 Hi, I'm Daza Greenwood, a scientist at the MIT Media Lab, where I run law.mit.edu and also a co-chair of legal hackers chapters that are putting on the open media legal hack this October, where we're going to be exploring how technology can put more ownership control in the hands of musicians and artists. And I'm joined by Daniel, who is the CTO of Endor, one of the sponsor companies who's making a really innovative, blockchain-based data and analytics tool available for teams that might want to use this technology for hacking projects at the Hackathon. And I'm just very grateful that you're taking the time to join us, Daniel, and to sponsor the Hackathon and to introduce yourself today and talk us through a little bit what your tool is and how it works. Welcome, Daniel. Thank you, Daza. So my name is Daniel. I'm the CTO of Endor. Endor is the main platform of doing predictive analytics on top of user behavior. Specifically for this Hackathon, we're going to be focused on predicting music, any predictions in the music field. And today, we're going to emphasize on showing a bit about our tool at Endor and what we developed specifically for you. So each one of the Hackathon participants can run their predictions on their own laptop or using their own notebooks or clouds. Great. Okay. So if you don't mind doing a little screen share, maybe you could walk us through how it works. And perhaps provide a little bit of background and context about Endor and what the purpose of your blockchain is. Then when you're done, I'll do a little screen share and show people where they can find more information on the Wiki and the documentation on our site. Sound good? Sure. Okay. So at the core basic, Endor looks at data science and machine learning in a different prism. What we do is instead of looking at sets of data as features and running standard models on top of them, we're actually clusterizing people by their behavior. And then we simply look for other people who might also act the same as the true positives, meaning that if we have a lot of data encrypted data, of course, we only run on encrypted data of a specific bank or transactions or any other usage in the bank's app website, then we are able to tell who is most likely to take loans in the next few months, who is most likely to increase his engagement with the bank, and so forth. This specific use case is very specific for bank and for financial vertical, but actually we're not limited to this vertical. We also have a few other use cases where we predict traffic jams using data from taxes, detecting terrorists and general bad guys from phone calls, and many other use cases. Pretty cool. And boy, it's so timely for this hackathon, we're looking to transform the music industry. One of the things that you'll find out more about when you get to MIT is the Music Modernization Act that Open Music Initiative has been working on, and it's going to really change the music industry in the United States and beyond by allowing for bulk licensing and simplifying and streamlining all of the streaming and other nature of how music exists as a digital asset in the current economy. And so that's going to really put a premium on more tools to use music as data and then to get more insights and ways to analyze and have more automated and data-driven processes. So very, very timely and can't wait to dig in. So do you want to open up, I don't know, a terminal or something and show us how to install, config and run this? Sure. So the assets we provide for this hackathon are kind of splated into two. The first one is the data itself, which would be shared a bit more closer to the hackathon. And the second part is the actual computation. The computation is quite controversial because it's the first time that Indoor allows people to run scores or predictions on their own laptops or on their own machines, and this is actually the first step towards being completely decentralized and being blockchainized as the vision of Indoor is. So we're going to start by looking at the Docker image, which is shared on the website, the hackathon website. This is the actual Docker that is being, the actual script for the Docker to be run on top of that is shared in the website. As we see in the output directory, we don't have anything at the moment. If we go back, we can look at the input directory, which basically has all the data that is actually going to be transferred and predicted on. So specifically, if we look at the Cornell, Cornell CSV, we can find a lot of IDs. These IDs actually represent transactions from users to the actual musics. In this case, we're talking about users listening to specific tracks. In this Cornell, you have a few columns where the main ones are whites. White in Indoor terminology is the true positives, is the people that we actually want to look for similarities in the actual results. If we have people who like rock music, we will also find people who also like or likely to like rock music. So the important thing here is to actually configure the whites prior to actually predicting or running predictions. And once we have the whites, we can go back to the root folder. Okay. Here we can look at the run as H script. The run as H is essentially what it does is it downloads the latest version of the Indoor square and runs it immediately. So if we run it, I already downloaded it for us not to have some busy waiting right now. So if we run the script right now, we can see that the docker runs. Awesome. It does a bunch of stuff, it crunches data, it's actually looking at the clusters we made from the user transactions. It looked at the whites we configured in the kernel CSV