1 00:00:00,120 --> 00:00:03,439 Speaker 1: Imagine a technology that promises a utopia, you know, to 2 00:00:03,640 --> 00:00:06,040 solve all our medical mysteries, write all our code, run 3 00:00:06,040 --> 00:00:09,560 all our errands. Now imagine that exact same technology is 4 00:00:09,599 --> 00:00:13,000 also causing your local utility build to spike, making your 5 00:00:13,039 --> 00:00:16,920 favorite brands produce these really bizarre, low quality ads, and 6 00:00:17,440 --> 00:00:20,960 potentially putting entire sectors of the white collar workforce out 7 00:00:20,960 --> 00:00:25,280 of a job. That massive chaotic friction point, that tension 8 00:00:25,320 --> 00:00:29,399 between accelerated progress and diccelerated public trust. That's precisely where 9 00:00:29,399 --> 00:00:32,600 we are heading in twenty twenty six. Welcome to Thrilling Threads, 10 00:00:32,640 --> 00:00:34,759 the deep dive into the information you need when you 11 00:00:34,840 --> 00:00:37,799 need it. Today, we're unpacking a stack of sources that 12 00:00:37,880 --> 00:00:41,399 offer truly fascinating, and to be honest, a deeply sobering 13 00:00:41,399 --> 00:00:44,840 look into the immediate future of artificial intelligence. We're specifically 14 00:00:44,840 --> 00:00:48,439 focusing on the deeply chaotic landscape predicted for twenty twenty six. 15 00:00:48,719 --> 00:00:50,479 This isn't just about what the next model can do. 16 00:00:50,520 --> 00:00:53,079 It's about what society does back to the models. 17 00:00:53,200 --> 00:00:56,000 Speaker 2: It really is. The sources you provided are absolutely critical 18 00:00:56,039 --> 00:00:59,079 because they map out eight key shifts in the AI world, 19 00:00:59,280 --> 00:01:03,079 and they move far beyond just simple model upgrades. This 20 00:01:03,200 --> 00:01:07,359 material focuses entirely on the second and third order consequences, 21 00:01:07,439 --> 00:01:10,280 the ripples that actually hit the general public. Our mission 22 00:01:10,319 --> 00:01:12,799 today is to move past that initial tech hype and 23 00:01:12,920 --> 00:01:16,680 understand what happens when AI truly hits the mainstream workforce, 24 00:01:16,760 --> 00:01:20,359 the political arenas. We're covering everything from political fallout and 25 00:01:20,400 --> 00:01:26,840 corporate dominance to a surprising and maybe very necessary rebirth 26 00:01:27,000 --> 00:01:28,560 of certain types of physical labor. 27 00:01:28,680 --> 00:01:32,439 Speaker 1: Exactly, we're diving into job displacement. We're exploring the overwhelming 28 00:01:32,480 --> 00:01:35,239 structural strength of one tech titan over all the others, 29 00:01:35,599 --> 00:01:38,840 and analyzing what really looks like a major societal reckoning. 30 00:01:38,959 --> 00:01:41,760 We're going to explore how rapidly public sentiment is turning 31 00:01:41,760 --> 00:01:44,680 into a genuine political force, and how that might inadvertently 32 00:01:44,719 --> 00:01:48,519 halt progress in the West, creating this massive self inflicted wound. 33 00:01:48,319 --> 00:01:51,959 Speaker 2: Right while competitors like China just keep accelerating completely unabated. 34 00:01:52,120 --> 00:01:57,680 Speaker 1: It's a profound tension, rapidly accelerating technology and rapidly decelerating 35 00:01:57,680 --> 00:02:00,719 public approval. Okay, let's unpack this so and I would 36 00:02:00,799 --> 00:02:03,840 argue the most socially important prediction for twenty twenty six 37 00:02:04,280 --> 00:02:07,439 is that the backlash against AI isn't just going to increase, 38 00:02:07,680 --> 00:02:11,759 it's going to dominate public consciousness. Our sources suggest this 39 00:02:11,879 --> 00:02:15,080 hostility is going to reach and I'm quoting here completely 40 00:02:15,120 --> 00:02:18,120 off the charts levels, and the simple root cod seems 41 00:02:18,159 --> 00:02:20,159 to be that the average person is feeling all of 42 00:02:20,159 --> 00:02:24,159 the negatives without seeing any of the corresponding tangible benefits 43 00:02:24,199 --> 00:02:25,159 in their daily lives. 44 00:02:25,439 --> 00:02:27,879 Speaker 2: That's the critical insight. I mean, we who spend in 45 00:02:27,879 --> 00:02:31,000 our days reading about large language models and foundation models, 46 00:02:31,000 --> 00:02:34,240 we're going a tiny, tiny minority for the average user. 47 00:02:34,360 --> 00:02:36,520 The utility of AI right now, it kind of stops 48 00:02:36,560 --> 00:02:40,439 at a basic chatbot or maybe some image generation, and 49 00:02:40,520 --> 00:02:42,400 even those experiences are often pretty mixed. 50 00:02:42,439 --> 00:02:43,680 Speaker 1: Yeah, they can be pretty hit or miss. 51 00:02:44,000 --> 00:02:47,120 Speaker 2: Exactly if you look at the evidence, the mass user 52 00:02:47,159 --> 00:02:50,240 base simply is not being helped in a fundamental way 53 00:02:50,280 --> 00:02:54,479 that improves their quality of life. Instead, they're primarily experiencing 54 00:02:54,479 --> 00:02:57,000 what the source is called the second and third consequences. 55 00:02:57,400 --> 00:02:59,759 And you know, a lot of those consequences are actively 56 00:02:59,800 --> 00:03:02,840 making their lives worse or at least more inconvenient. 57 00:03:03,159 --> 00:03:05,960 Speaker 1: I see that reflecting most acutely in just the sheer, 58 00:03:06,280 --> 00:03:09,639 visceral sense of annoyance people have right now. It goes 59 00:03:09,680 --> 00:03:15,080 beyond some philosophical resistance to an existential threat. It's the practical, 60 00:03:15,319 --> 00:03:18,879 day to day irritation of overpromising. I mean, how many 61 00:03:18,879 --> 00:03:21,520 times have we heard that AI will solve all our problems, 62 00:03:21,800 --> 00:03:25,479 only for a specific app to fail or produce wildly 63 00:03:25,560 --> 00:03:29,159 inaccurate results half the time. That expectation gap, that massive 64 00:03:29,240 --> 00:03:32,000 chasm between the marketing hype and the reality of the 65 00:03:32,000 --> 00:03:35,360 tech's current utility, that is a huge source of public resentment. 66 00:03:35,639 --> 00:03:37,800 It just creates disillusionment at scale. 67 00:03:37,919 --> 00:03:40,680 Speaker 2: And this is intensely compounded by the feeling that AI 68 00:03:40,840 --> 00:03:43,879 is being forced into everything. The sources pointed directly to 69 00:03:43,919 --> 00:03:46,560 the experience of companies shoving AI down your throats. Think 70 00:03:46,599 --> 00:03:50,240 about the experience of just updating your operating system. A 71 00:03:50,280 --> 00:03:53,479 really concrete example of this is Microsoft Hole Pilot being 72 00:03:53,520 --> 00:03:57,680 pushed everywhere in Windows. It's integrated into every taskbar, every 73 00:03:57,680 --> 00:04:01,120 suite of tools. For a user who didn't explicitly ask 74 00:04:01,199 --> 00:04:04,039 for it, who may not see a clear, measurable benefit, 75 00:04:04,159 --> 00:04:07,199 it feels like nothing more than you know, digital bloatwear, or. 76 00:04:07,319 --> 00:04:09,840 Speaker 1: Just an unwelcome, invasive corporate intrusion. 77 00:04:10,039 --> 00:04:14,919 Speaker 2: Exactly this creates a generalized public annoyance that rapidly coagulates 78 00:04:14,919 --> 00:04:17,800 into a wider, more potent backlash. They are getting the 79 00:04:17,839 --> 00:04:21,120 inconvenience first and the promised utility second. 80 00:04:21,519 --> 00:04:24,240 Speaker 1: If at all, that's a terrible dynamic. I mean, it 81 00:04:24,279 --> 00:04:27,920 shifts AI from being this exciting optional tool to a 82 00:04:27,920 --> 00:04:30,920 corporate imposition that you just have to endure, which is 83 00:04:31,160 --> 00:04:34,399 arguably the worst possible brand position for a new technology. 84 00:04:34,920 --> 00:04:37,800 But the backlash isn't just driven by annoying product placements 85 00:04:37,879 --> 00:04:41,759 or disappointing chatbots. It's driven by material costs that hit 86 00:04:41,800 --> 00:04:43,480 people directly in their pocketbooks. 87 00:04:43,480 --> 00:04:44,959 Speaker 2: Oh absolutely, When you can feel. 88 00:04:44,759 --> 00:04:47,079 Speaker 1: The impact of a technology in your monthly budget, the 89 00:04:47,120 --> 00:04:50,079 whole tone of the conversation changes instantly. 