“It was like, wow, maybe we really should start a company now”
In a matter of a couple of years, the anarchic Web of Tim Berners-Lee was downright business friendly as Web sites took advantage of tools to keep tabs on their visitors and accept secure payments. These were important first steps in imposing some order, but they didn’t begin to address the fact that the Web itself was a shambles . . . by design. A growing, decentralized, seemingly chaotic mess, the Web had resisted attempts to be thoroughly mapped, either by people or by computers. By November 1997, there were an estimated hundreds of millions of Web pages,1 yet as Marc Andreessen observed early on, the Netscape development team simply skipped the hard navigation problems in order to get a browser to market quickly. “How do you search across the entire information space?” Andreessen asked rhetorically. “I don’t know. How do you know where you are going? Beats me. The sort of surprising thing is we ended up with something useful anyway.”2 To Sergey Brin and Larry Page, Stanford PhD candidates in computer science who met in 1995, such questions weren’t so disposable. If users didn’t have the means to discover what was on the Web, then it may as well not be there. How, Brin and Page asked, could the Web ever meet its potential as a vital new public resource?
At its start, the Web was just small enough to be approached on a personal level. Its contents could be organized by indexers, modern-day equivalent of librarians, who kept up with what new sites had arrived on the scene and discovered interesting themes and connections among sites. There were potential flaws with this system, which would be particularly obvious to computer scientists like Brin and Page. “Human maintained lists cover popular topics effectively,” they wrote, “but are subjective, expensive to build and maintain, slow to improve, and cannot cover all esoteric topics.”3 In Silicon Valley jargon, the problem with human indexers was that they couldn’t “scale” along with the Web. People were too expensive for the task, especially considering how little processing power they came with.
In a biting account of her time as an English major working at Facebook, Katherine Losse explained the importance of scalability to Silicon Valley. “Things were either scalable, which meant they could help the site grow fast indefinitely, or unscalable, which meant that the offending feature had to be quickly excised or cancelled, because it would not lead to great, automated speed and size,” she wrote. “Unscalable usually meant something, like personal contact with customers, that couldn’t be automated, a dim reminder of the pre-industrial era, of human labor that couldn’t be programmed away.”4 The answer to meaningful Web search would have to come through computers, Brin and Page were convinced, not squads of librarians. Up to that point, however, computers had been found deficient, too.
One notable attempt to search the Web with computers was AltaVista from the Digital Equipment Corporation. AltaVista began December 15, 1995, with an impressive 16 million Web pages indexed by its computers, which could be searched through to answer a query.5 AltaVista wasn’t very intelligent, however. It focused on keywords in Web pages, seeing how often and with what prominence those words appeared when deciding what results to return for a query. What keyword-based searches typically produced, in Brin and Page’s professional opinion, were “junk results” that “wash out any results that a user is interested in.”6 Not only was the technique unsophisticated, it was ripe for manipulation by unscrupulous Web sites, which added potential keywords to their text willy-nilly. If one mention of the word car on a page helped it show up in the results for a search about automobiles, the thinking went, then surely twenty cars would be even better.
Seeing a computer act so stupidly was an affront to Brin and Page. The tools of artificial intelligence, they believed, would allow a computer to understand what was on the Web and help guide users to exactly what they were searching for. Their insight was to use the clickable links between Web pages, rather than keywords. Links, it turned out, conveyed much more information about a site than parsing which words appeared most frequently. Each link announced, in essence, I’ve been to this other Web site and found something helpful there. In addition, the text underlined by the link explained what in particular was so helpful—one person’s attempt to map what she had discovered about the Web. Perhaps those millions of Web links and descriptions, interpreted by a clever algorithm, were all that was needed to produce an effective, automated online catalog. The best part was that those millions of links were just sitting there for the taking. Thanks to the Web’s open design, Brin and Page, or anyone for that matter, in most cases could freely collect as much of the material on the Web as they wanted, whether individual links or copies of entire Web sites.
Brin and Page viewed this project in strictly academic, philanthropic terms at the start. They were computer scientists planning to follow in the footsteps of artificial intelligence pioneers like John McCarthy, who was a living legend to Brin and Page, walking the same halls of the Stanford computer science department. Their search engine would be tackling familiar AI questions—how does a computer receive information about the world and can it be trained to interrogate the information it receives to make real-world judgments? Had all gone according to plan, the two young academics would have created their uncannily accurate search engine, published the results in an important academic paper, earned their PhDs, and become professors. At the same time, the search engine they developed—at first called BackRub and later Google, in reference to the absurdly large number called a googol—would remain noncommercial and freely available to the public through Stanford.
When Brin and Page reversed course, however, and reluctantly dropped out of graduate school to run Google as a business, they would be sharing with the world a different discovery—an important new technique for profiting from the Web. By this time, there were companies like Netscape and Microsoft, among others, that profited from creating the software used online. There also were companies like Amazon and eBay, soon to be joined by PayPal, which intended to take a cut from online commerce. Google, however, would pave the way as a big business that extracted value from the ordinary material folks posted online while they were going about their digital lives.
