How we used AI in the Capital Brief build — and what we learned from it
Some media companies have experimented with AI in news articles. We think the risks are too high, right now.
After months of planning, designing, testing and developing, the Capital Brief website went live in late August. We hope you are enjoying it.
One of the most interesting aspects of the build, from my perspective, was how our developer Ben White and designer Alice Roberts used artificial intelligence in the process. And the lessons from that experience have influenced the way we approach AI in our newsroom and coverage.
Here is Ben's explanation of the reasoning for using AI-generated content in our testing phase:
Alice and I had proposed a slightly unconventional way to categorise content, and we wanted to fill up the CMS with some example stories to see if the idea worked at scale. My first thought was to some kind of "lipsum generator" , but I couldn't find one that was configurable enough.
At this point, I was using ChatGPT almost daily as a Google replacement for coding questions, but remained skeptical (and ethically opposed) to using it for any form of creative work. But we had a pragmatic need here, and I was morbidly interested in how ChatGPT would handle being asked to generate content for a newsroom. So I put together a prompt, and set it to work, churning out stories on the hour to fill up our index pages.
