 I said, now you're all right. So I'm going to come to government and to the FedCats. Right? That's the call. You must be trying to do the work. You still think you're going to do it? Wait. I wonder if you want to say something. Thank you. So I'm going to give you a call. I'll let you do your thing. I'm going to ask you to do something. I'm going to put the button. Put the button. I'm going to put the button. I'm going to put the button. I'm going to put the button. I'm going to put the button. I'm going to put the button. This would be much better. This is the vision. We are the company. And we spend, I think, at least a year with a lot of people, big scientists and fashion engineers. We do this for the people that are now starting the world. It's fully cloud-based. It's really for the leading customers. Fortunately, I have a lot of companies. So I'm going to put the button. I'm going to put the button. I'm going to put the button. I'm going to put the button. I'm going to put the button. I'm going to put the button. I'm going to put the button. It is the next step of the revolution. But every person, Eric was visiting, giving ihnen, and meeting this company that he invested in, he was telling me, find a way to give it for free, to give it to anyone. Not just the companies will pay you a $1 million a year. And we said yep, how can we do it is heavy machinery working on the cloud and we need to sign contracts with our customers. How can we give access to everyone? everyone. And the answer is tokenization of the predictions. Issuing our own token and putting all the system on the blockchain. The data is still stored off-chain in an encrypted way source. Although the payment desynchronization, this is done on the blockchain. And then we allow people to consume predictions in a completely self-serve way. With an API we can now go buy tokens and exchanges that it's being traded and then use our API stuff that will show you how you do it and generate predictions. And then in an hour you get a prediction that is comparable to what a team of skilled incentives would be able to generate after two or three months of hard work. And then you get the result, you can change the question, ask it again, create five, ten, thirty different questions, have all of them run in parallel, go have lunch, come back after an hour, you have the results. And this is something that is opening the bottleneck in a fraction of the cost that today a single project would probably cost. So is your pricing based on complexity then or are you going to price the engine? So in theory, every question costs slightly different. But what's also important to understand is that the way we process data is very different than any conventional machine learning technique. In the standard machine learning world, you take the data, clean it, you pre-process it and then you create features. And then from these features you train a model. And then you use the model. And this is done for every problem from scratch. So you're trying to predict when you come, two, three, four months of work, you have a model. But now you want to find who is going to fly the model. Two, three, four months of additional work, right? Because you have to do all the work from scratch almost when you change the data of the question. What we do is completely different. We take the raw data and without even knowing what are the questions that you want to answer, we first extract what we call behavioral clusters, regions in the data that we can analytically show are two coincidental, right? Regions in the data that social physics tell us are highly unlikely to just spontaneously emerge. Which means that they indicate that the users that created this data share some commonality in the real world. Maybe they are from the fact-side family, work in the same workplace, they have some common hobby. So the first thing is to take the data and extract millions of those behavioral clusters. And then when you ask the question, the system is just matching your question with the clusters that already were extracted. Which means that 99% of the work is done in common to all the questions being answered. We take the data, we crunch it once, and then million people ask million different questions. 99% of the cost of answering those questions was already done. Because it was done once in the initial crunching of the data. Which means that there are slight differences in the complexity of the questions, but we charge the same price. And it's very cheap. It's 100 EDRs, which by today's rate is less than $7. So in less than $7, you can have an accurate prediction that would be what the skill teams and scientists would give you in many cases even much more. So this could, you can use Endor to ask any question in the world pretty much at some point. It depends on the data that you are onboarding. We'll touch that point. So you have the onboarded data? Just a second. So today you can either use data that will onboard the system. This is a publicly available data, or data that we harvest. Or use your data. For example, in the case of retailers, like Coca-Cola case, they can upload other shipments' data, logistics and sales, and then they ask questions like this. Or a combination. Or in the future, this is the roadmap. In the future, you will be able to upload data. The staff will be able to ask questions on a combination of your data, someone else's data, and our data publicly available. And then the staff statement will be allocated to the providers of the data. So this is what would happen at the end of the roadmap, and the data would still be stored in a encrypted way, so your privacy is not compromised. And to your previous question, you can ask any data that can be given in terms of an example. Here is a list of customer that took a long time to find the molecules. Here are a list of products that are really increasing sales, finding products that are going to increase sales. So you don't have to follow the people's instructions. You can. The system does it automatically. If you provide one data set, then you get data from this data set. If you provide several data sets, you will get predictions that compress insects from both. The data sets need to have the same user IDs, otherwise they cannot be... So it could be IDs, it could be product IDs, right? So it could be multiple stores, even competitive stores that sell the same product. And we'd like to predict product behavior, aspects, and we'll be able to do that. So a few different quotes. Can you take over? So I'm taking over back? Yeah. It's a good question. You refer to the works of Daniel. I think that if I had to describe these works in a simple sentence, I would say that they are trying to model the behavior of a single individual. We cannot predict the data of a single person. We must have a crowd. And we know that