Executing Brin and Page’s innovative ideas for building a high-functioning search engine posed a number of challenges to surmount. The computers that “crawled” the Web, making copies of pages and collecting the links, were expensive, but Brin and Page were not above cadging what they needed by hanging out on the loading docks to learn who in the computer science department might be getting free equipment he didn’t need.7 Brin and Page then had to devise efficient methods for searching through the oodles of information they were collecting to isolate Web links. The last challenge: how exactly should the algorithm interpret those links?
Page and Brin, the sons of professors, used a technique that was similar to how the academic world evaluates a scholar’s reputation. Say there were two scholars you wanted to compare; each has articles you haven’t read or that were beyond your understanding. A natural approach would be to seek the opinions of experts in the field. You might even automate the process by sending out a survey to many scholars and then counting how often they had cited each candidate’s research in their own. The higher the total, the more important the scholar, you might conclude. If you wanted to get fancy you might devise some formula—an algorithm—that didn’t merely count the number of citations but somehow simultaneously factored in the importance of the scholars whose citations you were using. After all, if you were trying to assess the quality of somebody’s work, you would do well to assess the quality of the assessors, too.8 Think back to Frederick Terman’s time as Stanford’s provost, when he was determined to raise the university’s stature in comparison to the top schools back east. Armed with a slide rule and similar kinds of formulas, Terman had no qualms about determining the precise numerical value of scholars far from his area of expertise—say, a proposed new English professor or the entire history department faculty. With these tabulations at the ready, Terman would then confidently rank the hiring of this person compared to the hiring of that one.
In many ways, a search engine faced a similar predicament to Provost Terman’s. It, too, was trying to assess the quality of something it didn’t or couldn’t personally “understand”—in the search engine’s case, a random Web page written by a human being. Still, Page and Brin figured that with the proper algorithm a computer could rank the importance of Web pages and their relevance to any particular query. The algorithm treated links like academic citations: a page that was referred to by many different sites must be pretty important, especially if the sites doing the linking had themselves been judged important. Likewise, a page that was frequently described with the word baseball most likely had something to do with the sport. The algorithm for ranking Web pages’ relevance was given the name PageRank in a nod to co-creator, Larry Page.
PageRank didn’t just produce results far superior to other search engines’, the two wrote in an academic paper about their project, but it produced results that agreed with “people’s subjective idea of importance.”9 That is, the algorithm’s recommendations appeared to jibe with what you yourself would have come up with if you were a search engine. This was true even though Google was giving generically useful answers. The search engine didn’t necessarily know you from Adam, but imagine if it did? The more Google knew about the person searching—where she lived, what she liked to do—the better its search engine would perform when trying to answer user queries. No matter how intelligent an algorithm is, if it doesn’t know if you live near Berlin, Germany, or Berlin, Ohio, it’s not likely to give the most useful answer to the search “bookstore in Berlin.”
Google would become aggressive in compiling digital dossiers about its users to keep perfecting search, remembering each query and which results were deemed relevant. But even before such tracking became regular practice, Page and Brin used John McCarthy’s sprawling Stanford Web site to run an experiment on personalization. McCarthy was freely sharing his papers and reviews, and links to other Web sites, as well as his short, biting opinions about current events. Using all this material, the two created a filter personalized to McCarthy; not surprisingly, when PageRank applied that filter to search results they were likely to be much more relevant to McCarthy than those produced by the generic algorithm. Personalized search engines, the two wrote in a paper coauthored with their academic advisors, “could save users a great deal of trouble by efficiently guessing a large part of their interests given simple input such as their bookmarks or home page.” For example, “while there are many people on the Web named Mitchell, the No. 1 result is the home page of a colleague of John McCarthy named John Mitchell.”10 McCarthy had collected an unlikely honor: he was perhaps the first of a billion or more people to be profiled by Google so that it could generate personalized search results.
Page and Brin indeed are direct descendants of McCarthy and his artificial-intelligence-inflected computer science, although when the two were getting started in the mid-1990s, the field’s ambitions had been severely clipped and McCarthy was known as a legend rather than a cutting-edge researcher. Like David Shaw’s gifted investing computers, PageRank was a highly practical use of artificial intelligence techniques based on faster calculating and access to more data rather than a breakthrough in independent thinking. The computers running PageRank appeared to think like a person by working efficiently and incessantly as only a silicon-based machine can. “We don’t always produce what people want,” Page said early on. “It’s really difficult. To do that you have to be smart—you have to understand everything in the world. In computer science, we call that artificial intelligence.”11
Again, this application of AI may have fallen short of the high ambitions of those who believed they were on the verge of creating a new form of intelligence, but in terms of how we live the impact of the uncanny Google search engine was immense. The forty years between 1956, when McCarthy came up with the term artificial intelligence, and 1996, when Google arrived on the scene, encompassed a digital revolution. The pioneers of AI began their work on huge, expensive, memory-deprived mainframe computers. Only a select group of experts in business, academia, and government were capable of programming these colossuses, and only a minuscule group of MIT undergraduates—the hackers—particularly wanted to. The general public experienced computers from afar, and were quite suspicious. IBM was well aware of this suspicion, McCarthy recalled with regret, and didn’t conduct any AI research from 1959 to 1983. Their slogan attempted to lower the stakes: “data processing, not computing.”12
By the time Google was invented, however, computers had become epically faster and cheaper; memory was becoming manageable as well. The PC had inserted itself into daily lives. Under those conditions—let alone today’s world of pocket-size devices—AI research oozed with real-world implications. Anything that made computers act more “human” was of immediate interest either to regular folks in their daily lives or the elites in finance, as well as the military and government, which had highly specialized tasks for computers to perform. There was abundant enthusiasm for Shaw’s computer algorithms, which discovered and exploited minor inefficiencies in the stock market to make billions of dollars, as well as for PageRank, which for the first time allowed hundreds of millions of Web users to go exactly where they wanted online.
