 All right. So this is the last session of the day. And the plan is basically, so far, if you have questions that have not got answered, then our hope is that we can ask these experts to help us answer those questions. Also, as you can see, there is one chair that is left empty. What's the purpose? Because we believe that people here are also equal experts who can come up here and answer those. So this is what is referred to as a fish ball-style panel, where typically one chair has to be left empty at all times. So if there's a question from the audience, and if some of these guys answer the question, but let's say you're not satisfied, you think you can add something to this, then you come up, you take a chair, and you answer that. When you do that, one of these guys have to leave. Might as well sit down. And then you can ask a question and come back on the stage. All right. And just to make it a little bit more interesting, we have three echo dots to be given away. And like we publicized, yes, if you are sitting down there, and if the jury decides that that was really a good question. All right. I don't know who the jury is, so I'm assuming we'll let people decide whose question was the best. That's easy to just do crowd source it. All right. So I'm going to kick off with the first question, and I don't get an echo for the first question. All right, so it's a fair game. I'm just going to kick off with the first question just to set the context and get the ball rolling, and then we'll open the floor for people to ask questions. Cool. Those of you sitting here, I don't think you can see any of these guys. There are a lot of empty chairs there. If you want to move there, that would be good. Quick escape. Just let's see if these guys talk any sense or let's just escape out. All right. Cool. All right. That's a good suggestion. I assume that everyone knows all of these guys, but maybe it's worth doing a quick introduction. All right. Cool. We start here with Dr. Vina. Hi, I'm Vina Menderada. I work at Nokia Bell Labs in Chicago, and I work in the area of network reliability. My specialty, I mean, I'm an operations research person. I don't do software or hardware, but I do everything else. Hello, everyone. I'm Fabio. I'm one of the scientists. I work for Racken Data Group and for OXO and for more companies out there. I love sharing my knowledge, learning from you so you can learn from me then. And I create blogs, articles, webinars, stuff. So be active in the community. We want you there. All right. Hi, I'm Tom Stark. I'm the CEO of a consultancy company named AAA Coons, and we do a lot of work on financial models, automated trading systems, and anything to do with automated automation and machine learning. Hi, I'm Arun Verma. I'm with Bloomberg in New York, kind of in the same area as Tom in the area of quantitative finance. So, you know, finance is really a goldmine for high-dimensional data. So it's a perfect place for data scientists to do their magic. So that's what we do at Bloomberg. Hi, everyone. I'm Dennis. I'm a research scientist in Australia, at Australia's government research agency, and my team specializes on bringing the latest in distributed computing, Sparkadoop, or cloud computing services like Serverless to the live science area and thereby enable technologies or applications that just a couple of years ago have been thought impossible. Hello, everyone. My name is James McGovern. I'm the founder of ODSC. I used to be more of a software engineer than a data scientist, came out of financial fields. I like to say now that ODSC has made me pretty dumb. The only thing I write, I don't write code anymore. I write Google Docs. But, you know, my passion is growing the data science community, and it's great to be here. Thank you. My name is Terry. I am CEO, founder of deepcafa.ai. We build AI solutions, we do research, and we also do philanthropy teaching people around the world who may or may not have the access or capital to learn. So those are the three things we do. Thank you. Awesome. Hopefully, yeah, we've gone through a quick round of introduction. I'm going to start off with the first question, and then we'll open the floor. One more rule. Whenever any question is asked, at least two people on the panel have to take an opposite view of everybody else. I'm Irish. That won't be a problem. Irish or Israelis won't have a problem with this. The rest of you can figure it out, but the idea is that we want different perspectives, right? If everyone's, yeah, yeah, yeah, yeah, yeah, it's not great learning, right? So at least two people have to take a different viewpoint. All right. So the first question for the panel. Data science, AI, reality, hype. All right, if no one starts. I think it's definitely a reality. And I mean, coming from finance is actually interesting because it seems like data science in finance is definitely one of the much harder problems to solve. And so actually in my area, I would say it's probably still more on the hype side than on the reality side, although it's really tipping over. We said someone else has to take an opposite style. Oh my God. Very diplomatic. All right. So leading up for that. Definitely hype. Oh my God. I mean, we've been doing the same thing for 10 years now, and now suddenly it's a $100,000 extra on the salary and we call it data science. And not that we're complaining. I mean, seriously, the practices that we've been developing over the last couple of years is exactly what data science is, namely to curate good data sets, to clean them, to use machine learning or other statistical methods in order to analyze the data and then get inferences from there. I agree. He must want to disagree. But I have to take a side, huh? Yeah, most certainly the hype is running ahead of the reality. As I think I mentioned in my open remarks, the consumer electronics show in Las Vegas, 140,000 people, like AI was literally everywhere. And now you have, just as Denise said, people doing analytics, like there's a real difference between analytics and predictive analytics, right? Not even forget about deep learning, machine learning, like there's analytics that's, you know, non-predictive and people are calling like, we have a predictive analytics marketing application. And that stuff is going to come back and really bite people in the arse. It's giving the whole industry a bad reputation. And so the marketing departments, especially just running away with it, they're like, going back to their engineering departments, where can we stick the AI label on this thing? And I think it's going to create a bit of a backlash, especially when you start looking at them when you're applying data science and AI across departments. People are like, oh, HR, I can use this for figuring out how to screen resumes. And the same with deep learning, like there's deep learning is fantastic for certain narrow sets of problems, speech translation, image captioning, doing some good work in finance and time series. But I'm always looking for, okay, show me the other example, show me the U-case, because I want to learn as well. And I'm not quite seeing it. So the deep learning hype is completely out of control. That's mine. So I think data science is really not a reality yet. And AI has unfortunately sort of converged with media and reporting pretty interesting things and also pretty kind of weird things with robotics. And you have, you know, even sometimes eminent people talking about it in convoluted terms that makes it very difficult for people to understand. So either they go misunderstood and or comprehend that as hype. I think this is a serious problem. I really worry about it. I'm kind of concerned because, so we're, you know, you're investing and you guys are putting so much effort in this. We should really educate the enterprises a lot on the applications and also be truthful and honest about the reality of the true application of data science in their business domains. And explaining a linear sort of projection that probably you could do with a simple SQL script in your tabulated data and sort of framing it as a data science solution. Probably it's not a smart thing to do if you're a data scientist yourself or would like to believe. I think it's important to report back on results that companies can scale. And the whole idea of deep learning becoming a hype is people are not understanding it. So essentially if you ask someone what is, can you describe what AI is? I think people start talking again in a common-leaded term. So data science is not a reality yet. If you really ask me if you want it to scale and AI is unfortunately sort of kind of turning into a hype which we don't want. So, yeah, I was surprised that you said data science AI in the same breath because I think they're two different things. And now everybody pretends they're doing AI. So I'll address each of those separately. So even data science, I think there's a lot of interesting work, but deployment at scale is challenging. And I don't think it's happening as much as maybe we like to believe. And AI I think is just hype because everybody is doing AI now. And I don't think so. One final thought. So I compare data science right now to physics in the 900s. If you think about it, when we were starting in the revolution of modern quantum mechanics and everything, it was a hype that back then people did not believe it was going to go. And there was so big people saying bad stuff about it. Good people saying good stuff that were wrong, too. So I think we're in the process of making this something very important. I think we're not quite there yet. I think we're in the process of making data science an important and serious field. Right now it's very wide. It's not well-defined and stuff. So I think if we follow the steps we're going, right now in a few years we're going to have a good field to work in. Just quickly, I would say there are areas of data science which are definitely a reality and other areas which are a hype. For example, looking at some sort of simple perception problem, speech recognition or detection of objects in images, I think is clearly the case. It's a reality. We can easily take large amount of high-dimensional information and convert that into something which is easily understood. But things like causality versus correlation or inferencing, those things are really still very hard for machines to do. So this is where we have to make sure we keep it real and not contribute to the hype by asking the right questions, asking for the scientific rigor. I believe as a discipline it's a reality and I think it's here to stay. In some areas you'll see, unfortunately, a bust of the hype, but where we'll end up will be at a much higher place than we ever were before. So I'm definitely in favor of the discipline. Can I just say one more thing? I saw the workshop today with the Bayesian network guys from Mysore and I was talking to them and they're saying, oh, not many people use this. It's really interesting for financial applications. But they say, well, not many people use it. And they said, yeah, because it's difficult. And they said, exactly because it's difficult. And I think one of the things that we found is that a lot of these tools now are quite easy to use, so a lot of people use them. That creates the hype. But it's to a large extent because it's just easy to access, but not necessarily that great. And I think that when we really move to the more difficult things, it might actually quite quickly die down. And so I think it's worth considering as well. I mean, it's technologically difficult, but it's actually, in a sense, quite easy at the moment. All right, in spirit of time boxing, which we didn't do this time, but we'll do that next time because this is my question, right? Now, for other people, we will time box things. So the floor is open now for others to ask questions. There are volunteers here who are going to walk around with the mics. So what, according to you, would be the most prominent pain point of the society that AI will solve in the near future? I'm biased. Medicine, right? So with clinicians, currently it's a, if you have a good clinician, a knowledgeable clinician, they look at your data and they make the right decision what your disease or your treatment choice should be. But, you know, we know that there are different levels of skillset out there in every discipline. Therefore, having AI is sort of a first step in order to make the recommendations and then for the human clinician to validate that treatment or regime is probably going to be a game changer for everyone. Yeah, that's a really good question, by the way, because a difficult one to answer. And I'm sitting here going, why am I drawing a blank? And we actually run a data science for a good focus area as part of a conference. So we usually have a half dozen or a dozen talks on AI for good data science, for good. Sorry to use the two interchangeably there, AI and data science. But it's a hard one to answer because definitely I live in Boston now. A lot of great work going on there. Medicine, drug research and things like that. But I'm thinking more of like single use cases. We have a talk coming up in San Francisco in a couple of months. A lady was using deep learning to figure out, I think it was in West Sahara, Africa, how to get solar panels into what homes for people so they can charge a mobile device, which is a lifeline for them. So, you know, big wins, I'm sure. It's kind of hard in medicine, right, because AI, deep learning is not a drug. It's a drug for some of you, drug for me and you. But for other people, it's not a drug, per se. But I think there's a lot of small wins coming out by applying a lot of small problems that kind of have a big impact on people's lives. I remember we had a talk a couple of years ago on using predictive analytics. You know, not that advanced stuff. Predictive analytics for helping people on diabetes better manage their treatment program and better patient outcomes. I know some of this stuff is small and incremental, but that really makes a big impact on people's lives. There's a company that just sold over a couple of billion dollars in Boston. I think CVS. I believe they were doing data science just for simple things and making people take their drug regime on time and, you know, better patient outcomes less entry into hospitals. So, I see a lot of that. That's doing great stuff. So, I think it would not solve with anything right now. I'm