When AI Misfires Police Get Froggy With False Reports

Call it a 21st-century problem. When you use artificial intelligence as a time-saver, you still have to check your work! Otherwise, you’ll end up as the feature story for some columnist in St. Cloud, Minnesota.

Case in point: contrary to printed media reports, police officers in Heber City, Utah, are not turning into frogs.

Get our free mobile app

Artificial Intelligence has invaded everything we do. I use it every day; once I finish an article, I use A.I. to generate summaries and other tedious website “stuff” that I need before I can publish. It’s a time-saver. But I have to go over the results very carefully, because my A.I. platform has a long history of hallucinating.

“Hallucination” is the term used to explain when A.I. uses information to generate incorrect responses. Critics are quick to point out that hallucinations are more than just funny side-stories, it can reinforce pre-existing biases and make the users less accountable for their work. Like when it put out a news release in Utah that claimed police officers were turning into frogs.

The Heber City police department was trying out a new program called Draft One to automatically generate police reports from body camera footage. The initial attempts went well; officers claimed the new program saved them hours a day. Until it published a story, picked up by the local newspaper, that claimed officers on the scene mysteriously transformed into frogs.

It seems the program, while writing a report based on bodycam footage, caught a glimpse of a television in the background playing Disney’s “The Princess and the Frog”. Artificial intelligence can’t tell the difference between reality and what’s on the TV in the background. In an effort to accurately report on what was happening, it included people turning into frogs.

There’s no official word yet if officers will continue using the program, but it’s safe to assume they’ll be checking the reports a little more closely before they publish them.