AI and the Future of Journalism

Notes from the Perugia Journalism Festival

This past week, I had the privilege of attending the annual journalism festival held in Perugia, Italy, one of the largest gatherings of journalists in the world. I spoke on a panel discussing the potential impacts of artificial intelligence on the industry.

You can watch the full discussion here, but the short version is that we attempted to dispassionately analyze Elon Musk’s argument that AI would fundamentally transform journalism, with his platforms, X and Grok, playing a vital role in creating a host of independent news creators.

My inclination is to take Musk’s proposals seriously but not literally, so to speak. In fact, Musk’s actions since taking over Twitter — like penalizing posts with external links and amplifying various hateful sub-communities — have *decreased* the utility of that platform as a host for both traditional journalists and the emerging ecosystem that I call the “new news media,” which I’ve been writing about for years.

That said, the Perugia festival itself illustrated the extent to which the directionality of Musk’s prediction is correct. Some of the best attended programs, with lines of waiting attendees running hundreds of meters out of the venue doors, featured news creators like V Spehar (UnderTheDeskNews, TikTok), Johnny Harris (eponymous, YouTube), and Taylor Lorenz (User Mag, Substack). Hundreds of capital - “J” journalists queuing up a bit like fans at a concert.

It’s understandable. Who wouldn’t be at least intrigued by their success? At a time when many legacy news outlets are closing and laying off journalists, these “new news” creators have built audiences as large as all but the biggest newspapers. Some are actively expanding; during his panel, Johnny Harris made a crack about how he just wanted to make explainer videos on YouTube and ended up stumbling into becoming a full-blown media company with dozens of employees.

They also illustrate another point I made during the panel, that information, insight, and expertise are widely distributed. Every one of the 8 billion people inhabiting the world knows things that are potentially newsworthy under the right circumstances. But most of that knowledge is lost because the institutions responsible for gathering, collating, and distributing it as news are centralized. The Austrian economist Friedrich Hayek called this the “Knowledge Problem,” although he meant it as a critique of attempts at centralized government management of the economy. But the insight can apply to all centralized institutions, newsrooms included.

Thankfully, innovations like artificial intelligence reduce that informational deadweight loss. Local knowledge and specialized expertise can bubble up from the masses via AI-powered discovery and distribution networks.

You never know, for example, who might have the odd pieces of knowledge necessary to make a connection between Joseph Smith’s teachings on coffee in the 1840s and the sudden popularity of Stanley Tumblers in the 2020s. And up until recently, it would have been impossible for that strange insight to reach an audience of millions without access to traditional news outlets.

Thus far, this makes AI similar to prior information revolutions, like those generated by the invention of writing, the printing press, and so on. All of these innovations effected distribution, dramatically expanding the reach of a piece of information. Information sharing was once restricted to the number of people who could crouch around a Paleolithic fire and listen to an oral history of the tribe. But today we have expanded the ability of a single author to reach thousands of readers via the printing press or a single speaker to reach millions via radio or television or the internet.

Reducing the cost of news distribution to nearly zero removes the largest barrier to entry for news creators. Journalists once needed news outlets in order to have an audience. They needed the printers, warehouses, delivery boys, and all the hundreds of networked people that made up the entity we call a “news outlet” to reach a large enough audience. But now, a single news creator can potentially be matched to a global audience in the millions via machine-learning algorithms. The internet began a digital distribution revolution that AI-powered discovery is completing.

Removing these barriers to entry has already expanded the pool of news creators beyond the limits of the old news system made up of traditional news outlets, journalism schools, and professionalized journalists. While some “new news” creators have a background in the legacy news media — like Harris at Vox and Lorenz at WaPo — most do not. For example, Pew recently found that 84% of news creators on TikTok have no news industry background. That includes Spehar, who was a chef prior to starting her account during the pandemic.

But where AI has the potential to dwarf prior information revolutions is that it affects not only the distribution end of the news creation funnel but also the front end, the news creation end. It could exponentially increase the pool of people who are able to create the news.

Let’s take Johnny Harris as an example. By reducing distribution barriers, the internet and simple algorithms made it feasible for him to go independent. He could create videos on his own Youtube channel instead of for the Vox website and still reach millions of viewers. But Harris needs a wide range of fairly specialized skills to do what he does. He needs to know video and audio editing, the stylistic particulars of the platforms he uploads to, how to market his content, the time to actually do all of the above, and, importantly, the personal gumption to give up a stable career and go independent. That’s quite a lot! How many people in the world actually possess all of those skills, attributes, and opportunities? Hundreds? Thousands?

But now think of that collection of needed skills through the lens of AI technologies that are already available or in development. There are AI editing tools that can reduce production time from days to hours of work. There will be AI tools that can take a single piece of content and optimize it for different platforms. And soon there will be AI agents that can stitch together all of these specialized functions into a single production process, dramatically reducing the amount of human throughput required to turn an idea into a news artifact. Johnny Harris needs a media company; but, in the future, AI-powered Johnny Harris 2.0 might not.

We are headed towards a future in which the global pool of news creators is many times larger than it currently is, although it will be less concentrated, less routinized, and less professionalized than the old industrial model of journalism.