Live data from Hacker News

The Internet Is Full of AI Dogshit

aftermath.site

511–520 of 611 posts

Re: The Internet Is Full of AI Dogshit

#511
post #49

Earlier quoted context omitted.

No, that's never happened before. You're crazy.

> No, that's never happened before. You're crazy. Start filling up Discords with insane AI-generated garbage, and maybe you can devalue the data to the point it won't get sold. It's probably totally practical too, just create channels filled with insane bots talking to each other, and cultivate the local knowledge that real people just don't go there. Maybe even allow the insane bots on the main channels, and cultiva…

This idea has been talked about enough that we call it "Habsburg AI". OpenAI is already aware of it and it's the reason why they stopped web scraping in 2021.

Re: The Internet Is Full of AI Dogshit

#512

Earlier quoted context omitted.

It's probably people who waste others' time with baseless complaints like this that completely ignore substance that have ruined the internet, and not the fact that authors of interesting substantive content that actually gets consumed, whom also ask for some form of support that have ruined the internet.

It's not a baseless complaint to observe that the internet was better when you could simply click on a website and read it, as opposed to dismissing several popups about tracking cookies or like-and-subscribe.

I think it is, especially as a response to the idea that the internet is starting to lack substance

Re: The Internet Is Full of AI Dogshit

#513
post #94

Earlier quoted context omitted.

At this rate many exams will just become oral exams :-)

The paradigm is changed beyond that. Exams are irrelevant if intelligence is freely available to everyone. Anyone who can ask questions can be a doctor, anyone can be an architect. All of those duties are at the fingertips of anyone who cares to ask. So why make people take exams for what is basically now common knowledge? An exam certifies you know how to do something, well if you can ask questions you can do anythi…

That's not how knowledge works. Think of exams where you could have your textbooks and use them.

Re: The Internet Is Full of AI Dogshit

#514

I hate to keep beating on this drum but is _anyone_ working on a search engine yet that filters out anything with ads and/or a tracker? The root cause of the problem is content monetization so please let's just stop indexing anything that tries to monetize content and/or track readers. And I'll say this for all the overly-capitalistic folk on here: the amount of money I'm willing to pay for access to a search engine…

That would solve some problems but not all; spam sites would be filtered out but also sites with good quality information that monetize with ads would also be filtered out. There is no silver bullet(one way solves all) for SEO/ad/affiliate spam.

Re: The Internet Is Full of AI Dogshit

#515
post #393

Earlier quoted context omitted.

No, SEO made things noticeably worse from the very beginning. This is almost tautological, when you think about it. Agree googles machinations have not helped, but disagree with your 90/10 split.

You can't really separate SEO and the Google algorithms. SEO is a product of the Google's ranking algorithms. The entire effort of SEO is to either to follow Google's official guidelines, or reverse-engineer how things work to exploit their algorithms. Those whole point of SEO is to score higher on ranking algorithms.

> Those whole point of SEO is to score higher on ranking algorithms.

Which is why they typically make things worse for the end user.

Agree it's a coupled problem wrt ranking algorithm. SEO messes with the algorithms, algorithms change to account for it, lather rinse repeat.

More recently though, google make presentation changes not as simple as ranking, which made things worse. Which I think is what GP was referring to.

Re: The Internet Is Full of AI Dogshit

#516

Earlier quoted context omitted.

I remember when the angry mob, lead by the Penny Arcade guys came after Scott McCloud for advocating the end of the ad funded Internet in favor of a micropayment driven one. Turns out, he was right!

I don't think micropayments can work, either. The vast majority of Internet content—and I mean stuff that does get seen, not obscure sites with a dozen visits a years—isn't even worth a penny to most (not all, but most) of the people who "consume" it, but is worth a lot more than micropayment-amounts to some smaller set of people. Enforcing micropayments wouldn't, for nearly all sites that rely on ad revenue now, res…

I don't think micropayments can work, either.

What I meant, was that Scott McCloud was right, in the sense that ad supported content had some serious problems.

Re: The Internet Is Full of AI Dogshit

#517
One thing I find increasingly frustrating is reading an article where bits of it are plausible but the whole thing feels "off" and not being sure why. It's too verbose or vague. Was it AI generated? Someone working in a content farm? A non-specialist hired to write posts for SEO? A mixture of all of the above?

However, I don't think we've reached the end of the world yet, because if you just have the patience to learn what the reliable sources are, and the wherewithal to pay for a few of them, then you can ignore most of the junk, just as you can mostly ignore social media.

Also, y'know, books. Physical books. Public and university libraries. Yes some books age poorly or don't deserve to be published, but consider how many more checks most books go through before they end up on a shelf, including curation by librarians.

Re: The Internet Is Full of AI Dogshit

#518

What’s the chance that some of these comments are AI bots? Genuine question. When will OpenAI create an AI PR team?

I just did a demo this week for someone who asked how well GPT4 would write forum replies in their voice. The results were startling to be honest. The only way I would (maybe) tell the comments were AI is that they were slightly better quality replies than the author themselves would have written. I wouldn't go forward with it because it felt unethical, but it was a really fascinating thought experiment.

I’m sure we’ll see a lot of this soon with the US presidential election.

Re: The Internet Is Full of AI Dogshit

#519

That there has always been terribly-rendered dreck on the Internet. Badly-written copy by semi-literate self-appointed "authorities." Kitbashed plagiarism that only half-understood its sources. Conspiracy theories galore from garbage-minded gasbags. Totally muddle-headed hot takes. But it was manually generated. Or semi-automatic at best (like bad machine translations of foreign language sources). The issue with AI i…

>Not all AI-generated content will be horrible. And not all human-generated content will be trustworthy. There will need to be a way to mark reliability based on community-oriented standards. For AI-generated content (and human-generated content too) there will need to be ways to cite sources.

Exactly; if AI can for example summarize a research paper or a book better than human then it is hell of a more useful than any person could possibly be and that's the point of AI(that it possibly can exhibit super human skills).

Web search engine is a massive data mining and data analytics software solution powered by graph theory and data science algorithms for information retrieval. No search engine can know what information is per se better and no search engine can know what you know and know what you want but it can use graph theory and data science to approximate and show you what it thinks that you want.

Re: The Internet Is Full of AI Dogshit

#520
post #49

Earlier quoted context omitted.

> No, that's never happened before. You're crazy. Start filling up Discords with insane AI-generated garbage, and maybe you can devalue the data to the point it won't get sold. It's probably totally practical too, just create channels filled with insane bots talking to each other, and cultivate the local knowledge that real people just don't go there. Maybe even allow the insane bots on the main channels, and cultiva…

This is an intellectually fascinating thought experiment.

> This is an intellectually fascinating thought experiment.

It's not a thought experiment. I'd actually like to do it (and others to do it). IRL.

I probably would start with an open source model that's especially prone to hallucinate, try to trigger hallucinations, then maybe feed back the hallucinations by retraining. Might make the most sense to target long-tail topics, because that would give the impression of unreliability while being harder to specifically counter at the topic level (e.g. the large apparent effort to make ChatGPT say only the right things about election results and many other sensitive topics).

Post reply on HN