Hacker News new | ask | show | jobs
by beardedwizard 834 days ago
This is already happening and has almost nothing to do with llm, this is Google and Facebook business model because these are the marketable details.
2 comments

Is it happening? Is there anybody who would claim that Google and Facebook are showing them better-targeted content now than before?

Google and Facebook care only about shoving monetizable garbage in your face, be it through tainted search results, mindless "recommended" IG Reels/YT Shorts or FB Pages. That has driven people away from these platforms. Maybe not at the levels where their bottom line is being hit, but that is a lagging indicator anyway.

> mindless "recommended" IG Reels/YT Shorts or FB Pages

Only these clearly do work for the majority of people. They're ubiquitous in popular culture (and in fact culture is often built within them) and are essentially the whole business model for those services. If they don't work for you, great, it means you've successfully avoided using those services enough that they can't profile you accurately, but that doesn't mean that they don't work at all.

> claim better targeted content

Any such claim is irrelevant. The personal opinion of the receiver of the content is not relevant, only that the content delivered somehow makes money for the sender.

> That has driven people away from these platforms.

Platforms? These entities do not derive profits only from visits to their own domains. Please inspect the source code of any random site you read next. On the majority of web sites in the Western hemisphere you will find either a Facebook script or a Google script, or both. Often more than these two.

I don't think the current systems used by Google and Facebook can integrate and infer new details as well as the LLMs can; LLMs can do more with less.
I know LLMs are pretty wild technology, but you're wrong about this. LLM's are just models of language. To some extent they can do some of the things you're talking about, but almost by the very nature of the problem other models could do the same better and with much less by way of resources.
What makes you think the details provided by LLMs would be accurate?