Google 成立

1998 年 9 月，佩奇和布林注册了一家叫 Google 的公司。公司刚成立时，他们的产品只是一个搜索引擎，但藏在这个产品里的核心思想，后来改变的不只是搜索——它定义了「用数据与算法理解信息」这件事本身。

1990 年代中期的互联网，信息开始爆炸。门户网站靠人工编辑把网页分类，用户顺着目录一层层找；搜索则停留在简单的关键词匹配，结果常常又乱又难用。佩奇和布林在斯坦福想的是另一个问题：网页之间互相链接，这些链接关系能不能告诉我们「哪个网页更重要」？

于是有了 PageRank。它的思路朴素而有力：一个网页被越多重要的网页链接，它就越重要。这就像学术引用——被重要论文引用的论文更可信。把整个互联网当作一张巨大的引用网络，排序问题就从「匹配关键词」变成了「理解结构」。两人甚至用乐高拼了一台服务器集群来跑这个想法。

Google 的搜索结果一出来，干净、精准、几乎没有广告干扰，和当时的门户搜索形成鲜明对比。用户开始用它替代门户首页，搜索也从「浏览入口」变成了「直达工具」。而支撑这一切的，是对数据与算法的重度投入——这为后来的一切埋下了伏笔。

二十多年后回头看，1998 年的 Google 已经是 AI 版图的核心参与者：深度学习、TPU、BERT、Transformer、Gemini。这些都不是偶然，它们共享同一条逻辑——用规模化的数据与算力去理解世界。而这条逻辑的起点，是那个用乐高拼服务器、想把网页排好序的车库公司。

In September 1998 Page and Brin incorporated a company called Google. At founding, their product was only a search engine, but the core idea inside it would change far more than search—it defined "understanding information with data and algorithms" as a discipline.

In the mid-1990s the internet was exploding. Portal sites had human editors classify pages, and users navigated directory by directory; search meant simple keyword matching, with messy, unusable results. At Stanford, Page and Brin asked a different question: pages link to each other—can that link structure tell us which page matters more?

Hence PageRank. The idea is simple and forceful: the more important pages that link to a page, the more important it is. Like academic citations—papers cited by important papers are more credible. Treat the whole web as one giant citation network, and ranking stops being "match keywords" and becomes "understand structure". The two even built a server rack from LEGO to run the idea.

Google's results were clean, precise, almost ad-free, a stark contrast to portal search. Users began replacing portal homepages, and search shifted from "browsing entry" to "direct tool". Underneath was a heavy commitment to data and algorithms—which set up everything that followed.

Two decades later, Google in 1998 already foreshadows a core player in the AI landscape: deep learning, TPUs, BERT, Transformer, Gemini. None of these are accidents; they share one logic—understanding the world through data and compute at scale. And the origin of that logic is a garage company, building servers from LEGO, wanting to rank the web better.