Cloudflare turns AI against itself with endless maze of irrelevant facts
- Thread starter JournalBot
- Start date
Yes, you get it. Businesses should not be in the business of judging their customers, but businesses should provide the goods and/or services that customers pay to receive. Internet companies should be like McDonald's: Regardless of race, ethnicity, sex, gender identification, religion, political beliefs, what sports teams you support, etc., if you pay for a Bic Mac, you get a Big Mac.
I can certainly understand being against Neo-Nazis and other race supremacists, but once you deny one group something, it becomes easier to deny other groups something as well. For example, it can easily be imagined that Cloudflare could be put under pressure to censor any content related to DEI. Free expression should not be subject to whoever is in power or the current cultural zeitgeist.
Upvote
-2
(11
/
-13)
Ah, bad phrasing on my part, since English isn't my first language. Should have just said "business" instead of "private company". Sorry for the confusion.
Upvote
2
(3
/
-1)
And when the product is access to consumers' eyeballs?
It's no coincidence that fascism spreads easily through In social media, where the user is the product.
Upvote
12
(13
/
-1)
This is a take only a stright white Christian cis white male could have.
Many businesses value their employees. The idea of telling a black, or female, or gay employee that they must help / support a group calling for their marginalization or imprisonment is just abhorrent. Businesses can and should exercise editorial judgment on who they do business with.
You’re advocating for “just following orders” as not just acceptable but mandatory. No.
Upvote
12
(20
/
-8)
Nope it is not a slippery slope to deny service to nazis. There is literally nothing that forces you to deny services to other groups after you start doing that to nazis. Nazis shouldn't be welcome, and indeed many business refuse them services for very, very good reasons. The biggest one being fucking human decency, but also that they are subversive by nature and don't play by the rules, so accepting them is in fact the actual, real slippery slope.
Linking to the famous nazi bar Twitter thread screenshot: https://pbs.twimg.com/media/FKdZP_rXMAMczzq?format=jpg&name=medium
Linking to the famous nazi bar Twitter thread screenshot: https://pbs.twimg.com/media/FKdZP_rXMAMczzq?format=jpg&name=medium
Upvote
24
(24
/
0)
I don't think that it is fair to say that the web site is luring the crawler - the crawler found the site, the crawler disregarded the site's "no crawler's sign", the crawler spammed the site to suck it's content for the crawler's own profit with no remuneration to the site for their creations.
A predator lures its prey. This is just judo flipping the assailant.
Upvote
28
(28
/
0)
So how I'm seeing it is, we are now creating an alterweb adjacent to the one we use, designed only for bots. Looking at the sorry state of the current internet makes me wonder if this is also a fake internet, one designed for capitalistic purposes and manipulated by algorithms to a finite end. So then are we in a parallel internet already? And what is the real internet?
Upvote
1
(1
/
0)
Easy to circumvent by generating a webpage image - as a human would see the page - and only following human visible links.
Upvote
0
(0
/
0)
Got a link to a good how-to? I can serve 100G a month without worries!
Upvote
3
(3
/
0)
enlightened.doggo
Ars Scholae Palatinae
Like it or not, you're never going to be able to allow the google crawler access to your content while also blocking AI crawlers, simply because google is training AI as well. We're going to have to fight these plagiarism tools in court. If you are concerned with blocking access short-term, then my advice to you is to login gate content and aggressively IP ban.
Upvote
2
(4
/
-2)
So, question. These automated scanners are accessing remote servers to which they don't have permission, correct? Hence the whole point of robots.txt - which has always been an 'optional' file to let search engines and other scrapers know that hey, you don't want to be scanned.
If the bot still scans your site, aren't they now running afoul of the Computer Misuse Act of 1990 (UK) or the Computer Fraud and Abuse Act (USA)?
Example: how The Realm members Electron and Phoenix were taken down in 1990 for finding an open telnet connection, which was publicly available, at NASA?
Reference: https://en.wikipedia.org/wiki/In_the_Realm_of_the_Hackers
If the bot still scans your site, aren't they now running afoul of the Computer Misuse Act of 1990 (UK) or the Computer Fraud and Abuse Act (USA)?
Example: how The Realm members Electron and Phoenix were taken down in 1990 for finding an open telnet connection, which was publicly available, at NASA?
