Perplexity was planning revenue-sharing deals with publishers when it came under media fire

Perplexity, an AI-native search engine that launched in August 2022, was the first generative AI product to cite sources within the text.

ChatGPT, for instance, initially relied solely on outdated information that was used in the training of the AI model. (Srinivas worked for OpenAI as an AI research scientist.)

Perplexity was originally available via the Discord chat app and later the web. Rather than answer questions based on training data, Perplexity goes out and searches the internet for results, citing the original sources within the answer.

Last month, it launched a new “Pages” feature. Users can take a search result and turn it into a page, similar to a Wikipedia entry.

Then, at some point, a Perplexity user created a page from a search about an article by Forbes about Schmidt’s secretive new drone business.

John Paczkowski, a Forbes executive editor and veteran tech journalist, was livid. “Our reporting on Eric Schmidt’s stealth drone project was posted this AM by @perplexity_ai,” he posted on X. “It rips off most of our reporting. It cites us, and a few that reblogged us, as sources in the most easily ignored way possible.”

“Perplexity’s knock off of our reporting feels like a crystallization of the ‘Can journalism survive AI?’ convo. The company took our paywalled work, without our permission, and competitively broadcast it across web, video, mobile — as though it were itself a media outlet,” he wrote in another X post.

Perplexity’s new Pages feature doesn’t replace or compete with original journalism. It competes with the news aggregation business.

And humans have plagiarized so many of my articles, I’ve lost count. Sometimes the story isn’t even rewritten, it’s just posted under another byline on a new website. Rewriting content and using it to generate web traffic is part of the fabric of the internet.

When I was an early employee at The Information, our scoops would get aggregated by a myriad of websites and blogs. Often, those same blogs would warn their readers that The Information had a paywall, just so readers wouldn’t waste their time clicking through. How dare anyone charge money for news?

I have also lost count of the times reputable publications simply refused to credit my and others’ work. And any local news reporter can tell you what it feels like to be a victim of parachute journalism. When a national outlet swoops in, rips off your reporting and pretends to “own” the story.

For decades, media outlets have tacitly gone along with this kind of “plagiarism,” mainly because, as long as the aggregated article included a link, it would send traffic back to the original publication.

In fact, Perplexity’s use of the Forbes article did send some traffic back, albeit a small amount, according to Forbes Editor-In-Chief Randall Lane.

Perplexity did owe a credit to Forbes high up in the article, and perhaps a better rewrite of the original text. That’s something the company fixed quickly and can refine with more time.

The media business, on the other hand, isn’t so easily fixed. And as storied institutions lay off employees and face economic turmoil, there are a lot of raw nerves in our profession.

Lane described Srinivas about as uncharitably as possible: “But in the hands of the likes of Srinivas — who has the reputation as being great at the PhD tech stuff and less-than-great at the basic human stuff — amorality poses existential risk.”

That sentence does not match the crime. Srinivas didn’t even train the algorithms. He used them to summarize what’s out there on the internet and linked back to the original source. Yes, he could have done it better.

But turning him into public enemy No. 1 of the media is not really warranted.

I also wouldn’t discount the revenue-sharing idea Perplexity is planning to launch. It offers a promising alternative to a failing click-based news economy that incentivizes cheaply-made content with great headlines.

But if news organizations in the AI age can earn revenue from the actual substance of articles, the incentives are much more aligned with the mission of journalism itself. And that would be progress. It would also be an ongoing revenue stream, compared to the one-off training data deals that media companies are now making with the big foundation model companies.

In this particular case, I think we’ve made a mountain out of a molehill and an enemy out of a potential ally.

Here’s more of Lane’s argument about why Perplexity’s move was so wrong:

Perplexity then sent this knockoff story to its subscribers via a mobile push notification. It created an AI-generated podcast using the same (Forbes) reporting — without any credit to Forbes, and that became a YouTube video that outranks all Forbes content on this topic within Google search … As we dug, we found a similar rip-off of a second story at Forbes. And other stolen scoops — all the information, negligible citation — from Bloomberg and CNBC.

So what can be done? A few people can make a difference. For instance, the $174 million that Perplexity has raised so far includes a check from Jeff Bezos, who himself has made a huge investment in the need for journalism through The Washington Post. How does Bezos square that with companies like Perplexity which steal reporting that subscribers pay for at other publications (including Forbes and the Schmidt story) and then offer it for free in their ecosystem?