 Thank you very much. That's a very friendly welcome here and thanks for like making this meet up here today possible Justin. So I think I just like get right into the project and I will show you a bit what we do and tell you about the background first and how we actually came to this project and where we are heading. So Susie what's it about? Well actually it's not just a chatbot and it is artificial intelligence for chatbots and the question is what is artificial intelligence at first right because everyone's like wow really they're gonna do that and yeah we realized actually there are a lot of different definitions for artificial intelligence out there and a simple one is actually that artificial intelligence can solve a problem. It will help you to solve a problem. So based on that we think we are able to do that and we thought a bit like okay what are these projects are doing like Siri Google Allo here Cortana and the song. What are they doing? Actually a lot of the things that these projects are doing is quite simple I think like play a song from iTunes. It's actually just a search and an activity yeah and so a lot of these things can be done with rules based queries yeah and that's what we thought. But actually what was the start and how we thought we do is Susie AI and I don't know how many of you have been to previous false Asia meetups but we had a project here that we introduced previously which was a lock lock and lock lock was our start into this area. With lock lock what we actually did is we collected tweets or we're still collecting it so it's a service to collect tweets at the beginning but you can collect a lot of other things as well and then what we do is we enrich this tweet data with additional information for example sentiments yeah positive negative like there are different libraries available and you can teach to have like more sentiments out there. So with lock lock we are able to collect large number of tweets because it's a peer-to-peer network and yeah everyone can install it we also have like an Android application. So then what we see here like let's make a query of lock lock type in something simple here across Asia how will the outcome look like. So it's a JSON and this JSON has a lot of information even sometimes geolocation information here for example place country United States a lot of different information you can use this information to provide answers to questions. So that was our first idea we can actually use tweets to provide answers to people and we have a lot of different tweets. So that is a bit the background here of Susie and then the next step what we did is we integrated more services so you can integrate other social media engines you can integrate YouTube you can use a lot of APIs we even found undocumented APIs for example of the Urban Dictionary I don't know if anyone here is using the Urban Dictionary who knows about it yeah so the answers that you get are not always so safe like let's say for kids so right now we don't have that API inside we're not using it so you can people try it out because we also wanted to showcase that on businesses and yeah for businesses so that's the thing but actually it showed us that you can use a lot of sources and okay so what you do when you build an AI you need a lot of data so how can we get a lot of data there are so many services out there we can just like make use of them and there are two ways to use these services one way is we can build our own index and for example of like search queries or I don't know if that is correctly formulated but like the thing is we collect the data from from searches like here we call it we built our own date a database and then we can query that database but another way is also to query services directly for for instance so we have two ways here and actually we realize that it's quite a pretty data intensive so we spend several thousand dollars on servers per month yeah so and like for example we have one server running with eight giga isn't no eight terabyte a terabyte right of Twitter data so we'll give a lot of good answers like a lot of results but actually we realized for for if we now compare with with Siri for example we we can do things much easier like a lot of what people do with Siri is just simple it's not actually in like advanced definitions of AI it wouldn't be real like something complicated yeah like like play a song or tell me the time or like show me the way to go somewhere things like that so for that we can actually query like existing services so that's what we're doing at the moment so let me maybe show show you some slides slides are not always the most interesting but why not a little bit so well actually what does Susie stand for no Google okay maybe not so it stands for social universal super intelligence yeah okay what what can it what can it do here a few ideas so so our goal is to build an AI that is even better than Cortana that is better than Alexa how can we do that I mean for example how can we build something better than aloe how many engineers does Google have working on that 500 a thousand I don't know but like how so how can we be better so our idea is to build a service and that uses the the principle of Wikipedia where people can contribute in a way like Wikipedia Wikipedia I mean the the Britannica it was it was a huge thing it was like the encyclopedia but today I don't know who has the Britannica at home no okay nobody in this room okay but it used to be really like the thing right and then Wikipedia came so so our idea is this so we are there I talked to you about this so we need in order to build an AI AI we need a lot of data we need a lot of services so the services there's so many I don't know I don't know what services they are out there in Korea for example yeah I don't know what is out there in Mongolia but we want to build an AI that can answer questions and everywhere in the world so we need people to be able to plug in their services so we call them console services right and for that we want the wiki principle but even like we can go a level easier and say like what else can we do we have simple questions and answers so if we have an AI everyone thinks about machine learning these days but you can already like answer a lot of questions I play me a song from iTunes they are not so many different