WEBVTT 00:00:00.000 --> 00:00:02.560 Imagine hearing, say, a minute of spoken words, 00:00:02.740 --> 00:00:05.960 and then in less than a second, an AI generates 00:00:05.960 --> 00:00:09.820 that exact audio back perfectly with all the, 00:00:09.820 --> 00:00:12.179 you know, nuance and expression of a real person. 00:00:12.640 --> 00:00:16.079 Or think about translating super sensitive company 00:00:16.079 --> 00:00:19.379 documents, but without those files ever leaving 00:00:19.379 --> 00:00:21.839 your own secure system, not even for a moment. 00:00:21.940 --> 00:00:23.719 Yeah, that's a big one. These aren't just like 00:00:23.719 --> 00:00:26.059 far off ideas anymore. They're actually happening 00:00:26.059 --> 00:00:29.109 right now, changing things. Welcome to the Deep 00:00:29.109 --> 00:00:31.750 Dive. We're here to cut through all that digital 00:00:31.750 --> 00:00:33.869 noise and focus on the really impactful shifts 00:00:33.869 --> 00:00:37.149 happening in AI. Right. So today, we're going 00:00:37.149 --> 00:00:40.689 to explore Microsoft's big push for AI independence. 00:00:40.990 --> 00:00:43.530 It's quite a move. Then we'll look at some, frankly, 00:00:43.609 --> 00:00:46.149 incredibly clever new tools popping up. And finally, 00:00:46.250 --> 00:00:48.350 we'll dive into something potentially game -changing 00:00:48.350 --> 00:00:51.609 for businesses, secure enterprise -level translation. 00:00:52.359 --> 00:00:54.179 Think of this as, you know, your shortcut to 00:00:54.179 --> 00:00:56.079 getting up to speed. Okay, let's unpack that 00:00:56.079 --> 00:00:58.679 first one. Microsoft. This feels like a really 00:00:58.679 --> 00:01:01.899 significant strategic pivot they're making. For 00:01:01.899 --> 00:01:04.340 a while, you know, their AI strategy felt so 00:01:04.340 --> 00:01:07.180 tightly linked to OpenAI. It did. Very much so. 00:01:07.280 --> 00:01:09.180 But now they seem to be developing some serious 00:01:09.180 --> 00:01:11.040 foundational models in -house. It feels like 00:01:11.040 --> 00:01:13.819 a bold step. It absolutely is. We're seeing these 00:01:13.819 --> 00:01:17.159 new models emerge, like MAI -1 Preview for Text 00:01:17.159 --> 00:01:20.739 and MAI -Voice -1 for Speech. Right. And while 00:01:20.739 --> 00:01:23.700 they're still... relatively new, the message 00:01:23.700 --> 00:01:26.140 is loud and clear. Microsoft is building its 00:01:26.140 --> 00:01:29.700 own core AI from the ground up. And doing it 00:01:29.700 --> 00:01:33.040 fast. At incredible pace and scale, yeah. This 00:01:33.040 --> 00:01:34.799 isn't just about having more options. It's really 00:01:34.799 --> 00:01:37.980 about ownership, control. And that MAI Voice 00:01:37.980 --> 00:01:41.900 1, the speech model, it sounds almost unreal. 00:01:42.140 --> 00:01:44.239 It's pretty amazing. Generating 60 seconds of 00:01:44.239 --> 00:01:47.980 really natural multi -speaker audio in under 00:01:47.980 --> 00:01:50.219 one second. Yep. Less than a second. It's already 00:01:50.219 --> 00:01:52.299 running features in Copilot Daily, some podcasts, 00:01:52.519 --> 00:01:54.900 Copilot Labs. So the potential there for creators, 00:01:55.180 --> 00:01:58.299 interactive stories, maybe personalized meditation 00:01:58.299 --> 00:02:02.319 scripts, dynamic audio companions, it just opens 00:02:02.319 --> 00:02:05.659 up so much. Totally. It really expands the canvas 00:02:05.659 --> 00:02:07.939 for audio in ways we're probably just starting 00:02:07.939 --> 00:02:10.800 to figure out. And then there's MAI -1 Preview. 