X's AI Bot Is Reporting Joke Posts as Actual News | Hacker News Reader

X's AI Bot Is Reporting Joke Posts as Actual News

lifehacker.com

24 pointsmarban15 comments

Well, it's doing a good job of supporting the Russian effort to generate so much bogus info that any real newsgathering becomes meaningless.

Indeed, it's also supporting the (electoral) strategy of sowing distrust in every news media, so everyone is free to build their own comfort-zone away from facts that don't fit / they don't like...

Which is exactly the Russian strategy.

Features like this desperately need to be an AI-curated but human-driven operation. i.e an AI can aggregate some headlines from the latest news but a human needs to be the one that says "good to go" or "pass" on each. It's not shiny, but it would work (presumably better).

This just proves (to me) for the umpteenth time that AI can help us with our work immensely but company after company refuses to hire people who work well with it and instead focus on trying to employ it in a myriad of different unsuitable functions.

It makes no sense to summarize the news from the content of tweets that are about the news, no matter how it is done. You cannot produce a correct output from the wrong input: garbage in, garbage out.

AI summarizing the correct information could work with minimal human supervision.

Fully agree. The idea here doesn't seem to be to create quality content, but to satisfy people's itch to be informed about the world without actually paying a reputable source.

Next: Create real-time reports of sports events by just combining "hearsay" from random people on Twitter into ever-evolving "factual" news. Maybe with a user-option to select which team you're rooting for, so the hodgepodge of hearsay is automatically curated to create a favorable outcome for you personally...

News aggregators already exist. What value add from AI are you thinking of? As a user, I don't care about getting news 30 minutes faster. I also don't care if news aggregators miss out on some stories that AI would catch. I already don't use news aggregators because it's too much news.

The problem that needs to be solved is noise reduction. I want more thoughtful and original content. I want to slow down my media consumption. If someone could build a media aggregation/filtering product that serves me instead of advertisers, I'd use it, but AI wouldn't be the point of that product, just a means to an end.

>News aggregators already exist. What value add from AI are you thinking of?

Not parent commenter, but: Lower payroll figure.

Since when are those models (any of them) able to understand what they analyse/write? Isn't this completely expected behaviour?

If you asked me to summarize the news stories using nothing but top tweets that are about the news stories, I would also come up with crap about cops being deployed against earthquakes. What else could I do, given the requirements of the task? I could refuse to do it, insisting that you find someone else to do this stupid thing. I believe AI can be configured to similarly refuse.

I think you probably could build this if you had it corroborate stories with legitimate news outlets, eg chatgpt can search and summarise websites and could probably tell you if it also reports X or not. But if you want to just take Twitter posts from everyone as the only source and produce something useful then I doubt it’s possible with any current model.

It is; this is just yet another case of someone using an LLM for something for which it is unfit.

The "50,000 police ordered to shoot the earthquake" story is just the tip of the iceberg. The Grok bot also falsely reported that "Iran strikes Tel Aviv with heavy missiles" based on a network of blue-check accounts posting misinformation.

https://mashable.com/article/elon-musk-x-twitter-ai-chatbot-...

Which is ultimately more dangerous, because somewhat believable.

>no reasonable person would ever believe that Grok's summary of the news was accurate.

It's accurate in the sense that journalism these days is bullshit, specifically sensationalist bullshit manipulating fear and anger for quick and easy(?) monies. A cancer that is infesting society.

Unless someone is just quoting something verbatim from an actually credible source (eg: USGS, for earthquakes) and not an inch more, it's bullshit.

Reply on news.ycombinator.com