Transform your leadership with AI strategy for executives

Description of the video:

WEBVTT 1 00:00:00.705 --> 00:00:02.915 [Kim Allison] Alright, well, it's one o'clock, 2 00:00:03.135 --> 00:00:04.675 so we might go ahead and kick it off. 3 00:00:04.835 --> 00:00:07.755 I wanna be respectful of everyone's time, so, um, 4 00:00:07.785 --> 00:00:09.715 just wanna say hi everyone. 5 00:00:09.765 --> 00:00:11.755 Thank you so much for joining us today. 6 00:00:12.165 --> 00:00:13.755 We're so thrilled to have such a great 7 00:00:13.755 --> 00:00:15.155 group, um, in this session. 8 00:00:15.515 --> 00:00:18.395 I think last time I checked we had over 700 people 9 00:00:18.995 --> 00:00:20.755 registered, so that's really fantastic. 10 00:00:21.295 --> 00:00:22.795 So thanks, thanks for taking an hour 11 00:00:22.815 --> 00:00:23.915 out of your day to join us. 12 00:00:24.335 --> 00:00:26.835 Um, I'm Kim Allison with Kelley Executive Education. 13 00:00:27.065 --> 00:00:29.275 It's my pleasure to introduce our speaker today, 14 00:00:29.335 --> 00:00:33.475 Dr. Alex Lopes, who will be diving into the future of AI 15 00:00:33.575 --> 00:00:34.795 and impacts on business. 16 00:00:34.815 --> 00:00:36.795 So, really exciting topic for today. 17 00:00:37.375 --> 00:00:40.675 Dr. Lopes is a clinical professor of information systems 18 00:00:41.255 --> 00:00:44.275 and Grant Thornton Scholar at the Kelley School of Business. 19 00:00:44.735 --> 00:00:47.155 He also serves as associate chair 20 00:00:47.215 --> 00:00:49.275 for Kelley Executive Education Programs, 21 00:00:49.275 --> 00:00:51.395 where he leads our Online MS programs 22 00:00:51.395 --> 00:00:52.635 and professional certificates. 23 00:00:53.025 --> 00:00:55.795 With that, I'd like to hand it over to Dr. Lopes, 24 00:00:55.815 --> 00:00:58.315 and thank you so much for your time and joining us today. 25 00:00:59.425 --> 00:01:00.815 [Alex Barsi Lopes] Thank you so much, Kim. 26 00:01:00.975 --> 00:01:03.575 I hope everybody can, um, hear me well 27 00:01:03.795 --> 00:01:07.415 and also, um, see the slides as we go through. 28 00:01:07.715 --> 00:01:11.015 Um, this weird accent that you hear right now is, 29 00:01:11.115 --> 00:01:12.615 I'm from Brazil, so, 30 00:01:12.835 --> 00:01:14.895 and actually I just came from Brazil yesterday, 31 00:01:15.035 --> 00:01:17.415 so my accent might be a little bit stronger than 32 00:01:17.415 --> 00:01:18.775 usual, so sorry for that. 33 00:01:19.595 --> 00:01:21.015 Um, so let's start. 34 00:01:21.275 --> 00:01:24.895 And, uh, so we kind of, uh, try to title this webinar 35 00:01:25.565 --> 00:01:26.615 more provocative, right? 36 00:01:26.615 --> 00:01:30.855 The Future is AI, uh, Artificial Intelligence for Business. 37 00:01:31.195 --> 00:01:35.055 But having said of that, if you're expecting, you know, 38 00:01:35.165 --> 00:01:36.935 some, like, you know, cool demos 39 00:01:37.035 --> 00:01:41.375 and some like, you know, hype about AI, um, sorry, I'm going 40 00:01:41.375 --> 00:01:44.895 to disappoint you because this is really, I believe, 41 00:01:44.895 --> 00:01:48.495 an opportunity for us to take stock about where we are 42 00:01:48.835 --> 00:01:52.775 and, uh, to really think about how do we move from, 43 00:01:53.475 --> 00:01:56.575 you know, the, the, hype to something that is actually 44 00:01:57.095 --> 00:01:59.855 valuable for organizations, really understand, you know, 45 00:01:59.855 --> 00:02:02.615 where we are right now, what are the challenges, 46 00:02:02.885 --> 00:02:05.175 what are the risks associated with AI? 47 00:02:05.675 --> 00:02:07.255 And eventually try to come up 48 00:02:07.255 --> 00:02:11.255 with some recommendations about how to navigate those risks, 49 00:02:11.555 --> 00:02:13.735 the kind of things that you need to pay attention 50 00:02:14.555 --> 00:02:18.495 if you are, you know, involved in somehow in your, uh, 51 00:02:18.555 --> 00:02:20.735 the AI journey of your organization. 52 00:02:21.515 --> 00:02:22.735 So this is the agenda. 53 00:02:22.785 --> 00:02:24.895 We're going to do a little bit of a recap, right? 54 00:02:24.995 --> 00:02:27.415 Uh, essentially where we then talk about 55 00:02:27.425 --> 00:02:30.815 where we are right now in both applied and GenAI. 56 00:02:31.355 --> 00:02:35.135 Uh, what are the main risks? Why do most projects fail? 57 00:02:35.915 --> 00:02:39.135 And, um, how to succeed, uh, in the, in that, 58 00:02:39.135 --> 00:02:40.175 in the, in that scenario. 59 00:02:40.485 --> 00:02:44.095 Okay? Um, so, um, one of things that you're going to see is 60 00:02:44.095 --> 00:02:47.175 that, uh, you know, I'm going to present several, uh, 61 00:02:47.235 --> 00:02:51.295 in my slides that have lots of references, um, for you 62 00:02:51.295 --> 00:02:54.335 to explore later on, some no tech, you know, trains 63 00:02:54.335 --> 00:02:56.215 and some additional things that we can see. 64 00:02:56.635 --> 00:02:59.565 Um, so hopefully you'll find some additional resources 65 00:02:59.625 --> 00:03:03.005 to explore further after we are done. 66 00:03:03.505 --> 00:03:07.525 Uh, my target's about 40 minutes or so of me talking, 67 00:03:07.625 --> 00:03:10.605 and then I can answer all, any, any questions remaining. 68 00:03:10.985 --> 00:03:13.605 Um, or, but, you know, in the end, 69 00:03:13.625 --> 00:03:17.205 before passing back to Kim, uh, for, for the wrap. 70 00:03:17.865 --> 00:03:20.765 So let's start with this idea about, uh, you know, 71 00:03:21.175 --> 00:03:24.925 where does all this, this AI stuff, uh, come from, right? 72 00:03:25.465 --> 00:03:28.565 And, and really if you think about AI, I have been working 73 00:03:28.675 --> 00:03:29.765 with AI, I, I, 74 00:03:29.965 --> 00:03:33.325 I did like neural networks like back in the 90s, uh, 75 00:03:33.355 --> 00:03:35.565 when things were just much more like, you know, 76 00:03:35.565 --> 00:03:36.765 experimental, right? 77 00:03:36.765 --> 00:03:40.365 Nobody really knew what to do with those things 78 00:03:40.675 --> 00:03:44.845 because there are several pieces that were missing. 79 00:03:45.665 --> 00:03:48.725 And what we have seen in the last, uh, decade 80 00:03:48.905 --> 00:03:50.765 or so, a little bit more than that, is 81 00:03:50.765 --> 00:03:55.645 that essentially these three factors that really have led AI 82 00:03:55.905 --> 00:03:59.685 to this, this prominent position it has right now where a lot 83 00:03:59.685 --> 00:04:01.285 of people are excited about it, right? 84 00:04:01.825 --> 00:04:03.205 And the three factors essentially 85 00:04:03.205 --> 00:04:04.245 is the presence of big data. 86 00:04:04.425 --> 00:04:05.885 We have much, much more data. 87 00:04:05.935 --> 00:04:08.845 We're collecting more data in unprecedented levels, right? 88 00:04:09.545 --> 00:04:10.965 Any, anyone of us 89 00:04:10.995 --> 00:04:13.405 with a cell phone is constantly collecting data. 90 00:04:14.025 --> 00:04:16.285 Uh, we have much more computing power, right. 91 00:04:16.345 --> 00:04:18.925 One of the, the biggest companies in the world right now is 92 00:04:19.105 --> 00:04:21.125 Nvidia, uh, you know, for those of us 93 00:04:21.125 --> 00:04:22.885 that were gamers, right? 94 00:04:23.205 --> 00:04:25.245 Remember Nvidia for the graphic, you know, uh, 95 00:04:25.245 --> 00:04:26.405 processing units, right? 96 00:04:26.425 --> 00:04:28.325 And now it's essentially an AI company. 97 00:04:28.825 --> 00:04:31.885 So the computing power is advancing very, very quickly. 98 00:04:32.425 --> 00:04:35.525 And then you have the algorithms, uh, what exactly is, 99 00:04:35.525 --> 00:04:39.925 is is essentially mobilizing our ability to do things 100 00:04:39.925 --> 00:04:41.045 with AI, right? 101 00:04:41.045 --> 00:04:44.165 These algorithms, were all, a lot of them were kind 102 00:04:44.165 --> 00:04:47.525 of theoretical until we start really have the computing 103 00:04:47.525 --> 00:04:50.365 power and big data to actually, you know, train them 104 00:04:50.505 --> 00:04:52.845 and to start using them in, in real life. 105 00:04:53.345 --> 00:04:55.125 But the algorithms have been really 106 00:04:55.315 --> 00:04:57.165 advancing, uh, quite a lot. 107 00:04:58.815 --> 00:05:01.235 And, you know, again, just a recap, right? 108 00:05:01.455 --> 00:05:05.075 We have these two kind of main categories. 109 00:05:05.775 --> 00:05:07.955 Uh, if you think about AI, right? 110 00:05:08.125 --> 00:05:10.315 Algorithms, so you, you have the more 111 00:05:10.385 --> 00:05:11.715 statistical machine learning. 112 00:05:12.215 --> 00:05:13.875 We have things like supervised learning, 113 00:05:13.975 --> 00:05:16.355 unsupervised learning, like no cluster analysis. 114 00:05:16.895 --> 00:05:19.555 We have traditional reinforcement learning, um, 115 00:05:19.565 --> 00:05:21.315 which I'll talk a little bit about other types 116 00:05:21.315 --> 00:05:22.715 of reinforcement learning in a second. 117 00:05:23.135 --> 00:05:25.155 Um, and then you have all the deep learning, 118 00:05:25.255 --> 00:05:27.795 the more traditional deep learning approaches, right? 119 00:05:27.795 --> 00:05:31.555 Convolutional networks for, um, uh, image 120 00:05:31.655 --> 00:05:36.035 and video processing, uh, RNNs, GANs, diffusion models, 121 00:05:36.415 --> 00:05:39.075 things that have been around for, for some time 122 00:05:39.095 --> 00:05:40.755 and they have proved to be effective. 123 00:05:41.415 --> 00:05:45.995 Um, but they, they're kind of not very sexy, right? 124 00:05:46.315 --> 00:05:49.435 I mean, we have a lot of sexy stuff that's coming right now. 125 00:05:50.095 --> 00:05:53.035 Um, but one thing that I want to highlight to everybody is 126 00:05:53.035 --> 00:05:54.995 that those things work really well. 127 00:05:55.655 --> 00:05:58.805 Um, you know, the traditional non-sexy stuff, right? 128 00:05:58.865 --> 00:06:00.725 It still works extremely well 129 00:06:01.265 --> 00:06:04.685 and are widely used, uh, in many organizations. 130 00:06:05.305 --> 00:06:08.845 It is just that if you see where you are right now, we, 131 00:06:08.845 --> 00:06:11.165 we have this whole idea about transformers, right? 132 00:06:11.665 --> 00:06:16.005 And, and transformers are weak, have kind of revolutionized 133 00:06:16.965 --> 00:06:19.725 a lot of the stuff that, that are coming in, 134 00:06:19.725 --> 00:06:22.525 especially in the last four or five years. 135 00:06:22.645 --> 00:06:24.605 I mean, transformers were, you know, uh, 136 00:06:24.635 --> 00:06:26.005 developed before that. 137 00:06:26.185 --> 00:06:28.925 But the practical implications of that, right? 138 00:06:28.945 --> 00:06:33.005 And transformers in, in the very baseline are really this, 139 00:06:33.115 --> 00:06:36.765 this multidimensional matrices that are looking 140 00:06:36.825 --> 00:06:37.965 for correlations, right? 141 00:06:38.025 --> 00:06:41.885 If you define your basic large language model, essentially, 142 00:06:41.945 --> 00:06:43.645 is a model that is anticipating 143 00:06:43.645 --> 00:06:45.565 what the next word is, right? 144 00:06:45.825 --> 00:06:47.885 We know it's more powerful than that at this point, 145 00:06:48.265 --> 00:06:50.245 but that's how it kind of is started. 146 00:06:51.105 --> 00:06:54.165 Um, and the main idea here is that you have more parameters, 147 00:06:54.475 --> 00:06:57.645 more training, then the outcomes are better 148 00:06:58.235 --> 00:07:01.645 than if you have fewer of parameters and data, right? 149 00:07:02.105 --> 00:07:05.485 So the baseline of large language model are this type of, 150 00:07:05.505 --> 00:07:07.805 uh, on your network, this type of deep learning network, 151 00:07:07.805 --> 00:07:09.645 which are, you know, transformers. 152 00:07:10.425 --> 00:07:13.605 And again, they evolved, essentially dealing with text, 153 00:07:14.305 --> 00:07:16.885 but now they have been, um, essentially captured 154 00:07:17.105 --> 00:07:20.525 and extended to do with a variety of different types 155 00:07:20.585 --> 00:07:23.685 of problems, which makes our life very exciting, right? 156 00:07:23.705 --> 00:07:25.205 Is the, the, the sexy stuff, right? 157 00:07:25.205 --> 00:07:28.565 That's all the things that we have seen around in terms 158 00:07:28.585 --> 00:07:31.325 of new developments, advertisements, releases 159 00:07:31.325 --> 00:07:32.725 of new models, and so on and so forth. 160 00:07:34.025 --> 00:07:35.085 So that's where you are, 161 00:07:35.175 --> 00:07:37.685 we're in a very exciting time, right? 