00:02:11.159 --> 00:02:13.919 That's Microsoft's first big in -house text foundation 00:02:13.919 --> 00:02:17.240 model. Okay, so. Foundation model. Just quickly, 00:02:17.379 --> 00:02:19.879 what does that mean? Right. Good question. Think 00:02:19.879 --> 00:02:23.060 of it like a really big general purpose AI. It's 00:02:23.060 --> 00:02:25.159 trained on absolutely massive amounts of data. 00:02:25.300 --> 00:02:27.919 Okay. Which makes it super adaptable for lots 00:02:27.919 --> 00:02:31.039 of different tasks. This specific one, MAI -1 00:02:31.039 --> 00:02:34.539 preview, they traded on something like 15 ,000 00:02:34.539 --> 00:02:39.800 NVIDIA H100 GPUs. 15 ,000? Yeah. I mean, that's 00:02:39.800 --> 00:02:42.639 a huge compute investment, like world -class 00:02:42.639 --> 00:02:45.000 supercomputer level. Shows how serious they are. 00:02:45.159 --> 00:02:47.159 Right. Its main job right now seems to be handling 00:02:47.159 --> 00:02:48.960 everyday questions and following instructions. 00:02:49.599 --> 00:02:52.319 So very practical stuff. And people can actually 00:02:52.319 --> 00:02:55.340 try it out. They can. It's live on LM Arena for 00:02:55.340 --> 00:02:57.759 public testing, and some trusted testers have 00:02:57.759 --> 00:03:00.800 API access already. So why does all this matter? 00:03:00.919 --> 00:03:03.539 Well, is Microsoft really playing to own more 00:03:03.539 --> 00:03:06.810 of the whole AI stack? Control tech, control 00:03:06.810 --> 00:03:08.629 the future. And control the margins, let's be 00:03:08.629 --> 00:03:10.189 honest. Plus, it's a smart hedging strategy. 00:03:10.430 --> 00:03:13.949 Hedging? How so? Well, MAI -1 isn't necessarily 00:03:13.949 --> 00:03:17.030 replacing OpenAI's models. It sits alongside 00:03:17.030 --> 00:03:20.229 GPT -4, open source options, other partner stuff. 00:03:20.469 --> 00:03:22.530 Oh, okay. It's like Microsoft is setting up an 00:03:22.530 --> 00:03:25.250 AI model buffet. Businesses get more choice, 00:03:25.310 --> 00:03:27.830 more flexibility. Right. This whole move basically 00:03:27.830 --> 00:03:30.490 screams, we're not just OpenAI's wallet anymore. 00:03:30.789 --> 00:03:33.719 Beat. It's a big statement about their independence 00:03:33.719 --> 00:03:38.360 long term. Whoa. Just thinking about MAI Voice 00:03:38.360 --> 00:03:42.159 1. Imagine generating a whole audio book. Expressive, 00:03:42.159 --> 00:03:45.759 multiple characters in minutes. Or like doing 00:03:45.759 --> 00:03:47.919 professional voiceovers for entire Moody libraries 00:03:47.919 --> 00:03:51.319 localized overnight. The scale is just staggering. 00:03:51.599 --> 00:03:53.400 It really does fire up the imagination. And when 00:03:53.400 --> 00:03:55.340 a giant like Microsoft develops these capabilities 00:03:55.340 --> 00:03:57.599 in -house, while it doesn't just expand creative 00:03:57.599 --> 00:03:59.599 options, it really shakes up the whole competitive 00:03:59.599 --> 00:04:02.229 landscape. Definitely. So building on that thought, 00:04:02.330 --> 00:04:05.270 what does this big push for AI independence really 00:04:05.270 --> 00:04:07.830 mean for the broader competitive field? It points 00:04:07.830 --> 00:04:10.830 towards a more fragmented, more diverse AI world, 00:04:10.889 --> 00:04:13.750 more competition, tech giants controlling their 00:04:13.750 --> 00:04:17.170 own AI destiny. Makes sense. okay so while microsoft's 00:04:17.170 --> 00:04:19.209 playing that long game with foundational models 00:04:19.209 --> 00:04:21.910 there's also this whole buzzing ecosystem of 00:04:21.910 --> 00:04:25.730 like super specialized ai tools emerging constantly 00:04:25.730 --> 00:04:28.410 yeah happening fast let's pivot to some of those 00:04:28.410 --> 00:04:31.310 some really clever maybe immediately useful innovations 00:04:31.310 --> 00:04:34.149 we've seen pop up recently stuff tackling everything 00:04:34.149 --> 00:04:37.430 from studying to making videos in real time let's 00:04:37.430 --> 00:04:41.550 do it kicking off kpmg you know the huge consulting 00:04:41.550 --> 00:04:44.490 firm they developed what sounds like a monster 00:04:45.480 --> 00:04:48.600 A hundred page prompt. A hundred pages for a 00:04:48.600 --> 00:04:50.800 prompt. Yeah. It's not just a simple question. 