While we prepare for the cold winters and shorter days, here’s some need-to-know news about what’s happening at the intersection of AI and journalism.

AI’s role in newsrooms

Poynter reports that ESPN is going ahead with having AI help cover “under-served sports” for game recaps. These AI-generated summaries will be reviewed by a human editor to ensure accuracy and quality. In a release, ESPN said the move will help them “focus on their more differentiating feature, analysis, investigative, and breaking news coverage.” Not everyone sees it as a win.  In a column for Defector, Luis Paez-Pumar wrote that the sports giant, “is beefing up its coverage of ‘underserved’ sports… not by hiring people who can and already do write these kinds of stories, but rather by feeding existing soccer and lacrosse journalists’ work into a machine aimed at making them obsolete. 

In The Fix, Olle Zachrison, Head of AI and News Strategy at Swedish Radio, discusses using the text-to speech transcription model they developed to attract 1.4 million new listeners with hearing difficulties. He also shares insights on how they’re using AI to improve content recommendations.

Senior Editor for AI initiatives at The Economist Ludwig Siegele told Press Gazette that AI translation will never be a solved problem, although it has reached the point where it is good enough for the publication to have introduced AI-powered in-app translations in its budget daily briefing, Espresso.

How we got here

Charlotte Li, writing in Generative AI in the Newsroom, outlines concerns about adopting generative AI in newsrooms including a lack of ethics and safety standards, challenges in allocating limited resources and integrating AI, biases among readers, potential job losses, environmental impacts, and power imbalances with tech companies. On the brighter side, she notes that “these concerns also serve as a great source of inspiration for meaningful research questions in computational journalism.”

A longer piece in the journal looks at how the news industry scrambled to understand ways in which new technology would impact news gathering, production, distribution, business models, and intellectual property. Following the disruption caused by social media, newsrooms continue to grapple with how audiences will consume information in the future.

AI Assistance in Data Visualization

Nightingale road-tested AI apps for how they can help creating data visualizations. The results were a bit mixed, but they found

some of these are even capable of building dashboards.

We love Alberto Cairo and Simon Rogers’ well-named “Data Journalism Podcast.”  In a recent episode, they interviewed the AP’s Garance Burke about AI and data journalism, including the role of their style book, which is the Bible of usage for us here at Storybench and most newsrooms.

Generative AI and the law

Also in Generative AI in the Newsroom, Julia Barnett has the latest on over 30 copyright lawsuits against companies like Open AI and Microsoft, raising critical legal and jurisdictional issues. These cases have been filed for using copyrighted content to train AI models.

Open AI’s developments

“As generative AI matures, it’s critical that journalism created by professional journalists be at the heart of all AI products,” said Hearst Newspapers President Jeff Johnson. OpenAI will incorporate content from Hearst’s large portfolio of newspapers and magazines. 

Unlike Google and OpenAI, AI search engine Perplexity has revenue-sharing agreements with publishers like Time and Der Spiegel. Microsoft recently announced it will pay publishers whose content is used by its Copilot assistant. Varun Shetty O- Open AI-head of media partnerships said, the company intends to fully fold SearchGPT into flagship product of ChatGPT by end of the year.

OpenAI announced it will not share ad revenue from its SearchGPT product with the publisher, instead driving value through “incremental traffic from new audiences,” said Varun Shetty, head of media partnerships, at the Twipe Digital Global Summit. SearchGPT, set to be fully integrated into ChatGPT by this year’s end, surfaces content in natural language responses, aiming to balance user needs and publisher attribution. Unlike OpenAI, competitors like Perplexity and Microsoft have implemented revenue-sharing agreements with publishers whose content is used by their AI systems.

Spotting AI-generated fakes 

Identifying AI-generated fake content is a difficult task that governments and companies are increasingly tackling Wired covered China’s proposed new regulation that requires AI content to be clearly watermarked. YouTube is developing a pair of tools to protect creators’ voices from being simulated, according to the Hollywood Reporter. Want to learn how to detect fakes like an expert? Here’s an interactive story from The Washington Post on how to identify AI deepfakes of Harris and Trump’s voices.