AI in the newsroom: It’s here, but needs guardrails

I tapped out my first news stories at the Hatboro Public Spirit on a bulky Olivetti-Underwood manual typewriter roughly the size and weight of a 1970 Buick Regal. We typed our stories on cheap, thin gray sheets of newsprint and edited them with thick pencils — the kind you might have used back in the day to keep score at a bowling alley.

The sleek IBM Selectric was my next technological advance, but I wasn’t satisfied. I longed to land a job at the Courier Times in Levittown, where they used computers.

Those “advanced” computers were actually IBM “dumb” terminals, hooked up to a huge, finicky mainframe housed in a hyper-refrigerated room in the back of the building, tended to by a hive of nerdy acolytes.

We called our dumb computers “tubes,” and they were down more than they were up. Usually on deadline.

I was the newsroom liaison whose job it was to call the worker bees in the refrigerated room to ask whether we were down. You could usually tell we were down because the screens would freeze, but having one person to make the call saved the nerds the trouble of having to respond to multiple phone calls from ticked-off news editors.

I developed a suitably high-tech means for spreading the bad news. I purchased a wind-up monkey with a tin drum, and when everyone in the newsroom heard the monkey banging away, they knew for sure that we were down.

The technology never improved much while I was on staff, but I was on hand to be one of the first users of the TRS-80 Model 100, a comparatively primitive notebook computer with a horizontal screen roughly the size of four or five postage stamps end to end. It worked better than the tubes.

A few years later, along came the first desktops. My first personal computer purchase was an IBM 286. First, that is, if you don’t include the Commodore 64 that was mainly good for playing games of pong.

Laptops were next up, and I was an early adopter. By this time, it became clear that the internet was going to take over the world, and it kind of did. It was also apparent that my old friends in print were losing their jobs to the new technology.

It was the first time I had heard the term “creative destruction,” which to my mind meant that the new technology might create new jobs, but it could also cost my old friends in newspapers theirs.

Now, here we are with the next new technology, and we read about Jacob Coxon, a resigned researcher at the AI company Anthropic, who says some working in the field believe AI “could kill us all by the end of the decade.”

Set aside for a moment Coxon’s doomsday scenario. (If you can.) Artificial intelligence has been associated with a number of suicides — lonely people who previewed their own doomsday scenarios with a nonjudgmental chatbot. If you’re the parent of a kid who has succumbed, that’s plenty apocalyptic enough.

AI has even been associated with several murders. For example, a 17-year-old in Middlesex County, Massachusetts, is recently alleged to have employed a chatbot to search for “theoretical ideas or fantasy stories” about the killing of his mother and younger brother. The teen has pleaded not guilty.

There are far less lethal uses of AI, but no less concerning. Some college students use AI to cheat. In 2024, the web service Turnitin suggested that of 200 million papers reviewed by the service, more than 22 million were at least 20 percent generated by AI, and that more than six million were mostly written — 80 percent or more of the total text — by AI.

And we’ve just begun to talk about AI’s unquenchable thirst for electrical and computing power, requiring the construction of massive data center complexes. You’ve been reading about that in the Local.

Yet, the reality is this: Artificial intelligence is already well out of the box. While there should be clear guardrails — our elected officials are perilously perched behind the eightball on this — AI is here to stay.

Not surprisingly, the ethical use of AI is hotly debated in the Fourth Estate.

At 75, I remain an unrepentant geek. I spend a good deal of time on OpenAI’s ChatGPT, and to a lesser extent Anthropic’s Claude — perhaps more than I should — trying to figure out how artificial intelligence might aid the cause of local journalism.

A chatbot will write a whole story, soup to nuts, if you let it. Provide it with a few salient facts and feed it a couple of juicy quotes. ChatGPT will do a fair bit of research on its own and crank out a 2,000-word think piece in minutes, or less. The average reader might never know the difference. That’s a clear violation of journalistic ethics.

On the other hand, a chatbot can compile research with blinding speed, and that can save a reporter hours of online digging. It can fact-check. It can make useful suggestions. (You thought of this, but did you think of that?) It can crank out static and interactive graphics. (Although if you’re not careful, they can all look the same.)

My overall sense of the newsroom might be characterized in the following way: Not so fast, Jeff. And caution, when it comes to AI, is well advised.

But my own take is this: AI is one more tool to come along in my career, a career shaped by evolutionary technology at every step along the way. If nothing else, AI’s facility with numbers makes it an intriguing potential resource for data journalism, an enterprise that otherwise might not be available to a small local paper with a bare-bones budget.

There must be clear ethical guidelines for the use of AI in the newsroom, and we are talking about that in depth. We should be open about what those guidelines are. We should be absolutely transparent about its use in anything we publish.

But artificial intelligence is here, and we should use it.