1 00:00:00,000 --> 00:00:02,360 All right, welcome to another deep dive. 2 00:00:02,360 --> 00:00:05,160 Today, we're looking at something pretty fascinating. 3 00:00:05,160 --> 00:00:08,960 The energy demands of this whole digital world, 4 00:00:08,960 --> 00:00:12,440 especially with AI just exploding everywhere. 5 00:00:12,440 --> 00:00:14,920 We've got some great insights actually 6 00:00:14,920 --> 00:00:18,000 from energy expert Mark P. Mills. 7 00:00:18,000 --> 00:00:20,040 He was on the decouple podcast. 8 00:00:20,040 --> 00:00:20,880 And- 9 00:00:20,880 --> 00:00:22,240 Yeah, really interesting conversation. 10 00:00:22,240 --> 00:00:23,280 Oh, super interesting. 11 00:00:23,280 --> 00:00:25,600 So get this, some of these data centers 12 00:00:25,600 --> 00:00:29,280 are using as much power as a whole nuclear plant. 13 00:00:29,280 --> 00:00:31,640 I know it's mind-boggling when you think about it. 14 00:00:31,640 --> 00:00:32,480 It really is. 15 00:00:32,480 --> 00:00:33,640 And we're gonna dig into that. 16 00:00:33,640 --> 00:00:34,480 I mean, what's going on? 17 00:00:34,480 --> 00:00:35,400 Why is that? 18 00:00:35,400 --> 00:00:38,080 So, I mean, I think, you know, 19 00:00:38,080 --> 00:00:39,840 when we think about the cloud, 20 00:00:39,840 --> 00:00:41,400 we think of it as this kind of, 21 00:00:41,400 --> 00:00:43,000 I don't know, this intangible thing. 22 00:00:43,000 --> 00:00:45,720 Right, it's this sort of nebulous thing up there. 23 00:00:45,720 --> 00:00:46,560 Exactly. 24 00:00:46,560 --> 00:00:48,400 But you were talking about how it's actually, 25 00:00:48,400 --> 00:00:51,000 it's a very real, very physical network. 26 00:00:51,000 --> 00:00:51,840 Oh, absolutely. 27 00:00:51,840 --> 00:00:54,440 I mean, we're talking massive data centers, 28 00:00:55,640 --> 00:00:57,440 warehouse size buildings, 29 00:00:57,440 --> 00:00:59,520 millions of processing units, 30 00:00:59,520 --> 00:01:02,600 billions of miles of cables connecting everything. 31 00:01:02,600 --> 00:01:04,760 It's truly staggering. 32 00:01:04,760 --> 00:01:07,000 In fact, if you combined all that space, 33 00:01:07,000 --> 00:01:08,960 it would be larger than all the skyscrapers 34 00:01:08,960 --> 00:01:09,920 in the world combined. 35 00:01:09,920 --> 00:01:10,840 Oh, wow. 36 00:01:10,840 --> 00:01:14,080 Okay, so yeah, so it's not just all these ones and zeros 37 00:01:14,080 --> 00:01:15,360 floating around out there. 38 00:01:15,360 --> 00:01:18,600 There's this, I mean, this really power hungry 39 00:01:18,600 --> 00:01:20,120 infrastructure behind it all. 40 00:01:20,120 --> 00:01:21,160 Yeah, yeah. 41 00:01:21,160 --> 00:01:24,240 So why is AI so power hungry? 42 00:01:24,240 --> 00:01:26,200 Like, why does it use so much more energy 43 00:01:26,200 --> 00:01:28,440 than just regular computing? 44 00:01:28,440 --> 00:01:29,640 Well, the thing with AI, 45 00:01:29,640 --> 00:01:31,280 especially with machine learning, 46 00:01:31,280 --> 00:01:33,640 is that it needs these massive data sets 47 00:01:33,640 --> 00:01:35,040 to train the algorithms. 48 00:01:35,040 --> 00:01:35,880 Okay, so it's kind of like, 49 00:01:35,880 --> 00:01:38,400 it's like teaching a kid to recognize a cat, right? 50 00:01:38,400 --> 00:01:40,600 You're showing them thousands and thousands of pictures. 51 00:01:40,600 --> 00:01:42,000 Exactly, it's the same concept, 52 00:01:42,000 --> 00:01:45,160 but just on a ridiculously larger scale, 53 00:01:45,160 --> 00:01:46,400 and all that processing power. 54 00:01:46,400 --> 00:01:49,080 Well, it takes energy, lots of energy. 55 00:01:49,080 --> 00:01:51,400 And Mills was saying that AI uses something 56 00:01:51,400 --> 00:01:54,560 like 10 times the energy of traditional computing. 57 00:01:54,560 --> 00:01:55,400 Yeah, yeah. 58 00:01:55,400 --> 00:01:57,920 At least, and that's only gonna increase 59 00:01:57,920 --> 00:02:00,800 as AI gets more sophisticated. 