162 00:07:37.855 --> 00:07:39.885 We're very, uh, in a very good moment 163 00:07:39.885 --> 00:07:42.245 because there's a lot going on, right, 164 00:07:42.865 --> 00:07:44.365 um, in terms of where it came from. 165 00:07:44.385 --> 00:07:46.605 But now let's talk a little bit about where we are. 166 00:07:46.825 --> 00:07:49.965 What's the current state, uh, of these technologies 167 00:07:50.585 --> 00:07:51.965 in a very high level, right? 168 00:07:51.965 --> 00:07:54.805 We're not going to be, you know, doing demonstrations of, 169 00:07:54.825 --> 00:07:55.925 you know, transformers 170 00:07:56.665 --> 00:07:59.565 and post-training approaches, et cetera. 171 00:07:59.975 --> 00:08:01.405 We're going to kinda see, you know, 172 00:08:01.855 --> 00:08:04.725 where are the capabilities from a business perspective. 173 00:08:04.745 --> 00:08:08.125 And that's pretty much the main focus of this conversation, 174 00:08:08.865 --> 00:08:10.805 is we're looking at everything from the business lenses. 175 00:08:11.585 --> 00:08:14.725 So if you look in terms of where, you know, 176 00:08:14.755 --> 00:08:18.325 what are the latest and greatest in the AI investments. 177 00:08:18.345 --> 00:08:23.125 So McKinsey runs this, uh, annual, um, uh, survey, right? 178 00:08:23.125 --> 00:08:26.285 Looking at the macro, uh, essentially technology trains. 179 00:08:26.905 --> 00:08:28.845 And this was released, you know, um, 180 00:08:29.225 --> 00:08:30.525 couple of weeks ago, I guess. 181 00:08:31.225 --> 00:08:33.245 And, uh, what they're looking is, 182 00:08:33.535 --> 00:08:37.085 there are essentially these two kind of exciting areas. 183 00:08:37.265 --> 00:08:39.885 One is the whole of artificial intelligence, right? 184 00:08:40.425 --> 00:08:44.885 And you see essentially huge investments, a lot 185 00:08:44.885 --> 00:08:46.285 of job postings, right? 186 00:08:46.285 --> 00:08:50.285 That's all to have seen AI AI, AI, right? All the time. 187 00:08:51.265 --> 00:08:52.805 The new kid in the block, right? 188 00:08:52.945 --> 00:08:54.405 The exciting new thing, or, 189 00:08:54.425 --> 00:08:56.845 or maybe if we're more of a skeptical, right, 190 00:08:56.905 --> 00:09:00.565 the, the shiny new object is agentic AI. We're going 191 00:09:00.565 --> 00:09:02.325 to talk about agentic AI in a second. 192 00:09:02.825 --> 00:09:05.005 But this is where we start to see a lot, 193 00:09:05.005 --> 00:09:06.765 some investment come into life, 194 00:09:07.385 --> 00:09:09.285 and we start to see a huge expansion in terms 195 00:09:09.285 --> 00:09:10.325 of job postings. 196 00:09:10.325 --> 00:09:14.165 People are interested in really developing this agentic AI, 197 00:09:14.705 --> 00:09:17.205 uh, uh, processes and models to that. 198 00:09:17.225 --> 00:09:19.205 So this is where we are right now. 199 00:09:20.845 --> 00:09:23.425 Now, this is where we were just very recently, right? 200 00:09:23.485 --> 00:09:25.225 So I just want to make a comparison 201 00:09:25.775 --> 00:09:29.945 because if you look at the size of the investment in terms 202 00:09:29.945 --> 00:09:34.385 around $124 billion, um, that's kind 203 00:09:34.385 --> 00:09:38.305 of remains somewhat constant from 2024 to 2025. 204 00:09:39.005 --> 00:09:42.065 Um, but the numbers here a little bit spread in different 205 00:09:42.065 --> 00:09:43.505 directions because part is 206 00:09:43.505 --> 00:09:47.145 because, uh, agentic AI, it really has took, has taken off, 207 00:09:47.685 --> 00:09:50.865 um, in the tail end of 2024, right? 208 00:09:50.905 --> 00:09:54.225 That's when you start see a lot of, uh, agents 209 00:09:54.285 --> 00:09:55.465 and agents, right, 210 00:09:55.525 --> 00:09:56.825 and starting to come to life. 211 00:09:57.405 --> 00:10:00.625 But this is where we see right now, you know, uh, in terms 212 00:10:00.625 --> 00:10:03.665 of the distribution under that AI umbrella, right? 213 00:10:03.845 --> 00:10:05.105 The, a big AI umbrella 214 00:10:05.125 --> 00:10:09.265 of $124 billion investment, this is where it's kind 215 00:10:09.265 --> 00:10:10.265 of segmented, right? 216 00:10:10.865 --> 00:10:13.785 Some of it is generative AI, some 217 00:10:13.785 --> 00:10:16.185 of it is essentially machine learning, right? 218 00:10:16.185 --> 00:10:17.665 The more traditional approach 219 00:10:17.685 --> 00:10:19.665 for machine learning in industry, right, 220 00:10:20.065 --> 00:10:21.745 industry 4.0 and things like that. 221 00:10:22.365 --> 00:10:26.345 And a lot of it is just your general applied AI, right? 222 00:10:26.765 --> 00:10:28.265 Things that have been reliable 223 00:10:28.365 --> 00:10:30.505 before, things that have been working, right? 224 00:10:30.575 --> 00:10:34.305 Some machine learning, some algorithms that have been used 225 00:10:34.405 --> 00:10:38.105 to generate a lot of positive results, right? 226 00:10:38.565 --> 00:10:42.985 So, um, again, the main takeaway here would be loss of, 227 00:10:43.085 --> 00:10:46.785 of investment, loss of interest, um, 228 00:10:47.005 --> 00:10:49.625 but a little bit of variety in terms of 229 00:10:50.285 --> 00:10:51.665 what's exciting, right? 230 00:10:51.725 --> 00:10:56.265 And what's exciting in the, in the cycles of a lot of hype 231 00:10:57.125 --> 00:11:00.265 things should get a lot of, uh, focus, right? 232 00:11:00.605 --> 00:11:03.105 So you see applied AI huge, 233 00:11:04.125 --> 00:11:06.745 but essentially losing a little bit of, 234 00:11:06.745 --> 00:11:08.025 even like job postings, 235 00:11:08.205 --> 00:11:11.345 while GenAI you start to get more job postings, right? 236 00:11:11.405 --> 00:11:12.665 So it's a cycle 237 00:11:12.695 --> 00:11:16.105 that we have gone over many times in technology when 238 00:11:16.105 --> 00:11:19.465 sometimes, like the newest things gets a lot of coverage, 239 00:11:19.525 --> 00:11:21.425 and the things that have been working quite well 240 00:11:21.425 --> 00:11:25.745 before eventually do not get that same level of, 241 00:11:25.845 --> 00:11:26.905 of press, right, 242 00:11:27.005 --> 00:11:28.905 of coverage that others have. 243 00:11:31.025 --> 00:11:32.845 Now going beyond investment, 244 00:11:33.065 --> 00:11:35.965 so there's a couple things that happen that kind of, uh, 245 00:11:37.035 --> 00:11:39.845 kind of exciting and just again, as a, as a level, you know, 246 00:11:39.845 --> 00:11:43.045 setting here, I mean, some of you already aware of that, 247 00:11:43.625 --> 00:11:47.045 but want to make sure that we have a good understanding 248 00:11:47.575 --> 00:11:49.245 about what's the current state, right? 249 00:11:49.345 --> 00:11:52.485 And one of the big things that you have right now is data, 250 00:11:53.345 --> 00:11:56.725 is essentially, you know, really looking at the vast amount 251 00:11:56.725 --> 00:12:00.765 of data that we have available, uh, in the world, right? 252 00:12:00.885 --> 00:12:03.005 I mean, those, those, those systems, those models 253 00:12:03.515 --> 00:12:06.325 have pretty much consumed a lot of it, right? 254 00:12:06.625 --> 00:12:10.885 So we're not really generating, um, extreme amount 255 00:12:10.885 --> 00:12:12.645 of new data versus 256 00:12:12.955 --> 00:12:16.205 what is the systems, the AI systems, are consuming, right? 257 00:12:16.825 --> 00:12:21.085 Um, so the publicly available data is start to get to, 258 00:12:21.185 --> 00:12:24.285 to tap out in terms of that availability. 259 00:12:25.025 --> 00:12:26.805 Um, we start to explore more 260 00:12:26.805 --> 00:12:29.765 and more, uh, creating data, synthetic data, 261 00:12:29.945 --> 00:12:32.605 and that is another word, synthetic data essentially is 262 00:12:33.225 --> 00:12:37.645 systems generating data to eventually train systems, right? 263 00:12:38.145 --> 00:12:40.965 And as you can imagine, that is, there is a lot 264 00:12:40.965 --> 00:12:42.165 of opinions about that. 265 00:12:42.275 --> 00:12:44.645 They have to be extremely careful about how 266 00:12:44.665 --> 00:12:46.045 to use synthetic data. 267 00:12:46.615 --> 00:12:49.485 There is some really good positive outcomes. 268 00:12:49.485 --> 00:12:52.165 There's a lot of, uh, risks if not going done well. 269 00:12:52.865 --> 00:12:55.005 So, but that's where you are in the data, right? 270 00:12:55.185 --> 00:12:59.885 The other aspect you have is, uh, kind of using a lot of, 271 00:12:59.985 --> 00:13:03.405 um, uh, proprietary, uh, data, right, 272 00:13:03.645 --> 00:13:06.765 internal data to augment the capabilities 273 00:13:07.505 --> 00:13:09.125 of our, the public models. 274 00:13:09.185 --> 00:13:11.845 So instead of, you know, using public models at the time, 275 00:13:12.505 --> 00:13:15.085 you are combine those public models 276 00:13:15.085 --> 00:13:18.525 with things like your RAG right, that are kind of associated 277 00:13:18.525 --> 00:13:20.005 with the, with the big models, 278 00:13:20.225 --> 00:13:22.725 but to be very customized, to be much more narrow 279 00:13:23.305 --> 00:13:26.845 and in many cases, much more helpful for organizations. 280 00:13:27.065 --> 00:13:30.885 So I was attending, you know, a presentation from the, um, 281 00:13:31.065 --> 00:13:34.645 uh, university of Pittsburgh, um, AI, dean 282 00:13:34.665 --> 00:13:37.445 or vice dean there, and, uh, the medical school, 283 00:13:37.545 --> 00:13:40.365 And they're using a lot of, trying to start using synthetic data 284 00:13:40.425 --> 00:13:43.925 to help diagnose, um, new, you know, diseases, right? 285 00:13:43.945 --> 00:13:47.245 And train doctors about how to, to, how to diagnose the, 286 00:13:47.245 --> 00:13:48.845 uh, diseases in, in, in, you know, 287 00:13:48.845 --> 00:13:50.125 in the medical school there. 288 00:13:50.825 --> 00:13:52.565 So, very, a lot of potential there. 289 00:13:53.825 --> 00:13:57.805 The other interesting thing is that it, it is really hard 290 00:13:57.805 --> 00:14:02.645 to assess at times how good large language models, right, 291 00:14:02.725 --> 00:14:05.285 these new models are currently are in, in 292 00:14:05.305 --> 00:14:06.525 how far they can go. 293 00:14:07.305 --> 00:14:11.005 So we are, uh, you know, the press, right? 294 00:14:11.035 --> 00:14:12.605 This is a very timely event 295 00:14:12.605 --> 00:14:15.405 because the press is indicating that, uh, 296 00:14:15.405 --> 00:14:19.165 ChatGPT five might be released in next week, right? 297 00:14:19.825 --> 00:14:22.405 Um, so in a sense, it looks like, oh, 298 00:14:22.625 --> 00:14:25.045 has been quite some time since we had a new release, 299 00:14:25.345 --> 00:14:29.405 but we have always a lot of evolution in ChatGPT, 300 00:14:29.405 --> 00:14:32.045 and of course, all the different competitors like Gemini 301 00:14:32.185 --> 00:14:33.445 and Llama and 302 00:14:33.745 --> 00:14:36.245 and Claude, they all have been evolving, 303 00:14:36.515 --> 00:14:38.885 even though they might not have like a definitive, 304 00:14:38.885 --> 00:14:40.605 like a new version, right, 305 00:14:40.745 --> 00:14:42.445 an X point something. 306 00:14:43.025 --> 00:14:46.525 Um, so there always these new models, uh, 307 00:14:46.845 --> 00:14:49.605 becoming available and they tend to be very powerful. 308 00:14:50.105 --> 00:14:53.965 So one of the things that, um, that we are seeing is 309 00:14:53.965 --> 00:14:55.765 that now some regular tests, right, 310 00:14:55.885 --> 00:14:57.005 I dunno if you remember this, 311 00:14:57.005 --> 00:14:59.885 but two years ago, like by November 2022, 312 00:14:59.915 --> 00:15:03.485 when ChatGPT really became public, um, we saw a lot 313 00:15:03.485 --> 00:15:05.725 of discussion about, hey, uh, you know, um, 314 00:15:07.025 --> 00:15:09.245 AI can pass the bar exam, right? 315 00:15:09.425 --> 00:15:12.605 Or AI can run, you know, an essay and, and, 316 00:15:12.745 --> 00:15:15.965 and do something, uh, about like college degrees, right, 317 00:15:15.965 --> 00:15:19.485 college credits. Well, for now, having AI 318 00:15:20.265 --> 00:15:23.565 really sometimes doing qualifying exams 319 00:15:23.565 --> 00:15:25.525 for PhDs successfully, right? 320 00:15:26.105 --> 00:15:28.805 Um, we are seeing, for example, a couple 321 00:15:28.805 --> 00:15:30.165 of weeks ago, right, 322 00:15:30.225 --> 00:15:34.445 uh, Gemini got, uh, the gold medal in, um, 323 00:15:35.545 --> 00:15:38.085 in the, in the, in the international math Olympics, 324 00:15:38.665 --> 00:15:42.085 um, and had to use a lot of creativity, a lot 325 00:15:42.085 --> 00:15:44.285 of creating new solutions for problems. 