WEBVTT 00:00:00.000 --> 00:00:02.520 Imagine a future where nations secure their advanced 00:00:02.520 --> 00:00:05.820 AI capabilities, not just building things themselves, 00:00:06.000 --> 00:00:09.000 but through these massive global partnerships. 00:00:09.380 --> 00:00:12.460 We're talking billions of dollars, strategic 00:00:12.460 --> 00:00:16.140 alliances, all shaping tech for decades. Beat 00:00:16.140 --> 00:00:18.660 that future. It's not way off. It's actually 00:00:18.660 --> 00:00:24.160 happening right now. Welcome to the Deep Dive. 00:00:24.570 --> 00:00:26.789 Today, we're unpacking some really fascinating 00:00:26.789 --> 00:00:29.309 sources. They reveal just how fast the whole 00:00:29.309 --> 00:00:31.149 landscape of artificial intelligence is evolving. 00:00:31.370 --> 00:00:34.210 We'll journey from, well, the big stage geopolitical 00:00:34.210 --> 00:00:36.329 tech alliances, national goals, all the way down 00:00:36.329 --> 00:00:39.189 to some groundbreaking open source AI models, 00:00:39.390 --> 00:00:41.530 models challenging the giants. It really shows 00:00:41.530 --> 00:00:43.549 how quickly things are moving. Absolutely. Yeah, 00:00:43.590 --> 00:00:44.969 we've got a lot to cover. First up, we're going 00:00:44.969 --> 00:00:46.969 to pull back the curtain on this huge US -UK 00:00:46.969 --> 00:00:48.789 partnership. It's really shaping AI's future, 00:00:48.890 --> 00:00:50.570 you know. Then we'll switch gears, look at some 00:00:50.570 --> 00:00:52.710 exciting new tools, creative advancements. There 00:00:52.710 --> 00:00:54.250 was one moment that genuinely made me stop and 00:00:54.250 --> 00:00:58.450 just go, wow. And finally, we'll dive deep into 00:00:58.450 --> 00:01:00.990 an AI agent showdown, an open source challenger 00:01:00.990 --> 00:01:02.469 that could genuinely change everything. It's 00:01:02.469 --> 00:01:04.079 going to be quite the journey. Okay, let's start 00:01:04.079 --> 00:01:07.299 unpacking this first report. So U .S. tech giants 00:01:07.299 --> 00:01:11.019 are pouring over $40 billion into the U .K.'s 00:01:11.019 --> 00:01:14.260 AI infrastructure. This isn't just a simple business 00:01:14.260 --> 00:01:16.819 transaction. It's deeply strategic, a partnership 00:01:16.819 --> 00:01:19.760 being spearheaded, it seems, as a joint AI and 00:01:19.760 --> 00:01:22.579 quantum initiative. Between President Trump and 00:01:22.579 --> 00:01:25.219 the U .K. Prime Minister Keir Starmer, it feels 00:01:25.219 --> 00:01:27.579 like locking in critical alliances, almost like 00:01:27.579 --> 00:01:30.140 a digital arms race, but focused on compute power 00:01:30.140 --> 00:01:32.459 and innovation. Right. And what's really fascinating 00:01:32.459 --> 00:01:34.500 here, I think the centerpiece is something called 00:01:34.500 --> 00:01:37.939 Stargate UK. It's a sovereign AI compute project. 00:01:38.200 --> 00:01:40.640 Now, sovereign AI, that might sound like a buzzword, 00:01:40.700 --> 00:01:43.219 but basically it means a nation building its 00:01:43.219 --> 00:01:45.780 own dedicated digital infrastructure, its own 00:01:45.780 --> 00:01:48.519 kind of brain trust. This guarantees local infrastructure, 00:01:48.840 --> 00:01:51.299 local rules, and really crucial local control 00:01:51.299 --> 00:01:54.060 over its AI capabilities. It's like digital self 00:01:54.060 --> 00:01:56.120 -determination. And the big names are all in. 00:01:56.200 --> 00:01:59.129 You've got OpenAI, NVIDIA, Netscale. They're 00:01:59.129 --> 00:02:02.030 backing it already. Makes it serious. And the 00:02:02.030 --> 00:02:03.870 specifics of these investments, they're pretty 00:02:03.870 --> 00:02:06.489 eye -opening. Microsoft, for instance. They're 00:02:06.489 --> 00:02:09.189 committing to build the UK's largest supercomputer. 