Read an Excerpt
CHAPTER 1
At the Zoo
Early in the spring of 2005, Jawed Karim visited some friends in San Diego and met up with a high school pal at the city's famed zoo. While standing outside the elephant exhibit, Karim pulled out his point-and-shoot digital camera. "I had made up a list of random videos I wanted to do," Karim told me. "One was doing a silly introduction at a zoo." Karim switched on the camera's video mode and handed it to his friend Yakov Lapitsky, asking Lapitsky to film him. "All right, so here we are in front of the elephants," he says at the outset of the eighteen-second clip. "The cool thing about these guys is that they have really, really, really long, um ... trunks. And that's cool. And that's pretty much all there is to say." Yakov returned the camera to Karim and the old friends parted ways, neither realizing the puerile joke they had just recorded was destined to become internet history.
Five years prior in 2000, Karim, the son of a researcher from Bangladesh and a German biochemistry professor, left college early to join PayPal as one of the online payment platform's first developers. (He completed his computer science degree by correspondence a few years later.) It was there that he made two friends, Chad Hurley and Steve Chen, a friendship that would change the course of his life, my life, and maybe yours too.
By late 2004, Karim, Hurley, and Chen had all left PayPal and were meeting up regularly to discuss startup ideas. One opportunity that seemed enticing was creating a better video experience. In those days, you may recall, uploading and watching video online was kind of awful. For example, you generally had to host it on a website, if you had one, and provide people with a link they could use to download the file, praying they had the proper kind of video player installed to watch it.
Despite all the headaches people faced, a network of online clip sharing emerged, demonstrating consumer desire for a portable, discoverable, always accessible video solution. The reaction to the Janet Jackson "wardrobe malfunction" at the 2004 Super Bowl made this clear, as did the digital dissemination of an eviscerating exchange The Daily Show's Jon Stewart had with hosts of the cable news show Crossfire, a clip that, through BitTorrent and file sharing, had been watched millions more times online than when originally broadcast on CNN. Meanwhile, the tragic tsunami in the Indian Ocean in December 2004 proved that video-capturing devices like camera phones and digital cameras had become so ubiquitous that they could be used to docu- ment events all over the world with unprecedented scale and diversity of perspectives.
The pieces were there. Broadband access enabled higher-bandwidth streaming and Macromedia's Flash began supporting video, which allowed the YouTube team to ensure seamless integration into web pages. In the preceding years, sites like LiveJournal, Flickr, Wikipedia, and, oddly, Hot or Not had refined the basic elements of community-based online-content experiences. None were super complex in implementation. Karim, Hurley, and Chen would need to come up with an easy-to-use design, and then scale the infrastructure to support lots of people. Fortunately, Hurley had been PayPal's original designer and user-interface expert, and Chen and Karim had worked together on PayPal's architecture team, scaling the payment platform to millions of users. They weren't the only tech team with this idea, though; sites like Vimeo and Google Video had just launched. So, on Valentine's Day 2005, they quickly got to work, not even sure an infrastructure for such a site would be possible; hosting costs, the transcoding of hundreds of file types, and more presented legitimate challenges. Hurley later joked that had they known how complicated it would be, they never would have attempted it.
They weren't certain how people would use it either. They expected people might like to use it for video profiles about themselves and, famously, set it up as a dating site at first. In the initial design, you couldn't choose which videos you would watch; YouTube would pick for you randomly after you selected your gender and the gender/age window you were "seeking."
When deploying YouTube's first test server, Karim uploaded the video he recorded that day in San Diego. He titled it "Me at the zoo," because, well, leave it to an engineer to label something as efficiently and straightforwardly as possible. On April 23, the "Me at the zoo" test video became YouTube's first official clip. Two days later, Karim sent the following email to his friends:
From: Jawed
Date: Mon 25 Apr 2005 05:34:07-0400 (EDT)
Well guys, we just launched this site: http://www.youtube.com
Can you help us spread the word? Since we just launched, there are no girls on it ... YET.
Can you guys upload your own videos? Let me know what you think!
Jawed
But things got off to a slow start. There weren't many videos, and a number of them were merely clips of airplanes Karim had recorded. In an effort to increase uploads, they took out ads on Craigslist and put up flyers on Stanford's campus. They soon bailed on the whole dating thing because people weren't using it for that. They were using it to share videos of their friends and pets, funny sketches, unusual internet ephemera, and more. In June, the crew did some retooling, adding a "related videos" feature to help boost views, easier sharing tools, and critically, the ability to embed the YouTube player onto your own website.
