An editor’s flawed human ramblings on AI, journalism and the disappearance of truth, by David Horovitz

About a year ago, two very smart young men gave a presentation to our staff about how we could utilize rapidly improving AI technology to choose and write our stories.

The way I understood what they were saying, they had developed a process for scanning the internet, assessing the most important stories in the global news cycle, synthesizing the content that other outlets and writers had published on those stories, and thus enabling their media clients to quickly and cheaply produce their “own” versions based on that already published content elsewhere. An editor could tailor the product to meet his or her media outlet’s needs, including its particular style, gravitas, political orientation and more.

I stress, again, that this is what I understood them to be offering, and they indicated that they had a growing roster of satisfied clients. (It was an informative presentation, not a hard sell, and we have not spoken since.) But I may have it a little wrong. I was so awed and horrified by what they were saying that my mind immediately jumped to what I saw as the devastating implications of technology such as this for journalism, democracy and, well, truth. And that concern has only intensified in the year since.

The objection I raised loudest during our conversation was that their technology — and indeed any AI technology put to this use — would be cannibalizing the firsthand reported and written material on the internet. Any and every profit-driven media owner — or at least the many who would ultimately prioritize profit over reliable content — would be able to fill their publications, websites and social media platforms with apparently credible material, tailored to its readership, without having to employ a staff of capable reporters and editors to actually go out and see things, and write about, talk about or film them.

With time, I further posited, that very process of financially incentivized reliance on the tailored synthesis of other journalists’ original content would mean fewer and fewer journalists were being employed to produce original content. And thus the AI tools and technology would be scanning and summarizing and refashioning an ever-dwindling reservoir of actual, humanly reported words and pictures.

Gradually, in turn, the AI technology would simply be regurgitating versions of its own previously produced material, as published online by the growing number of media outlets no longer engaged in much firsthand journalism. And by still further extension, any clear, accurate sight of what the original news event was — what had happened, where, to whom, why — would become increasingly distorted and eventually obscured altogether. Finally, fewer and fewer news items and other developments of any and every kind would ever be reported by an actual person in the first place.

I’m conscious, writing this, that it sounds rather like the incoherent ramblings of an aging newspaper / news site editor. I’m conscious that my concerns may turn out to be touchingly naive, or understated and long overdue, or broadly misplaced. It could be that, ultimately, sufficient rigor and determination to chronicle humanity’s doings through journalism, and an obligation to inform the public of what is going on in order to safeguard the well-being of society, will marginalize a profit-led indifference to information flow and accuracy. But the way I see them, the signs are not encouraging.

Noticed but not prompting a vast global chorus of concern, to give one example, is the fact that, in recent months, the internet’s principal search engine, Google, the means by which we look for information, has encouraged the vast mass of humanity seeking knowledge online to use “AI mode” as its first resort. There it sits — not always, but increasingly — at top left of the Google results page, whereas, in the quaint old days of, say, two years ago, voyagers on the internet were directed to actual websites with ostensibly original, humanly produced, well-founded material.

Much of humankind may still be a little wary about using AI when hunting dependable information — many people are aware that AI chatbots such as ChatGPT can assure us that Donald Trump is the former president of the United States, produce legal precedents relied upon by lawyers in court that turn out to be fictional, or compile best books of the year lists containing volumes that were not in fact written. But such wariness will recede as the technology improves and as the temptation to accept AI as a simple, single-click, prime recourse for knowledge increases.

Remember those quaint old days of, say, five years ago, when Wikipedia was regarded as the unreliable enemy of rigorous scholarship? Nowadays, it’s one of the more trusted sources of information, despite the endless politicized battles over the accuracy of its entries.

Taking this back into the personal and journalistic, in recent weeks I’ve received job applications that give every sign of not being written by the applicant — not plagiarized, that is, but compiled with the help of AI. In one case, the applicant specified that this was what they had done — refreshingly honest, but not terribly helpful in trying to ascertain whether said applicant is actually capable of doing original, self-written work.

I was sent an opinion piece for publication, similarly, that AI detection tools indicated was 0% human — as in, AI-generated.

I was referred to a very lengthy article, published in a scholarly journal, by somebody whose writing I know well, and after briefly marveling at the extraordinary improvement of this person’s research and writing skills, I fairly rapidly assessed that there was simply no way that this person could have covered that subject — which had unfolded only recently — with this level of informed content, referenced material and dazzling articulacy, even had they spent months researching and refining it, without resorting to AI.

