By Sean Patrick Cooper
Since early-March, vague rumors have circulated online about the validity of Google producing a new smartphone Application capable of predicting the day you will die.
The rumors, it seems, are true. Two nights ago, I sat down to discuss this App, known as Word Count, with a friend of a good friend of mine, who works at Google. To protect his job security, he spoke to me on the condition of anonymity. We met at a bar on the lower east side to discuss how the App works, Google’s confidential in-house beta-testing of the App on its employees, the effect the App has had on his co-workers, and what a public beta-release might mean for the general public.
Below, you’ll find the full transcript of our conversation. Only slight adjustments have been made for readability and clarity.
8:38PM | NYC | Interview w/ An Anonymous Google Employee
Diane Page: When we spoke this afternoon on the phone, you mentioned that Word Count could empirically predict a person’s death. How is that possible?
Google Employee: Well—so, there’s a really long answer here, and it involves a lengthy discussion of algorithm analysis and other—
DP: boring stuff?
GE: [laughs]. Yes, a lot of it is. The short of it, and this is from a limited second-hand knowledge of it. I mean, I’m in the graphic design department and so I don’t deal with this on a day-to-day basis. My education comes mostly from lunchtime chats or talking in the hallway with the people who actually really know the algorithm and analysis stuff inside and out. That said, from what I do know, I can tell you that with enough data, an application or a piece of software can predict a lot about a person’s future. We’re pattern-oriented creatures, and we leave these quote un-quote pattern prints all over our digital communications. Word Count detects and decodes these patterns.
DP: So it can predict when I’ll die based on how many times I Google myself?
GE: Kind of.
DP: How is Word Count monitoring for these death-stamp patterns?
GE: It’s not looking for any, like, big clues that would indicate if a person is heading for some horrible, incurable disease in their late 50s or something like that. It’s more principle driven.
DP: It’s a morally stoic smartphone application?
GE: I doubt it. Word Count works by keeping tabs on everything you do online. When you have Word Count on your phone and in your Google account, every keystroke is up for grabs. It’s crawling, basically, for every message sent or received through Gmail and Gchat, including video too. It transcribes the audio from your video chats and analyzes the conversations. It monitors web searches and web history, so it watches what you search for and how well you find what you’re looking for.
DP: How well? It can judge that? It seems like using the Internet with Word Count would be some sort of sport, like competitive Google searching.
GE: When you start to use Word Count, you do become competitive, in a way, with yourself. Maybe with others, too, if you see how well they search. But people are already competent at searching. So it’s mostly going to be judging degrees of accuracy amongst the accurate. The goal of the App isn’t really sport, though, its efficiency and you want—
DP: To find things faster?
GE: It’s more about finding things better. When you first get Word Count, you have 40-days of data collection to—
DP: This is what everyone in your office went through, the trial testing?
GE: Right, so, for all of us at Google, basically 5,000 of us received an invite to test out the App. I think something like 96% accepted the invite. Only those about to go on long vacations or maternity leave passed. It was a big deal to get an invite; there was a good deal of buzz about Word Count before it was officially going down.
DP: What were people saying?
GE: There was just a lot of talk about how the engineers cracked some kind of code that could predict when a person was going to die. They’re a strange group, the engineers. I’m friends with some of them, but they mostly keep to themselves, socialize a lot together, stay late all the time. At parties most of them huddle in their own little engineer pods.
DP: You had a nickname for them on the phone…
GE: Right—The Grim Reapers. Once word leaked a bit, people started to call the engineers in the Google Labs Grim Reapers.
DP: The Lab is autonomous from the normal search engine operation?
GE: Partially. There’s some overlap but the Lab mostly works on new, experimental projects. Like Gmail was a Lab project, lots of testing with company employees before it went public and became supported by the primary engineer offices.
DP: And you were part of the Word Count testing?
GE: Yes, I still am.
DP: Okay, so tell me about the data collection you were mentioning.
GE: Right, so, at the start, you basically sign off on this waver to give the software access to everything. After a week, you forget that your phone calls are being monitored.
DP: That’s reassuring.
GE: So it’s basically doing that and monitoring everything else, making connections, making assumptions. What it’s trying to do, as it was explained to me, is measure how efficiently you communicate your wants and needs.
DP: This sounds like a Dr. Phil session.
GE: I don’t know who that—
DP: It doesn’t matter.
GE: Okay. Well, so, the example I was given was this. Let’s say you’re trying to rent a cabin in Maine for a week of skiing. You start searching for rentals, emailing with property managers, send texts to your girlfriend about departure times or whatever. You get a receipt in your email for the cabin, then you’re on the road and get to Maine. From there, you buy stuff, send out emails, etcetera etcetera. All of that can be quantified, like actually measured by its digital volume, based on when you started looking for a cabin and ending with when you actually got there.
DP: What if I never make it to Maine? What if there’s a snowstorm or my car breaks down, something outside my communications ability, and all I ever get is a black-snow weekend in New York?
