1 00:00:00,110 --> 00:00:02,990 in the data and then projecting those forward. 2 00:00:03,950 --> 00:00:09,630 And that works marvelous as long as the world you operate in never changes. 3 00:00:10,670 --> 00:00:16,110 But the minute your world changes from what it looked like historically, those models, they drift. 4 00:00:16,110 --> 00:00:18,990 You've heard data scientists talk about drift management. 5 00:00:18,990 --> 00:00:20,110 Models drift. 6 00:00:20,750 --> 00:00:28,590 So the problem is basing your models on historical trends and patterns means the minute those models are put into production, 7 00:00:29,390 --> 00:00:30,030 They're out of date. 8 00:00:30,910 --> 00:00:31,470 They're wrong. 9 00:00:32,390 --> 00:00:35,750 And you need constant tinkering by the data science team. 10 00:00:35,750 --> 00:00:40,510 It's called the Data Science Full-time Employment Act to constantly keep those models up to date. 11 00:00:41,710 --> 00:00:44,430 But AI is fundamentally different. 12 00:00:44,430 --> 00:00:45,710 It doesn't optimize. 13 00:00:46,910 --> 00:00:49,150 It learns and adapts. 14 00:00:50,030 --> 00:00:51,990 It's constantly learning and adapting. 15 00:00:51,990 --> 00:00:54,190 Think about how an autonomous vehicle works. 16 00:00:55,110 --> 00:00:57,150 It's not trying to make the optimization. 17 00:00:57,350 --> 00:01:00,950 It's trying to make the best decision in the context of the situation it's in. 18 00:01:01,070 --> 00:01:11,230 So autonomous vehicle is making, by the way, just a handful of decisions, probably 2018 decision it makes around braking and turning and windshield wipers. 19 00:01:11,670 --> 00:01:17,789 It uses 30,000 to 40,000 variables to help them make those decisions, but it only makes a handful of decisions. 20 00:01:18,110 --> 00:01:21,870 And the decision it makes right now, based on the context it's in, 21 00:01:22,830 --> 00:01:25,630 The decision might be very different 5 seconds later. 22 00:01:26,190 --> 00:01:31,230 A light turns red, it starts to rain, a ball rolls across the street, it sees a car pulling out of a parking spot, right? 23 00:01:31,550 --> 00:01:34,430 It sees a clown riding backwards on a unicycle, right? 24 00:01:34,670 --> 00:01:36,110 Everything's changing in the model. 25 00:01:36,110 --> 00:01:42,430 So what it does is it tries to make the right decision in the context of that moment. 26 00:01:43,789 --> 00:01:48,910 Constantly learning and adapting based on the context of the moment. 27 00:01:49,870 --> 00:01:58,990 This makes this technology very different and it makes it incredibly powerful because it's not held captive to what's happened in the past. 28 00:01:59,870 --> 00:02:12,670 It's using in the past trying to help it build the models and the variables and metrics, but it's using the context of the current decision, the current situation, the context of the current situation to make the right decision in that moment. 29 00:02:15,310 --> 00:02:16,350 In order to do that, 30 00:02:18,150 --> 00:02:20,829 What the model has to do, two key things up front. 31 00:02:21,550 --> 00:02:22,990 What are my intentions? 32 00:02:23,110 --> 00:02:24,829 What is it I'm trying to accomplish? 33 00:02:24,829 --> 00:02:26,910 And what are my desired outcomes? 34 00:02:27,870 --> 00:02:32,350 We run these workshops, we teach in the class, and we bring together a bunch of diverse stakeholders. 35 00:02:32,350 --> 00:02:36,829 When I was at Dell, this was a methodology we used at Dell to engage with customers. 36 00:02:36,829 --> 00:02:44,990 We bring together a broad range of stakeholders in a room about this size with all kinds of flip charts and post-it notes trying to dive down into the desired outcomes. 37 00:02:46,550 --> 00:02:47,630 We wouldn't have two or three. 38 00:02:47,630 --> 00:02:51,870 We'd have 50, 70, 80 desired outcomes across a wide range of stakeholders. 39 00:02:53,230 --> 00:02:54,270 All those diverse things. 40 00:02:54,590 --> 00:02:55,710 That's how we start. 41 00:02:56,230 --> 00:02:59,270 What are we trying to accomplish and what do we think is good look like? 42 00:02:59,270 --> 00:03:06,310 And then we want to start for each of those desired outcomes, what are the KPIs and metrics around which we're going to measure the desired outcomes? 43 00:03:06,310 --> 00:03:07,390 By the way, you don't want one. 44 00:03:07,790 --> 00:03:08,990 You want at least three. 45 00:03:09,350 --> 00:03:12,030 If you have one metric for desired outcome, you get a bias. 46 00:03:12,270 --> 00:03:13,790 You need 3 to triangulate. 47 00:03:14,270 --> 00:03:15,630 And so all of a sudden you go from 48 00:03:16,270 --> 00:03:23,150 80, 60, 90 desired outcomes to 1,000 variables and metrics. 49 00:03:23,350 --> 00:03:24,110 Thousands. 50 00:03:24,870 --> 00:03:40,670 And that's okay, because what's happened is these models are designed using a, this is a formula for a neural network, to process all those different variables and metrics, knowing what your desired outcomes, to make the right decision in this moment, given the context. 51 00:03:42,910 --> 00:03:45,630 Again, lots of work before we ever start putting 52 00:03:46,350 --> 00:03:47,470 science of the data. 53 00:03:48,390 --> 00:04:04,830 And in order for us to make certain that we have thought holistically about how we're going to make decisions, how we're going to look across our broad and diverse range of stakeholders and constituents to make decisions, you do need to think like an economist. 54 00:04:05,550 --> 00:04:07,550 And this is how an economist thinks. 55 00:04:09,230 --> 00:04:11,710 Economics and finance are not the same thing. 56 00:04:12,590 --> 00:04:19,870 AI models do a crappy job of optimizing on financial metrics because most financial metrics are lagging indicators. 57 00:04:20,350 --> 00:04:21,630 They measure what's happened. 58 00:04:22,029 --> 00:04:26,910 And pardon my bluntness, but AI does a shit job of optimizing around things that have already happened. 59 00:04:27,870 --> 00:04:32,110 So you need to think like an economist and start bracing across all these different variables and metrics. 60 00:04:32,750 --> 00:04:41,230 Think about how your organization creates value from a customer perspective, from an employee, stakeholders, community, ecosystem. 61 00:04:41,550 --> 00:04:45,870 partner, society, environment, workforce, ethical. 62 00:04:47,390 --> 00:04:49,870 Broad range of how value is created. 63 00:04:50,030 --> 00:05:05,310 By the way, in my 40 plus years of doing this, probably closer to 50 at this point, of doing this with thousands of companies, every company I've wrestled with, I've worked with, has wrestled with trying to define how they create value. 64 00:05:06,670 --> 00:05:10,910 We've got a finance mindset, and that mindset is 65 00:05:11,390 --> 00:05:13,790 the antithesis of what AI can do. 66 00:05:14,830 --> 00:05:16,790 So we're going to start by thinking like an economist. 67 00:05:16,790 --> 00:05:20,510 Again, you're going to get these slides, but you need to think more broadly. 68 00:05:21,030 --> 00:05:27,150 And so let me walk you through a really simple exercise about how the human mind makes decisions. 69 00:05:27,630 --> 00:05:34,150 Because when you're making a decision consciously or subconsciously, you are making some of these trade-off decisions. 70 00:05:34,150 --> 00:05:39,470 When you go to a restaurant, for example, you might choose local grown versus 71 00:05:40,030 --> 00:05:40,630 not local. 72 00:05:40,630 --> 00:05:45,470 You might choose a local-based restaurant versus a chain. 73 00:05:45,470 --> 00:05:49,070 You might choose a restaurant that pays your employee more. 74 00:05:49,630 --> 00:05:52,630 Each of you have a different set of variables and metrics you're going to do. 75 00:05:52,630 --> 00:05:53,909 In fact, here's what we're going to do. 76 00:05:54,750 --> 00:06:06,110 I want you to pair up with somebody, and I want you to think about what are the variables and metrics that you might want to consider 77 00:06:06,550 --> 00:06:10,750 when trying to decide where you're going to go eat tonight after the conference here. 78 00:06:11,550 --> 00:06:13,710 So pair up, get a piece of paper. 79 00:06:13,790 --> 00:06:15,230 We're going to take a couple of minutes. 80 00:06:15,390 --> 00:06:16,270 Get buddy, buddy. 81 00:06:16,550 --> 00:06:21,150 And I want you to identify what variables and metrics you might. 82 00:06:22,670 --> 00:06:23,430 Oh, I love the energy. 83 00:06:23,430 --> 00:06:27,070 We've got two minutes on this. 84 00:06:33,409 --> 00:06:34,370 Write them down. 85 00:06:35,570 --> 00:06:36,050 I love it. 86 00:06:53,150 --> 00:06:53,870 Right, right, right. 87 00:06:54,430 --> 00:06:55,070 Right, right, right. 88 00:06:59,550 --> 00:07:00,990 One minute, one minute. 89 00:07:19,950 --> 00:07:21,790 30 seconds, 30 seconds. 90 00:07:29,870 --> 00:07:30,670 15. 91 00:07:41,230 --> 00:07:47,550 All right, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1. 92 00:07:48,030 --> 00:07:49,950 Okay, can I get your attention? 93 00:07:51,150 --> 00:07:52,230 I want you to pause a second. 94 00:07:53,070 --> 00:07:54,110 I want you to pause. 95 00:07:54,990 --> 00:07:56,350 Did you feel the energy in this room? 96 00:07:57,630 --> 00:07:59,630 You see how everybody's sort of geared up? 97 00:08:00,430 --> 00:08:03,870 The human mind is a phenomenal machine. 98 00:08:04,790 --> 00:08:06,030 It's A phenomenal machine. 99 00:08:06,030 --> 00:08:13,470 When you allow the human mind to start to do its thing, the ideas that come out are phenomenal. 100 00:08:14,750 --> 00:08:15,550 Phenomenal. 101 00:08:15,950 --> 00:08:22,510 So let's go through, I'm going to point to a couple of folks and tell me, give me two or three variables and metrics that you thought were important. 102 00:08:24,030 --> 00:08:25,070 How many other people are there? 103 00:08:25,470 --> 00:08:25,630 Okay. 104 00:08:26,230 --> 00:08:26,990 Chain or independent? 105 00:08:27,150 --> 00:08:27,470 Good. 106 00:08:27,950 --> 00:08:28,590 Somebody else. 107 00:08:28,990 --> 00:08:29,390 Location. 108 00:08:29,550 --> 00:08:30,110 Location. 109 00:08:31,150 --> 00:08:31,470 Labor. 110 00:08:31,710 --> 00:08:32,190 Labor. 111 00:08:32,990 --> 00:08:33,429 Type of food. 112 00:08:33,549 --> 00:08:34,110 Type of food. 113 00:08:34,390 --> 00:08:34,669 Price. 114 00:08:34,830 --> 00:08:35,230 Price. 115 00:08:35,789 --> 00:08:36,150 Quality. 116 00:08:36,270 --> 00:08:36,830 Quality. 117 00:08:38,070 --> 00:08:38,230 Time. 118 00:08:38,510 --> 00:08:38,909 Time. 119 00:08:39,230 --> 00:08:40,590 They can be repeats. 120 00:08:40,830 --> 00:08:41,270 Yeah, time. 121 00:08:41,590 --> 00:08:42,110 Time, good. 122 00:08:42,510 --> 00:08:43,070 Reputation. 123 00:08:43,070 --> 00:08:43,710 Reputation. 124 00:08:43,909 --> 00:08:44,350 Proximity. 125 00:08:44,510 --> 00:08:45,150 Proximity. 126 00:08:45,790 --> 00:08:46,310 Who you're with? 127 00:08:46,430 --> 00:08:47,070 Who you're with? 128 00:08:47,350 --> 00:08:48,230 Convenience and time. 129 00:08:48,430 --> 00:08:49,270 Convenience and time. 130 00:08:49,270 --> 00:08:50,190 The atmosphere. 131 00:08:50,630 --> 00:08:51,710 Okay, one more. 132 00:08:52,190 --> 00:08:52,590 Cost. 133 00:08:52,670 --> 00:08:53,070 Cost. 134 00:08:53,310 --> 00:08:53,630 All right. 135 00:08:56,190 --> 00:09:03,630 So when I do this with high school students, I do this at the Waukee Innovation Learning Center where I get students together and we brainstorm. 136 00:09:03,630 --> 00:09:05,390 We have more time in that situation. 137 00:09:05,710 --> 00:09:07,390 They come up with about 120. 138 00:09:08,990 --> 00:09:10,110 Now think about this. 139 00:09:10,190 --> 00:09:11,430 Now I bet you went across the room. 140 00:09:11,430 --> 00:09:14,430 If I had given you two more minutes, fair minutes, you were to come up with 100. 141 00:09:15,550 --> 00:09:19,230 Your mind is weighing off all these different factors. 142 00:09:19,630 --> 00:09:22,030 You're weighing off factors about time, cost. 143 00:09:22,510 --> 00:09:29,230 quality, local, community events, maybe ethical, organizations that have ethical behavior. 144 00:09:29,390 --> 00:09:31,310 Your mind weighs these things off. 145 00:09:31,310 --> 00:09:34,750 Now you may not write them down as you do the process, but you're processing your head. 146 00:09:35,230 --> 00:09:38,750 So you've got 120 variables and metrics trying to figure out where to go eat. 147 00:09:39,470 --> 00:09:42,350 So how is it going to figure out which of these variables and metrics are most important? 148 00:09:42,350 --> 00:09:44,830 Remember, intentions, 149 00:09:45,390 --> 00:09:46,270 in desired outcomes. 150 00:09:46,270 --> 00:09:52,550 So for me, when I get done here, I've got to go over to the university and teach some faculty members how to build their own digital assistant. 151 00:09:52,550 --> 00:09:56,230 Then when I get done with that, I went some places quick and convenient. 152 00:09:56,230 --> 00:09:57,950 I'm going to be tired by the end of the day. 153 00:09:58,430 --> 00:10:03,310 And these variables here, the weights in these variables increase. 154 00:10:03,950 --> 00:10:07,350 I'm going to want affordability and convenience, and I don't want to have to pay for parking. 155 00:10:07,790 --> 00:10:10,150 So there's a number of these variables whose weights go up. 156 00:10:10,150 --> 00:10:13,790 Remember, a utility neural network looks at variables and weights. 157 00:10:14,550 --> 00:10:17,030 And uses those weights with those variables to help make a decision. 158 00:10:17,870 --> 00:10:20,670 So, I'm going to Chipotle, baby, right? 159 00:10:22,670 --> 00:10:27,790 Uh-oh, Friday night, date night, and my wife don't want to go to Chipotle. 160 00:10:28,350 --> 00:10:28,590 Right? 161 00:10:28,670 --> 00:10:30,270 So what's important to her? 162 00:10:30,430 --> 00:10:32,350 Like, I have any say in this whatsoever. 163 00:10:32,590 --> 00:10:32,910 Right? 164 00:10:33,310 --> 00:10:36,150 Well, she wants some place nice, got a nice ambiance. 165 00:10:36,150 --> 00:10:38,710 Somebody's going to, they're going to wait on her. 166 00:10:38,710 --> 00:10:40,150 You know, she can dress up. 167 00:10:40,150 --> 00:10:43,550 It's got, you know, maybe she can be seen by somebody or she can see friends, right? 