Washington Post's AI-Generated Podcasts Rife With Errors, Fictional Quotes (semafor.com) 35
The Washington Post's top standards editor Thursday decried "frustrating" errors in its new AI-generated personalized podcasts, whose launch has been met with distress by its journalists. From a report: Earlier this week, the Post announced that it was rolling out personalized AI-generated podcasts for users of the paper's mobile app. In a release, the paper said users will be able to choose preferred topics and AI hosts, and could "shape their own briefing, select their topics, set their lengths, pick their hosts and soon even ask questions using our Ask The Post AI technology."
But less than 48 hours since the product was released, people within the Post have flagged what four sources described as multiple mistakes in personalized podcasts. The errors have ranged from relatively minor pronunciation gaffes to significant changes to story content, like misattributing or inventing quotes and inserting commentary, such as interpreting a source's quotes as the paper's position on an issue.
According to four people familiar with the situation, the errors have alarmed senior newsroom leaders who have acknowledged in an internal Slack channel that the product's output is not living up to the paper's standards. In a message to other WaPo staff shared with Semafor, head of standards Karen Pensiero wrote that the errors have been "frustrating for all of us."
But less than 48 hours since the product was released, people within the Post have flagged what four sources described as multiple mistakes in personalized podcasts. The errors have ranged from relatively minor pronunciation gaffes to significant changes to story content, like misattributing or inventing quotes and inserting commentary, such as interpreting a source's quotes as the paper's position on an issue.
According to four people familiar with the situation, the errors have alarmed senior newsroom leaders who have acknowledged in an internal Slack channel that the product's output is not living up to the paper's standards. In a message to other WaPo staff shared with Semafor, head of standards Karen Pensiero wrote that the errors have been "frustrating for all of us."
Sounds like a scheme push out ads (Score:5, Insightful)
Re: (Score:3)
Clearly the people who thought it was a good idea.
Let's be honest, the minor pronunciation errors is not really a big deal if the errors are words that are seldom said. However certain words are always going to trip up TTS technology, and for the most part, a TTS is terrible to listen to, because it lacks the cadence of a human speaking. Like there is a video podcast around here that the Voice acting for the video is clearly a TTS, and the TTS trips up on native words and can't pronounce them worth a damn w
Everyone's bilding stupid junk (Score:5, Insightful)
The equivalent of all those LLM-spam "books" you see on Amazon.
They are product pumped out with no regard to quality control, dependent on potential consumers mistakenly thinking there's something like fact checking or editing going on because of the name on the tin.
Worse, this is all "hello world" style LLM programming - give it your cute little prompt ("That's where the real engineering goes!"), throw it a couple links to RAG in, and slap an ad on it. There's nothing here a vaguely competent teenager can't build for themself or the robot can't build for them.
That itself is a nested scam - pretending that any of this crap is difficult, that you need your betters at WaPo to write it for you.
And that's just the tool itself - the next problem is Bezos made it clear that anyone with integrity should hit the road, and those folks did. So all the source material may as well be robot poop already, quality-wise.
Re: (Score:2)
Continuing the speed-run towards being a junk-rag (Score:5, Informative)
We cancelled our subscription to the Post after the ownership kiboshed a the editorial team's presidential endorsement last year. Bezos is really pushing the editorial team into his point of view, and the paper is becoming more and more just a mouthpiece for him. This AI thing is amusing, but it's not like it's a surprise.
The only good thing left there is Carolyn Hax.
Re: Continuing the speed-run towards being a junk- (Score:3)
And here's a recent example.
https://www.washingtonpost.com... [washingtonpost.com]
Uncritically celebrating this administration's reverse-racism judo move to call any attempt to measure unintentional discrimination, harmful discrimination in itself.
Which, if everything was in a vacuum, leads to interesting thought experiments, but in reality, the same administration is persistently and openly pursuing "unintentional" race or sex or religious discrimination, in many of its policies. From gerrymandering congressional districts, to
Re: (Score:3)
Now who can argue with that? - I think we're all indebted to Gabby Johnson for clearly stating what needed to be said. - I'm particularly glad that these lovely children were here today to hear that speech. - Not only was it authentic frontier gibberish, it expressed a courage little seen in this day and age!
Re: (Score:2)
Was that from "Blazing Saddles"?
Re: (Score:2)
Nietzsche says: "out of chaos comes order."
The real news here (Score:5, Informative)
Re: (Score:3)
"... not living up to the paper's standards." doesn't necessarily mean surprise. You don't have to be surprised to still be appalled.
