On Dec. 7, 2022, seven days after OpenAI launched ChatGPT, a former OpenAI research scientist named Aravind Srinivas, who’d left the company just three months earlier, launched a competing A.I. chatbot called Perplexity. “Everybody was obsessed with ChatGPT. We were the only product that came and said, references and citations are important. So, from the beginning, we cared about it.” Srinivas, co-founder and CEO of Perplexity AI, said during an onstage interview at the TechCrunch Disrupt conference on Oct. 30.
Lately, Perplexity has been in a lot of hot water for exactly the problem Srinivas set out to address two years ago. Last month, the company was sued by the Wall Street Journal and the New York Post, both owned by News Corp, for plagiarizing their content in search results. A few days earlier, the New York Times sent a “cease and desist” notice to the startup demanding it stop using the newspaper’s content on its site.
Perplexity is at the forefront of the so-called A.I. answer engines, which aim at answering users’ specific questions by summarizing information on the internet, instead of just providing links in response to a few key words. Srinivas said the median query put into Perplexity’s answer engine is 10 to 11 words, versus Google Search’s two to three, suggesting that users come to Perplexity with more well thought-out questions.
Srinivas claimed Perplexity “always cites its sources” and “doesn’t claim ownership of any content.” “It’s just surfacing content from the web, summarizing it in a manner that the user can digest and then provide you where it’s getting all this information,” he said, adding that it’s exactly like how journalists do their job and therefore shouldn’t be considered plagiarism.
However, he admitted that, like other rapidly-evolving A.I. apps, Perplexity’s current safety guardrails are not perfect and could be easily bypassed using prompt engineering—a buzzing term describing the practice of designing inputs for A.I. tools that will produce optimal outputs.
The new publications that sued Perplexity claim the A.I. company is competing for the same audience as theirs using copyrighted content. But Srinivas said Perplexity users don’t come to the app to consume daily news, but to “make sense of what’s going on.”
“Like, how does that particular piece of news affect me? In the context of news, should I continue to buy more Nvidia stock? These are not the kind of questions you can come and ask TechCrunch, but you come and ask Perplexity,” the CEO said.
Acknowledging that reported news is essential in making Perplexity’s product valuable, earlier this year, the company launched a unique program to share advertising revenue with news publishers. It’s currently working with Time, Fortune and the German news site Der Spiegel.
But ultimately, Srinivas believes that no one should own the right to facts. “Our belief is that facts need to be universally distributed to everybody,” he said. “Imagine a world where scientists claim ownership over a certain fact, and other people cannot state it. Knowledge and truth cannot be disseminated in such a manner.”
Srinivas, originally from India and holding a Ph.D. in computer science from the University of California, Berkeley, was a research intern at DeepMind and Google. Before co-founding Perplexity, he was a research scientist at OpenAI for about a year.
