The slide that Atiq had been waiting for, “Touchpoints: Understanding Our Users,” appeared on the screen. Xiao spoke, “The Touchpoints project, led by Atiq Asad, typifies what makes Ubatoo truly incredible. The data mining group has been working for over two years in creating the infrastructure to, quite literally, harness the power of tens of thousands of machines to search for patterns—patterns in purchasing histories, in web searches, in e-mails, and in chat sessions. To date, in its pilot tests with some of our top advertisers, the revenue impact is estimated to be $65 million last quarter, and that’s before it’s even fully launched. Each one of you should find out more about Touchpoints; we’re going to be integrating it inside all of our products. And to think, all of this was done with only eleven people in the Touchpoints group. We had hoped the group would be forty people by now; imagine how much more they could have accomplished. But let’s keep our fingers crossed; hopefully we’ll see that potential realized next quarter. Let’s give Atiq and the Touchpoints group a hand for their incredible work—you’re about to know more about our users than you ever have before.”
The room reverberated with the boom of sudden applause. These weren’t the half-hearted courteous claps of disinterested employees. Far from it. Engineering and scientific achievements were always well received, and when large monetary numbers were associated with them, they garnered even greater enthusiasm. Atiq suspected that the rest of the Touchpoints team would be delighted by the exuberant response—both from Xiao and the audience. Despite the accolades from employees as they left the auditorium, Lynn and the other VPs who were more versed in the nuances of the passive-aggressive Xiao-speak, were more consolatory in their praise, “You’re doing great work. Don’t worry about Xiao” they assured him. Truth be told, Atiq wasn’t worried. This wasn’t the first time his expectations had been too high—and there was a high probability, statistically speaking, that it wouldn’t be the last.
-CHECKING IN-
March, 2009.
“Rajive here,” the voice on the other end of the telephone line announced.
“It’s Sebastin, from ACCL.” The line went silent. “Hello, Rajive, you there? This is Sebastin Munthe from American Coalition of Civil Liberties, ACCL. Hello?”
Rajive finally spoke. “Yes, Sebastin. I know who you are. You’re weeks late, again. I need you to be on time for this to work.”
He was doing Rajive a favor, not the other way around. It was frustrating working with him, barely worth the meager amount they were paying. “Do you want the update or not?”
Silence again.
Sebastin continued, “I finally got a hold of Dr. Atiq Asad at Ubatoo. Unfortunately, I don’t think that’s going anywhere. Everything I tried led to a dead end. Either he wasn’t getting any of my hints, or he wasn’t willing to ‘help.’ I’m going to have to find someone . . .”
“Do you have a plan yet?”
“I’m getting to that, Rajive. Ubatoo’s having some kind of intern contest this week. Atiq invited me to meet the interns at a party they’re having afterwards. Hopefully one of the interns will work out for us.”
“Names?” Rajive demanded curtly.
“I don’t know. Like I said, they haven’t even been hired yet.”
“Will they have access to all the data you need?”
“I imagine they will. ACCL worked with a few Ubatoo interns last year. From what I recall, they were plenty resourceful. I can’t imagine it’ll be any different this year. Besides, they will be part of Asad’s group, so I’m betting they’ll have access to everything. I’ll know for sure once I actually talk to them.”
“Any timelines yet?”
“Rajive, listen to me. I haven’t even met them yet. What do you want me to tell you? I can’t give you a firm timeline. I’ll call them a few weeks after I initiate contact. If you want this to work, I can’t scare them off. I’ll give you more details the next time we talk.”
“That’ll put you almost a month behind schedule. Any good news to report?”
“No.”
“Try harder with Atiq. He’s likely still your best shot. Make inroads with the interns. And, Sebastin, make your next update on time.”
“Always a pleasure, Rajive.” Sebastin hung up, exasperated.
Rajive left the conversation pleased with himself. In person, he usually had to play the good cop, but on the phone, he could do whatever he wanted. It was definitely more fun being the hard-ass. Despite any indications he may have given Sebastin, everything was fine.
-WORKING 9 TO 4-
March, 2009.
Wednesday, 9:00 a.m. It was the sort of visit that one actually brought cameras to and literally wrote home about. Few visitors were allowed in through the guarded gates without an explicit invitation. But, today, over 600 potential interns were being herded around the “grounds,” as Ubatoo’s campus was called, and taken on guided tours through all the usual tourist stops created for just this sort of thing. At every stop, whether in front of one of the five-star cafeterias, or in front of signs with names designed specifically with tour groups in mind (“World Operations Monitoring,” “Cyber Crime,” “Web Memory,” “Advanced Research Group”), flashes of cameras were steadily seen.
Of perennial interest to tour groups was the stop in front of two very large bulletproof windows. Here, onlookers were given the opportunity to see what most people not on the tour would find unconditionally dull. In the room beyond the glass were racks of several hundred computers. In there, the tour guide explained, 70 percent of Ubatoo’s public e-mail flowed through these computers. If you had ever used Uba-Mail, your message had flown through here before finding its way to its recipient. Despite the “No Photographs” signs, it was too hard for many tour members to resist, some hurriedly trying to snap pictures through the glass. Many had used Uba-Mail since it was first launched, and this was just one of the many things they wanted to memorialize with a photograph.