The new features worked, and as the site began to grow they studied the metrics and discovered something interesting. "Every two weeks there would be one video clip, which would be a totally random clip, that just went through the roof and it would be so popular that it would just dwarf all the other video clips," Karim recounted in 2006. As the rate of uploads increased, so did the probability of hits like these, until the gaps between them started to shrink, and YouTube became, arguably, the fastest-growing website in internet history.
"Me at the zoo" was never one of these hits, of course. The vast majority of the video's tens of millions of views occurred after Karim's bad joke became a legendary artifact in the museum of internet history. "Me at the zoo" doesn't represent the razzle-dazzle of what would come later, but the simple, unassuming video is a perfect example of the kind of personal experiences the platform would come to facilitate over the years.
I've often wondered if, in the gaze of history, Karim wished he had said something different in that first video. Surely, with more than a decade's worth of hindsight and the knowledge of the impact of the thing he helped create, Karim would have wanted to express something more profound than a pretty stupid double entendre about elephant anatomy. Right? When I finally got the chance to ask him about it, he simply said, "I don't mind it being the first video. It does get the point across that on YouTube anyone can broadcast what they want and the community decides what its value is." That it does.
SERIOUSLY, WHERE ARE THE VIDE OS, AND WHAT'S A VIEW?
YouTube has more than a billion users, but very few of them actually understand how it works. Almost all of today's major online platforms are intuitively useful; the people who use them don't need to know much about how they actually function. British science fiction writer Arthur C. Clarke famously declared, "Any sufficiently advanced technology is indistinguishable from magic." If we are going to talk about the impact we have on culture through YouTube, you should know a few basics about what makes the magic possible, and it all comes down to the infrastructure that supports it and the measurements that make sense of it.
A group of middle school students once asked me a simple question: "Where are all the videos?" Turns out, the answer is rather complicated. To sort it out, I went to Billy Biggs — one of YouTube's longest-serving and most respected software engineers, whose duties include overseeing YouTube's technical architecture — and asked him to explain. "An easy way to say it is, well, they get stored on hard drives," Biggs said. "But my god, that's a lot of hard drives!" When you upload a video file, YouTube's servers process the video by breaking it up into pieces and transcoding it into an array of different formats optimized for different screen sizes and connection speeds. Watching a video on a 4K television, a laptop with a high-speed connection, or a phone in a rural area could all require different types of video files. These files then get replicated and stored on servers all over the world so that someone watching "Double Rainbow" in Johannesburg does not need to access the same video file as the person watching "Double Rainbow" in Los Angeles. The more popular the video is, the more times it gets copied to all these different locations, meaning that for any popular video there might be a thousand instances of it out there in the world. So in that sense, YouTube videos are, quite literally, all around us.
Biggs has been at YouTube since 2006, when the company was growing rapidly. It was a thrilling time, and he was excited to see things thrive but was also aware that there was a possibility back then that there might not be enough servers configured to handle the traffic, especially on the weekends, when people have more messing-around-on-the-internet time. "When you have a system that's growing so much, it's going to reach some new peak every weekend," Biggs said. He kept a laptop and a 3G Wi-Fi card with him at all times in case something broke and once had to fix YouTube from the back seat of a car while on the way to the beach with his friends.
Today, things are far less tenuous than they were in those days, but YouTube is still constantly growing, expanding, and pushing limits. That constant evolution tests technological, legal, economic, and creative standards in ways that one can be only so prepared for. This all stems from the reality that we, YouTube's users, are changing all the time too, with our desires and interests ever hard to pin down or predict. In order to adapt to us and what we want from it as effectively as possible, the platform must be able to measure the actions of people who use it. The basic building block of measuring videos on YouTube is, of course, a view. The number of views indicates how many times something has been watched. But what is an actual view? This is another thing I'm often asked, so I went to the source. "It's a playback of a video where the viewer actually intended to watch that video," said Ted Hamilton, rather unsatisfyingly. Hamilton, or "Hammy" as he is known to my colleagues, was the product manager for YouTube's analytics department in Zurich, Switzerland, for more than six years, making him an expert on how we organize the data we publish about videos, channels, and more. Measuring a view, it turns out, is far trickier than you might think.
