A new AI tool developed for Reuters is pioneering automated video news reports.
The London-based news corporation plans to inject a little bit of artificial intelligence into English Premier League football match reports. Last week, it announced a prototype platform that mixes pre-recorded footage of a news presenter and automated data feeds to report on the matches.
The AI was developed in collaboration with Synthesia, a London-based AI startup.
“The system has two parts,” said Reuters’ Head of Core News Products Nick Cohen for Forbes. “Firstly, we use an algorithm to combine Reuters real-time match photography and reporting with a minute-by-minute data feed of what has happened in the game.”
“This allows us to automatically generate a script for any match report, combining the words describing the event with the relevant picture.”
The AI then draws on the prerecorded-footage to create the virtual presenter, a process very similar to how deepfake videos are created. This virtual avatar can quite realistically ‘read’ the script in question “within the set parameters,” adds Cohen.
Reuters already uses algorithms to monitor live data feeds and compile text summaries of events ongoing at sporting events, mostly football. However, this will be the first automated presenter-led news summary ever. So far, early-stage testing has been very promising, according to the company.
Synthesia’s CEO Victor Riparbelli says that the process of creating the digital presenter is similar to what “you would do in Hollywood to create a digital character”, but handled in hours rather than months. While it still relies heavily on prerecorded material, Riparbelli explains that the AI can be trained to simulate entirely new video of the presenter.
Behind the coolness factor, Cohen says the AI showcases the abilities of “synthetic media”. In theory, there’s nothing preventing this AI from being adapted to much larger scales — it could monitor tens, hundreds, thousands of matches, simultaneously, across the world. Currently, the prototype is limited to use on sports, but the company would be open to expansion into other topics as well.
Reuter’s prototype is just the latest in a trend of machine learning and AI taking on tasks from the sphere of the humanities. The essay writers of yore are quickly being replaced by text-writing algorithms such as EssaySoft, potentially posing “the next big challenge to academic integrity”, according to Edsurge. Publications are increasingly looking towards AI to trawl through massive amounts of data and write first copy; a paper published last year found that “journalism without human journalists is already possible”. Finally, automated helpers such as sciNote’s Manuscript Writer can already do much of the heavy lifting of writing a scientific paper.
By this point, all these systems have severe limitations. Professors can spot AI-generated texts with ease, and Manuscript Writer only produces a basic layout of a paper after a user filled in all the relevant data. Still, it does make me curious, somewhat excited, and a little bit concerned, to think how similar systems will look in 5, 10, or 20 years.