“Rapidity is the essence of war: take advantage of the enemy’s unreadiness, make your way by unexpected routes, and attack unguarded spots.”
—Sun Tzu, The Art of War
Can journalism survive AI?
Some of the most powerful reporting entities in the world, from universities renowned for their journalism programs to top nonprofits focused on protecting journalists, are asking this question.
Nearly two years since the launch of ChatGPT in late 2022, the threat AI poses to an industry already battling accusations of corruption — and notoriously dramatic staff cuts on behalf of major media groups — is one not many could have anticipated.
I am in my second year of graduate school at one of the world’s oldest journalism institutions: the University of Missouri. There has been a growing discomfort among both students and faculty regarding how professional writers are meant to compete against fast, efficient language models that do not require time off, benefits, or empathy.
My courses have started assigning exercises specifically designed to acquaint students with ChatGPT. One asked that we paste rough drafts of our theses into the language model and ask the AI to evaluate the working product. Students voiced collective discomfort with this assignment, due to language models’ murky privacy policies: my peers worried OpenAI would feed off our developing research to improve its mechanisms and possibly distribute portions of our findings to other users.
To some people, this is an ethical and productive use of new technology. After all, as Aimee Rinehart of the Associated Press’s Local News Artificial Intelligence Initiative declared, “Those who resist the change get left behind. They don’t get promotions. They’re not included in things.” Rinehart suggests language models can “transcribe interviews, generate story ideas, and suggest headlines, among other tasks” if newsrooms are willing to take on the costs and training.
Even so, the widespread fear of obsolescence has seeped into people with Pulitzer Prizes, PhDs, decades of experience and the most earnest ambitions. This has led to classrooms incorporating AI into curriculums — before we have developed a thorough understanding of, or way to control, the obscured data storage and distribution methods that allow artificial intelligence to function and grow.
We are currently living in the Wild West of large language models. They have simply developed far faster than they can be regulated. It is irresponsible at best, and at worst unethical, for newsrooms to begin employing a tool that develops no new ideas of its own and is prone to producing factually inaccurate results, especially in a tumultuous time when credibility matters more to the industry than ever.
LLMs might replace journalists themselves in the newsroom, or as the public’s chosen source of information (should people choose to abandon their newspaper subscriptions in favor of going directly to chatbots). In the future, they may even become standard tools in the backgrounding stage of reporting. Currently, however, there is simply not enough regulation to keep in check the capacities of invention — certainly not enough for AI to be trusted with any sort of role in the news cycle yet.
Still, even the best in the industry do not yet know how to maintain an edge against the twenty-first century’s newest defining innovation. In the news media’s uphill struggle to remain a relevant part of the future, it seems the only option is to look to the past for a way forward.
The late-twentieth-century approach to journalism introduced by Hunter S. Thompson is best known for his most lasting work, “Fear and Loathing in Las Vegas.” The eclectic and countercultural writer stands in history for his challenge to the notion of objectivity: “gonzo journalism” embraces the reporter’s perspective as a part of their process. “Fear and Loathing” did so to the extent that a fictionalized alter ego of Thompson himself served as its narrator. His coverage of sporting events focused more on his experience watching them than the final results — at the heart of the gonzo piece is, unflinchingly, the writer’s point of view bringing something wholly original to the table.
Thompson’s rabid, sensationalistic approach to phenomena like gang violence and the Kentucky Derby may not be the most objective (it certainly was not the most sober-minded). It also cannot be the only reporting style that carries into the future. Beats such as war, politics and criminal justice require a more factual standpoint to bring safety-related and educational content to the public.
On those beats, journalists who want to continue offering something language models cannot would be bolstering their work by sourcing (or providing their own) original insights. Newspapers that do choose to replace writers with LLM-produced content will be able to churn out the results of any given election and call it a day. Those articles will be bland, vague, and stylistically repetitive. But reporters who seek the additional predictions or evaluations of industry experts — and are able to inject healthy amounts of perspective into public occurrences — are more likely to keep offering what LLMs cannot.
The ubiquity of LLMs may lead us to abandon many of the standard stylistic guidelines, like the famous ‘The Elements of Style’ by Strunk and White, which shaped many of the stylistic practices in both journalism and copywriting throughout the past century. If we constantly default to tools like Grammarly, we will be solely conforming to those conventions, and we will be losing our agency in how we express ideas.