when a crowd becomes large enough, say 500 people or more, then our rules of social physics start to emerge. And then using the ambience of the crowd, you can find anomalies. Then using the crowd, you can start to do clustering, justification. So it's fundamentally different in this sense. Yes. Exactly. My PhD was on the topics of ants and drones, and how you can generate stores of drones that use algorithms from the way ants navigate in order to optimize the sites. Sounds good. So once again, I remind you, you can stop us at any time if you have questions. Go ahead. So for example, if I want to ask your algorithm what stack I should buy if I want to make 10% profit. Good question. It's always the one. First, the data needs to be given. You have to have the data on user-specific levels, and preferably you can talk about users, not about stocks or coins or commodities. It's too aggregated. But if you want to ask, for example, we onboard the ERC-20 blockchain the public data source that we make available, and anyone can use our tokens in order to ask questions about it. So you can provide a list of wallets that recently increased their activity and ask which other walls are more increasingly. If I am, for example, in exchange, and I have customers who just joined my service, and all I know is they're wallets. And I want to know if they're going to be really heavy users, then I can use the wallet number of existing app users to find which one of my new users are going to become the way that I like them very much. So these are good examples. I think that you can already look at them. And just really quick, so I guess there is no limited use cases or industries that can use. It's pretty much to the average person to decide. How come a bunch of entrepreneurs and enterprises to use for any question that you have, whether the data is there or the data you want to be uploaded? It's a very good point. Because one of the things that we are targeted into is to provide our engine and create this catalyst ecosystem that people will come up with different applications. As a matter of fact, while we are doing the hackathon, the hackathon allows you to wrap with your application our engine and create a new trick advisor. If you have the data, you will be able to create behavioral oriented predictions which places you would like to go to. And so insurance companies would be able to decide to ensure based on this data, hospital opinion to see how much at risk are their patients are for cardiac issues because of their data. So you don't have to see that. You don't have to see that. You don't have to see what's going on. We have a bit of a change of plan. We will be repeating a few points here for the presentation. I think it will be much more clear now that you've heard it from me. Starting from the point that everybody analyzes data. We have a lot of bits and bytes that we are aggregating. The big resolution of big data already passed and it's currently existing and we aggregate more and more data from people. We try to create some insight based on that. Normally, we take the person and try to break it down into special features. He's married. He has three kids. He didn't take his loan. He did something that is very, very semantic oriented. And based on that, create some sort of models to go my business or sell more exactly. So one of the things that they miss is the fact that the specific individual changed their behavioral problems. What happens any day, any second? Somebody, for instance, had an accident. Somebody was married to somebody else. Somebody is stuck in the traffic. Somebody else is just staring at the blue board. And there is some commercial that subconsciously goes into his mind because everybody sits in the same traffic jam. Now they are all affected with this correlation or pattern that will disappear in the next row. So this is something that you should take into account. And this is the strength of our engine. Picking up those small, very, very kind of classic correlations that don't stay in the data for a long time. And this is the more essential of how we bring the project. Building a single model allows us to answer multiple questions. We don't even dwell on the question when we build this model. So we're thinking in a way of how we can foster people according to what happens now. And it could be changing tomorrow. So we would be able to be in a specific group of people that are identified as somebody and get something happen tomorrow and you're already not anywhere to be in the same group. So now the model that you've created is also, you have to be very dynamic about it. And breaking people, like firing people, groups of people can happen all the time. This is something that is normally not in the data. Don't really know what's happening. So here specifically I wanted to catch word with the music industry. A lot of people try to resolve the music industry with the blockchain and the blockchain allows us to have a lot of information out there. Everything now is out there. You have a blockchain. It centralizes all the things that involve performers and labels and songwriters and publishers and so forth. And this is kind of an ecosystem that currently is not being managed well because customers, like new artists, are being fooled using labels that take their money and then steal that money and they end up being even poorer. And taking advantage of rights, you'll be taking a song and then rip it off and then change the of the song and now it looks not the same now how do you match a song to another match. So blockchain will allow to resolve a lot of those issues. Definitely, I think that once it will happen it will have a very rich data set. But the thing that it does not solve is what's going on with the victim aspect. And once you google the blockchain, I know that the fact that blockchain is almost fully anonymized. It's all bits and bytes. It's ideas that are ever changing at least. It's all there. You have a super little bit of bullet proof and we don't have any aspects of trust there but you don't have any semantics. So data scientists that will look at this data will be very, very lost. So how do I deal with that? How do I create a predictive model based on this data? So we have a very rich data set of ceramides and data. And I think it's a little stupid because I'm blindfolded. And I would like to kind of point this specific use case of new artists how they can engage the target audience. So this is one of the aspects that you would like to know. As a new artist you would like to kind of break out from your small circle of audiences that you really like. And we would like to find engaging audiences that will purchase your albums. So how can you predict whom you should address in order to spend your small amount of fortune that you aggregated or took from your parents in order to break out? So this could be done for instance with this enter. Now Yaniv talked about the way that we're kind of our vision is to make it as simple as possible but as simple as possible and let's say cheap as possible in our perspective because the model here the financial model here is kind of everybody kind of puts a buck and they create a very strong ecosystem of data, of predictions we're using predictions a lot of people ask the same questions but slightly in a different manner so they will be able to utilize those in a fraction of a price because it's very important when you build a model once again and you have a data scientist who must because he has to plan the data he has to kind of organize everything and this is 70% of the time that he spends on in order to just kind of set it up and then you have 30% of the data of the time taking to create model kind of understand what kind of strong features do I have in the data in order to create something to predict the better So this is the way that we're doing it so this is how we're doing it now in a second to show you how it's working and then we will be able to discuss it as I said as an open session and how it can be very utilized but it is as simple as that you have to upload a sample of a group of people that you like to rank so you have a lot of people that from, let's say, Boston who like to take the entire population of Boston and I would like to upload an example of my friends for instance that like my music, I would like to find who acting the same way and please target me and rank me the entire population of Boston according to the tendency to behave the same manner so once again it's all based on the data that is currently being available so if you have any data type of phone calls for example if you have social media presence everything that is available but it's something of a blueprint as I said that you are living after yourself in a digital world could be used in order to eventually see if there are those creations that happen in this time that you are asking for and create for it and the other thing that is very important is that the predictions normally are worthless if you are doing nothing about it so if you know the future and you don't do anything about it you don't target the audience that you got from this predictive reward and nothing happens you just know that you are most likely the people that should not target you so the system allows the essence of the prediction what is the gain that you are getting and this is the other thing that is very hard to do when you have a single mention so this artist has an engagement he said well how much would it cost me to engage with a specific person I would like to send a disk I would like to send you my USB key that I don't know branded as my name it has my track name so it's my cost I don't want to say send it to the entire population so I know how much does it cost to engage and I know what would be the conversion revenue if you are converted my album costs 13 bucks so I understand that I am getting 30 bucks back by giving you a 5 back black seven let's say a blue and this way will kind of allow organization and smooth organization as individuals to understand what is the essence of the prediction so we are ranking the entire population I remind you for this one thing here 1.1 million people right it's not vast area now we are asking questions that all customers who is likely to pick an additional more in the next month so it's kind of in a banking world but it's the same use case and how they behave and now I would like to find more people of that kind so I provide an example of people historically behaving in this way additional loan so for instance they already like to my track so this is the people that would like to use as an example and now I have ranked people of the entire population and their tendency and very similar people that I gave as an example now I can play with I can say let's target the 10% the top 10% of the list so it's a statistical model so the top 10% I should be having the most density of the people that are most likely to be the people that I'm working for the prediction itself has a point that it becomes less efficient because we are ranking the entire population so there are a lot of people that are irrelevant they just don't like music they don't listen to music normally and of course people that are more and more relevant to my listings so the top of the list I will be finding the people that would like to target and now with the second population I see that if I address 10% of the population I will be targeting 114,000 people right I'll be let's say engaging with $12 giveaway and I expect to get $5,000 so here we're talking about banking and this real data how much does it cost a community of pricing of buying a customer and what do we get if he's being converted we can easily see that if I go and I target the entire population of 1.1 million people then and there are only 1.3,000 1,000 people or 371 that will actually do that that I will be in tremendous losses right because I kind of gave those amazing giveaways and I didn't get anything in fact because the audience is fairly small people just don't like my music so maybe I should consider not doing that anymore so this is one thing another thing is if using Endor so you'll be able to target this audience and then we'll get those gains it's very very fast in order to understand what do we gain out of the prediction so we create a prediction you have this prediction if you act upon it then you will gain this specific ROI once again predictions are all statistical models so this is something that is estimated but normally this is what we're talking about now talking to the science guys in the room and there are a lot of metrics so if you'd like to understand better how the prediction is being calculated you'll have a lot of metrics that you can take and see how precise are you how you know that you are in this world but AUC is a political measure at one you can see the precision, you can see the recoil you can see the true positive you can see how it's being ranked within this you can see how it's going on with fewer numbers you can just test it and more of that you'll be able to test it in multiple points of time so you can ask it today, you can ask it tomorrow you can ask it the day after and create yourself a better understanding what is the size of audience now that I need to add so this is one of the things that allows the other scientists in the room to understand that something that is really transparent with regards to the metrics that we are measuring and if he goes and creates his own model we will allow him to just cover this and say I have a precision on this how does it compare to the precision of Andar in some cases we will have a poor model because the