B008J2AEY8 EBOK

by Jaron Lanier

  It’s not always necessary that the data be made absolutely unavailable; sometimes data can just be decontextualized enough to become less valuable. Facebook provides a fine example. If a great deal of personal creativity and life experience has been added to the site, it’s hard to give all that up. Even if you capture every little thing you had uploaded, you can’t save it in the context of interactions with other people. You have to lose a part of yourself to leave Facebook once you become an avid user. If you leave, it will become difficult for some people to contact you at all. Would you ever be willing to take the risk to sever a part of your own life’s context in order to disengage from a Siren Server that ogles you?

  Denial of Service

  Yet another way to create a punishing network effect involves control of routing and bandwidth. To understand this method, I refer you to your wireless bill. A particular Siren Server becomes the only way to connect to the information world. (Companies with proprietary hardware, like Apple, do this as well.) To sever, you must often pay penalties, purchase new equipment, and therefore potentially lose investments tied to the old equipment, like apps, only to get into a new long-term contract.

  Access-granting services need not be Siren Servers, since they could just be boring and bill for granting access, but they have caught the deliriously alluring scent of the game by now and are trying to become big data players as well. This has led to power struggles, such as whether a smartphone company or the wireless carrier is in charge of various services and revenue opportunities, and whether the principle of “net neutrality” will endure.

  There is often a cascade of hardware lock-ins that cumulatively corner a particular person. You might be locked into one service that connects your home to the Internet with a cable, another that connects your phone or tablet to the wireless signal, and yet another that provides the devices you use and key services like an app store for it.

  This demonstrates an interesting difference between Siren Servers and traditional monopolies. There is no reason that there can’t be a lot of Siren Servers. They form ecologies instead of company towns. The reason to be concerned about them is how they distort and shrink the overall economy by demonetizing more and more value. But they don’t necessarily turn into the only game in town in the way that an old-time railroad monopoly might have.

  Arm’s-Length Blackmail

  There are yet other punishing network effects that resemble a soft kind of blackmail. Some local retail review sites have periodically been accused of skewing or ruining the online visibility of local businesses that cease to buy “optional” premium placement services.2 Social networking sites will sometimes extract fees to make someone more “visible” on the site.3 This is particularly true for hookup services akin to the “where the babes are” app that was pitched by the Berkeley graduate students.4

  Readers of my previous book will recall an extended examination of how ideas and patterns of use and behavior get “locked into” networked software. This type of software lock-in is often employed to create or buttress a punishing network effect. If a small business designs its own processes and code around the cloud services from only one of the major cloud companies, then it can easily get locked into that company.

  Some sites have gotten fairly large with mostly rewarding network effects and barely any punishing ones. eBay is mostly based on rewarding effects, for instance. No one’s really punished for buying or selling elsewhere.* (This is in contrast to Amazon, which will sometimes lower prices on an item to undercut you if you sell the same item at a lower price elsewhere.)

  *Twitter’s lack of a plausible revenue growth plan as I write this is similarly due to offering carrot without a commensurate serving of stick. By the time you read this, that might have changed.

  When you are subject to someone else’s punishing network effect, every decision becomes strategic. If you plan to break out of the gravitational field of a Siren Server, you often have to swallow hard and go all the way. The burden of that big leap creates a new kind of social immobility.

  Who’s the Customer and Who Are All Those Other People?

  To understand a particular Siren Server, it is critical to distinguish between distinct populations connected to the venture in different ways. Siren Servers often pit these populations against each other.

  Once a Siren Server becomes dominant in its niche, after the Local/Global Flip, it treats those who connect with it as data sources and as subjects for behavior modification. However, there are usually sub-populations subject to different mixes of rewarding and punishing network effects. One sub-population might be shown carrot and stick in equal measure, for example, while another might mostly be offered carrots.

  In the cases of Google and Facebook, this difference tracks the distinction between users and customers. Some people, the users, are valued mostly as data and potential for behavior modification, while others, the advertisers, are also sources of money. It is crucial, obviously, to capture money if the Siren Server is to be a business.

  This bifurcation can lead to confusion, as when Siren Servers are scrutinized in the terms of old-fashioned antitrust. When a service like Google is evaluated, one of the first observations is that users are free to leave. That is true.* From a typical user’s perspective, Google is mostly carrot. But the other population—the true customers, the advertisers—is less free. It is captured because of punishing network effects.

  *True for search, that is. Not so true if a user has put personal data in Google’s tools.

  In the case of Wal-Mart, the captured population was the supply chain. Google’s true customers are the advertisers, who are captured. Wal-Mart’s customers weren’t the critical population for it to capture, however. Retail customers gradually became a little captured in some locations where retail choice was eventually reduced, but for the most part they could shop elsewhere if they were so inclined, but it was the optimization of the global supply chain through the use of punishing network effects that really empowered and enriched Wal-Mart.