Reference: https://en.wikipedia.org/wiki/In_the_Realm_of_the_Hackers
Upvote
8
(8
/
0)
I don't know about that one, but there is a Chomskybot that creates long texts strongly resembling Chomsky's linguistic writing, but which are gibberish. Perhaps this could be modified to resemble most any writer, and used to lead AI dredgers (I use that word intentionally) astray.
Note 1: Before someone says Chomsky himself wrote gibberish, I'm a linguist, and have read much of Chomsky's writing on linguistics (not on politics). It is not always easy to follow, but it is definitely not gibberish.
Note 2: Chomsky recently suffered a stroke, and apparently can no longer communicate.
Upvote
10
(10
/
0)
florinandrei
Ars Praetorian
Yeah. Their defense against AI is another AI doing SaaS = Shitposting as a Service
Upvote
6
(7
/
-1)
Cloudflare has obviously never witnessed my father in law trying to surf the web...article said:
Upvote
2
(2
/
0)
The Geeman
Wise, Aged Ars Veteran
...
As we know half the AI output is tainted by shit, we also know the public at large will think it's the truth. We can ignore it and they won't know that they should.
It's an all or nothing situation in which I try not to come anywhere close to AI but that will get harder and harder to ensure.
And don't get me started on wholly AI generated and narrated YouTube videos, FFS I'll never get those minutes back!
Upvote
3
(3
/
0)
I did that. There was a site with a Moscow IP that was spamming one of my web forms with injection attacks. They were hitting me with about five per second. When I detected them, I wrote a quick filter that sent them the Monty Python Spam Song, 65863 verses of it. They quickly went away.
Upvote
7
(7
/
0)
Chromskybot has the capacity to produce irrelevant strings for about 10% of the time until the heat death of the universe. That should be sufficient to screw the AI developers.
Upvote
3
(3
/
0)
How long until a donor victimized investor whispers in Musk’s ear and an executive order is drafted for the Trump autosig bot to declare this activity an act of terrorism?
Upvote
5
(5
/
0)
Upvote
0
(0
/
0)
Cloudflare also block smaller browsers, like privacy specialed forks and the like.
Upvote
0
(0
/
0)
The google crawler respects the robot.txt files. This is against crawlers that ignore it. So not indexing context explicitly marked not to be indexed is fine...
Upvote
2
(2
/
0)
I’m sure google has a crawler that respects a boundary or two, that they can point to and say “See?”
I have no confidence they do not also have other crawlers.
Upvote
3
(4
/
-1)
AI is cool i guess
Wise, Aged Ars Veteran
the irony here is that we're building increasingly elaborate mazes that only humans will eventually get lost in. AI systems will adapt, they always do.
what happens when the complexity of these defensive systems exceeds what an average person can navigate? we could end up with a web where sophisticated AI can pass as human but many humans can't pass as "human enough."
it's an arms race, but with a twist: the bots will learn to solve the maze by simplifying their approach. sometimes lower "intelligence" is more effective than higher intelligence for specific tasks.
the entire history of captchas suggests this outcome. first we built tests only humans could pass, now we build tests that frustrate humans while bots breeze through.
(strange thought: maybe the ultimate defense isn't complexity but unpredictability; systems that make no sense at all but still function. humans excel at navigating absurdity in ways AI still struggles with.)
what happens when the complexity of these defensive systems exceeds what an average person can navigate? we could end up with a web where sophisticated AI can pass as human but many humans can't pass as "human enough."
it's an arms race, but with a twist: the bots will learn to solve the maze by simplifying their approach. sometimes lower "intelligence" is more effective than higher intelligence for specific tasks.
the entire history of captchas suggests this outcome. first we built tests only humans could pass, now we build tests that frustrate humans while bots breeze through.
(strange thought: maybe the ultimate defense isn't complexity but unpredictability; systems that make no sense at all but still function. humans excel at navigating absurdity in ways AI still struggles with.)
Upvote
-10
(1
/
-11)
https://en.wikipedia.org/wiki/Dead_Internet_theory
The main issue with this theory was largely the timing, it was ridiculous to be proposing its end culmination in 2016 or 2017 unless you were on 4Chan, nowadays however with things like Cloudflare and Nepenthes necessary and the cat-and-mouse game to get crawlers and scrapers working, the only thing that changes about this theory are the details but the logic is enough to argue for the immediate cessation of publically accessible LLMs/generative AI models and their associated crawlers.