file. There it goes. And when it finishes, it goes to the output folder, which we can now see a single folder in. Yeah, I see some scoring of the population, so it's crunching the numbers. Right. So if we go to the actual output, we can see all scores. So right now, what we have right here in the output folder is the IDs of people in our example who are predicted to be very likely to like rock music as the whites we provided him with. Awesome. Yeah, and the second row is actually the probability of these people to be very similar to the ones inputted as whites. Okay. How do we see that second one? Is the second part after the comma? Oh, okay, I got you. Okay, thanks. Sorry, I'm just seeing this for the first time and just trying to parse through it with you. So what can we do from here? I mean, you know, it's a bunch of numbers, the very basic of it. But what can we predict from here? It can range from people who like rock music to people who like specific bands, to music trends, to what's trendy in the UK, or what's going to be trendy in the UK in 2019, for instance. Yeah. We can run so many predictions on top of this data and essentially at the very end of the prediction, we can kind of join the IDs we can see here with actual tracks, actual albums, actual artists. That's so fantastic. Very, very timely. You know, one of the things that's going to come up I think in the future, which maybe people could extrapolate from the data that we have right now into emerging business models when all the music's available under the Music Modernization Act, more easily for more automated or at least programmatically addressable licensing of music, being able to see for maybe for other people that have made videos of this type or done, you know, works of this, commercials of this type, what kind of music have they used for this and what music kind of related to that or that might have similar attributes or features might I want to look to license or other things like that, because it's going to be a data world in music imminently now. So boy, this is just so, so very timely. Okay. Well, is there anything else that you wanted to show on just, you know, how to use this or is that does that kind of wrap up for now what the preview of the application is? So this is actually the core essence of this year's hackathons. We are planning to release so much more, so many more tools for data validation for statistics. And of course, the full-fledged, the full-endored products include so much more than this. For this specific hackathon, we have only once per, but here internally at Endor, we use hundreds of them per prediction. Of course, we run it on a decentralized kind of clusters or distributed clusters. This is not something we can share for people to run on their own laptops rather than on full clouds or full solutions. Understood. Yeah. And that really gets down to the power and the potential of Endor as you kind of described in the white paper and on your website. And actually speaking of that, I'm going to take the screen over for one moment. And I just want to show people, if you go to legal hackathon.org and kind of scroll down, down, down, you'll see under resources, Endor tools. And you can click on this video that you're watching now if you want to see it again and relive the moment. But also we've got this other Wiki page that we've set up for Endor and that we'll keep updating that just goes through that first part that Daniel mentioned about how do you get the Docker image? We've got a link here to some script that will let you kind of get the image and unpack it. And then here's just the steps that you take to kind of kickstart the process to be able to do the analytics that Daniel just showed you. So that's how you get to that on legal hackathon.org. And then we'll be keeping that kind of evolving as more documentation comes online in the days and weeks leading up to the hackathon event. And then also if I'm not mistaken, we're going to be, I'll take my screen sure off, we're going to be very lucky to have people from Endor at the Media Lab for the hackathon itself and you'll be available to answer questions. And also we've got a telegram channel and Endor has a telegram channel so we'll be able to get a little bit extra support if people want to try to push the tool a little and come up with some creative projects. So with that said, actually could I invite you to maybe, oh there you go, you did it. Okay, I just wanted to see your face again while we're closing out. So I just wanted to thank you so much Daniel and your whole team at Endor for being such great collaborators and making this Docker image available for the hackathon. Well, I know you've got so much going on as a startup and you've got really big goals and it's been really great working with you so far and being able to see the fruits of your labors now in a way that people can get their hands on. And the Docker image part of this at least is open source and we've taken great pains to find data sets that people can have rights to use and so I just want to encourage everybody that's hearing this when it comes to the hackathon and beyond even after the hackathon, if you discover this video, give it a try and see what it can do. And it does a lot more than we're able to cover in this short video and read the white paper to learn more about some of the privacy protecting aspects of it and some of the other data-driven model-based potential for how this tool can be used and amplified with the distributed architecture that blockchain makes available. So thank you Daniel and thank you to your whole team. Thank you so much, Deza, for your help and support.