When Page and Brin started asking hard questions about search, seemingly everything online was taking a turn toward the crass commercialism best exemplified by the blinking banner ads that had multiplied across the Web. In 1993, barely 1 percent of servers belonged to a .com domain, meaning commercial; by 1997, that share was 60 percent.13 Web search was no different, filling up with opportunists looking to make a quick buck. Many of the leading companies preferred to manipulate results on behalf of advertisers rather than improve the underlying technology. In their paper introducing the ideas behind the Google search engine, Brin and Page express dismay that these companies wouldn’t share their raw data so that the two researchers could test their ideas. Not surprisingly, the companies that had the information considered it too valuable to part with and weren’t necessarily interested in what a couple of up-and-coming researchers might have to contribute.14 Google would be different, they promised, with its data organized so that “researchers can come in quickly, process large chunks of the web, and produce interesting results that would have been very difficult to produce otherwise.”
Entrepreneurs may have been content to treat search like just another service ripe for pillaging, but Page and Brin saw the stakes as so much higher. A search engine, more than other Web tools, only worked if it had a user’s trust, which required transparency. Too much of what a search engine did occurred out of sight. If it steered you away from a certain site to please an advertiser, how, precisely, would you know? Even computer experts were hard-pressed to determine if the commercial search engine they were using was playing fair when it published its results. Again, Google would be different, a project run by academics who were committed to improving search for its own sake, not to make a quick buck. There would be no ads. In a paper the two delivered to a conference in Australia in 1998, Brin and Page took pains to explain why advertising inevitably corrupted search technology and had to be kept out. They offered both ethical and practical arguments.
The ethical obligation to run a Web search engine without advertising reflected an academic’s belief in the importance of public access to information, which could be a matter of life and death. In a remarkable appendix to that 1998 paper, “The Anatomy of a Large-Scale Hypertextual Web Search Engine,” Brin and Page offered an example of the danger to the public from search results that were tainted by advertising. As a test, they typed in the query “cellular phone” at all the prominent search sites. Only the Google prototype, they reported, returned a top result that was critical of cell phones, specifically a cautionary study about speaking on the phone while driving. PageRank didn’t return the link in order to do the right thing, the two explained; it was simply conveying to its users what the Web thought were the most relevant links to someone interested in cell phones. The better question to ask was, Why didn’t the other sites link to that study? Page and Brin’s answer: “It is clear that a search engine which was taking money for showing cellular phone ads would have difficulty justifying the page that our system returned to its paying advertisers.”15
Practically speaking, as well, advertising led to an inferior product. Brin and Page didn’t point to obvious examples of manipulation like a search engine that promoted an advertiser’s business undeservedly to the top of results, or, as in the cell phone example, buried results that might offend. Instead, they were struck by the opposite effect: that a search engine might tamper with useful results to punish an advertiser. There was the case of one prominent search site that appeared to intentionally ignore an airline in the results of a search of the airline’s name. The airline had paid to have a prominent advertisement linked to such a query, and the owner of the search site must have feared that if the airline discovered that its name already appeared in front of users who were looking for it, then it might not bother to advertise at all. This meant, incredibly, that under certain conditions, more relevant search results were actually bad for business. What could be more offensive to a pair of artificial intelligence researchers than a system with a feedback loop that made things worse?
Brin and Page had put their finger on the Catch-22 of the search business: if results improved so much that they became uncannily accurate and precise—like a chess computer arriving at the single best move for a certain position—then advertising will have lost much of its purpose. You would be shown where to go based on the consensus “best result,” and thus should have little interest in hearing what an advertiser wanted to tell you. OK, there might be a few ads to introduce a new product, or to try to persuade you to switch between brands, but this wasn’t the basis of growing business. A search engine needed to sell something valuable—like reaching customers in a way competitors couldn’t—if it wanted to make a lot of money. The bad incentives were clear: search companies would stop trying to improve their services for business reasons, which is why Page and Brin toward the end of their paper made the following assertion: “We believe the issue of advertising causes enough mixed incentives that it is crucial to have a competitive search engine that is transparent and in the academic realm.”16
The Google search engine from the start was a monumental improvement on the other search engines and the good news spread across the Web purely by word of mouth, just as was the case with the Mosaic browser a few years earlier. Traffic to Google grew 50 percent a month, month after month.17 Here was another Internet-defining creation from students barely in their twenties, who were working out of a university computer lab and reaching a worldwide audience through the university’s wires and cables. Google was operating with a nest of different computers housed in cooling cabinets made out of Legos and through an Internet connection at the Bill Gates Computer Science building, a $38 million complex that opened in 1996 spurred by a donation of $6 million from Gates.18 When Google’s Internet connection appeared to reach its breaking point, a member of the team suggested switching a toggle so its computers could tap into the T3 line serving the entire university, which relieved the pressure for a time.19
A Web phenomenon like Google didn’t remain a secret for long to the university (with that toggle flip, the search engine grew to use as much as half of Stanford’s total bandwidth) or the news media or potential investors. What would the young computer scientists say to those who wanted to commercialize their magnificent new Web tool? Histories of Google tend to rush past what was an epic decision to abandon their conviction that advertising corrupted Web search. One account of these events put it this way: “Soon, the temptation to spin it off as a business was too great for the twenty-something cofounders to bear.”20 In his insider account of Google’s history, In the Plex, the writer Steven Levy doesn’t even credit the pull of temptation. “Page and Brin had launched their project as a stepping-stone to possible dissertations, but it was inevitable that they began to eye their creation as something that could make them money.”21 Inevitable.