326 00:15:44.745 --> 00:15:47.565 And that's a very, very tough questions, right? 327 00:15:47.685 --> 00:15:49.205 I mean, only a small percentage 328 00:15:49.205 --> 00:15:52.445 of people can actually do those things, certainly not me. 329 00:15:52.825 --> 00:15:54.165 Um, and, uh, 330 00:15:54.185 --> 00:15:56.485 and essentially it is even to understand 331 00:15:56.705 --> 00:15:58.445 how far these models have come, 332 00:15:58.475 --> 00:16:00.125 have become a little bit complicated 333 00:16:00.125 --> 00:16:01.685 because it's hard to test, right? 334 00:16:01.685 --> 00:16:04.565 It's hard to test for intelligence, it's hard to test 335 00:16:04.685 --> 00:16:06.085 for capabilities, um, 336 00:16:06.115 --> 00:16:08.565 because again, they keep evolving, okay? 337 00:16:09.425 --> 00:16:10.885 Now, one of the biggest things 338 00:16:10.885 --> 00:16:13.605 that you have right now is this idea about 339 00:16:14.145 --> 00:16:16.245 AI agents, right? 340 00:16:16.385 --> 00:16:21.125 Uh, and AI agents are something relatively new, 341 00:16:21.225 --> 00:16:24.685 and the big kind of transformation is that 342 00:16:25.305 --> 00:16:28.685 we have started training those models, right, 343 00:16:28.685 --> 00:16:30.205 those, those large language models, 344 00:16:30.785 --> 00:16:32.445 in a different approach, right? 345 00:16:32.465 --> 00:16:36.045 So instead of doing more of the imitation learning, we start 346 00:16:36.045 --> 00:16:37.805 to use more and more post-training 347 00:16:37.865 --> 00:16:39.285 and reinforcement learning. 348 00:16:39.345 --> 00:16:42.605 And that's something that really changed quite a bit, 349 00:16:43.225 --> 00:16:45.165 the capabilities of those models, right? 350 00:16:45.195 --> 00:16:47.765 Because in many cases, in previous attempts, 351 00:16:47.765 --> 00:16:50.605 like in the early history, like by early history, I mean, 352 00:16:50.605 --> 00:16:52.285 like one year and a half ago or so, right, 353 00:16:52.865 --> 00:16:55.765 you saw essentially models created making mistakes 354 00:16:55.865 --> 00:16:59.045 and compounding those mistakes over time, right? 355 00:16:59.505 --> 00:17:02.205 So, um, you know, kind of you're running, you know, 356 00:17:02.225 --> 00:17:05.085 if you are, if you're driving something right in a race, 357 00:17:05.585 --> 00:17:08.365 and essentially you're moving things around, so instead 358 00:17:08.365 --> 00:17:11.965 of trying to adjust back, you know, making a mistake, you go 359 00:17:11.965 --> 00:17:13.285 to the side of the road, right? 360 00:17:13.545 --> 00:17:15.525 It would let more mistakes and more mistakes, 361 00:17:15.525 --> 00:17:18.605 and a sudden the car would just fall from the cliff, right? 362 00:17:19.915 --> 00:17:23.605 That has changed because you are training models, uh, 363 00:17:24.065 --> 00:17:25.925 in a different way now, right? 364 00:17:25.945 --> 00:17:27.885 If a lot of post-training, if a lot of, uh, 365 00:17:28.205 --> 00:17:30.845 reinforcement learning, and you're using actually, uh, 366 00:17:31.225 --> 00:17:33.605 models to validate models, right? 367 00:17:33.605 --> 00:17:35.445 To see if the accuracy of the, 368 00:17:35.675 --> 00:17:39.005 what the models generating is actually good enough to, 369 00:17:39.025 --> 00:17:40.165 to pass master, right? 370 00:17:40.745 --> 00:17:42.925 So the consequences of that is 371 00:17:42.925 --> 00:17:45.405 that now we have essentially some solutions 372 00:17:45.955 --> 00:17:49.005 that can take more complex instructions, right? 373 00:17:49.110 --> 00:17:52.125 In the past, you, the more complex, there is a certain point 374 00:17:52.125 --> 00:17:55.405 where the more complex you became in terms of your prompts, 375 00:17:55.405 --> 00:17:59.125 in terms of your, your requests, the more likely the model 376 00:17:59.645 --> 00:18:02.165 navigate out of the acceptable range. 377 00:18:02.375 --> 00:18:05.325 Right? Now, our seeing models that are, that are, 378 00:18:05.785 --> 00:18:07.845 can keep track, right? 379 00:18:07.955 --> 00:18:12.735 They require more, um, more ideas, um, um, 380 00:18:13.075 --> 00:18:16.375 you know, to get, to get, uh, you know, some better 381 00:18:17.045 --> 00:18:20.895 solutions, better, uh, outcomes in that, okay? 382 00:18:21.315 --> 00:18:22.815 And the thing is that because of that, 383 00:18:22.815 --> 00:18:26.215 there's a potential opening, you know, in terms of a lot 384 00:18:26.215 --> 00:18:29.375 of social companies are investing on that, a lot 385 00:18:29.375 --> 00:18:32.895 of app applications with ServiceNow, Salesforce, 386 00:18:33.395 --> 00:18:38.215 and really going to this TKI approach, which is, has a lot 387 00:18:38.215 --> 00:18:39.535 of potential to execute. 388 00:18:40.435 --> 00:18:42.655 And of course, the final thing that I want 389 00:18:42.655 --> 00:18:46.775 to talk about in terms of where we are right now is we start 390 00:18:46.775 --> 00:18:49.695 to see a variety of different developments, right? 391 00:18:49.755 --> 00:18:51.695 We have seen deep seek, right? 392 00:18:51.695 --> 00:18:53.895 Deep seek seems to be around for quite some time, 393 00:18:53.955 --> 00:18:56.455 but was actually, you know, couple of months ago, right? 394 00:18:56.915 --> 00:19:00.175 Uh, Quan and other models that again, are looking for these 395 00:19:00.845 --> 00:19:04.175 different approaches, different ways of training 396 00:19:04.325 --> 00:19:08.095 that might be more cost effective in generate results 397 00:19:08.095 --> 00:19:09.935 that are very, very positive, right? 398 00:19:10.435 --> 00:19:14.295 Um, so, so there's a lot of things happening in the realm 399 00:19:14.295 --> 00:19:17.175 of algorithms, the realm of development of new techniques. 400 00:19:17.715 --> 00:19:20.295 Um, it's a very vibrant area right now. 401 00:19:20.635 --> 00:19:24.135 And, uh, and a lot of this potential, um, in these areas 402 00:19:24.805 --> 00:19:27.135 that can lead to better outcomes, right? 403 00:19:27.195 --> 00:19:29.535 And, you know, I, and even more, right? 404 00:19:29.935 --> 00:19:32.015 I was in Korea a couple of months ago talking 405 00:19:32.015 --> 00:19:33.335 to some AI companies there, 406 00:19:33.835 --> 00:19:36.455 and they're trying to say, Hey, how can you actually really, 407 00:19:36.545 --> 00:19:40.255 truly, uh, implement reasoning, right? 408 00:19:40.355 --> 00:19:44.695 Versus this imitation of reason that we have right now by 409 00:19:45.255 --> 00:19:48.095 creating, you know, uh, induction deduction abductions 410 00:19:48.345 --> 00:19:49.855 mechanisms in these models, right? 411 00:19:49.855 --> 00:19:51.855 So there's a lot of things happening right now, 412 00:19:52.315 --> 00:19:55.375 and it's a lot of exciting, uh, new developments in general. 413 00:19:55.805 --> 00:19:58.255 Okay? So I know the questions are coming, 414 00:19:58.355 --> 00:20:00.375 and please keep them coming, you know, try to address 415 00:20:00.435 --> 00:20:03.815 as much as possible, um, um, for that. 416 00:20:04.755 --> 00:20:08.495 Yes. Uh, Caitlin, thank you so much. I appreciate that. 417 00:20:09.015 --> 00:20:11.455 I will get, uh, to some of the questions in the end. 418 00:20:11.655 --> 00:20:14.015 I appreciate that. So let's talk about the risks. 419 00:20:14.155 --> 00:20:17.175 And actually, Lin, I might, Kathleen, I might actually get, 420 00:20:17.275 --> 00:20:19.135 uh, to one of the questions right now. 421 00:20:20.575 --> 00:20:25.155 Um, so let us, uh, 422 00:20:25.385 --> 00:20:28.995 talk a about, a little bit in terms of the ethics 423 00:20:29.615 --> 00:20:30.795 and regulations, right? 424 00:20:31.495 --> 00:20:35.115 And, um, the main idea here is, um, 425 00:20:36.535 --> 00:20:38.585 what is ethical and fair, right? 426 00:20:38.605 --> 00:20:41.705 In the world of ai? So, uh, just popping up in the, 427 00:20:41.705 --> 00:20:42.825 in the q and a, right? 428 00:20:42.825 --> 00:20:47.285 To see how do you actually ethically source data, right? 429 00:20:47.865 --> 00:20:52.285 Um, well, it's, if you look at all the, the, uh, 430 00:20:52.285 --> 00:20:55.645 the progression of sourcing data, you can see 431 00:20:55.645 --> 00:20:58.325 that there's a lot of statutes have a lot of agreements 432 00:20:59.115 --> 00:21:02.245 between, um, you know, the big tech companies 433 00:21:02.505 --> 00:21:04.165 and data providers 434 00:21:04.385 --> 00:21:06.325 and new sources, et cetera, 435 00:21:06.385 --> 00:21:07.885 to make sure that the data is used. 436 00:21:07.905 --> 00:21:09.805 But there is still a lot of a gray area 437 00:21:10.375 --> 00:21:13.885 where data is not necessarily, you know, being collected 438 00:21:13.885 --> 00:21:17.525 with the permission of those, uh, creators. 439 00:21:17.985 --> 00:21:19.885 And there's a huge discussion about that. 440 00:21:19.885 --> 00:21:23.485 There's a lot of legal, uh, issues around that right now. 441 00:21:24.065 --> 00:21:28.485 But even beyond that, right? What is ethical, right? 442 00:21:28.555 --> 00:21:31.925 What I mean, if you ethics is of course is a very, um, 443 00:21:32.555 --> 00:21:34.325 complex problem, right? 444 00:21:34.385 --> 00:21:36.885 If you ask, you know, 100 of us here 445 00:21:37.145 --> 00:21:40.605 and try to understand, you know, how many people we are 446 00:21:41.035 --> 00:21:45.765 that are, you know, uh, different approaches of ethics, um, 447 00:21:45.945 --> 00:21:49.845 we might have 250 different answers, right? 448 00:21:49.845 --> 00:21:53.565 Because are we talking about fairness in the processes, 449 00:21:54.125 --> 00:21:55.125 fairness in the outcomes? 450 00:21:55.785 --> 00:21:59.325 So expecting that reflection of ethics 451 00:21:59.905 --> 00:22:01.765 in systems, right? 452 00:22:01.785 --> 00:22:03.725 It is complex by its own nature. 453 00:22:04.345 --> 00:22:05.405 Uh, what you start 454 00:22:05.405 --> 00:22:07.645 to see right now is a little bit more regulations. 455 00:22:07.785 --> 00:22:09.645 The AI has more regulations. 456 00:22:09.645 --> 00:22:11.925 Google just, uh, had an agreement with ai, 457 00:22:11.955 --> 00:22:16.845 with the European Union to, um, to try to, you know, uh, 458 00:22:17.125 --> 00:22:18.805 navigate their regulations now. 459 00:22:19.385 --> 00:22:24.365 Um, but even then, how enforceable it is, right? 460 00:22:24.665 --> 00:22:28.085 How complex that, um, that might be 461 00:22:28.085 --> 00:22:29.925 to actually execute those regulations. 462 00:22:30.305 --> 00:22:34.765 So this still very, very much, um, um, undecided about 463 00:22:34.785 --> 00:22:36.565 how regulations are going to evolve. 464 00:22:37.145 --> 00:22:40.965 And if we're creating systems to serve customers, right? 465 00:22:41.075 --> 00:22:42.765 That becomes a problem, of course, 466 00:22:42.765 --> 00:22:46.085 because what's fair, what's regulated, what's legal, 467 00:22:47.675 --> 00:22:49.165 then you have biases, right? 468 00:22:49.225 --> 00:22:51.685 So nothing that you have are better than, 469 00:22:51.685 --> 00:22:54.085 than you should be, but biases are 470 00:22:55.055 --> 00:22:56.845 still very present in terms of 471 00:22:56.985 --> 00:23:01.485 how exactly is our data source, how exactly, um, you know, 472 00:23:01.945 --> 00:23:04.845 are we, uh, eliminating some biases 473 00:23:05.035 --> 00:23:07.965 that are in the data sets, biases in the selection 474 00:23:07.965 --> 00:23:12.325 of models, biases in the outcomes, um, that is still a lot 475 00:23:12.345 --> 00:23:13.445 to be discovered 476 00:23:13.445 --> 00:23:16.645 and discussed there, where much better dealing 477 00:23:16.645 --> 00:23:18.405 with biases than before, right? 478 00:23:18.545 --> 00:23:21.685 We are much more attentive to what the potential biases are, 479 00:23:22.185 --> 00:23:24.885 but we have to be very vigilant about 480 00:23:24.885 --> 00:23:26.125 that, uh, all the time. 481 00:23:27.485 --> 00:23:32.265 The other issue that we have seen is, you know, the lack 482 00:23:32.265 --> 00:23:34.065 of explainability at times, right? 483 00:23:34.325 --> 00:23:36.545 And the fact that as humans, 484 00:23:37.435 --> 00:23:39.425 we're not built, right? 485 00:23:39.525 --> 00:23:44.025 Our minds are not built to keep essentially monitoring, uh, 486 00:23:44.285 --> 00:23:46.045 AI all the time, right? 487 00:23:46.465 --> 00:23:49.165 But just not, eventually when you start seeing things 488 00:23:49.235 --> 00:23:51.045 working permanently, right? 489 00:23:51.045 --> 00:23:54.845 In a regular basis, we, we just say web, it's working. 490 00:23:54.995 --> 00:23:56.205 Know that 100 times 491 00:23:56.305 --> 00:23:58.725 before, it certainly is going to work in the future, 492 00:23:59.425 --> 00:24:02.325 and we start to pay less attention to that. 493 00:24:02.905 --> 00:24:04.685 So it's a big risk 494 00:24:04.685 --> 00:24:08.765 because essentially we are not very prepared individual 495 00:24:08.855 --> 00:24:11.645 basis to keep monitoring these outcomes. 496 00:24:12.345 --> 00:24:14.685 And then it starts to run into who is liable, 497 00:24:15.105 --> 00:24:16.645 who is liable, right? 498 00:24:16.665 --> 00:24:18.085 For failure, for oversight. 499 00:24:18.625 --> 00:24:21.925 Um, and we is even, um, responsible 500 00:24:22.005 --> 00:24:23.805 for auditing the outcomes. 501 00:24:24.465 --> 00:24:27.285 So there is an interesting article in our technical talking 502 00:24:27.375 --> 00:24:28.605 about, you know, like 503 00:24:28.605 --> 00:24:31.725 how the legal system might be completely unprepared to deal 504 00:24:31.725 --> 00:24:34.725 with fake, fake, uh, citations, making references 505 00:24:35.265 --> 00:24:38.085 in case law, because you don't know. 506 00:24:38.165 --> 00:24:41.565 I mean, how, who is actually constantly monitoring, right? 507 00:24:41.785 --> 00:24:44.125 Uh, case law that might be made up 508 00:24:44.125 --> 00:24:46.245 because essentially a model came up 509 00:24:46.245 --> 00:24:49.005 with something new was too creative, okay? 510 00:24:49.665 --> 00:24:52.085 Or even things like, uh, you might remember the, 511 00:24:52.105 --> 00:24:55.005 the Mecca Hitler, uh, sometime ago with grok, right? 512 00:24:55.525 --> 00:24:57.085 I mean, that was very blatant. 513 00:24:57.475 --> 00:24:59.965 Everybody was very obviously wrong. 514 00:25:00.345 --> 00:25:02.965 Uh, thank you know, with just no discussion about that. 515 00:25:03.635 --> 00:25:06.445 What happened if that was a little bit more subtle, right? 516 00:25:06.625 --> 00:25:11.125 Or happen if the bias was present but was not so blatant. 517 00:25:11.585 --> 00:25:12.725 How do we monitor that? 518 00:25:12.825 --> 00:25:16.765 How do you make sure that, you know, it is as unbiased, 519 00:25:16.985 --> 00:25:20.605 as impartial as possible in those scenarios? 520 00:25:21.545 --> 00:25:24.005 And that leads us to another situation who is, 521 00:25:24.945 --> 00:25:27.365 how do we even know how those things work, right? 522 00:25:27.675 --> 00:25:30.445 They are not necessarily, so one of the things that we want 523 00:25:30.445 --> 00:25:32.005 to always remember is 524 00:25:32.005 --> 00:25:35.165 that explainability is still an an issue, right? 525 00:25:35.215 --> 00:25:37.085 We're getting better and understanding 526 00:25:37.225 --> 00:25:39.205 how the internal work is of these models, 527 00:25:39.785 --> 00:25:43.085 but there is still a lot of, uh, complexity. 528 00:25:43.345 --> 00:25:47.245 You are not regular coding that you're able to track down 529 00:25:47.825 --> 00:25:50.445 and see exactly why the mistakes took place, right? 530 00:25:50.785 --> 00:25:54.045 We can see changes, Hey, this change led to, you know, 531 00:25:54.045 --> 00:25:57.125 these really big bad outcome, maybe you can revert back, 532 00:25:57.705 --> 00:26:01.085 but it's very hard to pinpoint what exactly went wrong. 533 00:26:01.225 --> 00:26:04.565 You know? Uh, Tropic has done some really good work about 534 00:26:04.565 --> 00:26:07.485 that, and even them, uh, still kind of lacking 535 00:26:07.835 --> 00:26:10.485 that precise understanding about what's the train 536 00:26:10.485 --> 00:26:13.925 of thought in that logic that essentially we understand 537 00:26:13.945 --> 00:26:16.165 how exactly things can go wrong, right? 538 00:26:16.585 --> 00:26:18.325 So it's very hard to debug. 539 00:26:18.395 --> 00:26:21.485 It's very hard to figure out, know how we can fix mistakes. 540 00:26:21.625 --> 00:26:24.605 And because it's hard to anticipate that it's also hard 541 00:26:24.605 --> 00:26:25.805 to catch them, right? 542 00:26:25.835 --> 00:26:28.085 It's hard to figure out what is wrong. 543 00:26:28.185 --> 00:26:29.885 How did you get this particular point? 544 00:26:31.475 --> 00:26:35.615 So why do most projects fail, right? So those are the risks. 545 00:26:35.875 --> 00:26:38.255 Um, and now let's think about the risks in general. 546 00:26:38.395 --> 00:26:41.295 Now, let's look at the risks for organizations, right? 547 00:26:41.595 --> 00:26:44.395 So first of all, we have 548 00:26:44.395 --> 00:26:45.915 to remember there's a lot of hype, right? 549 00:26:45.945 --> 00:26:49.955 This is, uh, um, um, you know, my sequence of, uh, you know, 550 00:26:50.115 --> 00:26:53.595 McKinsey, uh, has this, um, AI, you know, this status 551 00:26:53.735 --> 00:26:54.755 of AI in the world, 552 00:26:55.295 --> 00:26:57.675 and you can see, you know, some cases, again, 553 00:26:57.745 --> 00:26:59.555 many friends at McKinsey, great company. 554 00:26:59.975 --> 00:27:04.195 Um, but you can see in like, hey, 23, right, 555 00:27:04.635 --> 00:27:08.795 breakout year, 24 start to generate value, 25 556 00:27:09.025 --> 00:27:10.555 well, what is everybody, right? 557 00:27:10.855 --> 00:27:12.195 So there's a lot of hype, 558 00:27:12.655 --> 00:27:15.035 and that's something that I want to make sure in this, 559 00:27:15.135 --> 00:27:17.795 in this know, this conversation is that we try 560 00:27:17.795 --> 00:27:20.515 to minimize this, this hype to try 561 00:27:20.535 --> 00:27:23.555 to really be very solid in terms of what is 562 00:27:24.355 --> 00:27:26.995 achievable versus just, you know, trying 563 00:27:27.015 --> 00:27:29.115 to follow the latest trend. 564 00:27:30.135 --> 00:27:34.875 Um, so Infosys is an interesting study, uh, this year. 565 00:27:34.975 --> 00:27:37.195 And essentially they looked at interview hundreds 566 00:27:37.195 --> 00:27:40.555 of companies and interview a bunch of different companies 567 00:27:40.555 --> 00:27:44.355 as well to figure out what are the kind of five factors 568 00:27:45.425 --> 00:27:49.195 that are associated with success, right, 569 00:27:49.415 --> 00:27:51.155 uh, the readiness for AI. 570 00:27:51.615 --> 00:27:54.195 And they came up essentially with strategy, governance, 571 00:27:54.255 --> 00:27:56.235 talent, data, and technology, 572 00:27:56.455 --> 00:27:58.795 and only a tiny portion of companies 573 00:27:59.335 --> 00:28:01.875 are preparing all dimensions right? 574 00:28:02.495 --> 00:28:07.355 Now, the interesting fact is most AI projects fail. 575 00:28:08.215 --> 00:28:10.675 So, you know, we can explain that, right? 576 00:28:10.675 --> 00:28:12.635 To see how many of those, of those 577 00:28:12.735 --> 00:28:16.115 of those factors we are not really, had huge 578 00:28:16.195 --> 00:28:17.275 dominance among them. 579 00:28:17.975 --> 00:28:20.635 And, uh, the consequences that, um, you know, 580 00:28:20.635 --> 00:28:23.435 some estimates say that 80% to 90% 581 00:28:23.435 --> 00:28:26.955 of AI projects never leave the proof of concept, right? 582 00:28:27.785 --> 00:28:31.715 Gartner, um, uh, just did a study a couple of weeks ago, uh, 583 00:28:31.715 --> 00:28:34.555 saying that about 40% of, uh, 584 00:28:34.645 --> 00:28:37.875 agentic AI projects are going nowhere next year, right? 585 00:28:37.875 --> 00:28:42.195 They're going to be canceled. Um, only about 19% of projects 586 00:28:42.415 --> 00:28:45.875 beyond the proof of concept actually achieve expected value. 587 00:28:46.695 --> 00:28:50.995 So if you are in that AI, you know, let's go AI, right, 588 00:28:51.465 --> 00:28:53.955 it's very sobering when you see these numbers 589 00:28:54.145 --> 00:28:57.235 because the reality is that we are far 590 00:28:57.765 --> 00:29:02.155 right from have established practices that to make AI 591 00:29:03.175 --> 00:29:04.315 viable, right, 592 00:29:04.415 --> 00:29:06.875 and valuable for many organizations. 593 00:29:07.495 --> 00:29:08.955 So let's go over a little bit 594 00:29:08.955 --> 00:29:10.315 of them very quickly, you know, right? 595 00:29:10.555 --> 00:29:12.155 I mean, one of the things we have to realize 596 00:29:12.305 --> 00:29:14.795 that most AI projects already start 597 00:29:14.795 --> 00:29:16.195 with a losing proposition 598 00:29:16.505 --> 00:29:19.235 because the organization does not understand 599 00:29:19.305 --> 00:29:20.795 what the technology can do. 600 00:29:21.125 --> 00:29:22.675 There is no AI strategy 601 00:29:22.815 --> 00:29:24.835 and there are no business cases and metrics. 602 00:29:25.735 --> 00:29:29.755 So without that, any IT project, right, require those, 603 00:29:30.345 --> 00:29:33.395 without having that, why exactly are we doing this right? 604 00:29:33.585 --> 00:29:36.475 What is a strategy around that? What, what should we do? 605 00:29:37.145 --> 00:29:39.115 That already puts you in a disadvantage 606 00:29:39.345 --> 00:29:42.805 because what you are trying to achieve, the unachievable, 607 00:29:42.805 --> 00:29:44.405 because there are no targets to fall. 608 00:29:45.305 --> 00:29:47.765 Second part is governance. 609 00:29:48.345 --> 00:29:49.965 People don't understand the risks. 610 00:29:50.395 --> 00:29:52.965 What are the risks associated with AI? 611 00:29:52.965 --> 00:29:55.685 There is no understanding about that. 612 00:29:55.985 --> 00:29:59.665 Um, um, you know, that risks for personal risk, 613 00:29:59.985 --> 00:30:01.745 physical safety, financial performance, 614 00:30:01.895 --> 00:30:04.985 even national security can be affected, right? 615 00:30:05.645 --> 00:30:07.465 So what are we, are the risks, 616 00:30:07.685 --> 00:30:10.465 and if we don't have the proper governance, you know, 617 00:30:10.705 --> 00:30:13.305 structure, those risks are going to propagate 618 00:30:13.305 --> 00:30:15.465 and eventually get you later when you're not really 619 00:30:15.495 --> 00:30:16.705 expecting them anymore. 620 00:30:18.005 --> 00:30:21.225 Now we start to see more talent emerge. 621 00:30:21.645 --> 00:30:23.825 Um, so we start, 622 00:30:23.885 --> 00:30:27.065 but still we have this discussion about way we have 623 00:30:27.065 --> 00:30:29.305 to teach people more AI skills, right? 624 00:30:29.365 --> 00:30:31.625 And like, what, what, what are AI skills, right? 625 00:30:31.725 --> 00:30:32.825 Is it prompt engineering? 626 00:30:32.885 --> 00:30:35.905 Is it, you know, is it understanding the models, right? 627 00:30:36.805 --> 00:30:39.505 And, and what I have seen, I was talking with, uh, 628 00:30:39.505 --> 00:30:41.545 with an executive, um, in, um, 629 00:30:41.885 --> 00:30:45.105 in a large chemistry organization in the Europe, uh, 630 00:30:45.215 --> 00:30:47.025 last week, and they say, well, 631 00:30:47.035 --> 00:30:48.905 we're actually not really hiring many 632 00:30:48.905 --> 00:30:50.065 data scientists anymore. 633 00:30:50.485 --> 00:30:51.625 We need to hire more people who 634 00:30:51.625 --> 00:30:52.865 understand governance, right? 635 00:30:53.165 --> 00:30:56.465 So this patterns of hiring, the patterns of need 636 00:30:56.485 --> 00:30:59.145 for organizations of AI skills, 637 00:30:59.445 --> 00:31:02.585 AI talent, are really changing very dramatically. 638 00:31:02.685 --> 00:31:05.785 And of course, it changes with industries, uh, 639 00:31:05.965 --> 00:31:08.785 but it's also changing in terms of age, in terms 640 00:31:08.785 --> 00:31:09.785 of experimentation. 641 00:31:10.405 --> 00:31:14.145 So, um, um, um, the, um, Associate Press 642 00:31:14.725 --> 00:31:16.065 did a research, uh, 643 00:31:16.365 --> 00:31:20.625 the survey essentially recognized about only 37% of people, 644 00:31:21.365 --> 00:31:24.825 um, have actually used AI for work, right? 645 00:31:25.205 --> 00:31:27.985 We, we see all the hype and see, oh, everybody's using AI. 646 00:31:28.035 --> 00:31:30.065 Maybe they're all using in, in college, 647 00:31:30.165 --> 00:31:31.625 but not work as much. 648 00:31:32.305 --> 00:31:34.585 I was working with a company last week 649 00:31:34.645 --> 00:31:37.465 and they said about the same statistics, about only 40% 650 00:31:37.465 --> 00:31:40.225 of the employees are actually using AI in work. 651 00:31:40.805 --> 00:31:44.185 So a lot of companies who struggle in terms of what is 652 00:31:44.375 --> 00:31:45.825 that, what are those skills? 