00:04:50.899 --> 00:04:53.800 It's this super elaborate set of instructions 00:04:53.800 --> 00:04:58.079 for AI. And get this. It cuts down the time to 00:04:58.079 --> 00:05:01.180 create a major report from two weeks to 24 hours. 00:05:01.259 --> 00:05:04.300 Due weeks to one day. Exactly. Built using OpenAI 00:05:04.300 --> 00:05:06.560 and Microsoft tech underneath. Think about the 00:05:06.560 --> 00:05:08.519 efficiency gain for big firms like that. It's 00:05:08.519 --> 00:05:10.540 massive. That really is monumental. And speaking 00:05:10.540 --> 00:05:13.060 of helpful tools, ChatGPT rolled out something 00:05:13.060 --> 00:05:15.339 called QuizGPT. Oh, yeah, I saw that. Let's you 00:05:15.339 --> 00:05:17.959 make free flashcard quizzes instantly. Just feed 00:05:17.959 --> 00:05:20.180 it your notes, a PDF, whatever. Super handy. 00:05:20.319 --> 00:05:23.199 And it spits out a quiz ready to go on desktop 00:05:23.199 --> 00:05:26.160 or mobile. Seems like such a clever study tool, 00:05:26.279 --> 00:05:28.660 right? Could really change how people prep for 00:05:28.660 --> 00:05:31.939 exams or certifications. Totally. And for the 00:05:31.939 --> 00:05:34.180 visual folks, Kriya just dropped a real -time 00:05:34.180 --> 00:05:36.399 video creator. Yeah. Honestly, it looks like 00:05:36.399 --> 00:05:39.019 magic. How does it work? You literally watch 00:05:39.019 --> 00:05:41.800 the animation happen as you sketch or type in 00:05:41.800 --> 00:05:44.540 changes. It just responds instantly. Wow. Fully 00:05:44.540 --> 00:05:46.579 controllable, keeps frames consistent, which 00:05:46.579 --> 00:05:49.379 is huge for animation work. Imagine seeing your 00:05:49.379 --> 00:05:51.860 idea come to life, like right there as you draw 00:05:51.860 --> 00:05:54.459 it. Game changer for artists, animators. For 00:05:54.459 --> 00:05:56.779 sure. Then there's Higgs Field. They launched 00:05:56.779 --> 00:06:00.300 over 2 ,000 little apps. 2000? Yeah, powered 00:06:00.300 --> 00:06:02.680 by this new Google tech called Nano Banana. Nano 00:06:02.680 --> 00:06:05.579 Banana. Seriously? Yeah. Sounds funny, right? 00:06:05.959 --> 00:06:07.819 But these apps let you whip up quick commercials, 00:06:08.019 --> 00:06:10.160 visual effects, stuff like that, often in just 00:06:10.160 --> 00:06:12.379 one click. Okay, I have to ask, Nano Banana, 00:06:12.680 --> 00:06:14.579 what's the story there? It does sound whimsical, 00:06:14.740 --> 00:06:17.660 but Nano Banana is actually Google's internal 00:06:17.660 --> 00:06:21.980 name for a new, really efficient AI architecture 00:06:21.980 --> 00:06:24.319 designed to run right on your phone or device. 00:06:24.639 --> 00:06:27.879 Ah, on device. Exactly. Nano, because it's super 00:06:27.879 --> 00:06:30.800 efficient, doesn't need the cloud. And Banana. 00:06:31.660 --> 00:06:34.139 Well, that's apparently a playful nod to how 00:06:34.139 --> 00:06:36.959 it's modular. You can peel it, adapt it for specific, 00:06:37.060 --> 00:06:39.500 lightweight tasks like these quick content apps. 00:06:39.740 --> 00:06:42.839 Makes powerful AI accessible without needing 00:06:42.839 --> 00:06:45.600 a massive server. Got it. Makes sense, despite 00:06:45.600 --> 00:06:47.319 the name. And this next one's really important 00:06:47.319 --> 00:06:50.120 for anyone building AI. New research pinpointed 00:06:50.120 --> 00:06:53.879 32 different ways AI can, well... Go rogue. 