1 00:00:00,000 --> 00:00:02,000 Okay, so we got a lot to unpack today. 2 00:00:02,000 --> 00:00:05,600 You've sent over a really interesting mix of articles. 3 00:00:05,600 --> 00:00:09,480 Everything from Amazon's AI strategy to clear reactors. 4 00:00:09,480 --> 00:00:10,720 Yeah, clear reactors. 5 00:00:10,720 --> 00:00:11,840 I mean, it's pretty wild. 6 00:00:11,840 --> 00:00:12,920 Who saw that coming? 7 00:00:12,920 --> 00:00:14,360 It is an unexpected turn. 8 00:00:14,360 --> 00:00:15,560 Yeah. Right. 9 00:00:15,560 --> 00:00:20,480 But I think it highlights how AI is impacting not just the tech world, 10 00:00:20,480 --> 00:00:23,840 but like energy production and, you know, all of these things are. 11 00:00:23,840 --> 00:00:24,760 It's all connected. 12 00:00:24,760 --> 00:00:26,360 Inacted. Totally. 13 00:00:26,360 --> 00:00:29,640 So today, our mission is to kind of untangle these connections 14 00:00:29,640 --> 00:00:31,960 to figure out who the major players are. 15 00:00:31,960 --> 00:00:35,240 And most importantly, separate the AI hype from reality. 16 00:00:35,240 --> 00:00:36,120 Yeah. 17 00:00:36,120 --> 00:00:38,560 And I think a good place to start is with Amazon. 18 00:00:38,560 --> 00:00:39,560 Okay. 19 00:00:39,560 --> 00:00:45,680 One article you shared suggests that AWS, their cloud computing arm, 20 00:00:45,680 --> 00:00:48,400 might have fumbled a bit in the generative AI race. 21 00:00:48,400 --> 00:00:49,880 Yeah, that's right. 22 00:00:49,880 --> 00:00:54,040 While companies like OpenAI were making headlines with chat GPT, AWS 23 00:00:54,040 --> 00:00:55,800 seemed to be a little slow off the mark. 24 00:00:55,800 --> 00:00:57,920 And, you know, this wasn't just about missing a trend. 25 00:00:57,920 --> 00:01:01,120 It sounds like they may have underestimated the demand for these tools. 26 00:01:01,120 --> 00:01:06,640 And then to make matters worse, they lost a key player, Matt Wood, 27 00:01:06,640 --> 00:01:10,120 a top AI executive who recently left AWS. 28 00:01:10,120 --> 00:01:12,200 Okay. So not just missing the boat, 29 00:01:12,200 --> 00:01:13,720 but maybe some internal turmoil too. 30 00:01:13,720 --> 00:01:14,480 Right. 31 00:01:14,480 --> 00:01:15,880 Not a good look for them. 32 00:01:15,880 --> 00:01:18,400 So how is AWS responding to this? 33 00:01:18,400 --> 00:01:20,880 Well, their new CEO, Matt Garmin, 34 00:01:20,880 --> 00:01:23,280 seems to be aware of the need to catch up. 35 00:01:23,280 --> 00:01:23,880 Okay. 36 00:01:23,880 --> 00:01:27,720 They're aggressively acquiring AI startups like Anthropic, 37 00:01:27,720 --> 00:01:31,280 which is known for its work on the safer AI development. 38 00:01:31,280 --> 00:01:35,000 And they've been investing heavily in their own large language models, 39 00:01:35,000 --> 00:01:39,000 even claiming to have a multi-billion dollar generative AI business already. 40 00:01:39,000 --> 00:01:39,680 Wow. 41 00:01:39,680 --> 00:01:43,240 They're also promising cheaper and faster AI tools for developers 42 00:01:43,240 --> 00:01:46,560 trying to position themselves as the more cost effective option. 43 00:01:46,560 --> 00:01:50,720 So it's like they're trying to make up for a lost time with a full on AI blitz. 44 00:01:50,720 --> 00:01:52,880 The big question is, is it going to be enough? 