The more technically savvy members of the tour were less impressed, as it seemed highly unlikely that any machines “on the grounds” would hold such sensitive data as e-mail. It was much more likely that the data was spread to undisclosed datacenters, warehouse-like complexes with thousands or tens of thousands of computers, in remote locations around the world. And, of course, they were right. The computers behind the glass had been in the process of being decommissioned when one of the junior members of Ubatoo’s public policy team suggested this exhibit to impress a tour group of first graders—or so the legend goes.
Eventually, the tour guides deposited their charges in Ubatoo’s largest auditorium—a large room that the CEO, Xiao Ming, had personally taken an interest in creating. Like all the excesses of Ubatoo, this room, its cost, its designer, and its construction, took on a life of its own in the local newspapers of Silicon Valley—a geek’s episode of MTV Cribs. The peculiarity of the ornately decorated room, which gave the overwhelming impression of solid gold, was made complete with chandeliers, oversized drapes, and enormous oil paintings on three of the four walls of the Ten Tigers of Canton—a group of ten top martial arts masters in Southern China. The fourth wall consisted solely of a giant white projection screen, in front of which the speaker stood and any presentations took place. The room’s existence was, in every way, an ode to Xiao.
In time, Xiao’s excesses were forgiven, and the room became “Xiao’s Ballroom.” Xiao’s only comment on the room was to say that he wanted a place to entertain in style when actors came to share their worldly insights, charities stopped by to raise money among Ubatoo’s employees, or when, more importantly, the media was invited for a product launch.
As Stephen entered this room, he faced rows of cafeteria-type plain white tables set up with LCD monitors, keyboards, and computers. The tour guide instructed them to find a seat at any empty computer and get started filling out the forms displayed on the screen.
“A bit of advice. Get comfortable with your surroundings,” the tour guide said. “You’ll be
in here for a while.” With that, she left Stephen and the rest of the group standing in the entrance way. The room was strangely quiet given the number of people in it. A few times the staccato of a quick laugh broke through above the din, but these were few and far between. The members of Stephen’s tour group, who had been so closely huddled when the tour guide departed, soon dispersed in every direction.
Ubatoo was infamous for its intern selection process. Anybody could apply, whether a student, an experienced professional, or something in between. Based on resumés and entry questions, the initial screening was conducted. Those who passed were invited to participate in today’s event. Here, the remaining candidates were put into a single room and given a series of questions to answer. Sometimes the questions were programs that had to be written or sometimes they were riddles that required finding the one spark of insight that would unravel the entire puzzle. The answer was entered on the computer, and the cycle continued, for hours on end. The top scoring competitors were offered internships, and the others were sent home. Although nobody as yet knew how many people would be selected, it was a foregone conclusion that the competition would be difficult. No matter who you were, most likely you wouldn’t be coming back.
Stephen knew he should be worried about his chances, but he wasn’t. He hadn’t traveled more than a few miles, knew he was good at what he did, and was comfortable with the fact that he wouldn’t make it through. Moreover, he still wasn’t convinced that transitioning from running a company to being an intern wasn’t foolish. But Molly would be happy that he tried, and so would he.
Each computer screen offered a few multiple choice questions used to determine which department at Ubatoo was the best fit for the candidate—Ubatoo’s equivalent of Harry Potter’s Sorting Hat. After answering those, there was little to do but anxiously wait. The person seated in front of him was animatedly expounding to his neighbor about how the questions asked on the screen had a deeper meaning and that Ubatoo was probably already psychographically profiling the contestants. The ongoing conversation to his right, although not quite as overtly paranoid as the one in front, had its own set of quirks. Someone was describing why he was expecting questions on aerial image processing, and talking about satellites and Ubatoo’s ability to use the imagery to create detailed models of people’s houses. “It’s truly wonderful,” he continued in a thick accent, “I worked on this for my Ph.D., but I never thought anyone would actually be using any of it.” The person he was talking to was plainly waiting to tell his own tale, but by this time Stephen had stopped listening. The conversations weren’t helping his nerves.
On the far wall, on the stage in front of a large white screen, a man approached the microphone set up in the middle of the stage. “Good morning, everyone,” he started. “My name is Atiq Asad. It’s my pleasure to welcome you to our 5th Annual Internship Competition.”
The room didn’t just quiet down, it fell silent instantaneously. Atiq continued, “You all should be proud of making it this far. From the more than 3,700 applicants we received this year, based on your resumés and the sample work you sent us, we’ve invited 619 to take part in today’s competition. It’s truly an illustrious crowd; you have an amazing set of peers sitting with you. The brain power in this room is the stuff that Ubatoo was built on. It’s what keeps us always inventing, always creating, and always being the most innovative company in the world. Over 60 percent of the people invited have completed, or will complete, their Ph.D. this year. But don’t think that a Ph.D. is a requirement for success. Last year, one of our hires had not yet graduated high school. This year, we have two high school students competing. Next year, who knows, maybe elementary school. It’s truly an eclectic, brilliant crowd. Welcome to each and every one of you. Now, I’ll hand it over to Lynn Wiser, co-organizer of today’s event, to fill you in on how you will proceed. Again, good luck. I look forward to working with you.”
