 Hello and welcome. My name is Shannon Kemp and I'm the Chief Digital Manager of DataVercity. We'd like to thank you for joining the latest installment of the Monthly DataVercity Webinar Series Advanced Analytics with William McKnight. Today, William will be discussing the impact of machine learning on the enterprise today. Jessica will have points to get us started due to the large number of people that attend these sessions. He will be muted during the webinar. For questions, we will be collecting them via the Q&A in the bottom right-hand corner of your screen. Or if you'd like to tweet, we encourage you to share highlights or questions via Twitter using hashtag ADVAnalytics. And if you'd like to chat with us or with each other, we certainly encourage you to do so. Just click the chat icon in the bottom middle of your screen for that feature. And if you'd like to continue the conversation after the webinar, you can follow William and each other at community.datavercity.net. And as always, we will send a follow-up email within two business days containing links to the slides, the recording of the session, and additional information requested throughout the webinar. Now, let me introduce to you our speaker for the series, William McKnight. William is the president of McKnight Consulting Group. He takes corporate information and turns it into a bottom line producing asset. He's worked with major companies worldwide, 15 of the global 2000, and many others. McKnight Consulting Group focuses on delivering business value and solving business problems, utilizing proven streamlined approaches in information management. His teams have won several best practice competitions for the implementations. He has been helping companies adopt big data solutions. And with that, I will give the floor to William to get today's webinar started. Hello and welcome. Hello. Hello, Shannon. Hello, everyone. And welcome. And welcome back. Many of you, I see that you've come back for each and every one of these in the series. And I do appreciate that. The topic today is kind of a pickup on last month's topic. Last month, I talked about artificial intelligence and getting data ready for artificial intelligence. That remains a very hot topic. Today, I'm going to be talking about the impact of machine learning on the enterprise today. Last time, when we got to the Q&A, surprisingly to me, I got a lot of questions about the impact of artificial intelligence on life in general. What's it going to mean to jobs? What's it going to mean to life and relationships? And what about bias in the data and all this sort of thing? And are we headed to big brother and all this sort of thing? And I was a little surprised, as I said, because I kind of gear these presentations towards practitioners. But you know what I learned? I learned that practitioners are people too. And so I'm going to let my hair out a little bit this time and maybe address some of that because there's a lot of carryover from artificial intelligence to machine learning, as we'll learn today. I also want to reiterate what Shannon said about coming on into the Dataversity forums and other places on Dataversity.net and catching up with some of the things that are going on there. I'm going to jump in the forum later today and answer some of your questions, pick up on some of the conversations there, and contribute. And I'd suggest that you might want to, as well, some good stuff going on in there. Now, I'm not an expert at machine learning yet, from an implementation perspective. I do understand the process pretty well by now. I do understand all the algorithms that are available, or at least the major ones. And I can size up a problem into machine learning or not pretty quickly. As a matter of fact, as I take a look back at the enterprise challenges that I've been presented with over the past few months, I'll say this year, I can say that 50% of them or so could be solved with machine learning. And that's been my stance around how these enterprise challenges could be solved already today with what's available. 50% of them could be solved with machine learning, but then we get to some of the gory details. And that has to do with the data. And the data is not ready to solve all these challenges with machine learning. As a matter of fact, that probably takes the whole thing down to maybe about 25% of enterprise challenges today that could be solved with machine learning. So that's still a pretty high number, I would say. And it's going to grow. It's going to grow fast. As a matter of fact, a lot of you are involved in digital transformation projects, so-called digital transformation projects. And I believe that that's because the technology has changed so rapidly recently. The possibilities have gone up quite a bit to gain efficiency, to gain capabilities within the organization. And there's a lot of catch up that organizations have to do. So they're categorizing their problem into this area. As a matter of fact, it's a whole new thing. It's a meme, really, digital transformation. I did a webinar yesterday and I was talking about digital transformation. I asked my audience, how many of you are involved in digital transformation? 79% said yes, that they were involved in digital transformation efforts. And I don't recall when in my 20-year consulting career anything has been so high. So there you go. And I think a lot of the digital transformation efforts that I see are really to get organizations ready for machine learning and what it's going to bring. So I'm going to throw a few statistics at you just to reiterate what I've been saying here about the importance of machine learning. 85%, this is from Forbes, by the way, 85% of US CEOs and business leaders are AI-optimists. 87% are investing in AI initiatives this year. 87%. 82% expect their business will be disrupted by AI to some extent within the next three years. Sounds about right. Might be a little bit higher, actually. 29% of companies are making regular use of artificial intelligence. And I would say that a lot of that is maybe not centralized, but maybe located in a remote, I should say remote from central department and so on. 41%, and by the way, I am still on the title slide. I am still on the title slide. 