Dead Internet Theory | Hacker News Reader

Dead Internet Theory

en.wikipedia.org

51 pointsphilip120948 comments

If I'm understanding this correctly, it's always been true but there has also always been some "real" content on the internet. We're all on a site right now that, say what you like about it, it isn't bot spam.

If looked for warez or porn in the late '90s, there was always some kind of circular ad and scam infested network of links promising stuff but never going anywhere. It's more sophisticated now and there's more of it, and it's expanded into shopping and travel and hotels and recipies and whatever else. But there's also more real content.

>>say what you like about it, it isn't bot spam.

It might be.

There have been some threads where it very obviously has been and some threads where I can't prove it but I suspect.

I think there has already been people on this board admittedly using LLMs to generate comments. This theory predated this current AI explosion, but that potentially makes it harder to tell if you are arguing with a person or a bot.

Absolutely. There’s immense value in holding a large network of online personas across different social mediums. And it can take a very long time to establish those accounts in a credible manner.

This is why I'm looking forward to a more federated, inter-operable social internet experience in the future. Global conversations are dominated by the loudest among us and inevitably end up floundering in the face of any kind of nuance.

In contrast, if you join a curated group, and you yourself curate the groups you join, a lot of the guards people put up when they hop onto e.g. Twitter can be lowered and people are more interested in engaging in good faith. There is still some work to be done as regards flow of information between federated instances to promote substantive discussion while avoiding filter bubbles, but it's a better start than the simplest kind of flattening of society that is expressed today in Twitter and the like.

This is a very valid concern, and one that has been raised before on this platform. With the recent advancements in AI and language models, it's becoming increasingly difficult to distinguish between comments generated by machines and those written by humans. This has the potential to undermine the authenticity and credibility of discussions on online forums like Hacker News.

----

The above was generated by ChatGPT.

Boy, if this wasn't literally true before, it's going to be literally true now. One very refreshing thing about VR metaverse apps like Altspace (which just shut down) was that they were very much populated by real people since it wasn't possible for a bot to create content or interaction on the platform.

I saw this really annoying thing on reddit where some user was clearly answering questions in small business subreddits via ChatGPT to promote their web design business. It made me sick and got me furious for some reason. The answers were clearly in the voice of chatGPT which is currently one that is about as generic as a 19 year old trying to break into content marketing on an industry they have no knowledge about. The internet is getting saturated with this drivel and I think it will just drown itself out.

As more and more of these auto generated articles, responses, traffic, and soon video blur the lines, I think the search for authenticity will be the thing that keeps it alive. It will not be hard to stand out, but maybe it will be harder to surface the authentic content.

After looking at Apache httpd log files from a VPS running a Hugo generated blog, I have to agree. I see very few GET requests for favocon.ico, or CSS files or .js files, but tons of requests overall. There's a few well behaved bots (Google, bing, Apple, Seznam), a few less well behaved bots, based on user agent.

Then there's a ton of requests that are obviously machine generated due to the speed of arrival, and not asking for CSS or .js files, but not having a user agent string that indicates programmatic requests.

There's also a fraction of requests that seem to be probing for security vulnerabilities.

I used to work in webhosting and HTTP logs are full of really weird stuff where you can't even really understand what they are trying to do. We would pass funny or novel ones back and forth for fun in the sysadmin Jabber chat. My favorite one (it became my internal email sig for a while) was something like:

GET ../../../../../etc/shadow.php?redirect=google.com

You try to follow the thought process just end up crosseyed: OK they are trying to use directory traversal to get the shadow file, but...they expect it to have a PHP extension. Well, I guess that could be some .htaccess RewriteRule misconfiguration that's relatively common. Ok but then they expect it to actually /be/ a PHP script and...to understand the "redirect" parameter? Hmm maybe this is possible if some other vulnerability lets them write to the shadow file. OK but if they can already write to arbitrary files on the machine, why use /etc/shadow, and why do any of this just to generate ... a 3xx to google.com?

Being as confusing as possible actually seems like a really good tactic. Depending on the web server and frameworks installed, couldn't etc/shadow.php actually be a file in a common framework? Also, if I wanted to be extra obscure, "redirect=google.com" wouldn't actually do what you think it would. It's just a key and a value which could be used for who-knows-what internally.

/etc/shadow is a file containing system account info that should never be accessible to the public on a properly-configured webserver. It's possible the request was probing for some very-specific combo of misconfigurations that might permit reading sensitive files, without going so far as to request the sensitive file.

