WEBVTT 1 00:00:00.080 --> 00:00:02.080 You know, when you look at your phone or your laptop, 2 00:00:02.799 --> 00:00:06.639 there is this immense expectation of simple matters. 3 00:00:06.799 --> 00:00:07.360 Absolutely. 4 00:00:07.440 --> 00:00:09.320 Did you tap a piece of glass and a movie 5 00:00:09.320 --> 00:00:12.160 plays in high definition? Or you stretch a watch to 6 00:00:12.199 --> 00:00:14.800 your wrist and it tells you your heart rate, your 7 00:00:14.919 --> 00:00:18.480 sleep quality, how many steps you've taken. 8 00:00:18.640 --> 00:00:20.559 Right, it feels entirely clean. 9 00:00:20.960 --> 00:00:21.960 Yeah, it feels simple. 10 00:00:22.079 --> 00:00:24.839 It does. We're sort of conditioned to expect things to 11 00:00:25.000 --> 00:00:28.679 just work completely independent of the actual mechanics. I mean, 12 00:00:28.719 --> 00:00:32.880 it's highly comforting to interact with a smooth, frictionless surface. 13 00:00:33.119 --> 00:00:36.119 But then you peak behind the curtain, you look into 14 00:00:36.200 --> 00:00:39.840 the actual world of data, analytics and intelligent systems, and 15 00:00:39.960 --> 00:00:44.280 suddenly that simple magic reveals itself as well as an 16 00:00:44.359 --> 00:00:47.880 unbelievably complex, massive, humming web. 17 00:00:47.640 --> 00:00:51.359 Of decisions, decisions being made in literally milliseconds. 18 00:00:51.439 --> 00:00:51.920 Exactly. 19 00:00:52.039 --> 00:00:55.560 It is the absolute definition of an invisible infrastructure. We 20 00:00:55.640 --> 00:00:59.320 are essentially living inside a global nervous system. 21 00:00:59.320 --> 00:01:01.399 The nervous It's constantly taking in data. 22 00:01:01.280 --> 00:01:04.799 Right right, processing it, and then fundamentally altering our physical 23 00:01:04.840 --> 00:01:06.359 reality based on that data. 24 00:01:06.519 --> 00:01:09.079 Welcome to the deep dive. If you're joining us, you 25 00:01:09.200 --> 00:01:12.040 probably have this deep seated curiosity about how the world 26 00:01:12.040 --> 00:01:12.719 actually works. 27 00:01:12.879 --> 00:01:15.159 But you also probably don't want to be buried under 28 00:01:15.200 --> 00:01:17.200 an avalanche of computer science jargon. 29 00:01:17.319 --> 00:01:20.680 No, definitely not. You want the thorough knowledge, you know, 30 00:01:20.719 --> 00:01:23.799 you want those aha moments, but without the headache of 31 00:01:23.959 --> 00:01:27.959 info overload. Right, and today we've got a truly fascinating 32 00:01:28.000 --> 00:01:30.680 stack of sources for you. We're looking at a collection 33 00:01:30.760 --> 00:01:34.920 of research from the twenty nineteen International Conference on Cybersecurity, 34 00:01:34.959 --> 00:01:37.040 Intelligence and Analytics. 35 00:01:36.680 --> 00:01:40.280 Published by Springer. And admittedly that title sounds like the 36 00:01:40.359 --> 00:01:43.319 kind of dense computer science textbook that might just put 37 00:01:43.359 --> 00:01:44.359 you straight to sleep. 38 00:01:44.560 --> 00:01:47.280 It really does sound intimidating. But the mission of this 39 00:01:47.400 --> 00:01:50.280 deep dive is absolutely not to get bogged down in 40 00:01:50.359 --> 00:01:54.359 lines of code or complex acronyms. 41 00:01:54.599 --> 00:01:55.239 No, not at all. 42 00:01:55.359 --> 00:01:57.400 Okay, let's unpack this. We are going to frame everything 43 00:01:57.439 --> 00:01:58.560 around one central. 44 00:01:58.319 --> 00:02:01.799 Question today, which is how this invisible layer of data 45 00:02:01.959 --> 00:02:05.680 acting as our world's nervous system actually makes decisions for us. 46 00:02:06.000 --> 00:02:09.280 Right, We're going to uncover how these intelligent systems are 47 00:02:09.400 --> 00:02:13.280 quietly revolutionizing wildly different aspects of your daily. 48 00:02:13.080 --> 00:02:15.840 Life, starting from the intimate space of the human. 49 00:02:15.599 --> 00:02:17.680 Body, scaling all the way up to city grids. 50 00:02:17.800 --> 00:02:19.599 Looking at the invisible plumbing that makes it. 51 00:02:19.520 --> 00:02:23.560 All work, and finally, how it is completely rewiring the 52 00:02:23.560 --> 00:02:24.199 way we learn. 