66
votes
Boring Report: An app that aims to remove sensationalism from the news and make it boring to read, by utilizing the power of advanced AI language models
Link information
This data is scraped automatically and may be incorrect.
- Title
- Boring Report
I’ve been really enjoying news minimalist. It uses GPT to decide how significant the day’s news is, and then only sends you the ones that reach a certain threshold. Sometimes nothing significant happens and there are no articles. It’s great.
Gosh, this looks great! Thanks for sharing; I'm glad it has an RSS feed too :)
This is awesome. Thank you very much for this link.
I found this to be an interesting usecase of LLM's!
I guess it suffers from its own success that I barely open the app because it's not serving me short bursts of dopamine?
Has anyone else come across similar uses of LLM's?
I'm not sure if it's a thing nowadays, but there used to be a pretty big business selling edited versions of mainstream movies to religious extremists. Cut out the nudity, the swearing, sell a 40-minute-long incomprehensible VHS of Basic Instinct for double the price, and call it a day.
Imagine selling access to a machine instead, which has been trained to monitor your TV in real time and handle any objectionable content. It doesn't even have to cut it entirely. It can transform it in a way that maintains the integrity of the story.
The tech isn't there yet obviously, but it doesn't feel that far out of reach. And there's a market for it. Not a small one.
Now imagine a Chinese internet censored not through brute force, but through the subtle work of machines like this. No need to block speech when you can alter it to support you.
Social hackers finding clever ways to evade the censor-bots, pushing back with viral videos that are quickly tagged and transformed into pro-government messaging.
And a flood of users worldwide communicating through automated translation software, altered in ways that seem natural and genuine to either side. The incredible potential for political interference. All happening at a scale too massive to ever reasonably moderate.
The information wars have barely begun. You subscribe to a personal filter-bot that's advertised to spot false and altered data. You want to protect yourself. It's supposed to be trustworthy. But if the company controlling the service is bought by Coca-Cola, how would you ever know? They're now in a perfect position to control your ability to find out.
You buy a Coke (someone on Tildes mentioned it, and it just sounded good) and sigh in contentment. Everyone on the internet agrees with all of your fundamental political beliefs, and the world finally makes sense. Only one thing is bothering you... You already finished that drink. You need another. It will all be okay again as soon as that crisp flavor hits your lips.
...so I guess running all my news through a machine kinda fundamentally worries me... is what I'm trying to say. The world is only gonna get weirder from here.
I understand where you are coming from on this. The technology may not be there to alter reality, but machine learning has made text alteration easier to do at a large scale.
On the other hand, we cannot pretend that this is a train that can't stop. Reducing that dopamine rush from sensational news is a net positive in a world that is constantly trying to grab your attention.
We do probably need more tech literate people in charge of making appropriate policies for limitations and uses of this technology to head off the abuse of this technology (not Rockstar ceos).
I hate that I can see this coming to fruition.
This is part of my experience using tildes, and also something that reinforces that this is a good thing. The more dopamine-driving apps we can replace with these equivalents, the better off we'll all be.
Someone told me about a project that will use LLMs to dumb down official letters. Think of letters send by the government, that a large part of the population will have a hard time understanding. You would take a picture of the letter and the app would give you a simplified, easier to understand version in return.
Saw this on HN a while back and thought it was great. A good example is this headline: “ Submersible Incident Results in Casualties” which (granted its after the fact) summarises a pretty big media frenzy around it.
A good example of using LLMs, I wonder if there’s a way to make this on a spectrum of boring. And kinda dial up or down the ‘heat’ - perhaps based on engagement / usage - either to stop me spending time on it or to make me read more articles.
I find this language objectionable. "Casualties" is a military euphemism to remove the tragedy of death from language, dehumanizing victims in order to diminish their loss.
And no, I'm not being specific to the case at hand. I'm talking about how "casualties" is generally used.
I'm like two or three days late here, but it seems like employing vaguely disgusting euphemisms is all this app knows how to do. Like,
It's not like I'm pro-mass shooter, but using "Neutralize" to describe a cop shooting and killing someone is inherently suspicious behavior to me. That's the language cops use when they kill innocent people, too! It also just fails to communicate anything. You look at this headline and think to yourself that nobody except cops would ever use the word "neutralize" instead of "kill" unless the shooter survived, so why would he be indicted? But no, he was killed, the LLM just decided that was sensational language. The shooter being killed is the entire reason the grand jury the story is about is happening. The killing is essential to the story. Yet it's too sensational?