Brin and Page have come to terms with their change in priorities by concluding that ordinary rules don’t apply to brilliant engineers like themselves: Google would somehow remain a research project with noble aims even as it transformed itself into a business with a duty to make money. One explanation for how they managed this particular form of cognitive dissonance can be found in Page’s obsession with Nikola Tesla, the brilliant Serbian-American inventor who died in relative obscurity. Tesla’s superior ideas in electricity and other fields lost out to Thomas Edison and others, who had a superior business sense. Tesla’s life became a cautionary tale. “I feel like he could’ve accomplished much more had he had more resources,” Page said. “I think that was a good lesson. I didn’t want to just invent things, I also wanted to make the world better, and in order to do that, you need to do more than just invent things.”22 A great scientist needed to tend to business, if only to be sure his ideas had a chance to take hold. Placating your investors—in Google’s case by accepting advertising tied to search results—wasn’t ethical compromise but a way to make the world a better place.
There is another way of explaining why Brin and Page reversed course. They were students at Stanford, after all, and at Stanford, you don’t look for investors, investors look for you. Marc Andreessen cited his own experience at the University of Illinois working on the Mosaic browser to describe what was so different about Stanford. “Had we been at Stanford,” he said, “of course we would have started a company. In the Midwest, no way. Not a chance.”23 As discussed earlier, Only when Andreessen flew to Silicon Valley did his entrepreneurial horizons broaden. There, under the guidance of a former Stanford engineering professor, Jim Clark, was he educated in the ways of investment and entrepreneurism. Together, they built a browser business, Netscape, which got the dot-com era started with a bang.
Stanford meant something very different to Brin and Page, who first met there in 1995, a year after Andreessen arrived. After impressive academic careers at the state universities of Maryland and Michigan, Brin and Page were joining the academically prestigious Stanford computer science department to complete PhDs. They had committed themselves to an academic life. They weren’t blind, however, to the entrepreneurial legacy of Frederick Terman, what Brin called Stanford’s “big history of company building.”24 Even before they became fast friends and research partners, there was an example of a wildly successful Web start-up on the Stanford campus, Yahoo, which was built by a pair of electrical engineering graduate students, Jerry Yang and David Filo, working in their spare time.
Yang and Filo created their first Web index in January 1994, which quickly became “Jerry and David’s Guide to the World Wide Web” and then Yahoo, to keep track of the technical papers online related to their PhD research. Other electrical engineers found the index helpful and sent suggestions of cool things on the Web; the mission and the audience kept expanding. “All of a sudden both of them went from doing their graduate work to adding Web sites to their list for eight hours a day,” Tim Brady, Yang’s college roommate at Stanford, recalled. “As chance would have it, their thesis advisor was on sabbatical, so there was really no one looking after them, so it all worked. Had their advisor been there, it might not have happened.”25
On the face of it, Yahoo, which incorporated in 1995, was a very different project from the expensive engineering projects that Terman had encouraged Stanford graduate students and faculty members to pursue at first within the university lab and later in the private sector. Those projects typically earned government research grants, which included overhead payments to Stanford, before being brought to market in partnership with large multinational corporations. In the case of Yahoo, Stanford graduate students were working on a marketable project that had nothing to do what they were researching—actually took time away from what they were researching. This development path raised an interesting question: would Stanford claim to own the index, since it was created on university equipment, even if it wasn’t related to university research?
Brady, a recent Harvard Business School graduate and the first outside employee at Yahoo, recalled being nervous about this central, even existential, question for the company. But Yang and Filo reassured him. “Stanford is very progressive in that,” Brady said. “Yahoo is far from the first startup that originated there and will be far from the last one. It was new enough, and it wasn’t a specific technology; it was a brand.”26 Unlike the University of Illinois, which didn’t bend to the will of its students, Stanford as designed by Terman was content to bet on its students, even when they were playing academic hooky. Brady explained the Stanford philosophy, as it applied to Yahoo. “They were smart enough to know that anything they would do to stifle it would kill it,” he said of the university’s administrators, “so their best hope was to just let it go and hope that Jerry and Dave gave money back later, which they did. They optimized their outcome, trust me.”27
Indeed, after getting the go-ahead from the university, as Yang and Filo had predicted, Yahoo grew through an early investment by venture capitalists on Sand Hill Road. In April 1996, when Yahoo had its IPO, it reached a valuation of more than $700 million.28 The next year, Yang and Filo each donated $1 million to Stanford to endow a professorship at the engineering school—at ages twenty-eight and thirty, they were the youngest donors to make such a significant donation since the school kept records.29 And today, Yang is on the Stanford board of trustees; he and his wife, Akiko Yamazaki, have pledged $75 million to the university to expand interdisciplinary studies.30 All was going according to Terman’s plan. Stanford was encouraging its entrepreneurially minded students to start a business—playing the role, one might say, of the canny initial investor, providing crucial start-up equipment, overhead, and technical assistance. Instead of shares in Yahoo, the university produced goodwill that translated into donations, almost immediately. Stanford’s reputation was soaring, right along with its endowment.