correlations of human behavioral patterns are less strong in this specific case, for instance some cases are more oriented on behavioral patterns, some cases are irrational and doesn't have anything to do with people behaving so some questions will be less accurate than a specific cure of future which you can find in the data so this is a very important point the other thing is that I will try to aggregate the types of data that are available in the system and try to profile and allow you to understand much better what identifies the people at the top of the list so for instance if you have a lot of mental issues about people in Boston so you have information about their schools that they learn, their age and all of the semantics layers that you can aggregate on those people and then you can now see how it breaks down to see that your audience is kind of oriented to a specific age from specific origin people from Eastern Europe like your music more and their age range and so forth so all the semantics layers we try to take the prediction and enrich it with the semantics at the end of the day in order for you to allow to any marketing campaign anything that you would like to do to have it in a small way so kind of analyzing and doing the eye layer as well because normally people find it very hard to crunch data and when it comes to terabytes of data it becomes even impossible to write how do you do that nobody really understands how to analyze quantity of data unless you are not designed so this is kind of a I will stop on that so this is how they analyze data and I wanted to show you this is those are the parts that are being aggregated I will show you in a second the data that we are crunching there and I will join Daniel if you want to see the data itself it's taken from last FM last FM is a radio station I don't know if you are familiar with that sure a long time ago a long time ago a long time ago so they still exist and they were kind enough to share a lot of information about what is going on a lot of similar things Spotify is a more young solution although they have a lot of personalized aspects with them most of the audience here listen to Spotify or maybe Pandornos so they have tremendous power tremendous power comes with those rules that they have a lot of semantics on you they have a lot of power with regards to the artists who are engaged with them so they came kind of push forward take that word do whatever they like with the songs with the artists they are very wacky or very good at what you are doing and at the top of the list and it's unfair it's kind of unfair this is a platform this is a centralized solution we are talking about blockchain a lot of people raise their eyes when we are talking about this information this is what we are talking about the information will be available at the end of the day and we will just talk about it now we will be able to create those things at the tip of your hand without a doubt I think those companies will have a lot of power reduced they will have amazing projects amazing engineering solutions when it comes to new artists they will become less relevant because they will find alternative ways to grow and become those top artists in those lists so I will show you a very fast how it works this is the enterprise edition this is the version we are selling banks what they do is this is how it works specifically here they didn't hash the data because they said how it works so for instance here we have albums so I have the album I have here the type of the album and some information about it this is one of the data aspects I have tracks I have the name of the track very quick one in general you would tell them how to hash the data right normally what we do is we have an album that allows it to be done on the customer side if you are concerned about the way to do that no problem it does it on your end it comes to our infrastructure that we don't have the keys so kind of show you the car give you the key and then leave the car and the key is on your end you cannot do anything with it but you need to know how the hash functions in the background works it's alphanumeric values and from that point it's good enough for us we are looking at the correlated aspects of the data as I said, dates are very important because everything happens with the time frame and when it comes to data when you hash it let's say some say in a morphic way I don't want to get into it you can create data in different ways then you kind of preserve certain aspects of the data when it comes to let's say the key aspects of the data without disclosing the semantics I don't think you know what's going on but I know that something's going on something, using the end or end generator you know what to say to say that this something those people will not strongly correlate it now you give me an example of people who are purchasing the album and I see that they are strongly correlated with a different group of people those are most likely the people that will behave in the same manner which is kind of taking back to the cluster so once again this is the interface for the enterprise companies they kind of do it very very well you see it here, you have almost the cash data which you cannot understand what's going on here and the way problem the data is fairly easy this is what they do they upload the data it takes a couple of minutes then they validate that the data is with the quantity of aspects of the data is okay but the biggest difference is the amount of the expected get because export of the data is one of the painful things that you do not want to miss because once you stop from the poor point of data then you kind of lay down the poor precision and accuracy and from that point you start doing the magic so the create sphere this is the sphere that we are talking about this is the social sphere and when it comes to questions big companies it is very hard for us to give you an example we have tons of data taking those data sets and crunch them together so could you please assist us in doing it fairly easily so this is what we kind of did with the square builder we said what do we want to find we kind of build it in a very easy way so we have a complete language here we want to say artists from this area who behave in a specific manner and are not doing that and that and so forth so if you don't know how to crunch your data you will be able to do that within the tool itself and the same thing is from where so what people do you want to rank so I would like to rank people who are active in the past 30 days I would like to rank people who are in the Boston area I would like to do what may ask me here you have the semantics so you could tell me CT is A335007 I didn't know what it was A335007 but you are focusing on Boston I understand so with regards to the asking the