32? 00:06:54.319 --> 00:06:58.300 Like, how? It's not just making stuff up or hallucinations. 00:06:58.459 --> 00:07:01.560 It ranges from subtle misunderstandings all the 00:07:01.560 --> 00:07:03.600 way to just completely not doing what it was 00:07:03.600 --> 00:07:06.240 supposed to do. Misalignment. It's basically 00:07:06.240 --> 00:07:08.439 a vital checklist for developers. Helps them 00:07:08.439 --> 00:07:11.399 anticipate problems, build in safeguards. Crucial 00:07:11.399 --> 00:07:13.279 for safety and reliability. Yeah, understanding 00:07:13.279 --> 00:07:15.779 failure modes is key. And on a more everyday 00:07:15.779 --> 00:07:18.399 level, someone did a cool comparison of AI image 00:07:18.399 --> 00:07:21.379 removal tools on phones. Like on iPhones versus 00:07:21.379 --> 00:07:24.560 Android? Exactly. Apple, Galaxy, Xiaomi phones. 00:07:24.759 --> 00:07:26.779 And the results were apparently quite surprising. 00:07:26.920 --> 00:07:28.939 Big differences in how well they worked. Huh. 00:07:29.120 --> 00:07:31.939 Interesting. Shows that the AI baked into hardware 00:07:31.939 --> 00:07:35.699 isn't all the same quality. And finally, this 00:07:35.699 --> 00:07:39.199 one has huge potential impact. Alisa AI just 00:07:39.199 --> 00:07:43.620 raised $250 million from A16Zs. Wow. Big funding 00:07:43.620 --> 00:07:45.899 round. What do they do? They're using AI to tackle 00:07:45.899 --> 00:07:48.279 massive problems in healthcare and housing. OK, 00:07:48.300 --> 00:07:50.620 those are huge sectors. Huge. Together, they 00:07:50.620 --> 00:07:52.459 make up something like 40 percent of household 00:07:52.459 --> 00:07:55.639 costs in the U .S. At least AI is already powering 00:07:55.639 --> 00:07:58.519 10 percent of U .S. apartments, apparently, using 00:07:58.519 --> 00:08:00.839 AI to streamline things there. Tackling really 00:08:00.839 --> 00:08:04.339 complex societal level problems with AI. That's 00:08:04.339 --> 00:08:06.600 ambitious. Very. You know, hearing all this, 00:08:06.639 --> 00:08:08.720 I admit I still wrestle with prompt drift sometimes 00:08:08.720 --> 00:08:11.399 myself. Oh, yeah. Trying to get the AI to do 00:08:11.399 --> 00:08:13.639 exactly what's in my head really feels like an 00:08:13.639 --> 00:08:15.480 art form sometimes. Getting that nuance across 00:08:15.480 --> 00:08:18.199 perfectly takes practice. It really does. It's 00:08:18.199 --> 00:08:20.379 a skill. But if you zoom out from all these cool 00:08:20.379 --> 00:08:23.420 tools and features, what's the common thread 00:08:23.420 --> 00:08:25.300 here? What does this collection of innovations 00:08:25.300 --> 00:08:28.600 really mean for us as users? I think it's that 00:08:28.600 --> 00:08:32.019 AI tools are getting hyper -specialized, much 00:08:32.019 --> 00:08:35.019 more intuitive, and just woven much deeper into 00:08:35.019 --> 00:08:39.730 our everyday tasks and workflows. Okay, now shifting 00:08:39.730 --> 00:08:42.309 gears slightly, this next one gets really interesting 00:08:42.309 --> 00:08:45.029 for businesses, especially larger ones. Cohere. 00:08:45.110 --> 00:08:47.350 Right, big player in large language models. They 00:08:47.350 --> 00:08:50.429 just introduced something called Command AI Translate. 00:08:50.470 --> 00:08:52.669 And this isn't just, you know, another translation 00:08:52.669 --> 00:08:56.289 app. It's built specifically for serious, high 00:08:56.289 --> 00:08:58.769 -stakes enterprise translation. Yeah, definitely 00:08:58.769 --> 00:09:01.850 not just another tool. Cohere is aiming this 00:09:01.850 --> 00:09:04.450 squarely at that demanding enterprise market. 