Of course, stylistic conventions can be important (many of Strunk & White’s rules still hold up today, like the idea that adverbs are weak and should be limited.) And writing with our own voice, outside of existing stylistic conventions, is indeed much riskier—it takes much more work to sound clear and avoid being awkward. But because AI tools which can shape our writing style to previously established stylistic guidelines are becoming ubiquitous, the only way to maintain our authentic voices is to innovate on the stylistic conventions of the past.
Journalism can survive AI. The answer, however, does not lie in beating the machine at its own game.
Writing for the Poynter Institute, David Cohn — a leader in the journalism industry for over a decade — urged, “While we embrace the efficiencies and insights AI offers, we also preserve the irreplaceable value of human perspective and creativity.” Gonzo journalism does exactly that, proven by Thompson’s work standing the test of time long after his death.
Thompson’s influence would best be suited for intimate, personal stories that explicitly benefit from the value of human interest.
Articles that focus on people affected by the news, as opposed to events themselves, are great examples of where we might benefit from more gonzo — work to which the lived experience of the reporter adds substantial depth. It is not a universally applicable method of delivering news, but it is a style whose value we should be embracing more widely if reporters are to remain relevant alongside whatever disruptions AI may bring.
Thompson’s appeal came from the same place that articles detailing writers’ experiences with local restaurants or entertainment, and Vice News’ reporting, evoke in the socially-geared human brain: it was personal. Vice did not shy away from the notion of the journalist as a person immersed in the unfolding events.
Part of what alienates the journalism industry from the general public is the idea that its members are totally detached, “objective” in professional jargon. However, pretending not to be present does not fool the public into thinking our writing sprang from the ground of its own accord.
I mentioned “transparency” previously, and I believe this is what the public takes objectivity to mean. Nobody treats the newspaper like a catalog, hoping to be sold something or to have an idea pushed down their throat when they try to stay updated on local or global events.
Not knowing who provides money to newspapers, or the background of a given reporter, can seem disingenuous — especially when a published story might bear some conflict of interest. What is more deceitful than disclosing the viewpoints or relevant experiences of a reporter is pretending they do not exist, or pretending that the publisher’s agenda (bolstered by slanted language and blatant exaggeration) is equivalent to solid gold in the marketplace of ideas.
Transparency, not detached observation, lies at the heart of honest journalism.
Furthermore, a bullheaded infatuation with “objectivity” in stories with intimate personal value, outside of fact-centric weather and political reports, is a long-debated and tired notion; more so, it is a disservice to what journalism could be.
The reporting that sticks with me satisfies both the reader’s need for information and desire for a story.
I invite readers to think of the devastating Runner’s World article, “Twelve Minutes and a Life,” that builds to tell you, in the author’s own words, “What I want you to know about Maud.” In a survey of “how running fails Black America,” both the life and murder of Ahmaud Arbery are recounted in excruciating detail.
The author, Mitchell S. Jackson, did not hide his own Blackness in this Pulitzer Prize-winning feature. His writing is fortified through the insightful questions he raised regarding each man’s different path, asking why he himself was arrested and incarcerated — on the way to his future as a renowned journalist — while Arbery is no longer alive.
Think of the Wall Street Journal’s blank front page that ran on March 29, forcing readers to remember “HIS STORY SHOULD BE HERE” on the one-year anniversary of their reporter Evan Gershkovich’s detainment in Russia. That spread was run, and his parents were interviewed, by colleagues who felt the absence of Gershkovich for an entire “year of stolen stories, stolen joys, stolen memories.” They are not just showing you where his writing should be; that big white space is also an empty chair in their newsroom. The blank front page only works because the layout was conceived by people who miss him beyond words.
ChatGPT might have collected a surface-level biography of Arbery or a summary of the charges against Gershkovich, but these writers’ refusal to deny they care about their subjects strengthened their work undeniably.
Gonzo journalism does not need to be as fraught with artistry or misadventure as Thompson’s drug-addled escapades “journey[ing] into the heart of the American Dream.” But in the spirit of his claim that “absolute truth is a very rare and dangerous commodity in the context of professional journalism,” it is time to consider a more-than-viable approach to survive the strains of both the objectivity crisis and the introduction of AI into the mainstream workplace: acknowledging the journalist in the room.
EJ Haas is an author, a graduate journalism student at the University of Missouri, and a research assistant at Investigative Reporters & Editors, a nonprofit dedicated to improving the quality of investigative journalism.
—Bobby Becker, founder of The College Contemporary