question you have the transformation table at your end the unhashed values you can ask it with hash values inside the system this is once again an enterprise solution now I want to give you an example of how it works in SMB individually so this is the end of it all you have to do is actually kind of say a playlist of people that behave in a specific manner for instance here active customers list for 7 days I will show you how this looks like only I do this and then I say look likes once again list of people who took along I know for a fact that historically people took along I want to find some more of those this is an example that I will be giving so any questions that you might have would you now recognize if the data is insufficient or the data set is too small good question so normally as you said we need to have as substantial data in terms of users and a few dozens of data points in a hero pattern so it's not meta data I'm not interested in the fact that you are married or you have three kids I'm interested in the fact that you are traveling who are engaging with some sort of device when doing something that could somehow identify you as a person it's not a semantic issue it's a strong decision that everybody is going to be able to do and from center to center we have one or more kids less interested more interested how did you reach here what time did you enter how long did you sit on the chair how long did you pick up your phone this is what interests me because eventually it identifies the way you behave are you impulsive, are you quiet are you very engaged with what's happening if you are listening then you are not engaged so it's a part of me sorry I've got a couple of questions the next thing is you're talking about these patterns how do these kind of checklists you have beforehand are you looking at the data and then saying what could be a pattern what you said is very important to understand the thing that is special about social physics is those are mathematical equations once again it's targeted to a specific domain so I'm not tracking for specific things I'm tracking for correlations that happen, something happen I see an anomaly of social physics I understand that this anomaly could not occur if something really happened in the real world if me and you don't behave in the same way we will be below signal there but if you and me behave in the same way I'll give you an example something happened and in fact a terrorist or a fire in a specific building a lot of people started to behave in a very similar manner calling their relatives exactly now doing some specific things such as going out very fast from location to location so I don't know what happened maybe it's a party or maybe it's a fire breaks down from the building but if the data is encrypted we know that this kind of pattern is taking a phone call it's not really interesting because it happens in the same aspect so if it happens in the same location something changing in the correct manner I'm not really interested it's something that happens he doesn't know that the people were in a fire he just knows that the people are similar now you will have to tell me that it's a fire then you'll tell me we're involved in a fire I'll say well those people were involved in a fire so if you for instance have people that disappeared and now they are not there and now we want to track down to say if something happened I will be able to cross over and say okay we're slightly disappeared is not there but he now went to a drama or fire and maybe we should do something like that this is as simple as that you give your email account where do you want to send the prediction to specifically here we tied it the blockchain as I said we would like to democratize it so we're not kind of chaining ourselves a specific solution I think that we will support eventually traditional ways of painting as well and transform it to the blockchain way so you'll be able to pay whatever name that you find appropriate then you have to just say which deficits do you think are relevant for your question now eventually as we roam along I think that we will fuse this as well and we will not go to interacting we will choose the right deficits for your question and then if you were involved in the model itself then the relevant part is you will be kind of you know providing the funds so once you do that, that's it the question goes out to the ecosystem and once it's ready so it's submitted it started to run you have this request where was the question in there? specifically the question is the example it reminds you that you can only for you can ask any question you can give me an example of anything you can say well I would like to find people that will buy my product actually what you are looking for is people who will field you and whatever so I even don't know what you are asking I'm not interested in it because I'm finding correlations so we just give it a name you can fool me, do whatever you want if you want to hide your question it's very interesting by the way that we did we did quite a lot of things with social media for instance tracking ISIS members how do you do that so you have a lot of information a lot of data a lot of information talking about different governmental parties that are interested in the solution the problem is that we are talking about loud solutions it's kind of an everlasting fight when I have to go from my product perspective and saying that it's kind of changing my pricing model and changing my entire company profile so going back to the report itself as I said it's very very dynamic to play and see the bottom line right so you see the cost of engagement is $1 I'm going to get $1,000 I'm going to target 5% this is what you're going to get with my model this is what you're going to get with the run themselves as you go and just keep anyone in the room in the giveaway this would be your way of gaining from this process and if you have your own model then you'll be able to say well my model is 1.4 so this is what you can do so you can kind of track what under provides you and if you have that model then be it, it's fine so there is a killer feature that says strongly likes your product I can't do anything about that so there is a very strong signal you have a better base than me and this is the biggest case that people of sample actually are people that you are looking for so kind of what we're doing here we're rolling back in time hiding a portion of the data from the system and then seeing how well they can do and then as I said you'll be able to evaluate your predictions by saying I'll drink the future here for instance I created a prediction for today now I have people who purchased my track so I can upload it here and see how well did I do with my prediction once I do that it's amazing I can see really how good is the model but if it's bad then maybe I'll ask my question in a different way maybe I'll say on those maybe I'll reach my example I'll say I like this specific artist and I'm really really really kind of affected by his output so let's look for my my tracks and his tracks as well so there will be a much bigger example of audience that I like to track because now I have much more people that I can take as an example but it's like a different question because it's not on my track but it's somebody that I really like and I'm really influenced by so I'm kind of taking the leap of faith that those people who like mine are something like that it seems like excellent for problems where there are lots of examples but in something where it's maybe less clear like for example before people started making money on the internet no one knew if there would be a successful company like Google will this perform well on a problem like that where there aren't many successful examples or is that outside of the domain? Amazing question because one of the things that we do find ourselves doing later is taking marketing campaigns and saying what marketing campaign should we kill and this is exactly the point that you said because marketing campaign normally starts with 0 and you don't know how well it would do so we kind of purchase a few bunch of a bunch of doors and windows and a bunch of channels and we don't know whether people will give you the ROI that you're expecting and those emerging patterns of people engaging with your specific marketing campaign they have a very small data point so you have to be very, very, very fast so what we do is we were able to attract based on a very few days of activity well two or three days of activity and say well this is a bad campaign the engagement of people with this campaign is poor there are social aspects what's going on there nothing will lead to good places kill this it's very good the budget from here will have much more success this is touching on the point that we have an emerging product that you don't have a lot of engagement with once and it's ever changing because marketing campaigns are ever changing the landing pages that you give you have multiple landing pages for girls it's a pin for boys it's a pin so this will allow you data points in a very short amount of time where a human really kind of doesn't participate in that because you have to create some sort of hysterical data set in order to validate your prediction here you'll be able to validate it every day ask the question once again evaluate it it costs only $7 per prediction you have a strong model the next question is unrelated but how do you have encryption or hashing without destroying non-categorical data let's get to the question so it's a very very real data an amorphic way the future is something that preserves the data from all the aspects even categorically numerically it could be preserved in some way some companies make it to different levels of sensitivity some companies just say we have a hash function we want to have the keys we're not giving you the keys those hash functions change every root changing the keys every root it's a very simple way of engaging with it because hashing and data is very very simple encrypting data is by the way not a trivial thing because you have to roll through the entire data the alternative which could inflate in the amount of space it takes because every rate of value is kind of transposed to a very large number or a half an American representation so taking root points specifically it's a range that everyone kind of are capable of identifying whether it's a simple encryption hashing, very complex encryption where a fine would either of those aspects would do take the time to make sure that the categorical as you said values are preserved they don't have to see them but they have to be kind of coherent to what you gave was the question about non-categorical? yeah because yeah non-categorical so non-categorical sorry the other thing I wasn't really saying but I think it's important is that if you don't know what the different data represents you don't really know if it's categorical or not but for example if it's a time interval 0.1 and 0.2 you don't want them mapped to wildly different things because they're close to each other compared to like 12 correct correct so one thing that we do is we don't mix columns so every column think of it as kind of taking the lemon taking the juice of each column that you have in let's say so columns will not affect each other as long as you have income the same value the other thing you could have much more insight when you merge the rules kind of you take the time factor you take the product factor and now you close the insights together I think it will give you a much better solution because the foundation will be much stronger but once again we are kind of giving this freedom to the end user you want to have a very very hot encrypted data and you're losing data there it's fine to ask because you don't have anything now you definitely will have something with Android you want something stronger let's talk about stronger and give you some more data so the user picks when we were starting the contract this is the one thing that we discussed with the CISO company and the IT members of the company so our work was medical data and there's all this stuff about regulation and de-advocation I was just wondering are lab values, is that a behavior? lab values when it comes to people coming to do their tasks and their tests are not single point in time tasks for instance it's not usually checked because usually check is an attribute but if you constantly we have different aspects of behavioral aspects we have measures that you do lab values we have a lot of aspects of behavioral aspects we take two pills a day we take it in the morning, in the evening we sleep more we have blood pressure and specific for instance we had a news case with IoT a lot of smart devices now we track your behavioral patterns your walking everything is being transponded to the data points that say you have lower blood pressure higher blood pressure your heart rate so those are good things only the lab tests by themselves are not as strong because they do not possess as they are looking we had those discussions with the medical the name as a matter of fact we had this discussion because they said we understand that the data is very, very sensitive you are GPR compliant there is no PI-9 science it is close to you which is amazing because normally the company is full there normally the company will have a strong model amazing results and then the representative from the medical says I can give you the data I'm sorry here we have discussions we have a very interesting discussion in the world as well I think that we will be able to do very