Everyone on this road is driving fast.
Suddenly, a loud, sirenlike noise tears through the car’s interior. Simon pulls his walkie-talkie from his Gore-Tex jacket. A bright red light cries out in distress at rhythmic intervals.
“Pull over,” Simon commands. The driver applies the brakes, not exactly slamming them but not gradually depressing them, either, and steers the car to the side of the road. Like a surreal piece of choreographed theater, every other car on the road also slows and banks.
A moment later, we feel the ground beneath us rise and fall. This is a 6.0 tremor, large enough that—had we been traveling at our previous speed of more than eighty miles per hour—we likely would have crashed. The fishermen, Simon, the car’s other occupants, and I look around at one another. We share a silent acknowledgment that we have just barely avoided a terrible accident.
I’m alive today thanks in part to Japan’s Earthquake Early Warning (EEW) system, a network of more than four thousand seismographic sensor stations.1 These devices detect the low-level initial tremors called primary waves or P-waves that are released by seismic activity. An earthquake’s P-wave telegraphs the size of the secondary wave or S-wave, the tremors that crash cities and bring the fury of the sea to shore. The system computes the signals as input and issues output, the feedback of which takes the form of Simon’s phone going off.
The alert is issued automatically the second that the seismometer detects the signal and transfers it to headquarters.
Because earthquakes are a frequent occurrence in Japan, the alarm now goes off so often it has almost become background noise. In the moments before the 2011 earthquake hit, television broadcasts across the country were briefly interrupted by a crisp, telephonic ringing. A bright blue box appeared on every television screen showing the eastern coast of Japan and a large red X offshore depicting the earthquake’s epicenter. In one of the eerier video clips that emerged from March 11, 2011, members of Japan’s parliament can be seen debating a piece of legislation. Because they’re accustomed to the signals they’re slow to react to the warning at first. When they realize the size of the earthquake, they look nervously to the swinging chandeliers above them. The picture cuts to a flustered anchorman who warns of a possible tsunami off the coast of the prefecture of Miyagi.2
The Japanese have been applying creativity and resourcefulness to earthquake prediction for centuries. Historically, national myth held that earthquakes were caused by the movements of a giant catfish, or namazu, called the Earth Shaker. Though the idea seems ridiculous today, the Japanese took it very seriously at various points throughout their history. In 1592 the samurai warlord Toyotomi Hideyoshi issued what is perhaps the strangest building-code edict in history to the men constructing his castle in the Fushimi district of Kyoto: “Be sure to implement all catfish countermeasures.”
In the later Edo period small catfish were awarded a reputation as earthquake predictors. Strange fish behavior was thought to be an indication that the giant namazu was on the prowl for mischief.
Today, the idea feels fanciful. Several centuries of steady scientific progress have taught us to look for concrete causal relationships in order to understand how one physical entity might influence another. We know that the earth’s tectonic plates are affected neither by subterranean fish, nor the position of the constellation of Cassiopeia, nor the current level of God’s wrath but by physical systems of enormous complexity and limited accessibility. Our understanding of the world through the lens of science suggests that P-waves indicate S-waves, but there exists no physical mechanism by which a catfish could know of an earthquake days in advance. Anecdotal evidence to the contrary proves only that humans have active imaginations, because catfish don’t predict earthquakes.
Turns out, they almost do.
One of the key triggers of large seismic events is the buildup of pressure between rock formations in the earth’s crust. This pressure also releases electrical activity and will do so days before large quake events. Loose “defect electrons” rise up through porous gaps in the earth’s crust; they ionize when they meet the air. Under the right circumstances, this can cause subtle hydrogen peroxide increases in certain fault lines proximate to bodies of water, making such bodies just a bit toxic to very sensitive marine fauna.
British zoologist Rachel Grant observed this phenomenon firsthand when hundreds of toads fled a pond near L’Aquila, Italy, in the days just prior to an enormous 2010 earthquake. As Grant wrote in her paper that was published in the Journal of Zoology, “Our study is one of the first to document animal behavior before, during and after an earthquake. Our findings suggest that toads are able to detect pre-seismic cues such as the release of gases and charged particles, and use these as a form of earthquake early warning system.”3
Catfish, like toads, have extremely sensitive skin. But unlike toads, they can’t abandon a body of water that’s becoming toxic. They can only thrash about or behave strangely, like the Earth Shaker.4
In his book The Signal and the Noise: Why So Many Predictions Fail—But Some Don’t, statistician Nate Silver is rather hard on Grant. He suggests, though not explicitly, that she’s reached an insupportable conclusion, as her paper seems to assert that the observed toad behavior is “evidence that they [toads] had predicted the earthquake.” He describes her work as the sort of thing that “exhausts” real seismologists and notes dismissively, “Some of the stuff in academic journals is hard to distinguish from ancient Japanese folklore.”5
Silver is certainly a talented statistician deserving of the celebrity that’s been awarded him. He’s right to point out that history is littered with failed attempts to predict earthquakes, often by observing strange animal behavior. He’s also right to point out that statistical analysis of previous earthquakes is surely a far more useful signal than is toad behavior, at least for now.