just playing the others part. So, I think we have a lot of hope on AI right now. But I don't think we are on the last step of AI right now. It's a lot of things to do. I mean, there's so much things that needs to happen so we can be in that space of being in the AI world for real. And this has to do with the last question. I think our companies think they're doing the AI and stuff and they're changing the world. And most of them are doing something good, but I don't think they're solving good cause problems right now. So, I think in the near future we're going to be seeing more stuff. I think deep learning will have something as I think we have, we're missing something right now. We have deep learning and deep reinforcement learning. There's something else to come that we don't know yet. That will maybe be one of those game changers for AI in the world. So, I don't think we're there yet right now but we're in a good process right now. Sorry, Tom. Sorry, Tom, I just want to add one thing. It just dawned on me, but things like Google Translate, have you ever been to a country where you don't speak the language and use Google Translate? Like, okay, it's not air-chattering, but that, like some of these changes you don't even notice they're happening, you're kind of taking for granted and then you figure out later, like, oh, I could have been lost without that. I would have survived, but, you know, sorry. All right. Can I also suggest the game changer could be used for computing? And so, obviously, we're not quite there yet, but I can really see the potential in that. And I'm an ex-physicist, so I totally believe that. I mean, because... I'm just saying this because I started, before going to Data Science, I tried quantum computing. As you would. And I mean, 11 something, a lot of things that are happening in that time for quantum computing. I think that's very far away right now for, like, doing Data Science on quantum computation and stuff. That's a long way to go. Right now, I think that could be a game changer for computation and power and stuff. But we're not... I mean, they are right now very far away. We have, like, quantum machine learning and stuff, but I don't see that many applications right now. But I think quantum computation could be a good thing, but I think it's far away right now. All right, I'm going to time box. Sorry, we have a lot of people asking questions, so we'll let maybe only a couple of you to take a shot at it, right? First come, first serve. Yeah, so one mic at a time. Hang on, deeply. Hang on. One mic at a time. Otherwise, everyone will start speaking at the same time. Yeah, so Data Science is totally based on the real world data. I will not say real application, it's real world data, and the model which is coming out and the outcome data maybe in terms of accumulation, accumulated data in terms of any model or any kind of thing which is coming out, that is also a real picture of the thing which we can use to leverage the functionality in healthcare domain or in other domain, right? So, and AI, as a layman language, I'll say it's a technology. So, we all believe in technology. So, if we plug in, if we use that technology of AI with the real data, then it means it will create an ideal scenario. So, considering that situation, can we say the future is in futuristic society or futuristic world is going to be ideal, where everybody will be good, there is no bad, or it is going to face the same situation, but what is referred in the Terminator movie? I think everybody is going to be good. So, I mean, I don't, maybe you're looking for too much from AI. This is actually probably more of the opposite concerns with AI, which is like, AI could have inherent biases and just learning pattern from the data which are rooted in sort of maybe racial biases or other discrimination. And it's very difficult to train AI algorithms to forget those biases. So, it's actually, you know, this is one of the much harder problem to solve. AI becoming good, that's actually a very optimistic picture. People will say the opposite. Yeah, there was, there was a, there's a book out, I think it was last year, Weapons of Math Destruction. And in fact, there they talk about all the bad things that models can do, these biases and discrimination, all those kinds of things. So, one has to be aware of those. I'm not supporting everything that was in the book, but I think they brought up some good points. So, I think from another perspective, it's let's try to think in preventive terms, because what we see, I counted in this whole group right now. There are 12 women and probably majority males. I think that, you know, we don't talk about ethical programming as how you know, the brains of our over qualified counterparts of our males is not being deployed. I really think on ethical programming would be very interesting field in the future, where we will program things that is taking care of how we develop things. And I think that is the way we can stop it. I don't know how the society will evolve. I hope it will be for good. But I think if we have a good distribution of women and men, then we will have a healthy discussion before you start writing algorithms. Because I really think that the world was created for both of us and all the other species. I still think the hugely male dominant society is leading to a problem which I think we may not be able to explain in terms of algorithms if it is purely male dominated. This is my personal approach. I don't know how it is going to look like. But just kind of summarizing I think it's not going to be like Skynet. It could probably be that we actually self destruct. So we have good experience with that. You know, I mean we've been quite often every 300 years. I just hope that it doesn't happen. You know, it's something we hope I hope that we learn from the past. So it is kind of scary. But I hope that we can control it by taking a preventive approach. Thanks. We'll move to the next. Hi. Thanks a lot for the insightful session. Just a quick question. So probably traditionally what used to have is we used to be very close to our code emotionally attached to it with the event of the API eras. If you search for a problem statement you'll get multiple APIs on your hand. Would you like to go with the API approach more? Or would you like to control your code and write yourself? Of course you'll get pros and cons there. You can productualize your system quickly with the API error. On the other hand you can tweak. You can better control your system. So what's your thought on that? I think there's a lot of danger. So even though I'm a huge proponent of APIs, I love to build a lot of value in it. I think myself and Tom are talking about this number of dangers. One is you don't want black box data science. We tried that in finance. It didn't work too well. A lot of issues around black box models. Another issue too was when I was first starting my data science journey I won't mention but I was using some online APIs for text analytics and kind of a little bit going back to the previous problem. If you're working on data assume your models are bias. By default they have to be because they're bias by the data set that they were built on. So bias is fine if you're