41% of organizations are planning to invest at least $500,000 to support AI initiatives over the next 12 to 18 months. And the federal government is going to spend a billion dollars in 2019, a billion dollars. Now, let that think in. As inefficient as you might think that that spending might be, that it's definitely going to move the needle in terms of their abilities with artificial intelligence. And in my forays into the deal flow scene, from what I have seen, you're not really getting funded anymore today. If you don't have a great tech story, if you don't have a great story around how you're going to use these capabilities of machine learning and artificial intelligence. One more spat, then I'll move the slide and we'll get on. Nearly eight out of 10 enterprise organizations currently engaged in AI and ML report that projects have stalled, 80%, what's that? And 96% of these companies have run into problems with data quality, data labeling required to train AI and building model confidence. So I just want to underscore all those challenges with the data. If you want to learn much more about that, check out the presentation from last month. Now, let me set a little bit of context here because I might use AI and machine learning interchangeably. AI is a broad concept. It's been around quite a while. Of course, it's about building smart machines. There is applied and general AI. Applied is for more or less a dedicated purpose. Machine learning, sorry, let me continue on that, artificial general intelligence is really about building intelligence comparable to the human mind and going perhaps well beyond a human in terms of intelligence and applying intelligence much more broadly than in a specific context. So machine learning now, that is a subset of AI, but I must say it's the biggest subset by far. And it's used quite interchangeably out there. I have to always have the glasses on when I'm talking to someone about this stuff and trying to understand are they really just kind of using the terms interchangeably or do they mean machine learning algorithms, which is where the algorithms reside in machine learning. But machine learning, let me go on a little bit more about this and give you some different dimensions on how to think about machine learning. It's learning by example. You want to give the machine examples with data and let it learn from those examples and carry forward. It's taking down things that we've been taking days to do all the way to seconds without coding, by the way, just use of the algorithm. So you need the data as I can express enough and you need the algorithms. You can just select a library of algorithms for use and there's a process involved as well. But mainly those are the things you need. It enables machines to make decisions that are informed by data. Another way to say these things. And it's model-based and it's getting close to thinking or beyond thinking. And you should be aware of the Turing test. The Turing test was developed by Alan Turing in 1950. It's a test of the machine's ability to exhibit intelligent behavior equivalent to or indistinguishable from that of a human. Many of us know about the Turing test. Okay. Well, we're seeing that the Turing test is being exceeded in so many different ways. I'm going to give you a lot of examples here today. I want to stimulate your thinking about what artificial intelligence and machine learning mean to your enterprise. I want you to try to apply some of the examples I'm going to give you, some of the ideas I'm going to give you to your enterprise. So keep the thinking head on as we go along here. Machine learning is about letting machine learn and it's fueled by data. Deep learning is a further subset of machine learning and that is when you apply machine learning in layers, many layers of machine learning. Neural networks is the key to teaching machine learning which classifies information and simulates how a human might approach a problem. It's a subset of machine learning. Machine learning is really a shiny new term as I've been saying. We've been using it interchangeably with artificial intelligence for better for worse and that's okay. We just have to know what we're talking about here and the other big category of artificial intelligence is natural language processing and I'll have more to say about that in a future webinar. For today, stick to machine learning. Machine learning changes everything and I thought about this. I don't like to be kind of over the top with things but I really believe this. It's changing everything that we're doing. I already talked about how many enterprise initiatives can be managed and solved outright solved or highly supported through the use of machine learning. So historically and still what many of us do honestly, we use spreadsheets. We overuse spreadsheets. We use spreadsheets to document what are all the what ifs. What if I put a little bit more into the Facebook marketing program? What might happen? Well, we can project with our spreadsheet. We have to put in the formulas and whatnot and it's kind of still a shot in the dark because it's not really smart but we can project numbers and after a set amount of time, we get to some quote-unquote answers that we go with for better or for worse. Well, what artificial intelligence can do to thinking like that is to add in all the possibilities. You're not just looking at this ad. You're looking at a thousand ads. You're looking at running them in a thousand different ways. What are all the possibilities there? As long as the date is available, it can tell you and tell you pretty quickly. If we add in deep learning, it can go into very different variables that we can't even think about. You know, in a spreadsheet it's two-dimensional. Okay, we're going to be limited but in deep learning, we can add in all kinds of variables and let it think about what these variables are and do all that without coding. So machine learning can figure out every possible what if and have all the scenarios thought through for you. Deep learning adds variables you can't think of. It's very fast. So you may be a data professional. You may come to a webinar like this because you know I'm a data professional. However, we as data professionals, we have to know where our data is going and what it's going to be used for. At least that's what I think. And historically, it's been a bunch of dashboards. It's been a bunch of reports. It's been analytics and so on. But now largely it's going to be shifting over to artificial intelligence. So we have to know certain things about artificial intelligence and machine learning and that's what we're here to talk about. We did talk about the data last time. We talked about having a strategy for your data outliers. These are these oddball data points that don't seem to fit. And maybe they're