“I think he’s trustworthy, but you’re the boss. When you need his advice, let me know. I’d be happy to make the introductions.”
“Thanks, Stephen. Sounds like a good person to have in reserve. Anyway, everything you said sounds great. One more thing, can you also flag for me the people who have lost the most financially recently, I mean just of the 5,000 on your list?”
“That’s no problem. But, again, I don’t know if that will really be indicative of much,” Stephen said distantly. What Sebastin did with the data was only mildly interesting to Stephen. That was something Sebastin would have to figure out once he had the data. Stephen was already engrossed in thoughts of how to implement Sebastin’s request.
“Well, if you can, let’s just try it. I can only guess, but maybe people who suddenly experience an extreme downturn are the most susceptible to persuasion? If we have the data, we’ll figure out what to do with it.”
And with that, Sebastin picked at his pie and took a long sip of his now tepid tea. I bet he’s already working on the problem. That’s enough for today; don’t push too hard. Besides, if he’s half as good as last time, who knows how many extra treats I’ll wind up with.
Stephen sat in the booth without saying a word. He would stay that way for the few minutes it took to figure out the exact commands required to obtain the data Sebastin needed. This awkward interaction was exactly why he hated having these informal meetings in person. His coffee was cold and his apple pie neglected, the latter a poor substitute for Ubatoo’s handmade croissants.
-THOUGHTS LIKE
BUTTERFLIES-
July 24, 2009.
With only a few weeks left before the official end of their internships, the overwhelming majority of interns were getting increasingly anxious about their prospects of receiving a full-time offer. Every meeting an intern had with their sponsor came with anticipation and a hopeful but nervous stomach full of butterflies. Every meeting was a chance to hear good news. In reality, though, this almost never happened, notwithstanding the summer’s anomalies: Aarti, William, and Yuri.
Stephen was in a unique position among the crop of interns. He was the only one who had led a company from inception, through hiring and firing, through growth and decline, and had seen the cycle from every angle. Nevertheless, despite the advantage this gave him in recognizing the rat race with the other interns, Stephen was unabashedly caught up in it as well. His work had been solid this summer. He had mastered Jaan’s system. He had crunched numbers and scoured through petabytes of data looking for patterns in users, and in doing so had made many advertisers (and one non-profit organization) extremely happy. Nevertheless, as far as Atiq or Jaan was concerned, he had been doing exactly what he was hired to do. But who wasn’t doing that much? It was abundantly clear that only those who exceeded expectations had a chance of a job offer. That is what Yuri had done.
When Aarti asked for Kohan and Stephen’s help on an urgent project, there were many reasons that they were willing to offer their services in any way required. The project, with the three of them focused on it, only took a day to finish—they started at 1:00 p.m. and by 12:30 a.m. Kohan’s part was completed. Aarti and Stephen finished an hour later. The day had been spent improving a light beer advertising campaign aimed explicitly at people earning less than $35,000 in both the Carolinas and Georgia. The experiments were set to run. In twenty-four hours, the results would be tabulated and the success or failure of their analyses and campaign refinements would be known. There was nothing more to be done for that project, and it was too late to start anything new.
“I was going to see if Kohan wants to get some coffee. You interested?” Stephen offered to Aarti.
She took a book out of her backpack and held it up, this time with her hand deliberately placed across any exposed skin on an otherwise hot pink and lavender book cover. “Not tonight. I’ve got some important reading to catch up on,” she smiled.
They started to walk out of Aarti’s office. Both stopped before making it past the threshold. It was the only appropriate reaction. The sight of a gyrating Kohan, in full regalia, would likely have made anyone stop in their tracks. The hat. The boots. The mustache. An enormous set of headphones covered his ears, and tethered him closely to his computer on his desk. Despite his bobbing backside, his head had to stay still and slightly bent to ensure the headphones didn’t come unplugged. Though no music could be heard to anyone but him, Kohan’s urgently whispered rendition of Kid Rock’s “I’m a Cowboy” was probably music enough for the audience—which, as they discovered from glancing around the room, also included Yuri, who had been helplessly trying his hardest to ignore the spectacle.