interesting news cases in the future and as we go along and we collaborate with the different doctor's providers we will be having a lot of news cases that we will be able to share with people and then maybe push them to share more data and not be offensive about it so this is basically a long answer do you have an open API so if you wanted to rather than use this you had a project and you wanted to embed a script against it good question Daniel there so Daniel there is doing now a crunching session the crunching session does exactly this what we do is we say well you can use our engine without using the layout you can take it and create your own application as I said it's a catalyst problem we have by the way three catalysts already but onboarded some of them are doing some sort of predictions in the world of blockchain and you could create any applicative flows and you can engage once again engagement is very very fast upload list do whatever you want and those specific aspects that you are looking at here those are the business aspects as I said and break it down those are the main aspects that are quality over the top of the list those are areas and so forth things that you can use in order to create a rich application those are available for APIs as well so now Daniel you are a very kind of very kind of fast way of hacking of the music data if you have an interesting use case you could definitely do that by yourself as well question you just mentioned people in areas and everything so in general you can predict amount of the people in Boston who would buy the water right but for my advertisement campaign as always used to be spray and pray now I need to very granular so I know he is drinking in the morning he is drinking 10 o'clock 12 o'clock about 15 and to my advertisement money I give him $2 $3 and this one and your algorithm you just have in general but you cannot tell me okay, geographically incentivize and hours I will give you that the problem with this approach is that you are really reliant on the way that you kind of correct that you have the right model of incentivizing people and this is incorrect because the one that you gave money to and you thought that he is doing something the next day he doesn't do that but you still give him the same incentive the day after so those things are very dynamic if you want to be smart and talk about the things that you do with the marketing and this is what kind of touched here a few minutes ago you have to be very dynamic with the audiences that you are targeting so the audiences change as I said from different reasons they have this on their hands sorry, this on their hands all the time and everything changing life is dynamic and we cannot create women oriented models but this is what we are doing we are validating the granular to improve if it is I am sending him $1 and he is using $1 so the value is going up if I am sending him $1 and he is not using it is going without influence so it is a responsive way but I am always closing and evaluating one step back this is what I am saying you are always one step back you are getting burned and then say don't want that anymore I don't want him, I want him this is why we believe that there is a much better way to handle people better so first of all I am sorry but luckily through miracle science I will see it on YouTube in the future and then hopefully you guys can do the same and I love it can you speculate with me how will the technology I use be made on carbon credits so in that context there is one carbon credit and there is carbon registries if you are familiar with this it is and Mark has never really taken on it has been regularly there are registries and there are other issues too so there is a way to maybe fractualize the value of the carbon and consumers could take small actions and also carbon footprint of the car ride and watch a video where you have this small things like that how could you maybe find that by social physics so would it be something that you applied to to be able to join a group together how could you apply it based on your context interesting whose case and I am thinking about what kind of data can we use to solve this issue because taking the plain data and giving people credit we will have to start by some sort of privilege we would like to find more let's get some good audio on you we want to make sure we can hear you like a monkey right in your head so taking your point if you have an interesting basis which has a lot of aspects of behavioral behavioral let's say data points and then you take your program allocating specific say allocations for a person who will become better at consuming it or who are consuming it incentivize them so I understand you but I think what you need to do is as I said there are a lot of data sets to identify you as a person it would be if you tell me that we have CDRs people behaving in a way of interacting in the cellular world and then I give them some program for some sort of allocation of a specific thing that would like to optimize it right so then I will be able to say I can group people based on their behavioral buttons then you will say those are good I will go back see where they are correlated together those are good I understand so I think that was what we were saying what data sets would we use let's figure some out what if we have a what if we have a ride share company Uber or Lyft where it incentivizes people to offset the carbon footprint on their ride and then we can maybe do adjacent ways to like Airbnb or something offset the carbon footprint that everything you did in the city of the houses had gas still heating so we've got customer data date of staying time and date and location trip that kind of story the length of the script those are important as well just think about it people who take rides, short rides it's a different kind of people people that like walking this is very simple news cases and this is an example by the way I cannot disclose the two names of companies that do that in Boston as well you can tell us there are two companies one of which you've mentioned I don't know you can't disclose them so we can take this information very rich data because rides data really a good representation of what you're behaving in this world of how do you move in a physical world if you're a business person you don't take a lot of caps because you're very active if you're an office person you're on the same location if you are somebody who works long hours I'll tell you in a specific world most likely of industry because they are long working short working so those would be amazing assets to take advantage of and take your program and see how they engage identify very very fast the goods of the project engage with the product the way you want it to see that they're really kind of responsive to what you do