00:02:09.610 --> 00:02:14.460 And get this, powered by 24 ,000... Grace Blackwell 00:02:14.460 --> 00:02:17.240 Ultra Chips. Wow. Yeah. For context, these are 00:02:17.240 --> 00:02:19.560 NVIDIA's absolute latest top tier processors 00:02:19.560 --> 00:02:22.039 engineered for the huge demands of training and 00:02:22.039 --> 00:02:24.080 running the most complex AI models. They're like 00:02:24.080 --> 00:02:26.740 the super brains for these future supercomputers. 00:02:26.740 --> 00:02:29.219 Exactly. And beyond that, NVIDIA itself is deploying 00:02:29.219 --> 00:02:31.819 tens of thousands of its GPUs, the graphics processing 00:02:31.819 --> 00:02:35.259 units across the UK, specifically for AI and 00:02:35.259 --> 00:02:37.280 quantum research. Right. So when you add it all 00:02:37.280 --> 00:02:39.919 up, the announced investment from just Microsoft, 00:02:40.139 --> 00:02:42.500 CoreWeave, a few others, it's already over 32. 00:02:42.759 --> 00:02:46.099 billion. Over 32. It's almost hard to wrap your 00:02:46.099 --> 00:02:48.419 head around that much capital in tech. And it's 00:02:48.419 --> 00:02:50.080 more than just investment. Like you said, it's 00:02:50.080 --> 00:02:53.120 a strategic chess move. The UK basically becomes 00:02:53.120 --> 00:02:56.379 OpenAI's main AI for countries client. It's not 00:02:56.379 --> 00:02:58.539 just buying compute. It's a deep partnership. 00:02:58.759 --> 00:03:02.159 Gives the UK a real edge. Beat. And at the same 00:03:02.159 --> 00:03:05.759 time, NVIDIA just cements its role as the... 00:03:06.080 --> 00:03:08.580 Well, the backbone of global AI. Yeah. With these 00:03:08.580 --> 00:03:11.080 60 ,000 chips deployed there, it's really about 00:03:11.080 --> 00:03:13.340 solidifying US AI leadership through these key 00:03:13.340 --> 00:03:16.120 alliances. The UK gets infrastructure, jobs, 00:03:16.439 --> 00:03:19.080 top AI tools. The US ensures its companies stay 00:03:19.080 --> 00:03:21.039 central to global compute. It's a win -win for 00:03:21.039 --> 00:03:23.120 them. But it also means, you know, others like 00:03:23.120 --> 00:03:26.580 Meta, TikTok, even China. They're kind of on 00:03:26.580 --> 00:03:28.699 the sidelines of this particular alliance for 00:03:28.699 --> 00:03:30.139 now. Makes you wonder how they'll respond, right? 00:03:30.409 --> 00:03:32.189 So when we talk about this whole sovereign AI 00:03:32.189 --> 00:03:34.930 movement then, what's really the core idea driving 00:03:34.930 --> 00:03:37.169 it? Fundamentally, it's about national control 00:03:37.169 --> 00:03:39.250 over the critical AI infrastructure and the rules 00:03:39.250 --> 00:03:41.389 governing it. Self -reliance, basically. Okay, 00:03:41.469 --> 00:03:43.889 shifting gears a bit. Here's where things get 00:03:43.889 --> 00:03:45.610 really interesting on a mere immediate level. 00:03:45.770 --> 00:03:47.830 Day -to -day stuff. There are big events happening 00:03:47.830 --> 00:03:50.409 right now shaping AI's immediate future. Like, 00:03:50.430 --> 00:03:53.050 MetaConnect 2025 kicks off this week. We're all 00:03:53.050 --> 00:03:54.909 kind of waiting to see their latest innovations. 