90 00:04:49,879 --> 00:04:52,680 Speaker 2: It really does. We need to explore the financial and 91 00:04:52,759 --> 00:04:57,920 infrastructural drivers of this public resentment, and they are profound. First, 92 00:04:57,959 --> 00:05:01,839 you have high hardware prices and the resulting shortages people 93 00:05:01,879 --> 00:05:04,839 trying to buy high end graphics cards for legitimate professional 94 00:05:04,839 --> 00:05:08,319 work or even just for competitive gaming. They're facing historically 95 00:05:08,360 --> 00:05:12,560 inflated costs in massive scarcity, and they're making a direct 96 00:05:12,560 --> 00:05:15,439 link accurately back to the massive sums of money and 97 00:05:15,480 --> 00:05:19,079 physical resources being vacuumed up and invested in AI infrastructure 98 00:05:19,079 --> 00:05:24,120 and training clusters. That scarcity breeds a really palpable frustration, and. 99 00:05:24,079 --> 00:05:27,759 Speaker 1: The investment is staggering. We're talking about billions upon billions 100 00:05:27,800 --> 00:05:31,920 just poured into securing GPUs and building these specialized data centers. 101 00:05:32,199 --> 00:05:34,319 Speaker 2: And second, and this is perhaps more worrying for the 102 00:05:34,360 --> 00:05:38,759 wider non technical public, are the rising electricity prices. AI 103 00:05:38,879 --> 00:05:42,680 training and usage are notoriously brutally energy intensive. I heard 104 00:05:42,720 --> 00:05:46,079 this training a single massive model can use the energy 105 00:05:46,079 --> 00:05:49,600 equivalent of heating hundreds of homes for a year. People 106 00:05:49,680 --> 00:05:52,120 are starting to connect their growing electric bills, which are 107 00:05:52,120 --> 00:05:56,399 often skyrocketing due to infrastructural strain and increased demand. Back 108 00:05:56,439 --> 00:05:58,480 to this relentless AI expansion. 109 00:05:58,600 --> 00:06:02,920 Speaker 1: So when you feel that direct negative financial impact, your 110 00:06:03,000 --> 00:06:05,920 lights cost more because a company is training in LLM, 111 00:06:06,480 --> 00:06:09,399 the technological marvel stops feeling so marvelous. 112 00:06:09,399 --> 00:06:10,800 Speaker 2: It starts feeling predatory. 113 00:06:10,920 --> 00:06:13,519 Speaker 1: So the narrative transforms from AI will help us save 114 00:06:13,639 --> 00:06:17,800 time to AI is draining resources and spiking my utility bill. 115 00:06:18,040 --> 00:06:21,800 That shift is just toxic politically, which of course leads 116 00:06:21,839 --> 00:06:25,240 to the ultimate anxiety mentioned in the source material. What happens. 117 00:06:25,279 --> 00:06:29,480 If this massive investment bubble just crashes, We've seen unbelievable valuations. 118 00:06:29,720 --> 00:06:32,879 If the public benefit doesn't materialize rapidly enough, or if 119 00:06:32,879 --> 00:06:36,279 the economic upheaval caused by displacement is too severe. 120 00:06:36,040 --> 00:06:39,319 Speaker 2: The collective financial loss and disillusionment could be profound. It 121 00:06:39,360 --> 00:06:43,360 could rival past tech crashes. That inherent market volatility just 122 00:06:43,519 --> 00:06:46,560 fuels the fire of the public backlash even before any 123 00:06:46,560 --> 00:06:47,240 crash appens. 124 00:06:47,319 --> 00:06:49,199 Speaker 1: Right, That underlying anxiety is crucial. 125 00:06:49,399 --> 00:06:52,399 Speaker 2: The investment tweet we analyze summed up the situation perfectly. 126 00:06:52,920 --> 00:06:57,199 The backlash is fueled by cost shortages, massive investment, and 127 00:06:57,319 --> 00:07:00,800 energy prices. But the ultimate trigger, the point where things 128 00:07:00,839 --> 00:07:04,720 become off the charts, is when large scale job displacement 129 00:07:04,800 --> 00:07:08,319 begins with no perceived benefit ready to offset this. 130 00:07:08,839 --> 00:07:12,600 Speaker 1: That combination job loss hitting while infrastructure costs arising. That 131 00:07:12,720 --> 00:07:16,319 is politically and socially explosive. Okay, So if section one 132 00:07:16,360 --> 00:07:18,720 was about the build up of resentment, section two is 133 00:07:18,759 --> 00:07:21,800 where the real societal consequences start to hit. Let's start 134 00:07:21,800 --> 00:07:25,519 with the economic earthquake. Everyone is anticipating the impending job 135 00:07:25,560 --> 00:07:29,639 displacement crisis. The sources suggest the backlash will peak precisely 136 00:07:29,680 --> 00:07:32,920 when this job displacement starts with the very real prediction 137 00:07:33,000 --> 00:07:36,120 of potential literal civil unrest in the streets. Wow, that 138 00:07:36,160 --> 00:07:38,839 isn't a hyperbolic statement to grab attention. It reflects a 139 00:07:38,879 --> 00:07:43,000 deep seated fear about rapid, large scale economic instability. 140 00:07:42,680 --> 00:07:46,920 Speaker 2: And that prediction of potential civil unrest is directly linked 141 00:07:46,959 --> 00:07:50,000 to the sheer scale and the speed of displacement. The 142 00:07:50,040 --> 00:07:52,639 CEO of con Academy, who is someone with a clear 143 00:07:52,800 --> 00:07:56,920 view into educational and technological trends, he noted that AI 144 00:07:57,000 --> 00:08:00,240 will displace workers at a scale most people simply don't 145 00:08:00,279 --> 00:08:01,079 yet realize. 146 00:08:01,240 --> 00:08:01,439 Speaker 1: Right. 147 00:08:01,759 --> 00:08:07,439 Speaker 2: He actually offered a surprisingly proactive proposed solution. Companies benefiting 148 00:08:07,480 --> 00:08:10,319 from AI should devote one percent of their profits to 149 00:08:10,360 --> 00:08:15,720 retain or retrain displaced workers. His rationale was straightforward. Without 150 00:08:15,720 --> 00:08:20,240 this internal corporate responsibility, they will face tremendous public backlash. 151 00:08:20,560 --> 00:08:22,959 Speaker 1: That sounds like an admirable attempt at a private sector solution, 152 00:08:23,160 --> 00:08:25,720 you know, putting the onus on the wealthy companies causing 153 00:08:25,759 --> 00:08:28,399 the disruption to christion the blow. But I have to ask, 154 00:08:28,720 --> 00:08:31,160 what was the structural counter argument the sources brought up? 155 00:08:31,199 --> 00:08:34,519 Isn't this just pushing a government problem on to corporations precisely. 156 00:08:35,080 --> 00:08:37,960 Speaker 2: The immediate counter argument that was noted was that while 157 00:08:38,000 --> 00:08:42,559 corporate responsibility is vital, managing a workforce transition that affects 158 00:08:42,639 --> 00:08:46,960 millions of Americans and potentially hundreds of millions globally, that 159 00:08:47,120 --> 00:08:51,200 is fundamentally a government and policy problem. It's not when 160 00:08:51,240 --> 00:08:53,799 the private sector should be solely responsible for fixing. 161 00:08:53,919 --> 00:08:56,679 Speaker 1: You can't ask Google or open ai to craft national 162 00:08:56,720 --> 00:08:57,679 economic policy. 163 00:08:57,919 --> 00:09:02,639 Speaker 2: No. Managing this scale requires systemic policy changes, things like 164 00:09:02,879 --> 00:09:07,720 national retraining mandates, updated regulatory frameworks for labor, or discussions 165 00:09:07,720 --> 00:09:11,200 about the feasibility of mechanisms like a universal basic income. 166 00:09:11,639 --> 00:09:14,919 Placing the full burden on corporations might slow innovation, which 167 00:09:14,919 --> 00:09:18,200 has its own national security risks, without actually solving the 168 00:09:18,240 --> 00:09:20,399 deep structural problem of mass displacement. 169 00:09:20,559 --> 00:09:22,919 Speaker 1: And the displacement isn't just confined to the office worker 170 00:09:22,919 --> 00:09:25,960 who suddenly losing their capacity to write emails. We have 171 00:09:26,039 --> 00:09:29,879 to look far beyond just large language models disrupting knowledge 172 00:09:29,879 --> 00:09:31,039 work correct. 173 00:09:31,120 --> 00:09:34,639 Speaker 2: The scale of displacement is massive when you consider autonomous vehicles. 174 00:09:35,039 --> 00:09:37,720 We are on the verge of autonomous cars dominating ride 175 00:09:37,720 --> 00:09:41,559 sharing services and high level AI systems taking over long 176 00:09:41,600 --> 00:09:45,600 haul trucking and last mild delivery routes. These are millions 177 00:09:45,639 --> 00:09:49,559 of traditionally stable, blue collar jobs primarily held by men 178 00:09:49,879 --> 00:09:53,320 with families relying on them. This automation wave is predicted 179 00:09:53,360 --> 00:09:57,120 to hit faster and cut deeper across multiple labor sectors, simultaneously, 180 00:09:57,600 --> 00:10:02,120 leading to truly enormous sudden econo upheaval among traditionally secure 181 00:10:02,200 --> 00:10:02,840 labor pools. 