  I take the photos from my folder with the backs to me. I’ve gone to the extra effort shuffling them so I don’t know which one is which. I want Artice to give me his own opinion and not one influenced by my body language. We all think we’re good at hiding that, but chances are, the better you think you are, the worse you actually are.

  “Okay, I’m going to go through these one by one. When you see one that may be him, let me know and I’ll set it aside. I’ll warn you, some of these are pretty sketchy images.”

  “Got it.” He moves closer to the screen.

  “Nope. Nope. Nope. Nope.” He’s not even hesitating.

  “Do you need more time?” I ask.

  “Are your hands getting tired?”

  “Sorry. How about this one?”

  “Nope. Nope. Nope. Nope. Nope …” Artice pauses.

  I’d been unconsciously looking away from the screen, trying to avoid seeing a reflection of the image.

  “Artice?” I lower the photograph.

  He’s almost hypnotized. His eyes are wide and his mouth half-open.

  “Artice?”

  “That’s him.”

  “Let me set this aside, just in case, and go through the rest,” I reply.

  “You don’t need to.” He’s shaking his head slowly. “That’s him.”

  “Just to be sure?”

  He gives me a dutiful nod. “Fine, but that’s the motherfucker right there.”

  I show him the rest of the photos, getting a series of impatient noes from him. Finally we reach the last one.

  “Nope. It was that first one.”

  I turn the photo over.

  It’s the same one I’d pulled from the stack myself. I show the image to him again. “Are you sure it’s him?”

  Artist stares directly into the camera, bypassing my image on his screen so his eyes are looking into mine. There’s a coldness and a hurt in his expression. He’s been doubted for so long and now I’m questioning him.

  “The asshole even has his same white Cadillac,” Artice replies.

  “Cadillac?” I reply, flipping the image over.

  This was one of the photos caught in the reflection of the door. Sure enough, when I look at the upper corner of the image, there’s the unmistakable grille and front windshield of a white Cadillac.

  The nagging question in my mind is whether I was drawn to the photo because of the expression on the man’s face and Artice’s graffiti, or was it the fact that my animal brain, my early warning system that listens for noises and looks for the signs of predators, saw the Cadillac but never communicated it to me consciously.

  “Hey, Dr. Cray?”

  I look to the webcam. “Yeah?”

  “You got a name for him?”

  I shake my head. “It’s tricky. I don’t want him knowing I’m after him. If the store owner knows him, I don’t want him being tipped off.”

  “Okay … so what’s your plan? Send an anonymous tip to the police?” Artice says sarcastically.

  “No. Not that.” I’m searching the image for more details. Unfortunately the license plate is blocked by a bush. However, in the upper portion of the image, I can see something on the dashboard. It’s a parking receipt of some kind.

  “I’m going to try to find out what that is.” I hold the photo up for Artice to see.

  “You better have some Blade Runner–level shit if you’re going to see that. From here, it’s just a blur.”

  I glance over at the SD card the image came from and start thinking about the algorithm used to create the video file of the image and the one I’ll have to create to produce a clear photograph.

  “Wavelets, Artice. Wavelets.”

  CHAPTER FORTY-SEVEN

  FOURIER

  In the early nineteenth century, French mathematician and physicist Jean-Baptiste-Joseph Fourier, who had previously accompanied Napoleon on his expedition to Egypt, became fascinated by the concept of heat transfer and how two objects exchanged energy. Why didn’t one give up all of its heat to the other or vice versa? This led to a great number of other questions.

  One was trying to solve the riddle of why the earth wasn’t a giant snowball. When Fourier calculated the distance from the sun to the earth’s surface, he realized there just wasn’t enough energy hitting the ground to keep us warm. This led to the discovery of the role of the planet’s atmosphere and water vapor in moderating Earth’s temperature, and ultimately to the discovery of the greenhouse effect.

  But it was Fourier’s work describing the mathematical functions of energy transfer that gave way to the Fourier transform. Roughly summarized, it meant using arithmetic to reconstruct a larger signal by looking at only a smaller portion of it.

  Fourier transforms became the basis for computer compression, and they are the reason I can fit all of those surveillance images and videos on a tiny SD card. The processor doesn’t have to write everything to the memory chip—just enough that I can get a usable image.

  The problem with this kind of compression is that it loses information. Assuming even the tiny lens on my spy cam was able to refract a clear image of the Cadillac dashboard to the even smaller photo-sensor array, and that they in turn had the resolution to resolve what was on that slip of paper, by the time the processor squeezed that image down, any helpful data might have been lost.

  However, it was because of the lossy nature of Fourier transforms that mathematicians started looking at other compression and reconstruction techniques. Wavelet compression was based on the idea of using the whole wave of a signal and creating a lossless version of it by getting the precise function that created it. Although it’s much more processor intensive than a Fourier transform, it uses memory much more effectively.

  Sadly, my spy cams use the same Fourier-based algorithms behind lossy compression, an
d there’s only so much data I can pull from the photo.

  But on the other hand, the same math behind wavelet theory can also be used to reconstruct a signal across time.

  Photo software that can pull a clear image from a blurry photograph caused by camera shake works by calculating how long the shutter was open and measuring the amount of movement. By treating the blurs like paint strokes, it can essentially roll back time and figure out what the top of the brush looked like—or the shape of your eye.

  When our suspect walked through the door and triggered the motion detector on the spy cam, it recorded about four seconds of footage. Because I was using the motion JPEG compression method for video at fifteen frames per second, this means my little camera caught sixty photographs of his face and the front of his car.

  Because the door was moving, it caught his face from several angles, like a scanner used to make a 3-D image of his head.

  I’ve already got a 3-D model from those images using off-the-shelf software.

  While getting shape data was easy because his head was close enough to the glass, trying to get a clearer image of what’s on the Caddy’s dashboard is more challenging, because it’s a two-dimensional object photographed from sixty slightly different angles.

  But all is not lost. Part of the magic of wavelet transformation magic is that I can input certain known factors that give the software more information than is apparent in the image.

  While the grille of a Cadillac is just a rectangle to the algorithm, I know precisely how many centimeters across it is and can accurately estimate how far away from the door and the lens it was as the image was taken.

  By stabilizing the little paper square in 3-D space, I can overlay all the other images, adjust for specular changes, make estimates of refractivity of the paper, and even use a little artificial intelligence to build a best guess of what certain shapes are in the image.

  After five hours I have a copy in my hand of what’s on the dashboard—well, almost. I can make out what is supposed to be a bar code and might even be able to reconstruct that if I had other sample bar codes and could figure out what the blurry streaks might be, but the important part is the logo, which is quite clear and unambiguous.