Brin and Page were a different case, however. Their idea for how to navigate the Web wasn’t meant to assist their PhD research—it was their PhD research. They hadn’t stumbled on a business idea, they were methodically wrestling with difficult questions involving data storage, data retrieval, and artificial intelligence. Their initial discomfort in starting a business from their research reflected how seriously they took these profound questions. There had been a countervailing belief system within the Stanford computer science department, represented by John McCarthy, among others, that considered commercial success inappropriate for a serious academic. A deep thinker, a true scientist, should stay at the lab and try to change the world from there. Only the short-term students—dropouts, even—were focused on making a fortune from the ideas they had picked up at school.
Google, like Yahoo, required substantial Internet bandwidth, especially as it became more popular, but the PageRank algorithm was PhD-level research being conducted at a world-class university. It could hardly have been the most ambitious or costly research project operating there; consider the massive particle accelerator that Terman had helped land for the university in the early 1960s. Couldn’t support for Google be just another case of the university backing its scientists wherever their research took them, money be damned? Wasn’t studying the burgeoning World Wide Web a subject worthy of academic research and university investment?
From that perspective, Stanford’s backing of Google shouldn’t be seen as acting as a venture capitalist but rather as a nonprofit institution committed to supporting cutting-edge research. In fact, what turned Brin and Page into entrepreneurs wasn’t Stanford’s early “investment” in their research project, but rather its decision to stop supporting the fast-growing Google project. The withdrawal of support in the summer of 1998 meant that Brin and Page would have to fend for themselves. “The resources just weren’t there in the academic environment,” Brin recalled. “We decided, ‘Hey, there’s a lot of commercial interest here. People will give us a lot of money to solve this problem of search. Why don’t we go and do it commercially?’”31 Hector Garcia-Molina, Brin’s advisor at Stanford, has a slightly different view of the events of this period, seeing the Google founders as being pulled toward the market rather than being pushed there by Stanford. “It wasn’t like our lights were dimming when they would run the crawler,” he said of Google’s supposed overreliance on the university’s equipment. “I think it would have made a great thesis. I think their families were behind them to get PhDs, too. But doing a company became too much of an attraction.”32
In August 1998, when Brin and Page were exploring how to keep Google going, they sought the advice of David Cheriton, a professor wise in the ways of start-ups, who worked a few doors down on the fourth floor of Margaret Jacks Hall. Cheriton had made a small fortune a couple of years earlier when Cisco paid $220 million for a company he created with Andy Bechtolsheim, a former Stanford computer science graduate student who had dropped out to help start Sun Microsystems. Cheriton encouraged Brin and Page to seek funding for Google, looping in Bechtolsheim. “If you have a baby, you need to raise it,” the professor told the students.33 The next morning, Brin and Page took a meeting with Bechtolsheim at Cheriton’s house.
Brin and Page demonstrated Google on a laptop, and, as Brin recalled, Bechtolsheim interrupted to say, “Oh, we could go on talking, but why don’t I just write you a check?” He ran to his Porsche, cut a $100,000 check made out to Google Inc., and handed it to Brin and Page. They explained to him that there was no Google Inc., and therefore no account in which to deposit the funds. No matter, Bechtolsheim replied, “Well, when you do, stick it in there.”34 No terms were discussed. It was unclear what exactly he was buying with his money, or if it was only a loan. But just like that, Bechtolsheim had turned two graduate students’ curiosity about starting a business into something tangible and real. “It was like, wow, maybe we really should start a company now,” Brin said. “The check sat in my desktop drawer for a month. I was afraid I’d lose it. But until it really happened, until then, it had sort of been this intermediate state. Things hadn’t really happened yet. But when he wrote the check—well, it certainly does speed things up.”35
Google incorporated on September 7, 1998, and Cheriton would become one of four early investors in the company, each said to have written checks for $250,000. In addition to Bechtolsheim, who supplemented his initial investment, and Cheriton, there was Ram Shriram, an angel investor who was introduced to Brin and Page through a different Stanford computer science professor, Jeffrey Ullman, who also encouraged the two graduate students to go into business. You can always come back and complete your degree, Ullman assured them. Shriram signed up the fourth investor, Jeff Bezos.36 The seduction of Brin and Page wasn’t simply about greed, or having the resources necessary to “raise their baby” and thus avoid becoming modern-day Teslas, whose best ideas never reached maturity. There also was the thrill for these star students of getting A’s in real life, not just in the classroom. Cheriton, Bechtolsheim, Shriram, and Bezos were talented, wealthy men who were clearly impressed by the business potential of PageRank. The multimillion-dollar investment round that Brin and Page pursued next would be another big fat gold star right on top of their business plan, one that everyone could see.