00:03:56.039 --> 00:03:58.800 Probably in mixed reality, their AI assistance 00:03:58.800 --> 00:04:01.719 always generates a lot of buzz. Oh, yeah, definitely. 00:04:01.939 --> 00:04:03.919 And beyond the big company announcements, we're 00:04:03.919 --> 00:04:06.479 just seeing this explosion of creative tools 00:04:06.479 --> 00:04:09.439 for individuals. Take Reve, for instance. Launch 00:04:09.439 --> 00:04:12.039 four image tools in one go. Lets you mess with 00:04:12.039 --> 00:04:14.400 objects in an image, mix different images together, 00:04:14.580 --> 00:04:17.199 or even use an AI assistant with web search to 00:04:17.199 --> 00:04:18.980 help you out. It's like having a whole design 00:04:18.980 --> 00:04:21.620 studio right there. Makes complex stuff surprisingly 00:04:21.620 --> 00:04:25.240 easy. And YouTube jumped in too, right? New AI 00:04:25.240 --> 00:04:27.439 features for creators. We're talking auto ideas 00:04:27.439 --> 00:04:30.420 for content, smart dubbing for wider reach, even 00:04:30.420 --> 00:04:32.600 a VO3 shorts generator to whip up short videos 00:04:32.600 --> 00:04:35.639 quickly. It makes video creation way, way easier, 00:04:35.800 --> 00:04:38.339 more accessible. And it's not just the new tools. 00:04:38.500 --> 00:04:40.980 There's reflection happening too. OpenAI's board 00:04:40.980 --> 00:04:43.759 chairman, he recently gave this fascinating perspective, 00:04:44.120 --> 00:04:46.560 likened our current AI moment to 1999 for the 00:04:46.560 --> 00:04:49.459 internet. Suggesting it's a good thing, actually, 00:04:49.600 --> 00:04:52.160 implying we're on the edge of a massive boom 00:04:52.160 --> 00:04:55.360 like the dot com era, but maybe with similar 00:04:55.360 --> 00:04:58.120 early challenges, too. Yeah. And on the practical 00:04:58.120 --> 00:05:01.100 ethical side, OpenAI is taking steps. They launched 00:05:01.100 --> 00:05:04.139 new age prediction tech, strict content filters, 00:05:04.439 --> 00:05:07.439 parental control. Right. The idea is if it thinks 00:05:07.439 --> 00:05:09.680 you're under 18, you get sent to a teen friendly 00:05:09.680 --> 00:05:12.800 version. It's a clear move towards responsible 00:05:12.800 --> 00:05:15.639 deployment, trying to manage risks for younger 00:05:15.639 --> 00:05:17.500 users. That's super important. And those ethical 00:05:17.500 --> 00:05:19.779 thoughts are growing right alongside the tech's 00:05:19.779 --> 00:05:22.259 power. We're also seeing other big moves like 00:05:22.259 --> 00:05:25.079 GPT -5 codex integration. Apparently it's so 00:05:25.079 --> 00:05:27.180 good it's making users of cloud code rethink 00:05:27.180 --> 00:05:30.410 their whole workflow. That's a shift. Beep. and 00:05:30.410 --> 00:05:33.209 funding wise figure ai the humanoid robot company 00:05:33.209 --> 00:05:35.889 just pulled in over a billion dollars a billion 00:05:35.889 --> 00:05:37.750 yeah pushes their valuation to an incredible 00:05:37.750 --> 00:05:41.129 39 billion dollars backed by nvidia intel capital 00:05:41.129 --> 00:05:44.829 the heavy hitters whoa just imagine scaling to 00:05:44.829 --> 00:05:47.129 handle a billion queries a day that's that's 00:05:47.129 --> 00:05:49.990 wild shows massive belief in physical ai and 00:05:49.990 --> 00:05:52.850 speaking of physical ai open ai itself is hiring 00:05:53.339 --> 00:05:55.560 Robotics researchers like crazy. Yeah. Signals 00:05:55.560 --> 00:05:58.439 a clear push into AI that interacts with the 00:05:58.439 --> 00:06:00.980 real world, not just screens. Yeah. Makes you 00:06:00.980 --> 00:06:03.379 wonder what's next. You know, digital brains 00:06:03.379 --> 00:06:06.040 getting bodies. It's an incredible pace, just 00:06:06.040 --> 00:06:09.579 relentless. So if you had to boil it down, what's 00:06:09.579 --> 00:06:12.220 the single biggest takeaway from all these rapid 00:06:12.220 --> 00:06:15.160 fire developments? I'd say AI is maturing incredibly 00:06:15.160 --> 00:06:18.910 fast. It's changing creativity, raising new ethical 00:06:18.910 --> 00:06:21.769 questions and hitting valuations that show its 00:06:21.769 --> 00:06:24.670 huge potential. OK, so moving from the big picture 00:06:24.670 --> 00:06:28.329 geopolitics, industry shifts. Let's bring it 00:06:28.329 --> 00:06:30.350 down to the practical side. There are tons of 00:06:30.350 --> 00:06:32.769 new resources popping up to help you, our listeners, 00:06:32.970 --> 00:06:36.129 actually use AI effectively. Day to day work, 00:06:36.250 --> 00:06:38.579 personal projects like. There's this in -depth 00:06:38.579 --> 00:06:41.439 guide to Google Gemini's whole AI toolkit. It 00:06:41.439 --> 00:06:43.259 covers all the free and paid stuff for writing, 00:06:43.339 --> 00:06:45.480 research, everything. You know, personally, I 00:06:45.480 --> 00:06:47.360 still wrestle with prompt drift myself sometimes. 00:06:47.540 --> 00:06:49.120 Oh, yeah. Yeah, when the AI kind of wanders off 00:06:49.120 --> 00:06:51.600 topic after a few exchanges. So these practical 00:06:51.600 --> 00:06:53.699 guides, they're genuinely valuable for staying 00:06:53.699 --> 00:06:56.819 focused. Totally get that. Yeah. And building 00:06:56.819 --> 00:06:59.060 on the practical stuff, for developers listening, 00:06:59.240 --> 00:07:01.620 there's a guide on mastering the Gemini command 00:07:01.620 --> 00:07:05.379 line. Lays out 11 pro methods to streamline coding, 00:07:05.600 --> 00:07:08.779 automate tasks, basically supercharge your workflow 00:07:08.779 --> 00:07:11.300 with AI. Then for content creators, there's this 00:07:11.300 --> 00:07:13.920 no code guide. Shows how to build an AI factory, 00:07:14.000 --> 00:07:16.240 they call it. Imagine generating unlimited product 00:07:16.240 --> 00:07:19.360 ads without coding. Promising like 99 % cost 00:07:19.360 --> 00:07:22.199 savings. Wow. Yeah. Huge impact for small businesses, 00:07:22.379 --> 00:07:24.300 individual creators. It really is empowering. 00:07:24.540 --> 00:07:26.519 And for entrepreneurs thinking about jumping 00:07:26.519 --> 00:07:28.459 in, there's a blueprint now for starting a one 00:07:28.459 --> 00:07:30.899 person AI business covers everything. Finding 00:07:30.899 --> 00:07:32.800 a niche, landing your first clients really highlights 00:07:32.800 --> 00:07:35.240 how accessible this tech is becoming. And the 00:07:35.240 --> 00:07:38.060 sheer number of new tools emerging daily is just 00:07:38.060 --> 00:07:40.819 expanding constantly. We're seeing things like 00:07:40.819 --> 00:07:43.019 GradMay gives you super detailed website reports 00:07:43.019 --> 00:07:46.040 instantly, helps optimize your site, Thumblify 00:07:46.040 --> 00:07:48.699 AI turns ideas into YouTube thumbnails designed 00:07:48.699 --> 00:07:51.500 to get clicks, Aura creates hyper -realistic 00:07:51.500 --> 00:07:54.500 AI images pushing visual boundaries, and Canary 00:07:54.500 --> 00:07:57.420 PDF. Great little browser tool, extracts tables. 00:07:57.560 --> 00:08:00.079 Let's edit PDFs right there. Saves so much time. 00:08:00.180 --> 00:08:02.100 Plus, you hear these quick win stories all the 00:08:02.100 --> 00:08:05.060 time now. Like one user cut a $6 ,000 bill down 00:08:05.060 --> 00:08:08.399 to $300 using ChatGPT for negotiation. That's 00:08:08.399 --> 00:08:11.199 amazing. Right. And Google's Learn Your Way turns 00:08:11.199 --> 00:08:15.160 texts into interactive guides. And ChatGPT might 00:08:15.160 --> 00:08:17.040 soon try guessing your age, maybe ask for ID. 00:08:17.360 --> 00:08:19.720 So I have more regulation coming. It just shows 00:08:19.720 --> 00:08:22.060 how AI is weaving into, well, everything digital. 00:08:22.480 --> 00:08:24.519 With this whirlwind of new tools and guides, 00:08:24.759 --> 00:08:26.620 what's the common thread you see running through 00:08:26.620 --> 00:08:29.100 them all? The clear thread. AI is becoming way 00:08:29.100 --> 00:08:31.240 more accessible, much more user -friendly, and 00:08:31.240 --> 00:08:34.200 directly useful for tons of everyday tasks. For 00:08:34.200 --> 00:08:37.440 everyone. Midroll Sponsoree. So, what does all 00:08:37.440 --> 00:08:39.980 this mean for the bigger picture? The competitive 00:08:39.980 --> 00:08:42.919 landscape of AI? We've looked at national strategies, 00:08:43.220 --> 00:08:46.259 seen practical tools emerge, but now let's talk 00:08:46.259 --> 00:08:49.120 about a direct challenge. Something shaking up 00:08:49.120 --> 00:08:51.700 the status quo. A really significant development 00:08:51.700 --> 00:08:54.740 is Tangi Labs Deep Research. It just launched. 00:08:54.840 --> 00:08:57.679 It's the first fully open source web research 00:08:57.679 --> 00:09:00.240 agent. Yeah, this is big news because this agent, 00:09:00.279 --> 00:09:02.299 it's not just out there theoretically. It's going 00:09:02.299 --> 00:09:04.840 head to head with OpenAI's own deep research 00:09:04.840 --> 00:09:07.940 agent. And in some important tests, it's actually 00:09:07.940 --> 00:09:10.659 beating it. Really? Yeah. And what's fascinating 00:09:10.659 --> 00:09:13.980 is the efficiency. Runs on just 30 billion parameters. 00:09:14.320 --> 00:09:16.799 That's like the model's knowledge base. But it 00:09:16.799 --> 00:09:18.860 only needs 3 billion active during inference 00:09:18.860 --> 00:09:21.960 when it's actually working. Uses way less compute 00:09:21.960 --> 00:09:24.200 power than the big guys. Think of it like a huge 00:09:24.200 --> 00:09:26.539 library, the 30 billion parameters. But the librarian 00:09:26.539 --> 00:09:28.659 is so skilled, they only need a few key books, 00:09:28.840 --> 00:09:30.639 the 3 billion active ones, to give you the answer. 00:09:30.840 --> 00:09:33.559 Super fast, super efficient. And this isn't just 00:09:33.559 --> 00:09:35.720 academic, right? It's already being used in real 00:09:35.720 --> 00:09:37.940 applications and winning on benchmarks. Exactly. 00:09:38.500 --> 00:09:41.299 In its standard React mode. Tangi does better 00:09:41.299 --> 00:09:44.840 than OpenAI's O3 model on key tests. Like Humanity's 00:09:44.840 --> 00:09:47.600 last exam. That's a tough test of reasoning across 00:09:47.600 --> 00:09:52.519 subjects. Tangi scored 32 .9, OpenAI 24 .9. That's 00:09:52.519 --> 00:09:55.059 quite a gap. It is. And on Xbench DeepSearch, 00:09:55.100 --> 00:09:58.659 testing web research, Tangi hit 75 .0, OpenAI 00:09:58.659 --> 00:10:01.500 67 .0. It even outperforms in a more intensive 00:10:01.500 --> 00:10:04.299 heavy mode. These aren't small wins for an open 00:10:04.299 --> 00:10:06.759 source model. And it's not just scores it's deployed. 