GE: Word Count would take it into consideration, I think, and not give too much weight to your intentions. It knows the weather forecast, the basic measurements of the world via the Internet, just as you would. But say you got there. Then it can measure your efforts in satisfying your original intent of getting to Maine.
DP: Like the perfect boyfriend that doesn’t exist, this can clearly detect my intentions and hopes and then know the most satisfying way to attend to those intentions?
GE: [Laughs] It doesn’t know the best way to satisfy them, but it does measure how well they are satisfied. Think of it like a race. The start point is “book my vacation.” The finish line is “I’m on vacation.” How quickly can I do that, and how many words did I use to communicate to accomplish that? That’s what’s being quantified and crunched.
DP: But so what, then? So it knows it took me 5 days and a whole lot of email and craigslist to get a vacation booked. 30 emails instead of 20 emails doesn’t mean I’m going to die 10-years sooner.
GE: In a way it does.
DP: You understand how ridiculous that sounds.
GE: Look, it’s more complicated than that. But I agree. We all thought it was nuts when it started in the offices. This was one of those Lab products that’d never make it to the public, like the genealogy archive where you submit a DNA sample.
DP: No way.
GE: There were literally meetings about this.
GE: I know right. And this was what we thought Word Count would be, just another ridiculous Lab product. But then that woman in the Michigan accounting office died on her Estimated Time of Death day and everyone—
DP: You’ll have to clarify that for the interview because—
GE: Right, sorry. So, Estimated Time of Death, or ETD. It’s the name of the algorithm the Grim Reapers whipped up. It’s also the driving force, I think, behind Word Count.
DP: This is based on the Communications Principal of Efficiency you talked about before?
GE: Yes exactly. A few years ago, it was a big deal to a really small group of academics and people who care about this kind of thing. Dr. Sun became a sort of minor academic celebrity, if there can be such a thing. She tested dolphins and then people and came up with this basic idea that the better you communicate, the longer you live. The Reapers don’t have the ETD down to a specific hour of the day yet, but I hear they’re working on it.
DP: Also comforting. You talked about a Daily Average, too?
GE: Yes, right. So the App [editor note: GE is showing DP the App on his phone] has three sets of numbers. Top number is your Word Count, how many words, on average, you’re expending in a day.
DP: You’re talking about words used online?
GE: More than that. This is a prediction of all the words you use total. Like all the words we’re speaking right now, it includes that in the count.
DP: That’s not possible.
GE: It is when you see it. That’s what the data collection is for, to determine your patterns. After a few weeks, all that stuff shakes out, how you communicate across platforms, in different contexts. After about 12-days, the App has your whole set of numbers already predicted. The other 30-days are just fine tuning the predictions in case you had a bad week or were sick with the flu.
GE: And you see this everyday, how many words you use at Starbucks and in your work emails and calling your mom?
DP: Everything is totaled together. And every morning, there’s a new daily Word Count. It’s usually pretty close to the day before. Ideally it’s a little smaller. I’ve gotten mine down about 2,000 words since I started using it.
GE: You actively try to manipulate it?
DP: Yea you sort of have to. You can’t see this thing, knowing what it means, and not try to play with it. That middle number, that’s how many words I have left, forever. Like, that’s it. After that’s down to zero, I’m dead. And then the bottom numbers, the date by which I’ll no longer be alive.
GE: Your ETD?
DP: My ETD.
GE: But it’s an estimate—not any different from a horoscope or palm reading.
DP: Before the deaths, mostly everyone except the Grim Reapers thought that way too. We kinda joked about it after that 41st day when all our Apps were activated. We called it our “impending doom,” and the “Heaven or Hell Upgrade Date.” “Life—version 2.0.” You get the idea. But then the Michigan lady croaked one day from her ETD.
DP: It was a day off?
GE: Sorry, it was basically the same day. She died in her sleep the morning after her ETD, like 2:00am or something. Heart attack, totally peaceful. Her husband woke up to a dead 53-year old wife. I heard she was obese, but still, only 53. There was a rumor that the Grim Reapers got anxious that they missed it by a day.
DP: Yes, I hate it when my death predictions are off by anything more than 12-hours. Can you explain the benefit of this to me? I don’t see why Google would want to launch a phone App that’s going to freak so many people out. I can see certain people really loosing it over this.
GE: Well it’ll be an invite-only App for a while, which means some pre-screening. But even a fully public user-base isn’t that much of a gamble. If you look at it monetarily, it makes some sense. Literally millions of people use Google’s stuff for free. Those users put a big strain on servers, on Internet lines, on customer service. Right now, there’s seven fully fledged call centers around the world dedicated to nothing more than calls and emails for Google customer support and 60% of those quote un quote customers don’t pay a dime. So if you can get even 10% of those millions of people to be more efficient with their web searches, with their emails, chats, and everything else, then you’re saving money. Plus, Word Count should make some people live longer. So then these people are buying more, over time. And, most importantly, the Word Count users can be billed to advertisers as more valuable targets.
DP: Google users are encouraged to stick around for a few more years and then Google makes more money advertising to them?
