The Impact of AI on Journalism Integrity
This article discusses the transformative role of AI in journalism, highlighting its effects on integrity, news reporting, and ethical guidelines for a reliable media future.
Peter Hulm·
This article discusses the transformative role of AI in journalism, highlighting its effects on integrity, news reporting, and ethical guidelines for a reliable media future.
And two ways reporters can react.
Several of my well-read and educated friends are deeply glum about the imminent future of informative journalism.
In the face of today’s flood of what is known as Artificial Intelligence, swamping virtually all information services, they asked me how I thought serious journalists could respond.
My immediate answer was that we are in the first flush of reader-created news and AI-produced news stories.
Sooner or later, citizens would abandon click-bait television and newscasts — hopefully for Democracy Now! on YouTube, for example. Or even explore, as we do now, the non-news information videocasts to be found across Google’s broadcasting behemoth — and realize how little we need the news format to inform ourselves about the world.
One benefit of today’s Internet is that journalists and writers no longer need to depend on a rich media owner to give them space to publish articles, and these articles can be about any topic the authors wish, not what the owner decides will appeal to the public and bring in advertisers.
But what about all the trash that brings in audiences and fills their coffers? my friends objected. What can we do about them?
Not my problem, was my reply. People should have the right to read whatever they want. I know they don’t in Russia and China. But that, I argued, is for Russians and the Chinese to work out for themselves.
So what kind of journalism should writers committed to objective reporting be working towards if they want to create an informed public, however small?
Sure, I’d like the mass of news consumers to be properly informed. But that’s really their responsibility not mine. My friends, however, favour penalties against spreading knowing falsehoods and lucrative invasions of privacy.
And what happens when AI dominates all the regular news sources? my friends asked.
There are two contrasting responses, I declared.
One strategy is to write well-researched pieces whose AI summaries will send readers to the original: the historian Heather Cox Richardson on substack is a star of the first type of response, in my view. She has 1.4 million subscribers.
The other is to do what I have been trying to do for the past three years, since I first wrote about AI and news:
I try to write pieces that are impossible to summarize through AI.
How have I done?
To test my success I compared three recent articles I published on Global Geneva with the versions produced by new products from the AI world.
Because I have promoted the Brave Browser before, I used its AI assistant Leo as one main test device. I also used the Microsoft version CoPilot for comparison.
The three stories were a cultural review, a theoretical report and a travel piece.
Both posted reservations about what I asked them to do, but I don’t think it makes any difference to the tests I imposed.
Leo said:
The page is too long for Leo. Some context could be missing from the conversation.
Copilot said:
This conversation may have info that's not public. Conversation will not be saved after this point.
That was reassuring. But how either of these represented a conversation, I don’t know.
My overall judgement: Copilot, though more awkward to use, made more of an effort to produce a detailed account.
But I wouldn’t trust its boast about any summary as here about Jacob Collier:
"This summary captures the essence of the article, focusing on Collier's career, his interactive concerts, accolades, and musical creativity."
See below for why not.
Leo produced more of a summary but often missed essential points. Copilot did too, but Leo could more easily mislead you into believing it had covered all the article had to say.
You can stop here if that’s all you want to know. Otherwise I’ll take you briefly through each test.
Jacob Collier, the North London musical whizkid who has brought audience participation, funny socks, technicolor clothing and multi-tonal improvization to concerts and stadia around the world, got his Montreux consecration as an international superstar at the Jazz Festival on 9 July in the Stravinsky auditorium. Peter Hulm, a new fan, reports.
Brave’s AI bot missed essential aspects of my piece. For example:
-- Jacob has four Grammies from his first 4 albums, not 3
-- Conducting the audience in choral works is a distinctive part of his concerts
-- His musical abilities are the envy of other professional musicians and analysts
-- I wrote about attending his Montreux session but this was ignored.
So it missed pretty much everything that justifed the piece.
Copilot featured the audience participation aspect, and summarized the article adequately with most main points. But it did not provide a short-text substitute for what I wanted to say and was much too vague to be useful.
How much can science-fiction writers teach us about contemporary society? Not in the narratives, maybe. Rather, but in their picture of society and personal interactions. Peter Hulm explores some doomster authors and considers his current news sources. CERN held its 2nd Science Fiction and the Future Conference on 28 and 29 September.