While making the rounds to pitch the Google search engine to eager Sand Hill Road firms nearby, Brin and Page were impressed both by John Doerr of Kleiner Perkins, which had invested in Netscape and Amazon, and Mike Moritz of Sequoia Capital, which had backed Yahoo. What to do? Why not work with both? Bechtolsheim told them there was a “zero percent possibility” that this would happen: hypercompetitive, hypercapitalist investors, he explained, don’t like to share their quarry.37 Those words were all the motivation Page and Brin needed. When the road show was over, the two firms agreed to split a $25 million investment at a valuation of $100 million. Making the sale wasn’t exactly easy, Page said, but he and Brin had the advantage of “being at Stanford and having a product that was really good. . . . It was much better than anything else, and so we would show them and say, ‘Hey we’ve got this great thing. It works better than anything out there. What do we do?’”38
As these events illustrate, Google’s founders displayed a particular mix of iconoclasm and respect for status. They clearly didn’t care for the rules for how VC funding was done, but that rebellion was in pursuit of being the rare start-up that could win over two competing firms. Page and Brin set an idealistic goal for the company—“To organize the world’s information, making it universally accessible and useful”—and because of the difficulty of that task, they insisted on hiring applicants from a narrow range of top-tier universities who met the rigorous academic standards common to Silicon Valley. Google intended to measure the “pure” intelligence of the candidates for its jobs and liked to deploy riddles on the spot as a way of detecting genius.
It’s hard to know exactly where the attraction lay for the Termans or David Shaw or Jeff Bezos or the Google founders in trying to measure intelligence so exactly. That is to say, did they want to assign numbers to intelligence because they liked the efficiency of using an algorithm to make judgments about hiring and promotion, or did they sincerely think intelligence was best rendered as a number, the way a computer’s processor could be assigned a speed? Whatever the motivation, they were all sticklers about getting as much relevant data as possible about a job applicant’s brain power and acting on that information. Thus Terman would routinely step in to reject a candidate for a professorship in a field like biology based on his undergraduate grades in calculus. Similarly, Page and Brin would insist on reviewing the standardized test scores of each Google applicant. A candidate with low test scores could be instantly disqualified. As recently as 2010, Page had the ability to examine the dossier of grades and test scores for every Google applicant. Occasionally, he did a spot check on a candidate’s credentials, “to ask what is the real quality of person we’re hiring.” Google wallowed in grades and scores, whether ancient or newly manufactured. For example, each answer during the hiring interview was rated on 4.0 scale—just like college—and scoring an average below 3.0 meant you didn’t pass.
Again, the values of the computer lab were imported into a company. “We just hired people like us,” Page explained. There were classes, cafeterias, esteemed visitors like McCarthy, who showed up early on to talk to employees. Sometimes programmers would take the final leap and sleep at work. “The fact of the matter is that for some people living here makes sense,” says Eric Schmidt, the older computer science PhD who was brought in to run Google in 2001. “Their friends are here, it’s what they’re familiar with, and the things they do here are very similar to what they did in college.”39
In the year before the dot-com crash, there was a broad acceptance of and patience for the strategy of putting together the best computer lab, giving it a goal, and then figuring out how to turn a profit. The $25 million investment in Google Inc. came in June 1999, and at the sparsely attended press conference in the Gates computer building on the Stanford campus, Brin was asked how Google intended to make money. He replied, “Our goal is to maximize the search experience, not to maximize the revenues from search.”40 The next month Google made a grudging acknowledgment of those newly arrived investors by assigning its first staffer to study how to bring advertising to the site. Generally speaking, however, the company’s brilliant founders felt free to focus on making the search engine better and adding more users—as if nothing had changed from their university days. When the subject of advertising came up, Brin and Page insisted that it wouldn’t be allowed to threaten the integrity of PageRank, which was sacrosanct. Licensing the company’s search technology seemed like the better business. The purer business. Advertising remained an ethically suspect backwater.
In late 2000, however, after the dot-com bubble burst, Google was under real pressure to perform—the company still hadn’t registered a profit despite 70 million daily visitors. Cheriton, among others, made his disappointment known, joking that all he would have to show for his investment was Google swag, including what he called “the world’s most expensive T-shirt.”41 Some investors started to suggest that they might withdraw their money if others could be found to step in—the specter of failure for the first time hung over Google’s founders. Doerr and Moritz used this stressful time to again insist that Google have an outside chief executive, something Page and Brin had agreed to but pushed off at the start. This time, however, they had to accept, and after many interviews arrived at Schmidt, a forty-five-year-old engineer who had already led a large tech company.42
With investors suddenly skeptical of how Google was operating as a business, advertising became a higher priority and Page and Brin assigned a number of talented executives to come up with an answer. After a series of small failures, Google in 2002 developed AdWords Select, an improved system for delivering ads next to search results that seemed to have solved a multidimensional puzzle. The ads ran well to the side so no one would confuse ads for search results, a taboo for Brin and Page. Under Page’s prodding, AdWords Select embraced an automated sales procedure that could scale easily, stepping in for the company’s human sales team. Anyone with a credit card could use the system and advertisers were only charged by Google for a click-through, that is, when someone left the search results page for their landing page, as opposed to being charged based on how many people were shown the ad. The system tapped into all Google knew about Web pages and its users to guide a customer to the strategy that would work best. The auction system for deciding which ads would appear next to certain searches relied on feedback loops that produced better outcomes for everyone involved. Winning bidders were refunded the difference between their bids and the runner-up and, by treating bidders fairly in this way, Google gave them the confidence to be bold with their bids, spending as much as they could afford. Likewise, Google ranked ads in part based on how effective they were, not simply rewarding the highest bid. This encouraged advertisers to make ads more useful, thus pleasing Google’s users.