00:10:06.960 --> 00:10:10.059 It's live in Go Demonte that's an AI travel planner 00:10:10.059 --> 00:10:12.320 in China's Google Maps equivalent. That's also 00:10:12.320 --> 00:10:14.600 in Tangi Farui, a legal research tool citing 00:10:14.600 --> 00:10:17.240 real case law. Super useful. And the best part, 00:10:17.379 --> 00:10:19.820 developers, researchers, they can just run it 00:10:19.820 --> 00:10:23.120 themselves. Free. No API key needed. Beat. So 00:10:23.120 --> 00:10:25.559 yeah, OpenAI's O3 just got its first serious 00:10:25.559 --> 00:10:27.940 competitive open source rival. That fundamentally 00:10:27.940 --> 00:10:30.019 changes the game. accessibility, innovation, 00:10:30.340 --> 00:10:32.320 everything. So what's the biggest implication 00:10:32.320 --> 00:10:34.539 then, an open source model challenging, even 00:10:34.539 --> 00:10:36.960 beating, a proprietary giant like this? It means 00:10:36.960 --> 00:10:39.620 powerful AI could get a whole lot more democratized, 00:10:39.700 --> 00:10:41.899 more widely accessible, could speed up innovation, 00:10:42.100 --> 00:10:44.200 lower barriers for everyone. We've really covered 00:10:44.200 --> 00:10:47.379 a lot, seen the global strategic push for AI 00:10:47.379 --> 00:10:51.740 supremacy, that huge US -UK deal, sovereign AI 00:10:51.740 --> 00:10:53.980 initiatives. These are big geopolitical moves. 00:10:54.259 --> 00:10:57.659 We've also seen the just relentless pace of innovation, 00:10:57.980 --> 00:11:01.120 new creative tools, but also those critical ethical 00:11:01.120 --> 00:11:03.379 questions popping up. And then maybe most importantly, 00:11:03.460 --> 00:11:05.659 we explore these powerful open source challengers 00:11:05.659 --> 00:11:08.759 like Tangi's deep research. It's not just competition. 00:11:08.879 --> 00:11:11.120 It reshapes the whole landscape. It shows loud 00:11:11.120 --> 00:11:14.120 and clear that cutting edge AI powerful, efficient 00:11:14.120 --> 00:11:17.240 AI. It can come from anywhere, not just the biggest 00:11:17.240 --> 00:11:19.360 tech companies. It really levels the playing 00:11:19.360 --> 00:11:21.720 field. Yeah, this deep dive definitely shows 00:11:21.720 --> 00:11:24.539 a world in flux, adapting so fast to new capabilities, 00:11:24.659 --> 00:11:27.120 new players. What stands out most to you about 00:11:27.120 --> 00:11:29.059 this balance we're seeing between, you know, 00:11:29.059 --> 00:11:31.299 centralized power, the big national strategies, 00:11:31.399 --> 00:11:34.019 huge investments, and this decentralized innovation 00:11:34.019 --> 00:11:36.840 from open source and individuals? Well, consider 00:11:36.840 --> 00:11:40.940 this. As AI gets more powerful, sure. but also 00:11:40.940 --> 00:11:43.580 more accessible, more woven into our lives, what 00:11:43.580 --> 00:11:45.779 new questions do we need to ask? Maybe uncomfortable 00:11:45.779 --> 00:11:49.279 ones. About control, about ethics long -term, 00:11:49.320 --> 00:11:51.240 about how we collaborate across borders between 00:11:51.240 --> 00:11:53.379 companies to guide this thing responsibly. It's 00:11:53.379 --> 00:11:56.039 a field moving faster than ever. And those questions, 00:11:56.059 --> 00:11:58.580 they're getting more urgent every week. Thank 00:11:58.580 --> 00:12:00.340 you for joining us on this deep dive. Until next 00:12:00.340 --> 00:12:02.279 time, keep exploring. Outro music.