Context Collapse: Using AI to Make Journalism BETTER
AI for cool journalism stuff, not doomerism: CC #228
In this issue: Less AI doomerism & more using AI for cool stuff.
Welcome to Context Collapse, the world’s best comms newsletter. I’m Neal Ungerleider. I run Ungerleider Works and used to work as a reporter for Fast Company, write op-eds for the LA Times, and work as a senior copywriter for R/GA. This newsletter helps readers navigate the weird new world of media and gleefully ignores all the conventional wisdom about journalism, public relations, marketing, and advertising.
Have you heard the news that AI will destroy journalism? I know I’ve heard lots and lots of doomer takes lately that generative AI will replace the jobs of journalists, destroy web traffic to existing media outlets and flood the zone with shit.
Is some of that true? Probably.
But I’m more interested in the journalism problems AI will fix.
See? Here’s the thing: Generative AI is awesome. ChatGPT, Midjourney, Claude, Bing Chat, Adobe Firefly and all these other tools turbocharge creativity and give every ordinary worker the equivalent of a 24/7 intern.
That’s really cool.
Furthermore, the journalism industry in 2023 has lots of problems. Ad revenue is plunging. Audiences are microfragmenting. The weird conspiracy stuff is everywhere. The top-tier legacy publications are news for the decisonmaking class by the decisionmaking class, while good luck being able to make a living wage working at a local newspaper. At this point… Generative AI can only make things BETTER.
Over at New York, John Hermann—a writer who knows these things—posted a smart article about how AI will change the news business. Hermann’s take is this:
AI could replace journalism (unlikely, but there’s likely to be a ton of gen AI-generated crud content on the internet of the future scraping the bottom of the barrel for ad revenue).
AI will improve journalism (possible, also my take! which I’ll write about in a minute).
AI will swallow journalism (“this would turn news organizations into diminished wire services without the need to actually write whole stories for distribution — they would just be filling in blanks for some yet-to-be-determined AI-powered news-production apparatus.”).
Interesting! But journalism survived the rise of radio, the rise of television, the rise of the internet, the smartphone and a whole bunch of other technological developments. My money is on the journalism industry sticking around.
After the jump, three ways generative AI will improve journalism.
Read more of our expert articles for comms industries professionals with no paywall:
Subscribe to our free edition:
Share this article with someone who will appreciate it:
1. Gen AI will make the lives of actual journalists much easier.
Generative AI—for our purposes, tools like Adobe Firefly, ChatGPT and Claude that generate original outputs based on the data that they’ve been trained on—are supremely useful for one particular thing: Automating & streamlining knowledge work.
Journalism, of course, is knowledge work.
I worked as a journalist at publications like Fast Company for years, primarily as a business and technology writer, and saw firsthand how journalists struggle with too much to do in too few hours with not enough resources (A common professional story!).
AI in both its generative and non-generative iterations is already commonly used in journalism. Otter and Trint transcribe interviews automatically. Word and Google Docs autoedit text as you type. Grammarly streamlines copy. AI in Apple, Samsung and Google’s camera apps process newly taken pictures without the user even knowing.
But the real gold of streamlining journalist work with AI is yet to come.
Many journalists, especially business journalists, journalists at trade publications and investigative journalists, struggle with navigating and extrapolating facts from large data sources such as Excel sheets and PDFs. Current solutions include everything from machine learning tools like Python to good old-fashioned reading.
But generative AI makes this much easier. Imagine a world where you upload a PDF or a spreadsheet into an AI tool that automatically allows you to ask questions in everyday language, lets you make instant visualizations and automatically summarizes and contextualizes whatever you’re looking at.
As I write this in 2023, this technology is kind of finnicky and very imperfect. I can upload PDFs and spreadsheets to ChatGPT via code interpreter, but ChatGPT is finnicky and tends to choke if a PDF has too many images or multiple typefaces or if a spreadsheet is large. There are other tools like ChatPDF.ai, but they require a significant learning curve.
That said, OMG potential.
