Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

lajili.com

Post reply on HN
Print view View original

131–140 of 392 posts

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#131

Post by fnordpiglet »

px43 wrote: post #58

Imagine a world where the only content you see is from publishers that you trust, and that your friends trust, and their friends, to maybe 4 or 5 hops or so, and the feed was weighted by how much they are trusted by your particular social graph. If you start seeing spammy content, you downvote it, and your trust level from that part of your social graph drops, and they are less likely to be able to publish things tha…

This was how Epinions worked for products - you built a graph of product reviewers you trusted and you inherit a relevance score for a product based on the transitive trust amplifying product reviews. It was a brilliant model (it was a bunch of folks from Netscape including guha and the Nextdoor CEO, got acquired a few times and google shopping killed their model, eventually acquired by eBay for the product catalog taxonomy system - which I helped to build)

I would say the current model of information retrieval against a mountain of spam is already broken and LLM will just kick it over into impossible. I feel like we are already back to the world of Lycos, Excite, and Altavista where searches give you a semi relevant cluster of crap and you have to query craft to find the right document. In some ways I think the LLM chatbot isn’t a bad way to get information if it can validate itself against a semantic verification system and IR systems. I also think the semantic web might have a bigger role by structuring knowledge in a verifiable way rather than in blobs of ascii.

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#132

Post by jsemrau »

DebtDeflation wrote: post #22

An endless loop of AI generated content that gets posted to the web as original human generated content, with LLMs getting re-trained on this content and spitting out more content that also gets re-posted, resulting in a cesspool of BS masquerading as organic knowledge. I'm old enough to remember when Google provided meaningful search results rather than just SEO spam, the problem is about to get an order of magnitud…

What is content? Is content entertainment? If you look at entertainment we will soon be living in world where short clips can be autogenerated and personalized based on keywords and criteria.

Stable Diffusion provides already good enough images to cover a lot of visual content online.

If you look at image shares sites for the purpose of entertainment like Imgur, you will also notice that a large portion of the viral content are screenshots from Twitter or traditional media.

Is content opinions? Most people don't have opinions on every topic on the planet. Is Gobekli Tepe the place of Noah's Ark? What's going on with Hunter Biden's laptop? How will Meta's VR strategy work out? Will it rain tomorrow in Sydney, Australia? Depending on your area of interest you might or might not have an opinion about it which you may or may not publish online.

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#133

Post by dpflan »

DebtDeflation wrote: post #22

An endless loop of AI generated content that gets posted to the web as original human generated content, with LLMs getting re-trained on this content and spitting out more content that also gets re-posted, resulting in a cesspool of BS masquerading as organic knowledge. I'm old enough to remember when Google provided meaningful search results rather than just SEO spam, the problem is about to get an order of magnitud…

Infinitely-expanding snake eats infinitely-expanding tail.

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#134

Post by giantrobot »

danielvaughn wrote: post #80
DebtDeflation wrote: post #22

An endless loop of AI generated content that gets posted to the web as original human generated content, with LLMs getting re-trained on this content and spitting out more content that also gets re-posted, resulting in a cesspool of BS masquerading as organic knowledge. I'm old enough to remember when Google provided meaningful search results rather than just SEO spam, the problem is about to get an order of magnitud…

I don't have much knowledge around AI, but from what I can tell, it's dependent on inputs from across the web right? If so, then as the use of ChatGPT grows, it'll slowly get consumed back into the model. With enough iterations, it will start to veer more and more away from recognizable human speech/thought, like a recursive game of telephone. Unless I'm completely misunderstanding how ML works, which very well may b…

> Unless I'm completely misunderstanding how ML works, which very well may be true.

No, you got it right. Describing it as a game of telephone is a great analogy. This is exacerbated by the confidently incorrect problem. LLM output looks sophisticated and correct and may at time actually be correct. However some unpredictable percent of the time it will be incorrect and confidently so.

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#135

Post by avitous »

myth_drannon wrote: post #126

We will need something similar to the Biblical flood to flush everything away. And restart from local trust islands similar to what we had in 80's-90's with BBSs and possibly Fidonet. I don't know how it's going to work but I just don't see any future in Internet in its current commercial form.

Herbert's "Butlerian Jihad" keeps coming to mind; I wonder if something similar lies in our future.

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#136

Post by ozten »

Each webpage should express a piece of metadata that is content-quality.

* primary - The content on this page is a primary source

* secondary - The content on this page is high quality, but which is primarily research based and may thus be tainted by unknown sources

* bot - The content on this page is mostly automatically generated by one or more ML models with some human curated improvements

HTML elements can also have data-content-quality to override the page level metadata. So if you have a primarily sourced paragraph.

Search engines can index this signal. Sites that claim primary, but which frequently are provably wrong or bot generated can be penalized.

(edited for formatting)

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#137

Post by altdataseller »

isthisthingon99 wrote: post #50

Stop guys, stop. Be young in your mind. Be young. How are young people interpreting this? "Oh wow, I can get it to write or help edit essays" "I can use it as something to bounce ideas off of" "I can use it to take ideas from my head into the digital realm." Stop being old people.

Or just not play the game. If it's already this popular/hyped, it means it's too late to create a product and profit from it. Find the next big thing before it's overhyped

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#138

Post by dmonitor »

TIPSIO wrote: post #83
blhack wrote: post #76

Earlier quoted context omitted.

You could call it: "AI Kessler syndrome". The internet becomes so full of hallucinating AI output that it becomes impossible to train the model.

Or, the singularity.

the singularity implies reaching a point where the AI’s improvement becomes self sustaining. this is the AI choking itself to death with bad training data.

Re: Chat GPT is the birth of the real Web 3.0, and it's not going to be fun

#140

Post by giantrobot »

buzzerbetrayed wrote: post #104
danielvaughn wrote: post #80

Earlier quoted context omitted.

I don't have much knowledge around AI, but from what I can tell, it's dependent on inputs from across the web right? If so, then as the use of ChatGPT grows, it'll slowly get consumed back into the model. With enough iterations, it will start to veer more and more away from recognizable human speech/thought, like a recursive game of telephone. Unless I'm completely misunderstanding how ML works, which very well may b…

Unless it only reconsumes the “good” content. In other words, the stuff that got good reactions from humans. In which case, it will get better, not worse, at least at generating clickbait. But at least it will be coherent clickbait.

Not necessarily coherent. If the first generation model starts misusing a word or phrase or adopts a common misspelling, later generations of LLMs will pick up and amplify the error. Eventually you'll get purple Monkee dishwasher begging the question maps such as.
Post reply on HN