45 00:01:52,880 --> 00:01:55,240 Can they really catch up after such a late start? 46 00:01:55,240 --> 00:01:55,800 Yeah. 47 00:01:55,800 --> 00:01:58,360 Especially when the technology is evolving so rapidly. 48 00:01:58,360 --> 00:02:00,800 It's a classic case of better late than never, 49 00:02:00,800 --> 00:02:05,320 but whether it's enough to reclaim the top spot remains to be seen. 50 00:02:05,320 --> 00:02:09,320 Especially since their solution to powering all this AI 51 00:02:09,320 --> 00:02:13,320 involves a pretty controversial move, nuclear power. 52 00:02:13,320 --> 00:02:15,600 Well, nuclear power, that seems a little extreme, 53 00:02:15,600 --> 00:02:16,960 even for a company like Amazon. 54 00:02:16,960 --> 00:02:18,600 Right. But here's the thing. 55 00:02:18,600 --> 00:02:22,160 AI, particularly the kind of large scale AI 56 00:02:22,160 --> 00:02:24,640 that companies like Amazon and Google are building, 57 00:02:24,640 --> 00:02:26,720 requires a massive amount of energy. 58 00:02:26,720 --> 00:02:27,240 Period. 59 00:02:27,240 --> 00:02:29,560 And they're looking for ways to power these systems 60 00:02:29,560 --> 00:02:32,880 in a way that's both sustainable and reliable. 61 00:02:32,880 --> 00:02:36,720 OK, I get the need for energy, but why nuclear? 62 00:02:36,720 --> 00:02:37,800 I feel like there are other. 63 00:02:37,800 --> 00:02:39,680 Let's say controversial options out there. 64 00:02:39,680 --> 00:02:42,400 Why not just go all in on solar or wind power? 65 00:02:42,400 --> 00:02:43,720 Those are certainly part of the mix, 66 00:02:43,720 --> 00:02:47,480 but the reality is that renewable energy sources like solar and wind 67 00:02:47,480 --> 00:02:48,480 can be intermittent. 68 00:02:48,480 --> 00:02:49,120 Right. 69 00:02:49,120 --> 00:02:51,760 The sun doesn't always shine, the wind doesn't always blow. 70 00:02:51,760 --> 00:02:54,960 And for something as data intensive as AI, 71 00:02:54,960 --> 00:02:58,440 a consistent and reliable power source is essential. 72 00:02:58,440 --> 00:02:59,000 Yeah. 73 00:02:59,000 --> 00:03:01,880 That's where these companies see nuclear power fitting in. 74 00:03:01,880 --> 00:03:03,720 So they're saying it's about reliability, 75 00:03:03,720 --> 00:03:06,480 but are there specific types of nuclear power 76 00:03:06,480 --> 00:03:07,280 they're looking at? 77 00:03:07,280 --> 00:03:10,080 Because when most people hear nuclear, they think Chernobyl. 78 00:03:10,080 --> 00:03:10,400 Yeah. 79 00:03:10,400 --> 00:03:12,240 And that's not exactly reassuring. 80 00:03:12,240 --> 00:03:13,000 You're right. 81 00:03:13,000 --> 00:03:16,600 And both Amazon and Google have been investing 82 00:03:16,600 --> 00:03:19,720 in what they're calling a new breed of nuclear reactors 83 00:03:19,720 --> 00:03:21,760 called small modular reactors. 84 00:03:21,760 --> 00:03:22,760 Or SMRs. 85 00:03:22,760 --> 00:03:23,200 Oh, OK. 86 00:03:23,200 --> 00:03:26,960 The idea is that these reactors are smaller, more efficient, 87 00:03:26,960 --> 00:03:29,800 and potentially safer than traditional nuclear plants. 88 00:03:29,800 --> 00:03:31,960 They argue that this technology can provide 89 00:03:31,960 --> 00:03:34,720 a consistent carbon-free energy source 90 00:03:34,720 --> 00:03:39,280 to meet the growing demands of AI and other data-heavy technologies. 91 00:03:39,280 --> 00:03:42,800 OK, so smaller, more efficient, potentially safer. 92 00:03:42,800 --> 00:03:43,960 That sounds good on paper. 93 00:03:43,960 --> 00:03:45,920 But I can't imagine it's that simple. 94 00:03:45,920 --> 00:03:46,360 Yeah. 95 00:03:46,360 --> 00:03:49,000 I mean, are these SMRs even a proven technology? 96 00:03:49,000 --> 00:03:49,600 Right. 97 00:03:49,600 --> 00:03:51,320 You hit on some of the big questions there. 98 00:03:51,320 --> 00:03:53,320 SMRs are a relatively new technology. 99 00:03:53,320 --> 00:03:55,040 And while there are pilot projects underway, 100 00:03:55,040 --> 00:03:57,040 they haven't been deployed on a large scale yet. 101 00:03:57,040 --> 00:03:57,480 OK. 102 00:03:57,480 --> 00:03:59,960 So there's definitely a degree of uncertainty 103 00:03:59,960 --> 00:04:02,880 about their feasibility and their actual safety record. 104 00:04:02,880 --> 00:04:05,560 And of course, there's the whole issue of public perception. 105 00:04:05,560 --> 00:04:06,000 Right. 106 00:04:06,000 --> 00:04:08,000 Nuclear power has a lot of baggage. 107 00:04:08,000 --> 00:04:10,960 And it's hard to imagine that just shrinking the reactors down 108 00:04:10,960 --> 00:04:13,160 is going to magically erase those concerns. 109 00:04:13,160 --> 00:04:14,200 Absolutely. 110 00:04:14,200 --> 00:04:17,200 Gaining public trust and navigating regulatory hurdles 111 00:04:17,200 --> 00:04:19,320 will be major challenges for Amazon and Google 112 00:04:19,320 --> 00:04:20,560 as they pursue this path. 113 00:04:20,560 --> 00:04:21,160 Yeah. 114 00:04:21,160 --> 00:04:23,080 It's a gamble, for sure. 115 00:04:23,080 --> 00:04:24,840 But it highlights the lengths they're 116 00:04:24,840 --> 00:04:27,960 willing to go to secure their position in the AI race. 117 00:04:27,960 --> 00:04:29,720 But speaking of the AI landscape, 118 00:04:29,720 --> 00:04:33,000 while the big players are making these massive bets 119 00:04:33,000 --> 00:04:35,240 on nuclear-powered data centers, there's 120 00:04:35,240 --> 00:04:37,720 another trend emerging that's worth exploring. 121 00:04:37,720 --> 00:04:39,520 Small language models. 122 00:04:39,520 --> 00:04:40,600 Or SLMs. 123 00:04:40,600 --> 00:04:41,040 Right. 124 00:04:41,040 --> 00:04:42,560 Those were mentioned in one of the articles, too. 125 00:04:42,560 --> 00:04:44,200 Can you remind me what those are all about? 126 00:04:44,200 --> 00:04:47,040 Yeah, so think of them as specialized AI tools. 127 00:04:47,040 --> 00:04:47,560 OK. 128 00:04:47,560 --> 00:04:51,080 Instead of trying to do everything like chatGBT or Google's 129 00:04:51,080 --> 00:04:55,120 barred SLMs, our laser focused on specific tasks 130 00:04:55,120 --> 00:04:56,280 like translating languages. 131 00:04:56,280 --> 00:04:58,440 Generating different kinds of creative text formats. 132 00:04:58,440 --> 00:05:00,320 Or writing different kinds of code. 133 00:05:00,320 --> 00:05:01,800 And because they're more focused, 134 00:05:01,800 --> 00:05:04,000 they require less computing power, 135 00:05:04,000 --> 00:05:05,840 making them cheaper to develop and deploy. 136 00:05:05,840 --> 00:05:06,040 OK. 137 00:05:06,040 --> 00:05:08,520 So instead of one giant AI brain trying to do it all, 138 00:05:08,520 --> 00:05:10,960 we have these smaller, more specialized AI's. 139 00:05:10,960 --> 00:05:11,360 Exactly. 