Focus must move from what AI tech does to what it can do for local readers
Big Data For News Publishers | 20 May 2024
One-and-a-half years have passed since generative AI went mainstream. At United Robots, we’d already been working in the AI/newsroom space for years when ChatGPT came along. I admit, we were kind of side-tracked focusing on where our “old AI” tech fit with this new stuff.
But then the penny dropped.
Before I go into where and why that happened, just a quick recap of what it is our company does. United Robots builds text robots using rules-based AI, and we sell the automated content it produces. The raw material is structured, verified data. This means only facts available in the data set end up in the text. Hallucinations cannot happen because, as media analyst Thomas Baekdal put it, the type of AI we build is journalistically limited by design.
In contrast, generative AI based on current large language models (LLMs), like ChatGPT, simply looks for language patterns to create its texts. It is inherently unable to distinguish between fact and fiction. It is, however, fast, flexible, and creative.
The downside of building text robots, compared to using ChatGPT, is that it requires expertise. It is a complex process involving programmers, writers (because the robots don’t actually create the text segments; people do), data experts, and linguists. The trade-off for an output of safe-to-use content is that it takes a lot of time and work.
However, a focus on the workings and strengths of the respective AI tech is beside the point.
In April, a group of us attended the second annual Nordic AI in Media Summit in Copenhagen (many of the presentations can be watched on demand). Listening to the generative AI cases from across many newsrooms, it was apparent that we’re still at the very beginning of using this technology. The majority of the use cases were around gaining efficiencies by creating newsroom tools to assist with various text-based tasks.
As keynote speaker Professor Nick Diakopoulos of Northwestern University noted, we’re still in a production focused phase and have yet to move on to taking an audience focus: “Think more about prototyping to create value for individuals and society as opposed to focusing just on how you work in newsrooms.”
After the Summit, we went home and flipped the script from a tech focus to a (local) audience focus. We started to map out what value(s) different types of AI or automation can offer to news audiences, in the context of local content, as noted in the chart below.
Generative AI is fantastic at making local journalism accessible to more people by converting between media formats (like text-to-audio) or by converting content to different storytelling formats to attract, for example, younger audiences. Because of the efficiencies gained through LLM-based text tools, reporters should be able to spend more time talking to people in local communities — an AI-related goal recently illustrated by Amedia in Norway.
GenAI is also good at going through data to find hidden stories, as in the case from Sydsvenskan in Sweden, presented by award-winning journalist Inas Hamdan at the AI Summit.
Automating the content production process (including using United Robots’ rules-based AI), on the other hand, does other things for local readers.
This tech is generally used to produce large volumes of texts based on hyper-local data, such as house sales on a neighbourhood level, match reports from lower-division sports, or newly registered companies. That means each small community gets stories very close to home, relevant to them.
And, because the content can be auto-published, it’s possible to provide 24/7 instant updates of news, such as extreme weather warnings, as illustrated in this case from Advance Local in the United States.
What both of these types of AI tech have in common is that they can help free up reporters’ time in local newsrooms — a differentiating resource far more valuable to a local news brand than any AI.