A new program from Perplexity is about a year old. It’s undergone some expansion, and a fair amount of media coverage, and these days, it seems to be going well.
By creating its Publishers Program, Perplexity joined the ranks of AI companies paying news outlets and other publishing parties for the use of their content in feeding LLMs. This type of program often comes about in the context of questions about fair pay for content creators: reporters, publishers, and anyone in that food chain.
Paying Up
At the time, as they were rolling out the initiative, Perplexity spokespersons characterized the program this way:
“Every day, people turn to Perplexity with a wide array of questions. Our ability to provide high-quality answers hinges on trusted, accurate sources covering the topics people care about most. From day one, we’ve included citations in each answer, ensuring publishers receive proper credit and building user trust. To further support the vital work of media organizations and online creators, we need to ensure publishers can thrive as Perplexity grows. That’s why we’re excited to announce the Perplexity Publishers’ Program and our first batch of partners: TIME, Der Spiegel, Fortune, Entrepreneur, The Texas Tribune and WordPress.com. This program is designed to promote collective success and to equip publishers with new technology to engage their audiences.”
So there are citations, the giving of credit, and the actual doling out of money to corresponding value.
Expanding the Program
Around Christmas time last year, we got word that Perplexity was adding new partners to the program.
“We would not be able to serve factual, valuable answers without news organizations continuing to report on different topics,” Jessica Chan, head of publisher partnerships at Perplexity, said in a statement. “We’re excited to welcome these new publishers to the program.”
Now, the program has a large number of participants. At Imagination in Action, myself and my colleague Daniela Rus, director of the MIT CSAIL lab, sat down and asked Perplexity co-founder Johnny Ho about the publisher’s initiative, among other things.
“It’s grown exponentially,” he said of the program. “So we started with maybe 10 publishers, but now I think it’s at least 100 – I think the main thing there is that the user gets directed to link very directly (to) Perplexity at the top, and now, because of AI, we’re able to find the most accurate relevant information for the user, and attribute that revenue according to that query, (and) to the publishers that were cited.”
Ho also talked more broadly about being responsive to users, and how that fits into a company’s mission statement.
“Being grounded is what’s important, right?” he said. “Having the user feedback (that) goes straight to you, straight to your inbox every day, is pretty good. I’d say that’s the number one thing. And over time … it’s gotten pretty unmanageable, right? So you actually need another AI to summarize the user feedback for you. I think that’s pretty important. The other part that’s important is that all the founders are very technical … every so often, we just go back in and check the code, and check if the code that’s being written actually aligns with what the user is asking for.”
Later in the interview, he went back to that topic:
“I think I would just re-emphasize having that type of feedback loop,” Ho said. “It’s easy to get very high in the sky about, let’s say, valuations, or start bragging about your own technology, but I don’t think users actually care about any of that stuff. I think it’s very important to stay very grounded, and to keep listening. I think it’s mostly even an issue of asking the right questions.”
He gave an example.
“Let’s say you have all this data about what people are using what products,” he said, “but you’re not asking the right questions about what they actually want next, and just only purely copying other road maps or listening to the hype: you know, you’re not actually going to get anywhere.”
Other Initiatives
Of course, Perplexity isn’t the only company to do this: tech media estimates are that OpenAI, for example, pays publishers between $1 million and $5 million for content use.
The issue is fairly gray, too: how do you prove that the AI engine used your work?
Thus prompted, ChatGPT suggests:
· Watermarking and fingerprinting
· Prompt probing
· Data set matching
· Legal discovery
· Model attribution research
Some of those, for example, legal discovery, will only be available to larger publishers.
In any case, it’s good to see this idea catching on, especially as so many of us worry about data ownership in the AI age. As Will.i.am said while on stage at IIA last year, we have to own our personal data, for a free and fair twenty-first century.