3 ways newsrooms can enhance AI integration

As newsrooms worldwide explore the implications of using Artificial Intelligence, the conversation around its adoption continues to expand.

Pranav Dixit, Senior Editor at Engadget and a 2022 Nieman Fellow at Harvard, shared his insights on AI in newsrooms during our AI Summit in Bengaluru, drawing from his extensive experience with both the US and Indian media markets.

AI landscape in newsrooms

Outlining the trends that have shaped the use of AI in newsrooms in the past year, Dixit noted three key developments that stand out:

AI-generated articles without disclosure: Many media companies have experimented with AI to generate complete articles. This approach has often led to backlash, especially when the content was published without disclosing its AI origins. High-profile examples include CNET and Sports Illustrated, where AI-generated articles were initially published under misleading bylines, resulting in public complaints and corrective measures.

Publisher deals with AI companies: Many news organisations have struck lucrative deals with tech companies to let them train their AI models on journalism content. This has also raised concerns about data privacy and the future of content monetisation.

Data scraping by AI companies: There is growing anxiety over how AI companies scrape vast amounts of data from the web without consent, contributing to the models that power their services.

“Despite the anxiety, nearly every newsroom in the country – whether large or small, for-profit or non-profit – is beginning to experiment with AI,” Dixit said.

Nearly 75% of newsroom professionals in the USA and EU have used generative AI in some form. Source: AP

According to a recent study by the Associated Press, nearly 75% of newsroom professionals in the United States and EU have used generative AI in some form. The primary applications include generating text, social media posts, headlines, and even entire articles.

Drawing on a concept from Zach Seward, Director of AI Initiatives at The New York Times, Dixit said that “AI journalism fails when it is unchecked, lazy, selfish, dishonest, and opaque, but succeeds when it is vetted, rigorous, reader-focused, truthful, and transparent.”

Projects that made effective use of AI in newsrooms

Dixit highlighted several successful applications of AI in newsrooms that go beyond content generation:

Custom GPT: Filippino journalist J. Mark Tordesila trained a custom GPT to analyse audit reports for signs of corruption. This tool enables users to upload documents and identify potential corruption, making it easier for journalists to uncover stories.

BuzzFeed News: Dixit’s former newsroom (now shuttered) used AI to detect hidden spy planes. By analysing publicly available flight data and using pattern recognition, they uncovered surveillance activities by both government and private aircraft.

Realtime.org: Created by journalists Matthew Conlon and Harsha Panduranga, this website uses AI to generate real-time charts from public data sources. The AI writes headlines and summaries, providing valuable insights to readers with minimal overhead.

These examples demonstrate that AI can enhance journalism when used thoughtfully, according to Dixit. The common thread among these projects is the integration of human oversight with AI capabilities, ensuring that the technology serves as a tool rather than a replacement for human judgment.

BuzzFeed News had used AI to detect hidden spy planes by using publicly available flight data and pattern recognition.

Indian newsrooms are catching up with AI

In the United States, many news organisations have begun publishing AI ethics guidelines. These guidelines often specify that AI will not be used to write full stories.

Comparing the AI adoption in US and Indian newsrooms, Dixit said that while US newsrooms are more advanced, Indian newsrooms are beginning to catch up.

Newsrooms are centralising AI workflows and creating roles such as “Director of AI Initiatives” to oversee these efforts. Some Indian newsrooms are starting to adopt similar approaches, Dixit said.

Many Indian newsrooms are experimenting with generative AI, but these efforts tend to be less centralised and more piecemeal.

Dixit suggested three steps for Indian newsrooms to enhance their AI integration:

AI training workshops: Conducting workshops to familiarise newsroom staff with AI tools and their applications can foster more effective use of the technology.

Enhancing core storytelling: Newsrooms should explore how AI can support and enhance their core storytelling processes, learning from successful examples in other regions.

Defining goals: Identifying goals, such as increasing traffic or improving workflow efficiency, can help newsrooms use AI strategically rather than reactively.

About 100 media executives attended our two-day event in Bengaluru.

Balancing AI’s power with human insight in journalism

Looking ahead, there are still many unanswered questions about AI’s long-term impact on journalism. Dixit raised concerns about the potential effects of AI on search engine traffic and SEO, the problem of AI hallucinations (where AI generates incorrect or nonsensical information), and the implications of licensing deals between publishers and AI companies.

In response to questions from the audience, Dixit acknowledged that while AI is currently a significant factor in journalism, it also has limitations.

For instance, AI struggles with tasks requiring a nuanced understanding of humour or complex ethical judgments. He emphasised the importance of maintaining human involvement in journalism to ensure that AI remains a supportive tool rather than a replacement for human creativity and insight.

The post 3 ways newsrooms can enhance AI integration appeared first on WAN-IFRA.