The problem with the idea of "an internet on which the best answers to the majority of our queries come not from the vast, increasingly noisy expanses of the world-wide-web but from the concentrated knowledge and experience of its most articulate experts" doesn't exist on any sort of large scale. A site that gets you in conversation with intelligent insiders and hooks them in with upvotes used to be called Reddit. We see what happens to that on a large scale (which, in Quora's case, doesn't even seem to be in the cards).
Google and Wikipedia work because they don't limit who has a voice (initially, of course). They gather every contribution, and then make a decision. That way they can be both more comprehensive and often better sources of information than some guy who has a bit of cachet in a community.
Quora has the worst karma-whorage of any site out there and its more about who answers the questions rather than what they say. This has killed its credibility in my mind. (Contrast this with Stack Overflow, which is about high quality answers and doesn't seem to suffer from whoring for karma)
I agree there's something to your observation. And yet, Quora doesn't show karma totals or leaderboards, while SO does.
How does Quora get more distracting status/attitude-competition when answers are only ranked against each other in the context of a single question – a tiny subset of the gamesmanship-encouragement that SO implements?
Is it the greater subjectivity of Quora's topics? The vague pretentiousness of its seed community? Or something else?
IMO it has to do with the fact that Quora forces you to state your real name and credentials when creating an account. Although certain people will feel this is the way to ensure answers are given by the best possible authority on a subject, it actually hinders the meritocracy of user-driven content. Uneducated but well-informed Quora users may feel it's not worth their time to contribute, because they'll just be overshadowed by someone with better credentials. In essence, it's like paying for a game's DLC instead of putting in the time to earn it on your own, cheapening the experience for all.
I think the comparison should really be between Quora and the whole Stack Exchange family of sites.
I think there are two main reasons many of the areas in Quora are useless. First, Quora never got commitment from a group of verifiable experts to answer questions in an area before opening. Its a chicken and egg problem, experts won't join if there aren't already experts. SE goes halfway there by ensuring there is a community of people to use each particular area the site before bringing it online.
Second, question askers on Quora are lazy. They don't explain what they have tried to do to learn about the answer already. They often don't know things that could have easily been found on google in a few minutes. Nobody wants to spend more than a fraction of the time on an answer than the asker has spent looking for an answer. So when an expert goes on the site they see a community of people who just want to get their knowledge without putting in any work themselves. QA sites will never replace search (as many of the Quora fans claim) for this reason.
On SO this isn't as much of a problem. First,because its very hard to find answers to programming problems with search . Second, question askers have to explain carefully what they have tried in order to expect answers. However, on some other SE sites this is just as much of a problem as on Quora.
I think its because Quora is a tight-knit community that is not (well) aimed at a wide public audience. Increasing status can lead to tangible benefits. If you are popular within the community, its essentially networking which can lead to investment, job and other valuable opportunities.
SO on the other hand, is more oriented at helping people coming in through search engines. (I would like to say it has better answers because its a more hacker oriented community that reflects open-source communities, but that would probably be an over-generalization. probably...)
It is limited to the 10 styles of commands and it from what I have read it cant interpret different grammatical forms of similar search contexts.
I have written software that can do the barebones of what Siri can do (www.samir-ahmed.com/iris.html). It is far less efficient flexible and polished. But I have 6 months of software experience and Siri has 8 years of DARPA quality experience.
I dont think that google is going to suffer to much unless
A - Siri can integrate well with other apps so that app developers can contribute to its grammar and its ability to interface with twitter, foursquare, facebook, open table etc.
This is open the floodgates for Siri to basically be the ultimate iPhone utility, allowing people to use it to interface with their entire phone in a new way
B - Siri starts to get smarter - and I mean creepy smart. When your wife sends you an email, Siri needs to read it. Understand what your wife is like, and store that information. It needs to do this with every contact so that it has context. This will require an immense amount of machine learning and natural language processing.
What this will do however, is open the doors for a variety of new applications.
You can data mine with Siri
E.g -"What did my wife ask me to pick up from the supermarket"
When you query your contacts, it can make smart recommendations
E.g -"Remind me to buy my wife a present"
Siri can look into your correspondence and make recommendations.
Being able to do all these things will make Siri an order of magnitude more useful, open to door to advertising revenue too and kick Google in the balls.
Until Siri can do these things, Siri is will not be the future of search.
Samir, I read through your site and the sample video and it looks quite impressive. Put a good-looking front-end to it and I don't know how anyone would differentiate it from Siri, except for some differences in what API 'verbs' the system is capable (Siri knows how to manage my calendar, but Iris knows how to get content from YouTube, for example). Iris even knows how to concisely get some info that Siri doesn't perform well with; for example, it gives a concise summary in response to a question about who a particular soccer player is, whereas Siri would probably just punt me to the web or at best pull up a long WolframAlpha response.
Anyone know why Siri seems only as capable (and in some respects less so) than this 4 week project done by a junior in college? Or (besides the UI work) is there something majorly different?