“Sure you don’t want to join us for coffee, Aarti?” Stephen asked again.
She shook her head no, and with that Aarti departed, on her way to finding a comfortable spot on Ubatoo’s grounds to read for a few hours, and try her hardest to remove that last image of Kohan from her mind before sleep came.
At 2:15 a.m., instead of going home, as both of them should have, Kohan and Stephen walked to an all night café and ordered lattes from the single barista on duty in the nearly empty room. Stephen guided Kohan away from her, a precaution even Kohan found a bit odd. “I think I’ve come up with a project we can hang our hats on,” Stephen said.
“I’m in,” Kohan replied, without hesitating.
“Let me tell you what it is first, then you can decide.”
“Okay, but, really, I’m in. You know I don’t have a project of my own yet.”
“I’ve been working with this group, ACCL. Ever heard of them? Atiq introduced me to the head guy, Sebastin, at the first party we attended. They’re a non-profit that warns people when they are likely to be mistakenly flagged by the NSA/CIA/FBI/Homeland Security, or whatever government group is now doing all the wiretapping, tortures, and whatever else they do. You know, all your old bosses. Anyway, the ACCL tries to warn them, and the public, that this type of thing happens.”
“I’m not sure I believe that any group knows what the NSA, or any of those agencies you mentioned, looks for,” Kohan interjected.
“They obviously have some source I don’t know about. These guys are all well connected. Who knows who they talk to? They’ve certainly attracted the attention of Ubatoo and a bunch of the other big tech companies around here. I imagine they have the ear of plenty of people in Washington, too.”
“Okay, I’m skeptical, but go on. What’ve you been doing for them?”
“The same things we do for any advertiser who comes to us: sift through all of our logs and data, the usual. I just hand them a list of people who match the criteria they think are being targeted. Then, it’s up to them to contact the people. I think they’re setting up some lawsuits, too. But you didn’t hear that from me.”
“And Atiq is okay with you doing this? We don’t usually hand out specific people’s names to advertisers. We just usually use that information ourselves, right?”
“Well, it was Atiq who introduced us, and he asked me to do whatever they needed. They’re not an advertiser. It’s for a good cause. Besides, in the end, it’s going to save these people from a ton of trouble.”
“I suppose it will, if ACCL is right. So, what’s your idea?”
Truth be told, after hours of tortuous thought, Stephen hadn’t yet thought about a concrete project. He had wanted to brainstorm something with Kohan, but the unfortunate decision to start talking about this after 2:00 a.m., coupled with Kohan’s already clear skepticism, wasn’t boding well for a productive brainstorming session. But now he needed to continue with it anyway.
Stephen turned to the usual standby that is inevitably applied to anything done manually at Ubatoo—make it automatic. “Maybe we could automate the work I did for ACCL? Wouldn’t it be useful for anyone to be able to log on to Ubatoo and get a ‘How likely am I to be monitored score?’ We would look at what books you’ve bought, what products, medicines, lawn chemicals you’ve purchased, what web
pages you’ve visited online, what queries you’ve sent us, where you’ve traveled, who you’ve communicated with, and everything else we can get our hands on. We’d just crunch all that information and give you a single score, 1–100 of how likely you are to make it onto some watch list. It would be like a credit score, except rather than looking to see if you pay off your bills on time, we would look at a bunch of warning signals, telling you that you should be careful.”
Kohan’s expression portended his sarcasm. “If I get a score of 1, I’m basically Mother Teresa, and if I get 100, I’m Osama Bin Laden? Got to admit, Stephen, didn’t see this coming.”
“Something like that. If you’ve really got nothing to hide, and your score is high, you might want to know that. I’d certainly want to know, wouldn’t you?”