GE: Look, if you’re an efficient Google searcher, then that means you know exactly what you want when you’re Googling. Meaning, a Word Count user is going to find that French press coffee maker a lot faster than some high school math teacher in Arkansas ambivalently Googling for ‘glass coffee machine.’ Google can sell Ad Words for a higher price to French press companies if the ads are going to be put in front of Word Count searchers only. The ads are more valuable to Google and the company because the click through to purchase rate is so high.
DP: It just seems like such a gamble. The potential blowback outweighs the money Google would make.
GE: Well this was partially why they went with such a big private in-company beta test. 5,000 invites is a lot of people for a new Lab product and I think they did it because they needed to see how people would react.
DP: What was the feedback, in general, from beta-testers at the company?
GE: I guess that’s still being determined, to be honest. It was sort of split after that second death.
DP: This was the college kid?
GE: Right, so, once the Michigan lady died, people regrouped really quickly and dismissed the whole thing as a one-off fluke. Total coincidence, nothing to take seriously. But then word got out that there was this intern in the San Fran office with an upcoming ETD. The kid’s 22-years old, apparently a former high school track star with some state record for the mile, and interning at Google for his final semester of college. So he’s in incredible shape, peak physical condition, and word starts to circulate that he’s got a legit ETD about to go down in, like, 2-weeks.
DP: When was all of this happening?
GE: Right before New Years.
GE: So people start saying, ‘okay, if this kid goes, then this might be legit.’ Other testers were dismissive, but even those people, the ones who said, ‘Oh, forget it, this is bullshit,’ you knew that they were looking a little bit closer at their ETDs, waiting to hear what happened to the intern.
DP: So what happened to the intern?
DP: The SoHo marketing department started an office pool with like three grand to the winner but then it got shot down by HR, which was good, I think, because that was done in poor taste. Anyway, it’s a day before his ETD, and everyone that’s a beta tester and even now some others who are in the loop, everyone’s a bit quiet around the office. Sort of the quiet before the storm kind of thing. Which, ultimately, didn’t last long because the next day it spread like wild fire around the office that he was rushed to the ER for a brain aneurysm. 3-days later, he’s dead. 22-years old, peak physical condition, and the ETD was 3-days away from predicting his death.
DP: That’s impossible, though. How could the program possibly predict a healthy young male’s death, based on how well he communicates?
GE: To be honest, looking back on it now with some distance, it seems unlikely. A brain vessel popping isn’t something easily detected in a person’s speech patterns or whatever. But the timing of it, and the way everyone was waiting for it—at that point it didn’t matter how unlikely it was. It got people to take it seriously. And everyone started to look at the Grim Reapers in a weird way too. Like they were these interplanetary mystics or something, here to peep in on our GChats and predict our deaths.
DP: So is this good or bad for Google? Are they pleased enough with the feedback to go public?
GE: I’ve been told that they’re extremely pleased with it and that it will be out in some form or another pretty soon. I think the thing to keep in mind is that this is, in one very particular way, a tool. It’s a way to measure and improve a part of our digital lives. I think that’s how they want people to see it. It’s just an odd way to motivate people because we’re not a culture that talks openly about the inevitability of dying.
DP: Yes, it’s not a very popular topic. So did anyone else die during the testing?
GE: I don’t think so. Not yet. A friend of mine in my department has an ETD in like two years, and he’s only 33, so that’s something I’m sure he’s keeping an eye on right now. But I haven’t heard of anything sooner than that.
DP: And you’ve got 58-years to go?
GE: So it says. Since I started using the App, I’ve added 3-months to my ETD. I’m trying to push it up to 60-years soon. I’m aiming to live into my 80s, at a minimum.
DP: How do you improve your ETD? Less Googling, shorter emails?
GE: Not totally. It’s that but it’s more than that. I think it’s more about getting in tune with what you really want to say, what you really want to mean, and finding a way to communicate that, instead of all the extra crap that we often talk about instead. I’ve focused more on what it is that I need, and I think I’m doing that better now.
DP: What’s the mindset of your office after having, however accurate it really is, a sudden realization of when they might potentially be dead?
GE: It’s a weird shift in mood, and I think people are still grappling with it. There was definitely some blowback against the company, against the idea that Google would support this. Once the intern died, lots of people quit the testing program. Hundreds of people literally quit that day. And most everyone else, everyone else that didn’t quit, we were all a bit freaked out. Google and all our superiors, in every meeting, they always talk about ‘Don’t Be Evil.’ Well, maybe it wasn’t evil. But it was creepy. And it bugged people out. Some people even quit their jobs. Just put in their two-weeks and left the company. Others were upset they didn’t get invited. So it’s been a sort of weird atmosphere since it started.
DP: You were creeped out, but kept up with the testing?
GE: I stayed on, I’m still part of the beta-testing program. I can’t say if I was totally creeped out, one way or the other. I might have been in the beginning, because it was obvious that this was a little too powerful, this knowledge of something so big. But any knowledge, once you get familiar with it, once you get comfortable with its truth, you start to see things differently. You make better choices about your life. I think I’ve become a better person because of it.