And so Don Rodrigo sits, strapped to his horse. He may or may not be dead at the moment, but he is certainly dying. His horse marches into the battle, but even if our El Cid wins this conflict, it will be his last. The networks will try to keep up the fiction that he remains, but even that may become a hard sell.
5
The Runaway Train
If information is power, then those who master this digital chaos first, and derive meaning from it, will likely gain critical advantages. Intelligence professionals, whether in business or in service to the state, are therefore in a silent race to develop tools for mining and analyzing growing volumes of swiftly moving information and then to use it...
– Jennifer Sims, The Future of Counter-Intelligence
For some time, we've been telling anyone who would listen that the internet was becoming the greatest surveillance system in history, metastasizing into the darkest dream of any tyrant of any age. Yes, the internet has also been our great tool of emancipation, but hucksterism, foolishness, and the never-ending lust for dominance on the part of the castle-dwelling class has over-built the emancipatory function of the internet, and is slowly driving it out altogether.
The crucial thing to understand about this is that the lords of network power – the lords of data-derived intelligence – face a variety of all-or-nothing situations. They can either ride this train faster than anyone else, or they can be made irrelevant.
And so the braking mechanisms have been bypassed and a digital arms race is raging.
In the paper quoted above, Ms. Sims continues, noting that “It is not clear that all states win in the big data world.” To which we will add, it's certain that neither all states nor all networks can win. Like all battles between castles, the winners enjoy conquest over the losers.
Game Theory and Surveillance
Game theory studies how rational players behave with each other in certain situations. The situations include factors such as “can they communicate with each other?” “What kind of knowledge do they have about each other?” and so on. It seeks to learn what strategy a rational player would choose in each game and situation.
The cold war between the US and the USSR, for example, looks like this when examined with game theory:
The Russian plan was to use nukes only against military targets, and especially the nuclear facilities of the United States. Their goal was to decapitate and demobilize the US military. Population centers were not specifically targeted.
The United States however, chose a strategy of mutually assured destruction, meaning that in any situation both sides would lose. Their reasoning, straight from game theory, was that no rational player would start a war because he would seal his own fate. So, the situation was limited to only two choices:
Don’t start a war and you can rule your population as you wish.
Start a war and you no longer have a population.
This seems to be what the US communicated, but what the Russians heard was closer to, “the United States is willing to kill hundreds of millions of people to save itself.” This seems to have been a shock to the Russian generals and actually triggered leadership changes in Moscow.
The Russian response, so it seems, was to make sure that there was no way that the Americans could ever strike first or retain a secondary capability. That is what led to the Cuban missile crisis and the strategy of using submarines as missile launch facilities.
What this strategy did was not just communicate that it was stupid to attack, but to increase vigilance enormously... to a dangerous level. That this strategy worked is not attributable to game theory, or planning, but to luck. There were probably half a dozen incidents during the cold war where the only reason why missiles weren't exchanged was that people ignored the strategy. They placed their morals above their orders and the national strategy. If theory would have been followed, few of us would be left to analyze it.
Game theory has advanced since that time, of course, and it now includes factors such as chains of command, communications system, and the integrity of messages. Still, theories seldom deal very well with the complexities of human life.
The argument for global surveillance from game theory is this:
The technology for global surveillance exists.
Due to that existence, somebody will use it.
That somebody will have an advantage over everyone who is not using it.
Therefore, we must do it.
The rational strategy on this playing field is to engage in mass surveillance. That's the only way to mitigate the harm that might be done to you.
And, of course, everyone involved knows this.
So, every capable party, if they follow a narrowly rational strategy, must join the arms race and maximize their use of data-based intelligence, and as secretly as possible.
Dominance was once an issue of producing more cannons; now it's about who knows what about whom... which is the base definition of espionage. We are creating a world that is almost entirely centered on espionage and intelligence services.
Even in the field of commerce, to compete in the future will require people to treat their business as an espionage company. And since so many people are ethical wrecks, they will soon enough turn from the ethical position of defensive tactics, to the far less ethical position of offensive tactics, hoping for a bigger score.
Big Data
Cyber weapons are probably the best strategic weapons that exist. For one thing, cyber weapons have an enormous range of delivery. In theory, at least, they can be released almost anywhere and reach almost any point on the planet in seconds.
And because most of the developed world's critical infrastructure is highly dependent on networked computers, that infrastructure is vulnerable to cyber weapons. Switching off networked computers that control critical infrastructure would disrupt an entire country, and in a targeted way.