140 00:05:11,360 --> 00:05:13,280 That actually makes a lot of sense. 141 00:05:13,280 --> 00:05:15,920 What are some examples of how these SLMs are being 142 00:05:15,920 --> 00:05:17,520 used in the real world? 143 00:05:17,520 --> 00:05:20,240 One fascinating example is a company called Mistral. 144 00:05:20,240 --> 00:05:21,040 OK. 145 00:05:21,040 --> 00:05:23,360 They've created a suite of SLMs called 146 00:05:23,360 --> 00:05:26,800 Le Ministreux, that are designed to run on everyday devices 147 00:05:26,800 --> 00:05:28,360 like laptops and smartphones. 148 00:05:28,360 --> 00:05:29,440 Wow. 149 00:05:29,440 --> 00:05:31,000 These models are tailored for tasks 150 00:05:31,000 --> 00:05:33,320 like writing, marketing, copy generating code, 151 00:05:33,320 --> 00:05:35,120 and specific programming languages. 152 00:05:35,120 --> 00:05:37,440 Or even creating personalized music. 153 00:05:37,440 --> 00:05:37,840 Wow. 154 00:05:37,840 --> 00:05:40,120 The fact that they can run locally on your device 155 00:05:40,120 --> 00:05:42,280 means you don't need a massive data center 156 00:05:42,280 --> 00:05:45,240 to use them, making AI more accessible and affordable. 157 00:05:45,240 --> 00:05:46,400 AI in your pocket. 158 00:05:46,400 --> 00:05:47,120 Yeah. 159 00:05:47,120 --> 00:05:48,360 That's pretty amazing. 160 00:05:48,360 --> 00:05:52,040 But what does this shift towards smaller, more specialized AI 161 00:05:52,040 --> 00:05:53,600 mean for the average person? 162 00:05:53,600 --> 00:05:55,320 Are we talking about a future where everyone 163 00:05:55,320 --> 00:05:57,240 has their own personal AI assistant? 164 00:05:57,240 --> 00:05:59,600 Or is this more about businesses using AI 165 00:05:59,600 --> 00:06:01,080 in new and creative ways? 166 00:06:01,080 --> 00:06:02,600 It's probably a bit of both. 167 00:06:02,600 --> 00:06:05,600 On the one hand, we're already seeing AI systems become 168 00:06:05,600 --> 00:06:08,680 more commonplace in our smartphones and smart homes. 169 00:06:08,680 --> 00:06:11,440 These SLMs could make those systems even more powerful 170 00:06:11,440 --> 00:06:12,640 and personalized. 171 00:06:12,640 --> 00:06:15,560 But on a broader level, this trend towards specialization 172 00:06:15,560 --> 00:06:17,520 could be really transformative for businesses, 173 00:06:17,520 --> 00:06:19,920 especially smaller companies or startups. 174 00:06:19,920 --> 00:06:21,640 So instead of needing massive budgets 175 00:06:21,640 --> 00:06:23,760 to develop their own AI companies, 176 00:06:23,760 --> 00:06:25,800 we could just plug in these specialized SLMs 177 00:06:25,800 --> 00:06:29,200 to automate specific tasks or improve their existing workflows. 178 00:06:29,200 --> 00:06:30,720 I can see how that would be a game changer. 179 00:06:30,720 --> 00:06:31,840 Exactly. 180 00:06:31,840 --> 00:06:34,960 It lowers the barrier to entry for AI adoption. 181 00:06:34,960 --> 00:06:37,920 And it allows for more targeted and efficient use 182 00:06:37,920 --> 00:06:39,840 of this technology. 183 00:06:39,840 --> 00:06:42,880 And this brings us to another crucial question. 184 00:06:42,880 --> 00:06:44,080 The future of work. 185 00:06:44,080 --> 00:06:45,080 The future of work. 186 00:06:45,080 --> 00:06:45,920 Yeah. 187 00:06:45,920 --> 00:06:48,560 It feels like every time we talk about AI, 188 00:06:48,560 --> 00:06:50,400 that topic looms large. 