It's not so much the technical challenge of counting the billions of playbacks that occur each day and processing them; YouTube's systems have gotten very good at that. Determining whether or not those play- backs were intentional is more difficult. If someone misleads you with a title or thumbnail, did you intend to watch that video? What if you visit a web page and they have an embedded video autoplaying? Would you describe someone who replays a video five times in a row over a few minutes as five different views? What if videos are autoplaying while you're in another room making dinner? Certainly we would agree that the playbacks generated by automated systems shouldn't count, right?
Some of these scenarios are sneaky attempts to make videos look more popular than they are and some are totally above board, but all of them require the application of statistical analysis to help prevent questionable activity from corrupting the data. "Through the billions of views that we have every day, patterns emerge," Hamilton explained. The view count remains the metric most people use when discussing how popular something is on the internet, but I found out early at YouTube that trying to understand video popularity based on views alone can be very deceiving. First of all, how something gets viewed is often much more telling than whether it was viewed. The proactive interest displayed through a search-driven view suggests some- thing different from the reactive interest of a subscription-driven one. When something is watched can give us a sense of velocity, which helps us understand the dynamics of a video's popularity. For example, in 2015, the video for rapper Fetty Wap's "Trap Queen" took months to pick up steam while Adele's "Hello" immediately drew a huge audience. Both were among the year's biggest singles but were embraced differently within popular culture. Perhaps most important is how long someone watched, which indicates not how enticing something seemed to be before you started watching but how engaging the video was once you did. The sum of the amount of time people spend watching something is called "watch time," and it's become a metric even more important than view counts.
Available to anyone who posts a video, the video analytics tool — for which Hamilton was, in part, responsible — displays all of this information and lots of other important statistics too. All day every day, YouTube counts the views, likes, shares, comments, etc., amassed by each video, and then the platform's servers summarize all the data for videos and channels in ways that allow people to slice and dice the important bits, like where their audiences lie and even basic age and gender information. Hamilton's job is to bring all the pieces together to make sure this activity is accurately captured. The data is important to viewers, to creators who use it to adjust their strategies, and to YouTube's own systems too.
You might also say that Hamilton's responsible for a headier task: registering and validating the reality of our existence in the digital world. For video creators, rich analytics data helps turn anonymous bits and bytes into a real audience made up of real people. "In the internet age, we're very detached," Hamilton said. "And there are people out there, but you can't see them. We provide the feedback that the people are out there and they're watching what you put out there and liking what you do."
All this information also helps me interpret trends on YouTube. Many YouTube employees and the systems they've engineered analyze this data every day to find ways to make the platform better and more useful, which means that it's our behavior — yours and mine: our tastes, our voices, and our passions — that actually shapes the YouTube we use today and the one we'll use tomorrow.
HOW YOUTUBE IS FO R YOU
When you visit YouTube's site or use the app, the first things you see are likely video recommendations, one of YouTube's most important features. These recommendations, while not always perfect, can be pretty damn effective. Even the classic entertainment world has had to acknowledge that. In 2013, the National Academy of Television Arts & Sciences in the United States presented YouTube with a technical Emmy Award for its video recommendations.
One of the guys who accepted that prize was Vice President of Engineering Cristos Goodrow. A mathematician by training, Goodrow began as a business and manufacturing optimization consultant before moving into the world of search engines, managing search for Amazon and then Google's product search. He now oversees YouTube's search and discovery teams, an army of engineers — including many of the ones I work with on a daily basis — who are responsible for building and maintaining the systems through which YouTube recommends videos or channels to you.
"The fundamental thing that few people seem to understand about YouTube is that the whole thing is a giant crowdsourcing exercise," said Goodrow. "People often ask me about 'the algorithms.' The best we could hope for in the algorithms is that they quickly and faithfully reflect back the tastes and interests of the crowd and apply that information to your individual situation." As of 2017, YouTube's discovery algorithms were being automatically updated with over eighty billion new signals — views, likes, clicks, and more — from the audience every day. This helps make YouTube different from any other entertainment medium we've ever had, an extension of our individual tastes and collective psyche.
YouTube is a kind of experiment in human expression, and none of it was by design, really. YouTube's founders, and the many employees who've since followed in their footsteps, did not adhere to a carefully plotted plan, but rather worked to magnify the behaviors and interests of those who used the service. In other words, the experience you have on YouTube has been shaped directly by our individual and collective activity.
(Continues…)
Excerpted from "Videocracy"
by .
Copyright © 2018 Kevin Allocca.
Excerpted by permission of Bloomsbury Publishing Plc.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.