53 00:02:24.680 --> 00:02:27.879 The goal here is to spot the hidden patterns. I mean, 54 00:02:27.960 --> 00:02:31.159 we want to find the connective tissue between a surgeon's 55 00:02:31.199 --> 00:02:36.080 operating table, a city planner's desk, and a modern classroom. 56 00:02:35.680 --> 00:02:39.639 Because the underlying logic driving all of them is surprisingly similar. 57 00:02:39.759 --> 00:02:40.360 It really is. 58 00:02:40.560 --> 00:02:43.240 So if we're talking about a nervous system, let's start 59 00:02:43.240 --> 00:02:44.680 with our actual physical bodies. 60 00:02:44.759 --> 00:02:45.280 Makes sense. 61 00:02:45.439 --> 00:02:49.240 The research dives into how urology and CT scans have 62 00:02:49.360 --> 00:02:52.759 fundamentally evolved. Historically. You know, if you went in for 63 00:02:52.800 --> 00:02:55.759 a scan, the machine gave the doctor a flat, two 64 00:02:55.800 --> 00:02:57.159 dimensional image. 65 00:02:56.800 --> 00:02:59.719 Right, it was basically a sophisticated shadow exactly. 66 00:03:00.120 --> 00:03:03.439 But now algorithms take those flat slices from spiral CT 67 00:03:03.639 --> 00:03:06.960 scans and run them through rendering techniques things. 68 00:03:06.719 --> 00:03:10.719 Like volume rendering or VR, and maximum intensity projection. 69 00:03:10.800 --> 00:03:13.919 MIP right and multiplanar reconstruction yeah NPR. 70 00:03:14.159 --> 00:03:17.120 And what those do is build a fully navigable three 71 00:03:17.199 --> 00:03:19.360 dimensional landscape of your insides. 72 00:03:19.919 --> 00:03:23.120 What's fascinating here is this shift in perspective. The system 73 00:03:23.159 --> 00:03:24.080 isn't just taking a. 74 00:03:24.039 --> 00:03:27.879 Picture anymore, right, It is constructing a spatial reality. The 75 00:03:27.960 --> 00:03:33.719 algorithm actually calculates depth, density tissue relationships. 76 00:03:33.159 --> 00:03:35.719 Allowing a surgeon to essentially fly through a digital twin 77 00:03:35.759 --> 00:03:37.000 of your organs exactly. 78 00:03:37.080 --> 00:03:40.039 And practically speaking, if a patient have a kidney stone, 79 00:03:40.199 --> 00:03:42.639 the surgeon isn't just guessing its shape based on some 80 00:03:42.800 --> 00:03:43.919 blurry gray spot. 81 00:03:44.280 --> 00:03:46.960 Yeah. Using this intelligent three D rendering, they can see 82 00:03:47.000 --> 00:03:49.199 the exact size, the jagged edges. 83 00:03:49.159 --> 00:03:52.360 And precisely how that stone sits in relation to surrounding 84 00:03:52.400 --> 00:03:55.280 tissue or even surgical implants like double J. 85 00:03:55.360 --> 00:03:57.840 Two wow, before they ever pick up a scalpel. 86 00:03:57.520 --> 00:04:02.599 Exactly, which turns a highly un predictable physical exploration into 87 00:04:03.280 --> 00:04:05.879 a precise, mathematically mapped mission. 88 00:04:06.120 --> 00:04:06.800 It's wild. 89 00:04:07.080 --> 00:04:10.080 And this data driven approach isn't just for acute immediate 90 00:04:10.120 --> 00:04:13.360 surgical issues. The same logic of treating the body as 91 00:04:13.360 --> 00:04:16.759 a continuous data landscape is being applied to preventative care. 92 00:04:17.160 --> 00:04:20.079 Right. The research mentioned systems now that marry the predictive 93 00:04:20.079 --> 00:04:24.879 principles of traditional Chinese medicine with continuous wireless sensor networks. 94 00:04:24.639 --> 00:04:25.839 To monitor community health. 95 00:04:25.879 --> 00:04:28.360 Yeah, and they use zigbie technology for this, right Yeah, 96 00:04:28.480 --> 00:04:31.639 based on the IE eight H two point one five 97 00:04:31.680 --> 00:04:33.319 point four standard. 98 00:04:33.000 --> 00:04:36.279 Yes, And this part is so clever because of the 99 00:04:36.319 --> 00:04:37.319 hardware constraints. 100 00:04:37.319 --> 00:04:40.920 It solves because if you're monitoring a patient continuously, you 101 00:04:40.959 --> 00:04:43.160 can't ask them to plug themselves into a wall. 102 00:04:43.519 --> 00:04:46.240 No, and you can't have them changing watch batteries every 103 00:04:46.240 --> 00:04:47.480 six hours either, right. 