eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 17:54:27
So recently OpenAI has been stomping around my (silly, personal, non-commercial) website using GBs of bandwidth in a rather rude, arrogant & brutish manner.
We were going to block it, but Mr Tech pointed out that it had got caught in the Riddle-o-Matic, some silliness that we made about 25 years ago (for reasons unremembered) & it is consuming never-ending nonsense.
"It's basically ChatGPT poison" he said. "Shall we leave it?"
"Aye, may as well"
Bloody cheek though.
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 17:58:55
@FediThing I kind of like that I was DECADES ahead of AI when it comes to spouting rubbish on the internet ๐
Jargoggles (@jargoggles@kolektiva.social)
Föderation EN Do 12.09.2024 19:50:59
@eclectech @FediThing
Yes, but LLMs do it the Max Power way - the wrong way, but *faster.*
Föderation EN Fr 13.09.2024 00:49:03
@jargoggles @eclectech @FediThing When I read Max Powers, I first thought you meant โusing all the electronsโ, not a Simpsons reference. Maybe you meant both?
Jargoggles (@jargoggles@kolektiva.social)
Föderation EN Fr 13.09.2024 02:45:24
@birwin @eclectech @FediThing
Definitely meant it as a Simpson's reference.
Bill's in the shop for repairs (@wcbdata@vis.social)
Föderation EN Do 12.09.2024 17:57:31
@eclectech Brilliant. We should all have such a feature!
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 18:00:09
@wcbdata Honestly, the bandwidth it's used is EXTRAORDINARY & I was so annoyed to start with, then I saw all the lines in the log file ๐
Andy Hort (@Devonkiwi@mastodonapp.uk)
Föderation EN Do 12.09.2024 18:09:35
@eclectech This is wonderful!
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 18:12:54
@Devonkiwi It is quite pleasing. GBs and GBs of absolute bobbins!
CynthesisToday (@CynthesisToday@sfba.social)
Föderation EN Do 12.09.2024 18:09:54
Thanks for sharing proof of concept with your Riddle-o-Matic
I was wondering if there is a specific type of content that could completely poison the generative AI algorithm... something like people are beginning to find with clothing fabric designs and fooling facial recognition sw.
If the AI algorithm has optimazations in maths, it probably has de-optimizations in maths, too, beyond just eating its own generated crap.
Wondering, rhetorically, of course, could Riddle-o-Matic be that krytonite? Everyone just sticks a block of "Riddle-o-Matic" on their site in the most "eat me first" way and, ka-blooey, done
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 18:17:43
@CynthesisToday Hah, Riddle-o-matics for ALL! It has finally found its purpose.
In reality I suspect the sites full of AI generated rubbish will eventually do the job though.
sortius (@sortius@mastodon.social)
Föderation EN Fr 13.09.2024 00:33:34
@eclectech @CynthesisToday oh, they already are. I recall reading an article a week or two ago about how all the poor people in the global south being paid to "train" AI are using AI to train AI, because they're not stupid. It's hilarious, and doing similar to your Riddle-o-matic, but in a more creepy, exploitative, way ๐ฌ
eclectech (@eclectech@things.uk)
Föderation EN Fr 13.09.2024 01:08:02
@sortius @CynthesisToday Argh, FFS ๐คฆ๐ปโโ๏ธ
@now@n (@DoubleArobase@toot.aquilenet.fr)
Föderation EN Do 12.09.2024 18:10:42
@eclectech Wow, congrats for the poison, thank your website for its service!
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 18:14:22
@DoubleArobase It's nice to know that decades later the silliness finally has a use!
D. B. Stuck (@MyWoolyMastadon@toot.community)
Föderation EN Do 12.09.2024 18:22:03
Curious as to what the Riddle-O-Matic originally was used for. Silliness for visitors that created a never ending stream of content, or one of those traps to keep unwanted web bots busy for hours as a big FU to that sort of thing?
Stupid me. You had a link to it. ๐
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 18:28:53
@MyWoolyMastadon It was so long ago the specifics are lost in the midst of time, but we made a LOT of nonsense online playthings & I think it was mainly idle curiosity about how best to do it & whether any of them would 'work' in a human sense (answer - sometimes, in part).
We used to play with getting top spots on Google back then too, did a few newsworthy Googlebombs, & I think we had half a mind that it would be 'sticky' for anything automated.
hambach18 (@hambach18@kanoa.de)
Föderation EN Do 12.09.2024 21:08:32
@MyWoolyMastadon
I had a map on my website but instead of using javascript vor moving the map it can be moved and zoomed by clicking on links. The server script then generates a new site and calculates new links for navigation.
Because zoom is limited to 19 it can generate 1+4^18 sites.
https://trafficpixel.tk/kamera/viewmapOverlay.php I also had another map and a table with "show next page", both without a set limit.
Three bots scraped the maps pre-2020 and I could stop them with https://trafficpixel.tk/robots.txt
Mister Eel (@Mister_Eel@mstdn.social)
Föderation EN Do 12.09.2024 19:25:58
@eclectech I feel like @viticci would enjoy this!
LibertyForward1 (@LibertyForward1@beige.party)
Föderation EN Do 12.09.2024 19:35:49
@eclectech https://eclectech.co.uk/words?word=Zktgm&seed=910329
"My first is in hepatitis but not in pettish
My second is in agreement but not in magnate
My third is in salacity but not in catalytic
My fourth is in poesy but not in soppy"
*chefs_kiss.gif*
charvaka (@charvaka@hivemind.plus)
Föderation EN Do 12.09.2024 21:02:42
I love Riddle-o-matic!
