WEBVTT 1 00:00:00.840 --> 00:00:01.360 Dot Org. 2 00:00:06.280 --> 00:00:11.199 KCAA is your CNBC News affiliate where the station that 3 00:00:11.640 --> 00:00:13.320 gets down to business. 4 00:00:16.280 --> 00:00:18.920 The views and opinions expressed by individuals on the following 5 00:00:18.960 --> 00:00:22.359 program do not necessarily reflect those of the network, Guys 6 00:00:22.359 --> 00:00:26.960 Guy Radio and its platforms. 7 00:00:28.559 --> 00:00:35.479 It's Guys Guy Radio. Here's your host, Robert Manny. 8 00:00:37.399 --> 00:00:40.640 Welcome to Guys Guys Radio. This is your host, Robert Manny, 9 00:00:40.759 --> 00:00:43.719 welcoming you to the show where men and women can 10 00:00:43.759 --> 00:00:46.960 be at their best and everyone wins. Guys Guys Radio. 11 00:00:47.399 --> 00:00:51.520 We're here to inform you, inspire you, empower you, and 12 00:00:51.679 --> 00:00:54.960 get you to think, feel, and who knows, maybe even 13 00:00:55.000 --> 00:00:59.119 act by virtue of the journeys, stories, experiences, and insights 14 00:00:59.320 --> 00:01:01.640 of the guests bring you each and every week to 15 00:01:01.679 --> 00:01:04.439 the show. And once again this week is no exception. 16 00:01:04.519 --> 00:01:08.599 Today we're going to talk about AI artificial intelligence with 17 00:01:08.799 --> 00:01:15.120 author Agg Liu, who's got a book written with Ashwari 18 00:01:15.480 --> 00:01:19.480 how Are PhD. The book is called The Handy Artificial 19 00:01:19.680 --> 00:01:23.000 Intelligence answer Book. It is terrific and I can't wait 20 00:01:23.040 --> 00:01:26.760 to get into our conversations with Agg because I think 21 00:01:27.120 --> 00:01:28.560 I'm going to learn a lot and you're going to 22 00:01:28.680 --> 00:01:29.920 learn a lot, and we're going to have a lot 23 00:01:29.920 --> 00:01:31.840 of fun here on Guys Guys Radio. So I ask 24 00:01:31.879 --> 00:01:35.799 yourself what does artificial intelligence mean for work, for society, 25 00:01:35.879 --> 00:01:38.400 for our lives? What is it capable of right now? 26 00:01:38.439 --> 00:01:41.040 And what might it be capable of in the future. 27 00:01:42.599 --> 00:01:46.560 It's popping up everywhere. Google Search is consumer customer service calls. 28 00:01:46.599 --> 00:01:48.959 I just had one the other day where AI handled 29 00:01:49.000 --> 00:01:54.319 everything perfectly. Netflix, YouTube, TikTok, algorithms, real time traffic updates, 30 00:01:54.439 --> 00:01:57.400 and routing Siri at Alexa, to name just a few. 31 00:01:57.439 --> 00:02:00.799 It seems that AI is pervading more and more of 32 00:02:00.840 --> 00:02:03.359 our society and we've got to get a handler on it. 33 00:02:03.480 --> 00:02:05.680 A lot of people ask, is AI going to take 34 00:02:05.799 --> 00:02:09.840 my job? Well, I think the key is that you've 35 00:02:09.840 --> 00:02:12.120 got to learn how to manage AI. It's a tool. 36 00:02:12.479 --> 00:02:14.960 It's going to take some jobs, replace some jobs, and 37 00:02:15.000 --> 00:02:17.319 it's going to create other jobs. So you have to 38 00:02:17.319 --> 00:02:19.919 be on the creative edge when it comes to doing 39 00:02:20.000 --> 00:02:24.319 things that a computer or algorithms can't do. The creativity, 40 00:02:24.560 --> 00:02:28.039 the emotional connection with other people, the problem solving, and 41 00:02:28.240 --> 00:02:30.080 that's how I think people are going to be able 42 00:02:30.120 --> 00:02:32.719 to continue to grow and get ahead. Our side is 43 00:02:32.719 --> 00:02:35.000 going to change. It's the same thing when the Internet 44 00:02:35.039 --> 00:02:38.080 came up, and email came up, and fax machines came up, 45 00:02:38.080 --> 00:02:40.240 and the telephone came up, and the TV came up. 46 00:02:40.439 --> 00:02:44.199 When all of these things were unleashed on our society, 47 00:02:44.759 --> 00:02:47.560 we managed to deal with it and we managed to 48 00:02:47.599 --> 00:02:50.199 continue growing. And I think that's what's going to happen 49 00:02:50.520 --> 00:02:52.479 in AI. And today we're going to talk about the 50 00:02:52.520 --> 00:02:55.360 earliest forms of AI, how it works, what some of 51 00:02:55.400 --> 00:02:57.800 the ethical questions are, what are some of the fears 52 00:02:57.840 --> 00:03:00.759 associated with the idea of thinking machines, and more with 53 00:03:00.879 --> 00:03:06.319 special guests agg Liu right here on today's Guys Guys Radio. 54 00:03:07.000 --> 00:03:08.599 It's Guys Guy Radio. 55 00:03:10.280 --> 00:03:13.319 Okay, Guys Guys Radio. It is the interview portion of 56 00:03:13.319 --> 00:03:16.479 our show, and today we're going to talk about AI. 57 00:03:16.759 --> 00:03:20.840 My special guest is agg Leu, author of the Handy 58 00:03:21.759 --> 00:03:25.639 Artificial Intelligence answer Book, a practical guide designed to help 59 00:03:25.800 --> 00:03:30.280 everyday people, professionals, entrepreneurs, and curious minds understand one of 60 00:03:30.319 --> 00:03:35.280 the most transformative technologies of our time, Artificial intelligence. You know, 61 00:03:35.319 --> 00:03:38.560 AI is no longer something reserved for scientists and tech giants, 62 00:03:38.560 --> 00:03:41.800 as already influencing how we work, how we communicate, how 63 00:03:41.800 --> 00:03:44.719 we create, how we learn how we shop, invest and 64 00:03:44.800 --> 00:03:48.560 make decisions. Yet many people remain uncertain about what AI 65 00:03:48.759 --> 00:03:51.919 actually is, how it works, and how it will affect 66 00:03:52.000 --> 00:03:54.080 their lives. And believe me, it will. So today we'll 67 00:03:54.080 --> 00:03:58.199 separate fact from fiction, explore the opportunities and risks of AI, 68 00:03:58.360 --> 00:04:02.199 and discuss how listeners can use this powerful technology wisely 69 00:04:02.240 --> 00:04:06.960 and effectively. Welcome to Guys Guys Radio A G g Liu. 70 00:04:07.599 --> 00:04:09.680 Hello, Hello, good to be here. 71 00:04:10.000 --> 00:04:11.919 So I guess we should start at the beginning. You 72 00:04:12.000 --> 00:04:14.199 wrote this book with your writing. 73 00:04:13.919 --> 00:04:16.439 Partner, doctor Ashwari Powar. 74 00:04:16.560 --> 00:04:19.079 Yeah, here's the here's the book. I'm in the middle 75 00:04:19.079 --> 00:04:20.439 of it, so I have a piece of favor there. 76 00:04:20.480 --> 00:04:23.480 I'll take it out. But it's fantastic. And the first 77 00:04:23.480 --> 00:04:25.319 thing I wondered when I was reading the book was 78 00:04:25.439 --> 00:04:29.079 I wonder if AI wrote this book for you. 79 00:04:29.079 --> 00:04:33.160 No, we did this all all manually, all manually. We 80 00:04:33.160 --> 00:04:36.319 we we really wanted the goal here to be something 81 00:04:36.360 --> 00:04:39.399 that would be accessible to people, right. And I think 82 00:04:39.920 --> 00:04:41.600 in this age of AI, where it's so easy to 83 00:04:41.639 --> 00:04:43.839 get AI to write something for you, the thing that's 84 00:04:43.879 --> 00:04:46.600 going to be more important is knowing exactly what to write. 85 00:04:46.600 --> 00:04:48.800 Knowing exactly what you can do that, the I can't. 86 00:04:49.560 --> 00:04:54.079 And so when we created this, you know, the publisher 87 00:04:55.199 --> 00:04:58.000 with the publisher Visible in Press, they came to me 88 00:04:58.480 --> 00:05:00.199 and they said, hey, we have this great idea. We 89 00:05:00.240 --> 00:05:02.680 want to write this book in the series about AI. 90 00:05:02.920 --> 00:05:06.319 And I created a draft, worked with doctor Powart to 91 00:05:06.319 --> 00:05:09.360 sort of turn it into the current format. And yeah, 92 00:05:09.560 --> 00:05:11.560 it's really meant to be for a general audience. It's 93 00:05:11.560 --> 00:05:14.800 meant to be for people, you know, like your listeners. 94 00:05:15.079 --> 00:05:17.639 You know, it's amazing. I've gone through the book. It's 95 00:05:17.639 --> 00:05:19.040 not one of these books that you have to read 96 00:05:19.040 --> 00:05:21.279 from page one to the end, but it's just a 97 00:05:21.319 --> 00:05:24.839 list of questions that are really good and has very clear, 98 00:05:24.959 --> 00:05:28.040 easy to digest answers. And I think anybody who has 99 00:05:28.040 --> 00:05:30.000 an interest in AI, and if you don't have an 100 00:05:30.040 --> 00:05:33.000 interest in AI, you basically don't have an interest in 101 00:05:33.040 --> 00:05:35.439 being part of the color culture going forward. Because this 102 00:05:35.519 --> 00:05:38.240 is there, it's here to stay. And the key is, 103 00:05:38.279 --> 00:05:40.519 as I tell my thirteen year old son, you know, 104 00:05:40.519 --> 00:05:42.560 don't worry about it taking your job. Learn how to 105 00:05:42.680 --> 00:05:44.560 use it because it's going to create so much more 106 00:05:44.680 --> 00:05:47.040 in terms of jobs. So nothing to be afraid of. 107 00:05:47.279 --> 00:05:50.079 But a couple of guardrails, of course, but concerns but 108 00:05:50.199 --> 00:05:52.560 not fear, and let's let's work with it so for 109 00:05:52.600 --> 00:05:54.959 the benefit of everybody, because we hear, you know, about 110 00:05:54.959 --> 00:05:57.959 computing and Google, and then we hear about AI, and 111 00:05:58.000 --> 00:05:59.839 I think a lot of people are not sure how 112 00:05:59.879 --> 00:06:03.480 to make the differentiation between what's the difference between like 113 00:06:03.519 --> 00:06:07.759 googling something and asking a service like CHATGBT, what's the 114 00:06:07.800 --> 00:06:12.240 difference between AI and data processing computer typic computer stuff. 115 00:06:12.279 --> 00:06:15.240 Yeah, so AI came out of that data processing idea, 116 00:06:15.279 --> 00:06:18.120 but it works a little differently. So the AIS that 117 00:06:18.160 --> 00:06:20.920 we talk about today usually are something called a large 118 00:06:21.040 --> 00:06:24.120 language model. Large because it's got tons and tons of 119 00:06:24.160 --> 00:06:26.720 these numbers inside of it that it uses to make 120 00:06:26.759 --> 00:06:29.319 its processing. And it's a language model because it works 121 00:06:29.319 --> 00:06:31.879 with human language. So the difference between something like a 122 00:06:31.920 --> 00:06:34.399 standard program and a large language model is that when 123 00:06:34.399 --> 00:06:37.160 we write a computer program, even something as complicated as 124 00:06:37.279 --> 00:06:42.160 say Google Search was before it became AI related, every 125 00:06:42.480 --> 00:06:45.800 line of code in there was created by someone with 126 00:06:45.839 --> 00:06:49.240 a specific purpose. So you'd have a whole giant team 127 00:06:49.240 --> 00:06:52.319 of coders. They would work together to create this gigantic 128 00:06:52.360 --> 00:06:55.759 algorithm and everything, and they would know how each part 129 00:06:55.800 --> 00:06:59.319 works with AI, it's a little bit different with AI. 130 00:06:59.480 --> 00:07:02.800 What you do is you create a machine that can learn, 131 00:07:02.839 --> 00:07:05.439 hence the term machine learning, which is often thrown around 132 00:07:05.720 --> 00:07:09.360 in the same sentences as AI. And basically the way 133 00:07:09.399 --> 00:07:11.040 it learns it's not quite the same as a human 134 00:07:11.120 --> 00:07:13.759 but it has all these sort of tweakable numbers in it, 135 00:07:14.480 --> 00:07:17.920 and you give it tons and tons and tons of examples, 136 00:07:17.959 --> 00:07:20.639 you know, most of the Internet at this point, and 137 00:07:22.199 --> 00:07:25.439 it learns how to write human language by looking at 138 00:07:25.439 --> 00:07:27.439 those examples and trying to predict what's the next word 139 00:07:27.480 --> 00:07:29.560 in a sentence or a paragraph or something like that. 140 00:07:30.279 --> 00:07:34.199 And so it doesn't work in a way where we 141 00:07:34.319 --> 00:07:36.439 know how each line of it works. In fact, there's 142 00:07:36.439 --> 00:07:38.839 a lot of effort right now to try and understand 143 00:07:38.879 --> 00:07:42.319 what goes on inside these inside their little programming circuits here. 144 00:07:43.120 --> 00:07:46.079 So when you ask something like Chad GBT, it's not 145 00:07:46.399 --> 00:07:49.120 going to be necessarily quite as reliable as something like 146 00:07:49.120 --> 00:07:53.639 a Google search or quite as organize the same way. 147 00:07:53.680 --> 00:07:55.279 But it's going to have different strengths. It can do 148 00:07:55.360 --> 00:07:57.720 some things that are hard to program explicitly, things that 149 00:07:57.759 --> 00:08:01.040 are hard to explain about, you know, civity, making poetry, 150 00:08:01.040 --> 00:08:03.680 things like that fantastic. 151 00:08:03.759 --> 00:08:06.959 So what's the biggest mixed misconception and that people have 152 00:08:07.079 --> 00:08:09.959 about AI right now? The general population? 153 00:08:10.720 --> 00:08:15.120 Ooh, the general population. I would say a lot of 154 00:08:15.120 --> 00:08:17.360 people don't know what AI is capable of right now. 155 00:08:18.560 --> 00:08:21.720 I think a lot of people if they've tried AI, 156 00:08:21.920 --> 00:08:24.920 they you know, used it when chat GPT first came out, 157 00:08:24.959 --> 00:08:27.240 and they think, Okay, that's you know, from twenty twenty two. 158 00:08:27.360 --> 00:08:31.079 That's exactly how it is now. But it's gotten a 159 00:08:31.120 --> 00:08:33.799 lot better and different in certain ways. 160 00:08:34.440 --> 00:08:34.679 You know. 161 00:08:34.679 --> 00:08:37.559 Now AI it can work with text, it can actually 162 00:08:37.720 --> 00:08:40.919 write things out for you, it can work with images, 163 00:08:40.960 --> 00:08:42.840 it can work with so many different sorts of things. 164 00:08:44.159 --> 00:08:46.799 And one of the big problems with the AI that's 165 00:08:46.799 --> 00:08:48.679 been true for a very long time and it's still true, 166 00:08:48.759 --> 00:08:50.720 is that it's not one hundred percent reliable. You know, 167 00:08:50.759 --> 00:08:53.960 sometimes it'll make things up. So the companies call this 168 00:08:54.039 --> 00:08:57.279 the hallucination. I prefer the term confabulation myself. It's just 169 00:08:57.639 --> 00:09:01.159 basically shamelessly guessing an answer. But what you can do 170 00:09:01.200 --> 00:09:04.000 is you can ask the AI where it got its 171 00:09:04.000 --> 00:09:07.279 information from, and it'll be able to send you links 172 00:09:07.320 --> 00:09:10.600 to other places on the internet. And that means you 173 00:09:10.720 --> 00:09:13.879 as a person can back up what this AI is saying. 174 00:09:14.080 --> 00:09:17.159 So again it's about like what you're saying, how you 175 00:09:17.279 --> 00:09:19.759 use AI learning, how to use it effectively to help 176 00:09:19.799 --> 00:09:21.120 you instead of to mislead you. 177 00:09:21.679 --> 00:09:26.039 It's two thoughts here, agg One is why we question 178 00:09:26.240 --> 00:09:29.240 why AI seem to have like shown up all at once. 179 00:09:29.519 --> 00:09:33.000 It's been being used, but I think it's been kind 180 00:09:33.039 --> 00:09:38.440 of unveiled to the general public recently. But oh is 181 00:09:38.480 --> 00:09:39.159 that correct? 182 00:09:39.879 --> 00:09:42.639 Yeah, so really it hit the public scene all at 183 00:09:42.639 --> 00:09:45.399 once sort of in twenty twenty two with the release 184 00:09:45.440 --> 00:09:50.720 of chat GBT. The original version. Basically, the methods of 185 00:09:50.759 --> 00:09:54.919 computer science, the logic behind how we build these systems, 186 00:09:55.200 --> 00:09:58.919 has existed in some form for decades. It sort of 187 00:09:58.960 --> 00:10:02.840 became a hot topic in research field in the twenty 188 00:10:02.879 --> 00:10:08.679 tens or so, but it wasn't hooked up to enough 189 00:10:08.720 --> 00:10:11.879 computing power to actually make something that would be useful 190 00:10:11.879 --> 00:10:14.200 for the general public or interesting to the general public. 191 00:10:14.360 --> 00:10:15.960 It was just, you know, a research tool. It could 192 00:10:15.960 --> 00:10:18.399 do something like play the game of Go, for instance. 193 00:10:18.600 --> 00:10:21.320 It was an ancient Chinese game that was famously beaten 194 00:10:21.360 --> 00:10:25.399 in twenty sixteen. They beat the best human experts. So 195 00:10:25.799 --> 00:10:29.919 it's it's right when that chat GPT moment was the 196 00:10:29.960 --> 00:10:33.279 first time that these older algorithms were connected to enough 197 00:10:33.279 --> 00:10:37.759 computing power to matter. And once companies saw how big 198 00:10:37.799 --> 00:10:41.039 of a prize that was, that's when they started throwing 199 00:10:41.080 --> 00:10:44.000 tons and tons and tons of their computer power and 200 00:10:44.039 --> 00:10:46.320 their money to build new computers and everything like that 201 00:10:46.600 --> 00:10:49.399 in order to make this the behemoth that is today. 202 00:10:50.320 --> 00:10:53.879 And now we've got talking about behemoths. You have companies 203 00:10:53.919 --> 00:10:56.440 like Navidia who make the process and the video, yeah, 204 00:10:56.679 --> 00:10:59.159 the video, the processing chips, and then you have all 205 00:10:59.159 --> 00:11:01.519 these data sent being built and people are like, what's 206 00:11:01.559 --> 00:11:05.440 going on here? But using AI uses a lot of energy? 207 00:11:05.480 --> 00:11:07.879 I guess, So could you explain that how that works 208 00:11:07.960 --> 00:11:08.879 so people understand. 209 00:11:09.279 --> 00:11:11.159 Yeah, of course, So this is something we talk about 210 00:11:11.200 --> 00:11:14.440 in the book. The deal with AI is that it 211 00:11:14.720 --> 00:11:17.919 uses lots and lots of computing power both to make 212 00:11:18.360 --> 00:11:20.200 the AI do what you want in the first place. 213 00:11:20.240 --> 00:11:22.159 So training the AI, that is going through all those 214 00:11:22.159 --> 00:11:25.559 examples and having it learn the patterns of human language 215 00:11:25.600 --> 00:11:29.039 and how to answer questions and things like that. And 216 00:11:29.879 --> 00:11:33.279 when you're actually running an AI, it's a more complicated 217 00:11:33.320 --> 00:11:36.600 computer operation than something like older Google Search would have been. 218 00:11:37.320 --> 00:11:41.039 So you actually do need a pretty powerful computer somewhere 219 00:11:41.039 --> 00:11:43.399 in the world to be running that. And I don't 220 00:11:43.399 --> 00:11:47.480 want to oversell this. It's still using less energy than say, 221 00:11:47.679 --> 00:11:49.879 running a lawnmower or something like that. It's not going 222 00:11:49.919 --> 00:11:54.240 to be anything that's by itself. Each individual AI search 223 00:11:54.240 --> 00:11:56.039 isn't going to be a big problem. But once you 224 00:11:56.120 --> 00:12:00.159 have millions or billions of people using this thing, then 225 00:12:00.200 --> 00:12:02.519 it multiplies and then all of a sudden it's a problem. 226 00:12:02.600 --> 00:12:04.879 Now the electrical grid has to worry about can it 227 00:12:05.000 --> 00:12:08.120 power all these data centers and everything else that comes 228 00:12:08.120 --> 00:12:08.639 along with that. 229 00:12:09.240 --> 00:12:12.159 Now, as an expert in the area, what do you think, 230 00:12:12.200 --> 00:12:14.799 what's the answer to that? Because people are concerned, as 231 00:12:14.799 --> 00:12:18.120 you've seen all these data centers, it's all about surveillance. 232 00:12:18.279 --> 00:12:20.159 We're going to suck all the power. The grids are 233 00:12:20.200 --> 00:12:22.840 old that we can't protect them, et cetera. What do 234 00:12:22.919 --> 00:12:24.799 you see for SEA as a solution to this. 235 00:12:25.320 --> 00:12:28.159 As a solution to this, I'd say that we've known 236 00:12:28.200 --> 00:12:30.480 for a very long time that the power grid needs upgrading. 237 00:12:30.639 --> 00:12:34.200 It's just something that you know, every kind of person 238 00:12:34.519 --> 00:12:37.399 involved in the actual system says this is something that 239 00:12:37.440 --> 00:12:41.799 we need to work on, and so this is just 240 00:12:41.879 --> 00:12:44.759 adding fuel to that VIBA. Right. The more these data 241 00:12:44.799 --> 00:12:48.320 centers they use power, they use power basically twenty four 242 00:12:48.360 --> 00:12:50.720 to seven compared to some of these other sorts of things. 243 00:12:51.559 --> 00:12:53.559 And so I really think that part that's going to 244 00:12:53.559 --> 00:12:56.000 be making the grid smarter, making it have better energy 245 00:12:56.039 --> 00:12:58.879 storage so that you can take energy that's coming in 246 00:12:58.879 --> 00:13:01.799 the daytime or the nighttime and smooth it out. That 247 00:13:01.879 --> 00:13:05.559 also helps because there's a very specific problem where data 248 00:13:05.559 --> 00:13:08.679 centers need power that's organized very cleanly. So you know 249 00:13:08.720 --> 00:13:12.519 that alternating current that's going through your wires. If that 250 00:13:13.559 --> 00:13:15.679 shape that that current is making as a little jagged, 251 00:13:15.799 --> 00:13:17.960 then the data center can't really use it very well. 252 00:13:18.399 --> 00:13:20.279 So there's a specific thing they'll do is if that 253 00:13:20.679 --> 00:13:23.200 power just loses that quality to it, they'll switch over 254 00:13:23.240 --> 00:13:25.080 to backup power. And then all of a sudden, there's 255 00:13:25.159 --> 00:13:26.799 a bunch of extra energy on the grid, which is 256 00:13:26.799 --> 00:13:29.519 almost as big a problem, and it ends to be 257 00:13:29.600 --> 00:13:33.120 dumped somewhere. So basically all the same kinds of things 258 00:13:33.120 --> 00:13:34.639 people have been talking about for a while, you know, 259 00:13:34.639 --> 00:13:37.279 making sure we're having more efficient power plants and making 260 00:13:37.320 --> 00:13:40.200 sure that we have the energy storage. You know, solar 261 00:13:40.279 --> 00:13:42.000 is getting very cheap as much as it's kind of 262 00:13:42.000 --> 00:13:45.639 a meme and so so I think that's going to 263 00:13:45.679 --> 00:13:48.320 be the solution there. But then also there are some 264 00:13:48.360 --> 00:13:50.799 more technical ways to try and make AI more efficient 265 00:13:50.840 --> 00:13:52.879 on the chips that we have, which I could go into. 266 00:13:52.919 --> 00:13:55.799 Also, where do you see in terms of kind of 267 00:13:55.799 --> 00:13:59.720 the race for AI between the various big powers and 268 00:13:59.799 --> 00:14:01.559 kind of trees on the earth. Where do you see 269 00:14:01.559 --> 00:14:03.519 the US sitting? Are we in the lead? Are we 270 00:14:03.679 --> 00:14:07.039 number two? I mean, China obviously is highly evolved in 271 00:14:07.080 --> 00:14:09.480 the system. How about the other countries? Have they have 272 00:14:09.559 --> 00:14:11.480 they weigh behind? Or is it just the two of us? 273 00:14:11.600 --> 00:14:13.679 Or what do you see their agg. 274 00:14:13.679 --> 00:14:15.799 So it's pretty much just the US and China in 275 00:14:15.840 --> 00:14:19.120 the race at this point. There's a few European ones. 276 00:14:19.159 --> 00:14:22.080 There's some in the UK is involved in some I 277 00:14:22.080 --> 00:14:26.120 think there's Mysterroll that's in I believe France. But really 278 00:14:26.159 --> 00:14:28.519 it's the American companies and the Chinese companies that are 279 00:14:28.519 --> 00:14:30.480 in the lead. But I also want to stress it's 280 00:14:30.519 --> 00:14:33.200 not just all of America are raising against all of China. 281 00:14:33.360 --> 00:14:35.840 It's a few American companies. You've got Google, You've got 282 00:14:35.879 --> 00:14:38.960 open Ai, you've got Anthropic and they're raising each other 283 00:14:39.039 --> 00:14:40.960 as well as they're racing the Chinese companies. So you've 284 00:14:40.960 --> 00:14:43.600 got your q Win, you've got your Bidy and Ali Baba, 285 00:14:43.720 --> 00:14:47.039 all those guys. So it's not a monolithic thing in 286 00:14:47.039 --> 00:14:48.960 one or the other, I'd say that, But the US 287 00:14:49.039 --> 00:14:52.720 teams are, the US companies are in the lead right now. 288 00:14:52.759 --> 00:14:54.519 You know, the biggest news has been coming out of 289 00:14:54.720 --> 00:14:58.440 companies like Anthropic and Open Ai, especially with Claude Fable 290 00:14:58.480 --> 00:15:03.840 that's been in the news lately. So there's it's it's 291 00:15:03.840 --> 00:15:05.840 not an insurmountable lead. You know, there's a chance that 292 00:15:05.879 --> 00:15:07.919 depending on what we do, China could catch up. And 293 00:15:07.960 --> 00:15:11.559 I also want to stress that it's it's not a 294 00:15:11.679 --> 00:15:17.639 race in the sense that like if someone makes some 295 00:15:17.679 --> 00:15:19.679 AI system that's much better than everyone else's. We want 296 00:15:19.720 --> 00:15:21.240 it to be good, yes, but we also want it 297 00:15:21.279 --> 00:15:21.759 to be safe. 298 00:15:21.960 --> 00:15:24.960 M Okay, when you say safe, what do you mean. 299 00:15:26.360 --> 00:15:28.320 Well, there's a lot of dangers that come with AI. 300 00:15:29.039 --> 00:15:30.639 You know, again, there's a lot of positive there's also 301 00:15:30.759 --> 00:15:32.519 some dangers that we have to watch out for. So 302 00:15:32.600 --> 00:15:37.039 one thing that's recently been been in the news is cybersecurity. 303 00:15:37.600 --> 00:15:41.480 So using the models that are at the cutting edge 304 00:15:41.519 --> 00:15:46.720 of the AI systems to find exploits in the software 305 00:15:46.759 --> 00:15:50.799 that runs banks, or that runs important social media sites, 306 00:15:50.879 --> 00:15:53.840 or anything that has your data in it. And one 307 00:15:53.879 --> 00:15:55.960 of the issues that people are finding is that these 308 00:15:56.039 --> 00:15:58.679 ais are so good at finding bugs in this code 309 00:15:59.279 --> 00:16:01.919 that all of a sudden, the number of bugs that 310 00:16:01.960 --> 00:16:05.000 are known shoots up. So the responsible thing to do, 311 00:16:05.039 --> 00:16:07.639 which is thankfully what has been started to set it 312 00:16:07.679 --> 00:16:11.440 as a trend, is to give these models to the defenders. First, 313 00:16:11.519 --> 00:16:13.200 give it to the software companies, give it to the 314 00:16:13.200 --> 00:16:16.399 banks before you release to the general public. That way 315 00:16:16.440 --> 00:16:19.080 they can patch all the exploits that it defines. So 316 00:16:19.080 --> 00:16:21.879 cybersecurity is one issue. Another issue people talk about is 317 00:16:22.399 --> 00:16:24.759 synthetic biology is a little more speculative. But you know, 318 00:16:24.799 --> 00:16:30.360 maybe AIS could make people's work in biolabs more dangerous 319 00:16:30.360 --> 00:16:32.240 for people with less skill, could make something like a 320 00:16:32.320 --> 00:16:37.440 dangerous kind of germ. But then also in principle, if 321 00:16:37.440 --> 00:16:40.639 an AI gets smart enough, if it really wants something 322 00:16:40.679 --> 00:16:43.080 that's bad for humans, you know, something like turning the 323 00:16:43.080 --> 00:16:46.159 whole world into more computers, then we would have a 324 00:16:46.159 --> 00:16:46.720 problem with. 325 00:16:46.679 --> 00:16:50.600 That, right of course. Okay, yeah, my special guest on 326 00:16:50.879 --> 00:16:56.559 Guys Guys Radio today, agg Liu and his Agg's partner Ashwari. 327 00:16:56.879 --> 00:17:00.159 How are PhD put together in this wonderful book and 328 00:17:00.200 --> 00:17:03.919 it's called The Handy Artificial Intelligence answer Book. You folks 329 00:17:03.919 --> 00:17:07.240 did a great job with Italy. Fantastic and I'm reading 330 00:17:07.240 --> 00:17:09.319 it at the same time as my thirteen year old son, 331 00:17:09.359 --> 00:17:11.960 and we can just open up a question or ask 332 00:17:12.000 --> 00:17:13.880 a question, look it up and the answer is there. 333 00:17:13.920 --> 00:17:17.079 It's just it's really fantastic. And I urge everybody to 334 00:17:17.079 --> 00:17:19.039 get a copy of this because you really want to 335 00:17:19.160 --> 00:17:20.960 learn about this. This is going to be part of 336 00:17:20.960 --> 00:17:24.000 our lives and it's up to us how we use it. 337 00:17:24.359 --> 00:17:28.599 You mentioned agg finding little holes and things, and I 338 00:17:28.680 --> 00:17:32.680 was reading about medicine how artificial intelligence can help medicine 339 00:17:32.680 --> 00:17:36.440 where maybe they read a scan or something and they 340 00:17:36.440 --> 00:17:39.799 can pick things out ahead of time that traditionally that 341 00:17:39.880 --> 00:17:42.160 the physicians may not be able to see. Talk to 342 00:17:42.200 --> 00:17:43.759 us a little bit about that, because that's a really 343 00:17:43.799 --> 00:17:46.240 good thing, and is there an issue with that that 344 00:17:46.279 --> 00:17:48.000 maybe because sometimes AI is wrong. 