  CHAPTER 14

  Obscuring the Human Element

  Noticing the New Order

  Every tale of adventure lately seems to include a scene in which characters are attempting to crack the security of someone else’s computer. That’s the popular image of how power games are played out in the digital age, but such “cracking” is only a tactic, not a strategy. The big game is the race to create ascendant Siren Servers, or, much more often, to get close to those that are taking off and ascending in ways that no one predicted.

  Networked contests for wealth and power tend to follow a pattern. Each particular scheme launched over a network, each purported golden goblet, tends to follow a well-worn course. Networked information, when it is about business instead of science (or, if you like, about human behavior instead of nature), follows a characteristic life cycle.*

  *Oh how I hate using the term life cycle for something that isn’t alive. We are fascinated by ritualistic declarations that we have created new life in our artifices. In the case of big human data, it’s a mistake to perceive even an object, much less a living thing. This is the way these matters are talked about in my community, however, so I occasionally use the terminology despite my objection.

  Since I prefer to see the faces instead of the goblet, I find that following the ways in which servers obscure the real people who are the sources of value is also a good way of noticing how the struggle for power proceeds.

  Who Orders the Data?

  Some Siren Servers relish a world in which data starts out as a mess, decontextualized and mysterious, until it is brought to order by the server’s analytics. Google is probably the best-known example. A Siren Server in this position will do all it can to promote every manner of “open” activity. Data made available for free with inadequate documentation on the open Internet is the ideal raw material for such a venture.

  Later on I’ll describe how a remarkably simple idea in network architecture, which was
the motivation for the very first digital media designs, was lost, and how that loss created much of the chaos that search engines attempt to undo today.

  Other Siren Servers enjoy data that is ordered either at the time of entry or later on, but in either case for free. Facebook is a great example. Google must find patterns in chaos, while Facebook expects you to enter fairly contextualized information in the first place, essentially filling in the blanks of provided forms. However, Facebook also derives additional order through analysis, results that are hidden away in a dungeon.

  A “content” site in which almost all contributions are unpaid, like the Huffington Post, shares this quality with Facebook. Online retailers like Amazon and eBay are also examples, since they don’t have to pay for reviews or the design of product presentations. Those who sell through these schemes are mostly responsible for creating and tending their own presentations, unlike in traditional retail, where the retailer has to figure out how to present each product.

  This is a key sign of a Siren Server. The lowly non-Sirens are as responsible as possible, while the Siren Server presides from an arm’s length.*

  *Another example is Wikipedia. I am not condemning it, and in my previous book have discussed what I see as its strengths and weaknesses. As I argued earlier, however, it does reduce markets for certain kinds of scholars in the long term in order to demonetize scholarship in the short term, so it qualifies as a Siren Server. It creates the kinds of false efficiencies that thwart levees.

  Yet another interesting example is Craigslist. This is a fascinating, idealistic Siren Server that is mildly for-profit. It only charges for certain types of ads, such as from prospective employers, while offering most services for free. Craig Newmark could probably have built his business into a giant along the lines of eBay or Amazon. Instead, he created a service that has greatly increased convenience for ordinary people, while causing a crisis in local journalism that once relied on paid classified ads. To me, Craigslist has a tragic quality, since it is as modest and ethical as it can be, eschewing available spying opportunities, and yet it still functions as a Siren Server despite that.

  In some cases, ordinary people are persuaded to put extraordinary work into correcting and sorting the data in an Siren Server, at their own risk and expense. A fine and maddening example is credit rating agencies, which provide a labor-intensive path for people to correct mistakes in their own data.

  The Human Shell Game

  Computation done within a Siren Server occasionally still requires some human involvement from insiders to the scheme.* Today, for instance, Amazon has skilled, real people answer the phone to provide customer service.

  *There’s usually a ritual in place to make sure everything possible is done to avoid actual human involvement for as long as possible, even if it is inevitable. The cliché we’ve all lived through is that you call about, say, a problem in how an insurance or credit rating Siren Server has screwed up a key decision about your life. Perhaps you were denied coverage for needed medical treatment. After an hours-long battle with the maze of a robo call center, you finally talk to a real person, probably in India or the Philippines. This might be the first time real human eyes associated with the Siren Server have perceived your data.

  However Amazon is also exploring how to get non-elite service jobs out of the way of the Siren Servers of the future. The company offers a Web-based tool called Mechanical Turk. The name is a reference to a deceptive 18th century automaton that seemed to be a robotic Turk that could play chess, while in fact a real person was hidden inside.

  The Amazon version is a way to easily outsource—to real humans—those cloud-based tasks that algorithms still can’t do, but in a framework that allows you to think of the people as software components. The interface doesn’t hide the existence of the people, but it still does try to create a sense of magic, as if you can just pluck results out of the cloud at an incredibly low cost.

