Live data from Hacker News

You Can't Opt-Out of A.I. Online

newyorker.com

11–20 of 105 posts

Re: You Can't Opt-Out of A.I. Online

#11

The outcome is that spaces which are prone to AI slop and spam will wither and lose real users, eventually collapsing on themselves. For the social media platforms, good riddance

Hard to make any forecast about "spaces" "prone to AI slop," whatever that means, besides social networks that you personally do not like, in aggregate or individually.

TikTok and Meta apps are definitely growing faster, in relative or absolute terms, than the New Yorker and the New York Times audiences are, paid or unpaid. If you compare aggregate social media to aggregate news and magazines, the latter is shrinking, if you exclude their presence on social media itself.

If anything we already live in a post-scarcity world with regards to engaging content, starting a few years before the advent of generative AI. Why would AI generated content reverse that trend? Isn't TikTok's feed algorithm agnostic to whether a video is AI slop or user created? Isn't Instagram's? Are you really going to play No True Scotsman with "spaces" "prone to AI slop?" Shouldn't Kyle Chayka, who supposedly wrote a book on this, know that? Aren't there already too many good books, movies, TV shows, video games, operas, plays, etc. to consume?

The toughest thing about this article is it is longing for a world that hasn't existed for a long time. If anything, the New Yorker and the New York Times, by doing a bad job at being media companies, have reduced the amount of new narrative creative projects that can thrive, not increased it. They never look in the mirror. The fickle and sometimes vindictive personalities that work there are not allied with narrative creators.

The idea that discovering one or two diamonds in the rough offsets the incumbent cultural trends the New Yorker reinforces has long been dead - there is just way too much new stuff for any traditional media company to accurately review, report on and amplify. Everyone thriving on YouTube, Instagram, Steam, TikTok, hawking their shit, figured that out.

It's even crazier to me that Kyle Chayka, who wrote a book on this, misses the mark here - I mean he should know about non-negative matrix factorization, which basically was the beginning of the end of traditional media, he should be able to make the leap that invention that enabled accurate collaborative filtering at scale killed The New Yorker, not AI, or anything in between 2000 and now. He should know there's absolutely no reason that AI generated content would be treated any differently than any other bad content, by NNMF or whatever feed algorithm.

There's a possibility that the reason the NYTimes and Conde Nast have to reinvent themselves is because they do a bad job. To them, "A.I." is just another effigy, when the reality is that fewer and fewer people care what important New Yorkers think. Listen guys, it's not looking good for the writers, better to get your head out of the sand.

Re: You Can't Opt-Out of A.I. Online

#12
I thought it was going to be about escape generative AI output online.

Personally, I run a large-ish niche FB group and I haven't seen a threat of bots taking over there or other online spaces. Is there anywhere that people have seen the bots crowding out the people? My guess is Twitter but that place would already/always a sewer.

Re: You Can't Opt-Out of A.I. Online

#13

The outcome is that spaces which are prone to AI slop and spam will wither and lose real users, eventually collapsing on themselves. For the social media platforms, good riddance

When I pop open facebook and see some page has somehow gotten onto my feed with an AI generated children's science fair project I always notice that it has 60k+ likes. To me it says that the stuff is indistinguishable from real for enough people that discerning users are just helplessly along for the ride on every platform.

The article purports that the reason slop is here to stay is because people like it. The next iteration, IMO, will be AI generated children's science fair projects that also have a pepsi logo in them, still with 60k likes.

Re: You Can't Opt-Out of A.I. Online

#14
post #8

This is just the practice of invasive data harvesting taken to its natural conclusion, which any tech savvy user or computer geek brought up in the pre-postmodern social media era of the Internet (before Facebook) could have seen coming decades in advance. The only winning move here is not to play, accepting the unfortunate consequence that you will need to "limit the reach of your profile just to avoid participating…

I'm not sure why having an LLM use my output in particular is problematic - I'm pretty sure OpenAI trained GPT-3+ on every single bit of data they could find, so they probably included a bunch of stuff I've written here already (though it is minuscule fraction of their vast corpus of course). Losing anonymity is something I'd be much more worried about overall (not that it's entirely separate I don't think it's the same).