In each case, the ethical system proved the most effective. Google’s first profitable year was 2001: $10 million. The next year, when AdWords Select arrived, profits were $185 million. Revenue was growing fast, too. The next year, Google introduced another advertising system, AdSense, which allowed small-scale Web sites to carry AdWords-style ads based on the kind of material they published, guided by the data Google had collected about who clicked on which ads. Google took roughly a third of what the advertising brought in, with the rest going to Web site publishers, many of whom otherwise never would have had the wherewithal to run ads.43 In 2004, when Google went public and had to show its books, investors could see exactly how profitable the company was, even in what clearly were lean years for dot-com companies. With the company on firmer footing, Brin and Page’s iconoclasm returned during the IPO, which was designed to work around investment banks by selling shares directly to individuals through an auction. (The original filing sought to raise $2,718,281,828, an allusion to the mathematic constant e, 2.718281828. . . .)44
The two founders were confident they had threaded the needle, finding a way to bring in lots of revenue with ads that were distinct from the search engine proper, thus avoiding the harm they had so carefully documented in other commercial search engines that accepted advertising. The ads displayed next to Google search results were so well targeted that users didn’t even know the ads were there unless they were looking for them, Brin said. “Do you know the most common feedback, honestly?” he asked. “It’s ‘What ads’? People either haven’t done searches that bring them up or haven’t noticed them. Or the third possibility is that they brought up the ads and they did notice them and they forgot about them, which I think is the most likely scenario.” Google ran an A/B test to learn more, displaying an ad-free experience for some users, while others saw ads. The visitors who were shown ads, the test revealed, used Google more than those in the ad-free group, leading to the happy conclusion that people liked Google’s ads.
However, Brin and Page’s critique of advertising relied on some clear guideposts for judging if search was being harmed by advertising and the search for profits and whether someone who stayed on the site longer wasn’t one of them. One issue to investigate, according to the two when they were in graduate school, was whether advertising stifled innovation. In terms of queries about people, places, and things (that have nothing to do with commerce), Google has undoubtedly been improving, approaching the kind of uncanny results that the founders originally envisioned. Google effectively anticipates what you are typing and usually can provide a précis on any topic of your choosing, relying on Wikipedia and other deep online resources as well as the information the search engine has collected itself.
In today’s version of Google, however, ads are no longer automatically segregated from results; they frequently appear directly above them. Enter an open-ended search term like “bicycles” and Google will direct you to some neutral sites that provide some perspective on the subject, but there may also be ads, and above those ads may be a row of shopping opportunities. The effect can feel highly commercialized. Here is the explanation Google offers for the sponsored shopping deals that frequently appear: “Based on your search query, we think you are trying to find a product. Clicking in this box will show you results from providers who can fulfill your request. Google may be compensated by some of these providers.” There is a bit of a chicken-egg aspect to that explanation. Is Google merely reflecting the Web by focusing on commerce and assuming that someone who makes a generic search must be “trying to find a product”? Or is it helping to shape this reality through its decisions about how to organize information? A search engine was supposed to be an honest, disinterested broker between a seeker and what is available on the Web, according to Brin and Page. Google obviously isn’t that anymore.
Another point that Brin and Page warned against as graduate-student foes of commercial search was the lack of transparency. Research at a university is typically published so that it can be reproduced and scrutinized and improved, while research done for a business is done in secret. The latter leads to what the two back in 1998 called “a black art,” which can be used for good or ill depending on who knows the magic formula.45 Today, Google Inc. jealously guards its secrets, PageRank first among them, and has offered a couple of justifications. To start, there are competitive reasons. Schmidt had battled with Microsoft in his previous job at Novell, and thus advised Page and Brin to keep as much secret as possible to avoid giving a competitor any angle to exploit. Netscape had experienced how Microsoft could quickly take over a market—like browsers—by applying its tremendous resources to reverse engineer a problem. Even revenue breakdowns, by Schmidt’s lights, could offer clues as to how Google works. There were the practical arguments, too, for practicing secrecy as it related to search. If the public knew even in broad strokes how PageRank made its calculations some enterprising souls would be able to game it and search results would suffer. Without giving opportunists that advantage, Google still battles against companies and individuals who practice what is called “search engine optimization.”
Page and Brin have done a true 180 on the topic—transparency was once the path to the best results, and now they view it as an obstacle. Moving from a research lab to a start-up obviously has changed their perspective. If Google still operated transparently and noncommercially in a university setting, then when some companies tried to trick the search engine the consequences wouldn’t be so dire. An open system would be self-correcting: anyone keeping up with the project could spot the dirty tactics and propose a solution. Wikipedia, for example, has created a system that relies on human volunteers as well as “bots” to control vandalism or self-motivated editing. It tolerates some slip-ups but generally has succeeded in sapping the will of troublemakers.
