Gannett Halts AI-Written Sports Recaps After Readers Mocked the … – Slashdot

Become a fan of Slashdot on Facebook




The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
the AI-written news trope crashed and burned pretty fast it seems
The thing is, generating AI content is *incredibly* cheap while human generated content is *incredibly* expensive.
So I don’t think this is going to go away. I predict a combination of working to improve the AI with working to lower public expectations.
If they were doing an “experiment” then I wonder what they were experimenting on? Because if they generated stories like this and ran them past an actual editor, the editor should have said “this is garbage, we can’t print this”. Clearly that’s not what they were doing though. They were just directly publishing this stuff. So, that suggests that the subjects of the experiment were actually the readers. The experiment was to publish this garbage and see how the readers reacted.
The AI needs guidance, and this is exactly what needs to be done. An Editor that tells it to do it again differently. Basically everyone using AI assisted tools are doing this, you keep guiding the AI until you start seeing results you like (approve). Only later after a very solid long run of good results, is when you can start thinking of omitting the editor.
Clearly they thought they could axe everyone at once, not yet. AI is only ready after you train it well, like you would with humans.
Problem being I’ve seen a lot of folks spend more time trying to get the AI to do the right thing than it would take to just… Do the thing.
Now a lot of low quality “news” could let AI handle things fine. Even before generative AI, the name of the game was to extend a clickbait sentence to pages of vacuous text to leave plenty of places for ads. AI is just fine at writing that obnoxious crap, and I assume it must work given how much it seems to still be done.
editors.
those people that actually read the story.
and tell the writer what to rewrite.
and sometimes rewrite, and read again

I predict a combination of working to improve the AI with working to lower public expectations.

I predict a combination of working to improve the AI with working to lower public expectations.
I agree with you. I also expect that the ultimate result of those lowered expectations will be the further erosion of these firms income – “AI” is not actually going to solve anything for them.
Side note: I work in a STEM department. One thing I find interesting is that some smart people are very bullish about “AI”, while others think the current LLM technology has largely peaked and is unlikely to significantly advance much further.
I think the easy and impressive LLM demos will start to look less impressive as people get used to them, and then we’ll get to the hard part, which is learning how to use the tool effectively and refining the tool to that use.
In other words, the yellow-press reporters will lose their job.
I fail to see the problem.
You mean paying reporters shit so that only morons who couldn’t find other work would write most news articles (poorly) for the past ~15 years was actually a far-sighted plan to ease the reading public into lowered expectations from shitty AI?
Brilliant!
And here I was thinking it was just about padding the profit margins.
The gripping hand is that *verifying* content already took more effort than generating it, so Amdahl’s Law says you don’t gain much by automating the generation.
In addition, the generation of content is where you train the people to verify the content, so you need to have people doing it anyway.
The problem I see is that the AI does not know / understand exactly what happens on the field.
And someone just enters in a few details and “generate” whatever crap comes out.
As long as the AI system does not have “visual AI” so that it can “see” what happens and with an understanding of the rules of whatever sport, it will still be clueless.
The cycle will be: AI generates the flesh of the article, humans give it some touches, and then the AI will do the final check and automate the publication.
AI is better than human sports writers.

AI is better than human sports writers.

AI is better than human sports writers.
My dog is better than most sports writers.
My browser had no trouble resolving that to “The Worthington Christian Flaming Crosses defeated the Westerville North Mighty Saltines 2-1 in an Ohio boys soccer game on Saturday, last Saturday, this past Saturday, over the weekend….”
So there was a 6% staff layoff, but NOBODY bothered to proofread the AI posted articles at all?
Maybe they did- and let the machine’s awful versions pass in the hopes the machine would get fired.
Seems to have worked…
Given the situation, it doesn’t seem like any active sabotage needs to have occurred – the people who’d have normally proofread the copy are probably among the unemployed former staff.
You could argue that Gannett’s management effectively sabotaged the company, though.
The AI machines responsible for sacking the AI machines that wrote the articles, have been sacked.

LedeAI allows reporters and editors to focus on “journalism that drives impact in the communities they serve.”

LedeAI allows reporters and editors to focus on “journalism that drives impact in the communities they serve.”
Like… anything but ballgame scores.
Gannett papers will have to wait a little bit longer to fire that one salaried reporter making $21K a year.
The people still doing local reporting seem to be in their 70s, if Ganett waits a few years they won’t have to fire them at all.
One of my local newspapers, established for over a century, got rid of nearly all its journalists a few years ago and now most of the stories are about as well-written as a ChatGPT response. Writing the entire thing by AI wouldn’t make it much worse, because it was ruined already.
Fortunately another company that does local newspapers started up another newspaper covering the area shortly after the incumbent paper was enshittified, with actual journalism – news, analysis, even some breaking stories – and I’d prefer that one didn’t become AI drivel. I subscribe to that one, in the hope that they will see revenue and not turn to shit.
I like my local paper. If you see an obnoxious ad you can turn the page. If there’s a “clickbait” style headline you can quickly skim the article to see if there’s any actual worthwhile content in there. If there’s a photo, it was probably taken on the scene, not some generic shutterstock footage like the online news sources use to make an article “look right.” There are some real advantages to a newspaper that is printed on paper, compared to an online source.*
Not all of my local paper is brilliantly-writt
That should tell you how qualified it is at doing anything else…
Most human journalists are too dumb to write about sports. And yet you somehow trust them to inform you about that and everything else you’re not a direct witness to.
Realizations like this are what made me a believer in a version of the great man theory of history.
AI is even too dumb to write about sports.
We’re not really talking about something that requires a lot of analysis or understanding. It’s a score or ranking, you report it, you list the people who did something particular during the event. We’re not talking about reading between the lines of a press conference after a diplomatic summit and finding some subtle hints to analyze what they might have meant. You have a score to report.
You had one damn job, AI!
That would probably be something like Speedball.
The game [wikipedia.org], not the drug cocktail.
… than Howard Cosell.
You must be the pride of <SUBJECT HOMETOWN HERE>.
It just sounds like the cheesy wording is a decade or two older than the current pablum.
When you have human writers, somebody proofreads and edits their work.
Why did Gannett think they could skip these steps with relatively unproven AI technology?
I think there’s a place for AI writing news stories, you just can’t take whatever it writes and publish it without proper review, just like you wouldn’t take text written by a college intern (or even a seasoned pro) and publish it without review.
There may be more comments in this discussion. Without JavaScript enabled, you might want to turn on Classic Discussion System in your preferences instead.
After Hurricane-Caused Flooding, Some EVs Exposed To Saltwater Caught Fire
China Accused of ‘Coordinated Disinformation Campaign’ About Fukushima Waste Water in Multiple Countries
“Home life as we understand it is no more natural to us than a cage is to a cockatoo.” — George Bernard Shaw

source

Jesse
https://playwithchatgtp.com