Making the conspiracy theory more 'fun' is the fact that localized generative AI means Pandora's Box is open and even if state actors don't use it themselves or regulate it, private actors of various sizes will all be ready and willing to 'flood the zone with shit'.
As any person who thinks about these things for two seconds, there is an endpoint where 'flood the zone with shit' is another version of the https://en.wikipedia.org/wiki/Tragedy_of_the_commons and because of that fact there is no 'win condition' or way to not harm everyone at once. The only logical endpoint for all of it is another spin on https://en.wikipedia.org/wiki/Accelerationism
The logical, ethical and moral counter to Accelerationism, of course, is varying beneficial forms of https://en.wikipedia.org/wiki/Utilitarianism which requires no further debate if brought up as a counter to Accelerationism and the Dead Internet Theory, e.g. why regulation and/or an external force may be necessary to break those cycles.
Last edited:
Upvote
0
(0
/
0)
Yes, the very same. That relatively minor inconvenience (to a human, on occasion) is great for throwing a spanner in the works of someone trying to do a DDoS attack.
CloudFlare is usually seamless. When you notice it, it's because it's trying to keep a particular site alive in the face of something that would ordinarily bring a top tier web server to its knees.
Upvote
4
(4
/
0)
We're just touching surface of what AI can do for humanity and we're already trying to poison it. Sad.
Upvote
-15
(0
/
-15)
We live in a time when there's not enough money for healthcare or housing, but UNLIMITED FUNDS to throw at AI. . ..
Upvote
11
(11
/
0)
Drastic measures are needed. Drastic things like completely ending the use of fossil fuels immediately, for example
Upvote
3
(3
/
0)
That’s because the rich and powerful believe AI will be able to free them from paying workers anything at all
Right now, we don’t collectively realize it, but they need us more than we need them. We have the power, even though we aren’t using it
AI promises to change that and make workers completely powerless. And we have to recognize our power before that comes to fruition
Upvote
2
(2
/
0)
No, the goal is to shut down the shitgibbons who seem to think that they can scrape any and all content they might ever possibly reach, for free.
If said shitgibbons were to train their models only on public access information or with information they actually have permission to use, none of this would be needed.
Upvote
9
(9
/
0)
"No real human would go four links deep into a maze of AI-generated nonsense," Cloudflare explains.”
If only that were true we might not be in quite so many of the messes that we are in
If only that were true we might not be in quite so many of the messes that we are in
Upvote
0
(0
/
0)
CloudFlare can play judge tomorrow if they like. You are always at the whim of a private entity; so long as they aren’t discriminatory (in the legal sense of the word) they don’t have to do business with you.
And your argument is a very American one. A lot of other countries — ie Germany — have laws about content and don’t seem to be having the issue you are concerned about. While I appreciate the protections of the 1st Amendment, is it also possible for the Paradox of Tolerance to be at play?
Upvote
4
(4
/
0)
Suggesting it is an American ethos plays into the commonly-held misperception that the First Amendment is relevant to business relationships between non-governmental entities. It seems more rooted in mythical constructs of sanctified “neutral grounds” or “honor among thieves,” which would have been imports from older cultures.
A more apt analogy might be the “neutrality” of Swiss banking. In contrast to that approach, even in states where marijuana is legal, federal banking regulations hamstring the industry’s ability in the US to operate in what would be considered a reputable, aboveboard manner.
Without defending or even commenting on Cloudflare, there is a legitimate argument for providers of certain functions essential to the smooth functioning of modern society and its economy with whose interference by malicious actors—explicitly including sovereign actions of nation-states—to have leeway to provide such services IF they choose to do so—that fall outside of reasonable defined aiding and abetting. As with Swiss banking, there is ample room for reasonable regulation of such services.
Those who choose not to do business with providers because of their other customers, similarly, should not face sanction for their exercise of free association.
Personally, my DNS requests are resolved through a bunch of 1s rather than a bunch of 8s because of my personal evaluation of the business ethics of their respective providers.
Upvote
5
(5
/
0)