My first is in neocolonialism but not in comeliness
My second is in migraine but not in imagine
My third is in boost but not in boot
My fourth is in naked but not in dank
undead enby of the apocalypse (@enby_of_the_apocalypse@kolektiva.social)
Föderation EN Do 12.09.2024 21:52:26
@charvaka @LibertyForward1 @eclectech I think I got that exact same oneโฆ huh
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 22:07:27
@enby_of_the_apocalypse @LibertyForward1 @charvaka yep, the links with the code and seed on the end take you to a particular one, then you can regen on the page or generate one for a different word
Ken Butler (@nxskok@mastodon.cloud)
Föderation EN Fr 13.09.2024 04:45:48
@eclectech @enby_of_the_apocalypse @LibertyForward1 @charvaka load of arse (based on n = 3).
Frank Hightower (@FrankHghTwr@meow.social)
Föderation EN Fr 13.09.2024 01:14:27
@charvaka @LibertyForward1 @eclectech
Aww, a cute little riddle, let's try to solve it!
...blimey you got me good
gavinisdie (@gavinisdie@masto.ai)
Föderation EN Do 12.09.2024 20:08:38
@eclectech Modern Problems require Ancient Solutions I guess
Föderation EN Do 12.09.2024 21:02:37
@eclectech dang now I need to put up a page that just randomly throws Markov chain generated nonsense interspersed with links and let the automata poo on each other :D
David Michaels (@vwampage@xoxo.zone)
Föderation EN Do 12.09.2024 21:04:09
@eclectech this is positively delightful!
charvaka (@charvaka@hivemind.plus)
Föderation EN Do 12.09.2024 21:12:21
@eclectech This inspires me to run something like this on my silly personal website. Any chance you open sourced it?
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 21:21:51
@charvaka Heh, it's not (being very old nonsense that was of no practical use to anyone) but we have no problem sharing it.
It was originally php but had a quick and dirty Golang makeover when the site was updated a couple of years back.
I am told by Mr Tech it was "very quick and dirty" ๐
Timothy Wolodzko (@tymwol@hachyderm.io)
Föderation EN Do 12.09.2024 21:39:13
@eclectech we need more โAIโ honey traps like this ๐ค
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 21:46:06
@tymwol Time to fight back with silliness!
Avarna ๐ซ (@avarna@mastodon.social)
Föderation EN Do 12.09.2024 21:54:52
@eclectech reminds me of the Postmodernism Generator: http://www.elsewhere.org/pomo
Jonathan T (@JonnyT@mastodon.me.uk)
Föderation EN Do 12.09.2024 22:12:37
@eclectech Maybe you can successfully trap all the other AI scrapers on your site so that the rest of the internet remains free from their tyranny.
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 22:19:49
@JonnyT Hah. That sounds like such a worthwhile project.
Jonathan T (@JonnyT@mastodon.me.uk)
Föderation EN Do 12.09.2024 22:23:39
@eclectech Is there a cheap and effective way to publish billions of websites that are just full of Lorem ipsum gibberish that only scraper bots could and would discover?
eclectech (@eclectech@things.uk)
Föderation EN Do 12.09.2024 22:36:12
@JonnyT I feel like thereโs a very rude comment about the current state of the internet on the tip of my tongue here. I like the idea of all โproperโ sites having a little bit that the bots get trapped in though.
The Tired Horizon (@tiredhorizon@mstdn.social)
Föderation EN Do 12.09.2024 22:41:10
@eclectech I often go into an image AI tool and feed it bullshit.
Cluster Fcku (@clusterfcku@mastodon.social)
Föderation EN Do 12.09.2024 22:43:17
@eclectech what an arse
Oskar im Keller (@OskarImKeller@fnordon.de)
Föderation EN Do 12.09.2024 23:01:53
@eclectech ๐ @baldur you might enjoy this (if you haven't already seen it anyway).
Brendan (he/him) (@Brendan@phpc.social)
Föderation EN Fr 13.09.2024 01:28:02
This absolutely made my day.
Föderation EN Fr 13.09.2024 01:35:36
@eclectech if you don't want to block it, you could instead rate limit it.
eclectech (@eclectech@things.uk)
Föderation EN Fr 13.09.2024 12:50:14
@dev_ric Ta. Just keeping an eye on it at the moment, but that's definitely an option.
rateexportpilot (@rateexportpilot@infosec.exchange)
Föderation EN Fr 13.09.2024 03:37:26
@eclectech that's fantastic
Bruno Nicoletti (@bjn@mstdn.social)
Föderation EN Fr 13.09.2024 09:27:29
@eclectech @mawhrin I was musing on writing a scraper dungeon that did exactly this. Dynamically generate an infinite number of linked web pages filled with automagically generated content. Spot when something gets trapped wandering around in there and block or rate limit the IP addresses doing that.โNiceโ to see the idea in action, sorry about that them eating all your bandwidth.
Christ van Willegen (@cvwillegen@mastodon.nl)
Föderation EN Fr 13.09.2024 12:35:54
@eclectech
Couldn't be arsed to look at more than three...
@Binder
Cyber Yuki (@yuki2501@hackers.town)
Föderation EN Fr 13.09.2024 14:41:26
@eclectech This feels like the ending scene of War of the Worlds, but for AI.
I love it.
Lydia "Trivia" (@lydiafacts@strangeobject.space)
Föderation EN Fr 13.09.2024 16:26:58
@eclectech be careful, you might finally teach the AI to spell!
Charnock (@Printdevil@dice.camp)
Föderation EN Fr 13.09.2024 17:54:09
@eclectech What I find egregious is the way they ignore the robots.txt (etc) no matter how you phrase anything.