[Yemeth]

I'm not sure how LLMs are going to be immune to SEO spam and advertising. As if human nature would magically transform and people would stop buying stupid stuff.

LLMs are already being used to make the web even less useful, by shitting out vast ammounts of meaningless and even outright wrong text for SEO purposes. And don't forget the systems are being trained on the web in the first place… using a LLM that is able to utilize a web search already thinks listicles are useful information and not just a way to place affiliate links.

When the hype is over and the VC money dried out, companies will find ways to make the LLM interfaces and outputs an 'advertising friendly' affair.

[winwang]

I think the idea is that, if we (some of us) can figure out when something is SEO spam or rather, generally low quality, an LLM should be able to but faster and more quantitatively why.

[tempodox]

And who, pray tell, would have an interest in giving us that? The internet overlords are advertising companies. Would you pay for such an LLM? Those companies have gotten so powerful because everyone wants stuff for “free”.

[Terretta]

> Those companies have gotten so powerful because everyone wants stuff for “free”.

On the contrary, my experience of the Internet since pre-web is those companies have gotten so powerful because now everyone 'creating content' wants to get paid.

Put another way, the gwerns of the world may start by posting ad-free content just because they feel like sharing. "Those" companies can't profit from individual pamphleteering.

But usually, as soon as a gwern has a measurable readership, they imagine money, and start supporting the ad industry. Surprisingly quickly, chasing more ad revenue becomes the point of their content instead of just sharing whatever they had to say.

Today, it seems most people start by wanting to get paid, and come up with a type of content to create.

Curiously, and as a result of the type of targeted searching gwern describes here, you'll generally find the best content is that which is still published free (self hosted or as open papers) thanks to motivations of the content's creator.