Drones, Hellfire missiles, and sniper rifles may give you the illusion of supremacy and easy victory, but it’s only when you’re able to look the target in the eye that assassination really stands a chance of working.
The face of modern assassination that Washington would like us to see is one of cool, flawless, empirical efficiency. Like some all-seeing space-age pterodactyls, drones prowl America’s battlefields, relentlessly hunting down its enemies. We’re all but promised that nothing eludes a drone’s camera or its missiles. Which means that anyone with an X on his back and stupid enough to show his face in the open isn’t long for this world. It may sound like cruel geometry, but the point is we’re told that drones unerringly cull out the guilty from the innocent.
An official involved in drone assassinations in Iraq told me that a drone’s TV guidance system is accurate enough to drop a one-hundred-pound Hellfire missile inside a target as small as a medium-size pizza. It doesn’t matter whether the target’s moving or hiding behind a rock. It’s a big plus that a drone camera’s resolution is high enough to make out a man’s face from ten thousand feet above.
“The fucker won’t look up until he hears the incoming missile’s buzz,” my ex-colleague said, “and then you see this horror on his face as he realizes what’s about to happen. The last you see of him is a bright flash.”
It was thanks to drones that the U.S. finally ran to ground the Jordanian-born head of al-Qaeda in Iraq, Abu Musab al-Zarqawi. As the official told me, the initial break came when the man’s mentor, a radical Sunni cleric, was identified. From there it was as simple as keeping an eye on him, waiting for Zarqawi to pay him a visit. Which he couldn’t resist doing.
It did Zarqawi no good that he understood the threat of drones, and did everything he could to stay out of their sight. He moved locations on the hour, slept in a different house every night, and continually switched out cell phones, even then limiting his calls to a minute or two. In the end, he simply couldn’t beat the drone’s eye.
“I had working for me this woman named Margaret,” the official told me. “She lived and breathed Zarqawi. She spent eighteen hours a day glued to four flat-panel screens, telling the drone pilots what angle she needed, how long she needed it over a target. You should’ve seen her eureka moment when she finally put Zarqawi in our crosshairs.”
The honor of vaporizing Zarqawi fell to an F-16 because a bomb bigger than a Hellfire missile was needed to destroy the building he was hiding in. But the consensus soon became that it’s more efficient to arm a drone with a missile and cut down the time between lock-on and launch. It was a case of dumping the middlemen—the F-16 and its pilot.
Another advantage of drones over jets is that they’re able to hover over a target for days at a time. It would take a serried array of jets constantly up in the air to replicate them, burning up hundreds of thousands of gallons of fuel and running the Air Force into the ground. Drones are also more versatile, allowing an operator to easily make a last-second change to a missile’s course, if need be, diverting it in mid-flight and exploding it harmlessly in a field. Drones also are easy to maintain: They’re powered by high-performance snowmobile engines.
But what really sparked the love affair between Washington and drones was a promise that new sophisticated data analytics would bring near divine infallibility to political murder. It was a brilliant marriage between advanced algorithms, unlimited computing power, intercepts, and drone camera feeds that brought a speed and precision to assassination never before seen in the history of man. Instantaneous justice at the click of a mouse.
It was all the more seductive because data analytics had already started to drive everything that touches our daily lives, from managing electrical grids to long-term weather forecasting. As one computer executive told me, it’s now possible to run an algorithm through Twitter feeds commenting on movie trailers and then be able to tell a studio how many days it should keep a new release in theaters. Or whether it should let the movie go right to DVD. If data analytics are able to predict our tastes, why can’t it sort out the good guys from the bad guys?
Data analytics also took a lot of the moral hazard out of political murder. If IBM’s Deep Blue can outthink the world’s best chess players, it certainly can distinguish between right and wrong, who’s been naughty and who’s been nice. And woe to anyone foolish enough to suggest justice can’t be left to a computer, a modern-day deus ex machine; he’ll instantly be dismissed as a Luddite.
