you are viewing a single comment's thread
view the rest of the comments
[–] 31 points 16 hours ago (2 children)

AI generated news articles have become as terrible as they are ubiquitous. my work PC is pretty locked down and vanilla AF. sometimes i look at it to see what's up through that lens.

anyway, the irony (or absolute fucking inevitability) with the "news" is this narrative from like the 1990s about how journalism was too expensive to continue on as it had been, with "The Internet" cutting into ad revenues. so we've seen countless mergers and consolidations and closings over the last decades, until all that remains is a skeleton crew labor force of Zero Journalists keeping these legacy media content machines farting out noises where headlines reference pictures not contained or even linked in the article. just some generic slop layout template of [3 sentences], [picture], [5 sentences], [picture]... into eternity and 80% of the time, theres not even a common thread to follow through any of it. the old physical papers had constraints on spacing and layouts, but it's all become untethered as the incentive is to keep a story's copy going longer and longer to maintain the screen.

like, journalism under capitalism has always been a shit show, but now that they've gutted it of humans, it's not even baseline comprehensible. like eventually people won't even have a reason to look anymore, because it'll be incoherent symbols in no discernable order poorly representing a false reality with comments full of bots arguing passed each other.

and this will be the content that LLMs scrape to provide answers to people.

  • source
  • hideshow 2 child comments
  • [–] 2 points 5 hours ago

    Something I like to tell people is that we've been reading AI generated news articles as far back as 2013. I used to work for a company that was a publications services company that produced content and media mostly for textbooks but we had our fingers in a lot of things. One of the big initiatives there was implementing something called Wordsmith which would take structured data like a financial reports, stock tickers, sports scores, and weather reports and turn it into a narrative article that we sold/licensed to places like yahoo news. Sometimes humans had to step in to correct some schema analysis but it mostly just worked in a pipeline and generated text automatically. Looking back and researching the company recently they were actually on the forefront of what ChatGPT got credit for in 2022 and I'm sure a lot of my work at that company went to training the current models.

  • source
  • parent