Hacker News Reader

Comment by zeruch | Hacker News Reader

I cant foresee a scenario where an endlessly growing mass of AI-generated noise, stacked upon itself, will be useful to anyone. Humans will invariably find a less contentious path (maybe its seeking out more localized options, maybe its balkanizing into various specialized domains, maybe its $NEWTHINGYETTOBESEEN, or all of the above. But the current trajectory cannot hold.

Someone else mentioned the "reverse-takeover of Google by DoubleClick is the Fall" and I find that spot on.

The Dionysian appetites of adtech will be its own downfall.

View on HN

People keep saying the Internet is going to be flooded by AI-generated content. Is it going to be any different than the million Wikipedia/StackOverflow re-skins? A bunch of trash sites we learn to ignore in our search results?

Generative AI can also increase the amount of good content, too. If I knew that a particular human vouched for an AI-generated code snippet, I would trust it as much as a hand-written StackOverflow answer. You can bring the entire arsenal of community and reputation-building to have more people vouching for answers and double-checking others.

That's why the future of StackOverflow is more like Phind, the code-specific LLM that looks like a search engine. StackOverflow already has the community. Now, they need to keep people there to check AI-generated answers.

And there's second order benefits, too. You could have AI check old answers, even human ones, on a regular basis. Imagine if an answer is marked "obsolete" because the AI found a more modern solution, or if it found a security issue.

" A bunch of trash sites we learn to ignore in our search results?" Exactly. But when/if that stack shifts the signal to noise ratio at escape velocity into a toilet, behaviors will change.

I'm just curious as to what that might look like. We've already seen plenty of examples as to what folks will do when an online option becomes sour/untenable...new ones arrive. Sometimes they look like old options repackaged. Sometimes they are radically new.

Encyclopedia Britannica type of resources are probably gonna make a comeback. I recently was thinking how much of an opportunity for growth there is going to be for fact checked/expert verified resources in the next 10-20 years.

I've had similar thoughts.

Eventually; someone will create an index to sort through the AI output and determine what's true/relevant and return it to the consumer.

Call it 'airank' maybe?

yeah definitely not truth value.

probably a "exploit" trigger. hard to think how reality and predictive power are going to be mediated through some kind of source index.

Yahoo!

It's basically Wikipedia.

Take as axiom that there exists infinite content of apparently human-readable stuff in the universe at random IP addresses. It is grey static, that maybe at one point in the past used to be ordered information. Now, all entropy.

A human managed and manually curated safe zone of information is the refuge. Real people had real discussions and votes and decisions to result in what you see. This is true with Reddit to a lesser but still useful extent. Democracy can be gamed, but it's 1000x better than Engagement Signal Ranking that drives Google or Meta's products.

The same principles apply to education, science, and political organization of human enterprise.

Basically, institutionalize your info-world.

Reply on news.ycombinator.com