“You definitely think differently than most interns here. I could have easily imagined that you would come up with a self-serve product where advertisers can explore their own data and reconfigure their advertising spending. You know, automate the other 99 percent of the work that you and I do. But, instead, you come up with the Terrorist-O-Meter.”
“You’re still in, right?” Stephen said, trying his hardest to grin along with Kohan.
“First things first, Stephen. Where does this ACCL actually get their information? How do they know what books terrorists read or what web sites they visit? I didn’t know they had an official reading list.”
“Oh, come on, Kohan, you and I both know that once we have a few known terrorist support sites, we can infer dozens of others. We can watch who visits them and what other sites they visit, and what products and books they eventually buy, who they send e-mail to. For that matter, we could look at what soft drink they buy most or what car they drive most. You know this.”
“I do. We would have killed for that information at the NSA,” Kohan said, serious for the first time in this conversation.
“We have it all here, available to us. We have the opportunity to finally do something good with it, and get the full-time offers we want,” Stephen said, getting himself more excited about the project.
“Okay, it’s interesting. But I just can’t see it having the same impact that Yuri’s project did . . . Let me play devil’s advocate here. First, why would Ubatoo really care about the Terrorist-O-Meter? It doesn’t make them money and puts them in the spotlight for having this data. Second, technologically, it’s a nice, maybe even really nice, use of Jaan’s system. On the other hand, it’s just applying some new rules to the same type of data crunching we do every day, like you said. Third, who in the world is going to champion this inside Ubatoo? Who’s going to stick their neck out for this? I don’t know, Stephen.”
“Kohan, this is the right thing to do. Look at what we just did. We just finished using the world’s largest computation machine for micro-targeting crappy beer to destitute people in three Southern states so they can get drunk faster. Don’t you want to do something more meaningful?”
Kohan didn’t respond.
Stephen tried again. “If nothing else, it’s better than any other plan we have right now. Together we have a much better chance for success and I could use your help.”
Maybe he would say yes. Maybe Stephen had finally broken through.
But it was not to be.
“I think this one is going to be all you, Stephen. Having done an internship at the NSA, it might be a mistake for me to be too closely involved. You might want to think about it too, you know?”
In fact, Stephen did know. Despite trying to sound convinced in his ramblings, he was far from it. Kohan’s hesitation and reluctance derailed Stephen for days. Thoughts and ideas fluttered in and out of his mind like butterflies. They had no rhyme or reason, and each fleeting flash of brilliance held his undivided attention—but only until the next moment, when it was lost and forgotten.
-CORE-RELATIONS-
July 28, 2009.
Four days later, a rational justification of the idea remained elusive. Nonetheless, that hadn’t stopped progress. First, he committed himself to the project, despite any lingering trepidation. Anything else he had considered was too small, too incremental. And hadn’t he just promised himself to work only on projects that mattered, and that he could be passionate about? Second, the fact that the algorithms and programs required to make the idea a reality were within his grasp was a strong motivation to continue.
The full potential of what he was going to create was as exciting as the early days of SteelXchange. The system, which Stephen codenamed WatchList (instead of the Terrorist-O-Meter that Kohan had called it), was now just a component of a much bigger plan. If WatchList worked, it wasn’t an unreasonable leap of faith to foresee a world where knowing your WatchList score was as common as knowing your credit score, or even how much money you had in your bank account. It would be a score that everyone would want to track. There would be groups of users who disagreed with their score—but that was good. This is where he would truthfully point out to them that the score is based on the evidence that Ubatoo knows about you. The more details about yourself that you volunteer, the more accurately Ubatoo could predict your score. The less information you give about yourself, the less accurate the score. And this would be Ubatoo’s reason for supporting the project: it would be an incentive for users to provide even more information about themselves.
His grandiose plans didn’t end there. Once the system was created, it wasn’t hard to imagine it would be self-perpetuating. Not only would there be an active discussion community around it, but the possibilities for collaboration with other similarly minded non-profits were enormous. This had the chance of becoming a clearinghouse for all things related to rights preservation, everything that ACCL had hoped to do, but on a much bigger scale. And Stephen would help architect it from the beginning.