182 00:10:03,159 --> 00:10:05,840 Speaker 1: That level of economic shock leads us directly into the 183 00:10:05,840 --> 00:10:09,200 political arena. I mean, when millions of people are feeling 184 00:10:09,240 --> 00:10:12,799 economically insecure, anxious about their job security, and angry about 185 00:10:12,840 --> 00:10:16,399 their rising electric bills, they look to their elected representatives 186 00:10:16,399 --> 00:10:19,519 for answers. And the data from the sources suggests AI 187 00:10:19,639 --> 00:10:23,279 is already becoming deeply unpopular in America. It's a toxic subject. 188 00:10:23,320 --> 00:10:27,440 Speaker 2: It's rapidly turning into a political liability that crosscuts ideological lines. 189 00:10:27,840 --> 00:10:31,480 The source is noted a very specific political phenomenon. Republican 190 00:10:31,559 --> 00:10:34,000 senators and House members are afraid to even mention the 191 00:10:34,000 --> 00:10:38,159 word AI publicly wow Yes, because when they do, their 192 00:10:38,200 --> 00:10:42,919 popularity ratings immediately plummet. This indicates that the narrative being 193 00:10:42,960 --> 00:10:46,200 pushed by the most skeptical voices connecting AI directly to 194 00:10:46,320 --> 00:10:51,360 job cuts, economic anxiety, rising consumer costs. That is gaining 195 00:10:51,519 --> 00:10:55,639 highly significant traction with the electorate. It's making AI synonymous 196 00:10:55,639 --> 00:10:56,519 with economic pain. 197 00:10:56,960 --> 00:11:00,639 Speaker 1: That political caution presents a major problem, though especially for 198 00:11:00,679 --> 00:11:02,600 global competitiveness. 199 00:11:02,000 --> 00:11:05,960 Speaker 2: It does. The major concern highlighted by investors and defense 200 00:11:06,000 --> 00:11:09,399 analysts is the risk of an own goal. If it 201 00:11:09,440 --> 00:11:12,919 becomes politically popular or necessary for politicians to push back 202 00:11:12,919 --> 00:11:16,200 against AI, the US might slow down its progress through 203 00:11:16,279 --> 00:11:20,879 overly restrictive, fear driven regulation or by diverting investment through 204 00:11:20,919 --> 00:11:24,879 punitive measures. The finding was stark. While the US may pause, 205 00:11:24,960 --> 00:11:27,200 competitors like China are not going to slow down. 206 00:11:27,759 --> 00:11:31,279 Speaker 1: So if the US hinders itself because of domestic political unpopularity, 207 00:11:31,360 --> 00:11:34,440 it becomes a severe problem for both national security and 208 00:11:34,519 --> 00:11:36,240 long term economic speriority. 209 00:11:36,440 --> 00:11:39,799 Speaker 2: Absolutely, we desperately need a political path forward that doesn't 210 00:11:39,840 --> 00:11:42,879 sacrifice competitive advantage in the name of political expediency. 211 00:11:43,200 --> 00:11:47,639 Speaker 1: That tension public fear versus global acceleration is the central 212 00:11:47,759 --> 00:11:51,000 defining conflict of twenty twenty six. But let's look at 213 00:11:51,039 --> 00:11:55,000 a consequence that hits the bottom line immediately, brand damage. 214 00:11:55,279 --> 00:11:58,360 Will the backlash reach a point where it financially hurts 215 00:11:58,519 --> 00:11:59,919 companies to invest in AI. 216 00:12:00,240 --> 00:12:03,080 Speaker 2: So that is the truly fascinating commercial question. I mean, 217 00:12:03,360 --> 00:12:06,200 the corporate world is driven by sales, public perception, and 218 00:12:06,240 --> 00:12:09,320 brand equity. If a brand's image is damaged by its 219 00:12:09,399 --> 00:12:12,679 use of AI, sales drop, and that provides an immediate 220 00:12:12,799 --> 00:12:17,159 tangible financial incentive to steer away from poorly executed rush 221 00:12:17,200 --> 00:12:18,120 to AI efforts. 222 00:12:18,240 --> 00:12:20,399 Speaker 1: And we already have a striking real world case study 223 00:12:20,480 --> 00:12:20,759 right now. 224 00:12:20,840 --> 00:12:23,759 Speaker 2: Yes we do. McDonald's whoch, a massive global brand. They 225 00:12:23,840 --> 00:12:26,799 actually pulled down a forty five second AI generated holiday 226 00:12:26,840 --> 00:12:29,879 ad because viewers widely deemed it cringey and unsettling. 227 00:12:29,919 --> 00:12:31,759 Speaker 1: And when we say cringey, we mean the ad was 228 00:12:31,840 --> 00:12:35,519 likely suffering from that uncanny Valley effect right, visual inaccuracies, 229 00:12:35,840 --> 00:12:39,039 or just lack the subtle emotional warmth that human creators 230 00:12:39,080 --> 00:12:40,360 instill in holiday campaigns. 231 00:12:40,399 --> 00:12:43,919 Speaker 2: It felt cheap and manufactured. Precisely, the lesson here is 232 00:12:43,960 --> 00:12:49,399 brutal but clear brand recognition, prestige trust These take years, 233 00:12:49,440 --> 00:12:52,600 often decades to build up. If companies are rushing to 234 00:12:52,639 --> 00:12:55,759 integrate AI merely to cut costs or speed up production, 235 00:12:56,120 --> 00:12:58,960 and the result is a negative brand signal like a cheap, 236 00:12:59,080 --> 00:13:02,279 off putting or confus using ad. The financial loss from 237 00:13:02,320 --> 00:13:05,279 reduced brand equity is immediate and measurable. 238 00:13:05,480 --> 00:13:07,080 Speaker 1: Nobody wants their luxury brand to. 239 00:13:07,000 --> 00:13:09,840 Speaker 2: Look cheap, so the core insight is that the financial 240 00:13:09,879 --> 00:13:13,279 and brand impacts of poor AI implementation will force a 241 00:13:13,320 --> 00:13:16,840 selective slowdown in certain consumer facing sectors. You can't just 242 00:13:16,879 --> 00:13:20,679 automate creative work without quality control. This creates a powerful 243 00:13:20,720 --> 00:13:24,519 commercial pressure either use AI expertly to enhance human output, 244 00:13:24,759 --> 00:13:26,720 or don't use it at all. Where your brand is 245 00:13:26,759 --> 00:13:27,399 exposed to the. 246 00:13:27,320 --> 00:13:30,960 Speaker 1: Public, If AI usage reduces their brand equity, the CEO 247 00:13:31,080 --> 00:13:33,360 has a clear non political incentive to. 248 00:13:33,320 --> 00:13:36,720 Speaker 2: Hit the brakes, and this widespread political and social resistance. 249 00:13:36,720 --> 00:13:40,799 We're discussing this feeling of AI fatigue and annoyance. It 250 00:13:40,879 --> 00:13:44,440 leads directly to one of the most interesting market predictions 251 00:13:44,480 --> 00:13:48,960 for twenty twenty six, a radical value shift toward human creativity. 252 00:13:49,519 --> 00:13:53,399 If AI becomes ubiquitous, fast and often perceived as low 253 00:13:53,519 --> 00:13:58,120 quality or esthetically bland, then content or products explicitly made 254 00:13:58,120 --> 00:14:01,080 by humans will be perceived as higher value, even if 255 00:14:01,120 --> 00:14:04,440 they are technically fundamentally worse in some measurable way. 256 00:14:04,799 --> 00:14:07,120 Speaker 1: This is the concept of the human premium, and it's 257 00:14:07,159 --> 00:14:10,480 a truly counterintuitive market force. The fact that a human 258 00:14:10,519 --> 00:14:13,759 took more time, more effort, and faced more difficulty to 259 00:14:13,840 --> 00:14:16,679 create a piece of content or a product that becomes 260 00:14:16,679 --> 00:14:18,120 the inherent selling point itself. 261 00:14:18,200 --> 00:14:19,679 Speaker 2: The value the friction exactly. 262 00:14:19,720 --> 00:14:22,799 Speaker 1: It's the digital equivalent of buying a bespoke, handcrafted suit 263 00:14:22,919 --> 00:14:26,039 versus a mass produced identical garment, but now it's being 264 00:14:26,039 --> 00:14:27,919 applied to design, writing, and music. 265 00:14:28,039 --> 00:14:30,799 Speaker 2: The sources provided a perfect early illustration of this in 266 00:14:30,840 --> 00:14:34,879 the advertising world. The Porsche car ad Kortia explicitly marketed 267 00:14:34,879 --> 00:14:38,120 its ad campaign as being made with no AI used. 268 00:14:38,279 --> 00:14:42,120 The public response was overwhelmingly positive, with comments noting that 269 00:14:43,159 --> 00:14:45,879 it's crazy that we've reached a point where not using 270 00:14:45,960 --> 00:14:46,639 AI as. 271 00:14:46,559 --> 00:14:48,480 Speaker 1: A flex so it's a selling point. 