But he’s misstating Grant’s intent. She’s not suggesting that the toad behavior is “evidence that they predicted the earthquake.” Neither the toads of L’Aquila, nor the catfish of Japan, nor even the EEW are actually predicting anything and Rachel Grant knows this perfectly well. These are feats not of prognostication but of detection. Grant and her colleagues acknowledge that testing the hypothesis outside a laboratory setting has thus far been impossible because they still don’t know when and where an earthquake will strike. And neither do the toads. When they’re in a pond with higher hydrogen peroxide levels they become uncomfortable and they leave. They are indifferent to earthquakes, to Nate Silver, and to the future.
It’s humans who predict things.
As we attempt to make use of this abundance of telemetric data, we’re going to make errors. One of the statistical traps Silver and other statisticians warn against is overfitting, or applying a specific solution to a general problem. In the case of earthquakes, this could mean watching toads rather than history because toad behavior lends itself to a very specific type of prediction method.
We are about to enter a golden age of overfitting, if such a thing can be said to exist. The sheer volume of data we now generate as individuals and institutions suggests that more people will be able to create more models with data points and observations that offer the false promise of certainty. We will model more and so we will make more errors, but an increase in modeling activity will not diminish the costs or consequences of those errors. Many small mistakes will feel extremely large particularly in the context of international stock and commodities markets. Overfitting also speaks to an impulsivity that’s in our nature. We gravitate toward evidence, data, and facts that support a conclusion we’ve already reached or bolster the argument we’re trying to make. Finally there’s enough data to lend some support to virtually any argument, no matter how crazy. To overfit is human.
But the fact that electrical activity from pressure increases days before large seismic events is beyond dispute. It’s exactly the sort of predictor that coul
d reliably indicate an approaching disaster if only humanity could devise some cost-effective way to place millions of sensitive electron detectors deep beneath the earth’s surface near fault lines. It’s science fiction. But at one point, so was the idea of a sensor spiderweb that could detect P-waves.
We are turning our physical environment into a catfish.
A Global Nervous System Emerges
In 1988 a scientist at Xerox PARC named Mark D. Weiser put forward a novel vision for the future. Computer hardware, he said, would migrate from deskbound PCs to pads, boards, and “smart” systems that were part of the physical environment. The term Weiser gave this new sensing environment was “ubiquitous computing.”
This vision for the future speaks a lot about the man who came up with it. Weiser was not a typical computer hardware genius. Take a look at his informal writings and the accounts of people who knew him and you will not find a man who loved gadgets and code for their own sake but someone motivated by a passion for actual experience, a sensualist, a devotee of skydiving, rock repelling, and lead drummer in a punk band called Severe Tire Damage. Through ubiquitous computing he imagined a future in which humans interacted with computers on an unconscious level, through regular activity; a future in which computers served to remove annoyances and answer questions like “Where are the car keys?” “Can I get a parking place?” and “Is that shirt I saw last week at Macy’s still on the rack?”6 while keeping us connected to what we care about. Computers weren’t supposed to get in our way, or be constantly in our hands, or be connected to our ears through shiny white earplugs, or demand that we answer their every chirp and bell. As they became better they were supposed to become more numerous but also disappear into the background.
A decade after his death, it’s the “ubiquitous” portion of Weiser’s ubiquitous computing vision that’s becoming reality for most of us. The total number of devices connected to the Internet first exceeded the size of the global human population in 2008 or so, according to Cisco, and is growing far faster.
Cisco forecasts that there will be 50 billion machine-to-machine devices in existence by 2020, up from 13 billion in 2013. Today, we call ubiquitous computing by another name: the Internet of Things.
For large institutional or corporate consumers of information, the spread of sensors and computer hardware across the physical environment amounts to better inventory tracking and customer targeting, which will help bottom lines. The Internet of Things can be found most immediately in the RFID tags that have made their way onto everything from enormous inventory palettes to the clothing labels that Swiss textile company TexTrace7 sews into American Apparel clothing to track shipments. Most RFID tags that we encounter today are small squares of paper, plastic, or glass containing a microchip and an antenna at a cost of about twenty cents. The microchip holds information about the product (or thing the RFID is connected to). The antenna allows an RFID reader to access data on the chip via a unique radio signal. Unlike a simple printed bar or quick response (QR) code, the RFID tag doesn’t have to be directly under the reader to work. The reader need only be close by. This allows retailers to monitor the inventory in their store in something like real time. Some futurists have suggested that RFID could one day render the checkout station obsolete. In this future, when you saw a product that you wanted you would simply pluck it from the shelf and—so long as you had a user account or were identifiable to that store—walk out the door. The product’s RFID tag would tell the retailer the product had been purchased and your account would be debited. Sound far-fetched? Millions of Americans today buy access to toll roads through the dashboard-mounted RFID tags that are part of the E-ZPass system. The act of purchasing takes the form of a simple deceleration and a brief exchange of data between the RFID tag’s antenna and the tollbooth’s reader. And RFID is just one of the many smart or sensing tags and microchips that are making their way into our physical environment at rapidly decreasing cost.