Such deception is much more threatening to a commercial entity like Google, however. It goes right to the heart of the company’s pitch to advertisers and Web site owners alike that it runs a closed system that is beyond significant manipulation. Beyond outside comprehension, really. The algorithm does its job secretly and fairly—that is Google’s promise. Its results are above reproach and the only way to deviate from this preordained “correct” outcome—to make an extraordinary impact for your business in search results—is to pay for a reasonably priced, demonstrably effective ad. This is good for search, which is in the trusted hands of the experts who work for Google. And good for Google.
In this way, Google tries to cultivate in the public the same cognitive dissonance it displays about whether it is a high-minded research lab or a profit-minded business. The entity that collects all that information, Google assures you, isn’t some company spying on your behavior to sell ads, but a research lab trying to improve your search experience and only show you ads that you’ll want to see. The same goes for the broad definition of “fair use” that Google favors, which would permit it to collect all manner of copyrighted material—movies, music, books—in its computer storage to be included in its search results. This is not because it wants to profit from others’ work, but because of its high-minded purpose of organizing the world’s information and making it accessible. Same goes for its opposition to the European “right to be forgotten,” which requires that someone mentioned in an old news article be able to demand that that article be excluded from search results.
There was a final experiment that Brin and Page proposed twenty years earlier to gauge the influence of advertising on search: type in the query “cellular phone,” they suggested, and see what results appear. Running that search on Google (today cell phone is the preferred term) produces what the two predicted would be the case for any search engine dependent on advertising. You do not find a negative Web page anywhere near the top of the results. First, there is a banner of links to buy cell phones, with the label “sponsored.” Then there are two ads, clearly labeled as such, and then a map showing you a few places to buy phones. Then the proper listings begin—Wikipedia’s article, Walmart’s article. Then cell phones in the news. Then, about a dozen or so items down is a Scientific American article from 2008, “Mind Control by Cell Phone,” that is mildly cautionary. Brin and Page’s paper didn’t claim that the absence of critical links about cell phones directly proved corruption in the search process. They offered a rational explanation: “A search engine which was taking money for showing cellular phone ads would have difficulty justifying the page that our system returned to its paying advertisers.” This subtle point, coming from would-be academics, recognized that you can produce bad outcomes without having that intention. You may just be doing what it takes to make your business run and “being evil” in the process, without even noticing.
All of which raises some intriguing alternative history: Would Brin and Page have remained true to noncommercial search had they met at a top-class school with a less go-go culture, like, say, MIT? Could they and we have avoided this reimagining of Google’s relationship with its users? Or would we instead be talking about a different pair of brilliant Stanford graduate students who researched how to improve Web search and inevitably found their way to Sand Hill Road, while our alternative-universe Brin and Page would be obscure but distinguished computer science professors?
Looking back two decades, Google’s rise from research project to Google Inc. may represent the last gasp of the Terman model at Stanford. Certainly innumerable start-ups have their roots in Stanford, including Instagram, but the university’s resources aren’t nearly as significant. Less than a decade from Google’s founding, an ambitious hacker could make his mark from his dorm room, using just the money in his pocket. Mark Zuckerberg was a twenty-one-year-old chief executive of Facebook when he returned to his old Harvard computer science class to tell the students how lucky he was to have started his company in the era he did. “The fact that we could sort of rent machines for, you know, like $100 a month and use that to scale up to a point where we had 300,000 users is, is pretty cool and it’s a pretty unique thing . . . that’s going on in technology right now.” A decade earlier, he said, eBay had to run off of two $25,000 machines from the start. Google was somewhere in the middle of these two poles and had some serious bills to pay.46
If Google was a last gasp for Stanford, what a last gasp it was. Because PageRank was developed at Stanford, the technology behind Google is owned by Stanford, which exclusively licensed it to Brin and Page for a portion of Google stock, the last of which Stanford sold in 2005 for a total payout of $336 million.47 The Google-Stanford connections are many and varied, as Terman imagined—students, faculty, alumni, and the university all helping each other make a fortune. Google has returned those good turns from Stanford in so many ways, from $1 million a year directly donated to the computer science department, to money donated to honor the memory of an inspiring computer science professor, to hiring Stanford graduates, and on and on. The recently departed president of Stanford, John Hennessy, is on Google’s board.
But, just as the building named after Terman was torn down to make way for an entrepreneur donor, this Stanford model for success ended up being replaced by other systems offering financial and social support, whether that meant informal networks of already successful entrepreneurs or specialized tech incubators like Y Combinator and Tech Stars. One Stanford graduate, Peter Thiel, proposed cutting through the pretense. Why pay tuition if you weren’t there to study? “College can be good for learning about what’s been done before, but it can also discourage you from doing something new,” his organization, the Thiel Fellowship, explains on its Web site. “The hardest thing about being a young entrepreneur is that you haven’t met everyone you’ll need to know to make your venture succeed. We can help connect you—to investors, partners, prospective customers—in Silicon Valley and beyond.”48
This was Thielism in its raw form—no half measures or false solicitousness. Stanford already was breaking down the purpose of academia by being so eager to turn research into profits, which it would share indirectly or directly. Why keep up the charade about university life? Why not go all the way and accept that market success was all that really mattered?