Writers Unions Are Gearing Up for a Fight With AI

ChatGPT isn’t just making headlines, it’s also writing them. Now, reporters are getting ready to protect themselves from what comes next.

 

This is the first installment of Automate Me, an ongoing Study Hall series about how AI technology is transforming the media and publishing landscape.

Lately, it seems like every newsroom is announcing plans to experiment with generative AI, the class of artificial intelligence that can produce text that reads like it was written by a human. In January, Futurism reported that CNET had begun discreetly using the technology to produce stories riddled with factual errors, and two months later, Bloomberg revealed its own BloombergGPT tool. WIRED recently outlined potential plans to use AI to generate story ideas and brainstorm headlines, and by May, VentureBeat’s editorial director had told staff the company would “use AI to empower our reporting.” (The publications set a hard line on using ChatGPT to generate text.) 

Perhaps the most alarming sign of an AI takeover came when BuzzFeed CEO Jonah Peretti shuttered the company’s News division last month, telling staff in a memo that AI would play a role in BuzzFeed’s future. In the past, he’d suggested that the technology could be the foundation for a new media business model. 

Artificial intelligence tools are designed to imitate human language, and draw from content across the internet to do so. For that reason, they are capable of drafting stories, copy-editing, and even assembling simple emails and public records requests — though perhaps not without the oversight of humans who do these jobs already.

Still, these tools are not capable of doing research that goes significantly beyond what’s already available on the internet. There’s also no guarantee that the content produced by ChatGPT, or similar tools, like Google’s Bard chatbot, is accurate, unbiased, and hasn’t misrepresented someone else’s work. 

As newsrooms explore using the technology in the reporting process, writers’ unions are likely to find themselves wrestling with some of AI’s biggest flaws, which create myriad labor and intellectual property issues that the media industry will need to contend with. Chief among these concerns is that generative AI will be used to replace, or greatly diminish, the number of journalists working for a publication. As such, AI is emerging as a key issue for some unions. 

In March, SAG-AFTRA, which represents journalists, actors, and singers — as well as other performers — declared that digitally reproducing an artist’s “voice, likeness and performance” is an area that “must be bargained with the union.” Negotiators representing striking Hollywood writers have also proposed placing limits on the technology. And recently, the Financial Times Specialist union won the right to bargain in cases when new technology that will impact its members is introduced into the newsroom, according to the WGA-East, which represents them. 

The WGAE, now representing at least 20 newsrooms (including CNET), said AI is at the forefront of members’ minds as it prepares for contract negotiations in the next two years. 

“AI is an existential threat to the work done by Guild members and the WGAE has already begun fighting to build industry standards around the use of AI tools,” the guild told Study Hall in a statement. “We are committed to making sure all future collective bargaining agreements recognize that AI-produced content can never ethically replace journalism created by human beings.” 

Media outlets have utilized automation technology since at least 2014, when the Associated Press began using the software to produce minor league and college sports stories. Since then, newsrooms have used automation tools to report earnings, regional election results, earthquakes, and scoring during sports games. Automation can easily pull data from authoritative sources into simple story formats, filling in the blanks — Mad Libs style.  

But the latest generation of artificial intelligence does far more than previous iterations. The copy and images created by services like ChatGPT and DALL-E are trained with machine learning algorithms, which then create content by drawing from billions of original works that are available on the internet. These systems are designed to imitate human work — ChatGPT essentially works like an extremely advanced version of autocomplete — and predict what would seem realistic. 

The technology, thus, raises questions for some authors regarding liability, authorship, and transparency. Now, they are demanding some form of compensation, given that their works (including, possibly, copyrighted books) were used to build this extremely profitable technology. 

“[AI companies] think that 100 percent of the added value was in the software and 0 percent is a training corpus, which is nonsense,” said Edward Hasbrouck, a volunteer for the National Writers Union and its Freelance Solidarity Project, who has led the organization’s advocacy with the US Copyright Office. “Given that generative AI has generated a billion dollars, what fraction of that is owed to the creators of a training corpus? That’s a question we haven’t even begun to answer yet.”

Mary Rasenberger, the executive director of the Authors Guild, which represents writers and illustrators, said the organization is interested in finding ways to include freelancers in their negotiations because “you need the ability to collectively bargain” in order to take on artificial intelligence. 

The Authors Guild released a new model clause for potential contracts that prohibits using a writer’s work to train AI. AI could, however, ultimately play a significant, but hard to define, role in many creators’ works when they themselves use the technology. In an Authors Guild survey of 1,700 authors, 23 percent of respondents said they already used artificial intelligence in their creative process — and more than half of those respondents said they used ChatGPT. 

There is precedent for creatives successfully advocating for themselves against the seemingly inevitable forces of AI. Last December, Stable Diffusion, a popular AI image generator service, announced that it would allow artists to “opt out” of its training database, after creatives pushed back. 