Re: You Can't Opt-Out of A.I. Online

#15

The outcome is that spaces which are prone to AI slop and spam will wither and lose real users, eventually collapsing on themselves. For the social media platforms, good riddance

When I pop open facebook and see some page has somehow gotten onto my feed with an AI generated children's science fair project I always notice that it has 60k+ likes. To me it says that the stuff is indistinguishable from real for enough people that discerning users are just helplessly along for the ride on every platform. The article purports that the reason slop is here to stay is because people like it. The next…

And how many of those likes are AI-generated?

Re: You Can't Opt-Out of A.I. Online

#16

The outcome is that spaces which are prone to AI slop and spam will wither and lose real users, eventually collapsing on themselves. For the social media platforms, good riddance

Hard to make any forecast about "spaces" "prone to AI slop," whatever that means, besides social networks that you personally do not like, in aggregate or individually. TikTok and Meta apps are definitely growing faster, in relative or absolute terms, than the New Yorker and the New York Times audiences are, paid or unpaid. If you compare aggregate social media to aggregate news and magazines, the latter is shrinking…

> TikTok and Meta apps are definitely growing faster, in relative or absolute terms, than the New Yorker and the New York Times audiences are, paid or unpaid.

Imagine the doctor walking into the room and saying "Congratulations! The tumor is growing faster than your legacy cells."

Re: You Can't Opt-Out of A.I. Online

#17
post #10

The outcome is that spaces which are prone to AI slop and spam will wither and lose real users, eventually collapsing on themselves. For the social media platforms, good riddance

What I'd give to see the user retention stats at DeviantArt ever since they allowed it to be filled absolutely wall-to-wall with the most boring AI art you've ever seen. Not just allowed but encouraged by jumping on the bandwagon and integrating their own generator. Anecdotally it feels like there's not just an insane glut of AI slop drowning out everything else, but also less "everything else" over time due to user…

100%. It's intensely frustrating that they didn't at least firewall that vast amount of AI noise using a subdomain or something (e.g. ai.deviantart.com).

Re: You Can't Opt-Out of A.I. Online

#18
post #10

Earlier quoted context omitted.

What I'd give to see the user retention stats at DeviantArt ever since they allowed it to be filled absolutely wall-to-wall with the most boring AI art you've ever seen. Not just allowed but encouraged by jumping on the bandwagon and integrating their own generator. Anecdotally it feels like there's not just an insane glut of AI slop drowning out everything else, but also less "everything else" over time due to user…

100%. It's intensely frustrating that they didn't at least firewall that vast amount of AI noise using a subdomain or something (e.g. ai.deviantart.com).

They do have an AI tag that creators are supposed to set where applicable, and there's an option to hide everything with that tag, but the enforcement is weak so there's a lot of obvious AI spam which isn't tagged. Many of the AI spammers are using DAs monetization features or promoting a Patreon so they're incentivized to not set the AI tag to maximise their reach.

Re: You Can't Opt-Out of A.I. Online

#19

The outcome is that spaces which are prone to AI slop and spam will wither and lose real users, eventually collapsing on themselves. For the social media platforms, good riddance

When I pop open facebook and see some page has somehow gotten onto my feed with an AI generated children's science fair project I always notice that it has 60k+ likes. To me it says that the stuff is indistinguishable from real for enough people that discerning users are just helplessly along for the ride on every platform. The article purports that the reason slop is here to stay is because people like it. The next…

We need to move above that before it kills the web though - if the end state is that you want to look at any old random adorable puppy pictures so you just ask your machine to generate them for you then we might have a way out of this internet death cycle. I'd hope that if you can trivially produce your own better tailored slop on your local machine then the AI slop online will lose any value - I'm concerned that we're going to lose a lot of good content created by artists before we reach that level though.

I suspect the actual outcome will be a rise of manual curation and providence where a feed of adorable puppies or discussions on technology will rise or fall by how diligent the moderator is at keeping slop out of the feed.

Post reply on HN