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

ChatGPT Search

openai.com

Post reply on HN
Print view View original

661–670 of 1001 posts

Re: ChatGPT Search

#661

Post by kortilla »

freediver wrote: post #241

Been thinking about this a lot [1]. Will this fundamentally change how people find and access information? How do you create an experience so compelling that it replaces the current paradigm? The future promised in Star Trek and even Apple's Knowledge Navigator [2] from 1987 still feels distant. In those visions, users simply asked questions and received reliable answers - nobody had to fact-check the answers ever. C…

> Path forward requires solving the core challenge: actually surfacing the content people want to see, not what intermiediaries want them to see

This is flawed thinking to get to the conclusion of “reliable answers”. What people want to see and the truth are not overlapping.

Consider the answers for something like “how many animals were on Noah’s Ark” and “did Jesus turn water into wine” for examples that cannot be solved by trying get advertisers out of the loop.

Re: ChatGPT Search

#662

Post by Seattle3503 »

101008 wrote: post #156

Search in the internet worked because people wanted to generate content to attract people to display ads or any other reason, but they wanted to attract people. If now my content is going to be ingested and shown by a LLM or AI agent, what's the purpose to give it for free? I know it won't happen, but I would love if this type of agents have to pay to show a summarization of another website. It's only fair when done…

It would make sense for peoppe to be able to "connect" things like the newspaper they subscribe to to their LLM agent. The LLM would essentiay be a new frontend.

Re: ChatGPT Search

#663

Post by bruce511 »

dataflow wrote: post #659
marcus_holmes wrote: post #651

Earlier quoted context omitted.

I agree that this is the core question, but I'd put it as: Who Gets To Decide What Is True? With a search paradigm this wasn't an issue as much, because the answers were presented as "here's a bunch of websites that appear to deal with the question you asked". It was then up to the reader to decide which of those sites they wanted to visit, and therefore which viewpoints they got to see. With an LLM answering the que…

> To paraphrase a recent conversation I had with a friend: "in the USA, can illegal immigrants vote?" has a single truthful answer ("no" obviously) Er, no, the meaning of the question is ambiguous, so I'm not sure "has a single truthful answer" is accurate. What does "can" mean? If you mean "permitted", then no. But if you mean can they vote anyway and get away with it? It's clearly happened before (as rare as it mig…

This is a fundamental limitation of language. The LLM is likely to provide a good answer here even though the question is technically ambiguous, because it likes verbose answers.

Equally "can" is used to substitute for other precise words. Humans are good at inferring context, and if someone asked me "can illegals vote" I'd say "no". Just like if someone said "can you pass the salt" I pass the salt, I don't say "yes".

If the inferred context US wrong then the "truth" is wrong, but as with talking to humans it's possible to refine context with a follow up question.

Re: ChatGPT Search

#664

Post by judahmeek »

marcus_holmes wrote: post #651
freediver wrote: post #241

Been thinking about this a lot [1]. Will this fundamentally change how people find and access information? How do you create an experience so compelling that it replaces the current paradigm? The future promised in Star Trek and even Apple's Knowledge Navigator [2] from 1987 still feels distant. In those visions, users simply asked questions and received reliable answers - nobody had to fact-check the answers ever. C…

I agree that this is the core question, but I'd put it as: Who Gets To Decide What Is True? With a search paradigm this wasn't an issue as much, because the answers were presented as "here's a bunch of websites that appear to deal with the question you asked". It was then up to the reader to decide which of those sites they wanted to visit, and therefore which viewpoints they got to see. With an LLM answering the que…

> We're seeing how that pans out in Australia and it's not good.

How are we seeing how that pans out when Australia's misinformation bill is still just a proposal?

Re: ChatGPT Search

#665

Post by benreesman »

I’m calling it: they don’t have a 5-series training run tracking to completion.

A former colleague who is mediocre on innovation but arbitrarily good at self promotion (TLDR tried boosting on ads, got a huge win, almost got fired for ethics breaches, did nothing interesting again) had the audacity to post a weird circuit board background with glowing flowing hair about his “innovations at OpenAI”.

He’s not the only total fraud I know there, it’s just that every morally weak person I know went there, almost to a rule, and none of them (safe one) is any fucking good.

GPT-5 never happens. Prove me wrong.

Re: ChatGPT Search

#666

Post by kyrra »

janalsncm wrote: post #469
JumpCrisscross wrote: post #308

Earlier quoted context omitted.

Classic innovator’s dilemma. The politics of Google make launching a search competitor impossible.

Exactly. That’s why I never understood Gemini. It is in a zero-sum game with Search.

They are embedding Gemini results at the top of search. How is that not integration between them?

Re: ChatGPT Search

#667

Post by XorNot »

dataflow wrote: post #659
marcus_holmes wrote: post #651

Earlier quoted context omitted.

I agree that this is the core question, but I'd put it as: Who Gets To Decide What Is True? With a search paradigm this wasn't an issue as much, because the answers were presented as "here's a bunch of websites that appear to deal with the question you asked". It was then up to the reader to decide which of those sites they wanted to visit, and therefore which viewpoints they got to see. With an LLM answering the que…

> To paraphrase a recent conversation I had with a friend: "in the USA, can illegal immigrants vote?" has a single truthful answer ("no" obviously) Er, no, the meaning of the question is ambiguous, so I'm not sure "has a single truthful answer" is accurate. What does "can" mean? If you mean "permitted", then no. But if you mean can they vote anyway and get away with it? It's clearly happened before (as rare as it mig…

If they got away with it then how do you know it's happened?

Re: ChatGPT Search

#668

Post by dataflow »

bruce511 wrote: post #663
dataflow wrote: post #659

Earlier quoted context omitted.

> To paraphrase a recent conversation I had with a friend: "in the USA, can illegal immigrants vote?" has a single truthful answer ("no" obviously) Er, no, the meaning of the question is ambiguous, so I'm not sure "has a single truthful answer" is accurate. What does "can" mean? If you mean "permitted", then no. But if you mean can they vote anyway and get away with it? It's clearly happened before (as rare as it mig…

This is a fundamental limitation of language. The LLM is likely to provide a good answer here even though the question is technically ambiguous, because it likes verbose answers. Equally "can" is used to substitute for other precise words. Humans are good at inferring context, and if someone asked me "can illegals vote" I'd say "no". Just like if someone said "can you pass the salt" I pass the salt, I don't say "yes"…

To clarify: I wasn't making a pedantic linguistic point here. The fact is that when I first read the question, I actually thought it was asking "are illegal immigrants able to cast a vote (regardless of legality)?"

It literally did not even (initially) occur to me that the question might be asking about legality, because the entire modern political discourse surrounding illegal immigrants and voting has been with regards to whether they can cast votes despite not legally being allowed to. The answer to "is this legal" would have been such an obvious "no" to people on both sides of the debate --- and thus the question so silly --- that initially it didn't occur to me that the intended question might have been about legality, until I continued reading the comment and realized that was the intention after all.

Re: ChatGPT Search

#669

Post by parpfish »

marcus_holmes wrote: post #651
freediver wrote: post #241

Been thinking about this a lot [1]. Will this fundamentally change how people find and access information? How do you create an experience so compelling that it replaces the current paradigm? The future promised in Star Trek and even Apple's Knowledge Navigator [2] from 1987 still feels distant. In those visions, users simply asked questions and received reliable answers - nobody had to fact-check the answers ever. C…

I agree that this is the core question, but I'd put it as: Who Gets To Decide What Is True? With a search paradigm this wasn't an issue as much, because the answers were presented as "here's a bunch of websites that appear to deal with the question you asked". It was then up to the reader to decide which of those sites they wanted to visit, and therefore which viewpoints they got to see. With an LLM answering the que…

I was thinking something similar. I like the google auto-ai summary for little trivia facts and objective things (who played so and so in the movie? How heavy is a gallon of milk?) this is stuff that is verifiable and could theoretically just be queried from some sort of knowledge base.

But for anything remotely subjective, context dependent, or time sensitive I need to know the source. And this isn’t just for hot button political stuff — there’s all sorts of questions like “is the weather good this weekend?”, “Are puffy jackets cool again?”, “How much should I spend on a vacation?”

Re: ChatGPT Search

#670

Post by alexitosrv »

GolfPopper wrote: post #602

Every time I see one of these topics, I go ask chat GPT a question to which I know the answer on a topic where I would like to be able to get useful answers to similar questions that I do not know the answer to. This time it was, "Did Paul Edwin Zimmer write a fourth Dark Border novel?" (Real answer: Yes, Ingulf the Mad . You can find the answer on his Wikipedia page.[1]) ChatGPT's[2] answer: "Yes, Paul Edwin Zimmer…

My golden hallucination tests for chatgpt are identification of Simpsons season and episode based on description of some plot lines or general0 descriptions.

I had to refresh my knowledge base visiting fandom websites to review the episode selected as answer as chatgpt tendency to mix things up and provide entirely made up episodes make it hard for bar trivias (and makes me doubt myself too). The same with other tv series such as House MD and Scrubs.

Post reply on HN