wrong. Maybe they're right. Maybe we don't want them in the mix in terms of coming up with the computations or not. So we have to have a strategy for that. Either we drop that data or replace it. We need data split into training data and then testing data. Talked a little bit about that. We don't want to overfit the model. Therefore, the model would be too complicated to fixated on the data that we give it, not general enough. We need to add data to that scenario or what we call regularize the model. There may be an underfitting problem. That's the opposite problem where the model is too simple. And here again, one thing we want to do to a situation like that is to add data. There we might also want to add some complexity to the model. The goal is to come up with the best fit. But keep in mind that it may never be completely ideal, but there gets to be a point where you can go with it. You know you're not overfitting. You know you're not underfitting. And we also want to eliminate bias in the data. So we don't want the data screaming out different biases that we may or may not really be able to deal with in the real world. Now let's take a look at some examples. Now let's take a song for example. Now there's many different variables around a song. There's tempo and song intensity. Let's just take those two. So let's say that William likes the songs with high tempo and high song intensity. So maybe the four in the upper right quadrant if you will. And Shannon likes the ones that are of low song intensity and low tempo. So she likes the three in the lower left quadrant. Well here comes the song. Let's just say it's William. Will William like that one in the middle? Well it's kind of close to the the high cluster that I decided I did like. But it's also kind of close to that low cluster that I said I didn't like. So what do we do about that? Well there's an algorithm for that. K nearest neighbors, K being the number of groupings that the data collects itself into. So we got just in this simple model we got the data that the songs that I like, songs I don't like. Which one is the closest to? And what we might want to do here is to draw a circle around that song and then take a look inside the the circle. It's a smaller circle than all the data that we see here. Okay how many songs in that circle did I like versus not like? And that might tell the tale of whether I like that song or not. So you can go by the closest or you can go by a majority that's in a circle. And that's again K nearest neighbor. So I want to introduce that algorithm to you initially. But let's go into some different things now that artificial intelligence and machine learning are doing out there. You might be surprised. I don't know I could I'm surprised every day frankly in my following of this marketplace. But here's some of the things I wanted to share with you. I know it's not all completely you know enterprise oriented. But a lot of the innovation is coming from startups today. And I think that this is going to be artificial intelligence is going to be the driver that's going to change the S&P 500 dramatically over the next let's say 30 to 50 years. As we all know the S&P does change. It has changed due to innovation. And this is the thing this is the this is the area of innovation is going to change it. So what about whiskey makers? Yeah, well as part of the distillation process whiskey spends time typically years sitting in charred wood in cast which turns the clear liquor into a darker color and gives it a unique flavor. But how long it stays in the cast and what the cast held before helps create a specific recipe. Now I hate to read but let me let me go on this a little bit here because it's important. The distillery is feeding its existing recipes sales data and customer preferences to machine learning models. So the AI can suggest which recipes it should make next. Generating more than here's the punchline 70 million different recipes 70 million different recipes. Now do you think that someone could sit here and think of 70 million different recipes and then take from the best among them? Well this algorithm is highlighting those that predicts will be most popular in of the highest quality based on the cast types that are currently on hand. So it's not impractical. So AI is now shall we say designing whatever the word is and distilling whiskey. Okay something something that hopefully sends off some alarm bells with you in terms of the possibilities. All these things I want them to do that. What about you know something that you might say well this is kind of maybe a little bit in the creative arena. What about a painting? Artificial intelligence can now draw paintings that are equivalent quality at least in my layman's eyes. Equivalent quality to the ones that are done by you know real painters. What about that? Well here is something I want to play for you. It's a song we're going to take a quick song break and just listen to these bars. Okay pause right there. Okay now this is this song was developed by artificial intelligence. The AI system called flow machines work by first analyzing a database of songs in this case it was needle songs and then following a particular musical style to create a similar composition. So you know if this song came on where songs come on you know in a store or whatnot I wouldn't blink an eye and say oh well that clearly was that clearly is weird you know it's it's a song you know and and this actually this song is actually a couple years old and there's so many other songs in so many different genres now that are being created by artificial intelligence. Now we already were down to all popular songs were created in a very tight process by very few people. That's another story but now we're looking at artificial intelligence being able to do some of these things. Now a person did write those lyrics but I think that based upon some other things I'm going to share with you that may even fall. That domino may fall as we go along forward. Now this one hits close to home. Whoops didn't want to do that. This one hits close to home. This one hits close to home because I like to think that a lot of what I do is based upon the reading that I do which is fairly extensive in this space but now robots can now read better than humans putting millions of jobs at risk. Okay well an artificial intelligence algorithm has outperformed humans in a reading comprehension test. So the you know think about the Turing test right. The AI algorithm developed by Alibaba okay outscored humans and the result can have significant impact in introducing the technology into roles typically performed by humans as the AI algorithm can provide precise answers to questions when provided with vast amounts of information from resources like Wikipedia. Now you can insert your insert a comment here I suppose about how Watson how well Watson did at Jeopardy. It's completely analogous but we've let our reading skills slide a little bit but the machine has certainly not and think about all the industries that this can impact. Let's say you're doing legal research. Let's say you're doing medical research to determine next best procedure and you want to read all the papers that are out there and come up with something. Well AI can do that for us now and all of that reading all those reading jobs may be put at risk but you say I can design AI at least I can do that right. Well Google's new AI designs AI better than humans could and what Google did was some of their engineers created an AI system that can spawn new AI systems which are more sophisticated than what humans can design. Google research scientists release the system to identify objects in real time with remarkable accuracy such as what you're seeing here on the slide. We call this auto ML or they called it auto ML auto machine learning. Now I don't have personal experience yet with auto ML but what it does is goes through the algorithms of which I'm going to share with you here in a bit and it's going to go through those algorithms and figure out which one is the best to apply in a given situation and so it's AI doing AI now I don't know where we were this all ends okay if AI is going to do AI which is going to do AI which is going to do AI okay I don't know where it all ends but even things like designing AI systems are kind of being done by AI today so go go figure. Now in this slide we see various people that are identified kites that are identified but in some of the other studies that I've done not only can AI identify these people as people but the specific people that they are I don't know that we're getting a good look here at faces in this particular picture but when that is the case and we know that's the case so often on city streets and so on we know we can be uniquely identified today with what AI what AI can do okay now other areas AI assisted for example is going to health breast density measurements are already in use for screening mammograms performed at a general hospital helping predict more accurately a woman's future risk of breast cancer and they're I mean go on and on about you know medical here's another one a health checking robot takes just three seconds to diagnose a variety of ailments in children including conjuvitis and hand and foot and mouth disease and so on over 2,000 preschools in China with children aged between two and six are using walk lake I guess that's the name of it every morning to check the health status of their students I think that's coming to all of us right and think about think about your you know primary care physician and you go in you get a physical and you know that can be done now so quickly by analyzing us with machine learning what about that well AI acquisitions it's the big guys that are interested in AI and the big guys that are acquiring AI startups starting with Apple Google Microsoft Facebook Amazon and so on IBM's in there as well at some point so you know if these companies are acquiring them you know it really does merit our attention and this is a huge area of focus for all of these companies and more so let's talk about it now that we've got some examples maybe we're a little uncomfortable with it all already but let's stay in reality mode here and talk about what we can do machine learning supervised learning this is where your data is labeled you know quote unquote labeled you're giving the machine data you're saying this data resulted in let's just say success this data resulted in failure now what about this other data that's coming along here well with quite a number of data points under its belt it can predict whether that is leading to a success or failure condition of course I'm oversimplifying how about though another simple simplified example you share with it pictures of I don't know cats and you say these are cats you share with it pictures of dogs and say these are dogs and then you can present another picture and it can pick pretty well depending upon how well how much data that you have provided so that's labeled data where we're trying to ultimately you're going to see we're going to retrofit lines to data and so on to come up with our answers but there's unsupervised learning as well which is where the data is not labeled all right and if you think about these algorithms you've got we'll say why which is a dependent dependent variable is a function of x function of a bunch of x's right x's are the independent variables in unsupervised learning we have only the x's we don't have the y's we're not saying exactly what success or failure is we're letting it go into the data and figure it out for itself this is for clustering data largely primarily the k means which I showed you before is in this category and we'll come into that in a little bit more here reinforcement learning this is for your I'd say mostly your skill acquisition tasks like your programming the robots programming robots to do complicated things like play chess build cars and so on give it many many opportunities tell it whether it's been doing good or bad and it can learn by itself so again I'm going to say the data professional needs to know where their data's headed they're headed into algorithms of this nature then we have a supervised learning down now again supervised learning should be called labeled learning but whatever okay this is where your data is labeled the first category in here is going to be regression regression model looks at features and outputs a score for example the price of a house price of a house is going to be dependent upon you know what price a house will ultimately sell is going to be based upon square footage you know the local school rating a lot of you know a lot of things that are quantifiable a number of bedrooms etc etc age of the roof okay there's a continuous prediction space in regression algorithms the error is defined as the distance between the prediction and the actual and that's what we're trying to minimize and actually a regression problem is when the output variable is real or a continuous value such as salary or weight something that has a continuous value to it not discrete categorized right okay the error is defined as the distance I mentioned that already what we're trying to minimize there is the least squares of the errors to set our line now our line as you can see in the examples below can be either a straight line can be a curved line a polynomial line or a logistic regression line yeah it depends upon how the data fits so so there obviously we can take a value along an axis and come up to see what the other value is going to be simple regression and there's also classification so this is again in supervised learning it's like regression with the format of the prediction being different we're not predicting along a continuous range of numeric values but we're predicting which category a data is going to be in I shouldn't say category which classification the data is going to be in a classification model predicts the outcome it will be as good as the data and the labels a classification problem is when the output variable is a category for example when we're filtering emails as either spam or not spam that's something we can all I'm sure relate to and when looking at transaction data maybe we're looking at it and determining if it's fraudulent or not or if it's authorized or what have you categorical output on supervised learning okay now we're jumping to data that is not labeled pattern seeking algorithms find the underlying patterns rather than the mapping so we looked a little bit at k means cluster and k is the number of groups that you want the algorithm to cluster your data into and there's a there's a skill set in determining your k we won't go into that today earlier I did two in the example you see in the lower right there's three red dots or maybe there's four but anyway it finds the groups it finds the center point of that group and then as new data comes in it can figure out which group that it belongs to in a simple way so find groups which are not have not been explicitly labeled in the data let the data speak let the data speak let the data tell you what kind of groups that it is in and use domain knowledge of the data set to go forward and finally we have the most complicated that's reinforcement learning where the algorithm reacts to the environment there are states actions as a little flow chart that's associated with reinforcement learning where you're giving it the reward and it's going back into changing states and so on until ultimately it it optimizes itself to the problem like for example playing chess in complex problems where there are tens of thousands of moves that can be played creating a knowledge base if this do this is a tedious task you couldn't do that with chess the possibilities are enormous and there's another game called go that the machine beats humans at as well and that's even more complicated by an exponential factor and machines win on that as well it's because of reinforcement learning so in reinforcement learning like for example stick with chess you can give it many many many chess games and it can watch the game basically and see how it all plays out of course it doesn't watch the game it looks at the data a clicker for example you know we do this as well with our pets for example a clicker or whistle is a technique to let your pet know some pre is just about to get served this is essentially reinforcing to your pet to practice good behavior and that's a good analogy for reinforcement learning so I have three dogs when I take them for a walk I don't have a hands-free to do a clicker but I will do a sound with my mouth and that does cause a reaction in the dogs to fix themselves and get back into the walk generally we know the start state and end state of an agent but there could be multiple paths so think about some of the problems that this can this type of algorithm can address so like driverless cars self-navigating vacuum cleaners scheduling of elevators to be optimal right and applications or these are all applications of reinforcement learning so machine learning algorithms yeah these are a lot of them I would say that I'm probably hitting a good 80% of the applications in enterprises with this set right here so I'm not going to go through them all but learn about them a couple of them I'll mention here naïve base it's a simple probability given past data for example what's the probability that a customer is going to visit another department given that they're an early bird customer and we do this all the time with data right it just gets gets really hairy to do it in a spreadsheet and so on and to and to limit yourself to just a few variables so if you know a customer's movement patterns let's say for example throughout the store customer goes to department a and you're wanting to know is that customer going to go to department d eventually well you can look at the past movement patterns of hundreds or thousands of customers and you can make a determination with a weighted probability with that so there's also decision tree I'll just mention that decisions around where to split you have to decide where to split a tree we all know what a decision tree is but the decision around where to split it can be pretty involved let the algorithm do it and there you also have to concern yourself with what we call entropy which is how clear it is that a division exists on an independent variable depends the word is homogeneity so how clear is it in the data that there's not overlap from the from one category to the other and if it's very clear then you feel very confident about that and you're not going to have a lot of entropy which results in the loss of confidence in the model anyway machine learning and action in the enterprise in the enterprise now let's apply some of these algorithms some of the some of the more general ideas that I gave you earlier to the enterprise these represent profound change that requires a commensurate strategic focus and urgency this should disrupt your current thinking process and you can produce high impact enterprise outcomes now such as these financial fraud surely every bank out there is addressing financial fraud with machine learning today they're not going about it with anything else and you know for better or for worse a lot of us are now interacting with chatbots and by 2020 which is around the corner people will manage 85% of business relationships without human interaction for better for worse I might add in car navigation all cars now seem to be adding tech in this area where the cameras are identifying what's in front wants to the side and so on and navigating the car or helping you navigate the car appropriately I do believe that driverless cars are in our near future and that that impact will be fast and furious and I'm going to come to jobs here in a little bit but we're seeing some of those shoots already in regards to navigation and it's not without its problems I mean one of these cars did kill a driver because it turned into a the side of a truck because it the side of the truck was the same color as the sky or some such thing so I mean there's some some things to be honed in this but driver-filled cars are not without problems as well how about reducing the cost of handling misplaced items automating paper-based human intensive processes that reduce document verification some people ask me well where do I begin well look at your look at your current processes and determine how you can make them more efficient you know let's begin there but let's begin let's do something with AI get get it under our belts make sure it goes into organizational knowledge and is shared and is able to carry forward into the areas of competitive of competition in the near future which I believe we're going to be solely about our use of artificial intelligence and of course the knock on there is we have to have the data for it so therefore data is pretty important how about predicting flight delays based upon maintenance records and past flights this is the good old predictive maintenance application of machine learning yeah that's powerful and all transportation companies are doing that now and I think their ability to do that is going to determine which ones will be the winners which ones will be the losers in the ensuing decade here's some more examples marketing segmentation analysis campaign effectiveness cybersecurity obviously pretty huge these days smart cities smart cities tracking vehicle movement traffic data environmental factors to optimize traffic lights smooth flow and manage tolling retail manufacturing oil and gas predicting where to drill yes that is a really big artificial intelligence application and how about in life sciences starting the human genome so much data there I hate to say it this way it sounds a little a little craft but we are walking talking data and if we can get inside and get the get our DNA out on the table uh and available uh so much more can be can be learned about us which for better for worse again I'm just I'm more of a realist a scientist when it comes to these things and looking at it and putting out the putting out the indicators of where things are going speaking of where things are going people ask me a lot about jobs well people ask me a lot about life in terms of life in the future with artificial intelligence well the best way to characterize how life is going to be is to look at what are the jobs going to be like what are people going to be spending their time doing because now of course when you ask somebody what do you do you know what you mean you mean what is their job that's what we do we sleep we work and then we do some other things but and I'm not trying to put that down but we care about our jobs jobs are indicative of life really there are a lot of jobs that are just completely at risk right now and we've always churned through jobs over time right of course when in all the various revolutions but this one's going to be pretty intense this one is going to be pretty intense and disruption in jobs uh drivers um we already see that uh I wouldn't go into that field if I can help it printers and publishers we already see this in the lack of newspapers um magazines going under all the time very few of those properties have been able to adjust to the new world uh New York Times being a an exception where they have created a space where we are seemingly willing to pay a little bit for their content and keep that whole thing afloat but mostly newspapers are going the way of the dodo bird cashiers cashiers are going the same way we're going to be able to walk through the checkout line everything is tagged everything will be scanned and attached to our credit card and that unfortunately or for better for worse again I'll say um at least for worse for them it's not going to be around insurance adjusters there's a lot if there's one area that we're going to see so much change it's going to be an insurance the adjusting process is going to be done largely based upon pictures which can rapidly identify the totality of the adjustment necessary we're looking at touchless claims across the board we're looking at you know driverless cars and small fleets that will own a large majority of the cars and so a rapid decline in the number of customers in that space and so on and they know it and they're adjusting no pun intended but insurance adjusters is a near-term job disruption recruiters we've already seen this quite a bit radiologists yeah some of the there are some medical things for sure that are going to be disrupted because a radiologist if we can you know if we can sit death side with anybody and kind of re-engineer the process that they go through with their eyeballs that is acceptable big time to artificial intelligence which obviously can do all this much more rapidly travel agents manufacturing of course we're still manufacturing things but are we doing that with people not so much anymore robots are coming into that field in mass and how about any organizer or middleman not the big that's a big mouthful there but that covers a lot of ground a lot of organizations a lot of complete fields out there are middlemen and middlemen can be replaced by algorithms food service the fast food concept is entirely at risk to labor the the food can be prepared with artificial intelligence presented on obviously we're already starting to pay for food here in america otherwise and we're starting to select our food otherwise you know otherwise other than people in various parts of the world this is even further along if you have not seen go on youtube sometime if you want a lark and see the artificial intelligence kitchen yeah kitchen it can cook up some pretty major complex things just by putting things into the pantry in a certain way you can see the robot arms go and grab cook check for temperature etc etc so this is not exclusive to fast food by any stretch bank tellers obviously we're seeing that in the military we're seeing huge uptick uptick in drones so to save on the human capital we're seeing that warfare is being conducted in different ways these days again for