168 00:10:43,950 --> 00:10:46,350 We're going some place way, way too expensive. 169 00:10:48,030 --> 00:10:49,950 This is how the human mind works. 170 00:10:50,990 --> 00:10:55,710 A large number of variables and metrics, many of them conflict with each other. 171 00:10:56,910 --> 00:10:59,990 This is what we want this is not. 172 00:11:00,070 --> 00:11:06,830 not traditional machine learning where you're trying to get down using principal component analysis down to what are those four or five most variable metrics. 173 00:11:06,830 --> 00:11:07,950 No, I want thousands of them. 174 00:11:08,550 --> 00:11:10,430 And I want the ones that conflict with each other. 175 00:11:11,070 --> 00:11:13,310 I don't avoid collinearity. 176 00:11:13,870 --> 00:11:14,910 I embrace it. 177 00:11:15,310 --> 00:11:23,070 Because the more variables and metrics I have, the more granular, the more relevant, the more meaningful outputs I'm going to get. 178 00:11:23,950 --> 00:11:25,230 So #1, 179 00:11:25,710 --> 00:11:31,550 If you want to become an AI solution engineer, step number one, learn to think like an economist. 180 00:11:32,150 --> 00:11:40,270 And think about the broad range around how organizations, companies, people, society measure value. 181 00:11:41,150 --> 00:11:43,230 Okay, got a good start. 182 00:11:45,150 --> 00:11:47,230 Let's talk about Yoda. 183 00:11:48,590 --> 00:11:49,190 Yoda. 184 00:11:49,190 --> 00:11:50,110 And I love this quote. 185 00:11:50,590 --> 00:11:52,910 You cannot build a system that thinks well 186 00:11:53,670 --> 00:11:56,430 without building one that questions well, which that was my quote. 187 00:11:56,430 --> 00:11:58,430 That was given to me one of the classes I had taught. 188 00:11:59,030 --> 00:12:02,030 And we're going to build a thinking system with Yoda. 189 00:12:02,590 --> 00:12:07,950 But we need to understand before we dive into this the realities of generative AI. 190 00:12:08,670 --> 00:12:14,190 Now when people throw the word around AI today, they almost always mean generative AI. 191 00:12:14,670 --> 00:12:17,870 And generative AI has a fundamental problem. 192 00:12:19,070 --> 00:12:27,150 It's based on correlations or statistical averaging across a wide volume of data. 193 00:12:27,150 --> 00:12:32,190 So think about your average large language model having the equivalent of 440 million books in it. 194 00:12:33,190 --> 00:12:37,150 And 90% of those books, 95%, 98 are coming from social media. 195 00:12:37,550 --> 00:12:40,830 It means you're getting a whole lot of books about the Kardashians and Taylor Swift. 196 00:12:42,270 --> 00:12:43,710 And so your answers 197 00:12:44,430 --> 00:12:46,430 are averaging across all those. 198 00:12:47,310 --> 00:12:52,590 The outputs from a generative AI represent the most common data trends. 199 00:12:54,830 --> 00:13:00,830 And most of our data is highly questionable, skeptical, biased. 200 00:13:01,510 --> 00:13:05,230 And so what it does, it's a correlation-based tool. 201 00:13:05,790 --> 00:13:07,030 It's not a causation. 202 00:13:07,030 --> 00:13:08,430 It can't tell you why. 203 00:13:08,750 --> 00:13:11,110 It can only correlate about what it's observing. 204 00:13:13,390 --> 00:13:15,550 gives you very average generic answers. 205 00:13:16,590 --> 00:13:18,910 This is because it struggles with outliers. 206 00:13:19,150 --> 00:13:27,230 It immediately takes second and third standard deviation items that are the items that might be most valuable and averages them right out. 207 00:13:29,350 --> 00:13:32,350 And finally, it has a bias towards historical data patterns. 208 00:13:33,310 --> 00:13:35,870 It's wed to that and it's averaging across that. 209 00:13:36,030 --> 00:13:41,470 And so what happens, you get this regressive model, this collapse 210 00:13:41,790 --> 00:13:45,390 We're getting averages of averages of averages of averages. 211 00:13:45,870 --> 00:13:48,790 And pretty soon you're getting results that at best are average. 212 00:13:48,790 --> 00:13:55,310 Remember, if you're making your decisions based on averages at best, you're going to get average results. 213 00:13:56,510 --> 00:13:59,790 I can guarantee you everybody in this room here is better than average. 214 00:14:00,910 --> 00:14:03,190 If you were below average, getting to average is probably great. 215 00:14:03,190 --> 00:14:06,990 But everybody in this room, all of my students, they don't have aspirations of being average. 216 00:14:07,590 --> 00:14:12,110 And I can guarantee that people who've come to this session and who are spending time are not here to be average. 217 00:14:12,110 --> 00:14:13,230 So what do we do? 218 00:14:14,030 --> 00:14:16,830 What do we do with a tool that's got value to it? 219 00:14:17,230 --> 00:14:18,310 It's not causal. 220 00:14:18,310 --> 00:14:20,190 We're going to talk about causal in a bit. 221 00:14:20,670 --> 00:14:22,550 But it is correlation-based. 222 00:14:22,550 --> 00:14:24,670 How do I get this to work for me? 223 00:14:25,230 --> 00:14:30,670 Well, we're going to turn it into Yoda, your own digital system, by going through a five-step process. 224 00:14:30,670 --> 00:14:33,150 Now I'm going to go through this process fairly quickly. 225 00:14:33,950 --> 00:14:35,790 Again, you're going to get the slides. 226 00:14:36,030 --> 00:14:40,990 If you follow me on LinkedIn, I'm constantly publishing more content about Yoda. 227 00:14:42,510 --> 00:14:47,550 The first day of class, we spend the entire class setting up and building Yoda. 228 00:14:48,030 --> 00:14:55,230 And then throughout the entire 13 weeks of the class, the students pick a company and a problem to go after. 229 00:14:55,230 --> 00:15:02,350 So this year we picked nurse retention, we picked claims processing, we picked childcare, 230 00:15:04,110 --> 00:15:05,950 selection optimization at the university. 231 00:15:06,430 --> 00:15:11,950 Previous cases, we've picked things like product rationalization and such. 232 00:15:11,950 --> 00:15:17,350 We pick a company, we pick a problem, and we immediately start training Yoda on that problem. 233 00:15:17,390 --> 00:15:19,310 We're going to do the same thing here in a second. 234 00:15:19,470 --> 00:15:21,470 But then we go through a five-step process. 235 00:15:21,630 --> 00:15:26,590 Throughout the semester, they're training Yoda on the problem they're going after. 236 00:15:26,910 --> 00:15:29,310 Throughout the semester, they're taking, and what it does 237 00:15:29,710 --> 00:15:32,670 By doing this problem, oh, I don't think I have the slide here anymore. 238 00:15:32,830 --> 00:15:33,230 Bummer. 239 00:15:33,630 --> 00:15:49,150 We basically take that 550 million books out there and we condense it down to the 30 or 20 or 15 books that are relevant to the problem I'm going after. 240 00:15:49,710 --> 00:15:50,510 I'm training it. 241 00:15:51,150 --> 00:15:53,870 I'm training to say I care about nurse retention. 242 00:15:54,590 --> 00:15:56,430 Give me all the research on nurse retention. 243 00:15:56,430 --> 00:15:58,030 Give me all the information on nurse retention. 244 00:15:58,030 --> 00:15:59,150 Give me all the comments on... 245 00:16:00,670 --> 00:16:01,790 Kardashians go bye-bye. 246 00:16:01,790 --> 00:16:03,150 Taylor Swift go bye-bye, right? 247 00:16:03,790 --> 00:16:06,110 Chicago Cubs, sorry, you always go bye-bye, right? 248 00:16:06,510 --> 00:16:09,670 So I'm telling it what's important. 249 00:16:10,030 --> 00:16:16,670 And so I'm tricking this tool to think like me, to understand what's important to me. 250 00:16:16,990 --> 00:16:18,150 So here's how we do this. 251 00:16:18,150 --> 00:16:22,670 We're going to have a simple example that I did recently with a farming co-op. 252 00:16:23,150 --> 00:16:28,830 about how do we use a tool like generative AI to help us figure out what crops to plant. 253 00:16:30,110 --> 00:16:45,230 So here's a scenario where step one is to define our intentions, our desired outcomes, our boundaries, our constraints, give it as much information as possible about what it is we're trying to achieve about who I am. 254 00:16:45,910 --> 00:16:49,350 what kind of farm I am, what size farm, how long has it been in the family, where is it located? 255 00:16:49,750 --> 00:16:51,390 As much detail as possible. 256 00:16:51,390 --> 00:16:54,870 I want to train it to define my intent and my context. 257 00:16:54,870 --> 00:17:00,350 So I'm A 10,000 acre farm located outside Charles City, Iowa, my hometown. 258 00:17:00,990 --> 00:17:02,430 And these are my objectives. 259 00:17:02,430 --> 00:17:03,630 So we summarized it, right? 260 00:17:03,630 --> 00:17:08,910 So in reality, it's probably a lot longer than this, but I need to balance as a farmer 261 00:17:09,550 --> 00:17:16,990 selecting what crops I'm going to plant, but I've got to look at profit maximization and risk mitigation and resource efficiency and blah, blah, blah. 262 00:17:17,310 --> 00:17:20,750 Again, these things all conflict with each other. 263 00:17:21,950 --> 00:17:27,150 AI is a marvelous tool for providing transparency and making these balancing decisions. 264 00:17:27,910 --> 00:17:28,910 That's its real power. 265 00:17:28,910 --> 00:17:31,950 So I'm going to immediately set up conflict. 266 00:17:32,590 --> 00:17:33,790 I want profitability. 267 00:17:34,430 --> 00:17:35,790 I want water quality. 268 00:17:36,350 --> 00:17:39,470 I want environmental, I want long-term sustainability, right? 269 00:17:39,470 --> 00:17:39,990 I want it all. 270 00:17:39,990 --> 00:17:41,270 Let's make a little commercial, right? 271 00:17:41,270 --> 00:17:41,950 I want it all. 272 00:17:42,270 --> 00:17:43,790 But how do I make those trade-off decisions? 273 00:17:43,790 --> 00:17:44,830 So that's step one. 274 00:17:45,390 --> 00:17:47,310 Define what it is I'm trying to go after. 275 00:17:48,590 --> 00:17:49,310 Takes time. 276 00:17:50,110 --> 00:17:51,710 We do this, we did this at Dell. 277 00:17:51,870 --> 00:17:54,750 This was a two-day workshop. 278 00:17:55,310 --> 00:17:57,350 just making sure that everybody's on the same page. 279 00:17:57,350 --> 00:18:01,990 And by the way, we brought in people, diverse people, even people who didn't like each other. 280 00:18:01,990 --> 00:18:09,910 In fact, I preferred to have people who didn't like each other because there's probably a basis for why one person had a position and another person had a position. 281 00:18:09,910 --> 00:18:14,270 And once you get by opinions with people and you get down to rationale, now you got meat. 282 00:18:15,150 --> 00:18:16,110 Now you got meat. 283 00:18:16,630 --> 00:18:18,670 All right, so we've defined our problem. 284 00:18:18,670 --> 00:18:20,350 Now we need to start training this tool. 285 00:18:21,230 --> 00:18:23,230 We've got it focused, and I'm going to do two things here. 286 00:18:23,390 --> 00:18:33,470 I'm going to leverage outside trusted validated research, and I'm also going to start mining that critical domain knowledge organizations have. 287 00:18:34,270 --> 00:18:40,390 So in the farming example, we went out and found 16 different research studies. 288 00:18:40,390 --> 00:18:41,150 There are four of them, right? 289 00:18:41,470 --> 00:18:43,790 We went out and found validated research studies. 290 00:18:44,590 --> 00:18:46,670 Use Google search as the best tool for this, by the way. 291 00:18:47,230 --> 00:18:47,950 I can see it. 292 00:18:47,950 --> 00:18:48,910 I can validate it. 293 00:18:49,550 --> 00:18:58,990 I can go out there and I get a bunch of credible resources that talk about crop selection, soil conditions, irrigation, all the factors that go along with that. 294 00:18:59,550 --> 00:19:08,270 I'm loading it with credentialed information and telling it, find me more research like this. 295 00:19:09,390 --> 00:19:14,670 Credible, peer-reviewed, in some cases legally liable resources. 296 00:19:15,870 --> 00:19:16,350 Easy to do. 297 00:19:16,350 --> 00:19:18,350 There's a lot of great resources out there. 298 00:19:18,630 --> 00:19:29,150 And so whatever problem I'm going after, whether it's nurse retention or customer retention or inventory optimization, get a bunch of research I can get about that problem in that industry. 299 00:19:29,790 --> 00:19:31,750 But now this is where things get really interesting. 300 00:19:32,910 --> 00:19:37,670 The real secret sauce of this process is organizational domain knowledge. 301 00:19:37,670 --> 00:19:42,390 And so there's a book, a textbook we use in our class called The Art of Thinking Like a Data Scientist. 302 00:19:42,670 --> 00:19:44,990 It's an eight-step process where we take through 303 00:19:45,910 --> 00:19:52,110 And we bring stakeholders together and we brainstorm across eight different steps, think design, thinking kind of concepts. 304 00:19:52,430 --> 00:19:55,750 We're starting to mine all that incredible domain knowledge. 305 00:19:55,750 --> 00:20:01,230 So for example, here, summary, here's improved crop selection effectiveness. 306 00:20:01,230 --> 00:20:05,310 We've got our desired outcomes, maximize crop yield, optimize timing of planting, blah, blah, blah. 307 00:20:05,310 --> 00:20:09,950 So you start, remember, if you get a lot of people together, you don't have 12, you have 80. 308 00:20:11,230 --> 00:20:12,070 What are the benefits? 309 00:20:12,070 --> 00:20:13,470 What are the potential impediments? 310 00:20:13,830 --> 00:20:15,350 What are the failure ramifications? 311 00:20:15,350 --> 00:20:18,110 What are the potential unintended consequences, and how are we going to measure that? 312 00:20:18,910 --> 00:20:23,550 And then we basically take a page out of the design thinking journey book. 313 00:20:24,350 --> 00:20:24,950 Journey maps. 314 00:20:24,950 --> 00:20:28,110 We create journey maps in each of our stakeholders, understanding who's involved. 315 00:20:28,190 --> 00:20:30,270 By the way, more stakeholders, the better. 316 00:20:30,590 --> 00:20:32,750 The more granular the stakeholders, the better. 317 00:20:32,910 --> 00:20:34,950 The more desired outcomes I have, better. 318 00:20:34,950 --> 00:20:36,910 The more KPIs and metrics I have, better. 319 00:20:36,910 --> 00:20:40,670 I want more, more, more, more around the problem I'm trying to solve. 320 00:20:41,310 --> 00:20:42,590 This is a lengthy process. 321 00:20:43,150 --> 00:20:52,990 We spend one week on each one of those eight steps to really make sure we've really well defined the problem and captured all that valuable domain expertise. 322 00:20:53,630 --> 00:21:01,710 Side note, companies that are laying off their employees are watching their domain expertise walk out the door. 323 00:21:02,830 --> 00:21:04,670 What a huge failure. 324 00:21:05,470 --> 00:21:10,510 Short-sighted, because there's a big difference between productivity gains by laying people off 325 00:21:11,550 --> 00:21:15,230 Productivity gains are not differentiated because anybody can copy them. 326 00:21:15,870 --> 00:21:25,630 What is differentiated is taking that domain expertise of your people who have been in industry for two years, 20 years, 40 years, and turning that into insights that AI can collaborate with. 327 00:21:26,510 --> 00:21:27,310 Inside note. 328 00:21:28,750 --> 00:21:30,190 All right, drink of water here. 329 00:21:30,670 --> 00:21:32,310 So we've got a great foundation. 330 00:21:32,670 --> 00:21:35,830 We have a lot of external data to give us perspective and validation. 331 00:21:35,830 --> 00:21:38,190 We're now starting to mine all that internal information, right? 332 00:21:38,190 --> 00:21:39,870 We're starting to capture all that domain knowledge. 333 00:21:44,270 --> 00:21:49,550 Now, I need to teach this tool how to think. 334 00:21:50,510 --> 00:21:57,950 And there are four pillars around which I want it, when it has a conversation with me, I don't want it to give me answers. 335 00:21:58,590 --> 00:21:59,990 I want a friggin' conversation. 336 00:22:00,110 --> 00:22:01,710 So there's four things I'm going to do. 337 00:22:01,710 --> 00:22:03,870 Number one, I'm going to upload the Socratic method. 338 00:22:04,430 --> 00:22:08,110 The 7 questions that Socrates asked his students, he only asked 6. 339 00:22:08,110 --> 00:22:12,910 We had to add a 7th one because Socrates didn't care about sources of data. 340 00:22:13,390 --> 00:22:13,870 I do. 341 00:22:14,510 --> 00:22:15,870 I do care about sources of data. 342 00:22:16,190 --> 00:22:17,790 Very much care about sources of data. 343 00:22:17,790 --> 00:22:19,470 Some sources where I don't want to get data from. 344 00:22:19,710 --> 00:22:21,230 So we uploaded the Socratic method. 345 00:22:21,230 --> 00:22:25,390 So immediately, instead of giving me answers, it's answering me with questions. 346 00:22:26,470 --> 00:22:27,310 What are the perspectives? 347 00:22:27,310 --> 00:22:28,350 What are the rationales? 348 00:22:28,350 --> 00:22:29,710 What are the different diversity? 349 00:22:29,950 --> 00:22:35,550 I'm immediately getting, it's not a tool that gives me answers, it's a tool that starts to think better. 350 00:22:35,550 --> 00:22:38,430 If I want a tool that gives me better answers, I need to have a tool that thinks better. 351 00:22:38,670 --> 00:22:42,510 Number 2, I'm going to upload an ethical foundation. 352 00:22:43,390 --> 00:22:45,470 I upload the books of Matthew and Luke. 353 00:22:47,070 --> 00:22:52,350 Those books cover the parable of the Good Samaritan, which is, we can love the higher Bible if you want. 354 00:22:52,350 --> 00:22:54,110 But I like Luke and Matthew. 355 00:22:54,350 --> 00:22:55,990 I get the parable of the Good Samaritan, right? 356 00:22:55,990 --> 00:22:58,030 A huge difference between do no harm and do good. 357 00:22:58,590 --> 00:23:00,190 One's passive, one's active. 358 00:23:00,670 --> 00:23:03,830 And the parable and the lessons around how to help others, right? 359 00:23:03,950 --> 00:23:05,630 One of my students loaded the Quran. 360 00:23:06,670 --> 00:23:08,190 gives it an ethical foundation. 361 00:23:08,190 --> 00:23:13,070 So it's making decisions, making recommendations, having a conversation based on ethical foundations. 362 00:23:13,070 --> 00:23:19,470 When I upload it, I say, tell me, think like Jesus, think like Buddha, think like whoever you embrace, right? 363 00:23:20,590 --> 00:23:25,230 Number 3, humans make bad decisions. 364 00:23:25,870 --> 00:23:27,390 We're horrible decision makers. 365 00:23:27,710 --> 00:23:30,910 If you want any proof of that, just go to Las Vegas. 366 00:23:33,510 --> 00:23:46,190 And so we upload the 19 decision flaws that humans have and we tell it, help me to avoid making confirmation bias, make it recency bias, making fear of missing out, missing some, right? 367 00:23:46,350 --> 00:23:49,310 We train it to say, don't let me fall trap to that. 368 00:23:49,790 --> 00:23:55,070 And finally, we upload the UN sustainability framework so we have a sustainability perspective as well. 369 00:23:55,470 --> 00:23:57,630 So now when I'm having a conversation, 370 00:23:58,350 --> 00:24:04,510 With Yoda, I'm having a conversation with Socrates and Jesus and guys who don't go to Vegas. 371 00:24:04,670 --> 00:24:10,670 All right, now we're getting to the fun stuff. 372 00:24:12,750 --> 00:24:13,390 Experts. 373 00:24:14,270 --> 00:24:17,390 I can have this tool act as an expert. 374 00:24:17,390 --> 00:24:26,990 So for example, if I'm doing a product design conversation, I can say to Yoda, bring me together a panel of Jonathan Ivey, 375 00:24:27,470 --> 00:24:32,350 from Apple, Steven Jobs, David Kelly from IDEO, and Aristotle. 376 00:24:33,550 --> 00:24:34,270 Why not Aristotle? 377 00:24:34,270 --> 00:24:35,230 Throw them into the mix, right? 378 00:24:35,630 --> 00:24:41,150 I can say bring them together and I can start posting questions and getting answers from those experts, right? 379 00:24:42,270 --> 00:24:44,990 I can start adding expert data here. 380 00:24:44,990 --> 00:24:47,030 There's one gotcha here. 381 00:24:47,030 --> 00:24:48,430 I don't think it's in here. 382 00:24:48,910 --> 00:24:49,310 Sorry. 383 00:24:49,310 --> 00:24:53,470 This is a problem in only 45 minutes. 384 00:24:54,750 --> 00:24:55,230 One of the 385 00:24:57,390 --> 00:24:59,230 untruths about these Gen. 386 00:24:59,230 --> 00:24:59,870 AI tools. 387 00:25:00,350 --> 00:25:03,950 People say, well, the more conversations you have with this tool, the smarter it gets. 388 00:25:05,470 --> 00:25:06,190 That ain't true. 389 00:25:07,390 --> 00:25:08,030 That ain't true. 390 00:25:08,670 --> 00:25:18,030 You have a conversation with Steven Jobs and Jonathan Ivey and Aristotle and David Kelly and you get lots of great information, it's going to forget that. 391 00:25:18,630 --> 00:25:23,550 It's got a memory cap and it gets flushed out every couple of days, couple of weeks. 392 00:25:24,110 --> 00:25:25,950 In order to sustain that, 393 00:25:26,430 --> 00:25:37,390 You're going to copy that conversation, that narrative, their responses into a Word document or a PDF and upload it into Yoda with a prompt that says, factor these insights going forward. 394 00:25:38,510 --> 00:25:46,670 So it doesn't learn from the conversation, but we can turn the conversation, the feedback we get into knowledge that we upload back into Yoda. 395 00:25:48,270 --> 00:25:50,750 Pretty simple process, but don't think for a minute. 396 00:25:50,910 --> 00:25:51,950 I mean, I know my tool. 397 00:25:52,270 --> 00:25:56,590 I always like my slides in landscape format with a transparent background. 398 00:25:57,110 --> 00:26:01,230 And by the second or third day, it's all forgotten about transparent background. 399 00:26:01,670 --> 00:26:02,790 I'm like, why do you keep forgetting? 400 00:26:02,790 --> 00:26:04,550 And they'll say, well, I don't have a very big memory. 401 00:26:04,550 --> 00:26:06,270 Well, that's true, right? 402 00:26:06,270 --> 00:26:08,910 So I moved away from memory to storage. 403 00:26:09,190 --> 00:26:14,030 I'm creating a contextual knowledge base by loading more and more of these conversations up. 404 00:26:14,030 --> 00:26:15,310 So there's a way around this. 405 00:26:18,750 --> 00:26:24,270 So finally now, I want to synthesize, I want to validate, and I want to get an answer. 406 00:26:24,270 --> 00:26:33,550 So I say, okay, based on your research, I'm A 10,000 acre farm in Northeast Iowa, what should I plant? 407 00:26:34,350 --> 00:26:40,670 And it comes up with an answer and some rationale, some recommendations as far as why it came up with that. 408 00:26:40,910 --> 00:26:43,870 Right, and if you had a balance, it says, hey, I'm balancing 409 00:26:44,430 --> 00:26:46,030 profit and soil health. 410 00:26:46,030 --> 00:26:48,910 I'm improving input efficiency and building resilience or diversification. 411 00:26:49,230 --> 00:26:51,150 So it gave me an answer. 412 00:26:53,070 --> 00:26:54,670 I can start having a conversation with it now. 413 00:26:55,550 --> 00:26:56,590 I can say, okay, this is great. 414 00:26:56,590 --> 00:26:57,750 I can ask for more details. 415 00:26:57,750 --> 00:26:58,350 What about this? 416 00:26:58,510 --> 00:27:02,990 So I can say, well, what happens if we get into a trade war? 417 00:27:04,350 --> 00:27:05,870 How does that change? 418 00:27:06,750 --> 00:27:10,590 So I've got a prompt that says, what's the potential impact of a 50% trade on 419 00:27:10,910 --> 00:27:16,910 With Mexico and Canada, it gives me some feedback, it gives me a recommendation and a tariff strategy. 420 00:27:17,870 --> 00:27:19,990 I can turn this into a conversation piece. 421 00:27:19,990 --> 00:27:26,750 And of course, when I get this feedback, I copy it into a slide or a Word, and what am I doing? 422 00:27:26,990 --> 00:27:28,350 I'm uploading it back into Yoda. 423 00:27:29,150 --> 00:27:33,550 I am constantly training it by loading it back into my contextual knowledge base. 424 00:27:34,350 --> 00:27:37,390 So now I have a vehicle for having all these kind of what-if conversations. 425 00:27:37,790 --> 00:27:40,590 But I can have this with a high degree of confidence 426 00:27:41,070 --> 00:27:44,030 Because I have trained it on the problem I'm trying to go after. 427 00:27:44,270 --> 00:27:45,710 I have trained it on my intent. 428 00:27:45,950 --> 00:27:47,950 I have trained it on my desired outcomes. 429 00:27:48,110 --> 00:27:51,590 I've trained it on thinking like a Socrates and having ethical behaviors. 430 00:27:51,590 --> 00:27:52,990 I've trained it like thinking like an expert. 431 00:27:52,990 --> 00:27:54,670 I've done all this training. 432 00:27:55,790 --> 00:27:56,990 All this training to get here. 433 00:27:58,590 --> 00:27:59,430 A lot of work. 434 00:27:59,430 --> 00:28:01,950 A lot of work setting this up. 435 00:28:02,550 --> 00:28:10,350 But this is the difference between a tool that's going to give you the average answer versus the tool that's going to give you the average of experts. 436 00:28:10,950 --> 00:28:14,510 because I've trained it on the problem and the experts who I seek guidance from. 437 00:28:15,550 --> 00:28:16,830 I got 10 minutes, okay? 438 00:28:17,710 --> 00:28:18,510 Last section. 439 00:28:19,150 --> 00:28:21,590 I promised causal AI, and we're going to talk about causal AI. 440 00:28:21,590 --> 00:28:25,670 We're going to talk about a healthcare example. 441 00:28:25,670 --> 00:28:27,590 And here's the example we're going to walk through. 442 00:28:27,590 --> 00:28:38,670 We're going to talk about how putting AI in the middle, using causal factors, allows me to drive a better relationship between my doctor, nurse, 443 00:28:39,230 --> 00:28:40,430 patient and treatments. 444 00:28:41,230 --> 00:28:46,270 I'm getting down to the ultimate of 1 to 1 on my engagement. 445 00:28:46,270 --> 00:28:50,350 I'm not making decisions based on correlations, based on generalities. 446 00:28:51,030 --> 00:28:58,870 I'm going to make decisions based on the individual behavioral performance propensities of the doctor, the nurse, the patient, and the treatment effect. 447 00:28:58,870 --> 00:29:04,510 We're going to put AI in the middle to help us, not replace us. 448 00:29:04,510 --> 00:29:05,470 Here's how it works. 449 00:29:05,950 --> 00:29:10,830 It's a concept I developed back when I was at Yahoo many, many, many, many, many years ago called nanoeconomics. 450 00:29:11,710 --> 00:29:20,350 And that is, when you think about my challenge at Yahoo back 20-some years ago, I had 500 million visitors a month coming to my site. 451 00:29:20,990 --> 00:29:23,790 And I had a fraction of a second to figure out what ad to show them. 452 00:29:24,670 --> 00:29:33,230 I had to know not only what ad to show them, but what the value of that ad was, what the recency or the urgency of the need was for them, because in some cases I was bidding for those views. 453 00:29:34,670 --> 00:29:36,590 Now think about what I knew about you on Yahoo. 454 00:29:37,070 --> 00:29:38,590 I knew every site you went to. 455 00:29:38,910 --> 00:29:40,990 I knew every click you made. 456 00:29:41,070 --> 00:29:42,390 I knew every ad you clicked on. 457 00:29:42,390 --> 00:29:43,710 I knew every ad you didn't click on. 458 00:29:43,710 --> 00:29:44,950 I knew what you put on social media. 459 00:29:44,950 --> 00:29:46,030 I knew what you put in searches. 460 00:29:46,590 --> 00:29:48,030 I knew what you wrote on your e-mail. 461 00:29:48,030 --> 00:29:49,550 Sorry, you should have read the fine print. 462 00:29:50,030 --> 00:29:53,630 So I had this great level of detail about what I thought you were interested in. 463 00:29:54,390 --> 00:30:00,270 And we built a detailed propensity model on each and every one of those 200 million people. 464 00:30:01,190 --> 00:30:08,830 So when they came to my site, I had a fraction of a second to figure out what's their value, what's their urgency so I can make the right bid. 465 00:30:08,830 --> 00:30:16,990 For example, if I knew you were interested in vacations or car, getting you to click on an ad for a vacation or car gave me 19 bucks. 466 00:30:17,470 --> 00:30:20,670 If I knew you were interested in a cup of coffee, that paid less than a penny. 467 00:30:21,310 --> 00:30:26,990 So differentiation was built by my knowing more about each of these 200 million visitors. 468 00:30:27,270 --> 00:30:30,830 And by the way, these propensity models, these scores, I kept them on your cookies. 469 00:30:32,190 --> 00:30:33,070 Pretty simple game. 470 00:30:33,550 --> 00:30:35,150 You erase your cookie and I was blind. 471 00:30:35,950 --> 00:30:37,310 But most people didn't erase cookies. 472 00:30:38,110 --> 00:30:39,790 So this is a concept we're going to embrace. 