104 00:04:47.920 --> 00:04:51.240 So they use a specialized low power network protocol that 105 00:04:51.360 --> 00:04:53.720 relies on three distinct types of nodes. 106 00:04:53.879 --> 00:04:56.800 You have a coordinator, a router, and a terminal device. 107 00:04:57.199 --> 00:04:59.920 And to understand why that specific division of labor is read, 108 00:05:01.199 --> 00:05:04.560 you have to look at how data collection usually drains power. 109 00:05:04.879 --> 00:05:08.319 Normally, a device is constantly pinging a network saying here's 110 00:05:08.360 --> 00:05:10.480 my data, here is my data, which. 111 00:05:10.319 --> 00:05:13.560 Just burns through a battery in days exactly. So let's 112 00:05:13.639 --> 00:05:16.600 use an analogy here. Think of this wireless sensor setup 113 00:05:16.800 --> 00:05:19.480 as a highly efficient neighborhood watch program. 114 00:05:19.560 --> 00:05:20.000 I like that. 115 00:05:20.399 --> 00:05:23.279 The coordinator builds the network, It defines the borders of 116 00:05:23.279 --> 00:05:27.040 the neighborhood. The router acts as the messenger passing information 117 00:05:27.079 --> 00:05:30.800 along the streets. Okay, But the terminal device, the actual 118 00:05:30.879 --> 00:05:34.160 sensor attached to the patient, it doesn't waste its energy 119 00:05:34.199 --> 00:05:37.040 patrolling the streets. Twenty four seven. It sleeps, it sleeps, 120 00:05:37.279 --> 00:05:41.720 It just sits there completely dormant, conserving power while quietly 121 00:05:41.759 --> 00:05:44.480 tracking your pulse, your blood pressure, in your temperature. 122 00:05:44.519 --> 00:05:47.560 And because it is sleeping, it isn't transmitting. Transmission is 123 00:05:47.560 --> 00:05:49.000 what costs all the energy. 124 00:05:49.120 --> 00:05:51.920 Exactly. It only wakes up the very second it spots 125 00:05:51.920 --> 00:05:52.800 a target anomaly. 126 00:05:52.879 --> 00:05:55.519 So if it sees something suspicious in your vital signs 127 00:05:55.600 --> 00:05:56.920 that breaks the normal. 128 00:05:56.639 --> 00:06:00.680 Pattern, it instantly snaps awake, yeah, and raises the alarm 129 00:06:00.959 --> 00:06:04.839 through the routers. And because it utilizes that deep sleep mode, 130 00:06:05.319 --> 00:06:08.279 the battery on that tiny terminal device can last up 131 00:06:08.319 --> 00:06:08.920 to two years. 132 00:06:08.959 --> 00:06:13.360 Two years, that's incredible. What's fascinating here is if we 133 00:06:13.399 --> 00:06:16.120 connect this to the bigger picture. Treating the human body 134 00:06:16.199 --> 00:06:21.680 as a continuous low energy beta stream fundamentally shifts healthcare. 135 00:06:21.279 --> 00:06:23.680 From a reactive discipline to a proactive one. 136 00:06:23.839 --> 00:06:26.759 Yes, we stop waiting for the machine to break down. 137 00:06:27.399 --> 00:06:31.480 By monitoring these specific parameters over a long timeline, the 138 00:06:31.519 --> 00:06:35.720 intelligent system can predict the residual life of your physiological structures. 139 00:06:35.800 --> 00:06:38.199 It catches the anomaly incredibly. 140 00:06:37.680 --> 00:06:42.800 Early, predicting risks based on those traditional medicinal patterns long 141 00:06:42.879 --> 00:06:44.839 before you ever feel a symptom. 142 00:06:45.160 --> 00:06:48.920 It's essentially a tiny invisible doctor living on your wrist, 143 00:06:49.360 --> 00:06:52.720 patiently watching your vitals for two years straight without needing 144 00:06:52.759 --> 00:06:53.399 a coffee break. 145 00:06:53.439 --> 00:06:54.439 That's a great way to put it. 146 00:06:54.639 --> 00:06:58.079 But this raises a massive logistical issue. Oh, if we 147 00:06:58.120 --> 00:07:00.920 can use sensor networks to map them micro environment of 148 00:07:00.959 --> 00:07:04.480 a single human body, what happens when we apply that 149 00:07:04.560 --> 00:07:09.040 exact same intelligent mapping to the macro environment? Ah, what 150 00:07:09.120 --> 00:07:11.920 happens when we scale this up to entire cities and economies? 151 00:07:12.120 --> 00:07:13.920 The volume of data must be staggering. 