←back to thread

757 points headalgorithm | 2 comments | | HN request time: 0s | source
Show context
yakhinvadim ◴[] No.42951867[source]
I tried to solve this problem by making AI rank the stories by significance and rewriting the news titles in a boring, factual style.

I think it worked quite well, there's only about 10 headlines a day (out of 15k+) that get a significance rating higher than of 5.5 out of 10.

It also helps avoiding the overfocus on western issues and actually learn what's happening around the world.

https://www.newsminimalist.com/

replies(6): >>42952431 #>>42952930 #>>42953033 #>>42956978 #>>42958943 #>>42960245 #
1. j_bum ◴[] No.42952431[source]
I love the idea of this tool, but there are serious issues with using LLMs to summarize articles and text. Re: Apple’s Notification Summary Debacle

For example, this headline with a score > 5 is flatly incorrect.

“China launches innovative flying robot to explore Moon's south pole for water resources”

Every article listed in the summary says the launch is planned for 2026.

replies(1): >>42952620 #
2. yakhinvadim ◴[] No.42952620[source]
Thanks! Good point.

I think there will always be some hallucinations until they're solved on a model level, but I'll also try to nudge AI now to be more precise with the headlines.