Leo missed my key point, right at the top of the article:
Is doomster fiction coming true? If so, maybe Geneva has the answer
Half the piece was about Geneva’s good news for those anxious about the future:
And much more from the international world of organizations in Switzerland.
Leo also missed this judgement from 1990:
“American culture is now insane. In its cultural, social, and political behaviour, the United States resembles a baby […] almost uneducated and increasingly uninterested in questioning and education.”
My judgement of its summary: booorrinngg.
Link to Leo on the science fiction article
-- Copilot described Kathy Acker as a science fiction writer. Not my description.
-- It did catch the CERN conference, but not its interesting projects.
-- Remarkably, it noted speeches by UN Secretary-General António Guterres.
I have to give it credit for noting the sustainable finance summits and efforts to address global hunger.
But it kinda missed the Three-Body-Problem TV series and its Chinese background — my reason for writing the piece.
Then I checked a touristic piece:
The teaser started with the weekend’s festivities and indicated that the Valais was worth exploring for events like Zermatt’s classical music festival or Niouc’s Village Festival offering free axe-throwing lessons.
Leo‘s summary missed the main focus of my article: St-Luc and its history, though my first paragraph stated:
-- One TripAdvisor affixionado says: “The Val d’Anniviers is a Swiss Valley like no other” and St-Luc, 20 km from Sierre, is its most famous stopping-place.
-- Nor did it take in Chandolin, though I said: "Chandolin, despite 19th-century and early 20th-century fame with artists, wasn’t linked by road to its ski-trail partner and neighbour St-Luc until 1960."
-- Nor did it mention the local legend that the region was invaded by Huns (hence my title).
— Somehow Leo also omitted any reference to how the Val d’Anniviers became French- rather than German-speaking: the 14th century collapse of the Illgraben rock which cut Anniviers off from the German-speaking Valais.
— Worst of all, it dropped St-Luc’s present claim to international fame: its Observatory, accessible via the Internet from anywhere on Earth.
— And as for Val d’Anniviers, Leo didn’t think it worth telling you that after 50 years of neglect, the local managed water course known as a bisse or Suonen reopened in 2020 as the first in Switzerland to be rehabilitated for agriculture.
— In these circumstances, perhaps it doesn’t matter that my reason for visiting St-Luc (a jazz concert)and my guide to events there was passed over in silence.
Copilot did better. But hardly in a way that gives you confidence.
-- It did recognize that the piece was about Valais festivals (not festivities as it reported) and noted events. But its paragraph on St-Luc and its "breathtaking views" did not mention these include many of the highest Alpine peaks.
-- Sure, it picked up on the Huns reference, "distinct Christianization timeline", "funerary cheese wheel" and the still unbuilt solar power park project (which it called just a solar park) but it was hardly an inspiring invitation to visitors.
-- And the novel bisse project and its 8km walk didn't make the cut.
-- Chandolin somehow dropped out of the summary, as did the "picnic of the dead" and Grimentz's key part in these traditions, as well as the major point that many of these traditions are recent.
-- The jazz/experimental music festival in this tiny region failed to impress Copilot at all.
-- Finally, the Observatory and its global links, a key reason for visiting now, didn't rate a mention.
Readers might need more help than a simple headline or summary to clue them into the essence of stories that give them the most useful information. So at nusereal.com I began flagging what I saw as a good article on a news topic, at the risk of passing on errors, and only then link to the ground.news summaries. That way, readers don’t have to explore for themselves if they don’t want to. It’s not a complete solution, but it seemed a step in a better direction.
Not for long. Ground.news, in addition to using AI to summarize its stories, seemed to be using the same formula that offered a repetition of the same elements from various publications, with little search for variations. So, at the beginning of July 2024, I stopped putting its links and just featured the most interesting piece I found. I flagged this change on the index page.
Here’s a 20-minute AI version from Google’s NotebookLM from my Medium text about my article on FINCA and Rupert Scofield. I’m just amazed at the quality (though Rupert resigned in 2022 not 2006, and it’s FINCA not FINSA).
Just for fun, I asked LEO to give me the week’s main stories from around the world. This was its choice:
I missed all of them. This is what I found: https://www.nusereal.com/nuseum/2024/20240630nuseweekly.php. Which tells you more of what you need to know? I hope it’s my selection.