On AI Predictions for Journalism, a Plea for Modesty

Yes, big changes are coming, but uncertainties abound

Welcome to Second Rough Draft, a newsletter about journalism in our time, how it (often its business) is evolving, and the challenges it faces.

In the last technological revolution to transform journalism, beginning in the mid-1990’s, our industry was seized by the mantra, imported from Silicon Valley, that “information wants to be free.” It was taken from Stewart Brand, he of the Whole Earth Catalog from the Sixties. What the mantra recklessly omitted was Brand’s book’s very next phrase: “information wants to be expensive.” The consequences included a raft of strategic errors, most significantly a decade of delay in deploying paywalls where they are viable—and they would likely have been viable in many more places absent the delay.

Now, another technological revolution is upon us, and I have no doubt that it will be at least as transformative as the consumer internet, perhaps more so. But the contours of that transformation for our business are only just beginning to emerge. Despite this, just as was the case in the 1990’s, predictions and typologies and visions offered with great certainty are increasingly grabbing the attention of leaders in the field.

Not the best TED Talk

This week I want to offer a note of caution and a plea for some modesty. I know that a deterministic tone makes for a better TED Talk and is more likely to make an essay go viral, but that way could also lie a new set of strategic miscalculations. I’m not telling you not to consider some of the important new thinking about AI and journalism. I’m looking hard at it, and you may want to do so as well. (I especially recommend this recent column on using AI for writing specifically and learning generally.)

There are, to be sure, big things we know about AI and journalism as we come up on four years from the debut of ChatGPT:

• Search is being radically transformed;

• Agentic AI is just beginning in earnest, and will become much more important;

• Certain tasks and functions previously performed by humans are already now more efficiently performed by machines, and more will be soon. Some of these efficiencies will make economically feasible desirable products and services that previously were not affordable;

• As a consequence of all this, some jobs will be destroyed, and others created; and

• Many of the shortcomings of AI will be overcome as products and services improve, and as users understand how to generate better results.

There are other big things, however, that seem to me not yet established, including:

• How average humans—not early adopters or tech geeks or news junkies—will choose to interact with new capabilities as they emerge. People are entitled to guess about this, and some guesses will surely be more educated than others. But if you think even the most clever analyst can draw a straight line from here to, say, 2035 or even 2030, well, I’ll take the under on that;

• Relatedly, so far AI has been a revolution largely confined to software. Will it also spawn fundamentally new hardware? I would guess so. If so, what will be the implications of that?;

• Will there be a shake out in the emerging AI platforms, or will they all just merge or grow to the sky? In the last great technological revolution, there were a lot of early casualties: AltaVista, AOL, Excite, Friendster, Gateway, the New Century Network, the Newton, Palm, Pathfinder and Pets.com to name just a few. (If you’re too young to remember those, you might get a kick out of looking them up.) If there will be a new list to go with that old one, what will be the consequences?

• Is the apparent growing political backlash against AI, and data centers in particular, just a lot of noise, or will it have impact? Will European regulatory efforts amount to anything outside of that region? What will be the technological implications of increasingly bipolar geopolitics with a second power center in China?; and

• What will be the impact on politics and philanthropy of the trillions in new wealth being generated by AI-related investments, and how much of that wealth will be sustainable?

One of the greatest known unknowns about the future of AI and journalism is the resolution of a set of legal cases brought by content providers against the platforms, mostly centered on a question of copyright law. (In an important speech earlier this week, the chairman of The New York Times Co. said they had already spent more than $20 million on these cases.) The lawsuits have moved more slowly than I had first expected, but unless the Congress suddenly rediscovers its ability to craft groundbreaking legislation (which seems to me quite unlikely to occur in this decade), the cases will eventually need to be decided in the Supreme Court.

The median justice of the Supreme Court has been on the federal bench—a pretty isolated existence, not least technologically—for 20 years, since roughly the advent of consumer broadband. Three justices have been on the bench since before the first internet browser. To confidently predict how these people will view the factual issues and resolve the critical legal questions in these cases—and thereby potentially rebalance the leverage between publishers and platforms—strikes me as foolhardy.

That said, get on with it

None of this plea for modesty in prediction and projection should be taken as a call for complacence. It remains my view that newsrooms are doing too little in the way of AI-driven product development, and devoting relatively too much of their AI efforts to internal processes. I worry increasingly that a sort of newsroom Luddism, much of it encouraged by unions, is retarding necessary investment and progress.

The future, as ever, is coming. And for journalism, as for so much else, artificial intelligence will be a big part of it. It’s just not as clear exactly what that means as some people would have you believe.