answers the action is play a song from iTunes and then perform the search it's actually a simple rule so already we could do a lot of things with simple rules people can contribute rules so now imagine an interface where you can contribute a rule in nine steps to an AI yeah so that is possible that is possible and that's what we are doing so we don't have the interface yet but we have the back end and the way how to do that so the interface is just like the next step so we call that interface easy D AI rule development tutorial it's just like I think you just updated that an hour ago so it's really like here what we're doing is really like nearly live and here's a tutorial how you can contribute to Susie how can you add easy write easy dialogue I don't know easy D it's like German people have these ideas to come up with like some words like that or some short ones easy D conversation rules so I think we don't have enough time to go here into detail how to do that but the next step is we will build a simple UI where you can actually wear wear students or even kids we're running this with first Asia we're running this code in contest yeah with with with Google but we're also running on contest like code eat and other things and like we have hundreds of people we have even like sometimes thousands of people contributing yeah and they can do a small thing like one rule one improvement like in Wikipedia they could do that and so we already have the tutorial how you could do that in nine different steps and so if you come to first Asia I think we will have more time there to show you in detail how to do that but I want to point you to this already so I think there will be lots of questions like okay so you have a rule based systems like okay then you only get one answer and but actually no you can have different answers and it's also possible to implement that here with this system you can have better variables you can have star I bet you like this beer brand beer yeah okay people always about beer and I think we can go into detail also after the meeting if you're interested so how much time how much more time do you have left five more minutes okay good so and so where do people communicate they they use the mobile phone yeah I mean I don't know I'm I don't know how many people here are developers actually yeah okay so I don't know it's tough maybe so and how how many people have a laptop with them how many people have a laptop with them so it looks like I should have asked the other thing how many people have a mobile phone because laptop it looks like like it's not even half but how how many people have a mobile phone with them right everyone right so so we want of course an Android app yeah so the Android app project that is actually just started as a side project because for us it's just the interface we want to see that the logic works and but we get a lot of people from all over the world they're contributing to it and yeah it's working so if you want to try out because I'm not showing it life here to you if you want to try it out you can just go to our github repository and you can download it there we have an AP APK branch so on every commit the the Android app is built automatically so if it doesn't work maybe wait one hour and it will work we have like several commits a day and and you can try it out so it's basically like a chat interface where you can see Susie I think I had a thing is okay yeah here for example who am I you you am no one too bad you always like I knew okay so yeah it's not perfect but but it works you can ask all kinds of questions and especially always like my partner Michael he always likes these questions about the universe yeah I mean like the universe and nothing yeah a lot of people like post this on on YouTube and so on what does Siri answer to some kind of random question so we have an Android app here that can do that but I want to tell you it's a software project at the moment it's not a service so so like one hour it might work the next hour it might not work so we haven't deployed this as a service yet it's on a heavy development right but we also have like a web app that we are trying out at the moment and yeah so the web app is here and yeah and and just now the map box key expired so we put up a new one but like it's not deployed yet it's it's a guitar pages we love like not to run service so we run the back and somewhere but the front end on on guitar pages right now need to deploy yeah okay so tell me a question let's see what Susie answers so give me question Martin if a question to like cats cats okay yes I love them cats okay okay any other question let's see if Susie comes up with yeah how big is the universe okay big is the universe okay bigger than a pin okay yeah okay how to write George Clooney clue like this is it I double oh okay let's see like it would have been similar to my answer okay so yeah very good question so where is it getting the data source from so we have currently implemented I think like eight different data sources and Wikipedia and like also wiki data yeah and Twitter for example sometimes we have hard-coded something but like with several answers in the rule-based system yeah yeah new data source you could add it maybe like with you level yeah depends develop like what level you have but I think you could add it in one to two days exactly so the scrape through well there are two different ways to do it so one way is we scrape the data source and we store the data ourselves the other way is we actually use the just use API or for example the search page and we query live on the data source so yeah these these two options are there so of course like if you store all the information yourself you need a bigger server at the moment we don't have the resources we already like with Twitter we already done yeah so we would rather try to implement APIs and but the next thing is that you actually need to code the questions and answers so if you use wiki leaks you need a relevant question in order to get a relevant answer right but like imagine you have like the amount the number of volunteers that wikipedia has then you can do a lot right okay but like so what we understand here that we have a rule space system and now where does the AI come in yeah because now the people say okay is AI so of course you can think that