272 00:14:48,679 --> 00:14:52,919 Speaker 2: People viewed this human only approach this guaranteed authenticity as 273 00:14:52,960 --> 00:14:55,120 the ad campaign's biggest selling point. 274 00:14:55,600 --> 00:14:58,799 Speaker 1: That completely flips the script on what innovation means in marketing. 275 00:14:59,360 --> 00:15:03,240 Instead of dating technological acceleration, the ad is celebrating human 276 00:15:03,240 --> 00:15:06,759 effort and time and the lack of automation. We're back 277 00:15:06,759 --> 00:15:08,320 to valuing the human signature. 278 00:15:08,600 --> 00:15:11,919 Speaker 2: The critical insight here gleaned from the expert commentary is 279 00:15:11,960 --> 00:15:15,080 that normalized AI in high end advertisements makes a company 280 00:15:15,080 --> 00:15:15,759 look cheap. 281 00:15:15,960 --> 00:15:18,039 Speaker 1: Or lazy, or just dester to cut corners. 282 00:15:18,200 --> 00:15:22,320 Speaker 2: Exactly, if AI is cheap and fast avoiding, it signals luxury, 283 00:15:22,399 --> 00:15:24,720 exclusivity and genuine quality. 284 00:15:24,600 --> 00:15:28,600 Speaker 1: And this has massive implications for creative sectors. Real art, 285 00:15:28,960 --> 00:15:32,440 verifiable human creation is now becoming the baseline standard for 286 00:15:32,519 --> 00:15:35,320 brand seeking a luxurious or high entearity image. 287 00:15:35,399 --> 00:15:39,080 Speaker 2: This provides a genuine profit driven business case for keeping 288 00:15:39,159 --> 00:15:43,279 human creators employed in high end projects. It potentially establishes 289 00:15:43,320 --> 00:15:47,440 a profitable market segmentation where non AI creation demands a 290 00:15:47,480 --> 00:15:49,639 significantly higher price point, which. 291 00:15:49,440 --> 00:15:51,600 Speaker 1: Means companies will have to figure out how to authenticate 292 00:15:51,799 --> 00:15:56,120 human made work, maybe through providence tracking or certifications, just 293 00:15:56,159 --> 00:15:57,639 to command that premium price. 294 00:15:57,960 --> 00:16:02,360 Speaker 2: And this value shift towards tan human centric work has 295 00:16:02,440 --> 00:16:06,480 profound implications beyond just advertising. It leads directly to the 296 00:16:06,559 --> 00:16:11,480 second major social consequence predicted, a blue collar revival or renaissance. 297 00:16:11,639 --> 00:16:16,720 Speaker 1: This is a crucial, massive second order effect of AI's relentless, 298 00:16:16,840 --> 00:16:20,799 almost exclusive focus on knowledge, workforce automation. 299 00:16:21,000 --> 00:16:24,000 Speaker 2: I agree this prediction is incredibly compelling because it's driven 300 00:16:24,039 --> 00:16:28,360 by several deeply converging sociological and economic factors. First, the 301 00:16:28,360 --> 00:16:31,600 return on investment for a traditional, expensive four year college degree. 302 00:16:31,639 --> 00:16:35,440 It's been questioned for a decade, but AI displacement accelerates 303 00:16:35,480 --> 00:16:40,240 this doubt tenfold. Second, AI has everyone rethinking job security 304 00:16:40,279 --> 00:16:43,200 in the white collar workforce. Why spend four years getting 305 00:16:43,200 --> 00:16:45,440 a degree for a job that could be fully automated 306 00:16:45,480 --> 00:16:48,120 by a twenty dollars a month subscription service next year. 307 00:16:48,279 --> 00:16:51,840 Speaker 1: And that's compounded by a growing societal malaise. I mean, 308 00:16:51,840 --> 00:16:55,720 the stigma associated with blue collar work is rapidly fading. 309 00:16:56,039 --> 00:17:00,480 At the same time, the pressure, the sometimes meaningless cognitive load, 310 00:17:00,759 --> 00:17:03,840 and the relentless two hundred and forty seven demands of 311 00:17:03,879 --> 00:17:07,960 corporate culture are causing a fundamental rethinking of what success 312 00:17:08,119 --> 00:17:09,599 or stable life even means. 313 00:17:09,799 --> 00:17:13,039 Speaker 2: People are looking for stability, tangible results, and jobs that 314 00:17:13,079 --> 00:17:16,759 are inherently more resistant to immediate digital automation because they 315 00:17:16,759 --> 00:17:18,519 require hands and physical presence. 316 00:17:18,640 --> 00:17:20,079 Speaker 1: And here's the profound irony. 317 00:17:20,440 --> 00:17:23,839 Speaker 2: The very infrastructure required to run the AI that's displacing 318 00:17:23,920 --> 00:17:27,880 knowledge workers is directly fueling this blue collar revival. The 319 00:17:27,920 --> 00:17:32,680 prediction noted that the explosive, unprecedented demand for massive AI 320 00:17:33,000 --> 00:17:36,079 data centers, and you should think thousands of server racks 321 00:17:36,160 --> 00:17:41,319 stretching for miles. That directly fuels the need for electricians, plumbers, 322 00:17:41,559 --> 00:17:46,079 HVAC technicians, and construction workers. They're the ones physically building 323 00:17:46,119 --> 00:17:48,359 the new digital factories, making them some of the most 324 00:17:48,400 --> 00:17:52,200 sought after, highly paid, and stable talent in the entire industry. 325 00:17:52,319 --> 00:17:55,480 Speaker 1: So the economic instead of completely flips, AI is automating 326 00:17:55,559 --> 00:17:57,759 the job with the junior lawyer. But at the same time, 327 00:17:58,200 --> 00:18:02,440 it's guaranteeing a massive, profitable pipeline of work for the 328 00:18:02,440 --> 00:18:06,440 master electrician who needs to wire the server farms that 329 00:18:06,480 --> 00:18:09,160 host the LLM. This is a critical connection to make 330 00:18:09,279 --> 00:18:09,599 it is. 331 00:18:09,839 --> 00:18:12,920 Speaker 2: However, we have to introduce the crucial counterpoint, the necessary 332 00:18:13,039 --> 00:18:15,680 nuance here. Even though blue collar work is seeing a 333 00:18:15,720 --> 00:18:19,799 revival due to infrastructure demand and inherent physical difficulty, AI 334 00:18:20,000 --> 00:18:23,319 is not ignoring it entirely, Okay, Instead, it's assisting it, 335 00:18:23,480 --> 00:18:25,839 which radically changes the nature of the work itself. 336 00:18:25,960 --> 00:18:29,160 Speaker 1: That's the most important distinction. The sources shared a fantastic 337 00:18:29,200 --> 00:18:32,720 anecdote from an each VAC technician. He initially felt safe 338 00:18:32,720 --> 00:18:36,039 thinking AI couldn't touch his job, which relies on physical skill, 339 00:18:36,440 --> 00:18:40,119 but then he encountered this complex proprietary unit he could 340 00:18:40,160 --> 00:18:43,400 in troubleshoot. He joked, it was better off speaking Chinese 341 00:18:43,440 --> 00:18:43,640 to me. 342 00:18:43,920 --> 00:18:44,960 Speaker 2: Ah, I can imagine. 343 00:18:45,000 --> 00:18:47,559 Speaker 1: So what did he do? He took a picture, shot 344 00:18:47,599 --> 00:18:50,480 a video, explained the symptoms to an AI, and within 345 00:18:50,559 --> 00:18:55,000 thirty five seconds, the AI diagnosed the exact complex failure 346 00:18:55,319 --> 00:18:57,000 and provided the step will step solution. 347 00:18:57,240 --> 00:19:00,839 Speaker 2: See this doesn't eliminate the technician, but it transforms the 348 00:19:00,920 --> 00:19:04,039 nature of their expertise. It changes the work from being 349 00:19:04,079 --> 00:19:08,799 dependent on deep specialized proprietary knowledge that takes decades of 350 00:19:08,799 --> 00:19:12,039 field experience to acquire into a skill set supported by 351 00:19:12,039 --> 00:19:15,279 an instantaneous AI powered diagnostic engine. 352 00:19:15,359 --> 00:19:17,400 Speaker 1: So the AI becomes a copilot. 353 00:19:17,599 --> 00:19:20,599 Speaker 2: A copilot not for writing code or drafting legal briefs, 354 00:19:20,799 --> 00:19:24,200 but for complex physical troubleshooting in the field. The human 355 00:19:24,240 --> 00:19:26,759 need for the hands on skill remains, but the cognitive 356 00:19:26,759 --> 00:19:29,559 barrier to entry, the years required to build that internal 357 00:19:29,640 --> 00:19:32,400 database of knowledge, it drops dramatically. 358 00:19:32,559 --> 00:19:36,640 Speaker 1: Its efficiency not full displacement, and it's a necessary adaptation 359 00:19:36,720 --> 00:19:41,839 for the trades. Okay, we've established the intense societal friction 360 00:19:41,960 --> 00:19:44,440 happening on the ground, but now let's shift to the 361 00:19:44,480 --> 00:19:49,480 titans of the industry and their technological acceleration. The sources 362 00:19:49,519 --> 00:19:53,519 predict a clear, almost unassailable leader emerging in twenty twenty six, 363 00:19:53,839 --> 00:19:57,200 and that leader is Google. Now this resonates with a 364 00:19:57,240 --> 00:20:00,000 lot of professionals. I know the speaker and the source material. 365 00:20:00,039 --> 00:20:02,319 I'll mention that in their daily workflow they spend about 366 00:20:02,519 --> 00:20:05,519 fifty percent of their time with Gemini three, thirty percent 367 00:20:05,519 --> 00:20:08,400 with Claude, and only twenty percent with all others combined. 368 00:20:09,119 --> 00:20:12,319 Speaker 2: And critically, they noted that the old standard chat GBT 369 00:20:12,519 --> 00:20:15,759 is simply no longer their daily driver because in constant 370 00:20:15,799 --> 00:20:19,079 split tests conducted and professional in daily life, Gemini three's 371 00:20:19,119 --> 00:20:23,000 performance is just incredible. This isn't superficial praise. It's a 372 00:20:23,000 --> 00:20:26,079 reflection of model performance and reasoning capability that has taken 373 00:20:26,119 --> 00:20:29,480 the industry by surprise. They've clearly invested heavily and it's 374 00:20:29,519 --> 00:20:31,359 paying off in core intelligence metrics. 375 00:20:31,440 --> 00:20:34,200 Speaker 1: But the prediction of Google dominance isn't just about the 376 00:20:34,240 --> 00:20:38,640 current quality of Gemini three being marginally smarter in benchmarks. 377 00:20:38,920 --> 00:20:41,359 If that were the case, others could just catch up. 378 00:20:41,920 --> 00:20:45,400 The true strength lies in something much more fundamental and 379 00:20:45,519 --> 00:20:46,920 much much harder to replicate. 380 00:20:47,000 --> 00:20:51,720 Speaker 2: Their vertical integration superpower exactly. That is the core competitive moat, 381 00:20:51,839 --> 00:20:55,799 the structural edge that makes them almost impossible to compete with. Directly, 382 00:20:56,319 --> 00:21:00,400 Google owns literally every single layer of the AI stack, 383 00:21:00,440 --> 00:21:02,640 from the atoms and the chips to the pixels on 384 00:21:02,680 --> 00:21:06,359 your phone screen. Unlike the leading competitors, many of whom 385 00:21:06,400 --> 00:21:10,680 rely heavily on partnership models, external hardware supply, and disparate platforms, 386 00:21:11,039 --> 00:21:15,480 Google is a fully self contained AI superpower. This comprehensive 387 00:21:15,599 --> 00:21:18,640 end to end control allows them to optimize speed, reduce 388 00:21:18,720 --> 00:21:22,279 training costs, and increased capability in ways a fragmented stack 389 00:21:22,440 --> 00:21:23,720 simply cannot match. 390 00:21:23,960 --> 00:21:27,039 Speaker 1: Okay, let's break down that vertical integration because it's comprehensive. 391 00:21:27,079 --> 00:21:30,640 It goes way beyond just marketing slogans at the foundational 392 00:21:30,640 --> 00:21:34,480 infrastructure level. While every other major AI company is bottlenecked 393 00:21:34,519 --> 00:21:38,000 by relying primarily on Nvidia GPUs, which is the global 394 00:21:38,039 --> 00:21:39,279 supply constraint. 395 00:21:39,200 --> 00:21:43,839 Speaker 2: Google just sidesteps that dependency precisely. They have their own 396 00:21:43,960 --> 00:21:48,519 custom infrastructure, their own TPUs or tensor processing units designed 397 00:21:48,519 --> 00:21:52,319 specifically to accelerate machine learning workflows. They also develop their 398 00:21:52,319 --> 00:21:56,240 own specialized GPUs. Okay, this control over the hardware supply 399 00:21:56,319 --> 00:22:00,720 chain is invaluable for efficiency, security, and especially for cutting 400 00:22:00,759 --> 00:22:04,759 massive costs and training and deployment. When Nvidia faces a bottleneck, 401 00:22:04,880 --> 00:22:07,759 Google just spins up more of its own hardware. 402 00:22:07,359 --> 00:22:09,559 Speaker 1: And then you move up the technology stack, which they 403 00:22:09,599 --> 00:22:10,640 also own entirely. 404 00:22:10,799 --> 00:22:14,039 Speaker 2: They own the models. They have the flagship hyper powerful 405 00:22:14,079 --> 00:22:18,519 Gemini model, the lighter incredibly efficient Gemini Flash, and the 406 00:22:18,599 --> 00:22:22,000 open source model GEMA, which serves the wider developer community. 407 00:22:22,279 --> 00:22:26,440 They are not dependent on external providers for their core intelligence. 408 00:22:25,920 --> 00:22:27,720 Speaker 1: IP okay, so that's the models. 409 00:22:27,920 --> 00:22:30,400 Speaker 2: Then they own the platforms and tools. They have an 410 00:22:30,440 --> 00:22:35,599 extensive ecosystem for enterprise customers, including their cloud layer vert X, 411 00:22:36,240 --> 00:22:40,960 developer environments like Google Ai Studio, and internal knowledge tools 412 00:22:41,000 --> 00:22:44,720 like Notebook LM, Google Flow and Google Wrisk. This allows 413 00:22:44,759 --> 00:22:49,079 them to quickly prototype, test and deploy advancements internally before 414 00:22:49,119 --> 00:22:52,920 releasing them to the world, creating a massive internal feedback loop. 415 00:22:53,319 --> 00:22:57,440 Speaker 1: And crucially, they own the distribution, the massive audience pipeline. 416 00:22:57,640 --> 00:23:01,200 This is where the competition gets completely blown away because 417 00:23:01,240 --> 00:23:02,720 distribution is essentially free. 418 00:23:02,720 --> 00:23:05,839 Speaker 2: For Google, it is they are rolling out Gemini Everywhere 419 00:23:06,240 --> 00:23:10,960 deep integration into Android, replacing the traditional Google Assistant integration 420 00:23:11,000 --> 00:23:12,880 into Google workspaces. 421 00:23:12,319 --> 00:23:14,200 Speaker 1: Box sheets, slides. 422 00:23:13,799 --> 00:23:17,279 Speaker 2: Exactly, and perhaps most importantly, AI overviews embedded right into 423 00:23:17,319 --> 00:23:20,000 Google Search, which remains the largest data and user mode 424 00:23:20,000 --> 00:23:21,160 on the entire Internet. 425 00:23:21,319 --> 00:23:24,400 Speaker 1: And we absolutely cannot forget the long game the research 426 00:23:24,440 --> 00:23:25,880 division DeepMind. 427 00:23:25,720 --> 00:23:28,680 Speaker 2: No, we cannot. They are not just focused on iterative 428 00:23:28,799 --> 00:23:32,920 product updates. They are working on fundamental world altering projects, 429 00:23:33,119 --> 00:23:36,480 things like Alpha forty three, Alpha proof, which deals with 430 00:23:36,519 --> 00:23:41,400 automated theorem proving alpha geometry, and fundamental world models. These 431 00:23:41,400 --> 00:23:44,640 are the foundational technologies that others will need to license 432 00:23:44,759 --> 00:23:45,960 or copy years from now. 433 00:23:46,319 --> 00:23:49,279 Speaker 1: So the core reason Google is predicted to win isn't 434 00:23:49,319 --> 00:23:52,720 just that Gemini might be marginally smarter today. It's that 435 00:23:52,799 --> 00:23:57,279 Google can deploy AI cheaper, faster, and to billions of 436 00:23:57,279 --> 00:24:01,960 more people simultaneously than the partnership will OpenAI Microsoft model. 437 00:24:02,240 --> 00:24:05,920 Speaker 2: Open AI is essentially a specialized product company. Google is that, 438 00:24:06,119 --> 00:24:10,400 and then some a vertically integrated ecosystem. This internal ecosystem 439 00:24:10,440 --> 00:24:13,359 means Google can afford to innovate and iterate much faster. 440 00:24:13,880 --> 00:24:16,640 If DeepMind makes a breakthrough, they can immediately implement it 441 00:24:16,680 --> 00:24:20,960 across their proprietary chip infrastructure, refine it using internal tools 442 00:24:21,119 --> 00:24:24,279 and distribute it to billions of Android or workspace users 443 00:24:24,519 --> 00:24:28,240 without relying on external partners for chips or software integrations, 444 00:24:28,359 --> 00:24:32,519 all in house. That unified, cost optimized, and self reliant 445 00:24:32,519 --> 00:24:35,200 approach is what makes their predicted lead in twenty twenty 446 00:24:35,240 --> 00:24:36,720 six seem virtually locked in. 447 00:24:37,240 --> 00:24:39,279 Speaker 1: But let me play Devil's advocate here for a second. 448 00:24:39,599 --> 00:24:42,319 Google dominance sounds a bit scary, doesn't it? If one 449 00:24:42,319 --> 00:24:45,599 company owns the chips, the models, and the distribution, doesn't 450 00:24:45,640 --> 00:24:49,759 that inherently reduce the competition that ultimately benefits the end user. 451 00:24:50,240 --> 00:24:52,960 Speaker 2: That is a fair question, and the answer is nuanced. 