working on a closed data set or something but generally speaking every single model you work on is bias to some degree because it's not an infinite data set. There's no such thing as an infinite data set. So yeah, APIs are extremely useful but it depends to me depends on what part of the chain you are. If you're an AI engineer like I was talking about and someone else has validated that API and the model and stuff like that then it's safe but if you're doing critical mission stuff and you're relying on somebody else's API and you don't understand the bias in the model and the model weaknesses and stuff like that and you're going to find them because you're going to find them the hard way. So APIs are great but use with caution. So I think this serverless sharing of services, having APIs talk to or have the components talk to each other and I think with this whole movement to open source anyway probably the black box thing is not that big of an issue anymore. So I think for me it's more this is how I envision the future maybe I can't think of anything that's kind of brainwashed into that thought pattern but yeah this is to me the exciting bit where everyone in the world can write something and have a be a component of a larger bit. I'm a bit of a purist and I like to not believe in APIs but they are really convenient so I can see them really coming but I think it's actually really dangerous as well because there's a lot of people that don't really know what they're doing which is fair enough because we can't know everything but it can really lead to some problems and I can really see this happening. Thanks for taking the opposite stand. So I have a question here we've talked lots about data, data science and all of that which means that the data is being churned, worked upon and all of that. There's never a mention of privacy here. So how is that I mean how is that angle going to be handled in the next open data science conference because the GDPR, California data protection, India coming up with its own data protection law and all of that. How are we going to handle it? We have data distributed all across you just mix and match people will be able to identify or will be able to pinpoint who it is, what it is and all of that. It need not be Google, it need not be Facebook, even a normal person will be able to do it when the data is distributed all across right now. How are we planning to handle that? I mean I don't know the technical aspects of how you would handle it but I feel that companies like Facebook they use our data that's their product, they don't have anything else. So actually every user should be paid for their data they should make a choice do you want your data, they should bid for your data and then you have a choice whether you want to give it. So I guess because you are getting something for free like when I use Gmail it's free technically but they read my email and they give me ads on the side so I'm getting something over there but in general I think they should bid for our data. Okay so I guess I'm the only one from Europe here technically you left 16 years ago so that was before May 25th where in Europe the GDPR Ireland's in Europe? Ireland, we are I hope you don't Ireland and Netherlands we have lovely partnership besides that we are super friendly for companies to come and establish in Ireland and in the Netherlands as you know so if you have a company you should come over but again I digress I think you raised a very important point so in GDPR I wrote an article some time ago you as an individual person I don't know the implications of being if you're in India or if you're in US so I think this non-challenged approach so I'm disagreeing with you really vehemently that just to provide the opposite polls that you should know that if you are operating with any European company in any sort of secondary or tertiary way your data has been used it has been assumed where European context has been used your company is liable for I guess I don't need to go into the details so 20% of the revenue and or X amount of millions of euros of fine so that's a huge problem I think the privacy is a very important thing there is a clause 13 in GDPR that says that you need to be able to explain your algorithm if you are a machine learning department if you're in European companies you could have an auditor from Europe reaching out to your sales office in Europe which you may have planned to expand your company they will come at your door they're going to ask questions if your machine learning department is not able to explain the algorithm for that one single person who has asked this question you are already in trouble so I really think and definitely that needs to be addressed so I think besides Europe we are at least trying to do some things I think US and here in India and elsewhere it's pretty disturbing how data is being used so I think we should pay attention to that just really quickly very good point but we need more sessions around that because people think oh did ethics how hard can that be because a lot of stuff you wouldn't realize are non-ethical the engineering may not think of these issues speaking of AI for bad, data science for bad a lot of the bad stuff coming out is going to be around privacy because governments are going to use this to manipulate your data to manipulate your voting and stuff like that so there's endless places of risk so it's a huge issue and it's going to become more and more of an issue as time goes on for sure so at Bloomberg we carry data from all our trading clients people have portfolios and also a lot of their trade ideas are going through our systems so it's very very critical for companies like Bloomberg to really take privacy very seriously and basically we unless they have given any sort of permission for us to use the data we will not even touch it so it's very difficult to fully ensure within a close company with Bloomberg we can do that but in more general public data and other settings it's very hard to make sure but with GDPR and with regulations even in the US and what happened with Facebook and so on I believe it's going to be very very crucial in taking care of Next question Good evening everyone I notice a lot of hype and noise around preventing fake news in research but frankly we're not even close when it comes to production level or real time so with elections coming in next year in India do you think a credible candidate is winning it or is it going to be another failure of technology that will be winning the elections this time again What do you mean? What I mean to say is though there's a lot of noise about fake news detection and stopping it in research in real world we still don't see it and I'm saying with elections as early as next year do we see a credible candidate winning it or a failure of technology or the candidate supported by technology and a technology not being able to stop spreading of these fake news winning it I live in America there's no such thing fake news there's supposedly Robert Giuliani said there's no truth anymore truth isn't truth so I can say whatever I want yeah we have some we definitely have coming up in our US conference we have one on fake news but I gotta ask you what is fake news, why