we can use AI for a lot of different things I can give you a simple example and Susie will take like accounting they will take Susie will take into account what the user already interacted with Susie yeah so you can then say okay or let me give this example I'm in Singapore and maybe the weather service of Singapore is more relevant to me than the weather service from the US Susie will be able to analyze that and in order to to achieve a lot of course like we need feedback loops for example so we could have like some something like a teaching mode for Susie so where like switch on the teaching mode teach Susie something and Susie gives you for example three different answers from three different sources and you say to Susie okay this answer is correct so if 10,000 people do that we really have very good data what's correct what's not correct yeah and you could even then think okay something like what happened to to Microsoft this Twitter bot right like that people teach Susie something bad so you could establish a different layer like let's say an admin layer of very active users who say okay I confirm this yeah or registered users you could like say and you could even switch in the app like let's say I want all the answers of everyone or I only want the answers of registered users so like answers that are confirmed right and all these kind of information can be collected back and then help us to to give even better answers in the future right so that's where then machine learning comes in I don't know if that is like a too short wrap-up of the whole thing I see some people say yes some look a bit puzzled but maybe we can go also with other talks then in the future into more detail yeah exactly so this we didn't implement it yet and like we have something of it implemented but not everything yet so because Michael comes very much from a free software background and so he doesn't like the logins but I say okay but we need the logins right we in order like so I have another phone or something so so we need to lock in that usually can actually a Susie can actually get the data from from the past yeah it can already now if you interact if you say my like if Susie can also not just answer questions Susie can talk with you so for example like let's think of a shopping bot you say I want shoes Susie will ask me back what color yeah so you can you can set this up so and it will remember then what color you want it or you say hi my name is Mario and Susie will remember my name is Mario and I say I want shoes and next time you talk to Susie if you use the same IP at the moment yeah the same phone and so on it will already understand but I say like people change the IP so yeah so we need we need to lock in and one idea then of course like you all will have a lot of ideas is that we could for example use the history on Facebook and all these things right to get more data but like then it's also the question how much data we really need because some at some point maybe the more data you have about person life the like that could be difficult to understand what is relevant yeah but but these areas where we are like really excited about I think the big difference to the to the corporate world is that we are planning and and we have taken the first steps here we are we are we want a system and that can take user contributions on a big scale yeah that uses the Wikipedia principle for AI so it will not be just a simple chatbot it will actually really be AI yeah and at Force Asia we will be going into more details and about what we are doing here you have very Susie Susie thoughts ideas about the Susie mind yeah I mean sometimes Susie is answering something and I'm like thinking wow yeah where does it come from and the amazing thing is like it's not like making a simple web interface for something you sometimes you don't know where it comes from you really have to look so so it's it's pretty amazing what we can do with this so yeah but so there's a lot of detail to it and yeah we are we're happy we would be happy to to to get people on board call to action is check out our GitHub repository see where you can join install the project and at the moment our challenge is that the people who work together with are getting hired very fast and they get like very good job offers so so this is a challenge right so actually we are we are we have a also like a startup with this where we say okay so this project needs to have resources to continue yeah but we are dedicated to open source and we're dedicated free software like the practical part for me like I don't care so much if it's free open so whatever they call the practical thing because like people contribute if they can use it themselves I also see like opportunities like for everyone who wants to use it and right now we're talking to Siemens they have in their data centers computers that are not connected with the Internet they don't want to connect these with the Internet and but for example they want people to be able to call the computer by phone and talk to the computer so yeah so let's say you don't believe how many people sit in in one Siemens office and just take care of alarm management systems of buildings around the world you're in Brazil your alarm management goes on and then somebody takes the phone and calls and it's an office in Germany with like 50 people sitting there so you could they have this they have applications for this in their data center that are not connected to the Internet yeah so so this is always a big question if we are doing this also so we we are not focusing on this at the moment so for internal applications they are open source projects like a pocket Sphinx if it's on the Internet and we're using Android you can just use what Google is using and I also have a friend at Mozilla is there working on this so yeah we can just simply integrate it's just an interface for us right yeah so on on if you download the Android app you can already talk to Susie yeah but sometimes a bit slow but yeah but it's not a service as I said it's a software project right now and we're looking like also how to commercialize this yeah okay all right thank you very much if you want to find out more information about Susie Mario could you post in the media group yes yeah some of the links that you guys will have it'll be time the next week