653 00:31:45.965 --> 00:31:47.625 How do we train people, right? 654 00:31:47.625 --> 00:31:50.825 Should they review skills are necessary to use AI? 655 00:31:51.725 --> 00:31:53.825 And from the employee's perspective, many 656 00:31:53.825 --> 00:31:56.345 of them are waiting for guidance, right? 657 00:31:56.495 --> 00:31:57.825 There's a, a couple of, uh, 658 00:31:58.005 --> 00:32:00.945 new, our new surveys essentially saying, hey, I'm kind 659 00:32:00.945 --> 00:32:03.345 of willing to do it, but I dunno what my limits are. 660 00:32:03.585 --> 00:32:04.865 I dunno what the rewards are. 661 00:32:05.345 --> 00:32:07.425 I dunno what you know, I should be doing, right? 662 00:32:07.605 --> 00:32:10.905 And so leadership in organizations need to step up 663 00:32:11.365 --> 00:32:13.505 and kind of set up what their guard rails are 664 00:32:13.565 --> 00:32:16.945 and what objectives, what the strategy around AI 665 00:32:17.165 --> 00:32:18.345 to be, uh, successful. 666 00:32:18.345 --> 00:32:23.065 Data is the, one 667 00:32:23.065 --> 00:32:24.585 of the biggest things about the, 668 00:32:24.925 --> 00:32:27.785 the challenges in any AI project, right? 669 00:32:28.535 --> 00:32:32.505 What, uh, um, uh, [unintelligible], who is the guys that created the, 670 00:32:32.605 --> 00:32:35.705 uh, uh, Netflix algorithm, the recommendation algorithm, 671 00:32:36.265 --> 00:32:37.905 I was talking with him, um, a couple 672 00:32:37.905 --> 00:32:39.445 of us some last year actually. 673 00:32:39.785 --> 00:32:42.565 And it's like, hey, [unintelligible], he's a consultant 674 00:32:42.705 --> 00:32:45.805 now, he goes to companies and we can't define the problem. 675 00:32:46.705 --> 00:32:48.845 We can't define which model we should use, 676 00:32:49.545 --> 00:32:52.285 but then when you try to actually use the data to solve 677 00:32:52.315 --> 00:32:53.605 that problem, right, 678 00:32:53.715 --> 00:32:55.885 with that model, we don't have data. 679 00:32:56.705 --> 00:32:59.365 And that's something that organizations have really 680 00:32:59.565 --> 00:33:01.165 sidestep, uh, so far. 681 00:33:01.745 --> 00:33:04.365 And they're going to come to buy them 682 00:33:04.365 --> 00:33:06.645 because, you know, you don't have the right data, 683 00:33:06.645 --> 00:33:08.165 don't have the right data governance, 684 00:33:08.165 --> 00:33:09.845 you don't have the right data structure. 685 00:33:10.675 --> 00:33:13.125 That means that, you know, all of a sudden your, 686 00:33:13.125 --> 00:33:14.165 your models don't work, 687 00:33:15.065 --> 00:33:18.285 and all that investment gets wasted, right? 688 00:33:18.425 --> 00:33:20.445 So you, some cases you have to really figure out 689 00:33:20.445 --> 00:33:25.325 what my data status is before even trying to go into AI 690 00:33:25.435 --> 00:33:30.205 because you will not get the results that you're expecting 691 00:33:30.305 --> 00:33:32.765 unless you really truly understand the value 692 00:33:32.785 --> 00:33:34.725 of data in your organization. 693 00:33:36.475 --> 00:33:40.735 Now, um, the interesting thing is that what is emerging, uh, 694 00:33:40.935 --> 00:33:44.215 a lot right now is a little bit more those resources about 695 00:33:44.355 --> 00:33:45.535 the technology, right, 696 00:33:45.555 --> 00:33:47.855 the models. I, I, I got a, I, uh, 697 00:33:48.015 --> 00:33:50.935 I was giggling away if I saw this, uh, this, um, 698 00:33:52.075 --> 00:33:56.135 common reference architecture for, uh, GenAI at scale, 699 00:33:56.285 --> 00:33:58.335 that is a, a publication by McKinsey. 700 00:33:58.445 --> 00:34:03.175 Like, this seems extremely complex, uh, architecture for me, 701 00:34:03.635 --> 00:34:06.135 and believe me, I have been in technology for decades. 702 00:34:06.915 --> 00:34:09.775 Um, so there's nothing necessarily simple, right? 703 00:34:10.315 --> 00:34:13.095 Um, I like the attempt here of, of trying 704 00:34:13.095 --> 00:34:14.735 to create some overall model. 705 00:34:15.435 --> 00:34:19.135 But what we have seen is that from the more applied AI, 706 00:34:19.195 --> 00:34:21.335 the maybe the non-sexy AI, right? 707 00:34:21.715 --> 00:34:24.535 We start to have lots and lots of good models 708 00:34:24.715 --> 00:34:27.855 and good solutions that are within the market. We start 709 00:34:27.875 --> 00:34:32.135 to see more of the merging, uh, of the GenAI in terms 710 00:34:32.155 --> 00:34:34.415 of models, in terms of marketing availability 711 00:34:34.475 --> 00:34:36.575 of those models, such get better and better. 712 00:34:37.675 --> 00:34:40.295 The big issue is that, you know, for organizations now, 713 00:34:40.295 --> 00:34:43.135 they have to orchestrate a lot of those things, right? 714 00:34:43.715 --> 00:34:46.575 And, um, there's a lot of variations in options 715 00:34:46.995 --> 00:34:48.175 and contracts. 716 00:34:48.355 --> 00:34:52.775 Um, tokens. How many tokens are you going to need? 717 00:34:52.775 --> 00:34:55.375 Well, it's hard, right? To kind of anticipate those. 718 00:34:55.955 --> 00:34:59.135 So what the main organization is struggling right now is to 719 00:34:59.375 --> 00:35:00.375 actually establish, right? 720 00:35:00.475 --> 00:35:03.255 How, what's my relationship with vendors, with providers? 721 00:35:03.595 --> 00:35:07.215 How much do I pay for that? What is worth, right, 722 00:35:07.315 --> 00:35:11.375 in terms of the access to these tools. You can pay $250 723 00:35:11.515 --> 00:35:14.055 for individual license in some of the more refined models 724 00:35:14.435 --> 00:35:15.895 or pay nothing, right? 725 00:35:16.665 --> 00:35:18.895 Which, what's the range that you want? 726 00:35:18.895 --> 00:35:21.015 Which, which situations that you want 727 00:35:21.015 --> 00:35:22.975 to apply the super sophisticated model 728 00:35:23.615 --> 00:35:25.055 'cause that's the only one that's going to give 729 00:35:25.125 --> 00:35:28.335 that accuracy that you need versus the, 730 00:35:28.355 --> 00:35:29.535 the cheap ones, right? 731 00:35:29.875 --> 00:35:31.655 So that's the kind of things that, uh, 732 00:35:31.655 --> 00:35:33.415 your organization is struggling. 733 00:35:33.695 --> 00:35:36.295 I was talking again with a executive, uh, a couple 734 00:35:36.295 --> 00:35:38.365 of days ago and like, well, 735 00:35:39.525 --> 00:35:41.325 contracts have become a big issue, right? 736 00:35:41.395 --> 00:35:45.325 Because we don't quite know how much we, 737 00:35:45.465 --> 00:35:46.845 how much we assess, right? 738 00:35:47.315 --> 00:35:50.885 This investment, what's the ROI to establish relationships 739 00:35:50.885 --> 00:35:52.045 with vendors and 740 00:35:52.045 --> 00:35:53.045 how do you manage that 741 00:35:53.325 --> 00:35:57.445 ROI over time? Again, it, it's not unheard of, right? 742 00:35:57.825 --> 00:36:00.365 It is just that we are not quite there 743 00:36:00.365 --> 00:36:03.645 and having really super-established practices right now 744 00:36:04.075 --> 00:36:06.525 that we can rely on versus technologies 745 00:36:06.595 --> 00:36:07.845 that are more established. 746 00:36:08.165 --> 00:36:10.885 Although that is a marketplace for a lot of these solutions, 747 00:36:11.185 --> 00:36:12.365 uh, that's marketplaces 748 00:36:12.365 --> 00:36:13.645 you taking shape in terms 749 00:36:13.765 --> 00:36:16.805 of what's the best arrangement possible for a variety of 750 00:36:16.805 --> 00:36:17.925 different organizations. 751 00:36:19.925 --> 00:36:22.675 Now, let's go to the final part as you wrap 752 00:36:22.695 --> 00:36:23.955 and give more space, right? 753 00:36:24.095 --> 00:36:27.305 Um, how to succeed, okay. 754 00:36:27.325 --> 00:36:29.585 And, and I think, you know, to, again, um, giving credit 755 00:36:29.605 --> 00:36:32.865 to the service that McKinsey has and I made fun of some 756 00:36:32.865 --> 00:36:34.385 of them before in the hype cycle, 757 00:36:35.145 --> 00:36:36.665 I like the title of this one, right? 758 00:36:36.665 --> 00:36:39.545 Because essentially it's finally this realization that 759 00:36:40.165 --> 00:36:43.785 it is not only about the technology, right? 760 00:36:43.855 --> 00:36:47.785 Organizations have to change themselves if they want 761 00:36:47.785 --> 00:36:51.665 to bring some value, uh, from AI. Um, 762 00:36:52.085 --> 00:36:54.745 for many cases say, hey, let's just drop technology. 763 00:36:54.955 --> 00:36:57.905 Let's see what happens and pray for the best. Right? 764 00:36:58.445 --> 00:36:59.585 That's not possible. 765 00:36:59.655 --> 00:37:03.265 That does not work at all, um, in this scenario, 766 00:37:03.265 --> 00:37:05.505 because we're going to be facing disappointment 767 00:37:05.635 --> 00:37:08.465 after disappointment after disappointment on that one. 768 00:37:09.205 --> 00:37:13.815 So, um, um, one of the things that, uh, we, 769 00:37:13.915 --> 00:37:15.215 we, we see, right? 770 00:37:15.235 --> 00:37:18.575 Is that we need to really change the organization quite 771 00:37:18.695 --> 00:37:21.175 a lot more to actually get the value. 772 00:37:21.315 --> 00:37:22.895 So there's a couple of changes 773 00:37:22.895 --> 00:37:24.975 that might be more pertinent, right? 774 00:37:24.995 --> 00:37:28.255 One is really changing in the strategy, right. 775 00:37:28.635 --> 00:37:32.175 Um, I was, uh, giving a talk to some execs in, uh, 776 00:37:32.175 --> 00:37:33.295 executives in Indy 777 00:37:33.795 --> 00:37:37.895 and, uh, we're like, well, you know, talking about the fear 778 00:37:37.895 --> 00:37:39.135 of missing out, right? 779 00:37:39.315 --> 00:37:41.895 Uh, you know, everybody seems to be investing in AI 780 00:37:41.895 --> 00:37:45.415 because everybody else is doing that, so I need 781 00:37:45.415 --> 00:37:46.575 to do it as well, right? 782 00:37:47.525 --> 00:37:49.855 Well, FOMO is not a strategy, right? 783 00:37:50.475 --> 00:37:51.895 It is really, it really isn't. 784 00:37:52.055 --> 00:37:54.335 I mean, again, it doesn't really mean that you don't want 785 00:37:54.335 --> 00:37:56.335 to experiment with AI, right? 786 00:37:56.435 --> 00:37:58.975 You can't figure out what's the best approach is, 787 00:37:59.435 --> 00:38:02.495 but you still should try to formulate a strategy about 788 00:38:02.545 --> 00:38:03.575 where you are going. 789 00:38:04.435 --> 00:38:08.815 Um, otherwise a lot of that investment is going to be wasted 790 00:38:08.875 --> 00:38:11.375 and people is going to start a little bit more reticent 791 00:38:12.105 --> 00:38:13.655 about using AI in the future. 792 00:38:14.555 --> 00:38:17.615 The other thing that you have to think about is how 793 00:38:17.615 --> 00:38:19.975 to create this dual timeline, right? 794 00:38:20.375 --> 00:38:22.815 A timeline in which, you know, say, hey, 795 00:38:22.815 --> 00:38:25.935 those are my quick gains, my quick wins that I want 796 00:38:25.935 --> 00:38:27.255 to get right now, right? 797 00:38:27.995 --> 00:38:30.375 But what's my imagination, 798 00:38:31.215 --> 00:38:35.405 creative approach to AI that is going to say, hey, 799 00:38:35.545 --> 00:38:38.805 if you are successful in this, this is 800 00:38:38.905 --> 00:38:41.325 how our organization is going to change. 801 00:38:42.145 --> 00:38:45.165 And if you are successful in that, this is 802 00:38:45.165 --> 00:38:48.445 how organization is, go to iteration three, 803 00:38:48.865 --> 00:38:50.645 and this is how our organization going 804 00:38:50.645 --> 00:38:52.285 to iteration four, right? 805 00:38:52.745 --> 00:38:56.685 Is really looking at this realm of the possible, right, 806 00:38:56.705 --> 00:38:58.245 the art of the possible here, 807 00:38:58.515 --> 00:39:02.005 because it's just, you know, just looking at the short wins, 808 00:39:02.105 --> 00:39:04.285 the quick wins, it's not going to move 809 00:39:04.305 --> 00:39:06.245 you, move the needle quite a bit. 810 00:39:06.685 --> 00:39:08.885 I have classes where I have my students like, hey, 811 00:39:09.005 --> 00:39:10.725 I want you to pick one industry. 812 00:39:11.015 --> 00:39:13.605 Think about this industry 10 years from now, right? 813 00:39:14.625 --> 00:39:16.405 Try to come up with something that you see 814 00:39:16.795 --> 00:39:19.005 what happens in this particular industry in 10 years. 815 00:39:19.625 --> 00:39:22.725 Use what you know right now, expect the iteration 816 00:39:22.725 --> 00:39:24.645 that technology is going to come with, right? 