How far off is the day when killer drones will “own” America’s “battle spaces”? Or, as the NSA analysts colorfully put it, the NSA is now in the business of putting “warheads on foreheads.” Fighting war won’t be any harder than performing a Chopin sonata on a player piano—plug and play. Even better, no more of our girls and boys storming the shores of Tripoli; and no more Vietnams, Iraqs, or Afghanistans either.
Washington’s faith in computers goes back to American cryptographers’ informing everyone too late that the Japanese navy was about to attack Pearl Harbor. Since then, there’s been an unshakable belief that a better handling of chatter won’t let it happen again. Suck everything out of the air that’s to be sucked, instantly process it, and we’ll be as safe as a baby with a mother’s love.
Of course, 9/11 took a lot of wind out of that sail, but America’s intelligence agencies doubled down anyway. The National Security Agency turned its Big Ears on Afghanistan and Pakistan, scooping up trillions upon trillions of intercepted phone calls, e-mails, electronic money transfers, airline manifests, Facebook and Twitter feeds. Sophisticated new software allowed it to churn through the zigabytes upon zigabytes of intercepted data with lightning speed. NSA’s new, advanced algorithms ensured that the smallest speck of gold didn’t slip through its net.
If a prepaid cell phone in Karachi talks to a prepaid one in Peshawar, a digital red flag shoots up. If the Peshawar prepaid cell phone then calls a number in North Waziristan connected to the Taliban, another flag shoots up. From there, an algorithm brings divine clarity to it all: Innocent or guilty? Live or die?
The NSA hasn’t been too shy to whisper around Washington that its targeting is only getting better with time and money. Things like “self-learning algorithms” and “evolving databases” have mammothly improved the scope and accuracy of its software, instantly shaving off the random and the inconsequential. The implicit promise is that there’ll come a time when the NSA will be able to instantly and finely process billions upon billions of telephone calls and never introduce a human being into the process . . . until it comes time to pull a drone’s trigger. How far away can it be before the NSA’s last analyst is shown the door?
The same is true for monitoring the Web. At first, the NSA was able to read only e-mail “packets”—the so-called meta- or routing data. But now it can get into the “payload”—NSA’s shorthand for the text of a message. It’s also able to scan broadband Skype calls. With virtually free “cloud” storage, everything we ever write on a computer or talk about over a phone is somewhere out there in the ether, waiting for the NSA to scoop up and download into its giant “data farm” in Utah. Who knows what bit of random information will help call in a drone strike?
Similar advancements can be found in related fields, such as video forensics. By extracting digitally embedded data from images, it’s possible for a video sleuth to tell the temperature and humidity of a room where a DVD’s been filmed—whether it’s been shot in humid Karachi, for example, or the dry mountains of North Waziristan.
I’ve never myself seen it, but there’s apparently been an eerie science-fiction twist to it all, what’s called “data visualization.” The way I imagined it is that as algorithms pile suspicion on suspicion, a promising candidate for a kill list will go from a cold cobalt-blue dot to a canary-yellow one and then, if he’s particularly unlucky, to a bright burning-red one. Since, I’ve been told we’re not quite there with the color scheme. But the point is that while a human being may still decide whether to pull a drone’s trigger or not, it’s a computer doing the thinking—all of it—right before our very eyes.
The implicit promise with these fully automated killing machines is close to the Philip K. Dick classic short story “The Minority Report.” Only rather than predict future crimes it will mete out instant and terminal punishment. Is this what the end of history looks like, killer robots flying the skies and rendering justice?
John Brennan, President Obama’s first-term drone commander and now CIA director, has claimed that drones have broken al-Qaeda’s back. What’s left of its networks exists underground, completely neutered. And it’s all been accomplished with divine-like precision. Brennan: “There has never been a single collateral death because of the exceptional precision of the capabilities we’ve been able to develop.”
The machine victorious, political murder fully automated and scrubbed clean.
True, blasting al-Qaeda out of its lairs in Pakistan, Yemen, and every other rat hole they’ve scurried down into has all worked out nicely for us: no lost airplanes, no dead pilots. But in the headlong rush to automate political murder, we’ve forgotten one essential element of assassination: The more intimate the act, the more unambiguous the message.
