Quora ruined Quora. And it started well before the advance of LLMs.
Starting with the horrific UI where you don't know which question is the answer you're reading for
AI Ruined Quora
91–100 of 100 posts
Re: AI Ruined Quora
#92This story is really interesting for all the not-specific-to-quora parts. It’s pretty obvious that quora has a strange culture of know-it-alls trying to show off, but this AI product makes it clear that the act of showing off is the product . The answers are garbage and they obviously don’t care. The key action for sale is not finding an answer, but writing it. What’s really interesting is that Quora product team dec…
>look at all the top-level comments on this HN post with no engagement, by users who have repeated the same pattern discussed about quora. So responding to a post but not getting responses is the same as responding to slop with a financial incentive? I'm not sure I fully follow. I only very slightly agree in the fact that I rarely make top level comments unless I feel I have something unique to share. And I know that…
> So responding to a post but not getting responses is the same as responding to slop with a financial incentive? I'm not sure I fully follow
HN doesn’t have financial (?) incentives (but does Quora?). The point was mostly about the quality of content, the type of content, etc. I don’t mean to criticize anyone, I think it’s human behavior.
The root question of my claim is: When you create a (top-level or other) comment, how do you measure if it “contributes to the conversation”?
Most comments probably have zero engagement, while the top few comments drive the conversation. Part of that is the UI driving people towards the top. My claim is that most commenters, just like most quora users, are answering out of their own need to speak. I think the XKCD link describes the behavior well - and my claim is that it also includes sharing personal anecdotes and opinions and similar, not just “correcting misinformation”.
I have my own share of comments with no upvotes and no replies, which I can only assume means no one found it to contribute. I mean no disrespect towards others in similar positions. But does its presence in the database for casual passersby’s mean anything despite that? Despite the goals and intentions, does it matter? Should it just be routed to dev/null to silently fill my desire to speak? Should an LLM just entertain me with artificial conversation for entertainment?
Re: AI Ruined Quora
#93Earlier quoted context omitted.
It's the UFOs. They land anytime soon :DDD.
I know that was a joke, but the reality is almost as creepy. I worked for a corporation that drove marketing terminology to use their glossary, and saw the thinking firsthand. It's not paranoia. In fact, it makes perfect sense, and may not even be that underhanded. They submit huge reams of text, interspersed with their talking points, to any and all publications, news outlets, whatever. Some percentage of that gets…
I'm telling you it's the ufos. Phase shifting and other shit. It's real.
Re: AI Ruined Quora
#94Earlier quoted context omitted.
Quora was bad when I visited it in 2021. So what year it was good again?
Early 2010s. Basically around its founding. This is an article [1] from 14 years ago where Arrington characterized Quora with “People say they feel smarter after they use Quora” and “It’s kind of like Wikipedia [2]. This is the long tail of information” 1: https://techcrunch.com/2010/03/28/quora-has-the-magic-benchm... 2: https://techcrunch.com/2011/01/30/quora-is-really-about-a-be...
> Arrington characterized Quora with “People say they feel smarter after they use Quora”
But… they still do; as the article says, and it’s completely correct, it is essentially selling _feeling smart_. That’s not the same as _being_ smarter, though. I’m not sure there was ever _really_ a golden age; it wasn’t as bad as it is today on launch, certainly, but it was never great.
> “It’s kind of like Wikipedia […]”
Except without Wikipedia’s editing process, which is relatively good at keeping out incorrect things written by internet knowitalls. It _always_, right from the start, had a big problem with the confidently incorrect.
Re: AI Ruined Quora
#95The post shows limited understanding of what LLMs are. Ok, so Quora has a bad model that generates "Why doesn't Donald Trump shave his mustache?" type questions. That's Quora's problem. But in general, a frontier LLMs like the ones trained by Open AI would not make such silly mistakes in their output. Its "fancy autocomplete" is trained on millions of documents that include detailed description of Trump's face and th…
> But in general, a frontier LLMs like the ones trained by Open AI would not make such silly mistakes in their output.