473 00:30:40,030 --> 00:30:50,590 And so we're going to build very detailed profiles on each and every one of the players involved in a situation where we have a woman who's got cancer for the first time. 474 00:30:51,310 --> 00:30:54,350 She's older, she's very nervous about it, she's very scared. 475 00:30:55,350 --> 00:30:59,710 We're going to try to understand her situation from a causal factor 476 00:31:00,430 --> 00:31:03,710 in order to make sure we give her the best chance for the right kind of treatment. 477 00:31:05,150 --> 00:31:12,430 So the first thing we do, much like we did at Yahoo, we're going to build a very detailed model on her propensities. 478 00:31:13,230 --> 00:31:18,030 Treatment adherence and risk scores, behavioral response, social support. 479 00:31:18,670 --> 00:31:21,430 There's a whole bunch of variables and metrics we're going to capture. 480 00:31:21,430 --> 00:31:25,070 These are basically psycho scores, lack of a better word. 481 00:31:25,070 --> 00:31:27,630 We're building very detailed scores that are 482 00:31:28,670 --> 00:31:34,670 predictive of her performance and behavioral propensities. 483 00:31:34,950 --> 00:31:36,590 And we're going to turn them into a score. 484 00:31:37,230 --> 00:31:43,550 The more we have, the more detail we have about what makes this particular patient unique. 485 00:31:44,910 --> 00:31:46,910 We're not going to make decisions based on average. 486 00:31:46,910 --> 00:31:49,910 Based on average, women this age in this area get this kind of treatment. 487 00:31:49,910 --> 00:31:50,750 That's bullshit. 488 00:31:51,310 --> 00:31:52,670 That's not what AI can do. 489 00:31:53,150 --> 00:32:00,590 I want as much detail about her so I can personalize care, I can match, I can optimize, I can orchestrate real-time guidance. 490 00:32:00,750 --> 00:32:03,190 This is the nanoeconomics in reality. 491 00:32:03,190 --> 00:32:06,110 This is working to make sure we're providing the right kind of care. 492 00:32:07,870 --> 00:32:08,670 And guess what? 493 00:32:08,670 --> 00:32:10,110 I do the same thing for doctors. 494 00:32:10,590 --> 00:32:12,430 Not all doctors are created the same way. 495 00:32:12,910 --> 00:32:15,630 Some are more experienced, some are more empathy, more, right? 496 00:32:15,630 --> 00:32:19,710 So I'm going to build very detailed profile, quantifiable 497 00:32:20,110 --> 00:32:23,950 causal based profile on each and every one of my doctors and my nurses. 498 00:32:24,670 --> 00:32:28,190 So I can match the right doctor to the right patient. 499 00:32:28,670 --> 00:32:33,070 Given the patient's propensities and tendencies, I can match the right doctor. 500 00:32:33,470 --> 00:32:36,430 And same thing for my treatment methods. 501 00:32:37,230 --> 00:32:42,110 Very detailed profiles and side effects and efficiency and et cetera, et cetera. 502 00:32:43,470 --> 00:32:44,510 This is not hard to do. 503 00:32:45,470 --> 00:32:51,070 just means you have to have a different approach of understanding what are the variables and factors that are most important for care. 504 00:32:51,070 --> 00:32:56,430 What are the variables and factors that are most important deciding where you're going to go eat? 505 00:32:56,430 --> 00:32:57,550 A lot of similarities here. 506 00:32:58,190 --> 00:32:59,390 It just means that you... 507 00:33:00,070 --> 00:33:01,470 what I think the power of AI is. 508 00:33:02,510 --> 00:33:14,030 AI is ultimately going to make us more human because these are human conversations with the nurse, with the doctor to really understand where they are, what the propensities and behaviors are. 509 00:33:14,270 --> 00:33:17,870 This is where nurses and doctors, we excel. 510 00:33:19,710 --> 00:33:21,230 This is not what AI excels at. 511 00:33:21,310 --> 00:33:22,310 AI is great at matching. 512 00:33:22,990 --> 00:33:25,390 And so once we've got this, AI sits in the middle 513 00:33:26,110 --> 00:33:31,190 And it acts as this matching service, provides guidance, provides feedback as it goes around. 514 00:33:31,350 --> 00:33:40,110 We've got these very detailed profiles, and it's not only matching who the right doctor is for the right treatment and the right patient, but it's monitoring what works, what doesn't work. 515 00:33:40,430 --> 00:33:45,430 So it's going to give you a scoreboard. 516 00:33:45,430 --> 00:33:48,830 Let's see what the scoreboard looks like, shall we? 517 00:34:05,070 --> 00:34:06,590 Try to make it bigger, seat in the back. 518 00:34:06,910 --> 00:34:12,429 So this is a model that we're going to use to understand our patient in more detail. 519 00:34:12,429 --> 00:34:16,110 So you can see here's Sarah, her age, her situation. 520 00:34:17,630 --> 00:34:19,150 Propensity scores we built on her. 521 00:34:19,310 --> 00:34:21,150 We built and we update them all the time. 522 00:34:21,870 --> 00:34:23,670 She may become more tolerant to pain. 523 00:34:23,670 --> 00:34:24,870 She became less tolerant to pain. 524 00:34:24,870 --> 00:34:26,350 We want to know those changes. 525 00:34:26,989 --> 00:34:31,989 We can use language models to help us track that, but we've got to always turn it back into causal factors. 526 00:34:31,989 --> 00:34:35,030 We have to understand causal factors give us the why. 527 00:34:35,030 --> 00:34:36,670 And we need to understand why. 528 00:34:36,830 --> 00:34:39,790 If we understand why, then we can do interventions and counterfactuals. 529 00:34:39,790 --> 00:34:43,550 Interventions mean we can project before we make a decision what the likely outcome is. 530 00:34:43,550 --> 00:34:48,909 If we know the why, we can understand what kind of outcomes we think it's going to derive. 531 00:34:49,870 --> 00:34:50,190 So 532 00:34:51,070 --> 00:34:52,270 Here is Sarah. 533 00:34:52,590 --> 00:34:54,350 You see you got a whole bunch of information about her. 534 00:34:54,590 --> 00:34:55,949 You can kind of scroll down here. 535 00:34:56,190 --> 00:34:58,590 The doctor, as they have conversations, can change. 536 00:34:58,990 --> 00:35:07,230 Her pain sensitivity is not as high as we thought, and her emotional resilience is actually, where's the, she's got a really strong network, much stronger than we thought. 537 00:35:07,230 --> 00:35:09,070 She's got family nearby, et cetera, et cetera. 538 00:35:09,070 --> 00:35:17,110 So we can, as a doctor who's having these conversations, the doctor can use their expertise to help not necessarily override, but train this tool. 539 00:35:17,110 --> 00:35:20,270 All right, now we're going to match to a doctor. 540 00:35:20,990 --> 00:35:27,470 is to take a look at these propensities and say, across these doctors, here are the different factors we look at in doctors. 541 00:35:27,950 --> 00:35:29,470 Let me kind of scroll down here. 542 00:35:30,950 --> 00:35:32,270 And we think that Dr. 543 00:35:32,270 --> 00:35:34,750 Vasquez is the right one for us for the following reasons. 544 00:35:34,950 --> 00:35:38,790 And you can see that you've got matches here that talk about we think it's empathy and fear. 545 00:35:38,790 --> 00:35:43,390 If I select a different doctor, we can see that there's a risk regarding the patient fear management. 