Re:The real news here (Score:4, Interesting)
They were probably most concerned with ensuring that it wouldn't question the primacy of supporting free markets:
https://www.pbs.org/newshour/p... [pbs.org]
Re: (Score:2)
The real surprise here is that the people who run WaPo think the AI-generated posts are different than their regular WaPo-generated posts...
Actually, they knew that all along. They just had to issue disclaimers to redirect the prols to the preexisting sources of misinformation. And to run interference, adding an apparent air of legitimacy to the human-generated content.
Instant and Permanent Loss of Credibility (Score:5, Insightful)
I can accept a little bit of bias in the press because it's truly difficult to be completely neutral.
But when you take a technology that is known to hallucinate and publish content with it... you're a cheap entertainment/fiction slop shop. Even opinions are supposed to be based on facts, so any argument that "it's just podcasts" is absurd and wrong.
I don't know if anyone had much respect for the Washington Post lately, but it should be gone now. If they had tested it before going live, they would have known. But they didn't because they don't care about anything besides money.
Re: (Score:1)
I can accept a little bit of bias in the press because it's truly difficult to be completely neutral.
They taught us in school that everyone has a bias. The quality of journalism these days has deteriorated, with opinions leaking into what should be just facts. A lot of conjecture, a lot of "this is what this means"... Rush Limbaugh used to say it was because journalists want to change the world, they don't want to just 'report the news.' Journalism is just a vehicle for change, in their minds.
Don't worry (Score:4, Insightful)
Exactly as many people are listening to them as recording them.
Re:Don't worry (Score:4, Funny)
People get their AI to listen to the podcast and summarize it for them.
those people (Score:2, Informative)
Jeff didnt buy WP to improve journalism, he bought it so a certain shithole 3rd world middle eastern country can "control the narrative", why do you think that USA's media landscape is all run by a particular tiny (relative to population size) ethnicity ?
USA have exactly none that don't contain at least 1 sympathiser in the C-Suite, NYT, WP, CBS, Paramount, Disney, CBS, CNN, literally all of them, in fact Benny is still pissed that Elons X isnt "his people", he managed to get Larry to steal tiktok so that p
Why alarmed only on AI? (Score:1, Troll)
Re:Why alarmed only on AI? (Score:4, Informative)
If you don't like attacks, then you must hate Donald Trump, who continually attacks others with his nasty little nicknames he comes up with and shows the worst of humanity at every chance he gets.
Re: (Score:2)
How it probably went down (Score:5, Insightful)
PHB1: "We have to do something AI-ish, everyone else is!"
PHB2: "Here's one, have bots compile podcasts from our news articles."
PHB1: "Brilliant! Make it so."
[months later]
PHB2: "Um, the podcast bot has been making silly errors. Should we keep it?"
PHB1: "How is our competition doing with their AI?"
PHB2: "They suck also."
PHB1: "Okay, let's keep it so we can have AI on our brochures and resumes."
Only Frustrating with Unrealistic Expectations (Score:4, Insightful)
If they were to go into it as an experiment saying, "Hey! We're going to see how good it is and tell you where it failed," while jumping in and fact-checking it throughout the podcast, they'd have a winner.
Unfortunately, they were so poorly informed (maybe THEY don't read the news?), that they thought LLMs were actually artificially intelligent digital beings, capable of human-level thinking and retrospection, and now they feed disappointed and frustrated.
That's like saying, "I'm frustrated that my Tesla can't safely drive me everywhere I want to go without my interaction."
Can't be that bad if they didn't turn it off. (Score:2)
Cool! (Score:2, Troll)
This news from WaPo comes on the heels of Pete Kegsbreath announcing widespread deployment of AI in the US military. It's very reassuring to know that the Department of War is adopting the most advanced and most reliable fiction generator available to the entire chain of command throughout the Services.
OTOH, the Department of War's shiny new AI says that the recent attacks on fishermen off the coast of Venezuela are https://www.msn.com/en-us/news... [slashdot.org]">"unambiguously illegal", so maybe there's something p
Hardly surprising (Score:3, Informative)
But... (Score:2)
But are their AI podcasts any MORE rife with errors than their other coverage?
How is this remotely surprising? (Score:1)
It Didn't Take Long For AI To Go to Hell (Score:2)
Re: (Score:3)
no Author and no Publisher. There is no copyright, no index, no Table of Contents, no page numbers, no dedication and no identification of any kind
...
There is no logic, no story, no points being made nor arranged in any sensible order.
You bought The Bible?
Humans lie so much in 2025... (Score:2)