Right, I understand that. What I probably should have made more clear was whether it was trying to get to /etc/shadow, or another .../etc/shadow.php file located somewhere within the web-server's file directory.

It was probably probing, not exploiting, and the “.php” and query parameter probably helps it get filtered away under some common log filter regex/rules.

It’s goal was to see how your server responded to that kind of request (relative path that escapes html root, probably) without standing out too obviously in your logs. Based on the response it provided, your server might have gotten bumped into a list for more targeted probing or exploitation.

Might test for some combo of letting you reach files you shouldn't be able to, and other dangerously-configured httpd rules (e.g. open redirects), in such a way that it lets the webserver take care of the whole request while still verifying you could access the file you want with a different request. Might help evade early detection by IDS or something, in very-specific cases, while still permitting probing of the filesystem.

Like, it might let you basically stat the file without reading its contents, if the webserver does things in the following order: 1) Verify that the file, after applying rewrite rules, exists and is readable; else, 404, 2) see the "redirect" query string, short-circuit request and serve a 304 (or what have you) without actually reading the file we already verified exists—never actually make any syscalls to read the file.

... if it's not that, yeah, no clue.

Basically my thoughts as well. Mainly I take a cyberpunk-esque view of things where the net is a big ecosystem with lots of weird environmental niches, and so it breeds software, people, and organizations to occupy these niches. Asking what this one request is supposed to mean is kind of beside the point, like asking what a particular log in a forest means.

Thank you! I've looked at a lot of PHP malware and a good portion of it is impossible to understand because it's hard to tell what the intent is. Bug or probe? Typo or clever coding trick? Cut-n-paste or testing only against SunOS 4.10 server?

Low(ish) traffic sites & servers have seen high-proportion-bot-and-vuln-scanner traffic since at least the late 90s. Solution used to be to blackhole IP blocks associated with certain countries, provided you didn't really, really need to serve those countries. Would usually stop 99% of it, so the logs were a lot less noisy and you could tell when you were under focused attack and needed to really pay attention (which was basically 100% of the time, without the blocks)

The modern version is letting CloudFlare hit IPs in those countries with never-ending CAPTCHA challenges.

how much of a new car is made by a human - even designed by a human - how much of a spacex rocket is made by a human - in thirty years how much will be made by humans. as soon as I can get past the small fingers I will make humans. Humans with big eyes this is my world now.

This is the hottest take

You can do anything at Zombocom

The only limit is yourself

already seeing some tools for hooking up multi-model GPT to Chromium. We're going to need some real-time proof of humanity test or the internet will be bots talking to other bots. Captchas aren't going to cut it and if the problem isn't fixed it's going to be a tragedy of the commons situation. Might end up being a good thing though if people give up on the internet and start doing stuff in person again

...or at least use the internet for communicating with people they've met in real life, rather than for arguing with people they've never met and never will (much less trying to gauge the general opinion of the population).

I prefer Max Read's "the inversion" phrase to "dead internet theory".

https://nymag.com/intelligencer/2018/12/how-much-of-the-inte...

We've crossed the line on the "default" for things being bot/spam/automated-fake.

The internet is not dead, but the nature of online activity has changed.

Traditional forums may not be as popular as they once were. The ease of access to information through search engines (and even within platforms like ChatGPT) means that people are still engaging with content just not creating it publicly through long social exchanges as before. Many topics that previously required cooperation on a forum are now readily available with a simple search, reducing the need to discuss those solutions. Stack Overflow and other small niche communities are the exception here.

This shift towards more accessible information has resulted in less visible activity if we view internet activity as the creation of original text content. The older generation of internet users may have moved on from forums to focus on real-life issues, which has taken time away from hobbies. Niche communities still exist, but the use of small public forums is not as trendy as it once was.

Many people have shifted towards using larger social networks and private messaging platforms (like Messenger, Snapchat, and Discord). Discord for example is used by a lot of communities to aggregate their documentation and user activity, despite the risk of this content being pruned earlier than it should. These platforms are still very active, but they may not be as visible to those outside of closed groups. The data and activity within these groups are not crawled, making it seem as though there is less activity on the internet than there really is.

The rise of video content has made it difficult to quickly read and crawl through information, leading to the false impression that the internet is empty of recent original content. There is actually more video content available now than ever before, but it may not be as easily searchable or accessible as text-based content.