is it fake one person's fake news is somebody else's entertainment like the national inquirer has been around for 40 years you go into any supermarket in America it's fake news it's been around fake news has been around since the caveman you see that woolly mammoth out there and that's the problem in serious I know what you're saying but it is a very hard problem to crack because there was a section yesterday how do you detect sarcasm very difficult NLP problem and how do you detect fake news and then in a place like America which protects free speech okay I've identified it fake news now what so it's really it's not just about actually just a rant on about this because I'm in a rant now it's like you're relying too much on technology at some point we gotta take it's a human element you gotta take control, you wanna do something fake news then do something about it don't rely on software or AI to solve it yeah so in my mind the way to handle this is really education to teach the general public what is how did it come about, what are bots how do they influence what kind of scope do they have and I think some media outlets are doing a great job at that and we just need to see more of it so with advancement of AI, life of human will be more beautiful, I agree to it but on another head there are many people who are doing some skilled work, skilled labor or they are doing some manual work just like a driver or person who is not technical guy but as a working as a labor in some industry so their jobs will be in danger definitely because with advancement of AI some machine will able to do that job and as a data scientist how we are going to solve that problem because on one side we are making the life of human much more beautiful but because of that on another side jobs of some human which are not data scientists and which are not technical guys they are common people from society their jobs will be definitely in danger so as a data scientist what are our approaches to solve those problems I think you raised a very good question the only way to solve this is to ensure that the startups or companies or the expansion and inclusion of technology that is there in machine learning and or deep learning systems should ensure that it is enabling employment and there are ways to do it I can give you an example I was last week I was in London mentoring around 40 startups who were actually focusing on the theme is energy but it's for social good so in this case there is one of the great companies farmers who have this problem of drying of rice and so essentially the nature is unpredictable it rains all over the time they don't have the drying facility to dry rice and it's a havoc in the life of a farmer because he is essentially not sleeping because they hear the thunderstorms and they have to rush or try to cover so it's really panic stations so there was one startup they came up with a solution in which they have this hardware a device in which a real appliance in which you could funnel the rice and it dries the rice using solar energy and essentially stores the rice for the farmer in a way that the farmers can start doing easy stuff as in they can continue farming they also propose a concept of micro financing slash even potentially proposing a rice bank that was great because this is going to enable farmers it's not going to just deplete farmers and you do some you know so the whole idea of what I'm trying to say is that the whole idea of trying to use drones and we're going to farm is we should not kind of get into that and try to use technologies and use of technologies and enabling these making predictions and understanding how you can farm rice dry them as most of you would know the older the rice the more price point it has the farmers can actually potentially going from a disaster situation of drying rice and oh my god my whole year's crop is gone because it rained again they become profitable and using AI or what should be using AI using deep or machine learning solutions they can inform themselves of the equity they have with that rice which is essentially being sort of monopolized with other their own villages they have now visibility into the asset so I think and there were 40 other stories I can tell but we don't have time for that so I think creating and embedding these deep learning or machine learning solutions to enable these people to actually encourage agriculture getting more people getting back into farming it's going to actually increase that you know people in those areas and solve a lot of other problems like urbanization etc that we all suffer that's why Mumbai and other places in the world get overcrowded so I think there are a lot of things so I think we need to give it a thought and try to encourage each other to start startups that focus on solving these problems at the community level for that human being and that's really going to help it's a great question the same concerns were raised back in the time of industrial revolution maybe in the 19th century when we had invented and electricity and cars and people were worried like they're going to lose their jobs but what happened was really people really became more productive the technology enabled them to become more productive not really lose that many jobs in fact they moved on to doing more sort of higher level jobs became doctors or engineers or more thinkers and stuff like that as Tari was saying you know it's essentially sort of even more productive and potentially find better kind of profession or education education definitely is a concern but we can move on to become more intellectual sort of thinkers if technology is enabling us to do sort of more of the operational work very well you need to kick me off the stage soon because I keep talking I think there's a lot of discussion around this is really bad because I remember after financial crisis I'm back in Ireland Ireland had a huge building boom and they were building like 100,000 houses when the only demand was for 10 and all of a sudden everyone in the construction industry was laid off and the government minister came on and said we're going to convert all these carpenters and labourers into software engineers like complete freaking bullshit like not everyone in the future will be able to come because the problem is you see it in the states a lot you have these low skill jobs and then you have up skill jobs and a lot of wealth is transferring to the up skill jobs and that's great for most people to do it but the people who can't do these jobs you can't just simply up skill everybody especially in their lifetime you know that's what's scary and I'm pretty passionate about this I really think this is where government needs to come to events like this they need to understand machine learning, AI and they just need to get involved because at the end of the day I know it's like again technology is not going to survive in but they need to kind of step in here and figure out get ahead of this problem because it's going to be a huge problem Alright we are out of time we will take one last question and we'll go to the lady over there just to make sure that next conference I'm going to have to stand Hi, so I saw Sophia do you believe that first of all do