189 00:06:50,400 --> 00:06:53,560 Is this where we finally see widespread job displacement? 190 00:06:53,560 --> 00:06:54,680 It's a valid concern. 191 00:06:54,680 --> 00:06:56,880 And it's definitely something we need to be mindful of. 192 00:06:56,880 --> 00:06:59,000 But I think it's not necessarily 193 00:06:59,000 --> 00:07:01,560 about AI replacing jobs outright. 194 00:07:01,560 --> 00:07:03,200 It's more nuanced than that. 195 00:07:03,200 --> 00:07:05,080 Remember those neural symbolic AI systems 196 00:07:05,080 --> 00:07:07,480 we talked about earlier, the ones designed to mimic 197 00:07:07,480 --> 00:07:08,840 the reasoning of a lawyer? 198 00:07:08,840 --> 00:07:10,760 Right, the ones that could potentially revolutionize 199 00:07:10,760 --> 00:07:11,720 legal work. 200 00:07:11,720 --> 00:07:13,640 But you're saying it's bigger than just one profession. 201 00:07:13,640 --> 00:07:14,600 Exactly. 202 00:07:14,600 --> 00:07:17,960 This type of AI, which combines intuitive reasoning 203 00:07:17,960 --> 00:07:20,240 with large scale data processing. 204 00:07:20,240 --> 00:07:21,880 Has implications for any job that 205 00:07:21,880 --> 00:07:23,920 requires analysis decision making? 206 00:07:23,920 --> 00:07:25,440 Problem solving. 207 00:07:25,440 --> 00:07:27,440 You know, that's a lot of jobs. 208 00:07:27,440 --> 00:07:27,800 Yeah. 209 00:07:27,800 --> 00:07:30,640 But instead of thinking about it as a job apocalypse. 210 00:07:30,640 --> 00:07:32,440 What if we considered it an opportunity 211 00:07:32,440 --> 00:07:35,000 to kind of rethink how we work all together? 212 00:07:35,000 --> 00:07:37,400 So instead of AI taking over jobs, 213 00:07:37,400 --> 00:07:41,160 it's more about AI augmenting human capabilities. 214 00:07:41,160 --> 00:07:42,680 Precisely. 215 00:07:42,680 --> 00:07:47,120 Imagine a world where AI handles the tedious repetitive tasks 216 00:07:47,120 --> 00:07:50,800 freeing up humans to focus on more creative strategic 217 00:07:50,800 --> 00:07:53,200 or interpersonal aspects of their work. 218 00:07:53,200 --> 00:07:55,440 This could lead to entirely new job categories 219 00:07:55,440 --> 00:07:56,880 we haven't even imagined yet. 220 00:07:56,880 --> 00:07:58,680 That's a more optimistic outlook. 221 00:07:58,680 --> 00:08:00,800 But it also raises the question of control. 222 00:08:00,800 --> 00:08:04,280 If AI is going to play such a significant role in our lives, 223 00:08:04,280 --> 00:08:07,240 who gets to decide how it's deployed and regulated? 224 00:08:07,240 --> 00:08:07,600 Right. 225 00:08:07,600 --> 00:08:09,320 And this is where things get really interesting. 226 00:08:09,320 --> 00:08:11,000 While companies like Amazon and Google 227 00:08:11,000 --> 00:08:13,400 are making these massive investments in AI, 228 00:08:13,400 --> 00:08:15,360 there's a growing movement advocating 229 00:08:15,360 --> 00:08:17,160 for a more decentralized approach. 230 00:08:17,160 --> 00:08:19,080 One of the articles you shared talked about a platform 231 00:08:19,080 --> 00:08:21,840 called Assister, which uses blockchain technology 232 00:08:21,840 --> 00:08:24,000 to create a network of these SLMs. 233 00:08:24,000 --> 00:08:24,800 OK, blockchain. 