60 00:02:00,800 --> 00:02:03,960 So, I mean, what are the implications of that then? 61 00:02:03,960 --> 00:02:07,240 Especially as AI becomes more and more integrated 62 00:02:07,240 --> 00:02:09,800 into our lives, I mean, we're already seeing it everywhere. 63 00:02:09,800 --> 00:02:12,560 Right, I mean, we're looking at a pretty significant 64 00:02:12,560 --> 00:02:14,400 increase in energy demand. 65 00:02:14,400 --> 00:02:17,600 And this is on top of the growth we're already seeing, 66 00:02:17,600 --> 00:02:19,320 from all the other digital services we use, 67 00:02:19,320 --> 00:02:21,240 streaming, online gaming, all of that. 68 00:02:21,240 --> 00:02:24,200 Right, right, so AI is almost like throwing 69 00:02:24,200 --> 00:02:25,880 rocket fuel on that fire. 70 00:02:25,880 --> 00:02:26,720 Pretty much, yeah. 71 00:02:26,720 --> 00:02:28,840 And it's not just a tech issue, right? 72 00:02:28,840 --> 00:02:31,600 This has geopolitical implications too. 73 00:02:31,600 --> 00:02:36,600 Oh, absolutely, I mean, access to cheap and reliable energy, 74 00:02:37,080 --> 00:02:40,520 that could become a major factor 75 00:02:40,520 --> 00:02:42,560 in which countries kind of end up leading the way 76 00:02:42,560 --> 00:02:43,680 in AI development. 77 00:02:43,680 --> 00:02:46,040 And Mills specifically points to China. 78 00:02:46,040 --> 00:02:46,880 Yeah. 79 00:02:46,880 --> 00:02:48,720 Because they're investing really heavily 80 00:02:48,720 --> 00:02:52,560 in both data centers and kind of surprisingly coal power. 81 00:02:52,560 --> 00:02:54,080 Right, it's interesting. 82 00:02:54,080 --> 00:02:57,600 Yeah, you think with everyone pushing for renewables, 83 00:02:57,600 --> 00:02:59,000 it's a little counter-occuative. 84 00:02:59,000 --> 00:03:01,840 It is, but it shows you the kind of tension 85 00:03:01,840 --> 00:03:04,720 that exists between immediate energy needs 86 00:03:04,720 --> 00:03:07,560 and these long-term sustainability goals. 87 00:03:07,560 --> 00:03:08,400 Right, right. 88 00:03:08,400 --> 00:03:12,120 China's prioritizing their position in the AI race, 89 00:03:12,120 --> 00:03:15,040 even if it means maybe relying on energy sources 90 00:03:15,040 --> 00:03:18,480 that aren't so environmentally friendly, at least for now. 91 00:03:18,480 --> 00:03:21,040 So, I mean, how does the US kind of compare 92 00:03:21,040 --> 00:03:24,040 in this whole energy AI landscape? 93 00:03:24,040 --> 00:03:25,160 Well, the US and Canada, 94 00:03:25,160 --> 00:03:29,600 they're still major hubs for AI innovation, 95 00:03:29,600 --> 00:03:34,080 but they rely more on natural gas and renewables. 96 00:03:34,080 --> 00:03:36,240 And those can be more expensive 97 00:03:36,240 --> 00:03:39,640 and less consistently available than coal. 98 00:03:39,640 --> 00:03:42,440 So, it raises some questions about whether, 99 00:03:42,440 --> 00:03:44,160 maybe over-regulation in the West 100 00:03:44,160 --> 00:03:47,520 could actually hinder its ability to compete in the long run. 101 00:03:47,520 --> 00:03:49,960 Interesting, now Mills also talks about this concept 102 00:03:49,960 --> 00:03:51,600 of in-silico research. 103 00:03:51,600 --> 00:03:53,560 Can you maybe explain what that is 104 00:03:53,560 --> 00:03:56,480 and why it's relevant to this whole discussion? 105 00:03:56,480 --> 00:03:58,960 Sure, so in-silico, it basically means, you know, 106 00:03:58,960 --> 00:04:02,760 performed on a computer or through computer simulation. 107 00:04:02,760 --> 00:04:04,400 So, in the context of research, 108 00:04:04,400 --> 00:04:07,480 it's about using computer models and simulations 109 00:04:07,480 --> 00:04:09,880 to study biological processes. 