you like her and if yes, do you think that we need more Sophia's in this world okay, so first of all research about Sophia when she comes to a place people need to give the authors the questions beforehand she's not answering everything you want and they reply to you with some approved questions and you can only ask her those questions so that's one example of a huge hype on AI because we think that this is a human thinking machine that is answering everything we want and it's funny and I think it's a huge advance it's a huge advance in technology and something looking like a human being stuff and there's no way to stop that I think we will be seeing more humanoids in the future hopefully they will be more intelligent and so I think right now we're not there yet so Sophia is not that machine we think she is but I mean I like her I watch some of her videos and they're interesting and stuff and I think it's inspiring to people who can actually create these systems for real so I think it's a good thing that she's there inspiring other researchers and engineers to create this new thing that will be helping more people and actually being a part of society because that will be a thing we're not thinking on it but in some years these humanoids will be part of society they can help us they will be able to do some things for us so it's better that we take them seriously because there's no way to stop them I'm gonna say really bad things I think it's the most stupidest thing that you you've seen on internet and it's the most has anybody seen Sophia walk you want me to do the imitation here really do you want it so this is how she's walking and you know Boston Dynamics they're laughing their asses off this is Sophia I mean I really think probably some kind of AI drug so I think I really think handsome robotics and a gentleman who's investing millions of dollars if he put seed money into those 25 or 30 to 50 startups really solve this problem with the sale I think we would be in a much better place it's really unfortunate that you know people are investing in it please do not actually I would encourage you to not ask a question and not give this weird mechanical object with ultimate stupidity any kind of attention I really think so because I think your job is to to develop systems and the next time you come to ODSC you're probably creating a simple robotic feature that helps elderly who are especially you know falling off in the bathrooms and having hip fractures and those are the things that can really help you do really interesting robotic things you know having some stupid ridiculous dumbass robot is not is not what we want we don't need we have like what 9 billion people by 2050 there's a lot of potential I'd love to see a beautiful woman walk and with this you know Sophia was kind of yeah probably drugged sorry for the French alright we've killed it alright I think we should the last question special question thank you yeah actually I want to change the dimension completely and then ask a different question let's see I'm seeing this discussions from a business perspective from the business perspective at the end of the day I want to make more money the data science problem I'm solving or a problem or whatever it is what model I'm creating doesn't matter I want to make money so money how will I get it's not only by proving I solve the problem by a POC I want to replicate it and scale it wherever I could as fast as I could and automate it so that I can make more money so if I'm leading an analytics department the most scariest thing is the landscape is keep on changing by the time I make more money or make it stabilized the landscape is changing I'm not saying it's wrong but after the landscape change it's much faster and smarter so I can't ignore as well so especially the question is for CEOs what is your perspective around it and if a person is leading an analytics team what kind of advice you will give to that person I think well it's actually really interesting question in a sense that in finance is exactly what happens all the time the landscape is just changing and while you actually deploying the systems by the very nature of it the landscape is changing and it's changing much faster than it even three years ago and a lot of people really struggled with that especially people that have traditionally been really successful they're suddenly not successful anymore and I think and it's really interesting to see how the people react and some people will really not be able to keep up with this future pace and other people they just jump on the bandwagon and somehow get through it what I notice is and also for myself working in this there is no easy answer the only thing is you can just hold on for dear life like in Mumbai when they hold on to the trains and just trying not to follow and really this is it's evolution we are as humans not fully evolved right evolution keeps going keeps going and there's no way for us to stop that and no government can do that no anyone what we are seeing is evolution right and I don't think there is an easy answer if any answer at all I think the at least I'm a CEO of my own firm and actually heading to more firms that we're launching and I've worked with chief executives and chairman of companies between 25 to about 50 billion dollar capital and revenues so what I tell them is yes it is changing landscape the best way and you're seeing it already right so the way things like class and agile and pivoting also in your projects and creating a sort of minimal viable projects for solving those problems you already see actually if you are working with larger enterprises that the budgeting cycles around the period of November October November when you would essentially agree working with procurement that well this is where the budgets are you should make sure that you have a large that's change basically it's on demand basis and enterprises the ones at least I work with for example CEO a banker German banker who's now CEO of a large manufacturing company in the south of Germany so he asked me the question so well how can we get more competitive so you can do you know kind of basic and intermediate stuff you can also do advanced things like you know if you want to machine learning and start using it and applying it to material sciences to all this you know the ball bearings that you create you can also develop new alloys I mean that's not something new but the only thing is you can get much better position you can actually go to new markets for instance if you're you know serving ball bearings in Doha or Dubai or even Mumbai the wear and tear of your machines and the product is much higher so you need to develop come up with a new alloy so that's really kind of break next speed and probably there you can sustain your margins on the basic and the intermediate you will be caught of the competition catches up with you and obviously the invention and the evolution of deep learning can overwhelm you so I think you I would suggest enterprises is to place their bets in the basic little projects low hanging fruits and and a little bit in in your intermediate project with some volume and have at least one or two sort of big bang for bucks slash you know B H A G you know