152 00:07:14.199 --> 00:07:17.680 The volume explodes into the terabytes easily, which brings us 153 00:07:17.680 --> 00:07:19.040 to the mechanics of data mining. 154 00:07:19.199 --> 00:07:19.519 Okay. 155 00:07:19.600 --> 00:07:22.959 The foundational concept here is called knowledge discovery in database 156 00:07:23.040 --> 00:07:26.839 or KDD, and interestingly, this idea was first proposed in 157 00:07:26.879 --> 00:07:28.959 Detroit way back in nineteen eighty nine. 158 00:07:29.319 --> 00:07:32.240 Nineteen eighty nine, that is wild to me. We barely 159 00:07:32.279 --> 00:07:32.680 had the. 160 00:07:32.560 --> 00:07:37.319 Internet then, right, But they were already conceptualizing mining massive databases. 161 00:07:37.399 --> 00:07:40.560 So the conceptual framework existed, But today we actually have 162 00:07:40.639 --> 00:07:43.000 the computing power to execute it exactly. 163 00:07:43.360 --> 00:07:46.120 The core mechanism of data mining isn't just using a 164 00:07:46.160 --> 00:07:48.920 search bar to find a specific number. It is about 165 00:07:48.959 --> 00:07:55.560 extracting unknown, potentially highly valuable correlations from oceans of data. 166 00:07:55.120 --> 00:07:56.879 Like GB and TB levels of data. 167 00:07:57.120 --> 00:08:01.360 Yes, and it processes structured data like spreadsheets of financial 168 00:08:01.439 --> 00:08:04.240 numbers alongside completely unstructured data. 169 00:08:04.120 --> 00:08:07.800 Things like text documents, video files, clicks on a website. 170 00:08:07.360 --> 00:08:08.720 And geographical coordinates. 171 00:08:09.040 --> 00:08:12.079 The goal being to help city planners and economic decision 172 00:08:12.079 --> 00:08:15.680 makers shift away from making intuitive choices. Right. You know, 173 00:08:15.959 --> 00:08:18.319 I have a gut feeling this neighborhood needs a new 174 00:08:18.360 --> 00:08:18.920 power grid. 175 00:08:19.160 --> 00:08:21.920 Intuition is great for picking a restaurant, but it's terrible 176 00:08:21.920 --> 00:08:23.040 for urban infrastructure. 177 00:08:23.199 --> 00:08:26.160 Precisely, building a power grid in a new type urban 178 00:08:26.199 --> 00:08:31.079 area is massively expensive, highly dangerous, and largely permanent. 179 00:08:31.240 --> 00:08:34.639 You need a comprehensive evaluation system that turns a messy 180 00:08:34.679 --> 00:08:36.320 city into a solvable equation. 181 00:08:36.759 --> 00:08:40.000 And the research outlines how planners do this using something 182 00:08:40.080 --> 00:08:43.279 called the analytic hierarchy process or HP. 183 00:08:43.879 --> 00:08:47.440 Yes, and they weigh thirty four specific indicators across four 184 00:08:47.480 --> 00:08:48.960 main categories. 185 00:08:48.759 --> 00:08:52.320 Technical, economic, adaptability, and social resources. 186 00:08:52.799 --> 00:08:55.759 And they score these urban grid systems from level five, 187 00:08:55.840 --> 00:08:59.559 which is zero to sixty points basically a failing grade. 188 00:08:59.360 --> 00:09:01.799 Up to level one one ninety two one hundred points, 189 00:09:02.120 --> 00:09:03.120 the optimal. 190 00:09:02.720 --> 00:09:05.200 Grade based on how much power they save, and how 191 00:09:05.200 --> 00:09:07.600 intensively and efficiently they are arranged. 192 00:09:07.720 --> 00:09:10.399 And to process all these opinions and data points, they 193 00:09:10.480 --> 00:09:12.200 use something called the Delphi method. 194 00:09:12.399 --> 00:09:15.919 The Delphi method is a brilliant mechanism for consensus. How 195 00:09:15.960 --> 00:09:18.200 so well, Instead of putting a bunch of experts in 196 00:09:18.200 --> 00:09:20.960 a room where the loudest person usually wins the argument, 197 00:09:21.279 --> 00:09:25.000 the Delki method relies on structured anonymous communication. 198 00:09:25.279 --> 00:09:25.919 Oh that's smart. 199 00:09:26.000 --> 00:09:29.840 Experts answer questionnaires independently. The system aggregates the answers and 200 00:09:29.840 --> 00:09:33.039 then feeds that summary back to the experts for another round. 201 00:09:33.159 --> 00:09:36.000 So it iterates until the group converges on a mathematically 202 00:09:36.039 --> 00:09:37.759 sound consensus. 203 00:09:37.120 --> 00:09:39.360 Without ever being influenced by social pressure. 204 00:09:39.559 --> 00:09:41.720 That makes a lot of sense for the technical indicators 205 00:09:42.039 --> 00:09:45.120 like measuring voltage drops or channel costs. Sure, but I 206 00:09:45.200 --> 00:09:47.679 have to push back on one of the mathematical techniques 207 00:09:47.720 --> 00:09:48.200 mentioned here. 208 00:09:48.240 --> 00:09:49.039 Okay, what is it? 209 00:09:49.240 --> 00:09:52.879 The system uses something called fuzzy membership to determine the 210 00:09:52.919 --> 00:09:56.399 final score of a power grid. Ah, yes, wait, how 211 00:09:56.480 --> 00:10:00.919 does fuzzy math help build a concrete, highly dangerous selectrical grid. 212 00:10:01.600 --> 00:10:05.120 Fuzzy is literally the last word I want associated with 213 00:10:05.440 --> 00:10:07.679 high voltage power lines running past my house. 214 00:10:07.759 --> 00:10:11.879 I know it sounds incredibly contradictory, but fuzzy membership in 215 00:10:11.919 --> 00:10:15.480 mathematics is not about guessing, and it's not about being inaccurate. 216 00:10:16.039 --> 00:10:20.519 It is actually a highly precise mathematical language designed to 217 00:10:20.639 --> 00:10:24.120 deal with variables that are inherently not black and white. 218 00:10:24.360 --> 00:10:27.360 Give me an example of a variable that isn't black 219 00:10:27.360 --> 00:10:29.120 and white in a power grid. I mean, is a 220 00:10:29.120 --> 00:10:30.720 wire either safe or unsafe? 221 00:10:30.840 --> 00:10:34.399 The wires are binary? Yes, But remember those thirty four indicators. 222 00:10:34.720 --> 00:10:38.480 Some of them fall under the social resources or adaptability categories. 223 00:10:39.000 --> 00:10:42.559 How do you strictly quantify public attitude toward a massive 224 00:10:42.600 --> 00:10:43.279 new substation? 225 00:10:43.519 --> 00:10:44.039 Oh? I see? 226 00:10:44.320 --> 00:10:46.720 Or how do you measure the visual synergy between an 227 00:10:46.759 --> 00:10:50.000 industrial power cabin and the width of the residential road 228 00:10:50.039 --> 00:10:50.480 next to it? 229 00:10:50.720 --> 00:10:54.840 Ah, Human opinions and esthetic impacts aren't binary. A building 230 00:10:54.919 --> 00:10:56.840 isn't just a zero for ugly and a one for 231 00:10:56.919 --> 00:10:58.000 beautiful exactly. 232 00:10:58.320 --> 00:11:02.080 Traditional computers hate gray air. They only understand ones and zeros. 233 00:11:02.480 --> 00:11:06.000 But fuzzy membership allows the intelligent system to assign degrees 234 00:11:06.039 --> 00:11:06.480 of truth. 235 00:11:06.679 --> 00:11:09.320 So public approval might zero point seven, right. 236 00:11:09.360 --> 00:11:12.559 The visual synergy might be a point four. It translates 237 00:11:12.639 --> 00:11:16.279 the subjective gray areas of human society into a strict 238 00:11:16.399 --> 00:11:19.879 numerical format that the algorithm can actually calculate. 239 00:11:19.639 --> 00:11:21.919 Right alongside the hard, concrete economics. 240 00:11:22.000 --> 00:11:25.039 It ensures the resulting power grid isn't just technically sound, 241 00:11:25.399 --> 00:11:28.399 but socially adaptable to the humans living around it. 242 00:11:28.519 --> 00:11:31.399 That completely flips my understanding of the word fuzzy. It's 243 00:11:31.440 --> 00:11:34.720 actually a tool for extreme precision when dealing with messy 244 00:11:35.120 --> 00:11:35.960 human reality. 245 00:11:36.080 --> 00:11:36.559 Exactly. 246 00:11:36.639 --> 00:11:38.919 So, we've talked about all this massive data. Now we've 247 00:11:38.960 --> 00:11:42.960 got three D medical images of organs, terabytes of unstructured 248 00:11:43.000 --> 00:11:44.720 economic databases. 249 00:11:44.360 --> 00:11:47.679 Huge city greed plans that incorporate the fuzzy math of 250 00:11:47.759 --> 00:11:48.480 human opinion. 251 00:11:48.879 --> 00:11:52.519 But from a purely physical standpoint, how do we actually 252 00:11:52.559 --> 00:11:54.519 transmit all this heavy data quickly? 253 00:11:55.000 --> 00:11:56.120 That is the big question. 