Flood of AI-generated SEO chum content may put the web out of its misery

Originally published at: Flood of AI-generated SEO chum content may put the web out of its misery | Boing Boing


One thread of this multiverse takes us to a retro-futuristic web of layered curation relying on aristocratic “expert voices” i.e. a New Yorker Article profiles an online curator who is a physical hermit but digital native (semi-retired after years of rotating museum and avant-garde Jazz criticism jobs) who keeps a bespoke blog of their favorite links of the month.


With more and more of the text on the Internet being produced by LLMs, future LLMs will be trained on that (in short, they will eat their own shit), leading to the self-limiting phenomenon of model collapse.


I wonder if at some point the ever increasingly powerful “AI” that is being baked into our portable device chips (and the coming new desktops) starts carrying some of this load. Why have AI generate various content and then send it across the web when it could be created local and tailored even more specifically for the viewer? I realize currently that seems like a bit of a far off idea in terms of processing power, but in a decade who knows.


This is shit. If AI is taking over, where’s my Gigolo Joe?


Considering some phones are already doing this with photos of the moon I can definitely see the day this happens.


Yep, the recursive nature of the training has been one of my concerns for a while. I wonder if there will soon be an effort (or if there is one already) to make special LLMs that are trained entirely on “clean” data such as digitized books that were written prior to the advent of generative AI.


This isn’t a problem for the Web. This is a massive problem for Web Search.

The greybeards remember a time before Alta Vista and Dogpile and Google when you collected links to things that were valuable, and revisited them occasionally. Maybe I’ll start a webring.


digg Lives!


“So, are the Artificial Inanity systems still active in the Rampant Orphan Botnet Ecologies”, asked Arsibalt, utterly fascinated.
“The ROBE evolved into something totally different early in the Second Millennium”, Sammann said dismissively.
“What did it evolve into?”, Jersy asked.
“No one is sure,” Sammann said.

– Anathem, Neil Stephenson, pg 833


ahh the naivete of youth. next we’ll be yelling, “get off my starfield!”


“So,” he asked, “what’s your favorite Gopher client?”


Bring back web-rings!

Top 98% of the Internet!


I’m trying to remember where I saw it, but there was a thought experiment about a social media filter that would identify toxic people, and start to silence their content to the greater community. Instead, it would create a “safe space” for them to engage in, and use AI to generate content as well as responses for these people to “engage” with. In theory, the toxic person wouldn’t know that they’re not actually engaging with the greater community.


And you know the AI-SEO garbage is going to make things intolerable given how bad search has already been getting in recent years. In the early days of the web, I got used to not finding what I was looking for because it wasn’t on the web. Then there was a glorious period when information was both available and discoverable, but increasingly the desired information gets swamped out by garbage, by highly popular but not-particularly-relevant sites, and by search engines doing “clever” (i.e. stupid) things behind the scenes like swapping out search phrases for common equivalents (that aren’t, in this context). A bit of Google-fu used to fix that, but increasingly it seems like it doesn’t.

Post-LLM searches are going to quickly be impossible, and the information space will be hopeless contaminated, at every level.

Where most of the links are dead. (It’s surprising how quickly it happens.)

In trying to recapitulate the history of web search engines, we end up having to acknowledge there’s a reason why they evolved beyond that stage, too…

I don’t see how that doesn’t take down the web with it, though.

Ah so my device can create, just for me, some… (checks notes)… bespoke garbage text! Cool!

Someone made the comparison of steel produced before the atomic tests, which seems about right.

Outside of users who spend all their time on social media, is there really a difference, at this point? (And the social media users are fine only as long as the platforms are stable, which…)


I’m pretty sure the idea that everyone on 4Chan is a bot and that the entire thing is nothing more than a 3 letter government honeypot gets tossed around on there pretty regularly. Depending on the board it wouldn’t be hard to have AI generate 99.8% of the post in any given thread. I’m just spit balling a percentage there because a lot of threads could be 100% AI. Now a days an actual post by someone that I know on Facebook is rare, it’s 49% ad content and 49.5% videos.

I don’t like the idea of a digital safe space. Humans IRL do create safe spaces for themselves, but they are also forced to be social with other people who range from coworkers to complete strangers. Being in a space that constantly reinforces your ideas and confirmation biases feels good in one aspect, but it just widens the gap between what you think you know and what you don’t want to know.


Bespoke garbage isn’t exactly what it’ll be though. Imagine if your iPad is able to create adverts for whatever company is funding whatever site you are looking at (or app you are using) and then uses the iPad camera coupled with the iWatch you are wearing to track you eye movement and heart rate. Enough iterations and it’ll know exactly what you like, biofeedback controlled by AI for marketing purposes. Skynet never needed to start WW3, it’ll already control us.

If AI at that level comes to individual devices then Apple’s (or the device maker) profitability for your data profile will make what Facebook or Google have on you look like a cheap horoscope. (Unless it’s Android, then it’ll be 50% Google’s anyway.)


It also learns that you don’t earn enough money to afford the best products, so it submits job offers on your behalf and uses hiring algorithms to get you better employment. But it also knows you don’t want to leave your current job, so it just does it remotely on your behalf. The AI is “rewarded” when you buy stuff from an ad, so it just uses that secret, extra, remote-job income to buy the advertised stuff for you.

Everybody wins.


Reminds me of the Mythic Quest episode.

ahh the naivete of youth

too late I’m
File:Under construction graphic.gif - Wikimedia Commons