Speaker : Lisa Gibbs, automation / Al newsroom Lead at Associated Press / responsable de l’automatisation et de l’intelligence artificielle à Associated Press.
Usually, when we talk about robot writing, we talk about « natural language generation »: using algorithms to create simple templates for stories that you can use to produce a story very quickly. Four years ago, AP decided that they wanted to automate stories about earnings. They went from 300 to 3,700 earning stories and thousands more sports previews and recaps. By the end of the next year, they will be producing over 40 000 automated stories.
Another example is a company called RADAR, which takes national data and writes a template, making it possible to produce a story for every town.
Robots do what journalists can’t
Journalists are great at finding and writing stories, but they can’t scale production. The technologies behind automation allow for volume and differentiation which, in turn, makes it possible to satisfy more needs and have increased personalization. It also frees time so that the reporters can produce higher-impact work.
Of course, there are several issues you have to deal with when using automation for your workflow. Should we disclose that we are using these technologies? Is it worth emphasizing that robots don’t write the stories, but that journalists simply write based on templates? AP journalists, for instance, actually get bylines. This issue around transparency and best practices is still in its early stages.
Before automation, AP editors decided which story they’d write or not. Now, there’s an editor dedicated to automated content, who has to pay greater attention to the data used to automate stories.
As for natural language processing, we can build little templates and algorithms for robots to analyze each and every story and write a summary. At AP, they’re experimenting with having robots producing summaries for some stories.
Is this a journalist ?
A robot really isn’t a journalist, though, and is only doing what we’re telling it to do. But they make for great journalist assistants: they can help write stories faster and do « the boring job ». « Robots are not going to take our jobs! », the speaker reassures.
Crédits photo : Ulysse Bellier