“The online is a group of information, however it’s a multitude,” says Exa cofounder and CEO Will Bryk. “There is a Joe Rogan video over here, an article over there. There is no organization. However the dream is for the net to feel like a database.”
Websets is geared toward power users who must search for things that other serps aren’t great at finding, akin to sorts of people or corporations. Ask it for “startups making futuristic hardware” and also you get a listing of specific corporations lots of long fairly than hit-or-miss links to web pages that mention those terms. Google can’t do this, says Bryk: “There’s a variety of invaluable use cases for investors or recruiters or really anyone who wants any sort of information set from the net.”
Things have moved fast since broke the news in 2021 that Google researchers were exploring using large language models in a brand new type of search engine. The concept soon attracted fierce critics. But tech corporations took little notice. Three years on, giants like Google and Microsoft jostle with a raft of buzzy newcomers like Perplexity and OpenAI, which launched ChatGPT Search in October, for a bit of this hot recent trend.
Exa isn’t (yet) attempting to out-do any of those corporations. As a substitute, it’s proposing something recent. Most other search firms wrap large language models around existing serps, using the models to investigate a user’s query after which summarize the outcomes. However the serps themselves haven’t modified much. Perplexity still directs its queries to Google Search or Bing, for instance. Consider today’s AI serps as a sandwich with fresh bread but stale filling.
Greater than keywords
Exa provides users with familiar lists of links but uses the tech behind large language models to reinvent how search itself is completed. Here’s the fundamental idea: Google works by crawling the net and constructing an unlimited index of keywords that then get matched to users’ queries. Exa crawls the net and encodes the contents of web pages right into a format generally known as embeddings, which might be processed by large language models.
Embeddings turn words into numbers in such a way that words with similar meanings grow to be numbers with similar values. In effect, this lets Exa capture the meaning of text on web pages, not only the keywords.
Large language models use embeddings to predict the subsequent words in a sentence. Exa’s search engine predicts the subsequent link. Type “startups making futuristic hardware” and the model will provide you with (real) links that may follow that phrase.
Exa’s approach comes at cost, nevertheless. Encoding pages fairly than indexing keywords is slow and expensive. Exa has encoded some billion web pages, says Bryk. That’s tiny next to Google, which has indexed around a trillion. But Bryk doesn’t see this as an issue: “You don’t need to embed the entire web to be useful,” he says. (Fun fact: “exa” means a 1 followed by 18 0s and “googol” means a 1 followed by 100 0s.)
