Hacker News Reader

Comment by ballooney | Hacker News Reader

This is such a grim thing to read.

I disagree. I personally consider these my biggest problems with the Web:

  - Bias, specifically commercial bias
  - Webpage formatting: every website looks different, hides the information I want in different places. Sure, it "looks pretty", but I don't want pretty; I want info
  - Scams/SEO/etc...

The LLMs are very good at reading from multiple sources, parsing, and presenting only the data in a consistent format. There are many examples of this, but if you want a good one to try for yourself:

Google "How to make ham fried rice"; you'll get 10,000 articles, most pretty good recipes. But they're all different; most of them are just bait for ads. And most of them, the 10-line recipe is hidden between 50 useless paragraphs about how serving food is life's most important goal.

Now, ask an LLM to search for it, find the best combination, and list only the recipes. You get a perfect, 10-line recipe that doesn't waste your time.

And if you're lucky, it won't include rocks as an ingredient.

They're an important part of your diet if you happen to have a gizzard.

I've also been known to include some ground up rocks in my meals. I'm pretty picky though about them, I mostly just want a couple specific varieties.

LLMs are not unbiased. Nothing is unbiased.

Yeah I had to laugh at that. LLM's solve commercial bias? Coming from seemingly multiple the most valuable companies of all time?

Why even use a LLM, just buy a cookbook and you can even waste less time and know the information is valid. And again if your logic is consistent, google routes your search to an LLM automatically and renders the reply - that will save as much time and do the same thing since LLMs are so good at parsing and presenting information.

So why should I pick an LLM over google's inbuilt LLM or an actual cookbook on the desk? What wasted time is the LLM saving?

From Google's perspective you might as well be a bot because when you search there you don't see ads. If you don't have adblock installed, your parent comment applies because you get worse experience as LLMs don't serve ads.

I personally prefer to read a couple of recipes to understand the context, what people have tried, what they thought of it. Only when I have a good general idea of the recipe I choose the best one I found based on difficulty and the ingredients I have. I then do the plate, often not following entirely the recipe I chose, and after that and eating it I write down the name of the plate, the recipe link, the date, and a comment about what I did and how it went. Cooking is a lot more than executing a list of steps like a robot.

Well agreed, but it's not new, the web has been actively made user hostile by every website owner and their "273 partners".

Maybe, but it's reality. Not too long ago the rule would be "if it doesn't show up in google, it doesn't exist". Today agents+LLMs are better browsers than Chrome/Safari/Firefox/... If a webpage does not show up in an LLM it may as well not exist.

How is that grim? It's the dream of the Semantic Web coming true, just by different means than planed.

Because the economic, social, and climate impacts are at best uncertain and at worst devastating to the majority of human beings on the planet. It's always surprising to me that people don't intuitively separate the usefulness of AI from its risks. It feels like everyone has to be a doomer or a booster.

> at worst devastating

This is such a low bar literally anything would clear it. Even tomato farming.

I think you're going to have to flesh this out a little more, friend. I'm not sure what you're trying to say.

You know that the reason the semantic web didn't work out is that it only benefits a small number of participants (aggregators), right?

Why do you feel that ?

Because AI cannot be trusted as a reliable source of information. Information is more trustworthy when it comes from from primary sources on the open web.

Why not ? Can google be trusted ? Or facebook ? How do people have been accessing the internet for the last decade you reckon and how's this any different ?

> Can google be trusted ? Or facebook ?

No. Neither can be trusted as far as you can throw them. They're both incredibly invasive data brokers. Their customer facing products are just vehicles to show ads and Hoover up more PII and behavioral data on everyone.

I routinely run into Claude Opus and Fable making basic mistakes like misunderstanding a simple negation, which would be on top of whatever reliability issues there are with a source. I think that means it is functionally very different.

Google (as it originally) existed wasn't a source of information, it was an *index* of it. You were trusting google as a source for {search_query}, you were trusting the links it gave as a source (based on your own evaluation). LLMs are fundamentally different, because you are trusting the software to actually generate the information in a truthful way.

If you use google (sans AI), you're putting some trust in their page ranking algorithm. If you use it with AI, you're trusting the same algorithm (since that's how the model gets it's sources), but then you're trusting the model to evaluate the sources for credibility and extract the information you actually want.

Pagerank is just a popularity contest, doesn't say anything about trust. AI probably does a better job determining trustworthyness than pagerank.

Do you think the LLM reads every page on the internet before generating your answer? Of course not. What happens is that it use some sort of ranking algorithm to pick the pages that are most likely to answer your query and reads *them* (at best. At worst it just makes something up). You aren't avoiding the problems with ranking algorithms by asking an LLM, you're taking all of those problems, adding more problems on top, and pretending that this is somehow better.

> Do you think the LLM reads every page on the internet before generating your answer? Of course not.

The way LLMs are trained the answer is of course yes, but they don't remember them all exactly, of course.

Well then ask the LLM to go find and pull its information from primary sources. Don’t ever rely on its own training data.

100%

Reply on news.ycombinator.com