CLASSIFIED ADS
Advertise with Context Collapse! Reach out for our rate card.
2. Gen AI will help audiences understand news better.
One of the biggest obstacles that news outlets face is putting their coverage in context (See the name of this newsletter!). Journalists and publications often assume that readers/viewers have full understanding of the subjects of stories or of their beats… which is something almost impossible IRL.
Not everyone is glued to their screen and following Twitter for the Current Thing.
Journalism stories can be complicated. There are many topics which are extremely hard to encapsulate in 500 or 1000 words. Readers may not have time to read that 2500-word investigative piece or watch that half-hour video segment. So, let’s fix that.
Imagine a world where you read the New York Times and are able to use generative AI to instantly create a one-paragraph summary or a bulletpoint list of the article you’re reading. Imagine an embedded virtual assistant who can answer questions for you (What is an activist shareholder? What is this Supreme Court Case?) on the spot.
There are other easy implementations which should be doable by 2025 given current rates of technological advancement. One big one which I fully expect television news outlets to adopt is near-real time algorithmic reformatting of content—that is to say, Fox News or CNN segments being turned into text-based web pages or 30 second video clips on the fly. I also expect to see embedded AI factchecking in stories and instant citations for facts. This won’t be perfect (If there’s one thing people like to do on the internet, it’s making fake facts) but there’s lots of potential there.
Finally, there are weirder use cases. It’s possible to imagine gen AI-driven worldview rewrite buttons that automatically rewrite articles to reflect a certain political stance, like a supercharged Blindspot Report. Or being able to, God help us, have a conversation with an AI avatar of the journalist about the article. Future’s weird, yo.
3. Gen AI will give smaller publications more heft.
This is an important one.
The rise of the internet split the news media into three classes:
a) Incumbents who used digital to super-size and newer digital-native success stories (New York Times, Washington Post, Politico, CNN, etc.).
b) A media middle class of successful legacy publications that migrated to digital and smaller publications.
c) Smaller publications and larger legacy publications that have struggled with the transition to digital because of loss of ad revenue or unsuccessful digital transformations (Sports Illustrated, 90% of all daily local/regional newspapers in the United States).
Generative AI is a gamechanger for categories B and C—Media outlets that have defined audiences who are hungry for content, that cover areas which need coverage but which don’t have the resources to cover everything they should be able to.
And let’s be clear: I don’t mean terrible Generative AI-generated linkchurn articles like Gizmodo and G/O Media’s error-filled Star Wars listicles. The internet is already filled with terrible content written by humans that are designed to make pennies of display ad revenue; AI doesn’t change this.
But where Gen AI can truly help smaller outlets is in helping human reporters cover more in fewer hours.
Things like:
Templates for local sports matches, local government meetings (school boards, community boards, etc.), and other routine events that quickly turn notes into first drafts which can be edited by reporters into a finished product.
Automatic image enhancement and reformatting for multiple formats to be reviewed by reporters and photo editors.
Generating outlines and potential story approaches for longer articles.
Removing “uhms,” “errs” and other vocal artifacts from interview transcripts.
Generating draft copy for tweets, social media posts and other collateral based on the contents of an article.
In a situation where the internet has absolutely decimated American journalism’s traditional ad-driven revenue model and the public has consistently shown that they would rather have terrible content for free than pay for quality content (BTW: It’s true. Prove to me that I’m wrong.), this offers an alternate path for smaller publications to create the coverage their audiences expect.
This will also require journalists themselves to protect themselves against unrealistic employer expectations, but the lot of a professional journalist in the 2020s was already difficult pre-Gen AI… This just adds additional challenges to what is already a difficult professional path. But the good news is that Gen AI, indeed, gives workers superpowers.
But the question now is how long until media outlets start looking beyond the initial Gen AI hype cycle and move on to using Gen AI as a routine office tool that’s no less transformative than email or the smartphone.
Questions or something to add to the article? Leave ‘em in the comments or reply to this newsletter by email.
Read more of our expert articles for comms industries professionals with no paywall:
Share this article with someone who will appreciate it: