WEBVTT 1 00:00:00.160 --> 00:00:09.800 Families Teamsters nineteen thirty two dot org. The information economy has a ride. 2 00:00:09.919 --> 00:00:15.400 The world is teeming with innovation as new business models reinvent every industry industry. 3 00:00:15.480 --> 00:00:20.800 Inside Analysis is your source of information and insight about how to make the most 4 00:00:20.879 --> 00:00:25.600 of this exciting new era. Learn more at inside analysis dot com, Inside 5 00:00:25.640 --> 00:00:34.920 analysis dot com and now here's your host through Eric Kavanaugh. Yes, oh, 6 00:00:35.000 --> 00:00:40.200 yes, indeed, folks, welcome to the future. The future is 7 00:00:40.200 --> 00:00:45.039 here already. That's what William Gibson once wrote. It's just not evenly distributed. 8 00:00:45.479 --> 00:00:49.159 I have to say I heard that line years ago and it just blew 9 00:00:49.200 --> 00:00:53.439 my mind. I was like, what, that's pretty cool. The future 10 00:00:53.520 --> 00:00:59.079 is here already. It's just not evenly distributed yet. But that's happening right 11 00:00:59.119 --> 00:01:02.560 now. You see it with so for example, you see it with certain 12 00:01:02.679 --> 00:01:07.000 kinds of construction, skyscrapers. You look at the new skyscrapers and places like 13 00:01:07.079 --> 00:01:11.480 Dubai in the UAE and they're just amazing. And that is the forefront of 14 00:01:11.519 --> 00:01:14.480 the industry, and it's primarily because they have a lot of money. You 15 00:01:14.640 --> 00:01:17.560 want to build some cool buildings. And in the rest of the world you 16 00:01:17.680 --> 00:01:22.120 have older buildings that are still standing obviously, but I bring in this metaphor 17 00:01:22.319 --> 00:01:26.840 just to point out that we are at a very transformative phase or moment in 18 00:01:27.000 --> 00:01:30.680 time as human beings, quite frankly, and with technology, of course, 19 00:01:32.120 --> 00:01:36.000 it's always out there improving, usually improving our lives, doing cool things for 20 00:01:36.079 --> 00:01:40.959 us well. Artificial intelligence not new. It's been around for decades, fifty 21 00:01:41.079 --> 00:01:44.480 sixty years if you really push it all the way back even further depending upon 22 00:01:44.519 --> 00:01:49.879 your definition. But in terms of widespread adoption and use, boom, we 23 00:01:49.959 --> 00:01:53.120 are in the middle of that first major explosion. And you can thank the 24 00:01:53.120 --> 00:02:00.560 folks that open AI for sure with their chat GPT, which just the doors 25 00:02:00.599 --> 00:02:04.120 off. Now that even isn't all that new. GPT's been around for a 26 00:02:04.200 --> 00:02:07.280 number of years now, but this version that they unleashed on the public really 27 00:02:07.319 --> 00:02:13.479 opened eyeballs and it forced the game. It was a forcing function. But 28 00:02:13.520 --> 00:02:15.479 I have someone on the show today who knows a whole lot about AI and 29 00:02:15.520 --> 00:02:19.400 has known a lot about AI for a long long time, and he taught 30 00:02:19.400 --> 00:02:22.960 me a lot about this stuff too, Osama Fayad. He's got a company, 31 00:02:22.960 --> 00:02:25.360 Open Insights, but also he was the chief data officer of a company 32 00:02:25.360 --> 00:02:30.400 called Yahoo some of you all may remember those guys. They were the center 33 00:02:30.439 --> 00:02:32.439 of gravity in like the nineties and the early aughts. That was where you 34 00:02:32.479 --> 00:02:36.520 wanted to be as an engineer, as an AI person. And then of 35 00:02:36.560 --> 00:02:38.840 course it kind of shifted to companies like Cloud, Deerra and Google of course 36 00:02:38.879 --> 00:02:42.719 and now open Ai. But they're still out there. Yeah, he's still 37 00:02:42.759 --> 00:02:46.719 doing cool stuff and so is Osama Asama Fayad. He is now the executive 38 00:02:46.759 --> 00:02:52.800 director for the Institute for Experiential AI at Northeastern University. Sama, welcome to 39 00:02:52.840 --> 00:02:57.240 the show here. Thanks so much for your time and attention, and first 40 00:02:57.280 --> 00:03:01.080 of all, tell us about the Northeastern University Institute for Experiential AI. Where 41 00:03:01.080 --> 00:03:06.639 did this idea come from and what's your focus? Thanks? Thanks, Eric. 42 00:03:07.719 --> 00:03:16.280 Idea came from the president of the university, Joseph and the provost David 43 00:03:16.319 --> 00:03:23.599 Madigan, who's actually a well established statistician an AI expert in his own right, 44 00:03:23.800 --> 00:03:30.400 and some donors who gave some large grants saying hey, AI is going 45 00:03:30.479 --> 00:03:35.039 to be big. This was about This was five years ago, so around 46 00:03:35.560 --> 00:03:38.680 late two thousand they started saying, Okay, we've got a lot of funding, 47 00:03:38.759 --> 00:03:44.840 let's recruit for someone to head it up and define it the way we 48 00:03:45.479 --> 00:03:51.080 I'm the inaugural executive director for this, as well as joining as a professor 49 00:03:51.159 --> 00:03:59.319 in Computer Science College Cory College for Computer Sciences. So experiential AI, the 50 00:03:59.360 --> 00:04:02.039 first thing you need to do is name it. And we use the term 51 00:04:02.120 --> 00:04:06.840 experiential AI to meet AI with the human in the loop. Because the reality 52 00:04:06.960 --> 00:04:14.319 of my exposure to AI through my work, whether it was with Microsoft, 53 00:04:14.520 --> 00:04:20.560 NASA's GPL, my own startups, Yahoo, later with Barclays, and definitely 54 00:04:20.600 --> 00:04:27.439 with Open Insights, working with many companies around the world including AI in the 55 00:04:27.480 --> 00:04:33.040 healthcare space, etc. Has shown me one theme that is constant, which 56 00:04:33.079 --> 00:04:39.560 is every working AI has a huge element of the human in the loop, 57 00:04:40.040 --> 00:04:44.519 whether you're talking about the core of the Google Search Engine or we're talking about 58 00:04:45.120 --> 00:04:48.279 open AI and chat GPT in any of these large language models, a lot 59 00:04:48.279 --> 00:04:55.560 of feedback all the time from humans supplying what AI cannot supply. And like 60 00:04:55.639 --> 00:05:00.480 you mentioned, we've spent over seven decades trying to figure out stuff like common 61 00:05:00.480 --> 00:05:04.160 sense reasoning. One of the big learning lessons is you only can get that 62 00:05:04.240 --> 00:05:09.279 from humans today. You cannot get a machine or an algorithm to provide you 63 00:05:09.399 --> 00:05:13.839 that feedback, which is why there's a lot of human in the loop intervention 64 00:05:13.959 --> 00:05:20.079 in the operation of AI the other and that became a theme because what it 65 00:05:20.160 --> 00:05:27.439 means is you need to think about how AI help human intelligence and how human 66 00:05:27.480 --> 00:05:32.000 intelligence helps machine intelligence or AI become a reality. So it's about kind of 67 00:05:32.000 --> 00:05:36.800 the collaboration with algorithms and robots and so forth. So that's how we define 68 00:05:36.839 --> 00:05:41.240 it. And then we had to go look for, Okay, what are 69 00:05:41.360 --> 00:05:46.240 areas that need focus where as an institute we can make a large contribution, 70 00:05:46.399 --> 00:05:53.240 we can become the leader in certain areas, and quickly identified responsible AI as 71 00:05:53.279 --> 00:05:57.519 one big area of focus from a research perspective, defining what it is, 72 00:05:57.720 --> 00:06:00.560 how do you do it, how do you do risk assessment, et cetera. 73 00:06:00.879 --> 00:06:05.560 And then we also identified kind of what I like to call new generation 74 00:06:05.959 --> 00:06:14.000 large language models or new generation generative AI. So how do you reduce this 75 00:06:14.079 --> 00:06:18.040 obsession with larger and larger parameters in the model. Larger models that can so 76 00:06:18.160 --> 00:06:23.879 easily go out of control, start making errors hallucinating as Google calls it, 77 00:06:25.319 --> 00:06:30.879 or basically are extremely expensive to train because the more parameters you have, the 78 00:06:30.920 --> 00:06:36.279 more and of course capital equipment wise, that the training infrastructure, and the 79 00:06:36.360 --> 00:06:43.560 inference is so rather than how can we make models bigger, the question is 80 00:06:43.720 --> 00:06:49.079 how do I get away with the smallest possible large model and enhance it by 81 00:06:49.120 --> 00:06:54.319 stuff that we know prior knowledge? Do I have a knowledge graph about the 82 00:06:54.360 --> 00:06:58.759 domain? Do I understand who the people are, professional graph, a citation 83 00:06:58.920 --> 00:07:02.279 graph, et cetera. It is silly to try to let the AI rediscover 84 00:07:02.759 --> 00:07:08.959 from pure data, very very expensive data, what we already know. So 85 00:07:09.040 --> 00:07:13.199 can we supplement this and we call this better together? And then we picked 86 00:07:13.399 --> 00:07:18.199 certain areas and we are seeing better results indeed qualitatively and qualitatively when we apply 87 00:07:18.360 --> 00:07:23.959 these two things together. Finally, we had to pick kind of areas of 88 00:07:24.000 --> 00:07:29.120 focus for applications, which is critical for us. So applications, in my 89 00:07:29.199 --> 00:07:32.759 opinion, have been the driver in the developments of AI for the past decade 90 00:07:32.839 --> 00:07:39.040 or more. So. We looked at applied areas in AI plus health, 91 00:07:39.600 --> 00:07:45.319 AI plus life sciences, drug discovery and molecule design and things like that, 92 00:07:46.000 --> 00:07:51.920 antibody design. And then the last area we added was AI for climate and 93 00:07:51.959 --> 00:07:57.480 sustainability. How do we leverage the new generations of data sets that are kind 94 00:07:57.480 --> 00:08:01.639 of hyper local, very detailed, huge big data, if you will, 95 00:08:01.680 --> 00:08:05.800 in the world of climate and then use it to model both the climate and 96 00:08:05.879 --> 00:08:11.519 the environment and predict things like here are the impacts of a potential disaster or 97 00:08:11.600 --> 00:08:16.759 a change or what have you. So these areas define where we work and 98 00:08:16.839 --> 00:08:20.160 to drive the wheel forward. To start the flywheel going, we created the 99 00:08:20.160 --> 00:08:26.079 AI Solutions Hub. One of the big funders of the institute is Dave Rue 100 00:08:26.120 --> 00:08:33.039 who funded also the Real Institute up in Portland, Maine, in order to 101 00:08:33.120 --> 00:08:35.799 change the economy in Maine, and AI in his vision, was a big 102 00:08:35.840 --> 00:08:39.080 area. So we created this AI Solutions Hub, which is a team, 103 00:08:39.120 --> 00:08:43.799 a delivery team of data scientists, data engineers, AI practitioners whose job is 104 00:08:45.200 --> 00:08:50.080 to engage with our partners companies, organizations, governments and work on real challenges 105 00:08:50.120 --> 00:08:54.000 in the real world in order to drive that activity, figure out what the 106 00:08:54.039 --> 00:09:01.000 real research agenda should be and at the same time figure out how do we 107 00:09:01.519 --> 00:09:07.240 train our students and the corporate learners on AI in the wild AI happening in 108 00:09:07.279 --> 00:09:11.960 reality? Right? What are the real issues? How do you deal with 109 00:09:11.000 --> 00:09:15.720 it? You know, we all know that dependence deep dependence of data and 110 00:09:15.759 --> 00:09:20.120 generative AI, of generative AI on data, but The real issue here is 111 00:09:20.240 --> 00:09:24.000 what do you do? And the data is not in idealized circumstances. It's 112 00:09:24.039 --> 00:09:28.080 got quality problems, half of it is missing, a third of what's there 113 00:09:28.120 --> 00:09:31.080 is wrong. But you got to make the solution work, and you've got 114 00:09:31.080 --> 00:09:35.360 to make it work in the constraints of kind of real life and real businesses 115 00:09:35.399 --> 00:09:39.720 and real regulations and so forth. So all of this coming together gives us, 116 00:09:39.720 --> 00:09:43.720 if you like, the active lab that generates the research agenda and generates 117 00:09:43.759 --> 00:09:48.440 the chance to train our students and learners from the corporations on how to make 118 00:09:48.519 --> 00:09:52.799 AI solutions work. Yeah, that's I really love the vision here because you're 119 00:09:52.840 --> 00:09:58.840 appreciating the importance of research, but also the importance of training, the importance 120 00:09:58.879 --> 00:10:03.440 of understanding use cases, hence the solution portfolio or the solution hub. And 121 00:10:03.480 --> 00:10:07.559 you know, I want to drive into one key issue here that you brought 122 00:10:07.600 --> 00:10:11.399 up, because I think it's the most important one to make this stuff enterprise 123 00:10:11.519 --> 00:10:16.000 ready and business ready and consumer friendly, which is the truth. Right. 124 00:10:16.000 --> 00:10:20.159 We have one of our regular listeners, Alex Husky, super smart guy works 125 00:10:20.159 --> 00:10:24.320 for rex On Mobile. He sent me an email. We did an webinar 126 00:10:24.360 --> 00:10:26.960 a few weeks ago he said, we cannot allow these large language models to 127 00:10:28.120 --> 00:10:31.919 pattern recognize their way to the truth. And that's kind of what they're doing 128 00:10:31.039 --> 00:10:37.399 is it's basically a pattern recognition engine that predicts text based upon the prompt and 129 00:10:37.440 --> 00:10:41.120 all the corpus of text that it has consumed. Well, you mentioned citations 130 00:10:41.159 --> 00:10:46.639 for example, and this whole concept of embeddings comes into play. And that's 131 00:10:46.679 --> 00:10:50.559 what you need, right, that's the ground truth our citations. This is 132 00:10:50.559 --> 00:10:54.840 where I got the information, so you can verify that it's correct. That's 133 00:10:54.840 --> 00:11:01.080 a cornerstone of making this stuff doable and workable for business, right absolutely. 134 00:11:01.600 --> 00:11:07.080 And you know, citations are a beautiful example. You know, this autocomplete 135 00:11:07.080 --> 00:11:11.879 technology that we call large language models, it's trying to you know, build 136 00:11:11.879 --> 00:11:16.279 its output token at a time, and the token is part of a word, 137 00:11:16.559 --> 00:11:18.799 part of a group of pixels, depending on what your output looks like. 138 00:11:20.360 --> 00:11:26.559 Well, trying to recollect a precise citation based on an autocomplete is a 139 00:11:26.679 --> 00:11:31.320 very dangerous exercise and rife with room for making an error. You can get 140 00:11:31.320 --> 00:11:33.360 the year wrong, the author's wrong, you can make up, you can 141 00:11:33.399 --> 00:11:37.240 come up with a title, you could change the title, all based on 142 00:11:37.279 --> 00:11:41.600 your recollection well, why do all that? If you already have these citations 143 00:11:41.639 --> 00:11:46.039 in very precise format, how can you use them? And then tying the 144 00:11:46.080 --> 00:11:50.679 citations to the content to the meaning is the other big challenge because most of 145 00:11:50.720 --> 00:11:54.840 these models, I mean we call them technically stochastic parrots because they really parrot 146 00:11:54.960 --> 00:11:58.200 out. They don't know anything, they don't understand what they're saying. So 147 00:11:58.320 --> 00:12:01.960 tying that to kind of the sources of truth, as you say, is 148 00:12:03.000 --> 00:12:07.720 another huge challenge which doesn't exist today. Well, and also we have to 149 00:12:07.759 --> 00:12:13.240 realize there is a workflow to these models, and what I'm learning from folks 150 00:12:13.240 --> 00:12:16.759 is you can do one of two things. Basically, you can fine tune 151 00:12:16.799 --> 00:12:20.639 your model by embedding your corporate data into the model, so training the model 152 00:12:20.679 --> 00:12:24.440 itself on your corporate data. Or the other very common use case is to 153 00:12:24.519 --> 00:12:30.320 use a vector database and that's where you store your embeddings, and so embeddings 154 00:12:30.360 --> 00:12:35.200 would be your financial reports for your business, or your training modules or documents 155 00:12:35.279 --> 00:12:39.279 related to some process in your business. You want to populate all that and 156 00:12:39.360 --> 00:12:46.399 ideally just use the large language model for its text, generative and syntactical knowledge 157 00:12:46.480 --> 00:12:50.120 essentially, so it's you're using the engine to spin up the words, but 158 00:12:50.240 --> 00:12:54.759 the facts, if you will, come from the embeddings. Isn't that the 159 00:12:54.080 --> 00:13:00.360 general idea? Yeah, and well, you're touching on some of the more 160 00:13:00.519 --> 00:13:03.200 interesting aspects of large language models. So let me start by saying, first 161 00:13:03.240 --> 00:13:11.879 of all, no one understands why they work. Of why a very large 162 00:13:11.919 --> 00:13:16.120 model that traditional statistics will tell you should never work because it's over parameterized, 163 00:13:16.519 --> 00:13:20.360 it's got too many degrees of freedom it can fit the data, and therefore 164 00:13:20.399 --> 00:13:24.200 no chance of learning why it works. That's one interesting mystery, right that 165 00:13:24.240 --> 00:13:28.799 we don't write. The other mystery is this whole embedding space. Why is 166 00:13:28.840 --> 00:13:33.440 it that if I took a group of words, or a group of sentences, 167 00:13:33.559 --> 00:13:37.519 or a group of paragraphs and map them into this very large vector space, 168 00:13:37.559 --> 00:13:41.559 which is what's called embedding a bunch of numbers. Basically, why is 169 00:13:41.600 --> 00:13:48.480 it that those embeddings capture something about the domain that allows you to generalize, 170 00:13:48.519 --> 00:13:52.919 that allows you to build own knowledge. So that representation is another mystery eric 171 00:13:54.000 --> 00:13:58.639 that we have people working on trying to demystify and understand what the heck is 172 00:13:58.679 --> 00:14:03.120 going on. Those two aspects are some of the deeper aspects of the technology 173 00:14:03.480 --> 00:14:09.200 that represent an academic and research challenge trying to understand what's going on and how 174 00:14:09.200 --> 00:14:11.879 do these models work. The fact is, though they work, and therefore 175 00:14:13.600 --> 00:14:16.879 how do you best leverage them? And then how do you add these embeddings 176 00:14:16.879 --> 00:14:22.000 with the other generative stuff which is straightforward and well understood. The autocomplete technology, 177 00:14:22.000 --> 00:14:26.840 if you like, that kind of takes an input a prompt and produces 178 00:14:26.879 --> 00:14:30.559 the output. The other thing that I'll mentioned here, which is the third 179 00:14:30.639 --> 00:14:35.679 big mystery in the field, is the prompt themselves. How do you come 180 00:14:35.759 --> 00:14:43.120 up with the right prompt. We know there's a super high sensitivity between prompt 181 00:14:43.240 --> 00:14:48.039 variations and output variations, so small disturbances of the prompt may produce a completely 182 00:14:48.039 --> 00:14:52.240 different answer. Some people claim that, hey, these are the new hot 183 00:14:52.360 --> 00:14:58.120 jobs. There was a New York Times and Financial Times articles on this is 184 00:14:58.159 --> 00:15:03.399 the highest paid jobs his job. My colleague and a core member of the 185 00:15:03.399 --> 00:15:11.240 faculty of the Institute for Experiential AI at Northeastern, Byron Wallace, summarized it 186 00:15:11.399 --> 00:15:16.879 very nicely in our recent conference where he said, this is not from the 187 00:15:16.919 --> 00:15:22.759 engineering, This is more like incantations and black magic, right, And you 188 00:15:22.759 --> 00:15:26.200 know, you might stumble on the right incantation, and you might not. 189 00:15:28.519 --> 00:15:31.399 And he has a field of study that has to do with, Okay, 190 00:15:31.480 --> 00:15:35.399 how do we reduce that dependence, how do we reduce that sensitivity, because 191 00:15:35.399 --> 00:15:37.360 at the end of the day, to make it practical and useful, you 192 00:15:37.399 --> 00:15:43.159 can't have this huge variation and kind of weird stuff. And he says incantations 193 00:15:43.200 --> 00:15:50.440 because sometimes prefixes to prompts that you might you and I might think have nothing 194 00:15:50.480 --> 00:15:54.480 to do with it end up being very useful to guide the model to the 195 00:15:54.559 --> 00:15:58.600 right space. So it's people are adding weird stuff in the front and in 196 00:15:58.639 --> 00:16:02.080 the end of a prompt to make it work right. That is not a 197 00:16:02.120 --> 00:16:04.960 science, That is not engineering. We got to figure this out. That's 198 00:16:06.000 --> 00:16:08.039 funny, you know. I'm gonna close this opening segment here with a fun 199 00:16:08.639 --> 00:16:15.960 story and basically say that this is not as crazy and different than unique as 200 00:16:15.960 --> 00:16:18.879 some people might think. So I had a business partner for a number of 201 00:16:18.960 --> 00:16:22.519 years back in the year two thousand and I think earlier you were talking twenty 202 00:16:22.519 --> 00:16:25.240 men twenty twenty, I think when they started this whole thing. So it's 203 00:16:25.279 --> 00:16:27.559 five years ago. But back in two thousand, I was working for a 204 00:16:27.600 --> 00:16:32.039 company. It came gentled to call him a couple Russians, and the good 205 00:16:32.120 --> 00:16:33.799 Russian was named Nicola, like Nikol, let me, you gods. One 206 00:16:33.879 --> 00:16:37.519 thing he said to me, The funny thing about the web is that nobody 207 00:16:37.600 --> 00:16:42.720 know how it works. All we know is it works right, and it's 208 00:16:42.720 --> 00:16:48.240 the same thing. So now what's interesting is if you look at Internet and 209 00:16:48.399 --> 00:16:52.000 cloud and hybrid architecture and all this stuff that people are talking about today, 210 00:16:52.000 --> 00:16:56.799 containerization, Kubernetes, all that stuff, Well, what happened is Google came 211 00:16:56.840 --> 00:17:00.399 along with Kubernetes as this orchestration layer to sit on top of Docker as a 212 00:17:00.440 --> 00:17:04.400 way to containerize applications make them really really small, which is like the vision 213 00:17:04.400 --> 00:17:07.759 of service storrates at architecture, but a different bus, if you will. 214 00:17:08.279 --> 00:17:11.319 And what's happening is now we have all this data. We have tons and 215 00:17:11.359 --> 00:17:15.680 tons. It's called observability. So what's happening now is we're finally starting to 216 00:17:15.759 --> 00:17:19.200 kind of figure out how the web actually works. So I'm not saying we 217 00:17:19.240 --> 00:17:22.319 didn't know anything, but there are a lot of stuff that we didn't really 218 00:17:22.359 --> 00:17:27.039 fully understand about how the web actually works, like which ports are important in 219 00:17:27.039 --> 00:17:32.000 what particular process. I mean there's all this stuff going on. So my 220 00:17:32.079 --> 00:17:34.119 point view is that this AI stuff, You're right, we don't know how 221 00:17:34.160 --> 00:17:38.079 it works. It's not the first time we've found some engine that works and 222 00:17:38.079 --> 00:17:41.079 we're like, well, we don't know exactly how it works, but it 223 00:17:41.079 --> 00:17:44.000 sure does work. But folks, don't touch that dollar. We'll be right 224 00:17:44.039 --> 00:17:48.559 back talking to Osama Faiad of the AI Institute from the Experiential AI Institute. 225 00:17:48.559 --> 00:18:00.279 Will be right back. Expect Welcome back to Inside Analysis. Here's your host 226 00:18:00.839 --> 00:18:07.240 Eric Tavanaugh. All right, folks, back here on Inside Analysis, your 227 00:18:07.240 --> 00:18:12.359 host Eric Kavanaugh with doctor u Sama Fayad from Northeastern Universities, the Institute for 228 00:18:12.400 --> 00:18:18.079 Experiential AI, which basically just means humans in the loop. Humans are really 229 00:18:18.079 --> 00:18:22.200 important. Even with self driving cars, you have to give correction to the 230 00:18:22.319 --> 00:18:25.519 vehicle. I've actually driven a couple of these now, and it's kind of 231 00:18:25.559 --> 00:18:27.839 freaky when it starts taking the wheel from you and like, whoa, what 232 00:18:29.079 --> 00:18:30.759 are you doing. I didn't even know it was self driving. It was 233 00:18:30.759 --> 00:18:33.359 a rental car. I'm like, oh, what's going on here? Okay, 234 00:18:33.359 --> 00:18:37.400 it's taking over for me and seeing you shouldn't be doing something. We're 235 00:18:37.400 --> 00:18:41.599 getting there it's I mean, thanks to innovations, thanks to Elon Musk and 236 00:18:41.640 --> 00:18:47.359 Tesla and all these innovations, those folks are doing pretty cool stuff. But 237 00:18:47.480 --> 00:18:52.000 doctor Fayad, we wanted to dive into the critical success factors, and one 238 00:18:52.039 --> 00:18:56.359 absolutely positively is the data. You know, I keep cautioning people and we're 239 00:18:56.359 --> 00:19:00.920 going to probably dive into this more in our own business, this of consulting 240 00:19:00.920 --> 00:19:04.079 with organizations. Be very careful how you train your models. You want to 241 00:19:04.119 --> 00:19:07.160 train your model on your data. You want to give it curated, trusted 242 00:19:07.240 --> 00:19:11.519 data. You don't want to just point it at your whatever information system you 243 00:19:11.599 --> 00:19:15.000 have, because there's a bunch of junk in there. And anyone who's ever 244 00:19:15.079 --> 00:19:18.079 done data quality initiative knows. If you get an executive who doesn't believe the 245 00:19:18.160 --> 00:19:22.200 data is crap, just show them. Just open up any database and point 246 00:19:22.200 --> 00:19:25.599 to any corner of it and you'll find some problem somewhere. I promise. 247 00:19:25.640 --> 00:19:27.599 It's just everywhere, whether it's data in the wrong fields, data in the 248 00:19:27.640 --> 00:19:33.079 wrong format, data that was truncated during some ETL script or something, and 249 00:19:33.079 --> 00:19:36.880 it's just everywhere. There's bad data everywhere. So getting the right data is 250 00:19:36.960 --> 00:19:41.160 really important, Doctor Fiab how do you do that? What do you recommend 251 00:19:41.400 --> 00:19:44.440 of When people come to you and ask you what are the best practices, 252 00:19:44.480 --> 00:19:47.440 what do you tell them? Yeah? Great, great question. So so 253 00:19:47.599 --> 00:19:51.799 data is a whole ocean to open up. I mean, there's the collecting 254 00:19:51.839 --> 00:19:56.200 the data at quality, that's collecting data at the right level of granularity. 255 00:19:56.559 --> 00:20:00.000 So we have spent decades, as you will know, Eric, in the 256 00:20:00.000 --> 00:20:04.640 whole area of data warehousing and kind of information systems management and so forth, 257 00:20:04.680 --> 00:20:12.400 and data management thinking about data for use by humans. Now what the AI 258 00:20:12.559 --> 00:20:18.839 haves. The AI first companies or the companies who really are using AI today, 259 00:20:18.240 --> 00:20:22.039 they have a deeper understanding that, hey, machines needed data at a 260 00:20:22.079 --> 00:20:27.079 different level. They need high granularity. Not necessarily stuff understandable by humans, 261 00:20:27.319 --> 00:20:33.119 but every little detail of what happened, every log of every event, every 262 00:20:33.319 --> 00:20:37.880 kind of intervention, correction by a human, et cetera. Captured. This 263 00:20:38.039 --> 00:20:42.720 is not how most organizations or companies work today. So that is one big 264 00:20:42.759 --> 00:20:47.559 shift that needs to happen. The second thing is, if you notice a 265 00:20:47.559 --> 00:20:51.519 lot of the literature and lower in the world of data has to do with 266 00:20:51.559 --> 00:20:56.279 structured data. Well, according to Gardner and many others, the majority of 267 00:20:56.359 --> 00:21:00.119 data in any organization, in any company is actually unstructured data. If you 268 00:21:00.240 --> 00:21:06.119 actually talk to the IT teams and the technology teams and so forth, they 269 00:21:06.200 --> 00:21:08.599 only typically only know how to deal with structured data, and then they hit 270 00:21:08.680 --> 00:21:12.960 unstructured data, they're stuck. It's like in the headlights. We have to 271 00:21:14.119 --> 00:21:18.240 change this dynamic to understand that, hey, we got to embrace the unstructured 272 00:21:18.319 --> 00:21:25.039 data, the text, the images, the video, the audio interactions in 273 00:21:25.079 --> 00:21:30.279 the customer center, all the conversations and leverage that information properly. And that 274 00:21:30.400 --> 00:21:34.359 needs that. There exists many good tools out there. Most of them also 275 00:21:34.480 --> 00:21:37.559 fortunately are in the open source, which is something I'm very big fan of. 276 00:21:38.039 --> 00:21:41.559 But you can put these things together in solutions that work for businesses, 277 00:21:41.880 --> 00:21:45.599 and this is something we're very passionate about. How do we partner to figure 278 00:21:45.599 --> 00:21:49.200 out, how we help our partners figure out how to approach an area, 279 00:21:49.240 --> 00:21:53.519 how to capture the data, etc. Now that leads me to another very 280 00:21:53.640 --> 00:22:00.839 very important data that's information or aspect that's very related to data, which is 281 00:22:00.400 --> 00:22:04.839 kind of many companies out there think that all we have to do is wait 282 00:22:04.920 --> 00:22:11.640 for Microsoft or open ai or whoever to refine their models so that they can 283 00:22:11.680 --> 00:22:15.960 meet our needs. What happens then, is you become users of the technology. 284 00:22:17.240 --> 00:22:21.839 You don't have your own private language model. You don't have the ability 285 00:22:22.200 --> 00:22:26.599 to capture all these corrections. You mentioned corrections when you're driving right. Well, 286 00:22:26.640 --> 00:22:30.079 it's great to intervene and correct, but if you don't capture that correction, 287 00:22:30.279 --> 00:22:33.720 the context for that correction, why you made that correction, your model 288 00:22:33.799 --> 00:22:37.759 is never going to benefit from it unless you're willing to share it back with 289 00:22:38.400 --> 00:22:44.480 the company who's providing this public utility, this large language model. Often this 290 00:22:44.559 --> 00:22:48.519 information is sensitive, it's proprietary. I argue it's one of the most valuable 291 00:22:48.519 --> 00:22:53.599 sources of IP that any company operating in any space has, which is kind 292 00:22:53.599 --> 00:22:57.720 of the secret sauce. How do our processes work? Where do we intervene, 293 00:22:57.720 --> 00:23:02.519 Why do we fix what is way of doing things today? Companies let 294 00:23:02.599 --> 00:23:06.759 all of that go into the exhaust by never capturing it. In reality, 295 00:23:06.920 --> 00:23:10.039 you need to be able to not just capture it, but feed it back. 296 00:23:10.200 --> 00:23:14.519 And feeding it back leads us to another whole interesting area, which means 297 00:23:14.720 --> 00:23:19.480 you really need to have your own private language model. That model can't be 298 00:23:19.559 --> 00:23:25.160 too huge, like the likes of open AI or CHG, GPT or any 299 00:23:25.200 --> 00:23:29.000 of these. It has to be as small as possible to maintain its stability, 300 00:23:29.400 --> 00:23:36.359 to maintain its maintainability and its reliability and so forth. But it needs 301 00:23:36.400 --> 00:23:38.720 to get feedback. So you need to be able to capture all that feedback, 302 00:23:38.720 --> 00:23:44.480 all those corrections, and feed them right back and retrain that language model 303 00:23:44.559 --> 00:23:48.720 in an affordable way to maintain your own IP, your own knowledge and so 304 00:23:48.799 --> 00:23:53.440 forth. Now, many of the companies who offer the large language models will 305 00:23:53.480 --> 00:23:59.359 offer you the chance, of course, for money to kind of fine tune 306 00:23:59.480 --> 00:24:03.920 the model or create your own a tuned version of the model, and there 307 00:24:03.960 --> 00:24:07.400 we can talk about technically what happens. All they do is they cut off 308 00:24:07.480 --> 00:24:11.039 the output layer, they throw it out, and retrain an output layer they 309 00:24:11.039 --> 00:24:17.119 call these adapters that is kind of fine tuned just to your problem and on 310 00:24:17.160 --> 00:24:22.880 your feedback. But we know that that's less effective than retraining the model and 311 00:24:22.920 --> 00:24:26.960 reintegrating that data into the whole stack. So coming up with our large language 312 00:24:27.000 --> 00:24:30.720 model, your own private language model, is not impossible. It is very 313 00:24:30.759 --> 00:24:36.559 doable. Many tools are available out there, whether you're talking Falcon which came 314 00:24:36.599 --> 00:24:40.920 from Abu Dhabi or Lama or Lama too. I'm a big fan of at 315 00:24:40.920 --> 00:24:45.519 Facebook or Meta because they actually not only did they make it open source, 316 00:24:45.559 --> 00:24:49.880 they gave a commercial open source license a company I saw, Yeah, and 317 00:24:49.920 --> 00:24:55.519 that opens up a whole other approach and a whole other mechanism. But to 318 00:24:55.599 --> 00:24:57.279 do that and to take advantage of it, you need to have the talent 319 00:24:57.400 --> 00:25:00.480 and then know how of how do I use that language model? When do 320 00:25:00.559 --> 00:25:03.079 I find tune it? How large does it have to be? How small 321 00:25:03.079 --> 00:25:07.279 can I make it? All of these questions are the kinds of questions we 322 00:25:07.440 --> 00:25:12.599 partner with companies on kind of answering and figure out what is the right approach 323 00:25:12.640 --> 00:25:18.440 here in this very very confusing, and I might add fast evolving space. 324 00:25:18.480 --> 00:25:22.200 I mean I try to keep up and I am challenged. Right every week, 325 00:25:22.480 --> 00:25:27.359 every month there's new innovations, new ways of doing things. Somebody's crapping 326 00:25:27.400 --> 00:25:30.799 something old and saying, ah, it didn't work. Actually it was part 327 00:25:30.799 --> 00:25:34.319 of the system, but now we've got to do something completely new. All 328 00:25:34.440 --> 00:25:40.519 good, all healthy, but it's a challenge to keep up with the latest, 329 00:25:40.599 --> 00:25:44.039 greatest and the right best practice of how to use it. Yah. 330 00:25:44.400 --> 00:25:48.119 Start with, I just want to remind everybody that data that you're letting go 331 00:25:48.200 --> 00:25:52.880 into the exhaust by not capturing it systematically. That is gold that is actually 332 00:25:52.880 --> 00:25:56.640 more valuable than gold. It's not just oil. It's more valuable than gold, 333 00:25:56.799 --> 00:26:00.599 and you're letting it go out in the exhaust and losing all that ip 334 00:26:02.160 --> 00:26:04.400 that your teams, your employees are working on day in, day out and 335 00:26:04.519 --> 00:26:07.480 keeping operations going. Yeah, well, you know, I just as you 336 00:26:07.519 --> 00:26:11.599 were talking, I thought of an interesting metaphor here, and I'll throw it 337 00:26:11.599 --> 00:26:15.039 at you. I think you'll you'll find it interesting. So think about language 338 00:26:15.039 --> 00:26:19.240 and feeding a dictionary of English to something like one of these large language models. 339 00:26:19.279 --> 00:26:22.640 You feed it. The dictionary knows all the different words, but people 340 00:26:22.680 --> 00:26:27.160 don't talk the way they write, and people don't talk the way dictionaries are 341 00:26:27.160 --> 00:26:30.440 written. Necessarily, there's a lot of slang. There are lots of expressions, 342 00:26:30.480 --> 00:26:33.920 idioms, all these different things that don't fit neatly into the definitions of 343 00:26:33.960 --> 00:26:37.960 words. But it's how you actually speak when you're talking to someone. And 344 00:26:38.000 --> 00:26:44.160 that's the unstructured data, really, or that's the unstructured syntax almost of how 345 00:26:44.200 --> 00:26:48.480 people actually interact with words on the phone in person, and certain cultures and 346 00:26:48.519 --> 00:26:53.039 in certain regions, and so if you don't absorb that that contextual information, 347 00:26:53.559 --> 00:26:56.839 your engine will never speak like a normal person does. Right, So this 348 00:26:56.920 --> 00:27:02.279 is the nuance you have to appreciate sheet, is that what's on paper often 349 00:27:02.359 --> 00:27:06.000 is very different in the spoken word. What do you think? Absolutely, 350 00:27:06.039 --> 00:27:08.480 and it's critical in many applications, actually, I would argue in most of 351 00:27:08.519 --> 00:27:12.640 them, if you're dealing with conversational AI, if you're dealing with dealing with 352 00:27:12.799 --> 00:27:17.319 humans, whether you're doing it for marketing purposes, whether you're doing it for 353 00:27:17.440 --> 00:27:23.359 customer service, whether you're doing it for predictive maintenance. And it's not only 354 00:27:23.480 --> 00:27:26.720 kind of the idiom or kind of what the public how the public or the 355 00:27:26.759 --> 00:27:32.440 customer speed. It's often the culture within the company. Your engineering team has 356 00:27:32.480 --> 00:27:36.400 their own terminology where they use certain terms. If you come up in the 357 00:27:36.440 --> 00:27:41.000 right, they mean something, but they're using the company to mean a certain 358 00:27:41.039 --> 00:27:45.279 aspect of a process, a certain aspect of a system. Learning all of 359 00:27:45.319 --> 00:27:49.319 that and making it work in the right context is a real interesting challenge, 360 00:27:49.799 --> 00:27:53.279 and it's very possible. I mean, that's why the data is, That's 361 00:27:53.279 --> 00:27:57.240 why I describe it as gold or more valuable than gold, because it contains 362 00:27:57.319 --> 00:28:03.880 all these instances and gives you the keys to allowing these systems to have the 363 00:28:03.960 --> 00:28:07.359 bigger and correct impact on how you do acceleration, how you do automation, 364 00:28:08.000 --> 00:28:15.039 how do you leverage generative AI to kind of give you that unique advantage competitive 365 00:28:15.079 --> 00:28:22.960 advantage out there. I've experienced it personally in healthcare. Right if you're in 366 00:28:22.000 --> 00:28:26.000 an operating room or in an X ray room or in an imaging center, 367 00:28:26.480 --> 00:28:30.319 you know, the language being used is completely different than what's being used in 368 00:28:34.079 --> 00:28:41.759 you know, a pathology lab or a practice for physical therapy or what have 369 00:28:41.880 --> 00:28:45.240 you. So all of these things, all of these aspects are important. 370 00:28:45.319 --> 00:28:48.799 What contents to use, the words in, what they mean, new meanings 371 00:28:48.799 --> 00:28:52.519 that are not available in the dictionary. All of that is so important and 372 00:28:52.559 --> 00:28:59.559 so critical technically, socially and marketing wides. Yeah. Well, and there's 373 00:28:59.599 --> 00:29:03.720 also tone of voice. And I remember, going all the way back to 374 00:29:03.960 --> 00:29:07.799 like third grade or something, our teacher asked us a question and it stumped 375 00:29:07.839 --> 00:29:08.559 all of us, and I thought about it later, I was like, 376 00:29:08.640 --> 00:29:11.920 Oh, tone of voice, that's what he was talking about. You can 377 00:29:12.039 --> 00:29:17.000 say something and meet it honestly, or you can say it sarcastically, and 378 00:29:17.079 --> 00:29:19.079 usually it's the tone of voice and the context that will tip off a human 379 00:29:19.119 --> 00:29:23.119 being. But it's gonna be harder for a machine to understand that, but 380 00:29:23.160 --> 00:29:27.200 not if it has the intonation that if you capture the audio and it can 381 00:29:27.279 --> 00:29:30.519 kind of pick up on that sort of thing. But these are the sort 382 00:29:30.599 --> 00:29:36.599 of deeply understood aspects of communication that we just know as grown ups, that 383 00:29:36.640 --> 00:29:40.119 we've learned over time. I mean, I remember, this is funny, 384 00:29:40.119 --> 00:29:42.720 but probably most people don't remember this, but I remember the first time I 385 00:29:42.759 --> 00:29:48.839 really experienced sarcasm because I was like two or three or something and I'd spill 386 00:29:48.920 --> 00:29:51.240 milk and my mother and I knew it was a problem. Was like, 387 00:29:51.240 --> 00:29:52.799 oh, no, this is bad. And my mother looked at me. 388 00:29:53.079 --> 00:29:56.640 She goes, well, that's just great, And I was like, hold 389 00:29:56.680 --> 00:30:02.400 on, I know the great is good, but that tone is not good. 390 00:30:02.960 --> 00:30:06.559 What's happening here? And I was really defuddled as a child. But 391 00:30:06.680 --> 00:30:11.160 that was my first experience with sarcasm. Right, I'll give you here. 392 00:30:11.240 --> 00:30:15.240 You're touching on one of the biggest challenges that, for example, search engines 393 00:30:15.240 --> 00:30:18.599 today face, which are memes. Right, you hit a meme on the 394 00:30:18.640 --> 00:30:22.400 Internet, the search engine thinks it means one thing. Reality is, it's 395 00:30:22.440 --> 00:30:26.720 something completely different, and the only way you catch that, the only way 396 00:30:26.720 --> 00:30:32.279 you catch that today is human intervention. Search editorial team saying, wait a 397 00:30:32.359 --> 00:30:33.839 second, this shouldn't be in your search result. This is a meme. 398 00:30:34.079 --> 00:30:38.200 It means something completely different. It has nothing to do with biology or whatever. 399 00:30:38.359 --> 00:30:41.400 It's being used to refer to a group of people or a group of 400 00:30:41.440 --> 00:30:45.640 practices or what have you. So those are those are real things that matter 401 00:30:45.720 --> 00:30:51.119 a lot, and you have to have mechanisms for catching them. And you 402 00:30:51.160 --> 00:30:53.079 know, in a way, you can argue that a lot of the internal 403 00:30:53.119 --> 00:30:57.880 culture of a company's operations team, a company's engineering team, a company's business 404 00:30:57.920 --> 00:31:02.799 team, a company's sales team, a company's marketing team is filled with these 405 00:31:02.839 --> 00:31:06.200 memes. We don't call them memes, but they are there. That's funny. 406 00:31:06.359 --> 00:31:10.759 There's actually a great Seinfeld episode where George Costanza, he's working for the 407 00:31:10.839 --> 00:31:15.200 Yankees, and I think he was talking with it was like the Texas Rangers 408 00:31:15.279 --> 00:31:19.240 or something, and as a baseball team, and the culture of the executives 409 00:31:19.240 --> 00:31:23.480 for that team loved using curse words, so they were insulting each other all 410 00:31:23.480 --> 00:31:29.400 the time, but as a compliment and then George's boss walks by as he's 411 00:31:29.440 --> 00:31:32.839 on the phone using all these expletives, and the boss is like, oh 412 00:31:32.920 --> 00:31:34.079 no, what are you doing. This is terrible. In reality, he's 413 00:31:34.079 --> 00:31:38.519 bonding with these guys because the context was there. Classic dramatic irony, right, 414 00:31:38.559 --> 00:31:42.160 the one guy didn't know the context. But this is the importance of 415 00:31:42.279 --> 00:31:47.359 nuanced and understanding context. And to your point, just getting it back to 416 00:31:47.440 --> 00:31:52.279 training AI your own private language model, which is what most businesses are going 417 00:31:52.359 --> 00:31:56.519 to want to have. I'm pretty sure you have to bake in that kind 418 00:31:56.559 --> 00:32:00.799 of stuff. You have to have that human in the loop that goes okay, 419 00:32:00.839 --> 00:32:04.039 no, this is an example of sarcasm where these guys are just being 420 00:32:04.079 --> 00:32:07.200 goofy or whatever in their context. Calling you a jerk or a so and 421 00:32:07.279 --> 00:32:12.640 so is actually a compliment. That's the human invention part, right, absolutely 422 00:32:13.119 --> 00:32:19.400 absolutely, and I remember that episode. But the thing I wanted to mention 423 00:32:19.640 --> 00:32:23.720 is, you know, language, the language use is very different than kind 424 00:32:23.720 --> 00:32:30.079 of the formal language defined. And that is why I believe, like you, 425 00:32:30.160 --> 00:32:35.680 most companies will find out that achieve true competitive advantage. To have differentiation, 426 00:32:36.200 --> 00:32:38.680 you're going to have your own private language models, which opens up a 427 00:32:38.680 --> 00:32:43.880 whole new area. We are a company that's in the you know, quick 428 00:32:43.920 --> 00:32:47.440 service restaurant business, or we are a company that's doing manufacturing, what do 429 00:32:47.480 --> 00:32:52.000 we know about large language model AI, et cetera, and how do we 430 00:32:52.039 --> 00:32:55.000 approach it? And part of this demystification and kind of working with you on 431 00:32:55.039 --> 00:33:00.079 the AI journey is one thing our institute is very past it about because that 432 00:33:00.160 --> 00:33:06.279 actually puts us right up front on the front lines of getting the applications to 433 00:33:06.359 --> 00:33:08.839 work and helping our partners getting it to work. And through that we will 434 00:33:08.920 --> 00:33:14.519 learn what the real research problems are, where the real challenges are, and 435 00:33:14.599 --> 00:33:19.200 we will learn what should we be teaching our students and our corporate learners in 436 00:33:19.319 --> 00:33:23.039 order to be effective in that world. So that whole area of the infrastructure, 437 00:33:23.400 --> 00:33:29.119 the talent, the know how, you will need to build your own 438 00:33:29.279 --> 00:33:32.200 private, private language models to capture your own IP and your own knowledge. 439 00:33:32.640 --> 00:33:37.200 Right, it's going to be a whole huge new trend that no one out 440 00:33:37.200 --> 00:33:40.920 there is paying attention to in the business. They're focused on being of these 441 00:33:43.240 --> 00:33:44.960 right, No, that's the very very good point. Well, folks, 442 00:33:45.000 --> 00:33:46.799 we're talking to doctor Usama Fai will be right back don't touch that dout. 443 00:33:46.799 --> 00:33:59.880 You're listening to Inside Analysis. Welcome back to Inside Analysis. Here's your host, 444 00:34:00.079 --> 00:34:06.240 means Eric Tavanaugh. All right, folks, back here on Inside Analysis 445 00:34:06.240 --> 00:34:10.440 talking all about experiential AI. That's humans in the Loop with doctor Usama Fayad 446 00:34:12.079 --> 00:34:17.400 of the Northeastern University Institute for Experiential AI. Got to get the all clean 447 00:34:17.440 --> 00:34:21.760 and straight in my head. The Institute for Experiential AI, it just means 448 00:34:21.840 --> 00:34:24.519 humans in the loop basically. But we're talking about the power of these models, 449 00:34:24.559 --> 00:34:28.719 which is amazing. We're talking about how a lot of people don't really 450 00:34:28.719 --> 00:34:30.440 know how they work, including people who design them. We just see that 451 00:34:30.480 --> 00:34:35.920 they work, and so we're coming up with strategies to help companies really harness 452 00:34:35.920 --> 00:34:37.320 the technology. And I think that's the key, right is this is if 453 00:34:37.440 --> 00:34:42.320 unbridled, this thing will go wacko and cause you all kind of trouble. 454 00:34:42.360 --> 00:34:45.719 But if you harness it effectively, and that's your governance structure, it's your 455 00:34:45.719 --> 00:34:50.880 embeddings. It's using ontologies, and I mentioned my friends from Praxy Data who 456 00:34:50.920 --> 00:34:55.639 are using ontologies to really narrow the focus of their work and thus be able 457 00:34:55.679 --> 00:35:00.039 to deliver some tangible business value. And you know that's always been the case 458 00:35:00.039 --> 00:35:04.960 with machine learning and AI is you have to have a very narrow focus and 459 00:35:05.000 --> 00:35:07.079 you have to know what you're trying to accomplish. I mean, you can 460 00:35:07.159 --> 00:35:12.280 use deep learning to discover things, but even then you're trying to discover things. 461 00:35:12.320 --> 00:35:15.880 So you have to have a purpose and a mission and an objective upon 462 00:35:15.960 --> 00:35:20.920 which you train these models to be able to understand if you're getting anywhere right. 463 00:35:20.920 --> 00:35:24.360 Tell us a bit about that and why ontologies matter, Yosama, Well, 464 00:35:24.400 --> 00:35:30.840 I mean, ontologies give you the basics of understanding domains. There are 465 00:35:30.880 --> 00:35:35.199 many things you have to worry about. And by the way, not coincidentally, 466 00:35:36.119 --> 00:35:40.320 a lot of the work in generative AI now is using what we call 467 00:35:40.400 --> 00:35:47.079 the RAG framework, which stands for retrieval augmented generation, which is another form 468 00:35:47.519 --> 00:35:53.079 of kind of enforcing the model to stay within certain guardrails and to say here's 469 00:35:53.119 --> 00:35:58.679 what we mean. And it's aided by kind of retrieving the right references and 470 00:35:58.719 --> 00:36:01.679 all that to help it from straying, both in training and in answering. 471 00:36:02.039 --> 00:36:08.400 But anyway, you know, working in different domains, you need to figure 472 00:36:08.400 --> 00:36:15.519 out ways of bringing in that domain knowledge that domain into the problem in order 473 00:36:15.559 --> 00:36:19.400 to constrain the space in order to focus it. You know, I was 474 00:36:19.400 --> 00:36:22.920 talking about the secrets of making AI work, and we talked about the humans 475 00:36:22.920 --> 00:36:27.519 in the loop or must interventions are necessary? Data at the right level is 476 00:36:27.559 --> 00:36:32.000 absolutely necessary. Most data is being lost. But I will I will add 477 00:36:34.920 --> 00:36:39.039 the last bit is how do we bring in non domain knowledge and how do 478 00:36:39.079 --> 00:36:45.480 we narrow the problem down to a very narrow, very specific problem. Because 479 00:36:45.519 --> 00:36:52.559 trying to solve AI in general has proven to be an impossible task, a 480 00:36:52.679 --> 00:36:57.519 very difficult task for the past seven decades, and it will continue to be 481 00:36:57.639 --> 00:37:02.920 that, and our best prescription for fixing it is by narrowing the problem and 482 00:37:04.000 --> 00:37:07.920 resorting to what we call narrow AI versus general AI. General AI is still 483 00:37:08.000 --> 00:37:12.880 far out of reach. Narrow AI has worked and has been working for many 484 00:37:14.320 --> 00:37:19.400 years and decades actually in different domains. And the key there is domain knowledge. 485 00:37:19.440 --> 00:37:22.880 The key there is narrowing it enough so you know that not every possibility 486 00:37:23.000 --> 00:37:25.360 is here, and you don't have to worry about every language, and you 487 00:37:25.360 --> 00:37:30.199 don't have to worry about every variation and every meaning. No, restrict yourself 488 00:37:30.239 --> 00:37:35.119 to this domain. Here's what things mean. Try to get complete knowledge around 489 00:37:35.119 --> 00:37:39.199 that domain. And that's a secret of making it work that only comes by 490 00:37:39.199 --> 00:37:44.480 figuring out what are the use cases, what are the domains of applications, 491 00:37:44.480 --> 00:37:47.159 and how do I come up with a methodology that allows me to build solutions 492 00:37:47.559 --> 00:37:51.880 for each specific use case, Which is why we kind of focused on things 493 00:37:51.960 --> 00:37:58.199 like AI plus health. How do we get healthcare applications into AI? How 494 00:37:58.239 --> 00:38:01.639 do we leverage digital health? You know, we live in a world today. 495 00:38:01.719 --> 00:38:07.480 Think about at the highest level, you go see your doctor, hopefully 496 00:38:07.480 --> 00:38:10.000 once a year, maybe twice a year, right only when something is wrong 497 00:38:10.480 --> 00:38:15.480 other than the routine visits. And what happens between these routine visits is a 498 00:38:15.480 --> 00:38:21.599 total mystery to the provider, to your physician, or because who knows what 499 00:38:21.760 --> 00:38:24.400 happened to Eric during this one year. Now Eric needs to fill me in, 500 00:38:24.800 --> 00:38:29.119 and who knows whether Eric can fill me in completely or skip a whole 501 00:38:29.159 --> 00:38:31.360 bunch of important things. Well, guess what. We live in a world 502 00:38:31.440 --> 00:38:36.039 now where there's a lot of instruments. We've got these watches, we've got 503 00:38:36.039 --> 00:38:39.000 the mobile phones, we've got the you know, all sorts of devices, 504 00:38:39.039 --> 00:38:44.000 cameras, other things in our homes. They can actually capture a lot of 505 00:38:44.000 --> 00:38:47.039 the information that typically is only captured in a clinic. So how do we 506 00:38:47.159 --> 00:38:51.480 leverage that to close the loop at the highest level. Now we call that 507 00:38:51.559 --> 00:38:55.960 digital health. Another form of digital health is things like oncology, or actually 508 00:38:57.280 --> 00:39:04.280 more generally pathology. How do we bring these digital techniques down to a pathology 509 00:39:04.360 --> 00:39:07.559 lab where somebody is taking a sample, looking at it in a microscope and 510 00:39:07.639 --> 00:39:13.400 kind of storing it in this glass slide. Well, guess what digitization can 511 00:39:13.440 --> 00:39:16.480 help a lot. You can look at it with instruments and with algorithms instead 512 00:39:16.519 --> 00:39:21.719 of just human eyes, instead of just relying on staff to do you can 513 00:39:21.719 --> 00:39:27.519 strike capturing what's happening in the heads of these very talented pathologists who are looking 514 00:39:27.559 --> 00:39:31.039 at an image and concluding a result. This happened in radiology in a big 515 00:39:31.079 --> 00:39:36.280 way. Radiology has been digitized. It benefits from a lot of these things. 516 00:39:36.639 --> 00:39:39.519 In pathology, it hasn't happened. Pathology is still practiced like it was, 517 00:39:39.599 --> 00:39:43.519 you know, three or four decades ago. It can benefit from a 518 00:39:43.519 --> 00:39:45.559 lot of digitization, which brings in a lot of AI and a lot of 519 00:39:45.679 --> 00:39:52.440 use of these things. In life sciences, we have many use cases, 520 00:39:52.519 --> 00:39:58.360 whether it's the acceleration of drug discovery, how do you design how do you 521 00:39:58.400 --> 00:40:04.280 accelerate the design of the next antibody? Right, antibodies are very effective ways 522 00:40:04.320 --> 00:40:09.039 of coming up with cures, vaccines, ways to mitigate disease spread. But 523 00:40:09.159 --> 00:40:14.440 it typically takes about three years to come up with the right antibody for something. 524 00:40:14.519 --> 00:40:17.960 So it wasn't a viable solution for COVID back in twenty twenty or twenty 525 00:40:17.960 --> 00:40:21.960 twenty one because it took three years. But if I can get that three 526 00:40:22.000 --> 00:40:27.280 years down to three months by accelerating it through compute and through modeling, amazing 527 00:40:27.320 --> 00:40:32.559 things can happen. Right, take things like that happen in a hospital or 528 00:40:32.599 --> 00:40:37.280 a physician when you're being taken care of. In one of the projects we 529 00:40:37.320 --> 00:40:44.440 have is called heart, which expands to something involving AI. But what it 530 00:40:44.519 --> 00:40:49.280 is one of the applications and working with a large hospital system, is how 531 00:40:49.320 --> 00:40:52.119 do you determine when somebody should not be downgraded from critical care? They're in 532 00:40:52.119 --> 00:40:57.760 the ICU and things look good, or we can downgrade them, but the 533 00:40:57.840 --> 00:41:00.280 algorithms can tell you no, no, no, no, there's a huge 534 00:41:00.360 --> 00:41:04.519 danger here. If you downgrade that issues would recur, which is horrible for 535 00:41:04.559 --> 00:41:07.920 the patient. Horrible for the hospital, horrible for the physician, And you 536 00:41:08.039 --> 00:41:13.400 want to make that decision when you're absolutely sure that no complications will happen. 537 00:41:13.840 --> 00:41:15.840 Similarly, you don't want to keep them in ic you too long because that's 538 00:41:15.880 --> 00:41:21.360 not great for the patient and it's a use of a very limited resource, 539 00:41:21.440 --> 00:41:23.360 the ice you. So a lot of these problems come up in a lot 540 00:41:23.360 --> 00:41:28.639 of places. And there with my colleague Ray Winslow, who's also a core 541 00:41:28.760 --> 00:41:35.800 member of the faculty in the Experiential AI Institute, he leverages what's called computational 542 00:41:35.960 --> 00:41:38.840 medicine. How do we model these things? How do we model processes, 543 00:41:38.880 --> 00:41:43.199 how do we model patients? How do we model diseases and use those to 544 00:41:43.199 --> 00:41:46.239 come up with predictions that are hard to get otherwise. So those are applications 545 00:41:46.239 --> 00:41:53.320 in healthcare, in life sciences. And our newest area is the climate and 546 00:41:54.480 --> 00:42:00.000 sustainability. How do we leverage kind of new data sources and models. Yeah, 547 00:42:00.039 --> 00:42:04.480 and to drill real quick into this, the ICU example, which is 548 00:42:04.480 --> 00:42:07.119 a fantastic one I'd like to point out for our audience. And let's get 549 00:42:07.119 --> 00:42:10.480 some feedback on this. Se mean email info at Dmradio dot biz or info 550 00:42:10.559 --> 00:42:14.960 in Inside Analysis dot com both come to me. You know, at any 551 00:42:15.000 --> 00:42:20.360 point in time in a workflow, you can insert an AI suggestion, for 552 00:42:20.440 --> 00:42:22.800 example, or you can have it sort of waiting to make a suggestion. 553 00:42:23.239 --> 00:42:28.960 So anytime the clinician is there and they're typing information, they're doing stuff as 554 00:42:29.000 --> 00:42:32.519 you go to the screen where you're going to think about maybe downgrading this person 555 00:42:32.519 --> 00:42:37.480 from ICU, that's an excellent point at which to offer an opportunity for the 556 00:42:37.480 --> 00:42:40.039 AI to give a suggestion. Right, so it'll come in and say, 557 00:42:40.679 --> 00:42:45.159 I see you've just launched this screen, maybe you should reconsider because of X 558 00:42:45.320 --> 00:42:47.440 y Z. And I think that's really where a lot of this AI value 559 00:42:47.480 --> 00:42:52.280 is going to manifest is in suggestions at certain points in a workflow. Right, 560 00:42:52.159 --> 00:42:57.880 absolutely, Eric, And you've touched on a very important area where my 561 00:42:58.000 --> 00:43:04.639 colleague Professor Ray Winslow who I mentioned working also with Professor Jene Tunnik, who 562 00:43:04.719 --> 00:43:07.840 runs our AI plus health area for the institute. They like to call it, 563 00:43:08.480 --> 00:43:13.960 how do I make AI a valued member of the medical team. That 564 00:43:14.119 --> 00:43:17.239 word you use, the recommendation is key, Right. AI doesn't know everything. 565 00:43:17.360 --> 00:43:22.280 It may sometimes it's right, sometimes it's wrong. The feedback from the 566 00:43:22.320 --> 00:43:27.719 physicians or the care team, a nurse could be anybody would be very beneficial 567 00:43:27.840 --> 00:43:31.960 to both improve the AI and to allow that AI to contribute to the proper 568 00:43:32.000 --> 00:43:36.960 care of that patient. So I love that analogy, which is, you 569 00:43:37.000 --> 00:43:40.320 know, AI needs to be just another member of the healthcare team, making 570 00:43:40.360 --> 00:43:46.239 its own recommendations and learning from corrections, interventions or getting it right and wrong. 571 00:43:46.960 --> 00:43:50.360 That's right in learning. And here's the beautiful thing. You know, 572 00:43:50.400 --> 00:43:54.360 doctor Faiad mentioned you have to capture all this information. When you correct the 573 00:43:54.400 --> 00:43:59.159 engine or when you accept the recommendation, all of that should be captured as 574 00:43:59.199 --> 00:44:02.719 well, because for time you can see, well, doctor Jones never accepts 575 00:44:02.800 --> 00:44:07.400 the recommendation of the AI engine, and doctor Phillips always accepts the recommendation of 576 00:44:07.440 --> 00:44:12.199 the AI engine. Somewhere in between is probably where we want to be. 577 00:44:12.840 --> 00:44:15.159 But the point is, you know, as human beings, we are actually 578 00:44:15.320 --> 00:44:20.239 very predictable. And you will know that if you start looking at your device 579 00:44:20.320 --> 00:44:23.000 that giving you feedback and how often you are on this app or that app 580 00:44:23.039 --> 00:44:27.960 or all these different things, you will see a reflection of your behavior and 581 00:44:28.000 --> 00:44:30.840 you'll realize why I am pretty darn predictable. Every day at this time, 582 00:44:30.880 --> 00:44:34.440 I do X. But hey, hang out for one second. Uh, 583 00:44:34.840 --> 00:44:37.599 We'll be right back after this break. You are listening to Inside Analysis. 584 00:44:40.960 --> 00:44:45.519 Okay, folks, time for the podcast bonus segment here on Inside Analysis is 585 00:44:45.519 --> 00:44:49.159 always so much fun talking to doctor Fayad. And you made a point there 586 00:44:49.199 --> 00:44:53.599 after in the break about the importance of trust and building trust in the system, 587 00:44:53.639 --> 00:44:57.039 for the system, for the doctors, for the team. Go ahead 588 00:44:57.039 --> 00:45:00.760 with that. Yeah, So, I mean one of the crit pieces of 589 00:45:00.840 --> 00:45:06.719 making the AI a valued member of the medical team or the care team is 590 00:45:07.119 --> 00:45:12.360 that ability to kind of make suggestions and get feedback those suggestions, those recommendations. 591 00:45:13.000 --> 00:45:15.599 They get value, you know, they get evaluated by a physician or 592 00:45:15.639 --> 00:45:19.679 a nurse or whatever. They may get corrected every once in a while, 593 00:45:19.719 --> 00:45:22.440 which gives them a chance to contribute to the system and it's evolution and they 594 00:45:22.440 --> 00:45:25.239 feel that their baby. But at the same time they get kind of this 595 00:45:25.599 --> 00:45:30.880 incremental exposure to hey, this thing is right. It caught something I missed. 596 00:45:30.119 --> 00:45:35.239 So it's useful and by the way, applies equally well whether you're dealing 597 00:45:35.320 --> 00:45:38.440 with a medical team or a factory. A team of engineers in a factory, 598 00:45:38.760 --> 00:45:45.480 a team of operators in a monitoring center, every application where the AI 599 00:45:45.599 --> 00:45:51.280 can contribute, you need to gain the trust of the users and the domain 600 00:45:51.320 --> 00:45:53.880 experts, because that's the only chance you're going to have of getting acceptance, 601 00:45:54.119 --> 00:45:59.719 getting important data, getting feedback, and keeping the system healthy. Yeah, 602 00:45:59.800 --> 00:46:02.199 that's right. And you know what I've noticed about the AI that I've seen 603 00:46:02.280 --> 00:46:06.760 in practice, which I see in my email or my text messages and other 604 00:46:06.760 --> 00:46:08.679 places, it's like, oh, you haven't followed up on this text, 605 00:46:08.719 --> 00:46:12.840 you know, do you think there was three days ago? Basically that's a 606 00:46:12.960 --> 00:46:16.760 perfect example of a reminder that is helpful because, let's face it, even 607 00:46:16.800 --> 00:46:21.480 the best doctor in the world will forget sometimes to do one little process, 608 00:46:21.559 --> 00:46:24.119 one little piece, you know, and depending upon the use case, that 609 00:46:24.159 --> 00:46:28.119 could be a very serious thing or not so serious thing. But you know, 610 00:46:28.159 --> 00:46:30.800 you forgot to wash your hands, for example, you forgot to put 611 00:46:30.840 --> 00:46:35.000 the mask on, you forgot to put the scalpel back where it belonged, 612 00:46:35.320 --> 00:46:38.960 these little teeny tiny things because guess what, People get distracted, Even very 613 00:46:39.039 --> 00:46:44.079 focused people can get distracted. And so the AI is there as your little 614 00:46:44.079 --> 00:46:47.679 helper watching you, making little suggestions along the way, not wrapping you on 615 00:46:47.719 --> 00:46:52.119 the knuckles, getting you in trouble with your boss necessarily, but just giving 616 00:46:52.159 --> 00:46:53.960 you a little bit of a hint and a nudge. And it reminds me 617 00:46:53.960 --> 00:46:57.920 of one of my favorite expressions. That's an old cliche. I don't hear 618 00:46:58.000 --> 00:47:00.079 much anymore. But uh, and of course it's it's posed on man, 619 00:47:00.119 --> 00:47:02.719 but it means men and women. It said, man doesn't need to be 620 00:47:02.800 --> 00:47:07.000 taught so much as reminded. Right, we already learned. We just have 621 00:47:07.119 --> 00:47:10.320 to be reminded. Final thoughts from doctor Faytte go ahead, Yeah, and 622 00:47:12.000 --> 00:47:15.840 absolutely, and you actually need both. You know. Learning in the system 623 00:47:15.920 --> 00:47:20.559 is also critical because you know, no system is going to get it right. 624 00:47:20.599 --> 00:47:22.039 So every once in a while it's going to guess the wrong thing and 625 00:47:22.079 --> 00:47:25.079 it becomes annoying. It's reminding me of something I already know. Right, 626 00:47:25.440 --> 00:47:29.119 well, I need to give it that feedback, hey, in this context, 627 00:47:29.119 --> 00:47:31.760 don't remind me. And then if it adapts to that, Now you 628 00:47:31.800 --> 00:47:36.639 have intelligence. Now you have something you value and it's personalized. So now 629 00:47:36.679 --> 00:47:38.480 you want to keep it because now it's beginning to get to know you and 630 00:47:38.519 --> 00:47:44.119 your preferences versus me and my preferences, which could be completely different. So 631 00:47:44.199 --> 00:47:47.840 bringing machine learning into the loop and making sure these systems learn and adapt to 632 00:47:49.000 --> 00:47:52.960 the feedback they're getting is very critive. Yeah, this is wonderful stuff. 633 00:47:52.119 --> 00:47:55.400 Well, look this, gentleman up online and look up Northeastern University, the 634 00:47:55.480 --> 00:48:00.679 Institute for Experiential AI. That just means humans in the loop. Doctor Fiduction 635 00:48:00.960 --> 00:48:05.880 here, and you folks who put together a great program, and you understand 636 00:48:05.920 --> 00:48:12.480 the interplay between training and research and development and solution designed. All these things 637 00:48:12.559 --> 00:48:15.000 feed each other, and we'll feed each other over time. So you know 638 00:48:15.039 --> 00:48:20.599 how to train people, know how to get people excited about stuff and really 639 00:48:20.719 --> 00:48:23.280 nudge them and get them involved. Because the humans in the loop, they're 640 00:48:23.280 --> 00:48:28.599 the ones who are going to make all this stuff successful and worthwhile. And 641 00:48:28.679 --> 00:48:31.199 it's going to save lives in the healthcare space. I promise you it's going 642 00:48:31.239 --> 00:48:34.880 to save lives. It's going to save time and effort. It's going to 643 00:48:34.960 --> 00:48:37.920 do the most important thing I think for any business that has improved morale. 644 00:48:38.280 --> 00:48:42.199 When morele goes up, good things happen. When morel is down, good 645 00:48:42.199 --> 00:48:45.360 things do not happen. So just that's the Dallas Cowboys. Thank you, 646 00:48:45.519 --> 00:48:49.960 thank you, Eric, and thank you for getting our mission and vision for 647 00:48:50.039 --> 00:48:52.920 the institute. You got it perfectly. I love it. Well, we'll 648 00:48:52.920 --> 00:48:54.880 talk to these folks again. I'm sure folks don't touch that boy. Actually, 649 00:48:54.920 --> 00:49:00.960 we'll touch you next time. Info and Inside Analysis dot Com Bye bye. 650 00:49:01.079 --> 00:49:07.639 KCAA is your CNBC News affiliate where the station that gets down to the 651 00:49:07.679 --> 00:49:16.480 business open for takeout and delivery. Held Tapiac Mexican food restaurant in the Tri 652 00:49:16.639 --> 00:49:21.679 City Center of Redlands is back. Their entire family is on hand to serve 653 00:49:21.760 --> 00:49:25.519 up their delicious burritos, but chaka charizo, wavos, ranchero's steak and eggs 654 00:49:25.800 --> 00:49:30.920 just part of their mouth watering great food. Since fifteen thirty one, people 655 00:49:30.920 --> 00:49:34.199 have marveled at the miracle of El Tapiac, and now you can marvel at 656 00:49:34.239 --> 00:49:37.719 the great food the Lugo family has been serving up for over two decades, 657 00:49:37.840 --> 00:49:42.039 nestled quietly in the corner of the Tri City Center shopping mall next to Burlington 658 00:49:42.119 --> 00:49:45.199 Code Factory. Support them. They can't wait to serve you some of their 659 00:49:45.239 --> 00:49:49.679 delectable, authentic South of the Border Mexican fair at great prices, served up 660 00:49:49.719 --> 00:49:53.119 with love. Support the area's best loved Mexican food restaurant in these tough times, 661 00:49:53.320 --> 00:49:58.079 order up a tasty meal on the phone for delivery or takeout for breakfast, 662 00:49:58.119 --> 00:50:00.480 lunch, or dinner from ten am to six zo called nine to nine 663 00:50:00.559 --> 00:50:06.360 threeo seven zero zero one seven that's nine to nine three oh seven zero zero 664 00:50:06.559 --> 00:50:13.880 seventeen or google eltpiac Redlands and treat yourself to Hebot Club's original pure powdy Arco 665 00:50:14.000 --> 00:50:17.000 super TA helps build red corpuscles in the blood, which carry oxygen to our 666 00:50:17.119 --> 00:50:22.760 organs and cells. Our organs them cells need oxygen to regenerate themselves. The 667 00:50:22.800 --> 00:50:27.360 immune system needs oxygen to develop, and cancer dies in oxygen. So the 668 00:50:27.400 --> 00:50:30.079 T is great for healthy people because it helps build the immune system, and 669 00:50:30.119 --> 00:50:35.519 it can truly be miraculous for someone fighting a potentially life threatening disease due to 670 00:50:35.599 --> 00:50:39.119 an infection, diabetes, or cancer. The T is also organic and naturally 671 00:50:39.159 --> 00:50:43.880 caffeine free. A one pound package of T is forty nine ninety five which 672 00:50:43.920 --> 00:50:46.800 includes shipping. To order, please visit to ebot club dot com. To 673 00:50:46.920 --> 00:50:52.800 hebo is spelled T like tom, A h ee b like boy oh, 674 00:50:52.039 --> 00:50:57.079 then continue with the word T and then the word club. The complete website 675 00:50:57.119 --> 00:51:00.039 is to Hebot club dot com or call us at e one eight sixty one 676 00:51:00.159 --> 00:51:05.119 zero eight zero eight eight Monday through Saturday, nine am to five pm, 677 00:51:05.239 --> 00:51:08.960 California time. That's eight one eight sixty one zero eight zero eight eight tbot 678 00:51:09.119 --> 00:51:15.559 club dot Com with sixty years of fascinating facts. This is the man from 679 00:51:15.679 --> 00:51:22.920 yesterday and back in time. We go to this time in nineteen fifty three. 680 00:51:22.679 --> 00:51:27.480 He's back to civilian life. Harry S. Truman says he feels again 681 00:51:27.559 --> 00:51:30.159 like a country boy in the big city as he opens up private offices in 682 00:51:30.239 --> 00:51:38.079 Missouri and begins the long rule of adjustment to private life, even as you 683 00:51:39.960 --> 00:51:50.239 till average come. And from this time in nineteen eighty one in BC cancels 684 00:51:50.519 --> 00:51:53.960 Sanford and is going to replace it with The Brady Brides. The new series 685 00:51:54.000 --> 00:51:59.880 grew out of a reunion movie featuring the original cast of The Brady Bunch. 686 00:52:00.920 --> 00:52:09.360 Hey Good, I'm ready, And from about this time in nineteen seventy nine 687 00:52:09.760 --> 00:52:14.480 movie Grade, James Cagney is interviewed for the first time in nineteen years. 