254 00:11:56.399 --> 00:11:58.840 Why doesn't the Internet just buckle and break under the 255 00:11:58.840 --> 00:12:01.879 sheer weight of sending billions of three D renders and 256 00:12:01.960 --> 00:12:04.679 city grids across the globe every second? 257 00:12:05.000 --> 00:12:08.200 That is the ultimate bottleneck problem, and to solve it 258 00:12:08.279 --> 00:12:10.559 we have to look under the hood at the invisible 259 00:12:10.600 --> 00:12:12.759 mechanics of transmission. 260 00:12:12.360 --> 00:12:14.639 Specifically image processing and compression. 261 00:12:14.960 --> 00:12:19.279 Without these compression algorithms, the modern digital world would literally 262 00:12:19.799 --> 00:12:22.799 grind to a halt. We would be stuck waiting hours 263 00:12:22.840 --> 00:12:24.399 for a single web page to load. 264 00:12:24.679 --> 00:12:27.960 The research breaks down how computer desktop image compression works, 265 00:12:28.240 --> 00:12:29.840 and the mechanism is fascinating. 266 00:12:29.919 --> 00:12:30.600 Yeah, it really is. 267 00:12:30.840 --> 00:12:34.320 The algorithm takes your computer screen and divides it into tiny, 268 00:12:34.480 --> 00:12:39.000 microscopic blocks, usually sixteen by sixteen pixels that don't overlap. 269 00:12:39.159 --> 00:12:43.200 Right. Then it uses clustering algorithms things like partition level, 270 00:12:43.279 --> 00:12:47.120 density grid or model algorithms to inspect each block. 271 00:12:46.879 --> 00:12:48.919 To identify exactly what is inside of it. 272 00:12:49.039 --> 00:12:52.200 This sorting phase is crucial. The algorithm has to act 273 00:12:52.240 --> 00:12:55.279 like a highly trained postal worker sorting mail. It needs 274 00:12:55.320 --> 00:12:59.600 to know is this sixteen by sixteen block mostly texts 275 00:12:59.600 --> 00:13:03.440 and sharp graphics? Is it a natural photographic image or 276 00:13:03.519 --> 00:13:05.240 is it a complex mix of both? 277 00:13:05.559 --> 00:13:07.559 And the reason it sorts them is because it treats 278 00:13:07.600 --> 00:13:11.559 them entirely differently. Yes, if the block contains text, say 279 00:13:12.080 --> 00:13:16.440 a word document you are typing, this system applies lossless compression. 280 00:13:16.279 --> 00:13:18.759 Meaning absolutely no data is thrown. 281 00:13:18.559 --> 00:13:20.559 Away, which is why the edges of the letters on 282 00:13:20.600 --> 00:13:23.639 your screen stay razor sharp and legible exactly. But if 283 00:13:23.679 --> 00:13:26.320 the block is a natural image, like a photograph of 284 00:13:26.320 --> 00:13:30.000 a cloudy sky on your desktop background. It uses lossy 285 00:13:30.039 --> 00:13:31.320 photo compression. 286 00:13:30.879 --> 00:13:34.679 Techniques, things like H two sixty four interprediction or discrete 287 00:13:34.759 --> 00:13:37.200 cosine transform DCT. Right. 288 00:13:37.399 --> 00:13:39.320 Because a cloud does not have razor. 289 00:13:39.039 --> 00:13:42.360 Sharp edges, the system can mathematically average out the colors, 290 00:13:42.360 --> 00:13:45.000 effectively throwing away a tiny bit of the pixel data, 291 00:13:45.039 --> 00:13:46.720 and your eye will never notice the difference. 292 00:13:46.799 --> 00:13:49.720 Here's where it gets really interesting. The way they handle 293 00:13:49.840 --> 00:13:53.120 larger complex images. Oh yeah, they use a technique called 294 00:13:53.200 --> 00:13:56.120 wavelet transforms. And when I was trying to wrap my 295 00:13:56.159 --> 00:13:59.440 head around how a wavelet transform works, I realized it 296 00:13:59.519 --> 00:14:02.399 is exactly like a music producer's equalizer board in a 297 00:14:02.440 --> 00:14:03.240 recording studio. 298 00:14:03.399 --> 00:14:05.360 I really like where this is going. Walk us through 299 00:14:05.399 --> 00:14:07.120 how an image is like an audio track. 300 00:14:07.200 --> 00:14:10.519 Okay, So a wavelet transform technique takes an image and 301 00:14:10.639 --> 00:14:13.440 splits it into four sub bands. Right. The main one 302 00:14:13.600 --> 00:14:16.759 is the LL band, which stands for low frequency. This 303 00:14:16.919 --> 00:14:21.200 is the approximate overarching image. In our music analogy. The 304 00:14:21.519 --> 00:14:24.679 L band is your bass. It is the heavy foundational 305 00:14:24.759 --> 00:14:27.720 rhythm of the song. In an image, it's the broad 306 00:14:27.759 --> 00:14:29.080 strokes of shape and color. 