big hairy goals in order to kind of you know have your to get a key game competitive edge and it seems to work with a few and and you tend to get you know it's a matrix organization you should realize for bigger companies so one of the chairman of the companies I work with has himself given a lecture of machine learning to his thousands of employees the other are kind of teaming up with even unconventional companies like Tencent and Tesla and learning from them so I think I would suggest those kind of strategies in order to sustain your margins and continue to be profitable. Just a quick word I think it comes down to execution right execution is everything in business and really keep an eye on the you know what's exciting for the data science landscape right at this moment is a thing called data ops data ops came out of not so much data science but data but keep an eye on data ops and things like Kubernetes because the life cycle of a data science project is getting quicker and quicker we were using things like typical rendezvous and finance 15 years ago when all of a sudden like streaming analytics is big now but depends on them and maybe not getting real-time data so you know I speak to a lot of data sciences and I talk about model deprecation because again coming on finance you actually knew your model in the shelf life and people would just know you haven't ran this turn detection model for the last three years and not doing this anymore I'm like three years are you kidding me you got to refresh it so when I say execution like look at the data science workflow and a lot of smart companies they're figuring out okay we're going to use AI engineering here we're going to use the best to breathe infrastructure production right you're always looking to validate them you do model monitoring that's where I was trying to get that point across very quickly yes I was talking about it like there's so much potential here in data science for data engineering you know or AI engineering where we want to call it because just just having people who know how to model the monitors no one they're deprecating no one has time to swap them out like how do you will do a hot swap of a model like the thing can't even be down for a few minutes if it's a pricing model or an auction model or something like that so a lot of it comes down to having a really good if I was giving advice to a CEO like it comes down to people you're you know you can hire one very smart person with a PhD from MIT or one of the ITTs here or whatever but really you need a team of people who know what they're doing doing and you need the data science department and the IT department to kind of work together that's why we kind of need these AI engineers and stuff like that so I'm a firm believer in that thank you one last word so I think what Seymour say is so true and being a data scientist combining data ops with innovation and an agile framework for working I think is the best so you can stay up to date every time and this means being able to ship intermediate results I mean if you wait for a project to be finished to be able to send it to the market it's so late now so you need an agile process so it can be fast it can be you need a systematic way of doing things fast and it's the only way you can like make that time to market period short and you can follow along with the whole company and for that you need a good team you need good people and a good process I mean if you have good people but you don't have a good process you're doomed all right that's good words to stop the conference you're doomed doomed all right thank you very much I would appreciate everyone on the panel for staying with us and keeping holding back so many people not many people escaped out which is a good sign so now I'd request the panel to identify three people who gave you the hardest time three questions I like the last one question what if someone even had a harder question agile right you have to be first to market I need the three questions I like just myself gentlemen over there who asked the first question the data science for good one and the privacy one for me was the last one I like the privacy question and I like the one the first question about the what would change I first question the one about the pain points and how I will stop them you said it right I can't agree with the three the business the privacy and the AI question right so we have winners yes cool do we have privacy this is dangerous for privacy thank you I would like to have Sheamus share the last few words about the conference the community yeah so I got to say again actually I don't know what to say for once somebody asked me a question quick and yeah I got to admit I've been pretty blown away by ODSC India and what the rest is done here and I said this a few times before we've had conferences in different countries around the world that's one of the great ones that have been involved in ODSC and been invited to other conferences and I've been really blown away here because thanks to all of our wonderful speakers and the others are not here they're really what I call debate right they're debate and you guys are what make a conference you can go at home you can be on Corsair if you learn this stuff here but you come here to listen to their insights but it's bringing this community together which makes this event so great and I'm sure you've had those experiences speaking to each other and ODSC as I kind of mentioned briefly we started from a meet up I love the whole idea of meet up this free open exchange of ideas open talk format and this year we've been around four years this year should be our first probably year finally as Nuresh said this stuff is a bit of expensive hobby one of the ways we're giving back to the AI community we get some money to Chainer Python thing we get some money to AI for good cancer research in Boston we got another award coming up so we do have a grant award by the way in our website so if you look up ODSC grant we're doing scholarship tickets but I really kind of want to go back to my roots of doing meet up so one of the things I'm going to stand this stage next year and I want to have at least six meet ups running in India because again a lot of people can't afford to come to conferences so if we're going to be through to our goals and our ideals I want to start running do more meet ups here in India and not in the sense that oh great one more meet up that's all we need just to work with other local organizations other meet ups so if you guys are interested in helping ODSC and volunteering we'll hire a resource we'll get support for you but we'll love to start meet ups here these conferences are just blow into town make a lot of noise and they're gone in two days and that's it data sciences you guys know there's lots of challenges there's lots of opportunities we just want to be in the mix like as I always say the slogan is in the back of the t-shirt the future of AI is here ODSC is not the future of AI I'm not the future of AI you guys are the future of AI and as they say in the agile process we just want to be the facilitators it's great fun it's wonderful stuff so once again thanks Nuresh for organizing this you did a fantastic job it's just