Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I agree. My gut reaction to previous GPT releases was interest, but for this one (before even reading it) it was dread.

I think we're very close to an inflection point where functionally all information is polluted by the possibility that it's completely hallucinated or built on something hallucinated. We're already getting there in some ways - google vs. seo, astroturfed forums, fabricated publications, and this is just that but way worse. Probably orders of magnitude worse in terms of exposed information surface.

It's basically a pollution - and one that's nearly impossible to clean. The ecosystem of referential information now has its version of microplastics.




>an inflection point where functionally all information is polluted by the possibility that it's completely hallucinated or built on something hallucinated.

Actually, that's always been the case. This isn't something new. For a while (since the start of the information age at least) we've been able to accept information presented by media, the Internet or any other source as correct and true simply because the bulk of it has been. That's not saying anything good about humanity, it's just that people don't bother to lie about most things because there's no advantage in doing so.

Between the time when language and writing began and the advent of the Internet, there was less information being passed around and a greater percentage of it was incorrect, false, or otherwise suspect than has been the case for the last 50 years. So, it was critical for everyone to question every piece of information they received, to filter what they accepted as truth from the garbage. There was still bias involved in choosing what to believe, but critical thinking was a routine part of everyone's day.

I think it's going to be making a comeback.


I'm interested if you know of any historical research that talks about this. I can see that as a possible theory, but the counter would be that there's a fundamental difference in the nature of 'information' between now and pre-internet, where the combination of pure bulk of data and targeting means it's much much harder to actually filter than before.

It's difficult to fix this problem by interrogatin the validity of things when consuming the information in order to interrogate it causes you to have an implicit reaction. Consider advertising that operates on raw association, or curating information feeds that are designed to provoke a specific conflict/reward response.


The article literally says that their own model verifies the answers of their model.

Welcome to clown world. It’s clowns all the way down.


The Internet is already full of crap. You just have to know where to look for the 'good stuff'. Wikipedia will stay, arXiv will stay. Chill.


While there will definitely still be places that are less impacted - those two will probably be near the first to become heavily damaged in terms of credibility.

Wikipedia has multiple controls that facilitate quality and authenticity of content, but a lot of them break down in the face of synthetically polluted generated info.

The cost of engaging with the editorial process drops to functionally zero as sock-puppets are trivial to spin up that are near-human in quality. Run 50 of those for n-months and only then use them in a coordinated attack on an entrenched entry. Citations don't help because they rely on the knowledge-graph, and this pollution will spread along it.

Really what's left are bespoke sources that are verifiably associated with a real individual/entity who has some external trust that their information is authentic, which is tough when they're necessarily consuming information that's likely polluted by proxy.


This is an arms race, except the second player hasn’t shown up to the game yet.

The regulators must sponsor fact checking AIs. Bing Chat is a start. Alas, the regulator’s as usual have no idea what’s going on, except this time the rate of progress is so large even technologists can’t see further than a year out. Scary times.


> functionally all information is polluted by the possibility that it's completely hallucinated or built on something hallucinated.

This is already true of human curated information, not sure its really something new.


personally i'm just so grateful that I got to experience the "authentic" "unpolluted" version of the internet.


They specifically released it for the Khan Academy to think up at least one positive use case. To quell those feelings.


I feel a little the same way, but I am also a pessimistically inclined person.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: