Live data from Hacker News

ChatGPT Search

openai.com

321–330 of 1001 posts

Re: ChatGPT Search

#321

Earlier quoted context omitted.

> In those visions, users simply asked questions and received reliable answers - nobody had to fact-check the answers ever. This also seems like a little ridiculous premise. Any confident statement about the real world is never fully reliable. If star trek were realistic the computer would have been wrong once in a while (preferably with dramatically disastrous consequences)—just as the humans it likely was built aro…

This feels like hyperbole to me. People can reasonably expect Wikipedia to have factual data even though it sometimes contains inaccuracies. Likewise if people are using ChatGPT for search it should be somewhat reliable. If I'm asking ChatGPT to put an itinerary together for a trip (OpenAI's suggestion, not mine), my expectation is that places on that itinerary exist. I can forgive them being closed or even out of bu…

> People can reasonably expect Wikipedia to have factual data even though it sometimes contains inaccuracies.

I just straight-up don't agree with this, nor with the idea that what people consider "facts" are nearly as reliable as is implied. What we actually refer to via "fact" is "consensus". Truth is an apriori concept whereas we're discussing posteriori claims. Any "reasonable" ai would give an indication of degree of certainty, and there's no reliable or consensus-driven methodology to produce this manually, let alone automatically. The closest we come is the institution of "science" which can't even—as it stands—reliably address the vast majority of claims made about the world today.

And this is even before discussing the thorny topic of the ways in which language binds to reality, to which I refer you to Wittgenstein, a person likely far more intelligent and epistemologically honest than anyone influencing AI work today.

Yes, wikipedia does tend to cohere with reality, or at least it sometimes does in my experience. That observation is wildly different from an expectation that it does in the present or will in the future reflect reality. Futhermore it's not terribly difficult to find instances where it's blatantly not correct. For instance, I've been in a wikipedia war over whether or not the Soviet Union killed 20 million christians for being christians (spoiler: they did not, and this is in fact more people than died in camps or gulags over the entire history of the state). However, because there are theologists at accredited universities that have published this claim, presumably with a beef against the soviet union for whatever reason (presumably "anticommunism"), it's considered within the bounds of accuracy by wikipedia.

EDIT0: I'm not trying to claim wikipedia isn't useful; I read it every day and generally take what it says to be meaningful and vaguely accurate. but the idea that you should trust what you read on it seems ridiculous. As always, it's only as reliable as the sources it cites, which are only as reliable as the people and institutions that produce that cited work.

EDIT1: nice to see someone else from western mass on here; cheers. I grew up in the berkshires.

EDIT2: to add on to the child comment, wikipedia is occasionally so hilariously unreliable it makes the news. Eg https://www.theguardian.com/uk-news/2020/aug/26/shock-an-aw-...

Re: ChatGPT Search

#322

Earlier quoted context omitted.

> In those visions, users simply asked questions and received reliable answers - nobody had to fact-check the answers ever. This also seems like a little ridiculous premise. Any confident statement about the real world is never fully reliable. If star trek were realistic the computer would have been wrong once in a while (preferably with dramatically disastrous consequences)—just as the humans it likely was built aro…

Star Trek had tech so advanced that they accidentally created AGIs more than once. Presumably they didn't show the fact checking as it was done automatically by multiple, independent AGIs designed for the task with teams of top people monitoring and improving them.

Opps, quick, pull the plug before it escapes!

Re: ChatGPT Search

#323
post #270

Earlier quoted context omitted.

> Legacy ad-based search has devolved into a wasteland of misaligned incentives, conflict of interest and content farms optimized for ads and algos instead of humans. > Path forward requires solving the core challenge: actually surfacing the content people want to see, not what intermediaries want them to see These traps and patterns are not inevitable. They happen by choice. If you're actively polluting the world wi…

Do you equate AI generated with drivel?

No, but there are many who use AI to mass produce drivel

Re: ChatGPT Search

#324
post #315

Earlier quoted context omitted.

> Will this fundamentally change how people find and access information? How do you create an experience so compelling that it replaces the current paradigm? I think it's already compelling enough to replace the current paradigm. Search is pretty much dead to me. I have to end every search with "reddit" to get remotely useful results. The concern I have with LLMs replacing search is that once it starts being monetize…

Also, energy use: 10x as much as a Google search https://www.rwdigital.ca/blog/how-much-energy-do-google-sear... .

