August 19, 2023 may well have been the day that journalism’s enthusiasm for artificial intelligence began its descent into the trough of disillusionment. That’s when Gannett published a wince-worthy report on a high school soccer match that would have gone largely unnoticed if it had not been obviously written by an AI bot that knew little about sports…and may have also been drunk:

“The Worthington Christian [[WINNING_TEAM_MASCOT]] defeated the Westerville North [[LOSING_TEAM_MASCOT]] 2-1 in an Ohio boys soccer game on Saturday.”

“Worthington Christian edged Westerville North 2-1 in a close encounter of the athletic kind for an Ohio boys soccer victory on Aug. 19,” the report continued.

The mocking of that story — and others generated by a tool made by an outside firm and published by the news giant — started immediately. Gannett quickly paused its experiment.

For people familiar with the Gartner Hype Cycle, this was inevitable. The Hype Cycle is a visual representation of the ups, downs, and plateaus of a new technology over a period of time. First comes the Innovation Trigger — in this case, the rollout of generative AI for the lay public in November 2022 (though artificial intelligence had been around and in good use for years before then). Then came the Period of Inflated Expectations, when people could create art with a few simple text prompts (“Look Ma, I can create a green monkey dancing on the moon, and I can’t even draw!”), followed by the almost mass hysteria over ChatGPT when it was discovered it could write school essays, tighten up resumes, imitate voices, and create broadcast scripts. Almost every newscast reporting on the technology had some version of “That report you just heard was written entirely by a robot!”

Then comes the inevitable slide into the Trough of Disillusionment, when the hype folks discover that the tech is far from perfect. In AI’s case, there were the “hallucinations,” where the programs simply made up material, and the image generation tools that didn’t take direction well or correctly display the typical number of human fingers on a human hand.

But with most technologies that enter the Hype Cycle, there eventually comes the Slope of Enlightenment (“Wait, this technology can sift through a trove of documents and summarize it faster than I can, especially on deadline”) and the Plateau of Productivity.

I believe we’re entering the plateau, and this phase will accelerate in 2025. Journalists who were and remain rightfully skeptical of AI and its impact on our profession — from eliminating jobs to adding to the public’s souring on our trustworthiness — are starting to build stuff. Good stuff. Newsrooms such as the San Francisco Chronicle are rooting around in the archives to feed a bot that can surface restaurant reviews written by Chronicle staffers to compete with review apps such as Yelp. A company called Nota, created by former Los Angeles Times technologists, makes a WordPress plugin that can suggest headlines, art, and social media posts for harried editors — who are free to modify the suggestions or discard them altogether. Tools to help with newsroom investigations, such as data-sorting tools and summarizing long interviews, are just the tip of the iceberg, as we come around to the idea of AI being a partner rather than a boogeyman in the newsroom.

My students at the Walter Cronkite School of Journalism were largely wary of AI at the beginning of my Business and Future of Journalism course at the start of this past semester, but then in groups prototyped eight unique AI-driven bots or tools that help with fact-checking, bias detection, media literacy, and ethical decision making for photographers. Others came up with AI tools to detect fashion trends and gaps in reporting. They know the dangers of AI, but they also now see the promise of this technology. I predict many more journalists will join them.