The Washington Post uses AI to create personalized podcasts: NPR

It’s not your mom’s podcast, or your dad’s, or anyone else’s. The Washington PostThe new ‘Your Personal Podcast’ offering uses artificial intelligence to tailor podcasts to its users, combining the algorithm you find in a news feed with the convenience of portable audio.

The podcast will be “automatically personalized based on your reading history” from After articles, the newspaper says on the help page. Listeners also have some control: With the click of a button, they can change the topic mix of their podcast – or even swap the computer-generated ‘hosts’.

The AI ​​podcast immediately made headlines – and drew criticism from people who doubted its veracity and the motives behind it.

Nicholas Quah, critic and staff writer for Vulture and New York magazine WHO writes a newsletter about podcasts, says the AI ​​podcast is an example of this After‘s extensive digital experiments, but one that didn’t go entirely well.

“This is one of many technological, digital-oriented experiments they’re doing” aimed at “getting more audiences and entering new demographics,” he says. Those broader efforts range from a generative AI tool for readers to a digital publishing platform. But in this case, Quah adds, “it feels like it compromises the core idea of ​​what the news product is.”

On that help page, the newspaper emphasizes that the podcast is in early beta and “is not a traditional editorial podcast.”

Bailey Kattleman, head of product and design at the Aftercalls it “an AI-powered audio briefing experience” – and one that will soon have listeners talking to it.

“In an upcoming release, they can actually interact with each other and ask follow-up questions to dig deeper into what they just heard,” Kattleman said in an interview with NPR.

As technically advanced as that sounds, there are a lot of questions about the accuracy of the new podcast – even its ability to name the names of After journalists it quotes. Semafor reported that there were errors cited by employees of the After, included “misattributing or fabricating quotes and inserting commentary, such as interpreting a source’s quotes” as the newspaper’s own position.

In the newspaper’s app, a note advises listeners to “verify information” by comparing the podcast to the source material.

In a statement, the Washington Post Guild — which represents editorial staff and other staff — told NPR: “We are concerned about this new product and its introduction,” claiming that it After‘s mission and the work of its journalists.

Citing the newspaper’s long-standing practice of issuing a correction if a story contains an error, the guild added: “Why should we support technology that is held to a different, lower standard?”

So why is the After roll out an AI podcast? And will other news and audio outlets follow suit?

Here are some questions and answers:

Doesn’t AI podcasting already exist?

“The After has certainly come to the fore here among US publishers,” Andrew Deck tells NPR. But he adds that the paper isn’t the first to experiment with AI-generated podcasts in the broader news industry.

Deck, who writes about journalism and AI for Harvard University’s Nieman Lab, points to examples such as the BBC’s My Club Daily, an AI-generated football podcast that allows users to hear content related to their favorite club. In 2023, he adds, “a Swiss public broadcaster used voice clones of real radio presenters on air.”

News outlets have also long offered an automated feature that converts text articles into computer-generated voices.

Even outside the news industry, AI tools for creating podcasts and other audio are more accessible than ever. Some promise to streamline the editing process, while others can synthesize documents or websites into what sounds like a podcast conversation.

Why do publishers want to experiment with AI podcasts?

“It’s cost-effective,” said Gabriel Soto, senior research director at Edison Research, which tracks the podcast industry. “You’ve left out many of the resources and people needed to produce a podcast (studios, writers, editors, and the host himself).”

And if a brand can create a successful virtual AI podcast in today’s highly competitive podcasting market, Soto adds, it could become a valuable intellectual property in the future.

Deck says that if the AfterIf the experiment works, the newspaper could “significantly scale and expand its audio journalism offering without investing in the labor that would normally be required to expand.”

In an interview, Kattleman emphasized that the new product is not intended to replace traditional podcasts: “We think they have a unique and lasting role, and that won’t go away at the Post.”

What is unique about the After Do you have a podcast?

For Deck, the level of customization it promises is an innovation. Being able to tailor a podcast specific to one person, he says, “is arguably beyond what a podcast team in journalism can currently produce manually.”

In an example the After published, listeners can choose from voice options with names like “Charlie and Lucy” and “Bert and Ernie.”

Kattleman says her team worked with the idea that there is no “one size fits all” for audiences when it comes to AI and journalism.

“Some people want that really direct briefing style; some people prefer something that’s more conversational and conversational,” she says.

Quah says adding an AI podcast is an effort to make stories accessible to a broader audience.

He says that with the podcast the After seems to want to reach young people who ‘don’t want to read anymore, but only want to listen to the news’.

A key goal, Kattleman says, is to make podcasts more flexible, to appeal to younger listeners on the go.

The process behind the After‘s AI podcast, Kattleman says, “Everything is based on Washingtonpost journalism.”

An LLM, or large language model, converts a story into a short audio script, she says. A second LLM then checks the script for accuracy. After the final script is stitched together, Kattleman adds, the voice narrates the episode.

Will listeners embrace an AI news podcast?

Edison Research’s Soto says 1 in 5 podcast consumers say they have listened to an AI-narrated podcast.

But he adds that podcast listeners “many prefer the human connection, accepting AI tools to help create the content but not run or host the podcast.”

The new AI podcast reminds Deck a bit of the hyper-personal choices for users that TikTok and other social media offer.

“There’s a level of familiarity
and, arguably, comfort with algorithmic curation among younger audiences,” he says.

But while the younger audience tends to be tech-savvy, many of them are also attentive to authenticity and connection.

“Community is at the heart of why people listen to podcasts,” Soto says.

Then there’s the idea of ​​a host or creator’s personality driving engagement on TikTok and other platforms.

“These creators have built a relationship with their audience – and perhaps even trust – even if they haven’t spoken to sources themselves,” says Deck. “This kind of news content is a far cry from the disembodied chatter of AI podcast hosts.”

What are the possible disadvantages of AI podcasts?

A major potential consequence is the loss of jobs – and for companies, the loss of talent.

“The automation of it erases the whole kind of voice performance industry,” says Quah. “There are people who do this for a living,” he adds, and who “could produce higher quality versions of these recordings.”

There are also concerns that if AI chooses a story and determines how it is presented, it could create an echo chamber, cutting out the context or skepticism that a journalist would likely provide.

“AI-based news personalization tends to land firmly in the camp of offering audiences what they want to know,” Deck says.

Deck says he’s ready After‘s AI podcast some time to see how it turns out. But Deck does have an important concern: “I can say outright that generative AI models are hallucinating.”

And when AI models are wrong, he says, they often are.

Blurring boundaries between human and AI voices can also raise questions of trust – a crucial factor for a news organization.

As Soto puts it, “What happens when your audience expects content from the real you and finds AI instead?”