#Is there any way to prevent LLMs from destroying all online discussion?

1 messages · Page 1 of 1 (latest)

icy igloo
#

Basically every text-based medium of communication is being shitted up with LLM slop right now. Ebooks, forums, StackOverflow answers, etc. What's the solution to this? I see a couple things that might end up happening here:

  1. Major sites like Facebook, Twitter, Reddit, etc implement strong ID verification before allowing people to post. I really hate this idea for a lot of reasons, mostly to do with privacy and the chilling effect that this'll definitely have on any kind of discussion that you wouldn't want to attach your real name and face to.
  2. The major sites could fracture back into an internet that's more reminiscent of what we had in the early 2000s, where small and heavily curated spaces are the dominant way that people communicate online. Personally I wouldn't mind this, but I think that people are used to having a constant flow of a massive amount of content and a lot of people won't be willing to give that up, even if it improves the quality of discussion significantly.
  3. We somehow conquer the LLM tide, either through watermarking LLM output text, making laws that punish the use of non-human-generated content being passed off as human-generated, something like that. I don't see this being realistic at all, but I guess it's technically possible.

Is all online discussion just doomed to be full of LLM slop? None of the solutions I can think of seem both realistic and good at the same time.

undone cedar
#

I think that's a very valid concern. I found that in the AI forums already it's 99% bots and the web in general is becoming progressively more AI generated. The ultimate form of content generation would be to kill search entirely and simply have your search engine of choice generate the results for you based on what it was trained on. However, this also implies that the content has been indexed from somewhere. So if we continue on this trend we're going to have AI being trained on AI generated text which cannot be creative as you know.

Indeed I do imagine there will be some sort of fracture somewhere, but it's not going to be beneficial to us users because I can already imagine the marketing packages where you can purchase access to Google AI search for $10 a month, WhatsApp for another $5, and the boring old web for an extra $50. And it's going to be censored to bits of course through packet filtering or something similar.

Sadly, the early days of the Internet are long over and I am very concerned about this development.

icy igloo
# undone cedar I think that's a very valid concern. I found that in the AI forums already it's ...

yeah, I wouldn't mind so much if most internet users got replaced with AIs, if they weren't so censored. and I don't even just mean like, it won't say the N word or whatever, more the way that a friendly AI can insidiously push you towards particular views that aren't especially controversial.
so what do you think, we'll get a fractured internet or will everyone be more or less happy with the AI-slop they're fed? not to denigrate users of other sites but like, I can't see YouTube commenters being especially critical of whatever is being shoved down their throat by the website as long as they get the content they're addicted to.

undone cedar
#

Yes, I think you're right. There's another thing that could happen but I think it's going to get regulated real quick: uncensored AIs trained on material that others won't there to train their stuff on. The costs are astronomical but maybe that's only for now. But I think regulators will come after them.

unreal mist
#

or just make www2.0

icy igloo
#

you're going to solve 400 captchas per day and the people who make spambots will just pay third world workers pennies to solve the captchas for them so they can keep botting

inner pier
#

no

#

my SchizoGPT will replace every one of you

robust slate
#

I, for one, welcome our AI overlords.

warm mantle
#

There are models, such as the open-sourced YOLOv8, that can be trained to solve captchas. Captchas just get harder (see hCaptcha), and at some point, they will need to be modified, as even humans will not be able to solve them.