110 00:04:09,880 --> 00:04:10,720 Okay. 111 00:04:10,720 --> 00:04:14,800 So, imagine creating a digital twin of your body 112 00:04:14,800 --> 00:04:17,000 containing all your health data. 113 00:04:17,000 --> 00:04:17,840 Wow. 114 00:04:17,840 --> 00:04:21,440 And then scientists could use AI to run simulations 115 00:04:21,440 --> 00:04:23,920 and predict how different drugs or treatments 116 00:04:23,920 --> 00:04:25,080 might affect you. 117 00:04:25,080 --> 00:04:26,480 So, that sounds like, I mean, 118 00:04:26,480 --> 00:04:29,600 it sounds like something straight out of a sci-fi movie. 119 00:04:29,600 --> 00:04:30,600 It does a bit, doesn't it? 120 00:04:30,600 --> 00:04:33,520 It does, but what kind of energy demands 121 00:04:33,520 --> 00:04:36,080 would this in-silico research have? 122 00:04:36,080 --> 00:04:37,560 I mean, it sounds pretty intense. 123 00:04:37,560 --> 00:04:39,320 Oh, it would be very significant. 124 00:04:39,320 --> 00:04:40,760 You know, building and maintaining 125 00:04:40,760 --> 00:04:42,680 those detailed digital models, 126 00:04:42,680 --> 00:04:44,640 storing vast amounts of data, 127 00:04:44,640 --> 00:04:47,040 running these complex AI simulations, 128 00:04:47,040 --> 00:04:48,960 it all takes a lot of processing power. 129 00:04:48,960 --> 00:04:50,640 And like we've been talking about, you know, 130 00:04:50,640 --> 00:04:52,360 with any data heavy process, 131 00:04:52,360 --> 00:04:54,680 energy consumption is a key factor. 132 00:04:54,680 --> 00:04:56,720 So, we're not just talking about powering the cloud, 133 00:04:56,720 --> 00:04:57,560 as we know it. 134 00:04:57,560 --> 00:05:00,760 We're talking about a whole new level of energy demand 135 00:05:00,760 --> 00:05:04,720 as these in-silico applications become more widespread. 136 00:05:04,720 --> 00:05:05,560 Exactly. 137 00:05:05,560 --> 00:05:08,400 And that brings up some even more, you know, 138 00:05:08,400 --> 00:05:11,480 urgent questions about the future of energy production. 139 00:05:11,480 --> 00:05:13,960 You know, where will all this power come from? 140 00:05:13,960 --> 00:05:16,160 How can we, you know, ensure it's sustainable? 141 00:05:16,160 --> 00:05:17,000 Yeah. 142 00:05:17,000 --> 00:05:18,760 And those are the challenges that we're gonna be looking at 143 00:05:18,760 --> 00:05:21,000 more in-depth as our deep dive continues. 144 00:05:21,000 --> 00:05:23,160 All right, so we're back for part two 145 00:05:23,160 --> 00:05:27,720 of our deep dive into AI's energy appetite. 146 00:05:27,720 --> 00:05:29,720 And, you know, we've been talking about 147 00:05:29,720 --> 00:05:34,720 the energy-intensive nature of AI in-silico research. 148 00:05:34,880 --> 00:05:36,040 But what about solutions? 149 00:05:36,040 --> 00:05:37,800 I mean, where are we gonna get all this power? 150 00:05:37,800 --> 00:05:39,720 Well, that's the big question, isn't it? 151 00:05:39,720 --> 00:05:44,720 And Mills, he argues that we can't just rely on renewables, 152 00:05:44,840 --> 00:05:46,760 you know, for this base load power 153 00:05:46,760 --> 00:05:48,800 that these huge data centers need. 154 00:05:48,800 --> 00:05:50,400 Base load power, what's that? 155 00:05:50,400 --> 00:05:53,440 Right, so base load power, it's the minimum amount 156 00:05:53,440 --> 00:05:57,160 of electricity that a grid has to supply at all times, 157 00:05:57,160 --> 00:05:58,280 you know, to keep things running. 158 00:05:58,280 --> 00:06:01,640 And renewables like solar and wind, they're great, 159 00:06:01,640 --> 00:06:03,240 but they're intermittent, right? 160 00:06:03,240 --> 00:06:04,520 They're not always available. 161 00:06:04,520 --> 00:06:05,520 You can't always count on them. 162 00:06:05,520 --> 00:06:06,360 Okay. 163 00:06:06,360 --> 00:06:09,200 So data centers, they need this constant, 164 00:06:09,200 --> 00:06:11,360 this reliable energy source. 165 00:06:11,360 --> 00:06:13,480 And that's where, you know, all this talk 166 00:06:13,480 --> 00:06:15,040 about nuclear power comes in. 167 00:06:15,040 --> 00:06:16,040 Yeah. 168 00:06:16,040 --> 00:06:18,360 And it's interesting because Mills, 169 00:06:18,360 --> 00:06:20,960 he suggests that nuclear power, 170 00:06:20,960 --> 00:06:24,440 which has, you know, obviously been somewhat controversial. 171 00:06:24,440 --> 00:06:25,280 Just say the least. 172 00:06:25,280 --> 00:06:28,080 Yeah, he's saying that it might be making a comeback, 173 00:06:28,080 --> 00:06:30,320 not just for environmental reasons, 174 00:06:30,320 --> 00:06:32,720 but just out of sheer necessity. 175 00:06:32,720 --> 00:06:35,040 Yeah, I mean, he points out that some 176 00:06:35,040 --> 00:06:37,800 of these data centers are using as much electricity 177 00:06:37,800 --> 00:06:39,920 as a small nuclear power plant. 178 00:06:39,920 --> 00:06:40,760 It's great. 179 00:06:40,760 --> 00:06:41,600 It is. 180 00:06:41,600 --> 00:06:45,640 And as the demand for AI and computing power keeps growing, 181 00:06:45,640 --> 00:06:48,040 we have to consider all our options. 182 00:06:48,040 --> 00:06:50,680 And nuclear, it does provide that steady, 183 00:06:50,680 --> 00:06:53,840 that high output power that renewables, 184 00:06:53,840 --> 00:06:56,280 you know, at least for now, they just can't match. 185 00:06:56,280 --> 00:06:57,120 Yeah. 186 00:06:57,120 --> 00:07:00,080 I mean, but isn't building, you know, nuclear plants, 187 00:07:00,080 --> 00:07:03,360 isn't that a really lengthy and complex process? 188 00:07:03,360 --> 00:07:04,200 Oh, absolutely. 189 00:07:04,200 --> 00:07:05,040 Yeah. 190 00:07:05,040 --> 00:07:05,880 I mean, it takes years and years. 191 00:07:05,880 --> 00:07:07,720 And Mills, he acknowledges that. 192 00:07:07,720 --> 00:07:10,400 He says, even if we started today, 193 00:07:10,400 --> 00:07:14,600 it would be years before those plants came online. 194 00:07:14,600 --> 00:07:17,000 So he suggests that in the meantime, 195 00:07:17,000 --> 00:07:18,280 you know, we might have to think about 196 00:07:18,280 --> 00:07:21,320 some less ideal solutions, like keeping existing 197 00:07:21,320 --> 00:07:22,800 coal and gas plants running, 198 00:07:22,800 --> 00:07:25,880 maybe even bringing back some that were recently retired. 199 00:07:25,880 --> 00:07:29,160 I mean, that's bound to be, you know, unpopular, 200 00:07:29,160 --> 00:07:32,120 especially with everyone pushing for cleaner energy. 201 00:07:32,120 --> 00:07:34,720 Oh yeah, it's definitely gonna be controversial. 202 00:07:34,720 --> 00:07:38,120 But his argument is that it's a pragmatic approach, 203 00:07:38,120 --> 00:07:39,760 at least for the short term. 204 00:07:39,760 --> 00:07:42,800 You know, we can't just wish away the energy needs 205 00:07:42,800 --> 00:07:44,800 of the digital world. 206 00:07:44,800 --> 00:07:47,360 And while the long-term goal is to shift 207 00:07:47,360 --> 00:07:49,840 to more sustainable sources, 208 00:07:49,840 --> 00:07:53,280 we need practical solutions right now 209 00:07:53,280 --> 00:07:55,320 to kind of bridge that gap. 210 00:07:55,320 --> 00:07:58,120 Okay, so where does that leave us 211 00:07:58,120 --> 00:08:01,000 in terms of the long-term vision? 212 00:08:01,000 --> 00:08:04,360 Well, Mills, he believes that ultimately, 213 00:08:04,360 --> 00:08:07,840 it's gonna be a combination of nuclear power, renewables, 214 00:08:07,840 --> 00:08:10,360 and advancements in energy efficiency. 215 00:08:10,360 --> 00:08:13,800 That's what's gonna meet the demands of this digital future. 216 00:08:13,800 --> 00:08:17,280 And he's especially enthusiastic about the potential 217 00:08:17,280 --> 00:08:20,640 of small modular reactors, these SMRs. 218 00:08:20,640 --> 00:08:21,840 Oh, right. 219 00:08:21,840 --> 00:08:23,040 Which are, you know, they're smaller, 220 00:08:23,040 --> 00:08:24,400 they're more flexible, you know, 221 00:08:24,400 --> 00:08:25,720 they can be deployed more quickly 222 00:08:25,720 --> 00:08:28,000 and with less upfront investment. 223 00:08:28,000 --> 00:08:31,200 Okay, that sounds promising, but you know, 224 00:08:31,200 --> 00:08:33,760 this whole energy issue, it goes way beyond 225 00:08:33,760 --> 00:08:34,920 just data centers, right? 226 00:08:34,920 --> 00:08:37,520 I mean, think about, you know, electric vehicles, 227 00:08:37,520 --> 00:08:41,160 smart cities, all this tech that's changing our world. 228 00:08:41,160 --> 00:08:44,760 Oh yeah, it's a systemic challenge, for sure. 229 00:08:44,760 --> 00:08:46,800 And it has, you know, as we've been saying, 230 00:08:46,800 --> 00:08:48,920 these geopolitical implications as well, 231 00:08:48,920 --> 00:08:52,800 Mills points out that countries that have access to cheap 232 00:08:52,800 --> 00:08:55,280 and abundant energy, they could have a real advantage 233 00:08:55,280 --> 00:08:57,440 in this whole AI race. 234 00:08:57,440 --> 00:09:00,520 And we were talking before about China's, you know, 235 00:09:00,520 --> 00:09:03,040 investment in coal power. 236 00:09:03,040 --> 00:09:05,920 Are there other countries that are kind of positioned 237 00:09:05,920 --> 00:09:10,440 to benefit from this energy AI connection? 238 00:09:10,440 --> 00:09:11,280 Yeah, definitely. 239 00:09:11,280 --> 00:09:12,760 I mean, countries that have, you know, 240 00:09:12,760 --> 00:09:15,800 significant hydropower resources, like Canada 241 00:09:15,800 --> 00:09:18,120 or some South American nations, 242 00:09:18,120 --> 00:09:20,200 they could be in a good spot. 243 00:09:20,200 --> 00:09:23,200 The key is having a, you know, stable 244 00:09:23,200 --> 00:09:26,560 and affordable energy source to power all this, 245 00:09:26,560 --> 00:09:28,400 this massive computing infrastructure 246 00:09:28,400 --> 00:09:30,920 that's needed for AI development. 247 00:09:30,920 --> 00:09:33,000 And this is where, you know, the US, 248 00:09:33,000 --> 00:09:36,080 which has more of a diverse energy mix, 249 00:09:36,080 --> 00:09:38,080 this is where it might face a challenge 250 00:09:38,080 --> 00:09:41,680 because, you know, regulations and infrastructure limitations, 251 00:09:41,680 --> 00:09:43,720 they could slow things down, 252 00:09:43,720 --> 00:09:46,600 especially if energy costs start getting, you know, really high. 253 00:09:46,600 --> 00:09:48,760 So it's not just about, you know, 254 00:09:48,760 --> 00:09:52,080 having the smartest people or the most advanced algorithms, 255 00:09:52,080 --> 00:09:54,480 it's about having the energy resources 256 00:09:54,480 --> 00:09:56,640 to actually fuel those innovations. 257 00:09:56,640 --> 00:09:57,480 Yeah, exactly. 258 00:09:57,480 --> 00:10:00,000 And it really highlights the need for strategic thinking 259 00:10:00,000 --> 00:10:03,400 and long-term planning when it comes to energy policy. 260 00:10:03,400 --> 00:10:07,360 Yeah, one of the things that really struck me 261 00:10:07,360 --> 00:10:11,720 listening to the podcast was Mills' point about how 262 00:10:11,720 --> 00:10:15,520 our demand for information is fundamentally different 263 00:10:15,520 --> 00:10:17,760 from our demand for other resources. 264 00:10:17,760 --> 00:10:20,640 You know, we have a limit to how much food 265 00:10:20,640 --> 00:10:23,560 or material goods we can consume, 266 00:10:23,560 --> 00:10:28,160 but our appetite for information, it just seems limitless. 267 00:10:28,160 --> 00:10:29,880 Yeah, it's an interesting thought, isn't it? 268 00:10:29,880 --> 00:10:32,960 He uses the evolution of the words we use 269 00:10:32,960 --> 00:10:35,720 for large numbers as an example. 270 00:10:35,720 --> 00:10:39,240 You know, we've gone from kilobytes to megabytes 271 00:10:39,240 --> 00:10:42,200 to gigabytes, and now we're talking about 272 00:10:42,200 --> 00:10:43,400 zettabytes and yottabytes. 273 00:10:43,400 --> 00:10:45,960 I mean, these are numbers that are so large 274 00:10:45,960 --> 00:10:48,520 that it's hard to even grasp them. 275 00:10:48,520 --> 00:10:52,080 Yeah, he had that amazing analogy trying to visualize 276 00:10:52,080 --> 00:10:55,640 like a zettabyte of data as a stack of dollar bills 277 00:10:55,640 --> 00:10:58,000 reaching from the earth to the sun and back 278 00:10:58,000 --> 00:10:59,880 like 700,000 times. 279 00:10:59,880 --> 00:11:02,080 Yeah, which is just, I mean, it's mind-blowing. 280 00:11:02,080 --> 00:11:05,160 But it really underscores the sheer scale 281 00:11:05,160 --> 00:11:08,240 of this digital universe that we're creating. 282 00:11:08,240 --> 00:11:11,160 And as we generate more and more data, 283 00:11:11,160 --> 00:11:14,200 the energy needed to store it and process it, 284 00:11:14,200 --> 00:11:17,080 it's just gonna keep growing exponentially. 285 00:11:17,080 --> 00:11:19,280 Yeah, it sounds like we're really on the verge 286 00:11:19,280 --> 00:11:20,920 of a computing revolution, 287 00:11:20,920 --> 00:11:22,520 unlike anything we've ever seen, 288 00:11:22,520 --> 00:11:25,600 but one that comes with this massive, 289 00:11:25,600 --> 00:11:27,200 massive energy price tag. 290 00:11:27,200 --> 00:11:29,120 Absolutely, and that brings us back 291 00:11:29,120 --> 00:11:31,360 to the core question here. 292 00:11:31,360 --> 00:11:35,000 How do we balance this insatiable demand 293 00:11:35,000 --> 00:11:37,120 for information, for computing power, 294 00:11:37,120 --> 00:11:40,840 with the need for sustainable energy solutions? 295 00:11:40,840 --> 00:11:43,680 It's a tough question, with no easy answers, 296 00:11:43,680 --> 00:11:46,760 but maybe we can at least start by acknowledging 297 00:11:46,760 --> 00:11:50,720 the connection between energy technology and our future. 298 00:11:50,720 --> 00:11:51,800 Yeah, I agree. 299 00:11:51,800 --> 00:11:54,280 And that awareness needs to be reflected 300 00:11:54,280 --> 00:11:56,600 in our policies, our investments, 301 00:11:56,600 --> 00:11:57,920 our individual choices. 302 00:11:57,920 --> 00:12:01,960 We all have a role to play in shaping a future 303 00:12:01,960 --> 00:12:06,000 where AI and this digital world can thrive 304 00:12:06,000 --> 00:12:07,600 without damaging the planet. 305 00:12:07,600 --> 00:12:09,560 Welcome back to the final part now 306 00:12:09,560 --> 00:12:12,960 of our deep dive into AI's energy appetite. 307 00:12:12,960 --> 00:12:15,640 It's been quite a journey. 308 00:12:15,640 --> 00:12:16,920 It has, lots to think about. 309 00:12:16,920 --> 00:12:20,120 We've talked about the energy demands of AI, 310 00:12:20,120 --> 00:12:24,200 the potential of this in-silico research, 311 00:12:24,200 --> 00:12:25,560 the role of nuclear power. 312 00:12:25,560 --> 00:12:27,160 But before we wrap things up, 313 00:12:27,160 --> 00:12:29,720 there were a couple more key takeaways 314 00:12:29,720 --> 00:12:32,800 that I thought were worth digging into a bit. 315 00:12:32,800 --> 00:12:33,640 Yeah, I agree. 316 00:12:33,640 --> 00:12:35,600 We've definitely covered a lot of ground. 317 00:12:35,600 --> 00:12:37,400 But I think it is important to kind of step back 318 00:12:37,400 --> 00:12:40,040 and consider the bigger picture here. 319 00:12:40,040 --> 00:12:43,440 We talk about AI and the digital world 320 00:12:43,440 --> 00:12:45,760 as if it's its own separate thing. 321 00:12:45,760 --> 00:12:50,320 But the reality is it's very much tied to the physical world, 322 00:12:50,320 --> 00:12:52,240 especially when it comes to energy. 323 00:12:52,240 --> 00:12:55,000 Yeah, and Mills makes this really interesting point 324 00:12:55,000 --> 00:12:58,640 about how the infrastructure of the cloud, 325 00:12:58,640 --> 00:13:00,880 the data centers, the networks, the servers, 326 00:13:00,880 --> 00:13:02,960 it's almost like this hidden world 327 00:13:02,960 --> 00:13:04,800 that most people don't even think about. 328 00:13:04,800 --> 00:13:06,360 Right, out of sight, out of mind. 329 00:13:06,360 --> 00:13:07,280 Yeah, exactly. 330 00:13:07,280 --> 00:13:12,040 I mean, we interact with the digital world every day, 331 00:13:12,040 --> 00:13:13,120 constantly. 332 00:13:13,120 --> 00:13:17,160 But we rarely stop and think about this massive, 333 00:13:17,160 --> 00:13:20,200 this physical infrastructure that's making it all possible. 334 00:13:20,200 --> 00:13:22,040 Yeah, it's kind of like an iceberg, right? 335 00:13:22,040 --> 00:13:24,960 We see the tip, our smartphones, our laptops, 336 00:13:24,960 --> 00:13:26,000 the apps we use. 337 00:13:26,000 --> 00:13:28,520 But underneath, underneath the surface 338 00:13:28,520 --> 00:13:32,320 is this huge network of energy-intensive machinery 339 00:13:32,320 --> 00:13:33,720 that's keeping it all running. 340 00:13:33,720 --> 00:13:34,720 Right, right. 341 00:13:34,720 --> 00:13:38,080 And as we move towards even more complex 342 00:13:38,080 --> 00:13:40,360 and sophisticated AI applications, 343 00:13:40,360 --> 00:13:42,840 like the in-silico research we talked about, 344 00:13:42,840 --> 00:13:45,760 the size of that iceberg, it's only gonna grow. 345 00:13:45,760 --> 00:13:48,280 And that raises some really big questions 346 00:13:48,280 --> 00:13:49,200 about sustainability. 347 00:13:49,200 --> 00:13:52,680 I mean, can we keep expanding the digital world 348 00:13:52,680 --> 00:13:56,280 at this pace without putting an unsustainable strain 349 00:13:56,280 --> 00:13:57,600 on our energy resources? 350 00:13:57,600 --> 00:13:59,840 That's the, I mean, the million dollar question, right? 351 00:13:59,840 --> 00:14:00,680 Yeah. 352 00:14:00,680 --> 00:14:02,880 And there's no easy answer. 353 00:14:03,840 --> 00:14:07,960 Mills, he argues that we need this kind of, 354 00:14:07,960 --> 00:14:09,840 this multi-pronged approach. 355 00:14:09,840 --> 00:14:12,640 First, we gotta keep investing in renewables, 356 00:14:12,640 --> 00:14:14,520 solar, wind power. 357 00:14:14,520 --> 00:14:17,240 But as we've discussed, renewables alone, 358 00:14:17,240 --> 00:14:20,640 they can't meet the baseload power demands 359 00:14:20,640 --> 00:14:22,200 of these data centers. 360 00:14:22,200 --> 00:14:24,000 Right, because they need that constant, 361 00:14:24,000 --> 00:14:25,320 reliable source of energy. 362 00:14:25,320 --> 00:14:26,720 Right, exactly. 363 00:14:26,720 --> 00:14:30,280 Which is why Mills really advocates for 364 00:14:30,280 --> 00:14:33,680 a serious look at nuclear power. 365 00:14:33,680 --> 00:14:37,000 He sees it as a crucial part of the solution, 366 00:14:37,000 --> 00:14:40,880 especially as we try to move away from fossil fuels. 367 00:14:40,880 --> 00:14:42,880 He points out that nuclear plants, 368 00:14:42,880 --> 00:14:45,440 they have a very small carbon footprint 369 00:14:45,440 --> 00:14:47,640 and they can provide that steady, 370 00:14:47,640 --> 00:14:50,520 that high output power that data centers really need. 371 00:14:50,520 --> 00:14:53,920 I mean, but nuclear power, it comes with its own set 372 00:14:53,920 --> 00:14:55,280 of challenges. 373 00:14:55,280 --> 00:14:56,120 Right, yeah. 374 00:14:56,120 --> 00:14:57,160 I mean, the risk of accidents, 375 00:14:57,160 --> 00:14:59,440 the issue of nuclear waste. 376 00:14:59,440 --> 00:15:03,080 Absolutely, and those are serious concerns. 377 00:15:03,080 --> 00:15:05,560 And Mills, he doesn't shy away from that. 378 00:15:05,560 --> 00:15:06,400 He acknowledges that. 379 00:15:06,400 --> 00:15:10,480 But he argues that with advances in technology, 380 00:15:10,480 --> 00:15:13,360 in safety protocols, the risks associated 381 00:15:13,360 --> 00:15:16,560 with nuclear power, they're manageable. 382 00:15:16,560 --> 00:15:19,480 And he also believes that we need to invest more 383 00:15:19,480 --> 00:15:23,040 in research and development to find safe, 384 00:15:23,040 --> 00:15:27,200 effective solutions for dealing with nuclear waste. 