As Atiq left the stage, a brown-haired woman in torn jeans and a yellow baseball cap replaced him. “Welcome, everyone. Let’s get right down to business, shall we?” Her words began before she even reached the microphone. “By now, I hope all of you have finished answering the questionnaire on your screen. We used this to help determine which group you would be appropriate for. This year, based on your preferences and our needs, we’ll give you a set of tasks to accomplish. You figure out how to solve the task, design and write the program, gather the results, and submit them—as fast as you can. At 4:00 a.m., your results will be automatically analyzed and the scores tallied. The tasks chosen are as close to real world as possible. They’re the type of problems our groups are actually working on and solving today. You can use any tool you have to answer the questions. I don’t care if you phone a friend, use a lifeline, talk to the person next to you, or search the Internet. Do whatever you want; your answer is all that matters. Let me join Atiq in wishing you good luck.”
“By the way, the fridges in the hallway are stocked with all the caffeinated drinks, juices, waters you could possibly want. Coffee, tea, whatever—it’s all out there. You know the drill. Lunch and dinner will be served in the hall, starting in a few hours. The bio-break rooms, umm, excuse me, restrooms, are down the hall. Enjoy this, people; this is supposed to be fun.”
-PREDICTING THE FUTURE
AND 38 NEEDLES-
March, 2009.
The computer screen had come to life. Stephen could no longer hear Lynn or anyone else talking. Having been assigned to the data mining group, he now stared at the screen intently, puzzling over the first task:
Data Mining Task 1: Predict the Future
At Ubatoo, we strive to know our users better than anyone else. Last year, we launched a series of cell phones with the ability to browse the Internet on-the-go. To help our users, we would like to predict what they are going to search for before they type it. How good are you at predicting the future? Design and code a system that predicts what a user is likely to search for next.
You have access to a month’s worth of logs: all the search queries that 50,000 of our users typed into their cell phones. With each query, you are also given the geographical location of where the user was at the time the query was made, the time the user made the query, and the query itself.
Your TASK: Through the analysis of the logs provided, as well as any other information you can find, you must predict what the next ten queries typed into their cell phones will be for all 50,000 users. Points will be awarded for the number of matches between your prediction and their actual queries. We suggest spending no more than four hours on this problem.
This may have seemed more solvable when he was in his graduate student mode, but plugging in keyboard cables and fixing fax machines for the last two and a half years had done nothing to prepare him for this assignment. The only consolation he had was that he knew the task was do-able, or they wouldn’t be asking him to do it—hopefully. It wasn’t the number, 50,000, that was troubling. If he could accurately predict the search queries a single user would type, he could accurately predict them for 50,000. The problem was in developing the algorithm to make that first, correct, prediction. All the answers were in the logs kept of everyone’s searches. It was just a matter of figuring out how to extract the right set of information from the data—in under four hours.
Let’s start with the basics. Going back to his courses in grad school, Artificial Intelligence 101: To predict what a subject will do next, find other subjects like him, and see what they did in similar situations. He just had to match each of the 50,000 users with others who had also typed in a few of the same queries. By looking at similar users, he could find similar interests and other queries they might type. Once he had the idea, writing the program was easy; he was done before an hour passed. As he examined a few of the predictions, some seemed possible, but most seemed poor. For a moment, panic set in; he thought about trying to fix them manually. After fixing fewer than ten of them, he verified what he had known all along. There was a rea
son they had asked for predictions for 50,000 people—to ensure that any manual approach would simply be a disaster.
Stephen rose to get a drink, and glanced around the room for the first time. It was a scene he hadn’t been a part of in years, people working frantically, absolutely oblivious to their surroundings, and all of them here because they wanted to be, not because they had to be. He looked for telltale signs of stress—anyone with their head in their hands, or leaning back in their chairs—but noticed none of this. Maybe it was just too early in the day—though he had hoped to see some cracks in the competition.
He looked back down to his screen, already filled with horribly written computer code from the last hour’s work. It was ugly, it barely made sense, and it certainly was nothing to be proud of. He knew his college professors would probably take back his degree if they could see the mess of code and logic he was stumbling through right now. He also knew it was pretty unlikely that his undergrad professors would have made it into Ubatoo. Need to stop getting distracted with these thoughts, he cautioned himself. He sat back down without getting the drink. He put his hands on the keyboard, fidgeted with keys, and started thinking in earnest about what to do next.
When the epiphany did come, it was, as it always seems to be in retrospect, obvious. The fact that the users were on cell phones meant their location was probably an important clue. People searched on topics that were relevant to them, and if they were using a cell phone, they were traveling, and if they were traveling, then their location was probably important to them. If you were in California, you were more likely, based on empirical data, to search for sushi than if you were in Nebraska. Now the task became easier since the location of every user was recorded. When he searched for people who were similar to another, he added location to the profile he had built of the users (i.e., the only way Jane would be considered similar to John is if Jane was in a similar place, geographically speaking, to John).
The Silicon Jungle Page 4
