As the year winds down, it’s tempting to look back on what an explosive and tumultuous year it’s been for generative AI and how it’s affected the media industry — not to mention the larger universe of content creation. From the awkward first steps of generative articles to the landmark deal between Axel Springer and OpenAI, it feels like a transformation is well underway.
Reality check: From what most reporters and editors tell me, actual usage of generative AI in journalism is quite low. Although most everyone is aware of GenAI tools and many news organizations have put into place formal policies around the tech (though probably not nearly enough), few newsrooms have adopted it in systemized ways. Most of what’s going on is still more or less “experimental,” such as using various tools for things like social media copy and research.
That’s why, at least with respect to the application of GenAI to media, journalism and content creation (our universe here at The Media Copilot), I’d rather spend some time talking about what’s in store for next year. It’s no secret that 2023 has been a year most of the media industry would like to forget, and it’s unclear whether AI will make things better or worse in the short term. I can’t say for sure either way, but hopefully some of these predictions will help chart a path.
And don’t forget what I promised in the headline — stick around for the gift.
1. More AI Integrations
Barring an outright collapse in the AI hype cycle, it seems clear that, where 2023 saw a proliferation of AI tools, 2024 will be the year those tools evolve and become integrated into more workflows. Things will move away from continually writing and pasting prompts into tools like ChatGPT and toward buttons that simply create words or pictures that you want, often with options to choose from.
The power of generative AI to create original content from natural-language prompts is its greatest strength, but it’s also a weakness: It’s hard to put guardrails on a process where both the input and output are so open-ended. That puts a lot of pressure on any editorial team using an AI tool to essentially be prompt engineers, finding the right set of prompts and then adjusting their editing to account for the weaknesses of GenAI (most notably hallucinations).
In 2024, I expect more generative AI tools to be built directly into CMSes and other systems connected to editorial workflow. Prompt engineering won’t go away, but it’ll be a shared area between editorial and product/engineering teams, with the goal being the creation and integration of fully vetted “prompt to publish” tools (like the kind David Caswell imagines in his excellent essay on the future of AI in news).
2. Fully AI News Operations Will Fail, but Point the Way
When Channel1 unveiled its plans for an online news network where all the on-air talent is AI generated, reactions ranged from amazement to horror. The idea of being able to launch video news programming without actually shooting anything — or anyone — with, you know, a camera, means you no longer need to employ news anchors and build a studio to match the quality of a CNN or Fox News. That’s wondrous, but also understandably terrifying if you’re a TV journalist or camera operator.
In my view, I don’t think all-AI-all-the-time operations like Channel1 will succeed, and for a very simple reason: stigma. That is, the stigma of AI-generated content is a real thing, and I don’t think it’s going away anytime soon. People naturally value things more if they know a lot of human work went into them, and that’s simply not the case for most AI-made content. And as for the unreal avatars in synthetic photos and video, I have yet to see a case where the audience for those images made a “connection” with the fake subjects as much as they would a real person. Even S1mOne needed to pretend she was real to get popular.
For that reason, I believe the news media will stay mostly human-driven in 2024, but that doesn’t mean we don’t have a lot to learn from the Channel1’s of the world. AI-centered experiments will be valuable at identifying efficiencies: finding the places where the audience doesn’t care whether the content synthetic or not.
3. Content Scandals Will Continue to Evolve
It’s dismaying that the year is ending where it began, with another publisher — this time Sports Illustrated — caught posting what looks like poor-quality AI-generated content under the guise of a real byline. One might shake their head and conclude that there’s been no progress, but the SI “scandal” actually shows the problem is evolving.
If reps from Sports Illustrated owner The Arena Group are to be believed, the blame mainly lies with a third-party content provider. That brings good news and bad news. The good news is editorial teams still prioritize human-driven content, or at the very least content with humans “in the loop.” The bad news: publishers need to think beyond their own editorial teams and understand that the proliferation of AI tools affects their entire organization. And they need to put real effort into building guardrails.
So yes, I have no doubt these kinds of AI content scandals will continue. But they’ll also serve as a key motivator for media companies to wake up and face the future.
A Gift for Readers
Thanks for making it this far, dear reader. And I don’t mean just this article. This is the last edition of The Media Copilot for 2023, and I wanted to give you something for being a faithful supporter.
First a quick look back (it’s OK some of the time!): I launched The Media Copilot at the beginning of October, mostly as a way to stay on top of how AI is affecting news media. Although I’ve been a longtime tech journalist, I’d mostly been serving in behind-the-scenes strategic roles in newsrooms for the past few years. I didn’t know if anyone would care, or even notice, what I was doing in my corner of the web.
Well, all of you — which now number in the thousands, plural — have encouraged me that embarking on this journey was probably a good idea. I’m greatly looking forward to bringing you more in 2024, especially to provide more practical guidance on how to apply generative AI to creative work.
So in the spirit of the season, here’s a gift: One of the first conversations I had for The Media Copilot Podcast was with Matt Martel, the former general manager NZME, which owns New Zealand’s BusinessDesk. It was a fascinating discussion with someone leading a team that was at the forefront of applying generative AI to news workflows. BusinessDesk is a true pioneer in the realm of GenAI, and I think a lot of publications — particularly those that cover financial news — could learn a lot from what it’s done.
I had put my conversation with Matt behind a paywall because I previously had plans to create several paywalled podcasts, but I’ve put those plans on hold for the time being. And so, for your holiday listening pleasure, I’ve made the conversation with Matt completely free and accessible to all subscribers. You can check it out on Apple, Spotify, YouTube, and probably a bunch of other places.
That’ll do it for 2023. See you in the future.
Leave a Reply