Reporters' Roundtable Ep. 121: Wavii founder on the future of newsAdrian Aoun, the founder of the news-reading and summarizing app Wavii, talks with Rafe Needleman about the genesis of his technology and how far it can go.
Hey guys -- Rick Adelman in San Francisco to reporters' roundtable. -- last week I covered a really interesting company called wade VW a VII. And I just I thought the company was fascinating -- it is it has semantic analysis. Product that takes. News of the world all the news is coming in. And gives you a FaceBook like feed of what businesses and people and events -- -- are happening with things there are going on. Its own so bought this -- so. Company was acquired by such and such. End this system that does it is. As far as I can tell magic it's a computer that reads the news and -- -- -- for you. And I just think that's fascinating and because I think it's fascinating I have the CEO of the company. Adrian alone here with us to talk about his business wavy and what it's doing why it's doing it in the future of computers that read the news you don't have to. Adrian thanks -- -- the time to join -- roundtable. -- So give a -- a brief. Pitcher of what this business is it. Is and then let's get into that the science in the future here what is wavy and why does it exist. I checked so you can kind of hard to understand -- -- is creating context -- something that we all know what to space. And what FaceBook kind of guys. -- -- -- -- -- -- -- -- -- -- -- 32 visual updates about what's going on in the world right like Bob checked into a location and you know Jim Lee has a new job. Eric is now dating someone. And those nice visual updates are just really really I like to consume write their fast I get to decide what I -- -- set exceeding. You -- photos. I get to sit there and say hey I wanna click into these photos and learn a little more. Or -- get to say I'm not that insisted well let's say you start dating someone new like it to click on her and -- find out more about where we're going on common winter where she worked. So basic -- gives us is really nice interface of exploring the information about our friends. And it's it's incredibly good at keeping us up to date with all of our friends are -- in a few minutes a day you can keep up to date on -- Thousand people 15102000. People and so we based that we -- one of that same product. But for the world right I just wanna keep -- date with everything that's going on in the world in these nice little visual updates right in kind of click around and explore. And so that's what we set out to do the challenge there is that if you think about it let's say -- get it might be. -- is dating -- year you know -- I had an earlier Wright's name. Com and maybe get some photos that you instantly below it's the -- FaceBook was able to give me that is that. You know -- -- build out our relationship status and your wife -- confirmed dead and maybe you know what your friends as they are taking some photos maybe it's one of your other friends at home and -- see these photos. Act that it's art it's all very structured -- -- you're talking point structure to others. Exactly. -- the trick for us is. Unfortunately. That's structure isn't being provided by users for everything going on in the -- right satellite when -- but -- -- -- branded and changed our relationship status. -- -- when Whitney Houston died in chat and record that you know there's there's basically tons of data out there. -- that's not in a format or understand it's in a tweet it's in a blog post its in an article maybe it's in a YouTube video and so what we're doing is we're teaching the computer to -- -- read and understand that. Much the same way human led and then build these feed items that kind of -- what bird. Sounds. And and the product then is like a feed of the news now. Why do we need disseminate it in my field in technology. I read news.com of course I -- a couple of blogs and RSS feeds and I look at tech mean. Why do I need something that is giving the kind of a timeline view of the news when the news is all -- in front -- me anyway. Yet so the first aspect is hard to get through a lot. -- so odd when trees and guidance program. -- might -- you know 2000 times writes in -- assassins. Being -- -- I get from places and so what happens is for hard to seem that. Maybe there's different fragments to -- over in this article talks about. The purchase price in this article that had been made in this article talks a little about the background so you really kind of -- almost. -- single view a page -- DL have been cold out. -- -- second that. Aspects -- which it. Things it -- you care about that aren't -- so large and maybe your local ice cream shop -- new labor and you know what the local Paper hibernate -- sentences well that sort of stuff does it bear spray in it to an extent we also want circuits that information that people -- but then there's a third aspect which is putting things in contacts -- So I'd mention you're eating Mike and click on that person on the connect angle find out about her and when you're reading -- news article today you just don't have that ability to really get contacts or analyze and from each other really you only get kind of a -- you point it at each -- Think of something just really really simple and that's all right well let's say I want to know when Apple's release next I am pretty sure everybody wants -- -- -- so. So what I do like to read your article battery -- read something on -- some other sites. But wouldn't be great I just had some visualization -- it look fourteen sites said it's coming out in April. Hindsight said -- -- and here's the timeline. Of when they released and historically. An even better would be okay. Typically these subsites. After one and look at what they're saying there -- -- team I want to focus on. So we believe that kind of keeping users access to more information. And kind of visualizing that information. Is is a very compelling experience -- I don't think it replaces your news.com. I don't think it replaces your -- I think it ended a legal -- today at many different sites and that's okay. -- we just want to provide. And it. On sediment formation and experiences that users aren't getting outstripped that are out. So. When I saw this product I was very impressed by the demo -- I -- -- And I've been using an and it's still very impressive technology what it does is it looks -- all this information based -- upset that -- say you're interest and animal talk about that later. Ended this deals. The headlines. In some cases and better headlines than the original articles and puts them in front of you. That seems. Fairly magical -- -- how do you do that how do you figure out that. A story written by -- and are terrible rush. Is actually about such and such buying so and so or Google mail being down or something like that. -- cell I don't have my wizardry degrees so it's not actually magic though it would probably didn't ask it pretty lines. -- -- So think about stink about how odd children letter -- -- think about how you are. May be may be kind of one -- eight year your father comes calming sense I want -- -- watts. And then meaty you know later that a year other arms on the -- -- really anonymous. And so what you're doing is a little kid it's your kind -- hearing tons of language and then you're starting to discern patterns into the pattern in your hearing. And I want something I want something in your you're kind of get it. Figure out what to use media asks. Indicated we'll go upset. And what doesn't mean I want something and then -- can learn what wanting unique. And now think it is kind of like because he's learned the pattern. It can go ahead and can try and solve. It well I want to sandlot or Taiwan escalate tickets kind of glory that concept now -- -- -- -- well this is how we train our systems. So there. Our system is looking at all language you know that's a big problem just kind of ingesting tons of content -- -- And then our system using machine learning to discern patterns that will a lot of these things they all look the scene and -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- That means you now add this factor -- to this actors are igniting its response. And got it goes -- -- and it's probably guessing. Correctly most of the time. And then -- techniques and -- like it -- Barack Obama. And is engaged at edit word on when Lilly Lilly -- -- app and may be what it's solid rock Obama was engaged in its debate. And so what'll happen there is will help the system now it'll figure out what's -- machine learning -- figure out that. Being engaged to someone and being engaged in -- with someone are kind of two different things and so it's kind of -- -- its knowledge over. Looks like a little it is where little hit me say I want to happen. You'll correct that little snow on object -- can get better and better over well that's what our system it's getting better and better. -- people have had theories about this for com. Probably hundreds of years that that linguistics about meaning but that's the deconstructing language. And a human beings of course I even you know my five year old has -- before you know when he is learning a language. Has trillions of of neurons and enter connections and our brains -- fundamentally differently from the way. Binary computers work no matter what the programming is. There has got to be precedent here that you have and and and pass that scientists and linguists have gone down -- -- -- wrong and pass that have been right. What are you basing your technology on. -- -- the first and foremost thing to understand is that at the end of the day we are adults our computer. Our computer at the end of the day. It's running some math where it's recognizing patterns representing at and now I'm a ticket to the compliment -- you know a lot of people look at our product and they're anonymous -- -- on -- and it really is almost artificial intelligence but it's not it's really just first. A long long long long out of really building -- artificial intelligence now what people have tried to do in the past when teaching computers language and this is somewhat a generalization. But what they've tried to do is basically teach -- computer what you learn when -- in third grade fourth -- it's great which is around is teaching the computer though the rules of grammar so this is subject this is a -- this is an object this is a artists all you remember like all all that it -- -- as they eat up all that it -- you learn what we found. Is that it doesn't really make sense to try teaching language that way because that's not really how humans I -- think about it at three year old -- when -- when your child history -- usually speaking but at the end of the day that three year old doesn't know -- -- purposes or at least they -- they're smarter than I was candidates -- and so what we found was that. Those don't the grammar rules that we learn in school are very much an attempt -- retrofitting rules on top of language. And much the same way where you didn't learn all language in kind of your third fourth and that's great class and and by the time -- done -- -- -- -- -- saying well I don't know. A 100% of everything you're actually learning concept at a time right I'm teaching -- concept right now about. You know machine learning -- artificial intelligence and may be later today you're gonna teach any concept about journal race -- and so. These concerts that we learn. It's a never ending process we're gonna learn them until the you know until the day we are and that's how we decided to teach our computer rather than just trying to. Train all the schools up crimes once and -- You you come to this field. Honestly I believe I'd tell us about your your upbringing and in the lineage of what you've been doing here. Back to some of the great linguists -- of -- time. Yet so -- And kind of surrounded by linguistics it's that -- excellent site to keep psych out poking me since I was -- since I was really young -- as mine my -- Patents he studied at -- undertones. And aren't so I was constantly surrounded by my father and his friends and I'm one of the things that created -- -- -- -- current debate where their cards link honestly arguing about this language structure that language structure that happen is when I was growing up we just sit around the dinner table or you know my dad written by his friends over for drinks and they be arguing on. Various kind of -- exit of these rules which little is correct is that this ruler of this -- which covers more cases. -- -- -- They would often hide it on her knee as the -- -- it will which do you think is right because there's a degree to which our intuition and need an understanding of language option. More accurate and up over analyze and now I never wanted to be -- -- -- next logo on -- mailing list but. What with that experience -- -- with that language you know I spent a lot of time. It would taught -- that link which just isn't that hard and by that I don't mean that humans can't speak with incur. What I mean is that. We all learned it fairly intuitively. And so what that tells me it's like -- two year -- and speak we can at least get the computer to understand. Language at. At the two year old level -- year old level -- I don't think we're we're gonna do it at the fifty year old local. It certainly top mean that that language has some core elements that are fairly simple and combine that -- might focus more recently on trying to unlock meaning on the web. You know for better or for worse he only -- -- in the couldn't get -- think about the experience at least it gives you. It unlocked all the meaning about your friends -- can figure out which you know which my friends' lives in Seattle consider what's my friends. Work at this company which has been to a restaurant dating books they've really unlocked all of the information about my friends but if we want that same thing the Internet. All of the information on the Internet. For better for worse is in naturally try wanted to cooler all the celebrities and he lives two -- -- announcing a -- let's pretend. And how can I get that well. Any got to think that pretzels and working all day long -- that contents on the let -- in tweets and blog post. I wanna know what are all the series evaluations. That content is all on the wet. But the computer can't get that it's because the computer doesn't understands. Natural language so from our perspective. That was kind of the cost of doing business -- wanna build this product we teach computer to understand language. And for better or worse we didn't back down merely because we're kind of comfortable language on we've been surrounded -- it for awhile and -- wasn't. It wasn't that scary content to kind of tackle the problem. There are other. Things that signal importance the people in addition to the fact that you -- so -- doing pretty good job of extracting from unstructured via articles. And 01 of those signals is the social signal I -- if my friends read -- story. Then that's arguably important to me by Sam mentioned something by -- tweeting it on us on Twitter. Then that can be picked up as the signal. How does that play in two. The what the user of wavy c.s because one of the things about -- disposal site on here is. Then at the display of the -- page is not information -- in the same way that in New York Times front page or write tech mean page is where there's like. A hundred headlines. On -- you see a stream and its. Far fewer numbers of stories how do you decide. What the user -- -- -- so there's two aspects -- the deciding Helm. Which items should -- user and then there's second aspect of deciding how much to show the user and so deciding which items to show user. Really comes down to credited. I take three main aspects the first is kind of will we see in the world and by this I mean. How often are we seeing something and like how rapidly are we've seen it so we may see something -- for example odd -- Apple's release you know -- that Apple's releasing new iPhone we see it all the we don't -- very rapidly in one kind of expert right we see incredible all that's. It's time. -- and just