The Washington Post Is Secretly Planning to Start Publishing Articles Created Using AI

As the news media continues its often-disastrous pivot towards generative artificial intelligence, the most prominent publication yet has jumped on board: the Washington Post , it turns out, has quietly been building an AI tool designed to let underqualified writers publish content in its storied pages.

According to multiple sources familiar with the plan who spoke to the  New York Times , since April the newspaper has been secretly building a program known as "Ripple" that appears to be part syndication, part talent network akin to contributor schemes at HuffPost and  Forbes . 

The biggest difference from those prior contributor networks and this one, though, is the way the sausage will be made.

Along with partnerships with established writers, the new program will employ an "AI writing coach" called "Ember" designed to guide "nonprofessionals" through the article-writing process.

That tool will, per early prototypes described to the NYT , hand-hold aspiring columnists through every aspect of the writing process from start to finish. Its sidebar will instruct writers to devise an "early thesis," list out "supporting points," and provide a "memorable ending" — all while a live AI chatbot weighs in and a "story strength" tracker evaluates their progress.

At the end of it all, a human editor is said to review the columns before they go to publication — though in practice, similar promises have often ended in disaster, like at  CNET , which promised that editors were reviewing the scores of AI-generated articles it ran before it turned out they were riddled with errors and plagiarism .

Though the outputs from WaPo 's program wouldn't be entirely generated by AI, it sounds like everything but. And in practice, as we've seen time and again, the temptation with AI is to use it not as a thoughtful creative partner but as a speedrunning tool to churn out large quantities of low-quality material, or to cook up something that sounds confident but is shaky on the undergirding facts and logic.

In other words, it sounds like the program could be poised to take everything that's currently broken and controversial about newspaper opinion sections and amplify it using generative AI.

WaPo declined to provide comment for the NYT 's exposé. It's worth noting that the NYT  has been putting significant resources of its own into exploring how AI can be responsibly used in journalism — but while it's using certain machine learning tools for tasks like finding patterns in large datasets, it's pledged not to use generative AI to write any articles.

The revelation also comes during a period of broader crisis at  WaPo , with significant layoffs coming as its owner Jeff Bezos has increasingly exerted control over the content and ideological leaning of the paper's journalism.

Bob Woodward and Carl Bernstein, eat your heart out .

More on AI writing: Business Insider Did Something So Stupid With AI That We’re Reeling