Drone assassinations are like—I don’t know—phone sex. They solve the immediate problem, but they leave you unsure of what exactly you got out of it and hungry for more. Whether we’re talking about the little death or the big one, the act is only complete when you’re able to look your partner in the eye.
A couple of days before Hariri’s assassination, the head of Hezbollah, Hassan Nasrallah, and Hariri sat down for tea together. Tea, for God’s sake! We’ll never know, of course, whether Nasrallah was aware of the plans to assassinate Hariri. But if he was, I’d guess that rather than size Hariri for a coffin, Nasrallah was giving himself a last chance to decide if murder was the only recourse. Whether Hariri could have said anything that would have stayed his execution, we also don’t know. But we do know that Nasrallah and Hajj Radwan knew Hariri as well as they knew their own—all his failures and successes, the ins and outs of his crooked business deals, his acts of selfless charity, his brilliance and his stubborn and pointless grudges. They’d finely calculated what Hariri alive cost them, and how they’d be better off after his passing. In the end, it was murder, sure, but it was also a deeply personal affair, closer to the divorce from hell than letting an algorithm pick your next movie.
The other great assassinations in history were also close-up, personal affairs. Supposed friends and colleagues murdered Julius Caesar with daggers. Archduke Ferdinand’s assassin pulled the trigger from only five feet away. One plan to murder Hitler was for the assassin to wear an explosive and all but embrace the psychopath before blowing himself and his Führer to kingdom come.
And now that I’ve brought up that iconic attempt, isn’t assassination something of a twisted Freudian act of love? The true assassin, à la Charlotte Corday or Colonel Claus von Stauffenberg, is prepared to freely and gladly sacrifice himself in order to save society, to completely submerge his ego into the common good. In his attack on two American embassies and the Marines, Hajj Radwan employed suicide bombers to remove any doubt about the sacrifice, dedication, and intractability on his side. And by the way, paying the ultimate price has nothing to do with Islam per se. Japanese kamikaze pilots and Tamil suicide bombers also took their own lives for what they believed was a higher good.
Drones, algorithms, and Cray computers may have promised us Cartesian clarity, but in eliminating the human factor, they’ve greatly diminished the primeval force of political murder. But at bottom, isn’t it all beguilingly too easy, like shooting buffalo from a train?
—
Drone assassinations got off to a rocky start. The first attempt occurred on February 4, 2002, when the press reported that a drone flying over Afghanistan fired a missile at a man who appeared to be Osama bin Laden. Like bin Laden, he was unusually tall, and the two men with him were showing clear deference to him.
As soon as the Pentagon acknowledged the strike, reporters descended on the site. But rather than bin Laden, they found the bodies of three Afghans who’d been rummaging around for scrap metal. The tall man turned out to be only five-eleven, not bin Laden’s six-four. Bin Laden apparently had been nowhere near the site of the missile strike.
The botched attempt on bin Laden was noteworthy because it was the first and only time the American government publicly acknowledged the specifics of a drone assassination attempt. But lesson learned: No more details for the untutored.
The first apparently successful drone assassination came nine months later, in November 2002. A Predator drone flying over Yemen killed a suspected al-Qaeda member and five other passengers riding in a car. Among them was an American citizen. The attack was widely reported in the press, but there was no official confirmation other than whisperings here and there that the U.S. had gotten its man. But by no means did it mean that drones were now foolproof.
Among the many missed strikes, a drone attempt on bin Laden’s deputy Zawahiri at a place called Damadola, Pakistan, ended up killing eighteen local Pashtuns rather than Zawahiri. A Predator drone killed two Marines in Afghanistan, mistaking them for Taliban. Innocent Bedouin on the move in Yemen were cut down by a drone. One botched attack kept following another.
And then reports started coming out of Afghanistan’s and Pakistan’s tribal regions that suggested there were an alarming number of drone strikes widely off target. Wedding parties and houses full of children were being hit instead of al-Qaeda. Even reportedly accurate strikes were way off the mark.
Pakistani photographer Noor Behram had already pretty much decided on his own that politics was driving the strikes rather than anything like mathematical precision. He knew he couldn’t trust the wild rumors filtering out of the tribal areas or, for that matter, the official statements out of Washington and Islamabad. So he went up into the tribal areas to see for himself and photograph the results.
Behram started to show up at drone attack sites in 2008, doing his best to get to them before the bodies were dragged out and the sites were cleaned up. Before Behram could start taking pictures, he’d often get on his hands and knees to help dig out the survivors. It wasn’t risk-free work: Often there would be follow-up drone strikes at the same site. The locals admired his courage and welcomed him at the sites.
After three years of gruesome work, Behram claimed to have the proof on film that more civilians were being killed than the United States and Pakistan admitted. His pictures depicting dead children and old people seem to bear him out. (Custom in the tribal area will not let him take pictures of dead and injured women.)
“For every ten to fifteen people killed, maybe they get one militant,” Behram said. “I don’t go to count how many Taliban are killed. I go to count how many children, women, innocent people are killed.”
A lot of other stuff bubbling up in the press made one wonder exactly who it is we’re killing with the drones. In February 2011, for instance, The Washington Post reported that 581 “militants” had been killed the previous year by drones in Pakistan. But only two were on any “most wanted” list. Nine months later the Post reported that in the previous three months, sixty people had been killed in fourteen drone attacks. But the White House would put a name to only one victim, who, by the way, no one I knew in the intelligence world had ever heard of.
It’s hard to come up with a good explanation why the American public isn’t offered the identities of these people. The locals presumably comb through the rubble and charred cars and know precisely who’s been killed. Is the reason Washington can’t tell us is that it doesn’t know? Or does Washington know drones are killing civilians and doesn’t want us to know?
What Washington always seems to miss is just how tricky eavesdropping can be. When it’s good, it’s very good; I’d go so far as to say there’s no better firsthand intelligence. It’s just a fact that when two people are on a phone call foolishly expecting privacy, they’ll say things never meant to be said in the open or to a stranger. It’s even better when they’re angry and blurt out some hideous truth.
But for the unwary, chatter is a treacherous snare. Often what’s plucked out of the ether is two people speculating about something they know absolutely nothing about, like whether it’s going to rain the following week or not. Having myself been caught in a chatter trap, I have a good idea about the pitfalls.
One day something I took as unimpeachable chatter put Hajj Radwan in a certain hotel in Paris on a certain day. We even had a room number for him. We persuaded the French police to do the needful and arrest him. But when French commandos fast-roped down the side of the hotel and crashed through the windows, they interrupted a Spanish family taking an afternoon snack instead of Hajj Radwan. To this day I don’t know what went wrong. Did the French betray us? Or was it because we’d misread the intelligence?
One reason Washington fell in love with chatter is that it takes the hard work out of intelligence, letting it turn off its brain and abdicate all human agency and judgment. It’s all the worse these days because Washington is politically and financially invested to the hilt in drones, data analytics, and the fantasy there is such a thing as mechanical justice. Never mind that a lot of innocent people are being slaughtered.
According to a lot of accounts I’ve recently been reading in the newspapers, Washington doesn’t seem to care much that al-Qaeda has started to catch on and do things like constantly switch out phones. They’ve even taken to throwing their cell phone chips into a bag, shaking them up, and then redistributing them so the U.S. has no idea who is who. It’s left it with “signature” or “pattern of life” strikes. They’re based on nothing more sophisticated than a drone catching three young men in a field doing exercises, their AKs by their sides. If the three repeat it the next day and in the days after—giving the semblance of organized training—they’ve all but signed their own death warrants. Who cares that the condemned are without face or name.
The Darwin effect also seems to have kicked in. Smart al-Qaeda types now stick to Karachi and Peshawar, urban areas too densely packed to employ a drone missile. Even worse, many have stopped talking on telephones altogether. As one military intelligence officer responsible for drone targeting told me, in the years since drone strikes began in earnest, the leadership of our main enemy in Afghanistan—the so-called Haqqani network—only “spiked” once on cell phones. They’ve otherwise turned themselves into phantoms beyond the reach of our technology. Unconnected but alive.
And finally even our best software apparently sucks. In a little-noticed court case, two opposing software companies presented evidence in court that drone target–acquisition software was sending drone missiles off course by as much as thirteen meters (forty-three feet). In a built-up area that’s well within the kill radius of a lot of innocent people.
If there’s a moral to the story, I suppose it’s that the chances of automating assassination are about equal with the Plenty of Fish website finding us the perfect soul mate. But what do I know.
The more I heard about things such as data visualization and self-learning algorithms, the more it sounded like a bad science-fiction version of the Spanish Inquisition. I needed someone smart to show me where I was wrong.
I managed to run down the noted physicist Tsutomu Shimomura, who’s an expert in the mathematics of link analysis—the same algorithms and data visualization reportedly use to draw up kill lists. Among other credentials, Tsutomu studied under Richard Feynman at Caltech. After university, he went to work at Los Alamos. By the way, brains apparently run in the family: Tsutomu’s father won a Nobel Prize in chemistry.
Tsutomu earned a measure of fame when he helped track down the notorious hacker Kevin Mitnick. Mitnick had eluded the FBI for years by, among other things, cloning cell phones. But by comparing relevant metadata, Tsutomu was able to pinpoint Mitnick to his North Carolina apartment, where the FBI arrested him. Mitnick did five years in jail. The movie Track Down is based on the story.
I meet Tsutomu at a Berkeley organic restaurant where there’s rhubarb juice on the menu but no Coke or Pepsi. It’s unseasonably chilly, but Tsutomu is in shorts, hiking boots, and a T-shirt with an astral constellation print. He orders cauliflower soup.
I get right to it: Does data analytics work? Or are drones killing the wrong people?
Tsutomu’s answer is as short as my question: “Correlation isn’t guilt.”
For the next twenty minutes, he gives me a riff on “high-yield data” and “extensible” databases. I understand maybe a third of it. What I do get, though, is that traffic analysis tells you only where to look for suspicious activity, but it never determines guilt or innocence.
“To establish guilt,” Tsutomu says, “you need to dig a lot deeper, collect old-fashioned evidence admissible in a court of law.”
To make his point, he tells me that while he was able to pinpoint the location of Mitnick’s phone, the FBI ended up raiding the wrong apartment. Fortunately, their mistake was limited to pounding on the wrong door rather than sending a Hellfire missile through the wrong window.
—
The jury may still be out on drones, but shooting into the dark has never been anything other than an iffy proposition. Sooner rather than later you’ll kill the wrong person and end up making more enemies than you eliminate. When you think about it, with the charges against drone victims irrefutable (at least from Washington’s point of view), with body counts unverifiable, and with the definition of victory and defeat left to Washington politicians and bureaucrats who lie for a living, how could drone assassinations not have gone horribly wrong?
Drones are another reminder that scuttling an idiotic idea dressed up as a techno silver bullet is damn near impossible. Falling in dumb love with quants, algos, and their glittering promises, Washington was seduced by a quick, easy release from reality: war at a bargain, assassination without a downside. Extend the logic out far enough, and Washington will start to believe it can rule the world from a single thumb drive and a fleet of solar-powered drones armed with Hellfire missiles. It will have perfected murder, but it can’t remember why.
While all along it remains that to win at any game that counts you have to have real skin in it, see things with your own flesh-and-blood eyes. Murder isn’t some Shakespearean play where violence occurs offstage. Which means relearning quaint skills such as reading a map, getting your boots dirty, and summoning the guts to stick a dagger into the heart of an enemy. Like it or not, bare-knuckle reality will always trump an algorithm.
NOTE TO ASSASSINS: As soon as the act starts to look like a robot assembly line, packaged and corporatized, you’re heading down the wrong path.