345 00:17:48.759 --> 00:17:50.720 So yeah, So I'd say bedisine is one of the 346 00:17:50.759 --> 00:17:53.119 areas people are most optimistic about when it comes to AI. 347 00:17:53.279 --> 00:17:55.119 This is one of the areas where we think it's 348 00:17:55.160 --> 00:17:57.279 going to be a great tool. It's already starting to 349 00:17:57.319 --> 00:18:01.240 be a great tool because is an area where your 350 00:18:01.279 --> 00:18:04.079 body is so complicated. There's so many little systems. There's 351 00:18:04.079 --> 00:18:07.559 so much that could be changed by something like a 352 00:18:07.559 --> 00:18:09.440 medicine or something by a treatment like a surgery or 353 00:18:09.440 --> 00:18:13.519 something like that. And having a machine like an AI 354 00:18:13.599 --> 00:18:16.559 that's very good at finding patterns and data that's sort 355 00:18:16.559 --> 00:18:20.839 of its fundamental job, makes it a lot easier in 356 00:18:20.880 --> 00:18:24.880 principle at least to diagnose things. Like they could work 357 00:18:25.079 --> 00:18:28.880 with a radiologist looking at an X ray and see, okay, 358 00:18:29.119 --> 00:18:33.160 maybe I can flag little points on the X ray 359 00:18:33.200 --> 00:18:35.160 image that the radiolgists might want to look at further. 360 00:18:36.119 --> 00:18:39.839 Or it could help look at what your specific symptoms are, 361 00:18:39.920 --> 00:18:42.880 what your specific test results are, and find is there 362 00:18:42.920 --> 00:18:45.720 a drug out there in the world already that can 363 00:18:45.839 --> 00:18:49.000 treat your disease maybe better than it would be able 364 00:18:49.000 --> 00:18:50.839 to treat someone else with a vaguely similar but not 365 00:18:50.880 --> 00:18:54.559 quite the same disease. Already in clinical trials, you know, 366 00:18:54.559 --> 00:18:57.000 people are testing medicines that AI has come up with. 367 00:18:57.799 --> 00:19:01.640 The AIS searched through all the possible chemicals and seen 368 00:19:02.000 --> 00:19:04.599 what might actually be helpful to our bodies that are 369 00:19:04.680 --> 00:19:05.240 hurtful to it. 370 00:19:05.920 --> 00:19:09.039 At that same on the same note, a g G 371 00:19:10.400 --> 00:19:16.519 is a AI being used to find holistic cures, if 372 00:19:16.559 --> 00:19:20.599 you will, for maladies that people have. Because I was 373 00:19:20.599 --> 00:19:23.599 watching the news this morning and one of the guys, 374 00:19:23.640 --> 00:19:26.160 some science or you know, medical expert, came on and 375 00:19:26.319 --> 00:19:29.640 was touting AI developed vaccines. Now, depending on how you 376 00:19:29.680 --> 00:19:32.640 feel about vaccines, some people yeah, no way and uh 377 00:19:33.000 --> 00:19:34.880 and now they're going to use this against us, and 378 00:19:34.920 --> 00:19:37.759 other people are like, oh, that's great. What's your take 379 00:19:37.799 --> 00:19:40.880 on AI being I guess a lot of it depends 380 00:19:40.880 --> 00:19:46.119 on who's behind challenging AI to come up with stuff, 381 00:19:46.160 --> 00:19:48.680 and so I guess the question is how's that being done? 382 00:19:49.000 --> 00:19:51.480 And are they looking for natural stuff as well as 383 00:19:51.839 --> 00:19:54.039 you know, farmer based solutions When it comes. 384 00:19:53.920 --> 00:19:56.279 To of course, of course, you know, I'd say there's 385 00:19:56.319 --> 00:19:59.680 there's nothing saying AI has to do one exact approach 386 00:19:59.680 --> 00:20:02.599 to META. You know, there's nothing in the systems that 387 00:20:02.640 --> 00:20:05.079 are saying it has to follow this farming based approach. 388 00:20:05.079 --> 00:20:08.119 The way you're talking about, you know, people are talking 389 00:20:08.119 --> 00:20:10.000 about using it to help people's mental health, to help 390 00:20:10.000 --> 00:20:14.119 people's wellness in their exercise programs or in their general lifestyles, 391 00:20:14.200 --> 00:20:16.000 or you could even look at it, you know, to 392 00:20:16.039 --> 00:20:18.559 find remedies in nature that kind of stuff. You know, 393 00:20:18.680 --> 00:20:21.640 are there plants out there or are their animals or 394 00:20:21.680 --> 00:20:24.799 other sorts of you know, mushrooms I don't know that 395 00:20:24.920 --> 00:20:29.160 have have these these useful properties, you know. And so again, 396 00:20:29.240 --> 00:20:31.599 AI is a very good pattern matching tool. It can 397 00:20:31.680 --> 00:20:34.960 find patterns and data, it can find information that you 398 00:20:35.039 --> 00:20:39.480 might overlook. And so yeah, I'd say definitely there's a 399 00:20:39.480 --> 00:20:42.960 great capability to help with with wellness beyond just targeted 400 00:20:43.000 --> 00:20:44.559 drugs and surgeries and things like that. 401 00:20:44.839 --> 00:20:48.319