The magic he and Ubatoo brought to this project was how these analyses would begin. All he needed to initialize the system were lists of “seeds.” These were lists of any entities Ubatoo tracked—books, web sites, people, events, profiles, discussion groups—that were labeled as being “of interest” to any government agency, say the NSA. From this small set, he could “grow,” algorithmically of course, these seeds into a full set of profiles and individuals that were at risk of being labeled as “interesting” as well.
Fortunately, his work with ACCL had previously provided the necessary seed set. From the book list, he had found the people who read them. From the people who read them, he had found the web sites they frequented and built out their complete profile: Lucy. Then, he simply identified a larger set of people who were like Lucy—the list of 5,000 people he had already given to Sebastin. Everything was connected.
In academic circles, similar problems arose in statistics and artificial intelligence literature at every turn, something that had been the basis of decades of research in universities across the globe: the problem of extrapolation. Extrapolation, simply put, was the prediction of outcomes outside the range of the available data. Think about chickens. When you see a chicken on the side of the road, you assume it wants to get to the other side because every other chicken you’ve known has wanted to get to the other side. It’s a guess, just a hypothesis, but you can’t know for sure because you don’t know about this specific chicken, only of others that you think are like it.
So who cares about extrapolation? Warning sirens are tripped when it’s applied to people. Applied to people, it invokes ugly terms like “stereotyping” and “racial profiling.” Irrespective of the application area, however, the need for the tools to perform this analysis existed across all of Ubatoo, from advertising and tailored-searches to the recommendation of videos to watch and web pages to read, to how to organize your e-mail, and even what groups you might like to include in your social networks. It was an omnipresent form of analysis constantly being run, over and over, incorporating all the new information as it became available. The goal was to fill in holes in the knowledge about you by finding people who ma
y be like you.
To start, Stephen had already created a list of 5,000 people who were likely targets to be placed on watch lists. He meticulously filled in the profiles of these 5,000 with every piece of information he could find about them: what they searched for, what time of day they searched, what they bought, which forums they read, who their friends were and when they called them, what entries they put on their calendars, what files they had stored, how old they were, and even what their e-mails contained. He knew these people.
But it was only 5,000 people. He needed to apply what he knew to the more than 200 million people online in the U.S. There were two facts to remember when figuring out what these 5,000 people had to say about 200 million others. First, the simple reality expressed in the statement that “no man is an island”—everybody and everything was defined by their connections. Second, the expression of being “just another cog in a machine” needed to be updated. A web of shared interests, shared friends, shared traits, shared patterns existed between all people. This web, or more precisely, this graph, and one’s position in it, defines each person. Not cogs in a machine, just points in a graph.
Imagine yourself as a little black spot on an enormous blank sheet of white paper. Your mom is another little black circle, placed close to you. You call her. When you do, imagine that you’ve drawn a thin line between you and her. Next, you send her a photo attachment in your e-mail. Better make that line a bit thicker. Buy her a present online and have it shipped to her? Make that line thicker yet—maybe you should use a marker now.
She calls your Aunt Theresa in Florida who appears as another little black point on the piece of paper. A thin line between your mom’s point and Theresa’s point materializes. Theresa receives an e-mail from her son, Antonio, apologizing for not writing enough. He’s in Spain, doing his semester abroad. The line between Antonio and his mom, Theresa, grows thicker. Antonio chats with his girlfriend, Sarah (another point), at the University of Maryland, over some instant messaging program; their line grows stronger between them every minute they are online together. His girlfriend tells him that Amber (another black point), her twin sister (a strong line appears between Amber and Sarah), has just gotten her acceptance letter from Dartmouth and is starting next year. Months from now, Amber enrolls in a political science course in her first semester. Her textbooks, ordered on the web, arrive on her doorstep. She had ordered all the required reading and the supplemental reading, too: five textbooks and dozens of essays and novels.
The Silicon Jungle Page 22