For example, you could disrupt communication, power, transportation, media and so on. Then, the population might do your killing for you. In the scenarios for larger cities, the result is about a 90% population reduction within a few weeks, if things cannot be fixed. Cyber-Weapons have the additional property of being hard to trace to the user (cyber-attacks are hard to attribute) which allows for a low-intensity undeclared cyberwar. Furthermore, systems already controlled by an enemy a hard to detect, raising a sword of Damocles above each technologically advanced nation and introducing leverage that is not spoken of. And that makes cyber weapons, in theory at least, a very big thing.
The greatest of all new cyber weapons, however, is not offensive, like breaking a far-off power grid. Rather, it is analytical, and it is called Big Data.
As we've been saying the new age of intelligence differs radically from the old era. This difference is not superficial; it goes down the roots... all the way down to our assumptions of how we know what we know.
The foundation of all stable knowledge, from the 17th century through the 20th, was the scientific method: Start with the smallest, most clearly verified facts, then build on top of them, verifying each new piece along the way. Like nearly everything else, intelligence was built on this verify-and-build model.
In contrast to the verify-and-connect-the-pieces process of the scientific method, the new Big Data model is a kind of slow omniscience. If you remember the Deep Thought computer of Hitchhiker's Guide To The Galaxy, you'll have an image of the process: The petitioner comes to the machine and asks a question. The machine, through an unfathomable process, eventually spits out an answer.
Chris Anderson, in a seminal piece in Wired Magazine[20] summed this up by saying that “big is different.” In other words, when analyzing huge amounts of data, things are different. The scientific method has to be jettisoned and a new model used. Anderson went on, using Google as an example:
Google's founding philosophy is that we don't know why this page is better than that one: If the statistics of incoming links say it is, that's good enough. No semantic or
causal analysis is required. That's why Google can translate languages without actually "knowing" them (given equal corpus data, Google can translate Klingon into Farsi as easily as it can translate French into German). And why it can match ads to content without any knowledge or assumptions about the ads or the content.
In other words, Google does not translate an English word like “cat” into Spanish as “gato” because it consults a Spanish/English dictionary. Rather, Google's systems query a cloud full of data[21] – petabytes[22] of data – and conclude that “gato” is the word that is most likely correct.
This is indirect reasoning, and it is performed by huge masses of computers that work on sets of data that are constantly varying in size and content. (Data sets that are far too large to be analyzed piece by piece.) This indirect reasoning, however, works, as Peter Norvig, Google's director of research, has been quoted saying: “All models are wrong, and increasingly you can succeed without them.”
This is crucial: Models built carefully from verified facts are becoming passé, and we are thought not to need them anymore. We can succeed by going to every-changing oceans of data, querying that data cloud, and have a correct answer spat out to us. Again, this is rather like asking an omniscient but slow being. That is Big Data.
Big Data uses an analysis technique called statistical inference. It draws conclusions from sets of data that are subject to random variations.
These systems identify the mathematical model of a system, not by traditional analysis, but merely by watching what goes in and what comes out. If you can watch a process thousands or millions of times, you can also describe its results... not precisely, but predictably.
The new model is not one of understanding the gears, but of being able to predict what goes in and what comes out. Knowledge of the gearing is irrelevant to this process.
This type of analysis, from large datasets, is particularly good at revealing relationships and dependencies, as well as predicting outcomes and behaviors. Big Data allows analysts to ignore causation.
Ignoring causation, it needs to be appreciated, is a radical change, more or less reversing the scientific process that brought us to where we are now. And because this process works, understanding will soon enough become less than mandatory, and then abandoned, especially by the younger analysts. The omniscient Big Data machine will be fed questions and the analysts will wait for it to spit out answers. “It's faster, easier, and more accurate,” the young analysts will say; and they will not be incorrect.
Big Data is like a human that can only think intuitively – it cannot reflect on its own thinking or question itself. It produces conclusions that will usually be right, but there is no indication of why any conclusion is right, how it was arrived at, or why that conclusion reflects reality. It is a black box that spits out good results, but gives us nothing on how the result was achieved. The black box, located between the data gathering and conclusion, it never asks “why” or “how” but only “what”, and it does not know the effect of any single element in its processes.
Big Data has these characteristics:
It learns from feedback, creating information that it feeds back into itself.
It is based on hidden heuristics. (A heuristic is something that generates a short-cut answer.)
Big Data cannot be systematized.
It is unreflective.
It is hard to manipulate, mainly because it uses oceans of data; more than can be successfully polluted.
If Big Data fails, it fails fatally. This is because it is unreflective, has no systematization and is based on hidden heuristics.
Big Data systems are an absolute necessity under the game theory model. The rationalstrategy is to have more of them than anyone else, and for no one else to know that fact.
Big Data, obviously, is only of value where people rely on interactive environments. It would have very little impact on a country in Africa where only 1 in 50 people have internet access. But it works exceptionally well in developed countries.
