What can generative AI actually do for ordinary journalists?

Three journalists at a Newsrewired talk.

Lots has been said about generative AI since it hit the mainstream. But at Newsrewired last week (22 May 2024), we put the spotlight on three practical examples of generative AI for news in a show-and-tell session.

Speeding up local news delivery

Newsquest is a UK news group that has 14 AI-assisted reporters across 250 publications. This means that reporters are trained to use generative AI as part of their beat, and these reporters have publish more than 3000 AI-generated articles a month.

Its in-house draft checker works on a closed-loop system, accessing Chat GPT Microsoft Azure and connecting staight into its CMS. Reporters will place notes gathered from trusted sources into the checker, inputting a word count to receive a draft that will receive two checks: one more traditional check from the editor, and one from the AI systems themselves. 

Newsquest head of Editorial AI, Jody Doherty-Cove, says: “We hope this will alleviate the burden on important but mundane parts of the job to give time back to traditional reporters”.

He added this was not “AI for AI’s sake”, as freeing up journalists can forget about press release write-ups, and concentrate on court reporting, exclusives and investigations.

Welcome to the new age

Former BBC and Vice journalist Sophia Smith Galer built much of her name on vertical video platforms TikTok and Instagram. She built an AI chatbot, Sophina, to help others replicate her success. Though, fewer journalists use TikTok (30 per cent), than Instagram (57 per cent), according to her own research on journalists she has trained.

On closer inspection, journalists said the reason they do not engage with vertical video is because they lack the time (around 40 per cent), the skills and know-how (30 per cent) or they are simply camera-shy (25 per cent).

Sophina is trained on Smith Galer’s own successful scriptwriting technique, and produces more natural copy than a typical generative AI tool like ChatGPT. It is suited to those who wish to reduce workflows or who lack expertise in video creation. The tool helps content creators understand how long videos need to be and how to play to the algorithms.

“Sophina could very reasonably be used in the real world and is optimised to go viral,” she explains.

The chatbot was only made through significant personal investment and working with a company called BotStacks.

Standing out in a sea of beige

Daniel Flatt, co-founder of Flare Data, created an AI model that detects trends in data to discover where the stories might be. A typical news use would be interviews, as he said this would identify company-specific questions to bring to the table, making it hard for interviewees to dodge tough questions.

“We are able to tailor mass data to a specific goal so it really works for each individual person and organisation,” he explains.

Helen Philpot, managing editor of The Sun, said on a previous panel about public interest news that AI tools could create a large swathe of “beige content”, killing original reporting, something an audience member put to the panel.

All three panellists felt that with sufficient humans in the loop and at the centre of the creative process, the risk of this happening would be mitigated.

You might be interested in …