This is beginning to wear a bit thin. Every public application of AI produces obvious nonsense, and this doesn’t appear to be changing. The inevitable defence is “oh, they used GPT3.5 instead of GPT4/oh, they used GPT4, but not the version that was only publicly available for three weeks in 2023/oh, they used precisely the right GPT4 but, when sacrificing the goat while petitioning Roko’s Basilisk for useful output, they used the wrong sort of knife”.
Apparently, literally no-one can use these things properly, at least for publicly visible applications. Either that, or they’re just a bit shit.
Re: AI Ruined Quora
#96I always think people calling AI just ‘fancy autocomplete‘ haven’t really tried it yet.
I see LLM's as next-generation search. So much of everything created on the internet is useless, outdated, false or just irrelevant to the thing you want answered. It gets increasingly harder to find what you need to get your job done. A LLM can extract relevant information from the garbage pile faster than I can sift through various forums and mail threads via Google. The more I use it the more I feel that it's a gr…
> For coding it's easy, the code works or it doesn't.
Oh, wow. That’s a quote that could be written on a lot of figurative gravestones. Also some literal ones; see THERAC-25.
Re: AI Ruined Quora
#97I always think people calling AI just ‘fancy autocomplete‘ haven’t really tried it yet.
But that's literally what it is. The only reason you can have dialog-like interactions with language models is because they have been trained with special "stop tokens" surrounding dialog, so the model can (generally) auto complete something that looks like a reasons, and then the inference engine can stop producing text when the model produces the stop token.
Technically it is, of course, but the experience is completely different, and I get the feeling people call it that to downplay it.
Re: AI Ruined Quora
#98I wonder if it would really come down to AI if you did the user research, maybe focused on the “onboarding” process. I remember a time back when you would get a Quora result on a search query and you could click on it be find a few answers with comments on them to help you determine their merits. This is completely anecdotal, but these days I honestly can’t even find the answer if I click on a Quora result on a searc…
Yeah, I've noticed this too: seeing a Quora question that's exactly what I'm looking for, clicking it, and not actually finding the question which the SERP said was there. At least this used to happen back when I actually got Quora results, I'm not sure if its search ranking has dropped or if Quora results get caught by my banner blindness these days. There are some sites which will generate a bunch of pages which sh…
For some insane reason, Quora's default view is to show you "answers to related questions".
If you click the dropdown box in the upper left, you can change from "All related" to "Answers". That should turn up the answers that the search engine was seeing.
I can generally give you at least some explanation for why Quora behaves the way it does, but I'm afraid I just toss my hands up at this one.
Re: AI Ruined Quora
#99This story is really interesting for all the not-specific-to-quora parts. It’s pretty obvious that quora has a strange culture of know-it-alls trying to show off, but this AI product makes it clear that the act of showing off is the product . The answers are garbage and they obviously don’t care. The key action for sale is not finding an answer, but writing it. What’s really interesting is that Quora product team dec…
Quora knows that it's about attracting people to write answers, more than people with questions. They've had a number of tactics to get more questions so that people could write answers.
At one point, they paid people to do it -- which went about as well as you would expect. Users complained like crazy and it lasted long, long after it was an obvious failure.
Their idea wasn't completely insane. There are questions that people ask Google that they'd never ask a human. Stupid questions that people will nonetheless answer. And Quora wanted to be the place that Google directed them. Combined with a high reputation from its early days, it was well placed in SEO.
They never figured out how to turn that into money, and they've floundered about. They managed to have a bunch of smart people wanting to show off -- actually giving meaningful answers -- and they squandered it.
Re: AI Ruined Quora
#100Earlier quoted context omitted.
But that's literally what it is. The only reason you can have dialog-like interactions with language models is because they have been trained with special "stop tokens" surrounding dialog, so the model can (generally) auto complete something that looks like a reasons, and then the inference engine can stop producing text when the model produces the stop token.
Technically it is, of course, but the experience is completely different, and I get the feeling people call it that to downplay it.
I think understanding that helps me get more out of them. I feel like I am better able to provide information to the model with the expectation that it will need that information to autocomplete the dialog that I want.