546 00:35:43,390 --> 00:35:44,910 This doctor here is not really good at that. 547 00:35:44,910 --> 00:35:46,270 So we select a doctor. 548 00:35:46,270 --> 00:35:47,950 Hopefully you can see, this is 549 00:35:48,350 --> 00:35:54,390 really quick going through this about how understanding the doctor's tendencies and behaviors, matching that with the patient. 550 00:35:54,390 --> 00:35:55,470 This is what AI does. 551 00:35:56,230 --> 00:35:59,110 It's great and it can tell you why it's matching why. 552 00:35:59,110 --> 00:36:02,190 And it gives you a chance to override if you think that's not correct. 553 00:36:02,430 --> 00:36:04,910 And the tool is going to learn when you override. 554 00:36:05,190 --> 00:36:07,750 It's going to get smarter by you driving the overrides. 555 00:36:07,750 --> 00:36:10,710 And then next one, I'm running out of time here, so I'm going to be quick. 556 00:36:10,710 --> 00:36:11,710 We're going to select a treatment. 557 00:36:11,710 --> 00:36:14,710 So it's also going to look across, I only pick five, three treatments. 558 00:36:14,710 --> 00:36:16,430 It's going to recommend the right treatment. 559 00:36:16,830 --> 00:36:17,430 blah, blah, blah. 560 00:36:17,430 --> 00:36:18,990 And then I'm going to generate an advisory. 561 00:36:18,990 --> 00:36:23,230 And then I go into my large language model and I say, okay, generate for me a plan. 562 00:36:23,550 --> 00:36:26,830 And I click this button and it starts going to the internet and it starts finding current research. 563 00:36:26,830 --> 00:36:29,070 It starts pulling things together, right? 564 00:36:29,430 --> 00:36:36,510 Intelligence and external validities and research that's out there and it brings it all together, gives me some information. 565 00:36:36,510 --> 00:36:38,030 I can develop a care plan. 566 00:36:38,270 --> 00:36:40,510 So now I have a plan that's recommended for me. 567 00:36:40,670 --> 00:36:44,270 I can either accept points on here 568 00:36:45,670 --> 00:36:46,670 We'll accept this one. 569 00:36:47,470 --> 00:36:48,990 We'll ignore this for a second. 570 00:36:50,990 --> 00:36:54,750 I can modify this one and say blah, blah, blah. 571 00:36:54,790 --> 00:36:57,870 And this one I'm going to basically reject, and it's going to say why. 572 00:36:57,870 --> 00:37:00,150 I'm going to give it a reason why I'm rejecting it. 573 00:37:00,150 --> 00:37:13,630 It's going to factor all that in and create a customized care plan for Sarah that talks about here's what your doctor is, here's what's going to happen, here's what we know about you, the whole treatment thing, it's all pulled together, and printed off for her. 574 00:37:15,230 --> 00:37:17,550 based on causal. 575 00:37:18,590 --> 00:37:32,590 We feed it causal insights and we leverage that matching capability of AI to deliver a personalized care package for Sarah that gives her and the doctors the best chance for success. 576 00:37:34,430 --> 00:37:36,990 By the way, I vibe coded that. 577 00:37:38,550 --> 00:37:42,350 If you know what vibe coding is, when you've gone through and built a Yoda, my students 578 00:37:43,790 --> 00:37:47,150 like 2 paragraphs of instructions, it'll build this for you. 579 00:37:48,350 --> 00:37:48,830 Why? 580 00:37:48,830 --> 00:37:50,750 Because you've trained Yoda. 581 00:37:51,590 --> 00:37:53,550 And now Yoda knows the problem that you're trying to do. 582 00:37:53,550 --> 00:37:55,430 Yoda become a sort of mini-me of your mind. 583 00:37:55,430 --> 00:37:57,390 And it creates that for you. 584 00:37:58,190 --> 00:37:59,430 But this is how you bring together. 585 00:37:59,430 --> 00:38:02,910 This is where I think the big aha moment is going to be in the last two minutes here. 586 00:38:03,190 --> 00:38:06,590 These generative AI tools are great at providing context. 587 00:38:07,190 --> 00:38:12,030 It's great at finding all this information relative to what's important to me, but it doesn't know causality. 588 00:38:12,510 --> 00:38:15,870 If I use entity propensity models, EPMs, these are causal driven. 589 00:38:15,990 --> 00:38:28,390 If I bring causal together with my generative AI, I have that ability to deliver very, not only more personalized care, but understand the whys behind why it recommends this versus that. 590 00:38:28,390 --> 00:38:30,070 And then I can do interventions. 591 00:38:30,070 --> 00:38:31,710 I can say, what if I do this instead? 592 00:38:31,990 --> 00:38:34,430 And it can project what sort of outcome you're going to have from that. 593 00:38:35,630 --> 00:38:37,230 This is where the marketplace is heading. 594 00:38:37,790 --> 00:38:38,830 It's a game changer. 595 00:38:39,150 --> 00:38:43,390 It takes a lot more work than just throwing a bunch of random data into a generative AI tool. 596 00:38:44,230 --> 00:38:48,990 There's no easy button for this because the hard part about it is the domain expertise of your people. 597 00:38:50,270 --> 00:38:52,350 That is where the power is. 598 00:38:52,910 --> 00:38:54,110 It's in the people. 599 00:38:54,390 --> 00:38:56,270 It's not in the technology. 600 00:38:56,830 --> 00:39:00,350 And so anyway, I hope you found that useful, valuable. 601 00:39:00,350 --> 00:39:01,590 You know where the marketplace is going. 602 00:39:01,590 --> 00:39:02,590 You know where to find me. 603 00:39:03,150 --> 00:39:07,710 And let me wrap up with cool way slide. 604 00:39:15,110 --> 00:39:15,910 Here's how you reach me. 605 00:39:15,910 --> 00:39:17,150 You can find me on LinkedIn. 606 00:39:17,710 --> 00:39:20,430 By the way, I have a Dina Big Data GPT. 607 00:39:20,590 --> 00:39:24,950 If you use ChatGPT and you go to their library, it's like a library of GPTs. 608 00:39:24,950 --> 00:39:27,190 You can find Dina Big Data GPT. 609 00:39:27,190 --> 00:39:28,510 It's a mini me. 610 00:39:28,950 --> 00:39:30,470 It's got all my books, all my blogs. 611 00:39:30,470 --> 00:39:32,190 Anytime I write a blog, it's uploaded in there. 612 00:39:32,550 --> 00:39:35,910 It's smarter than me from a under-signed stuff, right? 613 00:39:35,910 --> 00:39:40,990 If I get ready to write a blog and then come back and say, geez, Mars, you wrote a blog about that about six years ago. 614 00:39:40,990 --> 00:39:41,710 Like, really? 615 00:39:42,110 --> 00:39:43,510 Six years ago, I was writing about this stuff. 616 00:39:43,550 --> 00:39:48,190 Anyway, so anyway, so if you have questions, and I hope you have questions, I'm going to be around about the next hour or so. 617 00:39:48,750 --> 00:39:51,150 Find me, ask me questions. 618 00:39:51,550 --> 00:39:53,790 I learn when you ask me questions. 619 00:39:54,190 --> 00:39:57,910 I teach because my students are asking me questions, which allows me to learn. 620 00:39:57,910 --> 00:40:00,030 And I'm here to learn, and I hope you're here to learn as well. 621 00:40:00,190 --> 00:40:00,910 Thanks for your time. 622 00:40:07,480 --> 00:40:07,880 Thanks. 623 00:40:10,040 --> 00:40:15,000 Correlation versus causation is one of those things that we like to go lecture everybody on. 624 00:40:15,000 --> 00:40:19,840 We yell at our families and that has nothing to do with them. 625 00:40:19,840 --> 00:40:24,720 But the actual application of it in this quantitative application.