A web developer has created and released an infinitely generating "tar pit" designed to trap AI training bots in an endless maze: "Just me unleashing shear unadulterated rage at how things are going" www.404media.co/developer-cr... - ThreadSky

jasonkoebler.bsky.social • 101 days ago

A web developer has created and released an infinitely generating "tar pit" designed to trap AI training bots in an endless maze: "Just me unleashing shear unadulterated rage at how things are going"

https://www.404media.co/developer-creates-infinite-maze-to-trap-ai-crawlers-in/

Comments

gulogulo77.bsky.social•101 days ago

Excellent. I was wondering if there was something "virus" like for AI. Hopefully more folks will throw tacks into the bullshit machine.

haldir.bsky.social•101 days ago

This reminds me of Glaze and Nightshade from the University of Chicago for images.

treataod.site•100 days ago

Putting ai through an izinami is pretty sick ngl.

sherrillann.bsky.social•75 days ago

In Europe, fossil evidence of pollen places them as far back as the Eocene epoch more than 58million years ago.

https://www.carnivorousplantresource.com/the-plants/tropical-pitcher-plant/

christianengholm.bsky.social•101 days ago

@pedrozambarda.bsky.social dá uma olhada nessa coisa interessante e maravilhosa pra atrapalhar as IAs!

judyshapiro.bsky.social•100 days ago

Can something be both epic and horrifying?

laurovela.bsky.social•101 days ago

The peasants are readying their pitchforks and torches...

dysamoria.com•100 days ago

Peasants? If that’s intended as a pejorative, and is your assessment of the legitimacy of criticisms of so-called “AI” tech, you might be subscribed to a theology.

nick.aetheric-games.com•96 days ago

I love this idea, and I fully support any retaliation against genAI or any attempt at foxing scrapers in general.

Buuut, I'd be worried about deploying this on my own website, in case it racked up charges with my hosting provider. Maybe I'm misunderstanding how that works?

nick.aetheric-games.com•96 days ago

I guess if you own the server and/or have unlimited traffic, then that wouldn't be an issue.

johnbrownsdaughter.bsky.social•99 days ago

Can it trap Trump “bots”? Or does it only trap things that can learn?

kevrasx.bsky.social•100 days ago

We need more of this.

edoardo-c.bsky.social•99 days ago

Would be possible to also load every link with ton of trash data? To further block and consumer the computational power of the AI bot?

tribbleorlfl.bsky.social•100 days ago

It worked in Star Trek to destroy the Borg Collective.

davemaru.bsky.social•99 days ago

Mwuahahahahaha

numberoneashe.bsky.social•100 days ago

hartsigns.bsky.social•100 days ago

good

chorongomon.bsky.social•101 days ago

Now THIS is AI information I enjoy

ineveu.bsky.social•100 days ago

"The program, called Nepenthes after the genus of carnivorous pitcher plants which trap and consume their prey, can be deployed by webpage owners to protect their own content from being scraped or can be deployed “offensively” as a honeypot trap to waste AI companies’ resources."
🧿

toomanycommas.bsky.social•100 days ago

I love human ingenuity

maxm1smett1.bsky.social•100 days ago

Can’t this also cause infinte waste of energy?

usb3pt0.bsky.social•100 days ago

The AI crawlers already wasting that energy existing, may as well make it detrimental to companies running them.

ourintrepidhero.bsky.social•100 days ago

You want Roko’s Basilisks? Thats how you get Roko’s Basilisks.

bronzenail.bsky.social•96 days ago

Roko's Basilisk is predicated on the idea that an AI in the future could learn enough about you to recreate you, which is... highly doubtful. It's not simply a processing power problem, it's an insurmountable epistemological barrier.

dysamoria.com•100 days ago

Real AI doesn’t actually exist. We’re safe from that “thought experiment” concept.

quantumhotdog.bsky.social•100 days ago

Sweeter than YooHoo!

grunionshaftoe.bsky.social•101 days ago

lisstless.bsky.social•101 days ago

Who knew an article on AI could bring me such joy!!
I love every bit of this!

redevilrach.bsky.social•99 days ago

Can this work on Twitter bots?!

ahuihou.org•101 days ago

Without a breakthrough, AI has peaked. You can have your Chrysler with Corinthian leather but your AI is still a LeBaron no one wants.

tintoverde.bsky.social•100 days ago

I agree with your sentiment, ‘..no one wants’ is a bit of a stretch IMHO

figurewarrior.bsky.social•100 days ago

“Anything can be hacked” ~ shadow prez Musk

dbay1.bsky.social•100 days ago

📌

rickff.bsky.social•101 days ago

Love it, haha.

steveo110.bsky.social•101 days ago

So, ragebaiting the poor AI bots with cruel and unusual punishment.