688 00:52:14.960 --> 00:52:19.440 James Cagney, who is now eighty was interviewed by David Hartman on Good Morning 689 00:52:19.480 --> 00:52:22.119 America. Haydi, my guts shall make you act like this. It'll do 690 00:52:22.280 --> 00:52:23.559 for the time being. Well, be that as it may. Let's talk 691 00:52:23.599 --> 00:52:27.559 sense. I'm out and taking your dirty yellow belly Rightawa. Give it to 692 00:52:27.599 --> 00:52:36.239 you to the door with more at man from yesterday dot com. It's that 693 00:52:36.360 --> 00:52:39.840 time of year again, No, not the holidays. Medicare open enrollment and 694 00:52:39.920 --> 00:52:44.760 if you have questions about Medicare, you should talk to the local experts. 695 00:52:44.800 --> 00:52:49.000 Paul Barratt and Associates. All of his agents are certified with plans that are 696 00:52:49.039 --> 00:52:52.960 accepted by most of the medical groups in our area. Call nine oh nine 697 00:52:52.119 --> 00:52:57.719 seven nine three oh three eight five. Their service is free and after forty 698 00:52:57.760 --> 00:53:00.039 two years of the business, their agents are trained to help you pick the 699 00:53:00.119 --> 00:53:06.679 plan that's right for you. Tune into The Faran Dozier Show, Music Marks 700 00:53:06.719 --> 00:53:09.639 Place in Time, the soundtrack to Life, Sunday nights at eight pm. 701 00:53:09.840 --> 00:53:16.320 Are kcaa radio playing the hottest hits and the coolest conversations Sunday nights at APM 702 00:53:16.440 --> 00:53:22.000 on The Ferroan Dozier Show. Within the ray of music, talk, sports, 703 00:53:22.199 --> 00:53:27.679 community outreach and veteran resources. The hits from the sixties, seventies, 704 00:53:28.039 --> 00:53:34.719 eighties, nineties, and today's hits. The Farandozier Show on KCAA Radio on 705 00:53:34.920 --> 00:53:40.280 all available streaming platforms and on a six point five m and ten fifty am 706 00:53:40.400 --> 00:53:59.679 The Farando Zier Show on KCAA Radio. Last July, several GOP senators combined 707 00:53:59.719 --> 00:54:05.679 their what intellects to charge that inflation was rising because of quote insane tax and 708 00:54:05.719 --> 00:54:10.480 spending spree of President Biden and the Democrats. Never mind that the insane spending 709 00:54:10.599 --> 00:54:15.559 is for such sensible, productive, and enormously popular national needs as childcare and 710 00:54:15.679 --> 00:54:22.159 childless benefits. Mitch McConnell's rapidly partisan flock saw the chance to politicize the public's 711 00:54:22.199 --> 00:54:28.079 legitimate worries about rising prices. You, poor consumers are made to pay more 712 00:54:28.119 --> 00:54:34.440 for basics, they squawked because of socialist Joe's investment, and grassroots people follow 713 00:54:34.480 --> 00:54:38.639 the ricocheting pinball of the GOP's logic. One, they say that helping hard 714 00:54:38.679 --> 00:54:44.440 hit families induces them to refuse to go to work. Two, This creates 715 00:54:44.480 --> 00:54:49.679 blockages in the global supply chain. Three, this causes shortages of everything. 716 00:54:50.199 --> 00:54:55.159 Four This forces corporate bosses to raise all prices, which five slams the middle 717 00:54:55.159 --> 00:55:01.920 class and poor. So six lazy workers cause inflation. Who rube Goldberg couldn't 718 00:55:01.920 --> 00:55:07.039 have dreamed up a more fantastical diagram to deflect attention from what's really happening, 719 00:55:07.480 --> 00:55:13.079 namely that instead of an inflation problem, we have a corporate greed problem. 720 00:55:13.719 --> 00:55:17.119 Of course, the greed moisters insist that their pursuit of excess corporate profits has 721 00:55:17.360 --> 00:55:22.280 not driven any price surges in our economy or free market competition. They snap. 722 00:55:22.599 --> 00:55:27.920 Prices are established by the law of supply and demand. It's the magic 723 00:55:27.960 --> 00:55:31.800 of the marketplace, they explain. But magicians don't do magic, they perform 724 00:55:31.880 --> 00:55:37.480 illusions, and the illusion of free market competition implodes when it hits the reality 725 00:55:37.559 --> 00:55:44.400 that our economy doesn't remotely resemble a competitive marketplace. This is Jim Hiitwer saying, 726 00:55:44.559 --> 00:55:49.679 for some forty years, corporate directed government policies have been transforming America into 727 00:55:49.719 --> 00:55:55.280 monopoly nation, letting the feud gouge the minute. That's where inflation comes from. 728 00:55:55.920 --> 00:56:00.000 Do you want to learn and get answers to questions, not addressing them 729 00:56:00.079 --> 00:56:02.719 as media, Then you want to hear my show business Game Changers with me, 730 00:56:04.119 --> 00:56:07.400 Sarah Westall, I have conversations with thought leaders who have the courage to 731 00:56:07.400 --> 00:56:12.800 address off limit topics. So you and your family have the tools to make 732 00:56:12.800 --> 00:56:16.760 the right decisions. Join me Wednesdays at four pm on ten fifty AM and 733 00:56:16.920 --> 00:56:22.000 one oh six point five FM. Right here at KCAA, that station that 734 00:56:22.119 --> 00:56:30.159 leaves no listener behind. KCAA Radio has openings for one hour talk shows. 735 00:56:30.280 --> 00:56:34.320 If you want to host a radio show, now is the time. Make 736 00:56:34.400 --> 00:56:38.119 kca your flangship station. Our rates are affordable and our services are second to 737 00:56:38.199 --> 00:56:43.559 none. We broadcast to a population of five million people plus. We stream 738 00:56:43.599 --> 00:56:47.000 and podcast on all major online audio and video systems. If you've been thinking 739 00:56:47.000 --> 00:56:52.400 about broadcasting a weekly radio program on real radio plus the Internet, contact our 740 00:56:52.480 --> 00:56:58.920 CEO at two eight one five nine ninety eight hundred two eight one five nine 741 00:56:59.039 --> 00:57:01.800 nine night. You can skype your show from your home to our Redlands, 742 00:57:01.840 --> 00:57:07.280 California studio, where our live producers and engineers are ready to work with you 743 00:57:07.320 --> 00:57:12.559 personally. A radio program on KCAA is the perfect work from home advocation in 744 00:57:12.599 --> 00:57:16.559 these stressful times. Just type KCAA Radio dot com into your browser to learn 745 00:57:16.599 --> 00:57:21.079 more about hosting a show on the best station in the nation, or call 746 00:57:21.199 --> 00:57:25.639 our CEO for details. Two eight one five nine nine ninety eight hundred Listina 747 00:57:25.719 --> 00:57:30.000 KCAA Lowolinda at one, O six point five, FMK two ninety three C 748 00:57:30.199 --> 00:57:37.199 F Brito Valley, NBC News Radio, I'm Chris Gragio. The GOP race 749 00:57:37.199 --> 00:57:40.000 for the presidential nominee is now down to a two person race. Florida Governorn 750 00:57:40.039 --> 00:57:45.599 de Santis dropped out and endorsed former President Trump today. DeSantis was considered a 751 00:57:45.599 --> 00:57:49.079 front runner early on, but finished a distant second in the Iowa caucus. 752 00:57:49.199 --> 00:57:52.400 That leaves former South Carolina Governor Niki Haley is the only remaining candidate against Trump. 753 00:57:52.639 --> 00:57:57.199 With the New Hampshire primary just over twenty four hours away. New Hampshire 754 00:57:57.239 --> 00:58:01.039 Governor and Nikki Haley surrogate Chris Sanunu was already calling the GOP primary a two 755 00:58:01.039 --> 00:58:06.960 person race hours before DeSantis dropped out. Speaking on NBC's Meet the Press Today, 756 00:58:07.000 --> 00:58:10.000 Sanunu said Haley knocked everybody out, even Ron. While he said Tuesday's 757 00:58:10.039 --> 00:58:14.440 New Hampshire primary wasn't a make or break for Haley, he did say that 758 00:58:14.519 --> 00:58:19.000 she'll have to start winning states. Michigan Governor Gretchen Whitmer says abortion is a 759 00:58:19.000 --> 00:58:22.639 major issue in the presidential election. It is very clear that abortion is on 760 00:58:22.840 --> 00:58:27.760 this ballot. Speaking on CBS's Faced the Nation, the two term Democrats said 761 00:58:27.800 --> 00:58:30.559 she believes a Republican win in November in the race for the White House means 762 00:58:30.599 --> 00:58:35.719 a national abortion ban, no matter who's the GOP candidate. Whitmer noted that 763 00:58:35.760 --> 00:58:38.639 although former President Trump hasn't committed to a national ban on abortion, he's taking 764 00:58:38.639 --> 00:58:44.199 credit for appointing three conservative judges to a US Supreme Court that overturned Roe v. 765 00:58:44.280 --> 00:58:46.280 Wade. In twenty twenty two. Whitmer ie re election in Michigan, 766 00:58:46.360 --> 00:58:51.480 largely campaigning on the issue of abortion. A crash of an air Vac helicopter 767 00:58:51.559 --> 00:58:54.280 in Oklahoma's killed three crew members. The crew was returning to whether Ford, 768 00:58:54.320 --> 00:58:59.719 Oklahoma, after transporting a patient to a hospital in Oklahoma City when the helicopter 769 00:58:59.760 --> 00:59:04.199 craw late last night. Official say law enforcement and members of the Air Evac 770 00:59:04.320 --> 00:59:08.239 Life Team work together to find the crash site. The NTSB will be investigating. 771 00:59:08.559 --> 00:59:13.039 Police are looking for suspects after four people were injured in a shooting at 772 00:59:13.039 --> 00:59:16.199 a Florida best Buy. It happened in Plantations shortly after eight pm Saturday, 773 00:59:16.400 --> 00:59:21.400 when the victims were approached by another group while leaving the store. One person 774 00:59:21.719 --> 00:59:24.719 was listed in critical condition. Police say the suspects fled the scene and investigators 775 00:59:24.719 --> 00:59:29.760 are asking anybody with information to come forward. I'm Chris Caragio, NBC News 776 00:59:29.800 --> 00:59:37.039 Radio, NBC News. I'm JCAA Lomelad celebrating our twenty fifth anniversary sponsored by 777 00:59:37.079 --> 00:59:43.960 Teamsters Local nineteen thirty two Protecting the Future of Working Families Teamsters nineteen thirty two 778 00:59:44.039 --> 00:59:52.880 dot Org. Morales Tiring Auto Electric in Ukaypas says it's hard being away from 779 00:59:52.960 --> 00:59:57.480 home during the holidays, and even harder if your soldier overseas securing our foods 780 00:59:57.480 --> 01:00:00.360 and letters and packages and show support to military families here at home during this 781 01:00:00.480 --> 01:00:05.960 holiday season. This message courtesy of Morales Tire In Auto Electric. Act three 782 01:00:06.119 --> 01:00:08.320 five zero one nine U Kuaipa Boulevard in u Kaipa