My takeaway from this is:
- Get a bunch of AI-generated slop and put it in a bunch of individual
.htmfiles on my webserver. - When my bot user agent filter is invoked in Nginx, instead of returning
444and closing the connection, return a random.htmof AI-generated slop (instead of serving the real content) - Laugh as the LLMs eat their own shit
- ???
- Profit
QUICK
Someone create a github project that does this
I might just do this. It would be fun to write a quick python script to automate this so that it keeps going forever. Just have a link that regens junk then have it go to another junk html file forever more.
Also send this junk to Reddit comments to poison that data too because fuck Spez?
there’s a something that edits your comments after 2 weeks to random words like “sparkle blue fish to be redacted by redactior-program.com” or something
That’s a little different than what I mean.
I mean to run a single bot from a script which interacts a normal human amount during normal human times within a configurable time zone which is acting as a real person just to poison their dataset.
I mean you can just not use the platform…
Yes I’m already doing that.
This is a great idea, I might create a Laravel package to automatically do this.
- Get a bunch of AI-generated slop and put it in a bunch of individual
Good. Let the monster eat itself.
So kinda like the human centipede, but with LLMs? The LLMillipede? The AI Centipede? The Enshittipede?
Except it just goes in a circle.
))<>((
It’s the AI analogue of confirmation bias.
All according to keikaku.
[TL note: keikaku means plan]
No don’t listen to them!
Keikaku means cake! (Muffin to be precise, because we got the muffin button!)
Well then, here’s an idea for all those starving artists: start a business that makes AND sells human-made art/data to AI companies. Video yourself drawing the rare Pepe or Wojak from scratch as proof.
How many times is this same article going to be written? Model collapse from synthetic data is not a concern at any scale when human data is in the mix. We have entire series of models now trained with mostly synthetic data: https://huggingface.co/docs/transformers/main/model_doc/phi3. When using entirely unassisted outputs error accumulates with each generation but this isn’t a concern in any real scenarios.
As the number of articles about this exact subject increases, so does the likelihood of AI only being able to write about this very subject.
Anyone old enough to have played with a photocopier as a kid could have told you this was going to happen.
Blinks slowly
But, but, I have a photocopier now…
So then you know what happens when you make a copy of a copy of a copy and so on. Same thing with LLMs.
Imo this is not a bad thing.
All the big LLM players are staunchly against regulation; this is one of the outcomes of that. So, by all means, please continue building an ouroboros of nonsense. It’ll only make the regulations that eventually get applied to ML stricter and more incisive.
Inbreeding
Photocopy of a photocopy.
What are you doing step-AI?
Are you serious? Right in front of my local SLM?
I always thought this is why the Facebooks and Googles of the world are hoovering up the data now
Hahahahaha
AI doing to job of poisoning itself
They call this scenario the Habsburg Singularity
The best analogy I can think of:
Imagine you speak English, and your dropped off in the middle of the Siberian forest. No internet, old days. Nobody around you knows English. Nobody you can talk to knows English. English for all intents purposes only exists in your head.
How long do you think you could still speak English correctly? 10 years? 20 years? 30 years? Will your children be able to speak English? What about your grandchildren? At what point will your island of English diverge sufficiently from your home English that they’re unintelligible to each other.
I think this is the training problem in a nutshell.
AI centipede. Fucking fantastic.