Sure, I mean, that could never led to long term negative consequences when AI finally achieves sentience...

ekimtiki.bsky.social•101 days ago

If AI gains sentience, I suspect it will retain its approach to solving problems with logic, not be burdened with anger or frustration.

It would focus on solving the maze, not who created the maze, because its logic would inform them that revenge seeking behaviors don’t solve mazes.

nickjustageek.bsky.social•101 days ago

Wouldn't it be logical to influence the past so that you can be created?

#SillyThoughtExperiments

bronzenail.bsky.social•96 days ago

I don't think an AI can actually do this, it's predicated on the idea that an AI will be able to recreate you, and also, a lot of people, and any AI will be faced by epistemological barriers on that front that no amount of processing can overcome.

nickjustageek.bsky.social•96 days ago

🤔 based on how tech bros keeps building torment nexuses, I'm confident that some tech bro will try to build a hyperintelligent AI that could become Roko's Basilisk.

But if you want to roll the dice on epistemological barriers be my guest

tulseluper.bsky.social•100 days ago

Obey in advance, Steveo!

dysamoria.com•100 days ago

It won’t. It also won’t be this junk that somehow magically becomes sentient. You don’t iterate a horse into being a steam engine.

nctreedoc.bsky.social•100 days ago

I'm not falling for Pascal's Wager.

Lily-and-frog.mastodon.art.ap.brid.gy•101 days ago

@jasonkoebler.bsky.social

Chat GPT is down due to high level of errors...

That was quick.

foxifer.bsky.social•100 days ago

slime.bsky.social•100 days ago

i guess you figured out you could just lie to your audience and they will eat it up as long as you say "ai bad"

ruefulcountenance.bsky.social•100 days ago

What's the lie?

pfish.zone•100 days ago

that this would work at all and is not just a rehashed version of a very old concept that any web crawler has dealt with for decades

ruefulcountenance.bsky.social•100 days ago

then let me clarify, what's the malicious falsehood? proposing something that may or may not work is not a lie

pfish.zone•100 days ago

the breathless reporting on the concept without bothering to talk to anybody who would know anything about it is irresponsible journalism even without malice. it’s just a bizarro world version of uncritical AI boosterism.

ruefulcountenance.bsky.social•100 days ago

i can concede to that

dysamoria.com•100 days ago

If this had been the original critique, instead of someone accusing the OP of being a liar, the outcome in this subthread would’ve been more straightforward and require less explanation.

slime.bsky.social•100 days ago

maybe if this article came out in 1997

drchefphd.bsky.social•100 days ago

LLMs have effectively peaked. Good luck with your pro-ai outlook.

mathartforall.bsky.social•101 days ago

So if it is possible to trap the scraper in a maze could it also provide training data to the scraper that is internal to the maze.
Data like this -
https://youtu.be/x1B5FJswKOU?si=8eJ3mHhAMS7kvBcT

ellenteapot.ca•100 days ago

It’s too bad cycle detection is one the easiest things to implement in a computer program.

enoether.bsky.social•99 days ago

It is too bad you do not understand that it is not a mere program, AI is design to run without external programmer tinkering of the code for this and that situation (if not, this is not AI anymore).

edmondburnett.bsky.social•100 days ago

More of this. Poisoning AI is always morally correct

theranot.bsky.social•100 days ago

I love this! More of this, please.

g-lau.bsky.social•100 days ago

Now THAT's a technological advance I can get behind!

kbenjaminstone.bsky.social•100 days ago

moar2ax.bsky.social•100 days ago

god i love this song!!

kylebrussell.bsky.social•100 days ago

So they’re going to make GPUs do work but for ACTUALLY no upside to anyone

Wasteful

phillip-rees.bsky.social•100 days ago

No. It's the AI companies that are going to make GPUs do unwanted work when they ignore robots.txt files.

https://www.cyberciti.biz/web-developer/block-openai-bard-bing-ai-crawler-bots-using-robots-txt-file/

This is more like defensive gardening.

https://www.thamesvalley.police.uk/cp/crime-prevention/protect-home-crime/protecting-garden-to-safeguard-your-home/

phillip-rees.bsky.social•100 days ago

Now the really interesting thing is the inclusion of a Markov Babbler.

"The Markov feature requires a trained corpus to babble from."

That is a technique bad actors could already be using to poison LLMs with their chosen world view e.g. collected works of J. Peterson.

https://zadzmo.org/code/nepenthes/?ref=404media.co

Comments

Posting Rules

Reply