307 00:14:29.240 --> 00:14:29.799 Makes sense. 308 00:14:30.120 --> 00:14:33.159 Then you have three other bands that handle horizontal, vertical, 309 00:14:33.159 --> 00:14:36.720 and diagonal details. HL LH and HH. 310 00:14:36.799 --> 00:14:38.200 These are the high frequency bands. 311 00:14:38.200 --> 00:14:41.039 This is your treble. It's the sharp edges, the fine static, 312 00:14:41.159 --> 00:14:43.559 the tiny textural details, And just. 313 00:14:43.519 --> 00:14:46.120 Like in a master audio track, the vast majority of 314 00:14:46.120 --> 00:14:49.399 the actual energy and identity of the file is concentrated 315 00:14:49.440 --> 00:14:52.360 in that foundation, the low frequency base, the LL band. 316 00:14:52.679 --> 00:14:55.759 Right, So the algorithm isolates the bass from the treble, 317 00:14:56.399 --> 00:14:59.039 and because most of the image's core identity is living 318 00:14:59.120 --> 00:15:02.720 in the base, the system can aggressively compress or essentially 319 00:15:03.120 --> 00:15:04.679 turn down the volume on the treble. 320 00:15:04.759 --> 00:15:08.279 It uses hard or soft mathematical thresholds to filter out 321 00:15:08.279 --> 00:15:09.399 that high frequency data. 322 00:15:09.519 --> 00:15:12.559 You end up saving massive amounts of digital file space 323 00:15:12.600 --> 00:15:15.840 without ruining the song or, in this case, the image. 324 00:15:15.480 --> 00:15:17.759 Which is a beautiful way to understand it, and it 325 00:15:17.840 --> 00:15:21.200 highlights the invisible magic trick these systems are playing on us. 326 00:15:21.279 --> 00:15:23.879 What do you mean, Well, the engineer is building these 327 00:15:23.879 --> 00:15:28.440 algorithms deeply understand human biology. The human eye is biologically 328 00:15:28.440 --> 00:15:31.480 wired to be highly sensitive to low frequency components. 329 00:15:31.679 --> 00:15:34.000 Right. We need to see the broad strokes of light 330 00:15:34.120 --> 00:15:37.360 and shadow to recognize a face or a landscape. 331 00:15:37.440 --> 00:15:40.679 But our visual system is incredibly forgiving of missing high 332 00:15:40.720 --> 00:15:43.879 frequency data. We simply do not have the processing power 333 00:15:43.919 --> 00:15:47.360 in our brains to notice every single microscopic edge or 334 00:15:47.440 --> 00:15:48.440 grain of static. 335 00:15:49.000 --> 00:15:52.120 So the computers are literally exploiting our biological blind spots. 336 00:15:52.200 --> 00:15:52.919 They really are. 337 00:15:53.200 --> 00:15:55.120 They know what we can't see, so they delete it 338 00:15:55.159 --> 00:15:56.519 to save Internet bandwidth. 339 00:15:56.720 --> 00:16:00.519 Exactly. It is technological sleight of hand, and the hardware 340 00:16:00.559 --> 00:16:04.279 executing these tricks is getting impossibly fast. Right. 341 00:16:04.440 --> 00:16:06.960 The research mentions infrared processing hardware. 342 00:16:07.080 --> 00:16:11.039 Yes. To process complex data like real time infrared video, 343 00:16:11.039 --> 00:16:14.279 where contrast is terrible and noises, high systems use a 344 00:16:14.360 --> 00:16:18.039 dual chip setup. Okay, They pair a DSP, a digital 345 00:16:18.559 --> 00:16:23.080 signal processor, with an FPGA, a field programmable gate array 346 00:16:23.240 --> 00:16:25.799 in a master slave configuration. 347 00:16:25.360 --> 00:16:26.840 A brain and a bronze setup. 348 00:16:26.919 --> 00:16:31.080 Precisely, the heavy lifter chip handles the raw, brutal work 349 00:16:31.120 --> 00:16:34.600 of acquiring billions of pixels of data in real time. 350 00:16:34.519 --> 00:16:37.559 While the smart processor runs complex algorithms like the retinext 351 00:16:37.559 --> 00:16:39.679 algorithm to instantly fix the lighting. 