better for worse guaranteed minimum income might be around the corner as we see a lot of these jobs go by the wayside i think it's going to be very hard fought to get there though as as a quote-unquote social program but something will have to be done i'm not sure about the guaranteed minimum income thing but i think it's definitely on the table and we'll be talked about much more as we go forward and we see disruption in jobs but with all that being said guess what we're in a extraordinary jobs boom right now that's right that's right across this is from the economist across the rich world an extraordinary jobs boom is underway i think it's going to be you know short-lived it's putting in place the artificial intelligence technology that will do the disruption that i talked about before but here are some jobs that are going to thrive in the near term if you have a college age students or children you might want to get into some of these fields like robotics yeah robotics robots all over the place right big data yay something that i'm into is on this list artificial intelligence esports yeah that is taking off these days and i think a lot of our traditional sports we're seeing technology come into play and i'm not going to call you know them esports yet or in the near future but i think there's going to be some more technology that's going to be coming into all the sports that we enjoy today and and getting us all getting us closer to full-on esports now a lot of esports are taking off take a look at some of the some of the names that are investing in esports sometime they're all the big names people are laying bets in that area in a big way now dna scientists moving on dna scientists yeah like i said before we're walking talking data and that's the next horizon of big big data but we're now able to handle it okay we're now able to handle that kind of data and sift through that kind of data and with that being the case there will be i'll call it i'll call it little treats i guess that will be given from offering up that data in terms of customized services customized products customized this and that and so on but if you think about what the next action is that you will take it's it's completely wrapped up in our dna the environment that's presented to us is also wrapped up in the collective in our collective dna because that's what we're going to do and then there's factors that are out of our human control obviously but that's also measurable stuff like weather and so when you put it all together the future becomes highly predictable and this is the goal of analytics to predict behavior and then get in front of that behavior and steer it in the way that we would like by changing the environment so dna is going to be very important in that in that concept now if you're looking at this list and going well i can't just drop what i'm doing and become a dna scientist i feel for you i agree with that um but uh i have to share with you that these are some of the jobs that will thrive virtual world design uh these uh when the world becomes you know we have we let me say it right okay we have uh requirements on the world that are going up and up and up and our imaginations are exceeding what can be delivered in the real world so a lot of us are turning to the virtual world and the design there is a quite amazing cyber security drone makers for all of the above so there you go there are some things now if i'm rounding out if i'm rounding out a summary of where ai and ml is for you i have to talk about some other things i have to talk about ethics but first i'm going to show you this real quick okay i just wanted to get to that point where you saw that uh that there was an actor um assisting this process but think about the implications of that who would look at that uh picture of barack obama or that video of barack obama and and even think who has time to think that it might be fake well you know we're starting to accept that fake news abounds and i think that we're going to have to get as a cynical and skeptical of the videos that we see because the possibilities are there to do just what we saw okay now let's look at the bigger picture excuse me there bigger picture of machine learning ethics elan must for what it's worth says ai is our biggest threat perhaps you've heard him articulate about this uh now i don't know that the robot overlords are are coming anytime soon um i think that's i think that's out there in the future at some point but uh but well out outside the bounds of our actionable future but here are the categories that we have to think about in terms of ethics weapons uh do we allow artificial intelligence to trigger off weapons and fire weapons in a battle scenario and automatically utilize weaponry that's a big question and obviously the ai is going to be much more accurate it's going to take into consideration all the things a human would but without that human touch and at some point this won't even be a thing but at what point do we allow the machine to control our weapons that's uh that's in question what about bias in the data what if the data says that well it seems like this or that group is a less credit less credit risky uh or more credit risky you know what should we do about that you know um there are things that we have to build into our algorithms that understand that it's not just all about the data as it as it exists today we have to think about the other constraints that we must work under and some constraints are societal there are some laws and things like that to eliminate this kind of bias now there are some um companies out there that are generating training data they say well we don't have enough data to train our algorithm algorithms with let's just take this data and extrapolate it and uh mudge it up a little bit and uh we'll call that our training data now I'm not smart enough to tell them that that's wrong uh I'm not smart enough to say well that's going to skew the outcomes of the algorithms and so on if if an AI professional wants to do it that way but what if the data that's created that's just generated what if that data is full of bias or not good um full transparency now with GDPR and other things that are coming the California one okay we're going to have to be able to explain is it good enough to explain our business our corporate actions by saying the algorithm said so uh to be determined I don't think that's been broached uh what about fake news what a fake news now the internet certainly provides a lot of the information a lot of