452 00:24:53,799 --> 00:24:57,440 While dominance is always a regulatory concern, the sources suggests 453 00:24:57,480 --> 00:25:00,920 that the presence of fierce competition like anthrop and open Ai, 454 00:25:01,400 --> 00:25:04,519 even if they lack Google's verticality, is good for the 455 00:25:04,640 --> 00:25:08,920 end user because it forces Google to constantly improve Gemini. However, 456 00:25:09,200 --> 00:25:12,119 the predicted dominance means that Google can pass on cost 457 00:25:12,119 --> 00:25:16,839 efficiencies to enterprise customers sooner, potentially undercutting rivals and accelerating 458 00:25:16,839 --> 00:25:20,519 the technologies adoption for better or worse. Because they control 459 00:25:20,599 --> 00:25:22,519 the entire cost structure. 460 00:25:22,119 --> 00:25:24,720 Speaker 1: So the question for regulators in twenty twenty six won't 461 00:25:24,720 --> 00:25:27,079 be about innovation. It will be about market access and 462 00:25:27,119 --> 00:25:30,559 competitive fairness. Exactly okay, moving from corporate power to pure 463 00:25:30,640 --> 00:25:35,119 technological advancement. The next major prediction focuses on unlocking a 464 00:25:35,160 --> 00:25:40,599 core frustrating limitation of current AI models, continual learning or CL. 465 00:25:41,119 --> 00:25:43,759 This is often cited as the key missing ingredient for 466 00:25:43,799 --> 00:25:45,960 truly dynamic evolving intelligence. 467 00:25:46,559 --> 00:25:48,680 Speaker 2: You are absolutely right. This is one of the most 468 00:25:48,680 --> 00:25:53,400 critical missing pieces for achieving true artificial general intelligence. The 469 00:25:53,519 --> 00:25:58,440 current paradigm is extremely expensive and inefficient. Models are frozen 470 00:25:58,480 --> 00:26:00,960 at the time of release. They are trained on a 471 00:26:01,000 --> 00:26:04,240 massive data set, fine tuned, and then launched as a 472 00:26:04,279 --> 00:26:05,160 static product. 473 00:26:05,279 --> 00:26:07,559 Speaker 1: So if new data emerges or if the world changes, 474 00:26:07,880 --> 00:26:10,720 you face what's known as catastrophic forgetting. Yes, and for 475 00:26:10,759 --> 00:26:13,599 anyone listening who hasn't heard that term, catastrophic forgetting sounds 476 00:26:13,640 --> 00:26:16,160 exactly as bad as it is right. It implies total 477 00:26:16,200 --> 00:26:16,960 knowledge loss. 478 00:26:17,240 --> 00:26:20,839 Speaker 2: It is total knowledge loss, and it's a massive technical hurdle. 479 00:26:21,359 --> 00:26:24,559 When developers try to update a large model with new information, 480 00:26:24,720 --> 00:26:28,000 say new stock data or recent world events, they use 481 00:26:28,039 --> 00:26:31,880 the standard method of training called back propagation. This process 482 00:26:32,000 --> 00:26:35,680 updates all the weights globally across the network. Because these 483 00:26:35,759 --> 00:26:39,079 updates are global. The old useful knowledge that the model 484 00:26:39,079 --> 00:26:42,279 spent millions of dollars acquiring gets overwritten and erased by 485 00:26:42,319 --> 00:26:46,319 the new training data. Wow, the model literally forgets critical 486 00:26:46,319 --> 00:26:49,480 things it was taught because the new data points override 487 00:26:49,480 --> 00:26:54,480 the entire architecture. It is inefficient, frustrating, and incredibly expensive. 488 00:26:54,880 --> 00:26:58,119 Speaker 1: So continual learning is the ability for the AI to 489 00:26:58,200 --> 00:27:01,599 update after the initial training supposed to mimic how humans. 490 00:27:01,400 --> 00:27:05,640 Speaker 2: Learn exactly, constantly evolving and building on new information incrementally 491 00:27:05,680 --> 00:27:07,680 without destroying the knowledge that came before it. 492 00:27:07,759 --> 00:27:10,119 Speaker 1: This would represent a fundamental change in the economics of 493 00:27:10,160 --> 00:27:11,000 running an LLM. 494 00:27:11,119 --> 00:27:16,160 Speaker 2: The urgency here is twofold theoretical advancement towards AGI and massive, 495 00:27:16,279 --> 00:27:19,799 massive cost savings for the companies running these models. Training 496 00:27:19,839 --> 00:27:24,279 processes cost tens, sometimes hundreds of millions of dollars. If 497 00:27:24,319 --> 00:27:26,400 companies can figure out a way for models to update 498 00:27:26,400 --> 00:27:30,839 their weights locally and incrementally, only updating the relevant parameters 499 00:27:30,839 --> 00:27:34,559 for the new information, rather than restarting the entire global 500 00:27:34,599 --> 00:27:37,160 training process every time they need new knowledge, they will 501 00:27:37,200 --> 00:27:40,720 save millions in computation costs and dramatically increase the model's 502 00:27:40,759 --> 00:27:41,599 shelf life and. 503 00:27:41,559 --> 00:27:46,200 Speaker 1: This challenge is driving intense competing research efforts globally. We 504 00:27:46,279 --> 00:27:49,039 saw Jeff Dean at Google discussing an exciting new approach 505 00:27:49,119 --> 00:27:53,960 using nested optimization for enhancing long context processing, a critical 506 00:27:54,000 --> 00:27:56,799 step towards this kind of dynamic localized learning. And this 507 00:27:56,839 --> 00:27:58,599 isn't just a Google obsession, not at all. 508 00:27:58,680 --> 00:28:01,359 Speaker 2: The industry consensus is at twenty twenty six is the 509 00:28:01,440 --> 00:28:04,200 year this problem gets cracked or at least substantially mitigated, 510 00:28:04,680 --> 00:28:07,440 and Thropics suggests that cl will be finalized by twenty 511 00:28:07,480 --> 00:28:10,920 twenty six. Meta released a significant paper detailing the sparse 512 00:28:10,960 --> 00:28:14,200 memory fine tuning approach, which is a method specifically designed 513 00:28:14,200 --> 00:28:17,920 for continual learning by isolating new knowledge. Even Elon Musk 514 00:28:18,000 --> 00:28:21,480 is reportedly banking Grock five's long term success on solving 515 00:28:21,640 --> 00:28:25,440 this specific breakthrough. It is universally seen as the next 516 00:28:25,559 --> 00:28:30,039 major hurdle necessary for the technology to truly evolve beyond 517 00:28:30,160 --> 00:28:31,200 static knowledge. 518 00:28:31,480 --> 00:28:34,640 Speaker 1: So fundamental technical challenge as the source is highlighted. It 519 00:28:34,680 --> 00:28:38,319 boils down to the need for a credible, efficient alternative 520 00:28:38,359 --> 00:28:41,799 to standard back propagation, something that allows for local updates 521 00:28:41,799 --> 00:28:44,440 instead of global updates across the vast network. 522 00:28:44,599 --> 00:28:47,440 Speaker 2: Exactly, we are looking for something that allows the model 523 00:28:47,480 --> 00:28:51,119 to partition new memory without corrupting old memory. And while 524 00:28:51,119 --> 00:28:53,839 the Deep Mind CEO believes scaling models will continue to 525 00:28:53,920 --> 00:28:57,279 yield impressive results, he still sees CL as one of 526 00:28:57,319 --> 00:29:00,720 the one or two truly missing breakthrough is required for 527 00:29:00,759 --> 00:29:03,519 AGI potentially still five to ten years away from true 528 00:29:03,559 --> 00:29:07,160 mastery in a generalized sense, but the massive progress predicted 529 00:29:07,160 --> 00:29:09,920 for twenty twenty six will transform how models are maintained 530 00:29:09,920 --> 00:29:10,480 and deployed. 531 00:29:10,720 --> 00:29:14,599 Speaker 1: And while engineers work feverishly on CL, the powerful models 532 00:29:14,599 --> 00:29:17,960 that currently exist are being put to new, much higher 533 00:29:18,000 --> 00:29:21,480 value work. This leads to the next major prediction, the 534 00:29:21,519 --> 00:29:25,359 shift of AI agents away from generic consumer chatbots and 535 00:29:25,400 --> 00:29:29,720 towards specialized, hyper efficient enterprise work. Companies are realizing the 536 00:29:29,720 --> 00:29:32,960 consumer use case is nearly saturated with basic tools. 537 00:29:33,000 --> 00:29:36,400 Speaker 2: The consumer market reached a saturation point quickly. The average 538 00:29:36,400 --> 00:29:39,240 person doesn't need much more than a simple voice assistant 539 00:29:39,279 --> 00:29:42,559 for scheduling or a basic chatbot for general inquiries or 540 00:29:42,640 --> 00:29:46,960 image generation. They don't need a powerful, multi billion parameter 541 00:29:47,039 --> 00:29:49,680 AI that can go off and execute complex multi step 542 00:29:49,839 --> 00:29:54,000 tasks like fixing an email resolution dispute across three different platforms. 