Craig RothThis blog post is being written by a flesh-and-blood industry analyst who had toast for breakfast and is consuming hot tea while writing this. That is now important to clarify since robo-writers (“robot” software that uses artificial intelligence and algorithms to write articles that mimic the style of human writers) are now getting a lot of attention (like here, here, here).
Journalists have, unsurprisingly, been all over the angle of whether these robots will replace journalists. Techies have been interested in how the algorithms work. The business press has written about the cost savings and target markets that are easiest to automate.
However, I am part of a team that analyzes communication, collaboration, and content and have other angles that I find interesting.
First is that the robo-writer sheds new light on the divide between structured and unstructured data. The line between unstructured freeform Word documents or email) and structured (Excel tables, databases, fields from forms) content has been getting more blurry over time. The way I see the robo-writing process, it converts structured data (box scores and play-by-play stats from a Little League game is the simplest example) into written stories. The question is why? If the whole story is in the data, why not just present the data? Is it a fallacy that people can’t absorb tables without help? Don’t businesspeople consume bulleted snippets every day in PowerPoint slides?
If the answer is that people want the big play highlighted, or the top performances, you can boldface those or annotate them. Would changing reader’s expectations to interpret the data themselves be easier than writing AI algorithms to convert a column of numbers into English? If an international affairs story is auto-generated based on a list of bulletpointed facts, would it be better to just read the bulletpoints? Sometimes narrative arc is important, but for a broad swath of stories we may demand an unstructured format out of habit, not need.
Picture this future extrapolation: as Narrative Science gets better at turning handfuls of data and millions of style examples into narrative stories, another software vendor creates AI that reads the news for you and boils it down into simple summaries for easier consumption. At that point you have the ridiculous situation of one company blowing data into stories and another trying to reduce them back. It’s like using a translation engine to translate a paragraph from English to Italian, then Italian back to English and reading it.
It seems to me that the biggest competitor to a robo-writing firm like Narrative Science isn’t journalists – it’s tolerance for direct use of structured information.
The second angle from my coverage area is that of portals – personalized delivery of information. It seems the news purveyors that use the robo-writing service gets to pick the tone of the articles, from fun and breezy to staid. Why not leave that decision in the hands of the end consumer rather than a newspaper? Maybe I want all my sports (and political news) to read like Dick Vitale was sitting next to me, rather than the mix of tones you’d get aggregating from multiple sources. And, for that matter, why not eliminate the newspaper as middleman and let readers go direct to the robo-writing firm for their news?
This type of journalism brings up many deep and interesting issues. I’m happy to see all of us writing about it while we still can.
Comments Off
Category: Content creation Tags:
