How Can Media Companies Thrive in the AI Revolution?

Artificial intelligence was the main topic of conversation, and it isn’t hard to see why. The rise of the internet and digital news platforms didn’t just disrupt traditional media companies; it devastated them. Publishers and other industry leaders have been understandably fearful that AI will finish them off.  

Over the last couple of years, we have seen a shift in this mindset, primarily driven by the emergence of AI solutions such as ChatGPT and other AI-driven open-source solutions. Some news organizations still have a fairly superficial approach to AI, looking at Generative AI as a tool to speed up news production and “summary-making.” But others are beginning to look at how AI-enabled tools such as Talisman can expand media organizations’ capacity to report, dig deeper, and offer higher-quality content. 

This was evident in this year’s Congress theme, “Shaping the Future of News Media in the AI Era.” The program featured panels and discussions on how AI will impact journalism and media writ large, and how organizations in the industry can deploy these tools. Some highlights for the FilterLabs team included the workshop “AI in the Journalism of the Future” by Fergus Bell and Tom Trewinnard of Fathm and one of the keynote addresses, “A Roadmap on AI and Future Trends,” given by Ezra Eeman, Director of Strategy & Innovation at NPO (Dutch Public Broadcasting). Another was the World Media Leaders Summit “AI in Tech and Media,” particularly Francesco Marconi’s remarks on the predictive capacity of AI for journalism. Marconi, of AppliedXL, reflected on the fact that as GenAI becomes more and more "literate" in data analysis, its capacity to generate not only content but also versions of possible development of events grows. 

And it wasn’t just the conference planners and speakers focusing on this issue. In our conversations with individual reporters, media executives, and professionals in all areas of media ops, questions of how they would adopt AI technologies across their operations were top of mind.

At FilterLabs, we believe that Large Language Models (LLMs)—especially their Natural Language Processing (NLP) capabilities—can strengthen media companies. We built our data platform Talisman to supplement existing forms of news gathering and dissemination, not replace them. So it was exciting to discuss Talisman with media professionals and explore ways that its capabilities could enhance their work and address challenges they face.

One of the things journalists voiced as we talked with them at the World Media Congress was skepticism regarding the accuracy and usefulness of attitudinal analysis. And this makes sense; many older sentiment analysis systems were fairly blunt instruments in terms of calculating sentiment scores, and users had to take the accuracy of the platform’s findings on faith. That’s why it was so exciting to talk with them about Talisman. 

Looking Beneath Sentiment Scores

One of Talisman’s signature features is the breadth and nuance of its sentiment analysis. 

Talisman gathers millions of stories a week from both traditional media sources and social media sources like Facebook, X (née Twitter), VKontakte, Odnoklassniki, Telegram, YouTube, and WeChat, as well as a variety of online forums and messaging platforms. It can then analyze narratives by frequency and sentiment. In other words, Talisman can spot which narratives are gaining traction, and whether stories about them have tended to be positive or negative. And media executives can allocate their resources accordingly. 

But in addition, Talisman also lets users go beyond sentiment scores and access the underlying artifacts. With Talisman, reporters can dig beneath a sentiment score to read the underlying posts, comments, and articles and discover what people are actually saying online. They get both sophisticated analysis of the widespread trends and granular access to the conversation. 

This level of transparency allows reporters to verify the information they use in a story or discover what the next story should be. And with access to the data underlying sentiment shifts, investigators can find and report the deeper stories and shifts behind the daily news. Within minutes of running a query, reporters can access hundreds of local sources that mention or discuss a topic in their story. 

With its use of LLMs, NLP, and transformer models, Talisman can achieve much more accurate sentiment analysis than legacy platforms. And since Talisman also includes behavioral data, reporters can supplement conventional sentiment analysis by putting what a population is saying online in context of what they are actually doing on the ground. 

A Hyper-Local View

Talisman is especially useful for coverage of foreign affairs. At a time when foreign bureaus have shuttered by the hundreds, Talisman has tools to help reporters cover the globe. Talisman’s LLMs can analyze stories from Russian, Hebrew, Arabic, and Chinese (for now—additional languages are in the works). It can help reporters identify breaking stories and emerging trends from around the world, giving them a head start on their own investigations. 

And because Talisman’s data is geolocated, it enables reporters to examine the data in a hyper-localized fashion in order to understand what is happening and what people are thinking about locally, not just nationally. Talisman offers regional breakdowns of the data, in some countries already to a county or district level. FilterLabs is constantly expanding its coverage and aims to provide world-wide coverage by the end of the year.

Talisman can also offer audience insights. Its analysis is specific not only to a particular region, but also to the audience that is consuming the information or creating it online.

Because of its capacity to detect shifts in sentiment and attitudes across a broad range of online discourse, Talisman also has a rare ability to gather data on what’s happening inside closed-off regimes. For example, Talisman’s analysis of messages on the messaging platform Telegram and various online forums enabled FilterLabs to help “Pierce Putin’s Propaganda Bubble,” in the words of The New York Times.

Continuous Coverage

With expanded access to continuously updated content, Talisman enables news organizations to provide coverage that stays abreast of topics and trends and even helps them to predict possible events in the short-term future.

Talisman’s sentiment analysis can spot emerging stories, and even uncover government propaganda campaigns as they unfold. And the combination of sentiment analysis and access to the underlying artifacts can improve both breaking news coverage and longer-form investigative reporting. The underlying data can help investigative journalists as they pursue stories that go beyond the daily news cycle. And for those wanting even greater access, FilterLabs can also allow newsrooms and media outlets to license direct access to raw data via a continuous data feed. 

Our Key Takeaways

It was a privilege to talk with so many media professionals and hear about their work, their needs, and their hopes and concerns about the role of AI in Copenhagen this week. (For FilterLabs research analyst Vasily Gatov, the Media Congress was also a wonderful opportunity to reconnect with colleagues, as he had served on WAN-IFRA’s board from 2006 to 2015!) We came back from these conversations reflecting on several key things that make Talisman so powerful for those working in media. 

Talisman can help news and media organizations track broad trends in people’s attitudes and priorities that have traditionally been supported by static polling. 

It can support investigative and deep reporting on human interest pieces through its vast aggregation of conversational and behavioral data.

It provides journalists and media outlets with access to valuable information from hard-to-reach or dangerous places, mitigating their risk. 

In short, Talisman can offer insights into the news of the day, the deeper stories behind the news, and into the audiences that watch and read the news and what matters to them.

We’ve built Talisman because we love all kinds of news media. Breaking coverage, investigative reporting, long-form narrative–we hope they are always with us. We believe that Talisman can help journalists and news outlets not just survive the AI revolution, but thrive in it.