But companies may not stick to their commitments regarding how they’ll use AI. We’ve already seen this happen. At the start of the year, BuzzFeed said it would use ChatGPT to help produce its popular online quizzes. In recent months, though, the company started using the technology to write basic guides, too, despite Peretti himself cautioning against using “AI to make content farms.” Additionally, writers may need other protections, including control over their byline and the right to oversee any work made by artificial intelligence, which, again, can produce false claims.  

At present, some of the most impactful work unions can do regarding AI may not be at the negotiation table, but in conversations with regulators and legislators. The Teamsters Union, representing many of the long-haul truck drivers in the US, have been active in policy discussions around regulating self-driving trucks — a potentially drastic new level of automation in their industry.  

There are some signs that media unions could pursue a similar strategy. Already, representatives of the NWU and the Authors Guild have voiced concerns to the US Copyright Office, which in March said that works created with AI could be copyrighted, depending on how involved a human was in the project, and recently launched an initiative to examine many of the new regulatory questions raised by AI.

It remains to be seen whether all these efforts will be enough. AI, with all its flaws, may seem appealing to venture capitalists and media CEOs looking to cut costs in an era where publications are struggling to turn a profit. But it’s always worth the fight, and we could very well win. Though humans are fickle and unpredictable, we are definitely more creative than any machine. 

---

LUKEWARM TAKE: CNN IS HANNAH HORVATH

Last Wednesday, CNN hosted former President Donald Trump for a town hall that was watched by over 3.3 million people. He quadrupled down on bogus voter fraud conspiracies, joked about the victim of his sexual abuse verdict, and called moderator Kaitlan Collins a “nasty person” in an homage to 2016 that no one asked for. The spectacle received so much backlash that Anderson Cooper delivered a bizarre, self-righteous monologue on air the following day defending CNN for forcing viewers to step outside of their “silo.” He then moderated a panel featuring none other than Anthony Scaramucci, an infamous veteran of the Trump White House. Everyone is once again having the same debate about whether or not a reputable news outlet should have platformed this particular Florida Man. It’s the same whirlwind of bullshit we’ve been trapped in since 2015. At this point, CNN is basically like Hannah Horvath in that “Girls” episode where a girlboss editor instructs her to snort cocaine so she can have something salacious to write about. Sporting a neon yellow mesh top, a coke-fueled Horvath ends up dancing at a Brooklyn nightclub: sweaty, deranged, and desperate to create clickbait by any means necessary. CNN, which has been struggling to juice up ratings, is basically giving the same vibe — sadly, minus the mesh. —Daniel Spielberger

---

COMINGS & GOINGS 

—Tami Abdollah is joining Noema Magazine as a senior editor. 
—Carly Olson is now the Los Angeles Times’ consumer economy reporter. She will be covering “retail, e-commerce” and “how the economy is affecting consumer behavior.”
—Jeremy Childs is joining the Los Angeles Times as a night reporter covering breaking news on the Fast Break Desk. 
—Matt Wille is now managing editor at V Magazine. 
—Matthew Zeitlin is joining Heatmap News as a correspondent “covering the energy transition.”  
—JP Mangalindan is joining People as a senior TV writer next month. 

---

EVERYTHING ELSE

—The Messenger, a media startup founded by media entrepreneur Jimmy Finklestein, launched on Monday. According to Axios, the outlet employs over 150 journalists who cover everything from politics to travel. On its first day, The Messenger blasted off over 140 stories (one of them was an “exclusive” interview with a certain former president who loves to hate the press).
—Former Fox News host Tucker Carlson is taking his pathetic, albeit lucrative, shtick to Twitter. In a tweet that has been viewed more than 130 million times because hell is a place on Earth, the demagogic pundit announced that he will host a new show on Twitter dot com. The website has been very buggy as of late, so best of luck uploading an MP4 file on a regular basis!
—Elon Musk named Linda Yaccarino, the former head of advertising and partnerships at NBCUniversal, as CEO of Twitter. “Looking forward to working with Linda to transform this platform into X, the everything app,” the Doge King, who is moving forward as the company’s  CTO, tweeted. What exactly will the “everything app” entail? I guess we will all find out together and cry!!!
—Paramount is shutting down MTV News as part of a wave of layoffs that will impact 25 percent of its US team.
—On Monday, Vice Media officially filed for bankruptcy. In 2017, the media company was worth over $5.7 billion dollars. Now, Fortress Investment Group, Soros Fund Management, and other creditors could acquire it for $225 million. As of now, the website — and its sister businesses, an ad agency and Refinery29 — is still running business as usual. Remember that time in 2015 when Shane Smith splurged on a $300,000 steakhouse dinner at the Bellagio Hotel?
—OpenAI CEO Sam Altman headed to a Senate hearing on Tuesday where he “optimistically” warned that AI could cause “significant harm to the world” and said that “if this technology goes wrong, it can go quite wrong.” That’s heartening!