For patients and graying baby boomers, the Internet of Things is ushering in a revolution in real-time medical care. It is alive inside the chest of Carol Kasyjanski, a woman who in 2009 became the second human being to receive a Bluetooth-enabled pacemaker that allows her heart to dialogue directly with her doctor.8 The first was former U.S. vice president Dick Cheney, who received one in 2007, but never activated the device’s broadcasting capability for fear of hackers.
The military uses the Internet of Things to do more with less. In Afghanistan it takes the form of the fifteen hundred “unattended ground sensors” that the U.S. Army is leaving littered across the Afghan countryside as the U.S. mission there winds down. These sensors, which are intended to pick up human movement, are intended to allow the Pentagon to eavesdrop on the countryside and detect how Afghans (or Pakistanis) are moving over their country.
It is, quite simply, all of the computerized sensory information that can be gathered and transmitted in real time about what is happening right now. When this happens to machines we call this big data. When it happens to us we call it sensing.
In many ways, this expanding, computer-connected environment is inconspicuous (as Weiser intended). The presence of sensors able to detect ammonia, a common component of explosive material, in the New York City subway is not something I devote thought to when I’m taking the downtown 6 train; I’m just glad it’s there.9
The Internet of Things is not a far-off dream; it’s here. We’ve been accepting the presence of more sensors in our environment for decades now. It’s impossible to argue against the usefulness of Japan’s EEW, or radon detection devices in subterranean structures, or home security systems that sense when a door is being opened and alert the police and homeowner. The average 777 has so many sensors on board that a three-hour flight can generate a terabyte of data. Twenty flights generate the data equivalent of every piece of text in the Library of Congress.
For the owners of the copper wires, the fiber-optic cables, the cell phone towers, and the servers on which the Internet runs, the growth of the Internet of Things means massive future profits. The firm Gartner has predicted that the global market for “contextually aware computing” will exceed $96 billion per year by 2015. It’s no wonder such companies as Cisco, IBM, and Verizon spend millions of dollars in ad, marketing, and grant campaigns to persuade the world that a “smarter” planet is so very good for everyone. And it is, in many ways. But first and foremost, a smarter planet is good for them.
Importantly, the Internet of Things is not solely the product of companies and governments. It’s become a homegrown phenomenon as much as a big telecom money machine, and it’s empowering regular people in some very surprising ways.
The Internet of Things, Three Vignettes
On March 11, 2011, engineer Seigo Ishino was at his office in the city of Kawasaki near Tokyo when the EEW system sounded. Like any rational person caught in a massive tremor, he crawled under his desk until the quake passed. He emerged a few minutes later unscathed but, as a result of the seismic event that had just occurred, his life was now far more complicated than it had been just that morning. News of technical problems at the Fukushima Daiichi nuclear plant spread quickly in those early hours through Japan and then around the globe. Tokyo was close enough to Fukushima (about 160 miles) that the meltdown posed a serious concern particularly for children and pregnant women, as radiation is most harmful to babies and kids. Seigo’s wife was eight months pregnant at the time. He was faced with some hard choices. Was the level of radioactive cesium and iodine spewing out of the plant dangerous enough to compel him to relocate his family farther south? If so, he needed to act quickly to get a train ticket, as the price was rapidly ascending. There was also the question of where they would stay, and how he would earn money because he would effectively be abdicating his duties at his present job and had no job prospects outside of Tokyo. Was the danger significant enough to warrant an evacuation from Japan? The many thousands of foreigners who w
ere attempting to leave that week were also driving up the cost of airfare and there were questions about how to obtain an exit visa. Alternatively, was it safe to stay where he was? What about the food and the water supply? He needed more information.
The Kan administration’s press secretary, Yukio Edano, began giving regular press conferences, clad in a bizarre blue jumpsuit, to inform the public that radiation levels were not dangerous and that the situation at the troubled plant was under control. The official messaging took a turn for the ridiculous on March 12 when the Kan administration assured the public that the pressure levels at the reactor had stabilized only to then admit, a few hours later, that a massive buildup of pressure had blown the walls off the reactor building. Yukio Edano again took to the podium to steadfastly affirm that the situation was improving as the reactor in the picture behind him smoked and fumed.
Seigo elected to stay but, like millions of other Japanese, he no longer trusted the official story that was coming from the government and from TEPCO, the corporate entity that operated the plant, both of which he regards as “most untruthful.”
Seigo was a member of an international group of community designers, engineers, hackers, and hobbyists who built sensors and installed them in buildings and other aspects of the built environment to monitor energy use. The community was centered around a platform called Pachube (now Cosm), which allows users with sensor data to share it in real time on the group’s site. Not long after the news of the meltdown spread throughout Japan, thousands of people across the country were tweeting Geiger counter data and hundreds of Pachube users were streaming their data directly to the Pachube site.
The Naked Future Page 2