385 00:15:27,200 --> 00:15:31,440 So it sounds like the future of AI and the digital world, 386 00:15:31,440 --> 00:15:33,680 it's gonna require this really kind of complex 387 00:15:33,680 --> 00:15:36,600 and nuanced approach to energy production. 388 00:15:36,600 --> 00:15:39,840 One that balances the need for clean energy 389 00:15:39,840 --> 00:15:43,440 with the need for this reliable, this base load power. 390 00:15:43,440 --> 00:15:44,960 It's a balancing act for sure. 391 00:15:44,960 --> 00:15:47,200 And it's not just a challenge for governments 392 00:15:47,200 --> 00:15:49,400 and corporations to figure out. 393 00:15:49,400 --> 00:15:51,440 It's something that's gonna require 394 00:15:51,440 --> 00:15:54,160 a collective effort from all of us. 395 00:15:54,160 --> 00:15:57,640 So what can individuals do? 396 00:15:57,640 --> 00:16:01,600 What can people like me, our listeners do 397 00:16:01,600 --> 00:16:02,920 to make a difference? 398 00:16:02,920 --> 00:16:05,880 I mean, it can feel kind of overwhelming. 399 00:16:05,880 --> 00:16:06,880 It can, yeah. 400 00:16:06,880 --> 00:16:07,920 It's a big issue. 401 00:16:07,920 --> 00:16:12,920 But even small changes in our own behavior can add up. 402 00:16:13,560 --> 00:16:16,240 We can be more mindful about our own energy use. 403 00:16:16,240 --> 00:16:20,440 We can support policies that encourage renewable energy. 404 00:16:20,440 --> 00:16:23,600 We can choose products and services from companies 405 00:16:23,600 --> 00:16:26,320 that are committed to sustainability. 406 00:16:26,320 --> 00:16:30,080 So it's not just about the technology itself. 407 00:16:30,080 --> 00:16:32,680 It's about the choices we make, 408 00:16:32,680 --> 00:16:35,120 both individually and as a society 409 00:16:35,120 --> 00:16:37,840 about how we use and produce energy. 410 00:16:37,840 --> 00:16:38,680 Exactly. 411 00:16:38,680 --> 00:16:41,920 And as our deep dive into AI's energy appetite, 412 00:16:41,920 --> 00:16:44,240 as it wraps up here, I hope it's clear 413 00:16:44,240 --> 00:16:46,680 that this isn't just a technical issue. 414 00:16:46,680 --> 00:16:50,320 It's something that touches on economics, on geopolitics, 415 00:16:50,320 --> 00:16:52,840 even on our own personal values. 416 00:16:52,840 --> 00:16:55,640 Yeah, it's a good reminder that the digital world, 417 00:16:55,640 --> 00:16:58,840 it's not this abstract, this intangible thing. 418 00:16:58,840 --> 00:17:02,720 It's built on a very physical infrastructure 419 00:17:02,720 --> 00:17:06,040 that needs real resources, especially energy. 420 00:17:06,040 --> 00:17:08,520 And as AI continues to evolve 421 00:17:08,520 --> 00:17:10,520 and become even more a part of our lives, 422 00:17:10,520 --> 00:17:13,720 it's really important that we have these conversations, 423 00:17:13,720 --> 00:17:15,360 these thoughtful and informed conversations 424 00:17:15,360 --> 00:17:18,040 about how to power this digital revolution 425 00:17:18,040 --> 00:17:20,160 in a way that's sustainable. 426 00:17:20,160 --> 00:17:21,960 Couldn't have said it better myself. 427 00:17:21,960 --> 00:17:25,920 The future of AI, the future of our planet, it depends on it. 428 00:17:25,920 --> 00:17:28,440 Well, thank you so much for joining us 429 00:17:28,440 --> 00:17:30,640 and for sharing your expertise with us today. 430 00:17:30,640 --> 00:17:33,200 This has been a really insightful discussion. 431 00:17:33,200 --> 00:17:36,120 And to our listeners, thank you for joining us 432 00:17:36,120 --> 00:17:37,200 on this deep dive as well. 433 00:17:37,200 --> 00:17:38,800 We hope you learned something new 434 00:17:38,800 --> 00:17:42,040 and that you'll keep exploring this really fascinating 435 00:17:42,040 --> 00:17:46,200 and ever-changing world of AI and how it's shaping our lives.