543 00:29:54,480 --> 00:29:57,359 The true high margin economically valuable work is in the 544 00:29:57,440 --> 00:30:01,880 enterprise backbone automating tedious knowledge heavy professional tasks. 545 00:30:01,559 --> 00:30:04,519 Speaker 1: And propig realized this early with their cloud code model, 546 00:30:04,799 --> 00:30:07,640 which managed to nearly monopolize the AI coding industry by 547 00:30:07,640 --> 00:30:11,799 focusing on that singular, deep task execution area. Its reputation 548 00:30:11,920 --> 00:30:15,079 is so strong that for complex, multi layered programming tasks, 549 00:30:15,240 --> 00:30:17,240 it has become the undeniable industry standard. 550 00:30:17,480 --> 00:30:20,559 Speaker 2: That success drove the realization across the industry that they 551 00:30:20,599 --> 00:30:24,720 need to focus intensely on automating meaningful, high value work. 552 00:30:25,160 --> 00:30:28,119 This is why we see OpenAI shifting its focus dramatically 553 00:30:28,200 --> 00:30:33,160 towards professional knowledge work, measuring success using the GPT valve benchmark, 554 00:30:33,240 --> 00:30:37,400 which covers task execution across as staggering forty four different 555 00:30:37,400 --> 00:30:39,559 white collar occupations. 556 00:30:38,920 --> 00:30:43,359 Speaker 1: Forty four Wow. They are targeting the heart of corporate. 557 00:30:43,000 --> 00:30:47,160 Speaker 2: Operations and the performance gains reported are staggering. The sources 558 00:30:47,200 --> 00:30:50,759 noted nearly a doubling in performance from previous iterations to 559 00:30:50,799 --> 00:30:53,839 the predicted capabilities of GPT five point two. In this 560 00:30:54,000 --> 00:30:55,160 professional knowledge work. 561 00:30:55,200 --> 00:30:56,960 Speaker 1: This is where the enterprise money is going to be spent, 562 00:30:57,119 --> 00:30:59,000 because the ROI is demonstrable. 563 00:30:59,079 --> 00:31:02,319 Speaker 2: Absolutely, we're seeing major improvements in state of the art 564 00:31:02,400 --> 00:31:06,720 long context reasoning, which is essential for digesting huge corporate documents. 565 00:31:07,119 --> 00:31:10,279 There are also significant leaps predicted in mundane but time 566 00:31:10,319 --> 00:31:14,759 consuming tasks like spreadsheet creation and formatting, and early impressive 567 00:31:14,839 --> 00:31:17,359 gains and high quality slideshow creation. 568 00:31:17,200 --> 00:31:19,960 Speaker 1: The backbone tasks of white collar professional life. 569 00:31:20,160 --> 00:31:23,359 Speaker 2: Exactly, the entire industry is moving away from the tedious 570 00:31:23,440 --> 00:31:27,359 back and forth cyborg central prompting where you constantly check 571 00:31:27,400 --> 00:31:31,200 and correct the AI's work, toward fully agentic work, where 572 00:31:31,200 --> 00:31:34,000 the AI can take a high level goal and execute 573 00:31:34,000 --> 00:31:37,839 the complex multi step subtasks required to complete a professional 574 00:31:37,880 --> 00:31:41,920 project autonomously. To achieve that higher level of agentic work, 575 00:31:41,960 --> 00:31:45,799 in general reasoning capability we just discussed, AI needs better 576 00:31:45,839 --> 00:31:49,680 methods for understanding consequences and simulating reality before it acts. 577 00:31:50,079 --> 00:31:52,720 This leads us to world models, which are predicted to 578 00:31:52,720 --> 00:31:56,680 be the secret sauce behind Google's current and future impressive capabilities. 579 00:31:56,720 --> 00:31:59,440 Speaker 1: Okay, we hear the term world model tossed around a lot. 580 00:31:59,799 --> 00:32:02,319 Is exactly is a world model in this context and 581 00:32:02,359 --> 00:32:05,000 why is it so fundamentally transformative for AI? 582 00:32:05,480 --> 00:32:09,559 Speaker 2: Fundamentally, a world model gives the AI a deep internal 583 00:32:09,759 --> 00:32:13,720 world memory and allows it to generate an interactive, predictable 584 00:32:13,759 --> 00:32:17,079 environment based on physics and logic. Think of it less 585 00:32:17,160 --> 00:32:20,000 like rendering a static image and more like simulating a 586 00:32:20,079 --> 00:32:24,759 miniature working universe, complete with rules, gravity, and consequence. Google's 587 00:32:24,799 --> 00:32:28,319 Genie three model, for example, generates interactive worlds that actually 588 00:32:28,359 --> 00:32:31,599 remember previous states. If you move an object, the object 589 00:32:31,599 --> 00:32:35,359 stays moved. It's not just AI generated nonsense where physics 590 00:32:35,400 --> 00:32:37,319 fails after the first frame. 591 00:32:37,200 --> 00:32:39,960 Speaker 1: So it's not just generating pixels, it's generating rules. 592 00:32:40,240 --> 00:32:44,519 Speaker 2: Precisely. If a model can internalize the fundamental rules of 593 00:32:44,559 --> 00:32:48,640 a simulated world, how gravity works, how objects react when 594 00:32:48,640 --> 00:32:51,279 they collide, how cause and effect plays out over time, 595 00:32:51,839 --> 00:32:54,400 it can plan and reason far more effectively in the 596 00:32:54,400 --> 00:32:57,839 real world. Instead of simply generating the next word in 597 00:32:57,880 --> 00:33:00,920 a sequence, it can generate the nextime logical step in 598 00:33:00,960 --> 00:33:02,000 a sequence of actions. 599 00:33:02,079 --> 00:33:03,119 Speaker 1: Okay, that makes sense. 600 00:33:03,160 --> 00:33:05,359 Speaker 2: World models are predicted to become a large part of 601 00:33:05,359 --> 00:33:09,039 how future reasoning models get exponentially more effective, forming the 602 00:33:09,119 --> 00:33:13,240 indispensable backbone for advanced planning and eventually AGI. 603 00:33:12,880 --> 00:33:16,599 Speaker 1: Systems, and that simulation capability is directly feeding into the 604 00:33:16,640 --> 00:33:20,759 next prediction, the use of generalized agents in complex virtual worlds. 605 00:33:21,240 --> 00:33:25,240 We're moving beyond the restricted, simple reinforcement learning environments we 606 00:33:25,359 --> 00:33:27,440 use in the past to train basic AI systems. 607 00:33:27,799 --> 00:33:30,799 Speaker 2: Yes, the focus is shifting to general AI agents that 608 00:33:30,839 --> 00:33:34,559 can play, learn, and reason in much more complex virtual worlds, 609 00:33:34,759 --> 00:33:38,640 as evidence by developments like Simo two. The significance here 610 00:33:38,720 --> 00:33:41,960 is profound, even if the training environment superficially looks like 611 00:33:42,000 --> 00:33:46,079 a video game. The goal is training general intelligence. 612 00:33:46,000 --> 00:33:49,480 Speaker 1: Because the real world itself, with all its unpredictability and 613 00:33:49,559 --> 00:33:52,880 interconnected systems, is, for lack of a better term, the 614 00:33:52,920 --> 00:33:53,880 biggest game. 615 00:33:53,720 --> 00:33:57,240 Speaker 2: Of all exactly. The general concept is that if AI 616 00:33:57,319 --> 00:34:00,359 agents are able to successfully work in reason within common, complex, 617 00:34:00,480 --> 00:34:04,839 realistic virtual world's, learning high level skills, strategy, long term planning, 618 00:34:04,880 --> 00:34:08,480 and advanced problem solving, there's no reason those learned capabilities 619 00:34:08,519 --> 00:34:12,440 couldn't eventually be transferred or used to power robotics, autonomous vehicles, 620 00:34:12,480 --> 00:34:15,159 and other complex reasoning tasks in the physical world. 621 00:34:15,320 --> 00:34:19,199 Speaker 1: So virtual worlds offer a safe, scalable, and instant reset 622 00:34:19,280 --> 00:34:22,440 environment for trial and error learning that the real world 623 00:34:22,760 --> 00:34:23,800 just cannot provide. 624 00:34:23,840 --> 00:34:26,880 Speaker 2: It's the ultimate cost effective training ground for real world 625 00:34:26,920 --> 00:34:28,920 physical and complex reasoning. 626 00:34:28,599 --> 00:34:31,280 Speaker 1: Tasks, and the source material predicts This will be a 627 00:34:31,480 --> 00:34:35,679 massive focus for Google and deep Mind and potentially yield surprising, 628 00:34:35,880 --> 00:34:39,760 unexpected hacks or outcomes as these agents begin to display 629 00:34:39,800 --> 00:34:43,239 emergent behavior when faced with complex, dynamic. 