occasionally. On so that's kind of the first -- that is what we think it is a world -- the second aspect that we get is just a priority knowledge of the. The concept we know that -- death. Is more -- stating that -- earth. Almost always we also know for example. That acquiring a company for billion dollars when you've only ever acquired companies and -- asked for tens of millions is actually really big deal. -- so. Because we have this underlying information we can place -- -- Kind of analysis games to really make sure surfacing relevant -- And the third aspect has to do it the user and their their world and so what this comes down to his. What had what is the user following irks -- right to argue following Barack Obama is -- something happening what Barack Obama. Well if it's something happening -- brought about -- pretty relevance. Or are you following. Michelle Obama but it's something happening with Barack Obama so -- may be slightly further away from your interest but it also might be relevant. Content or in the past when we shall new things about Barack Obama. Clicked on them or not. And then the course and you're discussing which is how much -- your friends engaging with its piece of content are your friends all like to commenting reading well. If so that's also gonna build heat in our system. So says that all goes into building a -- that decides. Hey what's the most important thing shall user now the second aspect is how much to -- the user. And you know do we get you ten items we -- -- hundred items we give you a thousand artists and this is a trade opt out. What's known as actors and recall the few worthy items to show you -- the lesser recall. Well pretty typically did better accuracy and by that I mean there are many things you care outs are are far more right to -- just showed you one item -- but I. -- the accuracy and really I -- likening her. -- is pretty low. -- you know there's tons of things you next exit this is the game more costly -- don't wanna get too much recall and edit in Hindi user what I did things they don't care -- but we also don't want them to nix things so what where what were doing. What exists in the products they will were also -- working on is basically trying to you look at how the user engages our contents. And figure out on a per user basis -- this user want more content as this user want -- -- and so if this is something that you see at -- but the more often you come to FaceBook the more stuff they get you in your stream -- in your. -- and we have this -- And finally. What is the role in your estimation as the reluctant linguists. The role for article writers who are putting their blood sweat and tears into crafting their work into gathering -- -- information gathering of the opinion and crafting together what is the role. Of the -- Yet so think of this kind of -- not like the readers. The first perspective is un. At the beginning at the very beginning like what what was a reporter -- there were reports. -- -- breaking the news. And our system. Needs our system -- that -- not looking to replace that right so. The first kind of -- you people that rake -- news that's particularly interesting arts. Now the fact that right now many many many -- articles printed on the web are just -- duplicates. In terms of the actual what happened. -- -- So we'd like to take all that you know -- fifty articles that say it is the but it's emerged that -- I don't want it. One -- -- but what is interesting let's think about the second. When your recording something rate may need to incorporate it. But what do you do when you're adding your analysis Europe in Europe -- -- And that's extremely valuable -- iTunes and it fifteen people adding their union analysis as it. Is still something that we still wants -- -- you know worldwide we don't want that we want people to do so we'd love to do is. In are central to our system we say -- -- race ops if you click on. Hey now I died in rates article kind of learn a little more about it now there's a third aspects -- And this has to deal with the fact that we Austin. -- -- Approximate our interest by its source is I did not adequate -- what I mean by that it. -- interest in -- I don't necessarily know what. Student so instead I follow what -- rights and rate is good it kind of introduced me to things it's almost a -- -- is experience that discovery. And we think that's extremely valuable to and that's why it over time we want people to kind of have. But we want our users -- as much control as they once -- if they want to follow. The recorder -- if they want to follow the company ordered a product or you know the story namely acquisition. Great -- we basically want to give users edit the -- control over that and let them define their experience that makes its. Okay hey Adrian thank you so much -- -- a time for us today. Very -- product you guys have to check out -- WA VI IE dot com is also mobile app for. And I just wanna point out for people who are have gotten this far this is a very interest in -- in from business perspective the news aggregators site was acquired by Ian and power said it was acquired the search companies -- -- was acquired by Microsoft. I don't know what's gonna happen wavy but -- get it while you still can and still independence is a really fascinating company to watch Adrian that Lothian thanks for the time. Thank you demonstrate there -- really appreciate it. By.