352 00:16:39.679 --> 00:16:42.279 In contrast, it is a perfect division of labor that 353 00:16:42.399 --> 00:16:46.279 ensures high definition video processes without a millisecond of lag. 354 00:16:46.519 --> 00:16:48.559 So if we step back and look at the journey 355 00:16:48.600 --> 00:16:52.120 so far, we have mapped the intricate three d landscape 356 00:16:52.120 --> 00:16:52.960 of the human body. 357 00:16:53.080 --> 00:16:55.919 We've used fuzzy math to plan our massive cities. 358 00:16:55.639 --> 00:16:58.720 And we understand the invisible plumbing the equalizers turning down 359 00:16:58.759 --> 00:17:01.480 the treble on our data so it can scream effortlessly around. 360 00:17:01.240 --> 00:17:03.320 The glow, which brings us to the ultimate application. 361 00:17:03.840 --> 00:17:08.720 Right, how are these vast intelligent networks fundamentally rewiring human 362 00:17:08.799 --> 00:17:10.599 learning and daily interaction? 363 00:17:10.960 --> 00:17:13.400 This is where the data leaves the background and steps 364 00:17:13.440 --> 00:17:17.519 to the forefront of human experience. Education is undergoing a 365 00:17:17.640 --> 00:17:21.839 profound structural transformation thanks to these systems. 366 00:17:22.079 --> 00:17:25.200 The research looks closely at teaching English via streaming. 367 00:17:24.839 --> 00:17:29.720 Media, Yes, highlighting a critical mechanical distinction between old digital 368 00:17:29.720 --> 00:17:32.359 models and what intelligence streaming allows today. 369 00:17:32.519 --> 00:17:35.599 Yeah. They map out the shift from the traditional model, 370 00:17:35.640 --> 00:17:38.559 which was essentially just downloading a video file like a 371 00:17:38.640 --> 00:17:43.519 multimedia plus network setup, to true interactive streaming. 372 00:17:43.240 --> 00:17:46.119 And the key mechanism here is the difference between progressive 373 00:17:46.160 --> 00:17:48.480 streaming and real time streaming. 374 00:17:48.160 --> 00:17:51.200 A vital distinction. Progressive streaming is like the early days 375 00:17:51.240 --> 00:17:52.200 of video on demand. 376 00:17:52.279 --> 00:17:56.240 Right, bod, you are downloading a file sequentially, block by block. 377 00:17:56.599 --> 00:17:59.720 If the video hasn't physically downloaded past minute five, you 378 00:17:59.759 --> 00:18:02.359 see cannot click ahead to minute ten, You just have 379 00:18:02.440 --> 00:18:03.759 to sit and wait for it to buffer. 380 00:18:04.000 --> 00:18:07.880 But real time streaming dynamically adapts to your exact Internet 381 00:18:07.920 --> 00:18:09.319 bandwidth in that split second. 382 00:18:09.559 --> 00:18:12.799 It requires specialized protocols, but the result is that you 383 00:18:12.799 --> 00:18:15.960 can instantly jump to any timestamp in the stream. 384 00:18:15.759 --> 00:18:18.599 And the server will instantly deliver the correct data packet. 385 00:18:18.720 --> 00:18:22.799 And this seemingly tiny technical difference completely changes the dynamic 386 00:18:22.839 --> 00:18:25.680 of a digital classroom. How So, because the media is 387 00:18:25.680 --> 00:18:29.880 no longer static and sequential, the teacher's role completely shifts 388 00:18:30.119 --> 00:18:33.599 from being the strict leader of a lecture to being 389 00:18:33.680 --> 00:18:35.759 a guide through an interactive. 390 00:18:35.359 --> 00:18:38.200 Landscape that makes total sense. And the architecture of how 391 00:18:38.279 --> 00:18:41.160 schools managed data is evolving just as fast. 392 00:18:41.279 --> 00:18:43.359 Oh with the school sports manument systems. 393 00:18:43.480 --> 00:18:48.519 Yes, the researchers describe a total departure from heavy localized 394 00:18:48.559 --> 00:18:51.240 software that you had to manually install on a single 395 00:18:51.319 --> 00:18:55.119 school computer. You know, the old CS or client server models, Right. 396 00:18:55.599 --> 00:18:59.240 They have moved to dynamic web based platforms built on 397 00:18:59.279 --> 00:19:01.119 a BS or browser server. 398 00:19:01.000 --> 00:19:04.839 Architecture utilizing MVC model view controller design. 399 00:19:04.599 --> 00:19:07.640 Which means the entire system is accessible from anywhere with 400 00:19:07.720 --> 00:19:11.720 specific access permissions for administrators, teachers, and students. But I 401 00:19:11.759 --> 00:19:14.839