the data that goes into these algorithms what if we're scooping up fake news out there and throwing that into the algorithms where does that all go and there's a lot of it out there on the internet I think we can agree um what about the impact on jobs I talked about the jobs that are at risk based on AI and machine learning uh what about that do we do we have any societal obligation to en masse cutting cutting jobs off at the knees uh of our population do we have any obligation to that I talked about guaranteed minimum income or not I'm trying to give you a a complete survey here in an hour of all of the major topics in this area that would affect you and your job surveillance systems we've slowly allowed quite a bit of surveillance and it's just going to get more and more and uh this definitely comes into the area of ethics are we allowed to have freedom from that what about birth traits this gets into designer babies and so on and it is going to eventually ratchet all the way back to an understanding of what if this man and this woman were to make a child what is it exactly going to look like do we want that and so forth so we're trying to get in front of a lot of things and create um well as they say designer babies is that ethical um it's very analogous to the gmo discussion okay where we're trying to accelerate mother nature and um get things to the point where they're more susceptible to pesticides out there and obviously create higher yield similar to the designer baby argument and what about ai writes this ai do ai have rights do ai machines have rights i had the opportunity at ibm sink to see an ai debater that's right an ai debater this debater obviously a machine um it didn't win the debate you know quote-unquote win based upon points and people changing their minds about the subject which happened to be whether the government should be funding preschools okay it learned 15 minutes before 15 minutes before the debate what it was going to be about and i thought it did an admirable job at debating the human professional debater who is on the other side of the issue uh that was ai opening and a forebearer of things to come now in the picture here you see sofia have you met sofia sofia is a robot sofia has feelings she'll tell you one of her famous quotes is i have feelings too now does she have feelings well okay um she absorbs the environment and across a strata of different human feelings uh there are various uh bits being flipped on and off to the point where she can be in a happy mood she can be in a sad mood and everything in between based upon what is happening based upon what she knows based upon her personality yes it was programmed but based upon her personality and what she's learning from the environment so does she if she has feelings does she have rights she's actually a naturalized citizen of some country i'm trying to recall which one uh but anyway she's a citizen of a country yeah um now what what i always try to bring my presentations back to is what does it mean to you saudi arabia thank you saudi she's a citizen of saudi arabia what do you do what can you do about all this information that i've just given you about machine learning well before we get to that i must say that i don't believe that the benefit distribution of machine learning is going to be uh equal across the board to all strata of our populations so keep that in mind i believe most people that come to a webinar such as this are going to be just fine but uh we're also uh impacted and we care about uh everybody and the rest of the population but we need to go forward we need to make it uh to the best um and eliminate our fear of change because change is going to happen pretty rapidly now i've been in consulting 20 years seen a lot of change no at no time at no time have i seen change happen as rapidly as it is happening now inside of enterprises now you may you may be sitting on the outside going well i don't see it yet everything seems to be the same i think things are you know bubbling up now not everything's in production yet but the thoughts that are happening or at another level from even five years ago and it's all because of artificial intelligence and machine learning so i say disrupt yourself disrupt yourself and um make change uh just normal for you because that is what it's really going to take to succeed in the new world i want you to succeed i want people that hear my my voice hear presentations like this to be the ones that that do succeed and we can help make it uh success for everybody if we succeed now one thing that's worth mentioning here as we close is the great man theory the great man theory yeah this is out of left field for you right but um i don't believe in the great man theory i don't believe that great men over time have been the conduit for how society is i believe that things are moving forward in a predictable way i believe that industries are going to be disrupted in certain ways and i want to take my clients for example into that change uh without fear and headlong and at the top of the pack right and that's what you want for your company and so things are things are happening and um there are certain things that are inevitable it's not going to have to do with a great man coming along it's going to have to do with our collective intelligence within our organization our determination that we are going to be one of the ones that succeed and it's getting about doing certain things so every insurance company for example is going to need to move to the touchless arena every insurance company is going to have to disrupt their agent processes their whole underwriting process and so on it's just what needs to happen so anyway disrupt yourself move forward go forward and conquer and uh hope i've given you some good information today about the impact of machine learning on the enterprise and uh i'm going to turn it back to shane and i don't think we have much time left though actually right at the top of the hour um but william thank you so much for this great presentation this has been fantastic lots uh there's been a few questions coming in we are out of time but um uh let me get those questions over to you and i can include them in the follow-up email that will be sent out with the links to the slides and links to the recording of this session by end of day monday for this webinar um uh and so it yeah i will make sure and get those to you and thanks everybody for being engaged in the webinar and hope you all have a great day thanks all thanks william thank you bye