234 00:08:24,800 --> 00:08:27,120 I'm going to need a little help understanding how that fits 235 00:08:27,120 --> 00:08:28,240 into all of this. 236 00:08:28,240 --> 00:08:32,480 So imagine a system where communities, not just corporations, 237 00:08:32,480 --> 00:08:37,080 can own, manage, and even profit from AI development. 238 00:08:37,080 --> 00:08:39,880 That's the potential of blockchain in this space. 239 00:08:39,880 --> 00:08:41,960 Assister, for example, allows developers 240 00:08:41,960 --> 00:08:44,720 to contribute their SLMs to a shared network. 241 00:08:44,720 --> 00:08:47,080 And then users can access those models 242 00:08:47,080 --> 00:08:48,920 through a blockchain-based system. 243 00:08:48,920 --> 00:08:49,680 OK. 244 00:08:49,680 --> 00:08:52,760 This means that instead of a handful of tech giants 245 00:08:52,760 --> 00:08:54,280 controlling the AI landscape. 246 00:08:54,280 --> 00:08:57,440 We could see a more open and collaborative ecosystem 247 00:08:57,440 --> 00:09:00,760 where the benefits of AI are more widely distributed. 248 00:09:00,760 --> 00:09:02,920 So instead of having to rely on Amazon or Google 249 00:09:02,920 --> 00:09:06,640 for AI services, you could have a decentralized network 250 00:09:06,640 --> 00:09:09,600 where anyone can contribute and everyone can benefit. 251 00:09:09,600 --> 00:09:09,960 Right. 252 00:09:09,960 --> 00:09:11,360 That's a pretty compelling vision. 253 00:09:11,360 --> 00:09:11,800 It is. 254 00:09:11,800 --> 00:09:14,240 And it speaks to the larger question of what kind of future 255 00:09:14,240 --> 00:09:15,880 we want to build with this technology. 256 00:09:15,880 --> 00:09:18,320 Do we want a future where AI reinforces 257 00:09:18,320 --> 00:09:20,360 existing power structures? 258 00:09:20,360 --> 00:09:22,800 Or do we want a future where AI empowers 259 00:09:22,800 --> 00:09:24,000 individuals and communities? 260 00:09:24,000 --> 00:09:24,600 Right. 261 00:09:24,600 --> 00:09:26,720 These are not easy questions, but they're questions 262 00:09:26,720 --> 00:09:27,960 we need to be asking ourselves. 263 00:09:27,960 --> 00:09:30,200 Because the future is being written in algorithms, 264 00:09:30,200 --> 00:09:31,120 as they say. 265 00:09:31,120 --> 00:09:31,800 Exactly. 266 00:09:31,800 --> 00:09:33,680 And the more informed we all are, 267 00:09:33,680 --> 00:09:36,240 the better equipped will be to shape that future. 268 00:09:36,240 --> 00:09:37,280 Well said. 269 00:09:37,280 --> 00:09:40,200 So to our listener out there who's been on this deep dive 270 00:09:40,200 --> 00:09:43,640 with us today, here's something to think about. 271 00:09:43,640 --> 00:09:45,840 As you go about your day, pay attention 272 00:09:45,840 --> 00:09:48,120 to where AI is already present in your life, 273 00:09:48,120 --> 00:09:50,600 where you think it might pop up next. 274 00:09:50,600 --> 00:09:52,800 From the apps on your phone to the way you work, 275 00:09:52,800 --> 00:09:54,840 AI is becoming increasingly integrated 276 00:09:54,840 --> 00:09:56,520 into the fabric of our lives. 277 00:09:56,520 --> 00:09:59,320 The choices we make today, both individually and collectively, 278 00:09:59,320 --> 00:10:01,520 will determine the trajectory of this technology 279 00:10:01,520 --> 00:10:07,560 and its impact on our world.