news

No wonder both Huang Renxun and Son Zhengyi are optimistic! Perplexity search volume surges: half of last year in one month

2024-08-11

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Cailianshe News, August 9 (Editor: Xiaoxiang)Since its launch, artificial intelligence search startup Perplexity AI has been regarded as a "strong rival to Google search" in the AI ​​era. And the growth rate of its business seems to be living up to expectations - after completing a new round of $250 million in financing, its monthly revenue and usage have increased by about seven times since the beginning of this year.

In a recent interview with the media, Dmitry Shevelenko, Chief Business Officer of Perplexity, said that the artificial intelligence search engine answered about 250 million questions last month, and the number of search responses in 2023 will be only about 500 million.

This new data highlights Perplexity’s growing market position in AI search today—It’s one of the fastest-growing generative AI applications since OpenAI’s ChatGPT launched in November 2022, even though the startup’s data collection techniques have long been controversial.

According to public information, San Francisco-based Perplexity was founded by Aravind Srinivas in August 2022, just three months before ChatGPT was launched. As its founder, Srinivas had internship experience at Google's DeepMind department and later worked as an artificial intelligence researcher at OpenAI. These experiences ultimately created favorable conditions for Perplexity to catch up with this round of AIGC wave.

Cailian actually introduced Perplexity and its business last year when the company was still relatively unknown. According to the official website, Perplexity uses advanced artificial intelligence models to provide direct answers to search queries, with complete citations and their sources, rather than providing a list of website links for users to search on their own like traditional search results such as Google and Baidu.

Over the past year or so, Perplexity's business has clearly grown significantly. It is estimated that Perplexity's annual revenue at the beginning of this year was only $5 million (the full-year revenue was calculated based on the sales in the past month), but according to company insiders, based on the same standard, its annual revenue is now expected to exceed $35 million.

That number, while not high, is what the company has achieved so far with only Pro subscriptions. Now, the startup is shifting its business model from subscriptions to advertising, putting it in more intense competition with Google, which dominates the $300 billion search advertising industry.

In addition to Google, OpenAI has recently launched an artificial intelligence search tool model SearchGPT for use by approximately 10,000 testers.

So, as more and more companies step up to integrate artificial intelligence capabilities into their search products, what advantages and characteristics does Perplexity have in this field at the moment?

Perplexity’s “Moat”

Shevelenko said, “At the end of the day, small companies in this space have two advantages: speed and focus. When it comes to Perplexity, our users and team only think about one thing: this is where you can get answers to your questions. Competition allows us to be more focused.”

To compete with its larger rivals, Perplexity recently raised $250 million in new funding from investors including Masayoshi Son’s SoftBank Vision Fund 2, people familiar with the matter said. Perplexity’s valuation also tripled to $3 billion from $1 billion in April.

The company’s existing investors include AI chip maker Nvidia and Amazon founder Jeff Bezos, as well as some of the biggest names in the AI ​​industry — such as OpenAI co-founder Andrej Karpathy and Meta chief AI scientist Yann LeCun.

Shevelenko said:Perplexity isn’t afraid of competition from tech companies with deeper resources, even if that includes Microsoft-backed OpenAI.

“OpenAI is doing a lot of different things… They’re not focusing on answering people’s questions with high-quality information sources,” Shevelenko said. “That’s why the side feedback from SearchGPT shows that it’s not as good as Perplexity.”

Perplexity recently announced that it will introduce advertising on its platform by the end of next month. Shevelenko noted that “unlike OpenAI, we always knew that our main monetization engine would be advertising.”

Shevelenko also said it would give a "double-digit" share of revenue from each sponsored article to the publisher of the news cited. The company has signed partnerships with Time, Der Spiegel and Fortune, among others.

He added that within two weeks of Perplexity launching revenue sharing, 50 publishers had requested to join the program. The company hopes to have as many sites as possible join. Shevelenko pointed out that "for Perplexity to be a useful product on the open web, there needs to be a good business model to publish up-to-date, real information about the world. If you want to align incentives (with journalism) in the long term, then revenue sharing is a more effective way than a one-time payment, and OpenAI is clearly taking the latter path."

Unlike Google and OpenAI, Perplexity did not build its own AI models, as this has become increasingly expensive. Instead, it licensed a combination of AI systems from OpenAI and other companies. Like many of Google's competitors, Perplexity's search engine was initially powered by a licensed version of Microsoft's Bing web index. But Shevelenko said it no longer uses Bing as a core system.

"We have our own proprietary search index and ranking system," Shevelenko said. "We use signals from various engines, but we also have our own crawler and ranking system."

Currently, Perplexity is targeting the press and academic communities because of its large amount of reliable information and data.Perplexity sees these source materials as an advantage over traditional search engines such as Google, which cite a much more diverse range of websites, said a person who worked at Perplexity earlier this year.

“This junk information is fed in and eventually spit out, which is actually a problem that plagues AI companies, so more data sources need to be referred to when training models,” the person said.

(Cailianshe Xiaoxiang)
Report/Feedback