630 00:34:43,000 --> 00:34:46,960 Speaker 2: Environments, and it connects back perfectly to Google's vertical integration 631 00:34:47,000 --> 00:34:50,960 we discussed earlier. They control the models Gemini, the research, 632 00:34:51,119 --> 00:34:55,320 deep Mind, and the deployment platforms. Simulating and training agents 633 00:34:55,360 --> 00:34:59,159 in virtual worlds allows them to rapidly iterate on complex, 634 00:34:59,320 --> 00:35:03,159 goal orient behaviors far faster and cheaper than training physical 635 00:35:03,239 --> 00:35:04,760 robots in the real world, and. 636 00:35:04,679 --> 00:35:07,920 Speaker 1: Once the agents are robust, the learned skills are optimized 637 00:35:07,960 --> 00:35:12,079 for transfer to their eventual physical or virtual deployment. Speaking 638 00:35:12,079 --> 00:35:14,639 of the physical world, let's wrap up the technological predictions 639 00:35:14,639 --> 00:35:18,159 with the biggest leap of all robotics. The sources are 640 00:35:18,199 --> 00:35:21,599 extremely bullish predicting that twenty twenty six will bring robotics 641 00:35:21,639 --> 00:35:25,159 its own chat GPT moment that is a high bar 642 00:35:25,280 --> 00:35:27,159 for cultural and technological impact. 643 00:35:27,599 --> 00:35:31,159 Speaker 2: That comparison a chat gypt moment is absolutely crucial because 644 00:35:31,320 --> 00:35:36,199 it implies a sudden, viral, undeniable leap in both capability 645 00:35:36,239 --> 00:35:39,719 and public awareness. It means the technology moves from being 646 00:35:39,719 --> 00:35:43,920 an academic curiosity, or a demonstration of slow, deliberate movement 647 00:35:44,280 --> 00:35:48,000 to an undeniable visible reality that appears to operate almost 648 00:35:48,079 --> 00:35:49,000 magically overnight. 649 00:35:49,199 --> 00:35:52,039 Speaker 1: The history of robotics demos is actually pretty funny if 650 00:35:52,079 --> 00:35:55,199 you're cynical. I remember the sources mentioning that some demos 651 00:35:55,199 --> 00:35:57,920 were so incredibly advanced that the public and media couldn't 652 00:35:57,920 --> 00:36:02,360 believe they were real, forcing the develops literally cut open 653 00:36:02,440 --> 00:36:04,559 the leg of the robot to reveal that there was 654 00:36:04,639 --> 00:36:06,320 not a human inside the shell. 655 00:36:06,480 --> 00:36:09,920 Speaker 2: That's the ridiculous standard of credibility physical automation has to meet. 656 00:36:10,000 --> 00:36:12,880 Speaker 1: It is a ridiculously high bar, mainly because the real 657 00:36:12,920 --> 00:36:16,440 world is infinitely more complex and unstructured than a lab floor. 658 00:36:17,119 --> 00:36:19,440 But the prediction is that twenty twenty six will see 659 00:36:19,519 --> 00:36:23,880 more incredible, fully autonomous robotics demos performed seamlessly in messy 660 00:36:24,280 --> 00:36:27,760 real world environments that finally overcome that deep skepticism. 661 00:36:27,920 --> 00:36:30,360 Speaker 2: The tipping point will be when the robots are operating 662 00:36:30,400 --> 00:36:33,719 with the fluidity and intelligence of a human being and complex, 663 00:36:33,920 --> 00:36:38,239 unpredictable domestic or industrial environments, not just following pre program 664 00:36:38,360 --> 00:36:39,480 factory paths, in. 665 00:36:39,440 --> 00:36:42,800 Speaker 1: Which company is predicted to lead this shift and deliver 666 00:36:42,920 --> 00:36:43,920 that viral moment. 667 00:36:44,119 --> 00:36:47,119 Speaker 2: The sources named Physical Robotics with their PI zero and 668 00:36:47,239 --> 00:36:50,639 subsequent PI one point five updates as the most bullish 669 00:36:50,760 --> 00:36:54,440 candidate to pioneer this change. This is significant because they 670 00:36:54,440 --> 00:36:57,719 are backed by billions of dollars of investment, including major 671 00:36:57,760 --> 00:36:59,960 commitments from companies like Amazon. 672 00:36:59,639 --> 00:37:04,199 Speaker 1: Ah and that Amazon investment provides a huge clue. This 673 00:37:04,239 --> 00:37:05,920 isn't about home cleaning robots. 674 00:37:06,000 --> 00:37:10,480 Speaker 2: No, this is about automated logistics, warehousing, and complex physical fulfillment. 675 00:37:10,519 --> 00:37:14,079 Speaker 1: So the expectation is clear. A robotics demo from a 676 00:37:14,119 --> 00:37:17,800 company like Physical Robotics will go completely viral, achieving the 677 00:37:17,880 --> 00:37:22,320 visibility and impact of a chat GPT moment for embodied AI. 678 00:37:22,159 --> 00:37:25,400 Speaker 2: And that sudden public awareness will make physical automation a sudden, 679 00:37:25,480 --> 00:37:28,679 unavoidable and immediate reality for the public and for investors, 680 00:37:28,920 --> 00:37:33,400 getting sceptors that have previously felt completely safe from digital displacement. 681 00:37:33,119 --> 00:37:36,519 Speaker 1: That, combined with the autonomous trucking we discussed earlier, will 682 00:37:36,599 --> 00:37:41,000 radically reshape the physical labor landscape overnight, adding immense pressure 683 00:37:41,039 --> 00:37:43,079 to the political backlash already building. 684 00:37:43,320 --> 00:37:47,920 Speaker 2: So if the knowledge workers displacement creates the political fire, the. 685 00:37:47,920 --> 00:37:51,960 Speaker 1: Robotics moment in twenty twenty six provides the physical, unavoidable 686 00:37:51,960 --> 00:37:54,480 evidence that AI is coming for all labor, which is 687 00:37:54,480 --> 00:37:57,000 why the backlash is predicted to reach those off the 688 00:37:57,119 --> 00:37:57,920 charts levels. 689 00:37:58,079 --> 00:38:02,159 Speaker 2: Exactly. The technological accel oleration in robotics is directly feeding 690 00:38:02,199 --> 00:38:05,320 the fear and resistance in the societal sections we discussed. 691 00:38:05,599 --> 00:38:08,840 Speaker 1: It's all interconnected, so if we synthesize these intense predictions, 692 00:38:09,000 --> 00:38:12,519 twenty twenty six is defined by two major contradictory and 693 00:38:12,559 --> 00:38:13,920 fiercely competing forces. 694 00:38:14,039 --> 00:38:16,880 Speaker 2: On one side, you have massive societal and political resistance, 695 00:38:17,199 --> 00:38:20,679 overwhelming public backlash fueled by cost and job fear, job 696 00:38:20,719 --> 00:38:24,280 loss potentially fueling civil unrest, and politicians running scared from 697 00:38:24,320 --> 00:38:26,800 the word AI. On the other side, you have rapid 698 00:38:26,960 --> 00:38:31,679 fundamental technological breakthroughs, Google's vertical integration locking and dominance, the 699 00:38:31,760 --> 00:38:35,719 core problem of continual learning nearing finalization, and AI agents 700 00:38:35,719 --> 00:38:39,079 shifting to hyper efficient enterprise work, culminating in the undeniable 701 00:38:39,159 --> 00:38:40,800 chat GPT moment. For robotics. 702 00:38:40,920 --> 00:38:44,079 Speaker 1: It is absolutely not a linear upgrade path. It is 703 00:38:44,119 --> 00:38:47,679 a chaotic tension between public fear what they are losing 704 00:38:48,079 --> 00:38:51,880 and corporate acceleration what they're gaining. And that profound tension 705 00:38:51,960 --> 00:38:55,000 is leading to surprising second order effects like the human premium, 706 00:38:55,480 --> 00:38:59,320 the realization that authenticity and human effort are now luxury goods, 707 00:38:59,719 --> 00:39:02,079 and the necessity of a blue collar rebirth fueled by 708 00:39:02,119 --> 00:39:05,760 the very infrastructural demands AI requires. Here's where it gets 709 00:39:05,800 --> 00:39:09,960 really interesting. Considering the extreme political resistance and economic acceleration 710 00:39:10,039 --> 00:39:12,960 predicted for twenty twenty six. If you were an executive 711 00:39:12,960 --> 00:39:15,239 at one of these major AI companies like Google or 712 00:39:15,239 --> 00:39:18,559 open Ai, what single prediction we discussed today would you 713 00:39:18,639 --> 00:39:22,800 prioritize addressing immediately to mitigate the public backlash. Is the 714 00:39:22,880 --> 00:39:25,719 risk of an own goal of slowing down progress to 715 00:39:25,800 --> 00:39:28,679 appease a fearful public worth the catastrophic cost of losing 716 00:39:28,679 --> 00:39:31,480 global competitive advantage. What's your stand? We want to know 717 00:39:31,480 --> 00:39:31,960 what you think.