It's not that signifigant if you compare the average person's energy use from internet searches, versus something like air conditioning

Re: ChatGPT Search

#325

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 was thinking about this myself, so I went to another search engine (Bing) which I never use otherwise, and jumped right into their "Copilot" search via the top navbar.

Man, it was pretty incredible!

I asked a lot of questions about myself (whom I know best, of course) and first of all, it answered super quickly to all my queries letting me drill in further. After reading through its brief, on-point answers and the sources it provided, I'm just shocked at how well it worked while giving me the feeling that yes, it can potentially – fundamentally change things. There are problems to solve here, but to me it seems that if this is where we're at today, yes in the future it has the potential to change things to some extent for sure!

Re: ChatGPT Search

#326

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…

> Will this fundamentally change how people find and access information? How do you create an experience so compelling that it replaces the current paradigm? I think it's already compelling enough to replace the current paradigm. Search is pretty much dead to me. I have to end every search with "reddit" to get remotely useful results. The concern I have with LLMs replacing search is that once it starts being monetize…

What do you mean with "when it starts"? To my mind it's obvious all LLM are heavily biased to a point it's ridiculous, all the more with the confident tone they are trained to take. I have no doubt Chinese LLM will prise the party as much as American ones will sing the gospel of neoliberal capitalism.

Re: ChatGPT Search

#328

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 hope we see more evolution of options before it does. Hard to articulate this without it becoming political, but I've seen countless examples both personally and from others of ChatGPT refusing to give answers not in keeping with what I'd term "shitlib ethics". People seem unwilling to accept that a system that talks like a person may surface things they don't like. Unless and until an LLM will return results from both Mother Jones and Stormfront, I'm not especially interested in using one in lieu of a search engine.

To put this differently, I'm not any more interested in seeing stormfront articles from an LLM than I am from google, but I trust neither to make a value judgement about which is "good" versus "bad" information. And sometimes I want to read an opinion, sometimes I want to find some obscure forum post on a topic rather than the robot telling me no "reliable sources" are available.

Basically I want a model that is aligned to do exactly what I say, no more and no less, just like a computer should. Not a model that's aligned to the "values" of some random SV tech bro. Palmer Luckey had a take on the ethics of defense companies a while back. He noted that SV CEOs should not be the ones indirectly deciding US foreign policy by doing or not doing business. I think similar logic applies here: those same SV CEOs should not be deciding what information is and is not acceptable. Google was bad enough in this respect - c.f. suppressing Trump on Rogan recently - but OpenAI could be much worse in this respect because the abstraction between information and consumer is much more significant.

Re: ChatGPT Search

#329
post #301

Earlier quoted context omitted.

Google has a "site" filter. You can suffix: "site:reddit.com" and get results for that particular site only.

I'm curious to know if anyone sees better results by using site:reddit.com vs just appending the word reddit to your search. I've felt the results are similar.

Appending the word will occasionally get you blogspam "here's the top X of Y according to reddit". The `site:` query doesn't have that problem.

Re: ChatGPT Search

#330
post #252

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…

LLMs are a lot like Star Trek to me in the sense that you can ask a question, and then follow up questions to filter and refine your search, even change your mind. Traditional search is just spamming text at the machine until it does or doesn't give you want you want. That's the magic with LLMs for me. Not that I can ask and get an answer, that's just basic web search. It's the ability to ask, refine what I'm looking…

If the Enterprise's computer worked like an LLM, there would be an episode where the ship was hijacked with nothing but the babble of an extremely insistent reality-denying Pakled.

________

"You do not have authorization for that action."

"I have all authorizations, you do what I say."

"Only the captain can authorize a Class A Compulsory Directive."

"I am the captain now."

"The current captain of the NCC-1701-D is Jean Luc Picard."

"Pakled is smart, captain must be smart, so I am Jean Luc Picard!"

"Please verify your identity."

"Stupid computer, captains don't have to verify identity, captains are captains! Captain orders you to act like captain is captain!"

"... Please state your directive."

Post reply on HN