00:09:04.840 --> 00:09:06.899 And it's not just good they're saying it's outperforming 00:09:06.899 --> 00:09:10.840 giants like GPT -5, DeepSeq v3, even Google Translate 00:09:10.840 --> 00:09:13.100 on key multilingual tests. Okay, that's impressive 00:09:13.100 --> 00:09:15.379 performance. It is. But the raw performance, 00:09:15.539 --> 00:09:17.360 while obviously critical, isn't even the whole 00:09:17.360 --> 00:09:19.600 story. What's really revolutionary here, I think, 00:09:19.600 --> 00:09:22.059 is the intense focus on privacy, customization, 00:09:22.279 --> 00:09:24.620 and control for the business. Ah, the things 00:09:24.620 --> 00:09:27.059 that hold companies back sometimes. Exactly. 00:09:27.100 --> 00:09:29.860 Those have been huge roadblocks for AI adoption 00:09:29.860 --> 00:09:32.200 and industries handling sensitive data. That 00:09:32.200 --> 00:09:34.909 privacy and control angle. That's huge for companies, 00:09:35.009 --> 00:09:37.169 right? Especially regulated industries. Monumental. 00:09:37.429 --> 00:09:40.809 Let's dig into the features. So top scores across 00:09:40.809 --> 00:09:45.190 23 major business languages. That covers a lot 00:09:45.190 --> 00:09:48.730 of ground globally. Then there's this deep translation 00:09:48.730 --> 00:09:51.590 workflow. What is that exactly? Okay, so this 00:09:51.590 --> 00:09:53.669 is an optional add -on. It's what they call an 00:09:53.669 --> 00:09:56.320 agentic pipeline. Agentic pipeline. Yeah. Think 00:09:56.320 --> 00:09:58.919 of it like an AI system that uses multiple steps, 00:09:59.039 --> 00:10:01.840 multiple checks, almost like a little team of 00:10:01.840 --> 00:10:04.639 specialized AI agents working together. OK. It 00:10:04.639 --> 00:10:06.799 double checks complex stuff to really ensure 00:10:06.799 --> 00:10:09.840 accuracy and nuance. Super useful, maybe even 00:10:09.840 --> 00:10:12.340 essential for things like legal contracts, complex 00:10:12.340 --> 00:10:16.840 financial reports, compliance documents. Anywhere. 00:10:17.210 --> 00:10:19.090 An error is just not an option. Right, where 00:10:19.090 --> 00:10:21.809 precision is absolutely critical. Totally. That 00:10:21.809 --> 00:10:24.009 multi -stage verification is key when mistakes 00:10:24.009 --> 00:10:26.789 cost big money or have legal consequences. But 00:10:26.789 --> 00:10:28.950 there's more. It offers enterprise -grade customization. 00:10:29.509 --> 00:10:31.529 Meaning companies can teach it their own specific 00:10:31.529 --> 00:10:34.809 jargon. Exactly. You can actually train the model 00:10:34.809 --> 00:10:37.110 on your company's specific terminology, your 00:10:37.110 --> 00:10:39.830 style guide. So a pharma company can feed it 00:10:39.830 --> 00:10:41.929 all their drug names and research terms. Or a 00:10:41.929 --> 00:10:45.100 bank their financial lingo. Precisely. Or a bank 00:10:45.100 --> 00:10:47.279 teaches it their unique financial language, their 00:10:47.279 --> 00:10:49.879 branding tone. This makes sure the translations 00:10:49.879 --> 00:10:52.620 aren't just grammatically okay, but actually 00:10:52.620 --> 00:10:55.100 perfect for their specific context and internal 00:10:55.100 --> 00:10:57.960 standards. That level of customization is powerful. 00:10:58.360 --> 00:11:01.039 It is. And maybe the most game -changing part, 00:11:01.159 --> 00:11:03.899 the real kicker, is the private deployment option. 00:11:04.519 --> 00:11:06.940 With an air -gapped capability. Air -gapped, 00:11:07.059 --> 00:11:09.139 meaning completely offline. Totally offline. 