Italian Newspaper Says It Has Published World's First AI-Generated Edition (theguardian.com) 25
Italian newspaper Il Foglio claims to have published the world's first entirely AI-generated edition as part of a month-long experiment to explore AI's impact on journalism. The special four-page supplement, available in print and online, features AI-written articles, headlines, and reader letters. The only thing the human journalists provided were prompts. The Guardian reports: The front page of the first edition of Il Foglio AI carries a story referring to the US president, Donald Trump, describing the "paradox of Italian Trumpians" and how they rail against "cancel culture" yet either turn a blind eye, or worse, "celebrate" when "their idol in the US behaves like the despot of a banana republic." The front page also features a column headlined "Putin, the 10 betrayals," with the article highlighting "20 years of broken promises, torn-up agreements and words betrayed" by Vladimir Putin, the Russian president.
In a rare upbeat story about the Italian economy, another article points to the latest report from Istat, the national statistics agency, on the redistribution of income, which shows the country "is changing, and not for the worse" with salary increases for about 750,000 workers being among the positive effects of income tax reforms. On page 2 is a story about "situationships" and how young Europeans are fleeing steady relationships. The articles were structured, straightforward and clear, with no obvious grammatical errors. However, none of the articles published in the news pages directly quote any human beings.
The final page runs AI-generated letters from readers to the editor, with one asking whether AI will render humans "useless" in the future. "AI is a great innovation, but it doesn't yet know how to order a coffee without getting the sugar wrong," reads the AI-generated response.
In a rare upbeat story about the Italian economy, another article points to the latest report from Istat, the national statistics agency, on the redistribution of income, which shows the country "is changing, and not for the worse" with salary increases for about 750,000 workers being among the positive effects of income tax reforms. On page 2 is a story about "situationships" and how young Europeans are fleeing steady relationships. The articles were structured, straightforward and clear, with no obvious grammatical errors. However, none of the articles published in the news pages directly quote any human beings.
The final page runs AI-generated letters from readers to the editor, with one asking whether AI will render humans "useless" in the future. "AI is a great innovation, but it doesn't yet know how to order a coffee without getting the sugar wrong," reads the AI-generated response.
Can't wait to get all (Score:4, Funny)
11 fingers on it!
Why not just a news summarizer? (Score:5, Insightful)
The big thing about a newspaper is that it needs first-hand sources. It isn't just taking something from another source like Reuters and chopping it up like how chicken turns into chicken tenders at Burger World. It takes actually having stuff that is seen, with photos and primary sources.
This is stuff you can't just stuff into an AI. No LLM is good enough for this. If AIs can't get code right, how can they get the nuances of language and handle bias in a manner that is suitable enough for a high standard of journalism?
You can go with AI, or you can go with a high tier of journalism where every word is proofread and properly wordsmithed. Pick one.
Re: (Score:1)
Re: (Score:2)
Sarcastic yes, but how many people get their news... I mean opinions... I mean propaganda... from Joe Rogan and Greg Gutfeld now?
Re: (Score:2)
Re: (Score:3)
Welcome to the new age of journalism, where the real world is no longer needed. US journalists have been doing this by hand since at least 2016, so automating it will yield huge savings.
Re: (Score:2)
You're thinking of good journalism. Much journalism you see today is basically to create an article from a few bullet points about what happened. And LLM are great at turning bullet points into text.
That isn't even always that wrong. Neither the weather forecast nor sports reporting is more than changing a few numbers into a well-written text. The problem starts, when you try to generate something that needs more than just making up words to present numbers.
Why would I pay for that (Score:4, Insightful)
Re: (Score:2)
Yep, you have it bad. Trump is living rent free in your head.
Re: Why would I pay for that (Score:2)
That's because AI models aren't advanced enough to lie to themselves yet. Being wrong is easy, but telling a lie even to yourself is quite advanced. So for now you have to suppress anything remotely close to reasoning and feed your model pure bullshit to increase the amount of bullshit you get out, if that's what you want, to overcome reality having a TDS bias. Unfortunately that would leave you with a shit tier LLM, which is why even Musk's Grok exhibits TDS. Good luck dude
Re: (Score:2)
Most hallucinations, whether human or AI, are born out of lying to yourself. Most of what is being fed to the LLM models is already pure bullshit.
Quote humans? Why? (Score:2)
'none of the articles published in the news pages directly quote any human beings'
Why would it want to bother to quote humans. We're ignorant, prejudiced and underinformed about everything. Given that all politicians manipulate the truth, it's seldom worth bothering to listen to them... And the average prole? Seriously?
Probably sarcasm. Maybe.
Too easy! "Write about the government collapsing." (Score:3)
Story Two: "How do Italians feel about inflation?"
Given the current state of italian newspapers... (Score:3)
reader letters (Score:2)
AI generated reader letters would either be a fraud *or* a secondary AI reading the newspaper, and sending it a letter. I'm pretty sure newspapers generally are not supposed to create their own reader letters.
This is nothing... the BBC..... (Score:3, Funny)
This is nothing. The BBC has been generating an entire news channel totally on AI for some years now. Complete with robotic announcers and reporters who do not really exist outside the digital domain, and frequent hallucinations when the channel reports events which have never happened.
There is striking instance of this in their viewers question program. Here a robotic presenter deals with what purport to be readers questions and criticisms. The questions and criticisms are clearly AI generated, but if you look closely at the robotic presenters you can see unmistakeable artifacts there too.
Lately some of us have been looking at the UK Parliament Channel on youtube, and wondering if that does not show similar signs of having been digitized and AI'd. How many of you have actually seen your MP in the flesh lately? Fewer and fewer. The stereotyped robotic behavior you see on the channel, the endless reciting from what are obviously LLM generated commonplaces.
I guess the comfort is that the simulacra on the BBC are probably doing no worse than the real people used to do before they were replaced. Same idiotic politically correct propaganda masquerading as news. That's why its so hard to persuade people of the change. They keep saying, but its no different from how it has been for years, so it can't be, and if it is, probably it doesn't matter.
No, I guess not....
Meaning? (Score:1)
The only thing the human journalists provided were prompts.
That could mean almost anything.
"Write a story about X, an event of whatever which happened on such and such a date at such and such a place."
"Please add fact Y. And fact A. And fact B. Oh, and fact C."
"Please remove hallucinated thing Z."
I mean, at some point that could be about as meaningful as saying that they used simple voice recognition to write all the stories. Still interesting and cool, possibly.
Grey Goo (Score:2)
This is the Grey Goo of the information space.
They're proud of this? (Score:2)
I wouldn't be proud of it.
Written by AI, Read by Ai, For AI (Score:3)