817 00:39:24.645 --> 00:39:25.925 That's naturally going to occur. 818 00:39:26.305 --> 00:39:29.245 And how you prepare organization, not for right now, 819 00:39:29.385 --> 00:39:32.605 but for 10 years from now when the world might be changing. 820 00:39:32.785 --> 00:39:34.485 How do you not get left behind? 821 00:39:35.065 --> 00:39:36.805 That's what you need in terms of strategy 822 00:39:36.865 --> 00:39:38.005 for organizations, right? 823 00:39:38.025 --> 00:39:40.645 And of course, also developing business cases 824 00:39:40.745 --> 00:39:42.605 and metrics that can guide people. 825 00:39:43.345 --> 00:39:45.725 It cannot be investment just for the sake investment. 826 00:39:45.945 --> 00:39:50.005 It has to come up, come back with real, real value. 827 00:39:51.415 --> 00:39:54.315 We also need changes in the skill and organizing, right? 828 00:39:54.465 --> 00:39:57.275 What are the true AI skills that employees need? 829 00:39:58.265 --> 00:40:01.635 What can AI do? What can't AI do? 830 00:40:02.455 --> 00:40:03.835 How to set up in those skills? 831 00:40:03.985 --> 00:40:08.155 What are the processes that need to change, right? 832 00:40:08.535 --> 00:40:10.555 So you can train people, right? 833 00:40:10.615 --> 00:40:12.595 We can use, you know, more AI. 834 00:40:12.655 --> 00:40:14.355 You can be talented in using AI, 835 00:40:14.775 --> 00:40:17.075 but if the process in the organization don't change, 836 00:40:17.175 --> 00:40:19.555 if you don't think about the ways that you can perhaps 837 00:40:20.435 --> 00:40:22.875 compress time or perhaps improve accuracy 838 00:40:22.895 --> 00:40:26.395 or perhaps benefit the customer in a way 839 00:40:26.475 --> 00:40:27.955 that is coherent, right, 840 00:40:27.955 --> 00:40:30.955 you're going to always to start these isolated mechanisms 841 00:40:30.955 --> 00:40:32.555 that cannot scale very well. 842 00:40:32.935 --> 00:40:36.555 So thinking about how exactly those process need to change 843 00:40:36.975 --> 00:40:39.075 to reabsorb AI in organization, 844 00:40:39.095 --> 00:40:41.435 and many organizations fail to do that, right? 845 00:40:41.705 --> 00:40:42.955 Like the future of work, 846 00:40:43.815 --> 00:40:45.715 but the future of the nature of work. 847 00:40:45.775 --> 00:40:48.475 What's the nature of work when you start embedding AI, 848 00:40:49.055 --> 00:40:50.795 and a lot of it is Agile, right? 849 00:40:50.945 --> 00:40:53.635 Agility. How to actually keep iterating 850 00:40:53.705 --> 00:40:56.835 through these new approaches, not only the solutions 851 00:40:56.835 --> 00:40:59.275 with AI, but how the processes themselves 852 00:40:59.895 --> 00:41:01.955 can change over time in a way 853 00:41:01.955 --> 00:41:04.195 that can benefit all the stakeholders. 854 00:41:06.095 --> 00:41:08.505 Data management. Again, very briefly, I'm not going 855 00:41:08.505 --> 00:41:09.545 to repeat myself, right? 856 00:41:09.685 --> 00:41:11.745 How do you deal with data? You need 857 00:41:11.745 --> 00:41:14.865 to really focus on data from the very beginning. 858 00:41:15.335 --> 00:41:17.305 Many organizations struggle with AI 859 00:41:17.305 --> 00:41:20.345 because they don't have good data management structures. 860 00:41:20.345 --> 00:41:22.105 They don't have the good data governance. 861 00:41:22.735 --> 00:41:24.985 They try things, it doesn't work. 862 00:41:25.925 --> 00:41:27.145 And then they regret, 863 00:41:27.685 --> 00:41:29.745 and then they say, well, you know, AI is, 864 00:41:29.805 --> 00:41:30.865 is for nothing, right? 865 00:41:30.925 --> 00:41:33.765 Why I'm doing this? And that's not really, I mean, 866 00:41:33.765 --> 00:41:36.405 you essentially, you already start in a position 867 00:41:36.405 --> 00:41:38.725 that unattainable because you not have the 868 00:41:38.725 --> 00:41:39.885 data to actually run this. 869 00:41:40.225 --> 00:41:41.365 So allowing yourself 870 00:41:41.425 --> 00:41:43.085 to create a good data infrastructure 871 00:41:43.265 --> 00:41:44.445 is absolutely fundamental. 872 00:41:46.055 --> 00:41:49.035 And finally, right, again, I want to keep some time for Q&A, 873 00:41:49.035 --> 00:41:52.715 is, is just thinking about governance and risk, 874 00:41:53.135 --> 00:41:54.395 right, risk management. 875 00:41:54.565 --> 00:41:58.195 Bring risk very early in the process 876 00:41:59.215 --> 00:42:03.595 so you can have, um, you know, solutions that are practical 877 00:42:03.775 --> 00:42:06.155 and, you know, minimize the amount of risk 878 00:42:06.185 --> 00:42:07.315 that is involved, right? 879 00:42:07.735 --> 00:42:09.755 So in the world, in technical world, right, 880 00:42:09.755 --> 00:42:12.395 to have sometimes things like DevOps, right? Dev, and 881 00:42:12.535 --> 00:42:15.635 now we're moving much more in DevSecOps, right-- 882 00:42:16.025 --> 00:42:17.155 development, security, 883 00:42:17.215 --> 00:42:20.075 and operations-- is thinking about this 884 00:42:20.175 --> 00:42:21.795 for AI solutions as well. 885 00:42:22.375 --> 00:42:24.515 What's my governance? What are my risks? 886 00:42:25.095 --> 00:42:29.795 Uh, one good thing is that responsible AI, a set of guidelines 887 00:42:29.855 --> 00:42:32.995 and rules, have become much more, um, 888 00:42:33.425 --> 00:42:35.995 much more disseminated than they were in the past. 889 00:42:36.355 --> 00:42:38.235 I teach about AI in business. 890 00:42:38.315 --> 00:42:40.155 I have been teaching AI in business for like, you know, 891 00:42:40.215 --> 00:42:41.275 six, seven years now. 892 00:42:41.695 --> 00:42:44.275 And I remember, you know, a couple of years ago, I say, hey, 893 00:42:45.185 --> 00:42:47.995 talking about your responsible AI initiatives in 894 00:42:47.995 --> 00:42:51.395 organizations, you know, get blanks, blank stares back at 895 00:42:51.395 --> 00:42:52.555 what we're talking about, right? 896 00:42:52.975 --> 00:42:55.995 And now I see, you know, my students start to really see, 897 00:42:56.335 --> 00:42:59.075 oh, well, yeah, we have this approach for responsible AI. 898 00:42:59.075 --> 00:43:00.075 You have this framework, 899 00:43:00.095 --> 00:43:02.075 that's what I use, you adapted from this company, 900 00:43:02.175 --> 00:43:04.715 so on and so forth. That is pretty much we're heading. 901 00:43:04.855 --> 00:43:09.115 So again, risk management governance is something 902 00:43:09.115 --> 00:43:10.595 that going to minimize the risks. 903 00:43:11.135 --> 00:43:14.635 And of course, that actually real value in that, right? 904 00:43:14.635 --> 00:43:18.035 Because customers knowing that you care about them, right? 905 00:43:18.035 --> 00:43:20.635 Stakeholders know that you care about the reputation, 906 00:43:21.005 --> 00:43:23.795 about the implications really, you know, is, 907 00:43:23.815 --> 00:43:27.115 is well-perceived in organizations. In, sorry, 908 00:43:27.455 --> 00:43:28.675 in the marketplace as well. 909 00:43:29.065 --> 00:43:33.725 Okay? So let's go in some questions, uh, in that, 910 00:43:33.985 --> 00:43:36.445 uh, let's see what we have here. 911 00:43:36.705 --> 00:43:40.285 Um, trying to come up with some again 912 00:43:40.285 --> 00:43:42.365 and again, keep that in about like 10 minutes 913 00:43:42.385 --> 00:43:43.805 or so, sorry, you know, 914 00:43:43.825 --> 00:43:46.925 and feel free to reach out to me, um, you know, uh, 915 00:43:46.985 --> 00:43:48.205 in the LinkedIn or so, 916 00:43:48.205 --> 00:43:50.445 and we can continue the conversation there if don't have 917 00:43:50.445 --> 00:43:54.925 the, the change that. Um, so in terms 918 00:43:54.985 --> 00:43:59.005 of the university, uh, the university itself does have, um, 919 00:43:59.355 --> 00:44:03.405 some infrastructure right now to provide for some resources. 920 00:44:03.825 --> 00:44:07.125 IU Innovates is something that happens. 921 00:44:07.385 --> 00:44:10.405 So if you don't have IU Innovates, so, you know, um, 922 00:44:10.405 --> 00:44:12.365 that would be, if you are related to IU, 923 00:44:12.365 --> 00:44:13.445 that's where I would go. 924 00:44:13.625 --> 00:44:16.605 Um, um, in terms of the office at Indiana University. 925 00:44:17.025 --> 00:44:19.925 Um, but essentially looking at, uh, 926 00:44:19.925 --> 00:44:22.165 looking at the ecosystem where you are, right? 927 00:44:22.165 --> 00:44:24.205 If related to university might be something 928 00:44:24.205 --> 00:44:27.685 that you want you to, uh, to address your local university 929 00:44:27.745 --> 00:44:29.005 and figure out, you know, what's 930 00:44:29.005 --> 00:44:30.085 the potential of doing that? 931 00:44:30.535 --> 00:44:33.445 Thank you, Kathleen, appreciate that very much. 932 00:44:33.745 --> 00:44:37.725 Um, we talked a little bit about the large language models 933 00:44:37.795 --> 00:44:39.525 that are ethically sourced data. 934 00:44:40.065 --> 00:44:44.605 Uh, I would check. Um, it's, it's still very controversial, 935 00:44:44.605 --> 00:44:48.405 unfortunately. There's a lot of, uh, uh, companies that [unintelligible], 936 00:44:48.545 --> 00:44:50.805 uh, are using data without permission. 937 00:44:50.905 --> 00:44:55.005 Um, don't even start talking about artists and creative arts 938 00:44:55.065 --> 00:44:58.005 and stuff like that, um, because that's goal. 939 00:44:58.105 --> 00:45:00.205 So, uh, I wish I had a better answer, 940 00:45:00.305 --> 00:45:03.365 but it is probably involves a little bit more research in 941 00:45:03.365 --> 00:45:06.605 trying to figure out that, um, to, 942 00:45:08.135 --> 00:45:11.165 let's see, uh, ba, ba, ba, 943 00:45:11.195 --> 00:45:13.165 contemplate the effect of coaching education. 944 00:45:13.405 --> 00:45:14.525 Artificial intelligence what kind 945 00:45:14.525 --> 00:45:16.765 of labor supply will be required in the field 946 00:45:16.765 --> 00:45:18.805 of financial services and banking. 947 00:45:19.635 --> 00:45:22.485 What kind of new graduate degree programs would be necessary 948 00:45:22.505 --> 00:45:24.325 to produce effective workers in the field 949 00:45:24.325 --> 00:45:26.525 of financial management in corporations? 950 00:45:26.995 --> 00:45:29.045 Awesome. Awesome question, right? 951 00:45:29.505 --> 00:45:32.445 Um, this is, previous questions were great as well, 952 00:45:32.585 --> 00:45:34.885 but this is something that actually we, we, 953 00:45:34.985 --> 00:45:36.605 we have a guest speaking from, uh, 954 00:45:37.075 --> 00:45:39.485 from a financial institution talking about AI at 955 00:45:39.485 --> 00:45:40.605 Kelley, uh, next week. 956 00:45:41.065 --> 00:45:43.285 And what we have seen, right? 957 00:45:44.005 --> 00:45:48.565 A lot of it is, is, um, from a genetic, genetic view, 958 00:45:48.905 --> 00:45:52.245 the way I trying to train my students, I know that, uh, 959 00:45:52.385 --> 00:45:54.205 you know, Jamie and others are here. 960 00:45:54.345 --> 00:45:55.485 Hi, Jamie, nice to see you. 961 00:45:55.945 --> 00:46:00.045 Um, so it's, it's figure out what's the art 962 00:46:00.045 --> 00:46:01.125 of the possible, right? 963 00:46:01.355 --> 00:46:04.325 What exactly can you think about 964 00:46:05.025 --> 00:46:08.125 how these technologies can be applied to 965 00:46:08.835 --> 00:46:10.565 different scenarios, right? 966 00:46:10.665 --> 00:46:15.085 So I push a lot my students to say, alright, this is 967 00:46:15.085 --> 00:46:19.485 what we know about, um, intelligent, you know, uh, 968 00:46:19.535 --> 00:46:21.925 interfaces, for example, just to give a topic, right? 969 00:46:22.355 --> 00:46:25.805 This is what we know about effective AI, right? 970 00:46:26.265 --> 00:46:30.165 How to read emotions, how to read, you know, uh, scenarios. 971 00:46:30.185 --> 00:46:33.565 How to read voice anxiety, how to read, uh, 972 00:46:33.625 --> 00:46:34.965 you know, the AI, right? 973 00:46:35.305 --> 00:46:39.165 How to understand what the potential, uh, implications of, 974 00:46:39.345 --> 00:46:41.845 you know, your eyes moving a particular way. 975 00:46:42.305 --> 00:46:44.165 So that's what we know, right? 976 00:46:44.705 --> 00:46:48.045 And, um, let's think about potential applications on that. 977 00:46:48.185 --> 00:46:50.525 And that's why I push my students a lot to say, 978 00:46:51.065 --> 00:46:52.645 get us a baseline of technology. 979 00:46:52.785 --> 00:46:56.365 You might not know everything about technology. 980 00:46:56.825 --> 00:47:00.125 Um, so I, but understand what the outputs 981 00:47:00.465 --> 00:47:03.845 of the outcomes are, the technology in thinking about 982 00:47:04.025 --> 00:47:05.965 how they can change processes. 