Why does any of that matter, you might reasonably ask? AI is an incredible tool, and if, in journalism and scholarship and arguably most every field beyond, it can enable greater speed, dexterity, coherence and success, and maybe accuracy, surely that’s a good thing.

And maybe it is… except, doesn’t it risk gradually shrinking humanity’s skill set, rigor, individuality and concern for truth?

We also, in recent days and weeks, here at ToI, published an article supplied by a news agency on the presence of American troops at the US-Israel Gaza coordination headquarters in Kiryat Gat that opened with a photograph, from the same agency, showing a uniformed US soldier speaking to a reporter in Kiryat Gat while eating falafel. Great photograph. Didn’t happen. The picture was AI-generated, as the agency’s caption had specified. I didn’t notice the caption before we used the picture. Once spotted, we took it down. (It’s published elsewhere, in some cases without a caption.)

The picture was itself taken from an AI-generated TikTok video, disseminated by the Kiryat Gat municipality (which we kept in the story, carefully described), which evidently had no issue sending around lighthearted content from its city that didn’t actually happen.

Should we care about that?

What should we make of the internet furor surrounding the glare Miss Israel supposedly directed at her Miss Palestine rival at a recent beauty contest? A video appeared to show our Melanie Shiraz giving Nadine Ayoub a nasty look at the Miss Universe event in Thailand, something she denied doing. Was the footage manipulated? Was she even standing where the clip had her standing, in a position to shoot the foul glance in the first place? We updated our story as we tried to understand what we were seeing.

Right now, there’s an incendiary AI-faked clip doing the rounds online of Israeli soldiers burning an American flag. Most people with even a superficial knowledge of Israel and a modicum of common sense would rightly doubt its veracity, and would therefore assume it has been created with AI technology and discount it. But not all. And some of those disposed to think the worst of Israel would tuck it away somewhere in that part of the brain that forms and revises beliefs and worldviews.

We all know that President Trump did not dump sewage from a plane on “No Kings” protesters last month, or sit on the beach at a Gaza Strip riviera with Prime Minister Benjamin Netanyahu in February — despite the White House-posted AI videos to that effect. It’s when the lines between fact and just-possible AI-generated fiction get blurred that the real trouble starts.

I said up high that this was rambling. Smarter people than I can better understand the implications of the contraction of original journalism and other sources of wisdom and knowledge, and the exponential spread of unreliable material. So I’m doing my best to sound an alarm for better minds to grapple with.

In my particular world of journalism, I see a rolling contraction of firsthand reporting — because, hey, it’s difficult and expensive. Less investigative journalism, acting in the public interest, exposing corruption. Fewer reporters at city council meetings and parliamentary committee sessions. Fewer writers slaving through data for days, weeks and months in the watchdog service of democracy. Fewer relationships cultivated with law enforcement. Fewer politicians who feel an obligation to subject themselves to difficult encounters with knowledgeable reporters. And the growing use by media outlets of AI tools — not to help edit their content but to produce it.

Beyond that, I fear a growing inability to distinguish between what’s true and what isn’t on our phone and computer screens. And, in time, a growing indifference to the distinction.

I see people of all generations moving to social media — rather than traditional journalism — for their news. And I know that social media owners will gradually shift further away from delivering even short news posts to users because let’s call it “non-fact-dependent” material can be so much more provocative, compelling and profitable.

I’d like to be sure that vast resources and intellect are being devoted to using AI and other developing technologies for good — little things like charting a path to tackling global warming, conflict resolution, ensuring the preservation of human life and the planet.

But I’m worried that we’re losing our capacity to establish the truth of what is unfolding around us. And that we’re well on the way to not even trying.

This piece was written amid my routine exhaustion, with all my trademark overloaded sentences and idiosyncratic formulations. All the flaws and errors are human. I did use a spellchecker.

• David Horovitz is the founding editor of The Times of Israel. He is the author of “Still Life with Bombers” (2004) and “A Little Too Close to God” (2000), and co-author of “Shalom Friend: The Life and Legacy of Yitzhak Rabin” (1996). He previously edited The Jerusalem Post (2004-2011) and The Jerusalem Report (1998-2004).

Courtesy: Times of Israel (ToI)