My understanding is that the speech recognition isn't special with Siri, Apple just licensed tech from Nuance, known for their Dragon Dictation product line.
If TC had any reliable source on a $1 billion Google offer for Quora, wouldn't that morsel have gotten it's own story, rather than being trickled out indirectly in a guest post?
Someone – maybe someone who'd like to make it true via self-fulfilling premature reporting – has been pushing the ill-sourced 'gossip' of such an offer since Quora's last funding round. Nicholas Carlson gave the gossip a more sober treatment back in February:
Just watch this video[1] of Siri asking for the user to choose Home or Work and then after the user, who has a heavily accented English, says "work" twice it offers up "walk" and "wall". There are only two possible answers, both his replies start with a "w", the first ends with a "k", and this "most sophisticate piece of AI to ever see the light of the consumer market" can't figure out he probably means work?
Edit: I work with, and know personally, a lot of geeks, hackers and researchers. I don't know anyone who uses Quora or thinks it's the key to anything.
He was sounding "walk" every time. Which still proves Siri has an excellent voice recog. And as long as that's true the rest can be changed by a product manager's instruction, when necessary.
I know it's not popular on Hacker News, but actually AI research has little to do with Lisp/the language of the moment or the Q&A website of the moment. It's to do with algorithms and core datasets. Watson being one example that's not just based on shuffling around email and setting alarms on your phone.[1] Andrew Ng's work at Stanford is another example of something that's feeding into really changing the world, not the HN stereotype of two ivy grads doing a ruby on rails app which gets an edge in a market dominated by a few lumbering megacorporations. Not that both aren't important, but there is a heavy bias on HN toward the latter, and new languages, rather than core research and algorithms.
This is interesting, but I'm not sure if it's really saying anything more than: Quora, as a knowledge-store, can power mobile search. I say mobile search, since that seems to be the implication of his dream of a siri-quora collaboration.
But is knowledge extraction from Quora really so game-changing with respect to the mobile web? Don't get me wrong--I love Quora and I think it's an incredible knowledge resource online.
But I think the power of siri comes from real-time knowledge (that is, knowledge or information that is useful to us while we're on the go). My experience on Quora has been so much more about intelligent, high-quality content than about real-time, practical, in-the-moment information.
The "future of search" is very intimately related to the "future of AI", and quite frankly, Quora has nothing to do with either, except that maybe it will be a decent source of answers to specific questions about startups...
As for the future of search and AI, the way I see things, there are eventually going to be two main AI camps that emerge (NB, I'm not saying that this is the state of the AI field right now, but I think it will segregate more along these lines in the next few years):
1) Big data AI: this is the Google style of AI, which usually assumes that the best way to do AI tasks is to use fairly naive algorithms and toss truckloads of data at them.
2) Deep model AI: this group is more about models, and believes that the best approach is to extract as much structure as possible from a limited data set. More data helps, to be sure, but the emphasis is more on getting as much as possible out of the data rather than getting more data.
The way things are going now, Google is going to win at big-data AI, no ifs, ands, or buts - they've got more data than just about anyone. So the way I see it, the only way someone is actually going to unseat Google is if it turns out that deep-model is a workable approach (thus far it hasn't had many serious successes), and Google doesn't focus on it early enough (because if they do, then their massive data availability will make sure they win anyways - that said, the folks like Peter Norvig driving the attention at Google are very explicitly in favor of data-based approaches, so I don't see Google leading the pack on model-centric research).
IMO there are some good reasons to think that deep-model approaches are viable: despite Google's massive billion-book data sets, humans are still better at doing, for instance, translation, despite the fact that a good translator might have only received the equivalent of maybe .1% of the input that Google leans on for its translation approach. The question is, will anyone actually figure out how to do it well? That's up in the air, but the fact that evolution figured out how to do it means that it's probably not terribly difficult, we just haven't looked in the right places yet.
My personal opinion is that the main pinch point in the typical big-data approach is that it's limited by statistical correctness, whereas human intelligence is not (we happily assume patterns exist in data even when we don't have enough data to make a proper statistical inference, and then we filter out incorrect assumptions later, also by using statistically incorrect heuristics and patterns - in other words, we're usually wrong, but sometimes we get lucky, and as long as we can eventually recognize that we're wrong, we do just fine). I think deep-model hopefuls would be wise to look more seriously at explicitly statistically unsound approaches if they want a shot at beating out big-data...
Deep model is progressing much more rapidly than the statistical card-trick algorithms, Boltzmann machines, and in specific Restricted Boltzmann machines already vastly outperform SVM (By which I mean to say both Structured and Support vector machines) in a number of hidden model discovery tasks related to CV and is taking a lot of the existing 'state of the art' backprop classifiers (classical Neural Net & k-means clustering) to town.
That said, traditional stochastic & markov driven approaches are more familiar and are cheaper to implement, which has hitherto driven the development of virtually all sciences.
Google and Wikipedia work because they don't limit who has a voice (initially, of course). They gather every contribution, and then make a decision. That way they can be both more comprehensive and often better sources of information than some guy who has a bit of cachet in a community.