On Monday, News Corp. filed a lawsuit against the artificial intelligence firm Perplexity, arguing that the AI-generated search company infringed on its copyrights by ingesting its work and repackaging it to its users.
On Wednesday, Perplexity CEO Aravind Srinivas appeared at a News Corp. conference, where he sought to address “the elephant in the room” right at the top of his conversation.
Srinivas was interviewed at the WSJ Tech conference, which is hosted by News Corp.’s flagship publication The Wall Street Journal. “They’re not the only publishers like that are worried about the AI products that are surfacing their content on their answers,” Srinivas said of the Journal.
“I would say that they reached out around June to talk to us, and we responded to them and said we are also open to talking to them, and are interested in a proper commercial discussion on how to use the content in the Wall Street Journal and other similar outlets owned by them. So we certainly were very surprised about the lawsuit, because we actually wanted a conversation,” he added.
Srinivas compared his company to Spotify, the music streaming service that helped transform the business model of music, which had seen declining revenues and influence from labels, until the streaming boom brought it back from the brink and into an era of growth.
“The whole point I’m here is like to make it very clear that I would love to have a commercial contract. And regardless of Wall Street Journal alone, we have a select publisher program that we announced months ago where we clearly said we’re going to do advertising on Perplexity, and whenever we make advertising revenue, we’re going to share that revenue with the content publishers in a manner inspired by Spotify, where the creators are still getting paid as long as Spotify keeps growing,” he said, adding that “I’m hopeful that we can figure something out here.”
Perplexity fashions itself as an upstart competitor to Google, helping users ask questions to get answers. But while Google (at least originally) served users by giving them links to other websites, Perplexity consumes a vast array of content and condenses it to deliver an answer. It includes citations, but many of the concerns from publishers is that it obviates the need for the user to visit the source material.
In the case of News Corp., the media giant argued that not only in some cases did Perplexity use some paragraphs word for word, but it also hallucinated some things, but cited its outlets.
“We try our best to post-train these models to make sure it’s not like regurgitating content, but it’s just trying to take multiple different sources and give a diverse perspective,” Srinivas said. “AIs are not perfect, we’re still working with a piece of technology that is constantly improving.”
“Whatever issues exist today are new set of issues that did not exist one year ago or two years ago. Similarly, hallucinations is another problem where media outlets are not happy if they’re cited in an answer that might be inaccurate, which is completely reasonable to want,” he added. “We are trying our best to be the most accurate chatbot on the market by trying to constantly ground our answers in sources and every single day, all the people in the company are constantly tracking all the inaccurate answers and constantly thinking of ways to address them. So we are doing our best to make sure that like the product keeps improving.”
For now, Perplexity is still raising cash, seeking to value itself at around $9 billion in its latest round. Srinivas told the conference that he hopes the company will be profitable in three to five years, though the uncertain legal landscape around generative AI is sure to play a role in that outlook.