What Is This Thing, Really?
In the physical world, Big Data consists of large groupings of fairly average computers, mounted in racks in large data centers. The size of these groupings is typically in the thousands of computers. These computers are connected in ways that allow them to operate in parallel. That is, thousands of them can process the same datasets at the same time; the data being broken into a thousands of parts and handled by thousands of machines simultaneously.
More or less all the big computing companies have moved into Big Data, including Oracle, IBM, Microsoft, Dell, Software AG, Fico, Raytheon, SAP, EMC, HP and many others. Big Data has been growing roughly twice as fast as any other computing sector over the past several years.
Big Data has been in use for some time. Google, for instance, has used it for many years, starting by identifying specific colors, layouts, and designs that made people more efficient internet searchers. They did this by slightly tweaking the pages their customers see for a few million searches at a time and then examining the subtle ways in which people react.
A famous instance of Big Data's results involved the American retail chain, Target. Their data systems correlated the purchase of certain types of products with pregnancy. So, they sent custom advertisements (through the mail in this case) to pregnant women. Uncomfortably for all involved, however, one of these pregnant women was still in high school and hadn't told her father, who called his local Target store to complain. (He handled the news quite well, once it was verified by his daughter.)
This Target case provides a very elementary view of this process:
Big Data didn't figure out why people buy what they buy; it just correlated what went in and what came out. In this case, it learned to identify pregnant women by their purchases.
The store had very complete set of data about this young woman, because she used a Loyalty Card, hoping to save a bit of money.
Using even the old and slow mediums of printed paper and mail, Target created and delivered a custom environment to the young woman. No one else in her family received such a set of advertisements, only her.
This story, while entertaining in its way, is just the most preliminary case. To illustrate the development of Big Data with just one more example, a few years ago Google spent £300 million ($510 million) to buy Deep-Mind, a British artificial intelligence firm which builds profiles on individuals, based on their internet activity.
More Is Double-Better
Far from choking on too much data, the intelligence of the 21st century wants more of it and functions better with ever-more of it. And so, it cannot be scaled back. Stepping backward ten percent on surveillance would result in far more than a ten percent loss in useful information, and probably twice that amount. The benefit from surveillance is not linear, it is exponential.
Before it hits a threshold, surveillance isn't of terribly much value, except when closely targeted, as it was in traditional police work. Past a certain point, however, surveillance (mass surveillance now) has a wide enough base of information that it gains predictive and causative value. And the more surveillance is done, the more value it gains.
Having enough surveillance data allows you to look at history precisely and find patterns in it. And again, these are not indications of causation, only patterns of correlation. But that's enough for predictions to be made. For example, when food prices rise faster than a certain percentage per year, riots and overthrown governments become far more likely. By surveilling food prices, you can predict riots. Having more data means that more statistically relevant correlations can be found.
There is also a fundamental difference between the simple surveillance of traditional police work, for example, and mass electronic surveillance, and it is a difference, not in quantity, but in quality:
Traditional surveillance is about creating a record of past behavior and to capture communication that reveals planning.
Mass surveillance has two qualities: It crea
tes a record to be looked at in hindsight, but it also sees all the individuals individuals it surveilles as a mass - as a single, abstract, object.
But even this term, to surveil, is outdated here. This type of life-long, ubiquitous surveillance is an act of ownership. It turns the mass of people into a swarm, a mob, a collection of non-unique things that do not exhibit individual will. The target persons are first conceived as a mass – without individual faces and sensations, and are seen moving as a swarm.
Once past that point of being seen as collectives rather than individuals, the surveilled masses can be treated as a single entity, leaving empathy unable to attach. Empathy, as has been known since time immemorial, connects between individuals, not swarms. This is why manipulators (or even non-manipulative communicators) almost always use individuals as examples. People just do not connect emotionally with large groups.
This is one of the most dangerous foundations for power imaginable, but it is a type of power that a surveillance operation is most unlikely to give up. Consider these comments, in which Thomas Drake, formerly a top executive at the NSA, likens the control of surveillance to mainlining heroin:
In the digital space, you’re “data drug” habit goes exponential, because there’s just so much. You can mainline this all day long. To me, there’s a psychology that’s not often written about: What happens when you have this much reach and power, and constraints of law and even policy simply fade into the woodwork... Which is made worse by the fact that you can’t get enough, there’s never enough, and there’s more coming... You’re high all the time. Because you’re plugged in. It’s now 24/7. There’s no relief from the addiction.[23]
A Few Final Thoughts
There is obviously a great deal to be said about Big Data, and we will be talking more about it in chapter seven, but we will conclude here with the thoughts of others.
The New Age of Intelligence Page 5