  The service is much loved and celebrated, and competes with other similar constructions. My techie friends sometimes suggest to me in all seriousness that writing books is hard work and I should turn to the Mechanical Turk to lower my workload. Somewhere out there must await literate souls willing to ghostwrite for pennies an hour.

  The Mechanical Turk is not really that different from other Siren Servers, but it is so up front about its nature that it stands out. Those who take assignments through it often seem to even enjoy the fun of emulating an intelligent machine for someone else’s profit.1

  The charade has a triply dismal quality.

  Of course there is the “race to the bottom” process that lowers wages absolutely as much as possible,2 making temp jobs in the fast-food industry seem like social climbing on-ramps in comparison. Yet there are people ready to step up and take such roles. More than a few recruits appear to be the live-at-home kids of middle-class Americans, whiling away their time.3

  Whenever there is a networked race to the bottom, there is a Siren Server that connects people and owns the master database about who they are. If they knew each other, comprehensively, they might organize a union or some other form of levee.

  The second dismal quality is that artificial-intelligence algorithms are getting better, so gradually it will become more possible to not even acknowledge the contributions of real people to the degree done now.

  Finally, the Mechanical Turk is often applied to the more pathetic tasks associated with Siren Server contests. One journalist found that 40 percent of the tasks on offer are to create spam.4

  CHAPTER 15

  Story Found

  The First Act Is Autocatalytic

  A newly launched Siren Server is like a tiny baby creature in a hostile ecosystem that must grow fast enough to survive in a world of predators. The most common means to survival is to route enough data fast enough so that by the time predators notice you at all, they won’t find it worthwhile to go after your niche.

  There are a variety of Siren Servers, ranging from consumer-facing Silicon Valley startups tempting people with “free” bait, to financial servers that skim the cream off the economy in relative obscurity, to providers of infrastructure who realize that they can also play the big data game, to governments and other entities yet to be discussed.

  In all cases, there has to be some way for a particular Siren Server to gain enough initial momentum to become the beneficiary of network effects. Therefore, the primary enemy of a fresh server is not competing wannabe servers, but rather “friction.”

  Friction is what it feels like to be on the bad side of a network effect. Even the slightest expense or risk might slow the initial growth spurt, so every possible effort is made to pretend there are no costs, risks, or even delayed gratifications. This can never really be true. Yet it feels true as you sign up for a social network or an app store for the first time.

  Since You Asked

  Here’s typical advice I’d give to someone who wants to try the Silicon Valley startup game: Obviously you have to get someone else to do something on your server. This can start out as a petty activity. eBay started out as a trading site for people who collected Pez candy dispensers. The key is that it’s your server. If you’re getting a lot of traffic through someone else’s server, then you’re not really playing the game. If you get a lot of hits on a Facebook page, or for your pieces on the Huffington Post, then you are playing a little game, not the big game.

  In some cases you can be the predator. You might start by noticing some other pretender to a throne that isn’t growing as fast as it could and overtaking it once it has identified a viable Siren Server niche to be won. This is what Facebook did to Friendster, Myspace, et al.

  In other cases you might form an offering out of whole cloth at just the right time and place. This is what Twitter did.

  Some part of me still wishes that serious technical innovation were more essential to hatching Siren Servers. Google was initially based on genuine algorithmic innovation. Facebook certainly has had it
s engineering challenges, mostly related to getting big fast without a reliability crisis, but it’s hard to see much computer science innovation in it, at least in its foundation.

  Why the Networked World Seems Chaotic

  Lately, the depths of pettiness seem unbounded. Why do so many people use Pinterest?* There were many competitors offering similar designs. By now Pinterest enjoys rewarding network effects so there’s no mystery. People now use it because others do. But why did Pinterest grow enough to win network effect prizes, instead of any of the many other similar infant creatures in the ecosystem?

  *It’s always tricky to write about these things since I must guess what points of reference will survive long enough to mean anything to this book’s readers years hence. Pinterest is a fast rising star among consumer-facing sites. You can copy photos and other data from around the Web onto virtual pin boards and share them.

  There’s a well-supported analytic class—statisticians and MBAs employed by venture capitalists, big companies, and private capital firms—that attempts to model the qualities of hopeful startup sites, in order to predict which ones will take off. This is like predicting the weather, a challenging kind of science. Some progress has been made, but there remains an element of chaos and unpredictability. No one can know all the little fluctuations that were in play that gave a site like Pinterest its window of opportunity.

  What makes one Siren Server take off while a seemingly identical one flops? This is like asking why some silly Internet memes rise and others fall. There are many factors, mostly uncounted.

  It’s entirely imaginable that Pinterest would have flopped if circumstance had been just slightly different. A butterfly might have flapped its wings on the other side of the world, as the saying goes. Of course, the proprietors of a site that takes off are always certain it was because they did exactly the right thing.