983 00:47:06.385 --> 00:47:08.445 And of course, think about the risks as well, right? 984 00:47:08.635 --> 00:47:10.845 Privacy risks and so on and so forth. 985 00:47:11.385 --> 00:47:14.765 Um, and the accuracy and the potential biases, right? 986 00:47:15.185 --> 00:47:16.645 And understand the whole package 987 00:47:17.675 --> 00:47:20.205 because this is what you want to have people 988 00:47:20.205 --> 00:47:21.965 who are innovating the processes 989 00:47:22.425 --> 00:47:26.245 and understanding what the processes can be once you apply 990 00:47:26.245 --> 00:47:28.925 these new technologies in place so 991 00:47:29.005 --> 00:47:32.005 that you can have a better way and perhaps where AI 992 00:47:32.665 --> 00:47:35.485 and humans are helping each other, right? 993 00:47:35.485 --> 00:47:37.365 Humans are a very good issue. 994 00:47:37.365 --> 00:47:39.685 Always going to be better than AI, at least 995 00:47:39.685 --> 00:47:42.325 for a long time in certain tasks. 996 00:47:42.505 --> 00:47:45.245 And AI is going to be very good in certain tasks, right? 997 00:47:45.945 --> 00:47:48.485 How can we design, redesign process there? 998 00:47:48.485 --> 00:47:50.605 So I think that anyone that kind of thinks about 999 00:47:50.605 --> 00:47:53.765 that creative in financial services, is 1000 00:47:53.775 --> 00:47:54.925 about processes, right? 1001 00:47:54.955 --> 00:47:59.085 It's about security, is about, uh, uh, uh, governance. 1002 00:47:59.235 --> 00:48:01.965 It's about auditing, it's about serving customers. 1003 00:48:02.425 --> 00:48:05.765 How can you create ways in which AI is not [unintelligible], right? 1004 00:48:06.115 --> 00:48:07.925 It's not put like, hey, you have AI, 1005 00:48:08.265 --> 00:48:09.685 but actually effective, 1006 00:48:10.195 --> 00:48:12.405 effectively change the process itself. 1007 00:48:13.425 --> 00:48:16.165 Ah, cool, cool, cool. So let me go over here. 1008 00:48:18.435 --> 00:48:22.855 Yes. So there's a great point about is smart employees know 1009 00:48:22.855 --> 00:48:24.535 they have to fact check the AI. 1010 00:48:24.955 --> 00:48:28.665 Um, oh, I love this question 1011 00:48:28.665 --> 00:48:29.905 because it's a question about 1012 00:48:30.005 --> 00:48:32.265 why don't you just do it instead of checking AI. 1013 00:48:32.685 --> 00:48:37.585 So there is specific situations where we use AI, 1014 00:48:38.565 --> 00:48:41.145 um, where even with all the mistakes, right? 1015 00:48:41.265 --> 00:48:43.345 I mean, idea generation is something 1016 00:48:43.545 --> 00:48:45.825 that AI can actually do quite well. 1017 00:48:46.365 --> 00:48:51.345 Um, hey, give me some, you know, um, potential titles for, 1018 00:48:52.125 --> 00:48:53.305 for a talk, right? 1019 00:48:53.845 --> 00:48:55.905 Um, speed up the process. 1020 00:48:56.385 --> 00:48:59.065 Anything that can speed up the process, um, 1021 00:48:59.435 --> 00:49:00.585 might be helpful. 1022 00:49:00.925 --> 00:49:04.025 Uh, for example, putting together a presentation, um, 1023 00:49:04.265 --> 00:49:05.905 I was teaching in my class last week, 1024 00:49:06.245 --> 00:49:07.465 and I ask people together, 1025 00:49:07.525 --> 00:49:09.905 to put together a little presentation for me in 20 minutes. 1026 00:49:10.405 --> 00:49:13.745 And every single group, right, they discuss the ideas before 1027 00:49:14.245 --> 00:49:16.105 and everything, and they say, hey, just get, Jim 1028 00:49:16.105 --> 00:49:17.985 and I should put together the actual presentation 1029 00:49:18.205 --> 00:49:20.065 and then eventually validate the presentation. 1030 00:49:20.445 --> 00:49:21.825 So a lot of those tasks 1031 00:49:22.215 --> 00:49:25.105 that you can evaluate the outcome yourself, 1032 00:49:25.525 --> 00:49:28.945 but can save you some time, those I think are really, 1033 00:49:28.945 --> 00:49:31.265 really good fit for, for AI, right? 1034 00:49:31.665 --> 00:49:34.345 I, the way that I always represent AI to me is like, 1035 00:49:34.345 --> 00:49:37.185 think about having a very smart intern 1036 00:49:37.935 --> 00:49:41.345 that you ask the intern to do something, the intern is going 1037 00:49:41.345 --> 00:49:42.945 to be joyfully do that, 1038 00:49:42.945 --> 00:49:45.625 and present a really good results and that, 1039 00:49:46.365 --> 00:49:47.585 but with two little things. 1040 00:49:47.845 --> 00:49:50.225 One is that you cannot really trust the 1041 00:49:50.225 --> 00:49:51.385 intern to run the show. 1042 00:49:51.765 --> 00:49:53.025 You have to validate that. 1043 00:49:53.525 --> 00:49:57.785 And two, your intern is also a big fan, a big fan 1044 00:49:57.805 --> 00:50:00.705 of Fortnite and might have spent the whole night playing 1045 00:50:00.705 --> 00:50:02.545 Fortnite and made up things from time 1046 00:50:02.545 --> 00:50:04.745 to time in the outcome, in the result, right? 1047 00:50:05.165 --> 00:50:07.785 So as long as you understand that it's a trust 1048 00:50:07.805 --> 00:50:10.745 but verify situation where essentially looking for parts 1049 00:50:10.765 --> 00:50:13.225 of the process that can, you know, shrink, um, 1050 00:50:13.695 --> 00:50:16.145 that is something that have like, you know, something like, 1051 00:50:16.145 --> 00:50:18.665 you know, uh, creating, writing, modifying the writing 1052 00:50:18.885 --> 00:50:20.065 of something that you create 1053 00:50:20.125 --> 00:50:21.465 to become a little bit more elegant. 1054 00:50:22.115 --> 00:50:24.825 Those things that save time, as long as you're able 1055 00:50:24.825 --> 00:50:28.165 to verify the outcomes, um, that's, those are great. 1056 00:50:28.225 --> 00:50:30.085 And I know this because I have my students, 1057 00:50:30.105 --> 00:50:31.565 you know, creating essays for me. 1058 00:50:31.625 --> 00:50:33.285 And from time to time I like, yeah, 1059 00:50:33.285 --> 00:50:36.885 that was totally generated for AI because you never read it. 1060 00:50:37.025 --> 00:50:38.165 Uh, and that is wrong. 1061 00:50:38.865 --> 00:50:42.085 But, so that is something that, uh, you know, saving time, 1062 00:50:42.205 --> 00:50:46.665 I think that's been approach here. Um, to, to, to, to, to 1063 00:50:47.125 --> 00:50:49.865 how are companies getting data from Sysnet not, um, 1064 00:50:50.045 --> 00:50:51.625 so Thomas, that's a great question. 1065 00:50:51.765 --> 00:50:54.185 Um, but I think that the data component, 1066 00:50:54.205 --> 00:50:56.065 the actual data, right? 1067 00:50:56.815 --> 00:51:00.025 This has both is a, there's technical solutions for that, 1068 00:51:00.085 --> 00:51:02.225 but there is always been a problem, right? 1069 00:51:02.335 --> 00:51:05.465 Sometimes have data that were not intended to do anything, 1070 00:51:05.925 --> 00:51:08.065 and they're running some analysis in that data. 1071 00:51:08.725 --> 00:51:13.345 Um, you know, data for, um, marketing that's now used for, 1072 00:51:13.965 --> 00:51:15.705 um, you know, recalls, right? 1073 00:51:16.895 --> 00:51:18.665 Reliability might not be quite the same. 1074 00:51:18.765 --> 00:51:21.625 So understanding data sourcing, understanding data 1075 00:51:21.785 --> 00:51:24.825 [unintelligible] in understanding kind of the mechanisms 1076 00:51:24.825 --> 00:51:28.345 that leads you to modify, uh, you know, to, 1077 00:51:28.365 --> 00:51:30.265 to assess the quality of the data. 1078 00:51:30.595 --> 00:51:33.385 Those are very important. That's a part of data governance. 1079 00:51:33.765 --> 00:51:36.025 So you have struggle with that, right, 1080 00:51:36.165 --> 00:51:39.185 for a while, but there is actually technical solutions 1081 00:51:39.485 --> 00:51:40.825 and management solutions 1082 00:51:40.825 --> 00:51:45.145 to deal with data that is migrated to different purposes 1083 00:51:45.145 --> 00:51:47.785 because we had to face that for quite some time. 1084 00:51:48.685 --> 00:51:51.945 Oh, a FOMO approach, sorry, Paul is a fear of missing out, 1085 00:51:52.165 --> 00:51:55.465 um, which is everybody jumps into AI just 1086 00:51:55.465 --> 00:51:57.585 because everybody else is doing that, and 1087 00:51:57.585 --> 00:51:59.425 because everybody else is doing that 1088 00:51:59.445 --> 00:52:01.905 so I have to do it, uh, without much that. 1089 00:52:02.605 --> 00:52:07.085 Um, so process 1090 00:52:07.545 --> 00:52:10.205 of, uh, um, categorization 1091 00:52:10.205 --> 00:52:13.445 of AI skills from nontechnical skilled workforce group, 1092 00:52:13.465 --> 00:52:15.005 in this case, business operations staff, 1093 00:52:15.005 --> 00:52:16.525 always done things that way. 1094 00:52:18.635 --> 00:52:19.975 That's a tough one, Jamie. 1095 00:52:20.155 --> 00:52:24.655 Um, that might be something we might discuss more in, uh, 1096 00:52:25.115 --> 00:52:27.295 in, uh, perhaps on LinkedIn or something. 1097 00:52:27.835 --> 00:52:29.695 Um, I think that we're, 1098 00:52:29.835 --> 00:52:33.225 but we're not quite there yet. 1099 00:52:33.285 --> 00:52:35.185 I'm, I'm, I see how I answered that. 1100 00:52:35.735 --> 00:52:39.905 That is, that is, I think that was a start to really set up 1101 00:52:40.575 --> 00:52:44.025 some skill sets that are valuable. 1102 00:52:44.045 --> 00:52:45.585 And this is part of the struggle in 1103 00:52:45.585 --> 00:52:46.785 university as well, right? 1104 00:52:47.005 --> 00:52:49.305 We are, we need to train students that are going 1105 00:52:49.305 --> 00:52:51.345 to be in the workforce where they're going 1106 00:52:51.345 --> 00:52:52.985 to be using AI all the time. 1107 00:52:53.965 --> 00:52:57.625 How can I, you know, making sure the students are getting skills 1108 00:52:57.735 --> 00:52:59.625 that are trained to brought to a variety 1109 00:52:59.645 --> 00:53:01.385 of different processes, right? 1110 00:53:02.065 --> 00:53:05.185 I think that is a lot of that is understanding capabilities, 1111 00:53:05.185 --> 00:53:06.785 understanding what is possible, 1112 00:53:06.785 --> 00:53:10.625 understanding, do like a responsible use of AI, understanding 1113 00:53:10.625 --> 00:53:11.905 that, you know, as you was said 1114 00:53:11.905 --> 00:53:14.305 before, sometimes there is, there is risks 1115 00:53:14.565 --> 00:53:16.345 and the outcome is not great. 1116 00:53:16.885 --> 00:53:19.025 So how can we understand, you know, 1117 00:53:19.025 --> 00:53:21.185 that is always should be a human in the loop, 1118 00:53:21.205 --> 00:53:23.345 at least at this stage of development of AI, 1119 00:53:23.805 --> 00:53:26.045 to make decisions about the validity 1120 00:53:26.105 --> 00:53:28.285 of the initial outcome, right, 1121 00:53:28.385 --> 00:53:31.725 in the assessment. So, I mean, I was, uh, talking to someone 1122 00:53:31.825 --> 00:53:32.925 of a faculty here 1123 00:53:32.925 --> 00:53:35.005 before, it's like, hey, if you are an expert already 1124 00:53:35.065 --> 00:53:39.485 and are using AI to minimize your time, that's pretty solid. 1125 00:53:39.485 --> 00:53:40.965 That's something that we actually, you know, 1126 00:53:40.965 --> 00:53:42.165 can do relatively well. 1127 00:53:42.625 --> 00:53:44.885 But if you are someone completely new 1128 00:53:44.945 --> 00:53:48.885 and you have no basis to evaluate the accuracy of outcomes, 1129 00:53:49.465 --> 00:53:51.045 that's a little bit touchy, right? 1130 00:53:51.045 --> 00:53:52.845 That's something that have, so I think that's something 1131 00:53:52.845 --> 00:53:54.045 that you have to evolve 1132 00:53:54.045 --> 00:53:56.325 with you in terms of how we go on that. 1133 00:53:57.985 --> 00:54:00.165 Who is best suited to lead this effort? 1134 00:54:00.505 --> 00:54:03.445 Uh, I, that's a, that's a great question. 1135 00:54:03.545 --> 00:54:05.165 [Kim] That's a great question. [Alex] Yeah. 1136 00:54:05.345 --> 00:54:08.845 Um, and, uh, it is just, I, I think that is, 1137 00:54:09.665 --> 00:54:12.525 so when I teach my students, my students, uh, many 1138 00:54:12.525 --> 00:54:14.205 of my students are MBA students, right? 1139 00:54:14.205 --> 00:54:16.605 They always tell, like, you know, hey, they are the ones. 1140 00:54:16.625 --> 00:54:17.725 So I, dude, 1141 00:54:18.105 --> 00:54:20.805 and I think it's really a collaboration between, um, 1142 00:54:21.345 --> 00:54:25.285 the technical people and the business people, right? 1143 00:54:25.555 --> 00:54:28.685 They have to be aligned. They absolutely have to be aligned. 