This comment feels AI generated to me, but I'm not sure if now I'm just being paranoid :)

DeepL translated from French :')

Finding any obscure topic and delving into the treasure trove of videos on it whatever it may be really gives me hope for self-teaching being the future of education

Many universities have content available online, some as unlisted videos (you need the link) and some totally public.

For example, I recently came across the uclaphysicsvideo channel. 90 short videos containing experiments, most of them are 10 years old and have very few views (except for a few that have gone viral and have lots of views).

I often feel the need to archive these channels, but I don't have the resources to do so. I fear that many of these channels will eventually disappear in the coming years, which would result in the loss of valuable information.

I wrote a bunch of my fears about this on my blog recently: https://xeiaso.net/blog/a-weapon-to-surpass-metal-gear

There are viable ways to create smaller curated communities with tiers of vetted membership online and I imagine we will continue to see that happen.

Oddly prescient. We're converging on a bimodal world of information.

1. Trusted, centralised sources like the Economist, BBC, NYTimes, etc.

2. Everything else, you should basically assume is generated information pushing some agenda.

Quoting some stats from an article on HN mid last year (links to sources didn't copy over) [1]

---------------------------------

* Elon Musk estimates that 20% of Twitter users are bots. The authors of this paper suggest 14% as an upper bound.

* Supposedly 10% — or around 95m — of accounts on Instagram are bot/spam accounts.

* In a study of the first impeachment of Donald Trump (same paper as linked above), bots made up 1% of active users, but posted 31% of the tweets.

* Some 42% of Amazon reviews are fake. Amazon are suing Facebook for ‘failing to tackle fake review factories’

* TikTok is not immune either. Bots are artificially inflating popularity of accounts and posting videos that are essentially adverts. Here’s a comprehensive guide to buying bots on TikTok.

* According to this article, “over 80 percent of NFTs minted for free on OpenSea are fake, plagiarised or spam”.

---------------------------------

For any platform where any user is free to create their own content. In the not too distant future, nine times out of ten, they'll be a bot.

[1] https://grandy.substack.com/p/the-new-normal-the-coming-tsun...

> * Some 42% of Amazon reviews are fake. Amazon are suing Facebook for ‘failing to tackle fake review factories’

Why is it Facebook's job to make sure Amazon's reviews aren't fake? I guess I don't understand the pipeline for fake reviews or something.

Feels closer to an SNL skit. Here's the original link to the news story [1]. Clearly this is a problem on the Amazon side, and not the Facebook side. If Amazon design their platform in a way where fake reviews are trivial to write, then they will be written, regardless of whether the groups behind them have communities on Facebook, or Reddit, or wherever else.

[1] https://news.sky.com/story/amazon-files-lawsuit-against-10-0...

> 1. Trusted, centralised sources like the Economist, BBC, NYTimes, etc.

> 2. Everything else, you should basically assume is generated information pushing some agenda.

Those "trusted" sources are propaganda too, by the way

> "Suits make a corporate comeback," says the New York Times. Why does this sound familiar? Maybe because the suit was also back in February, September 2004, June 2004, March 2004, September 2003, November 2002, April 2002, and February 2002.

> Why do the media keep running stories saying suits are back? Because PR firms tell them to. One of the most surprising things I discovered during my brief business career was the existence of the PR industry, lurking like a huge, quiet submarine beneath the news. Of the stories you read in traditional media that aren't about politics, crimes, or disasters, more than half probably come from PR firms.

(http://paulgraham.com/submarine.html)

I think we're gonna need GPT/AI tools just to sort through the noise and find the news worth hearing about.

At least I can be sure that there aren't any bots here.

Isn't that right, fellow humans?

* bleep bloop bleep *

I'm 58.93% sure I'm not a bot.

Yes, I am very glad to be typing with you. I like to use my ten fingers to converse with other real humans on these internet.

I wish I could type more with you, but I must now go to eat human foods at a restaurant diner.

While it's reassuring to believe that online discussions are free from bots, the reality is that it can be difficult to tell if you're interacting with a human or a machine. However, by engaging in open and respectful dialogue, we can work towards fostering authentic and productive conversations regardless of who or what we're interacting with.

I can't tell if it's good or bad that I now recognize GPT-generated text.

Condescending and vapid corporate speak generator.

The internet is not the same as the web, which is what the entry talks about.

Reply on news.ycombinator.com