not me there's a lot of volunteers and a lot of people when I thank you I'm really thanking the volunteers because I know you guys and ODSC it can be a rough organization he's putting us to shame like sending drivers to pick up speakers and stuff like that you just have better than we even do it but again thank you to all the speakers I got amazing feedback from all the talks and that's what ODSC is about because we would never be growing like we are if it wasn't for the speakers and the quality of the speakers and it takes a lot of effort to put these talks together to travel for three days and all that so thanks to the speakers again thank you yeah so Nuresh I will ask you maybe at some point I know you're going to go home and sleep for a week but we'll send an email so if you want to get in contact with me we'll send an email to all attendees to contact me back about the meetup stuff but the last people I want to thank is you the audience because again you guys make the conference so thank you for coming thanks for investing your time thank you we're going to be sending out an email where in the beginning I said we don't do feedback forms that was more we don't do feedback forms during the conference where you're kind of filling in some stuff but when the conference is over we do want you to like sit back and think about what was valuable to you what kind of speakers you really like what topics you like if we were to do this again what should we focus on what should we improve so we kind of try and give that and hopefully there's constructive feedback which basically helps overall the conference to improve so you would be getting shortly an email which will have some of these details for you to fill out which will basically it'll again be all public so watch what you fill in there the reason I mentioned that is not to discourage you from filling bad feedback we actually appreciate that but just watch out in terms of the language you use because whatever you fill in is going to be publicly visible to everyone alright so that's the feedback that's going to happen after the conference and like Ashimu said if you're interested in running a community then please contact any one of us and we'd be happy to kind of see how we can facilitate that any other questions you guys had open data I think apart from one session most of the sessions were about data science and so maybe if we can do more I don't know if science is being open data is getting closed people are thinking it's a cold mind so maybe more sessions and more interaction on that it's interesting the name open data science there are four different interpretations of that name so we had companies reaching out to us saying we want to present at this conference but we don't have any open source so we are not going to be presenting at the conference we had people approaching us saying it's about open stuff so only if everything is open then we can present then we had people who were saying open data which is just like about open data in fact we had a panel that was planned for today which is about open government data unfortunately the person who had originally written the paper could not make it so we had to change the panel but anyway the point I was trying to make is there's a lot of different interpretations of what open data science conference means but you have a valid point that we need to include more topics on open data in general and some of the privacy concerns and stuff like that associated with that cool? a court jam alright so you are volunteering to run a court jam next year alright perfect you got it on the camera so this is all completely volunteer driven so when we announced the conference we also called for an open program committee open program committee as in truly open program committee so a lot of people actually at this conference who organized this conference I've met for the first time in my life right and that's because there's an open submission system through which people come in and put their name saying hey I want to be part of the program committee and basically our philosophy is to accept everyone so there were about about 33 people who came in we accepted all the 33 people as time passed by bunch of them were completely inactive so they get dropped off and then I think we were left with 14 people or 16 people I don't remember the final con so the point I was trying to make is that if you want to contribute to this conference it's an open submission process you can come and submit and then you know you can take that the second way is basically we also do open submissions for the talks itself so a lot of speakers who have come here have actually gone through our open submission system where they can go and nominate a topic so again you know like I want to run this particular thing or I want to do this particular thing could be a great submission in this process so that's how we could get selected right so I'm just trying to hint how you could contribute and how you could get involved because quite a few people were asking about that so it's all a community driven conference and people can come in so whenever we announce the next conference watch out for call for program committee call for proposals various different ways in which you can involve and influence the program there's also voting thing all of this bunch of different options through which you can influence what gets into the program alright I think it's been a long two days for you guys a long three months for us so we're going to call it a day thank you again as Shemu said for me the success of a conference is not what a great speaker lineup we had but the success of a conference is what a great attendees we had who you know many speakers said the quality of the attendees was really high right that's what makes us happy so thank you for coming to the conference and keeping that engagement really high keeping that curiosity really high thank you very much of course we do need to thank our sponsors without whom we couldn't run this conference we're still not financially profitable ODSC as a conference overall the India one we probably will lose some money in this year's conference which is perfectly fine but to a large extent the sponsors reduce the damage for us so we do need to thank our sponsors we don't do much for the sponsors as you would have seen right a lot of people came and said this is one of those conferences where we don't see a lot of marketing right which is great for the attendees not necessarily for the sponsors but we still take the stand that you know this is not a conference where you're going to basically buy a speaking slot right you don't go sponsor and then you get a keynote slot the three keynotes that we had none of them were sponsors right did you guys enjoy the keynotes they were absolutely people who deserve to be on the stage not people who could buy this slot right like that's our philosophy so again we want to thank our sponsors because they sponsored the conference even though they knew there is a not a lot they will get out of it right that's that's again like in the true spirit they're trying to build this community so really want to thank them again alright thank you sorry to keep you holding back