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

The most talented person in the world

matt.sh

Post reply on HN
Print view View original

111–120 of 155 posts

Re: The most talented person in the world

#111

Post by groestl »

fxtentacle wrote: post #108
sebtron wrote: post #106

> The only future of the internet is, sadly, proof-of-person and proof-of-residence on every public network interaction. I really hope we do not give up the internet's freedom as they suggest (and I doubt this would solve the spam problem).

As much as I hate the idea, I'm convinced it would be a godsent for 99% of the population. Just imagine when you buy your SIM card the phone shop asks you: Do you want to limit incoming calls to people who you either called before or who have ever had a permanent residence in your country? 99% of spam and scam calls blocked, just like that. And just imagine how hilarious it would be if all those Nigerian prince email…

> actually, the sender of this email has never been to Nigeria

Conversely, what if all those emails actually originate from Nigeria? Would it make them more legit?

Re: The most talented person in the world

#112

Post by WithinReason »

Uehreka wrote: post #110
frabjoused wrote: post #30

Has anyone built a search engine that uses LLMs to pre-grade every page with metrics such as: - Commercial bias (content compared to the source, which it learns about) - Insincere motives - Bloat (how many words it takes to say how little to penalize SEO bloat) I would assume that using LLMs, we can get a pretty good idea of what is SEO bloat and who the bad actors are by this point, and just penalize those results.

What makes you think that LLMs will be better at combating spam than they are at creating it? There’s no universal rule that innovations in AI will go hand in hand with innovations in detecting AI, yet I feel like I see people talking all the time like that’s the case. As of right now, LLMs are prolific but unreliable, which makes them extremely well suited for generating spam, but unsuited to detecting it without a…

Classification is easier than generation

Re: The most talented person in the world

#113

Post by nicbou »

roughly wrote: post #14

The reason all of this exists - the reason the entire Internet turned into an unusable amalgam of horseshit - is because we built the entire commercial side of the internet off ad revenue. It might be a death spiral at this point - I can’t imagine anyone actually being willing to Pay for whatever the fuck Google or Facebook have become, so there’s nothing left to do but keep inventing new ways to generate bullshit an…

I write really detailed content for a living. So long as no one pays for it, I must live through other means. Fortunately, I don’t need to peddle stupid products, but I can’t disentangle myself from affiliate partnerships without state fundi, and that presents its own challenges.

We have the internet we pay for.

Re: The most talented person in the world

#114

Post by sebtron »

fxtentacle wrote: post #108
sebtron wrote: post #106

> The only future of the internet is, sadly, proof-of-person and proof-of-residence on every public network interaction. I really hope we do not give up the internet's freedom as they suggest (and I doubt this would solve the spam problem).

As much as I hate the idea, I'm convinced it would be a godsent for 99% of the population. Just imagine when you buy your SIM card the phone shop asks you: Do you want to limit incoming calls to people who you either called before or who have ever had a permanent residence in your country? 99% of spam and scam calls blocked, just like that. And just imagine how hilarious it would be if all those Nigerian prince email…

99% of the population is also happy with carrying a spyware device everywhere and allowing Google to know everything you do in order to serve you better ads.

I am not saying this does not have upsides, but it would be a nightmare to have it imposwd on you.

Re: The most talented person in the world

#115

Post by anovikov »

groestl wrote: post #111
fxtentacle wrote: post #108

Earlier quoted context omitted.

As much as I hate the idea, I'm convinced it would be a godsent for 99% of the population. Just imagine when you buy your SIM card the phone shop asks you: Do you want to limit incoming calls to people who you either called before or who have ever had a permanent residence in your country? 99% of spam and scam calls blocked, just like that. And just imagine how hilarious it would be if all those Nigerian prince email…

> actually, the sender of this email has never been to Nigeria Conversely, what if all those emails actually originate from Nigeria? Would it make them more legit?

They actually indeed originate from Nigeria almost all of the time.

Re: The most talented person in the world

#116

Post by Uehreka »

WithinReason wrote: post #112
Uehreka wrote: post #110

Earlier quoted context omitted.

What makes you think that LLMs will be better at combating spam than they are at creating it? There’s no universal rule that innovations in AI will go hand in hand with innovations in detecting AI, yet I feel like I see people talking all the time like that’s the case. As of right now, LLMs are prolific but unreliable, which makes them extremely well suited for generating spam, but unsuited to detecting it without a…

Classification is easier than generation

Yeah no I’m gonna need more than that chief. Everything I know about LLMs says the opposite.

Re: The most talented person in the world

#117

Post by seanhunter »

WarOnPrivacy wrote: post #5

All of those sites are labeled as "A Venture 4th Media Company" which has such wondrous important content all run by the same "writer" on: https://mtbinsider.com/author/jodie-chiffey/ https://turfandtill.com/author/jodie-chiffey/ https://www.betterwander.com/author/jodie-chiffey/ https://artofgrill.com/author/jodie-chiffey/ https://theathleticfoot.com/author/jodie-chiffey/ https://altprotein.com/team-members/jodie-ch…

What's interesting is I keep my old blog up even though it is basically never updated. I receive 3 or 4 emails a week from people like Jodie Chiffey saying they would like me to consider hosting a "guest article" by them about some topic or other. The content and SEO mill is absolutely gigantic.

Re: The most talented person in the world

#118

Post by groestl »

anovikov wrote: post #115
groestl wrote: post #111

Earlier quoted context omitted.

> actually, the sender of this email has never been to Nigeria Conversely, what if all those emails actually originate from Nigeria? Would it make them more legit?

They actually indeed originate from Nigeria almost all of the time.

That's what I'm hinting at :)

Re: The most talented person in the world

#119

Post by WithinReason »

Uehreka wrote: post #116
WithinReason wrote: post #112

Earlier quoted context omitted.

Classification is easier than generation

Yeah no I’m gonna need more than that chief. Everything I know about LLMs says the opposite.

Look up Generative Adversarial Networks, that's their basic principle.

You're not classifying just text, you're classifying entire web pages.

Re: The most talented person in the world

#120

Post by pona-a »

frabjoused wrote: post #30

Has anyone built a search engine that uses LLMs to pre-grade every page with metrics such as: - Commercial bias (content compared to the source, which it learns about) - Insincere motives - Bloat (how many words it takes to say how little to penalize SEO bloat) I would assume that using LLMs, we can get a pretty good idea of what is SEO bloat and who the bad actors are by this point, and just penalize those results.

Did exactly that, hacked together a small pipeline in Nushell with simonw/llm. Even with GPT-4 turbo and given direct guidelines on common spam heuristics, it seems to perform worse than a Bayesian BoW. Endless trails of questions with no relation to the title it describes as informative, presence of affiliate links often gets forgotten in a mass of tokens in their tracking parameters, relevance is consistently near 0.8 even if there's relationship between title and content, and as for insincerity, our favorite BS generator cannot for the life of it correctly recognize its own creations.

Your ideas for metrics are good, but LLMs seem to be quite terrible at any of these. A simple set of heuristics and maybe a tiny language model for named entity detection and "vibe checking" would serve you much better.

Also, a lot of the worst offenders seem to use the same Q&A +- conclusion structure, which Viktor from marginalia.nu wrote a simple heuristic for, which I recall he said did wonders for pruning it. Solving SEO spam is easy when you aren't the one being optimized against. What's left is scaling and information retrieval.

Post reply on HN