In a dimly lit garage, cluttered with wires, circuit boards, and the hum of old computers, a small group of visionaries gathered around a makeshift server. Their eyes sparkled with the excitement of a simple idea—one that would soon change the world. This humble garage, easily overlooked by anyone passing by, became the birthplace of Google, a company that would revolutionize how people navigate the vast and often overwhelming world of information on the internet.
What started as a modest startup quickly grew into a force that would challenge and eventually surpass even the tech giant Microsoft. Microsoft, with its deep resources and strong position in the industry, seemed unbeatable. Yet Google, fueled by relentless innovation and a vision that went far beyond what anyone had imagined, and soon came to dominate the search market. Today, Google controls an incredible 90 percent of the global search engine market, a testament to the power of being fast and innovative in an industry where speed is everything.
But now, even Google, which has become a giant in the industry, faces the same challenge from a new wave of disruptors. OpenAI, with its advanced work in large language models (LLMs), has taken the lead in an area that could potentially change the future of search and how we find information. What was once Google’s territory is now being disrupted by new approaches and fresh ideas from companies like OpenAI and Perplexity, who are looking at the future of search in ways that Google might not have fully expected.
The story of Google is a powerful reminder that in the tech industry, staying on top is harder than ever. The same drive and creativity that once pushed Google to the top now inspire new players eager to change the way we interact with the digital world.
How does search work?
Search engines perform a few key functions that are essential to finding the right information for users. It all starts with programs called spiders or web crawlers, which explore the internet to discover new web pages. These crawlers move from one link to another, following paths that connect different pages. When they find a page, they download it and carefully examine its code and content. If they find something new, they add it to the search engine’s index—a huge database that stores important details about each page, such as keywords, content type, and how recent the information is. This index is what helps determine the quality and relevance of each page.
When a user performs a search, the search engine uses this index to find the most relevant pages that match the user’s query. These results are then ranked based on how likely they are to answer the user’s question, considering factors like the user’s language, location, device, and search history. This process ensures that the best and most accurate information appears at the top of the search results.
Interestingly, the way large language models (LLMs) are trained on similar data that search engines gather to provide users with relevant information. A big part of the training data for LLMs is collected in the same way—by gathering information from across the internet. One of the major datasets used is the Common Crawl, which includes billions of web pages collected over more than a decade. This massive collection of data represents a large part of what we know and share on the internet, capturing the essence of human knowledge and experiences online.
A novel search experience
With this vast collection of knowledge, large language models (LLMs) gain a foundation that allows them to mimic human language and hold conversations that feel surprisingly real. When you interact with an LLM, you’ll notice that it’s especially good at summarizing information and giving you the main points of what you’re looking for. In contrast, when you use a search engine, you often have to click through multiple pages and sift through a lot of information to find what’s relevant, which can take much more time.
However, one challenge with the knowledge that LLMs have is that it’s often static and based solely on the data they were trained on. This means that these models might give users outdated or even incorrect information that they believe to be true.
To address this issue, both Perplexity and OpenAI have equipped their models with the ability to search the internet in real time, allowing them to provide answers that are more current and based on the latest information available online. Perplexity also goes a step further by providing users with sources where they can explore the information in more detail. It uses its own algorithms to rank these pages and ensure that the citations are from high-quality and relevant sources.
This approach has the potential to disrupt Google’s core business model, which mainly relies on users typing their questions into a search bar and then being shown paid search results first. It’s been rumored that Google has been slow to adopt large language models in its search engine because it could undermine its current business.
It’s still too early to say whether Perplexity will fully disrupt the search market or whether its new approach to search will catch on with a wide audience. However, its annual search queries are on the rise, and its valuation has reached $3 billion after a recent funding round. While Perplexity was initially seen as a competitor to OpenAI or Anthropic, it’s now being viewed as a challenger to Google Search itself.
Winner takes all
As the dust settles in the ongoing battle for search dominance, one thing is clear: the world of search is on the brink of another revolution fueled by LLM technology. Just as Google once emerged from a cluttered garage, fueled by a simple idea that would change the way we navigate the internet, new players like Perplexity are stepping into the spotlight with their own innovative visions. They, too, began as modest startups, yet they are now challenging the very giants that once seemed unbeatable.
The question remains—will Perplexity, with its fresh approach to search and commitment to real-time accuracy, become the next garage startup that transforms the way we interact with information?
One thing is certain: in the ever-evolving world of technology, even the mightiest can be dethroned by a simple idea born in an unexpected place.