I find myself deeply skeptical and wondering how this tool will talk about garden-variety police shootings rather than the irregular occasions when they stop mass shooters. Fortunately I don't have to scroll long to find out.
Conjure in your mind what this might possibly be trying to describe. For me, I'm thinking maybe excessive force, maybe an illegal hold or something. What actually happened was that the police broke into a home without a warrant, then tortured the two men inside with sex toys and tased them literally dozens of times over the course of 90 minutes, then put a gun in one of the men's mouths and fired. Is that sensational? Yeah, I guess I would say those facts are pretty sensational.
But the article title the LLM has butchered here is "Mississippi sheriff fires deputies accused of sexually assaulting Black men and shooting one in mouth". That communicates what happened in a pretty neutral way. Conversely, "Alleged Assault Incident" is a laughably bad way to describe what happened.
Does this app make the news boring? Yes, but it does it by taking literally all of the facts out of the news. And why is "making the news boring" a desirable goal? What is supposed to be the purpose of this? The website's about page says:
But this tool gets rid of both the catchy headline and the actual facts and relevant information.
But also "Three Elderly Individuals Found Deceased in Newton Residence", which should have been "Three Elderly Individuals Found Murdered in Newton Residence". It's extremely impressive how good of a job it's doing, but sometimes it makes things too boring.
Both of these cases make me wonder if LLMs are just trained to not use dead/murdered/imploded/etc...
Those words do tend to invoke feelings and the point is to not feel anything when reading these article summaries. That may or may not be a good thing, but it is an interesting thought.
Seems silly to make it an app. Probably would check it out if it was a site
It is a site too! You have to click through. Here's the direct link: https://www.boringreport.org/app
I agree, I'm not going to download an app just to view some AI text.
Whenever a simple thing like this gets shoehorned into an app and only an app, I become suspicious that the primary purpose of the app is to be an ad delivery mechanism or a data collection engine.
Sometimes that is probably inaccurate, but whatever.
Exactly
I have found that https://ground.news/ does something similar with the summarization of many different sources through ChatGPT.
That looks really interesting with the bias breakdown, although it seems to be extremely US-centric in what news sources it grabs. Still, really cool site.
I believe it's based in Canada actually, but yes a lot of the sources are from the US. I've been using Ground News for my main news source for about a year and a half and I love it. I don't have premium though I've considered it a few times. It's great to be able to read the different slants the opposing sides focus on in a particular story, or to look for a wire source on a particular event.
Edit: I just checked on the app and there are different regional versions you can select: US, International, UK, Canada, and Europe
Oh damn, I must have totally missed that. I'll check what the sources are it cites from then. TY!
Thanks, very interesting!
This seems quite helpful. Right now, I use the Wikipedia Current Events Page.
It does a pretty good job and I have it set as my browser's homepage.
+1 for this! I find it good for covering non-US news stories. :>
Also, even though Wikipedia doesn't natively have an RSS feed, I found an unofficial feed here: https://www.to-rss.xyz/ (specifically https://www.to-rss.xyz/wikipedia/current_events/)
I like that the feed updates only once a day, so it often feels more like a newspaper than something I obsessively check throughout the day.
This is excellent, thank you. Just popped it into my reader :)
Personally I'm enjoying Artifact for a similar purpose. It leaves titles as they are unless community deems them to be clickbait. You can mark titles as clickbait too which changes them before your eyes which is curiously satisfying to watch.
I would really like for RSS readers to implement this functionality because I feel I don't get enough control over what news I see though.
allsides.com is a good one that shows the bias and gives you the opposite leaning news site too.
Journalists are ENRAGED by this new app, click to find out why!
Looks interesting, though I wish they'd make it a bit easier to see what the original headline was. That extra click to go to the original article just takes alot of work, but it'd also be nicer to see them side-by-side for comparison.
This sounds like it would be a ripe target for prompt injection attacks.