1144 00:54:28.685 --> 00:54:31.285 This is a big challenge in many technical issues, is 1145 00:54:31.825 --> 00:54:34.965 how do you align people that actually doing both, um, 1146 00:54:35.335 --> 00:54:37.645 understand the business consequences 1147 00:54:38.065 --> 00:54:41.045 and understanding also the technology possibilities, right? 1148 00:54:41.265 --> 00:54:45.085 So creating, uh, structures that are really looking 1149 00:54:45.145 --> 00:54:46.845 for value streams instead 1150 00:54:46.845 --> 00:54:49.325 of just a translation from tech to business. 1151 00:54:49.745 --> 00:54:52.805 Um, this is always my recommendation in terms of, you know, 1152 00:54:52.805 --> 00:54:56.045 making sure that we're a able to get the value. To, to, to, 1153 00:54:56.045 --> 00:55:00.645 to, alright, so I'm going to pass, um, 1154 00:55:00.645 --> 00:55:03.725 because I just wanna make sure that I give the time to, um, 1155 00:55:04.885 --> 00:55:06.465 to Kim, uh, that, uh, 1156 00:55:06.565 --> 00:55:08.465 and by the way, if anyone wants to reach out 1157 00:55:08.465 --> 00:55:10.705 to me in LinkedIn or, you know, uh, via email, 1158 00:55:10.925 --> 00:55:12.225 et cetera, I love to discuss. 1159 00:55:12.225 --> 00:55:13.505 Those are really great questions. 1160 00:55:13.555 --> 00:55:17.185 Thank you so much, uh, really for, uh, for sharing those. 1161 00:55:17.205 --> 00:55:20.145 But I'm going to pass you, Kim, so that, uh, oh, 1162 00:55:20.145 --> 00:55:21.545 I forgot to actually move this slide. 1163 00:55:21.565 --> 00:55:23.425 The Q&A, we're in the Q&A, sorry for that. 1164 00:55:23.765 --> 00:55:26.025 Um, so Kim, I'll, I'll let you take over. 1165 00:55:26.045 --> 00:55:28.905 And again, please reach out to me, in LinkedIn or email 1166 00:55:29.125 --> 00:55:30.705 and we can continue the conversation. 1167 00:55:32.765 --> 00:55:36.665 [Kim] Thanks Alex. So, um, 1168 00:55:37.105 --> 00:55:39.905 I think next slide we have a little bit of information on. 1169 00:55:40.015 --> 00:55:44.105 Okay, so today's session, um, if you enjoyed Alex, 1170 00:55:44.115 --> 00:55:45.265 which I'm sure you all did, 1171 00:55:45.265 --> 00:55:46.265 and you'd like to learn more, 1172 00:55:46.445 --> 00:55:49.145 we have a really wonderful professional development course 1173 00:55:49.545 --> 00:55:51.665 offering starting on September 2nd. 1174 00:55:52.285 --> 00:55:55.905 It runs September 2nd through October 7th. 1175 00:55:56.565 --> 00:55:58.505 And the course title is Rise of AI. 1176 00:55:58.525 --> 00:55:59.785 And Alex, I'll let you jump in 1177 00:55:59.785 --> 00:56:01.705 and provide a little bit more detail around it, 1178 00:56:01.805 --> 00:56:03.945 but it's, it's a hundred percent online. 1179 00:56:04.405 --> 00:56:06.585 Um, we have live online sessions, 1180 00:56:06.765 --> 00:56:08.505 the duration's only about six weeks, 1181 00:56:08.565 --> 00:56:11.305 so a really great course for, for you if you're just trying 1182 00:56:11.305 --> 00:56:14.465 to learn a little bit more and, um, get like a certificate 1183 00:56:14.525 --> 00:56:15.745 or whatever you're looking for. 1184 00:56:16.045 --> 00:56:18.225 Um, Alex, do you wanna talk a little bit about the, 1185 00:56:18.565 --> 00:56:20.585 the content and what there's, it's you 1186 00:56:20.585 --> 00:56:23.925 and two other professors, so [Alex] yes [Kim] a great, uh, cohort 1187 00:56:24.145 --> 00:56:26.125 of faculty presenting on this topic. 1188 00:56:27.085 --> 00:56:28.655 [Alex] Yeah, sure thing. Thank you, Kim. 1189 00:56:29.035 --> 00:56:31.015 Um, so there are three of us. It's me, 1190 00:56:31.155 --> 00:56:33.815 I'm going to handle the AI for organizations, uh, 1191 00:56:33.835 --> 00:56:35.975 before, you know, uh, kind of, uh, you know, 1192 00:56:36.165 --> 00:56:38.415 extending much more about what we discussed here, 1193 00:56:38.795 --> 00:56:41.215 but with more details and more use cases. 1194 00:56:41.595 --> 00:56:45.215 Um, the first part is actually done, um, by Sagar Samtani. 1195 00:56:45.635 --> 00:56:49.895 Sagar is a bright, uh, very, the top AI researchers, uh, 1196 00:56:50.035 --> 00:56:52.335 uh, you know, that we, we see around in the US. 1197 00:56:53.105 --> 00:56:54.535 Sagar was also the director. 1198 00:56:54.675 --> 00:56:56.455 He took a leave of absence for one year 1199 00:56:56.455 --> 00:57:00.255 to become the director of AI for TSMC, so 1200 00:57:00.835 --> 00:57:03.455 the big Taiwanese, uh, uh, ship, uh, 1201 00:57:03.525 --> 00:57:05.215 chip manufacturing company, right? 1202 00:57:05.435 --> 00:57:08.775 So, um, um, he was there in industry, 1203 00:57:08.955 --> 00:57:10.095 uh, talking about that. 1204 00:57:10.195 --> 00:57:12.175 So a lot of, uh, a lot of great content. 1205 00:57:12.355 --> 00:57:15.175 And the third, uh, faculty going to be talking about the AI 1206 00:57:15.195 --> 00:57:17.535 for individuals is Alan Dennis. 1207 00:57:17.715 --> 00:57:19.735 And Alan is a researcher, has been working 1208 00:57:19.735 --> 00:57:23.375 with AI research from an individual-use perspective 1209 00:57:23.915 --> 00:57:25.335 for eight years or so. 1210 00:57:25.755 --> 00:57:28.695 Um, he has some really great stuff in things like, you know, 1211 00:57:28.695 --> 00:57:30.775 digital humans, um, you know, 1212 00:57:30.795 --> 00:57:33.415 and he's actually going to be doing some cool demos about, 1213 00:57:33.515 --> 00:57:35.135 uh, that area as well, 1214 00:57:35.195 --> 00:57:37.375 but also he's going to be talking about things like a prompt 1215 00:57:37.655 --> 00:57:38.895 engineering, right? 1216 00:57:38.895 --> 00:57:41.375 Things that are very practical, uh, as well. 1217 00:57:41.455 --> 00:57:43.335 I mean, I think everything is going to be, try 1218 00:57:43.335 --> 00:57:44.335 to be super practical, 1219 00:57:44.635 --> 00:57:47.135 but that's the definition of those, those course. 1220 00:57:47.155 --> 00:57:49.975 And I hope to see many of you there, Kim, back to you. 1221 00:57:50.305 --> 00:57:53.455 [Kim] Great, great. And we have a QR code here, um, on 1222 00:57:54.135 --> 00:57:55.215 previous slide, so feel free. 1223 00:57:55.355 --> 00:57:57.015 Um, we're gonna, we're recording the session. 1224 00:57:57.145 --> 00:57:58.255 We'll send out the recording. 1225 00:57:58.255 --> 00:58:01.255 So if you wanna scan the QR code for more information, 1226 00:58:01.275 --> 00:58:04.415 you'll, you'll see all of the course details there, uh, 1227 00:58:04.415 --> 00:58:07.295 pricing, dates, registration information, so, 1228 00:58:07.475 --> 00:58:08.695 um, you'll be all set there. 1229 00:58:10.755 --> 00:58:13.015 [Alex] So very quickly here before, back to Kim. 1230 00:58:13.235 --> 00:58:15.375 Um, so we also have new, uh, 1231 00:58:15.555 --> 00:58:18.135 online MS programs, uh, for both. 1232 00:58:18.445 --> 00:58:20.815 There's a lot of content about AI, so 1233 00:58:21.875 --> 00:58:24.455 that's when I have a whole course on that. 1234 00:58:24.835 --> 00:58:27.615 [Kim] Mm-hmm. [Alex] Um, uh, we also have Sagar teaching the program 1235 00:58:27.715 --> 00:58:30.855 as well, talking about, uh, generative AI and deep learning. 1236 00:58:31.435 --> 00:58:35.335 So, uh, both, uh, both the MS in IT Management 1237 00:58:35.435 --> 00:58:38.335 and the MS in Business Analytics, the QR codes are there. 1238 00:58:38.715 --> 00:58:41.215 So if you, you know, um, of course is, is, 1239 00:58:41.275 --> 00:58:43.135 is a deeper dive of course. 1240 00:58:43.395 --> 00:58:44.975 Um, but there's a lot 1241 00:58:44.975 --> 00:58:48.735 of content in about AI in those two courses from AI 1242 00:58:48.735 --> 00:58:51.655 for business, machine learning, with a lot of, uh, 1243 00:58:51.655 --> 00:58:54.575 practical experimentation, deep learning, right? 1244 00:58:54.575 --> 00:58:58.015 GenAI, actually, you, I was talking to Sagar last week, 1245 00:58:58.155 --> 00:59:00.055 and he plans to have projects 1246 00:59:00.055 --> 00:59:02.615 where people can actually create solutions in the 1247 00:59:02.615 --> 00:59:03.775 class, right? 1248 00:59:03.875 --> 00:59:06.455 To, to come up with some uses of large language models. 1249 00:59:06.675 --> 00:59:09.975 So really, really good stuff. And again, QR code available. 1250 00:59:10.195 --> 00:59:12.655 Uh, but if you are interested, just let us know 1251 00:59:12.835 --> 00:59:16.095 and happy to direct you to the people who can, you know, 1252 00:59:16.095 --> 00:59:17.255 give you even more information. 1253 00:59:17.375 --> 00:59:19.495 I can give information, but if you want more about the 1254 00:59:19.495 --> 00:59:22.285 process, et cetera, uh, Kim, back to you again. 1255 00:59:23.165 --> 00:59:24.775 [Kim] Wonderful. Thank you. 1256 00:59:25.245 --> 00:59:26.655 Okay, so thanks Alex, 1257 00:59:26.875 --> 00:59:28.375 and thank you all for joining us today. 1258 00:59:28.445 --> 00:59:30.775 Just before we close, I wanna provide just a little bit 1259 00:59:30.775 --> 00:59:33.895 of information about Kelley Executive Education in cor-, in case 1260 00:59:33.895 --> 00:59:35.495 you aren't familiar with our unit. 1261 00:59:35.995 --> 00:59:38.415 Um, our programs are offered in a 1262 00:59:38.415 --> 00:59:39.535 variety of different formats. 1263 00:59:39.655 --> 00:59:42.695 So online, in person, in Indianapolis, um, 1264 00:59:42.725 --> 00:59:44.455 also in Bloomington, hybrid. 1265 00:59:44.835 --> 00:59:47.095 Within our professional development portfolio, 1266 00:59:47.355 --> 00:59:49.655 we have courses on topics such as leadership 1267 00:59:49.655 --> 00:59:52.655 and management; AI and tech, which is rapidly growing, 1268 00:59:52.755 --> 00:59:55.535 so keep an eye on out for, um, new programs in that field, 1269 00:59:55.535 --> 00:59:59.655 that, that topic; operations; finance; um, 1270 00:59:59.675 --> 01:00:01.335 and number of different offerings. 1271 01:00:01.635 --> 01:00:04.735 Um, if you're looking for more on AI, I know some 1272 01:00:04.735 --> 01:00:05.855 of you commented, um, 1273 01:00:06.075 --> 01:00:08.935 do you provide certifications on topics around AI? 1274 01:00:09.195 --> 01:00:11.495 We, we have Rise of AI, September 2nd with Alex, 1275 01:00:11.595 --> 01:00:14.495 so we mentioned. Um, AI, AI Applications 1276 01:00:14.495 --> 01:00:16.695 in Marketing launching August 5th. 1277 01:00:17.155 --> 01:00:21.375 And then we also have a several other courses launching in 1278 01:00:21.375 --> 01:00:22.855 the fall that we'll be announcing soon. 1279 01:00:22.915 --> 01:00:24.135 So keep an eye out for those. 1280 01:00:24.715 --> 01:00:27.095 If you have any questions at all, um, feel free 1281 01:00:27.095 --> 01:00:28.175 to email me anytime. 1282 01:00:28.995 --> 01:00:30.175 Um, on the next slide, 1283 01:00:30.235 --> 01:00:32.335 you'll see there's a professional development, 1284 01:00:32.835 --> 01:00:34.175 uh, email that we have. 1285 01:00:34.435 --> 01:00:37.975 The, our, our advisors will follow up with you immediately, 1286 01:00:38.115 --> 01:00:39.655 or you can also shoot me an email. 1287 01:00:39.795 --> 01:00:42.695 I'm always happy to set up a call, um, talk you 1288 01:00:42.695 --> 01:00:44.055 through any questions that you have. 1289 01:00:44.635 --> 01:00:47.455 If you are an IU alumni joining us today, 1290 01:00:47.455 --> 01:00:51.575 which we hope you are, um, you can receive 15% off of any 1291 01:00:51.575 --> 01:00:53.375 of our professional development course offerings. 1292 01:00:53.515 --> 01:00:57.135 So you can use that code "SAVE15" for any 1293 01:00:57.135 --> 01:00:58.295 of the courses that we offer. 1294 01:00:58.395 --> 01:00:59.535 So just keep that in mind. 1295 01:00:59.555 --> 01:01:02.375 That's a nice little perk, um, for alumni and staff. 1296 01:01:03.315 --> 01:01:04.695 And you see the, 1297 01:01:04.755 --> 01:01:07.495 the QR code here will go directly to our websites. 1298 01:01:07.495 --> 01:01:09.455 You can learn more about all of our program offerings. 1299 01:01:09.915 --> 01:01:11.735 Um, but like I said, if you have any questions at all, 1300 01:01:11.845 --> 01:01:13.695 feel free to shoot me an email anytime. 1301 01:01:13.955 --> 01:01:16.175 Um, Alex as well, we're always happy to talk to you, 1302 01:01:16.635 --> 01:01:18.055 and thank you guys for joining us today. 1303 01:01:19.235 --> 01:01:20.665 [Alex] Thank you so much, everybody. 1304 01:01:20.805 --> 01:01:21.985 Uh, please keep in touch 1305 01:01:22.165 --> 01:01:24.545 and, uh, let's keep the conversation going.