with them and they do change their way of lives because you gave them those let's say incentives and then you'll be able to find more of those and then say and incentivize you even more because you'll have a growing effect growing example and so forth you've spoken yet I mean I've just been listening so can I out you a little bit basically this guy's got a good idea for a new kind of music an economy, a psychology or it's related to samples people have heard about it but it's not it's not it's not it's not it's not so maybe can we brainstorm for a minute and just to go off of what you were saying because it was interesting so basically you said you have an API that allows existing technology applications to pop on to music so Airbnb we're definitely using it to see where the demand what people need to stay where incoming tourists to San Francisco or Boston where they want to stay how much they look for the consistent Uber drivers when they wake up in the morning it's enough for the morning in Boston to get to work where should I be should I be in Cambridge should I be in the orchestra so I'm just showing you supply this is one of the things that are real cases the real cases of you have a few and now you want to be who did what you do because let's say you're measured by the answer I wanted to ask you if I have to wait for five minutes it's a poor service if I have to wait for one minute so kind of in this world of competition if you have this knowledge how to predict where will be that next how can I utilize my resources better those are real cases that are kind of working with those companies and they have a lot of ways for the food driver to ride from the end of the road in secret because he accepted where he was sometimes I understand because I see ten minutes I have to wait ten minutes for him to drive to pick me up I am sure that there is a driver five minutes away from me but sure I've taken this drive and spend less fuel polluting the environment so forth and wasting his time and they're always looking and Uber are always looking to prove their confidence and it just seems to be in the brain for them to integrate people into the industry because in fact you're making a big difference smart decisions for the company and for the food drivers so everybody wins so my opinion is as a medicinalist producing hip-hop it can be created organically you have a deal using whatever software you want but then there is a lot of artists majority of artists they have an ear for work they don't only have a musical knowledge they build an organically or they want to they hear something and inspire and they're like hey look this is a big channel track it sounds great but it's billions of dollars value are just stored away and not being utilized because of the fact that it's really hard to have artists be able to either use it pay for it, license it for example for me I can be watching Mr. Roeper I don't know Fight Club I'm not allowed to be so great for an intro or whatever the point is that there's limitless things that can inspire people to create music there's not really efficient ways to go about that if I want to sample something and I can't find it on vinyl I don't have the means to do that on YouTube I don't have the mp3 file that's compressed and kind of rip it using YouTube to have a free time to put a lot of them now because it's probably on there you can't do that so you're stuck with you can't sample what you want are you concerned with the back catalog or are you ready to just start over because the back catalog is owned by the dinosaurs and they don't want you to do this efficiently exactly and that's what I was talking about you didn't say normally it's not a single man show like you said I don't maybe I'm not adequate to do certain things I'm really good at doing something but if you had a multi-layered implementation of the song sampling and putting something on it and so forth finding people in one of the things that you can do for instance once again those are the people that I like to work with socially those are the people that I really connect with those are really killers in the world of sampling but they will never work with me because I know they don't want to come up with me finding some people who are very behavioral astronauts who are very active in the way that they interact with other people how they listen to the music whether there are a lot of things that are embedded in different lines of movies so you understand that this specific persona is just one of the same frequencies in you so you can collaborate with people that are not superstars they're most likely to be able to take your information at another level it's the same way of finding ordinances for your song finding people that you work with for me it's mainly building an API where it's embedded in Netflix or any broadcast TV or anything that provides audio because of the fact that they said to me I'm not inspired just by listening to a song a little in between transition from where it's just like you're in a movie or sitcom or show where you're transitioning from one scene like a coffee shop whatever it may be the point is that there's a lot of things like how someone can be inspired to remix something or make something that doesn't just come from music it's media in general there's a lot of those plays anything that's recorded and the only way to do that now is through samples of China and YouTube so I think it's like how do you build a protocol or API or something that you can program to Netflix or whatever it be HDL great soundtrack great soundtrack how can I get an actual permission high quality vial can you make a natural break where people can get pizza if they want to do it closer from that point I think that for the first answer for you to be I think that creating a service high quality embedding different parts in the music world or in the video it's really quite complex but if you take once again it's possible but if you take people who are good at it and specifically at that and you combine people that are very good that are very good and you can create an ecosystem of collaboration and say I'm not solving that automatically but I do and I connect people in a way that they will collaborate and create what you want automatically I think it would be what is that there's a lot of people but it's done there and it's real alright fantastic so clearly there's a lot to discuss with Endor and it's a deep technology it did come from MIT for you to think so in addition to pizza it's so fast not so fast come here come here you're probably saying I don't like pizza I'm always kind of troubling I'm living a different life but I'm traveling here in San Francisco so we'll be able to catch it show show show show show show show show show show show show show show show show show show show show show show show