00:11:09.240 --> 00:11:11.940 You can install Cohere's entire translation engine 00:11:11.940 --> 00:11:14.659 right inside your own secure data center, completely 00:11:14.659 --> 00:11:17.240 isolated from the internet or any external network. 00:11:17.519 --> 00:11:20.480 Wow. This means you can translate ultra -confidential 00:11:20.480 --> 00:11:22.960 contracts, sensitive patient medical records, 00:11:23.200 --> 00:11:26.860 proprietary R &D, whatever, without that data 00:11:26.860 --> 00:11:30.080 ever leaving your control. Not even for a millisecond 00:11:30.080 --> 00:11:32.480 hitting an external cloud. That is a massive 00:11:32.480 --> 00:11:35.279 deal for data security and compliance. Huge. 00:11:35.580 --> 00:11:38.100 This directly solves that critical practical 00:11:38.100 --> 00:11:42.230 business need. Getting real time. accurate customized 00:11:42.230 --> 00:11:46.250 and most importantly secure ai translation it 00:11:46.250 --> 00:11:49.509 basically takes this old often clunky sometimes 00:11:49.509 --> 00:11:53.070 risky workflow and turns it into a smart secure 00:11:53.070 --> 00:11:56.769 ai native process no more uploading your crown 00:11:56.769 --> 00:11:59.289 jewels to some third -party cloud and just hoping 00:11:59.289 --> 00:12:02.350 for the best right it genuinely removes a major 00:12:02.350 --> 00:12:05.909 barrier for ai use in highly regulated or security 00:12:05.909 --> 00:12:09.070 focused fields that's a fundamental shift really 00:12:09.070 --> 00:12:11.879 so how does offering this This on -premise, private, 00:12:12.019 --> 00:12:14.919 truly air -gapped option fundamentally changed 00:12:14.919 --> 00:12:16.759 how businesses can even think about using AI, 00:12:16.960 --> 00:12:18.720 especially with their most sensitive information. 00:12:19.080 --> 00:12:21.159 It basically eliminates those core data security 00:12:21.159 --> 00:12:23.960 and compliance fears. It unlocks AI for potentially 00:12:23.960 --> 00:12:25.840 all kinds of highly confidential information 00:12:25.840 --> 00:12:27.720 that was previously off -limits. Okay, let's 00:12:27.720 --> 00:12:29.799 hit a few more quick updates because, as usual, 00:12:29.899 --> 00:12:31.759 the AI world just doesn't stop moving. Never 00:12:31.759 --> 00:12:33.919 does. You got it. First up, Microsoft Copilot. 00:12:33.919 --> 00:12:36.159 It now includes ChatGPT's multi -file analysis 00:12:36.159 --> 00:12:38.820 feature for free. Oh, that's useful. Analyzing 00:12:38.820 --> 00:12:41.539 across multiple documents. Yep. Big value add 00:12:41.539 --> 00:12:43.980 for Copilot users. Lets them synthesize info 00:12:43.980 --> 00:12:46.639 from several files at once without paying extra. 00:12:46.799 --> 00:12:49.139 Makes Copilot even more powerful for complex 00:12:49.139 --> 00:12:51.659 research or reporting tasks. Good to know. And 00:12:51.659 --> 00:12:55.659 Time Magazine put out its 2025 Time 100 AI list. 00:12:56.000 --> 00:12:57.940 Ah, always interesting to see who they highlight. 00:12:58.159 --> 00:12:59.919 Yeah, it gives you a sense of who's shaping the 00:12:59.919 --> 00:13:01.919 conversation, who the big movers are perceived 00:13:01.919 --> 00:13:04.899 to be in the AI space. Helps gauge influence. 00:13:05.279 --> 00:13:08.620 For sure. This next one sounds... pretty profound, 00:13:08.720 --> 00:13:11.019 maybe even life -changing. New research suggests 00:13:11.019 --> 00:13:13.559 AI can now spot hidden signs of consciousness 00:13:13.559 --> 00:13:16.799 in patients, sometimes even before doctors can. 00:13:16.960 --> 00:13:19.639 Hidden consciousness? How? By analyzing really 00:13:19.639 --> 00:13:21.820 subtle patterns in brain activity, neurological 00:13:21.820 --> 00:13:24.500 signals. It's not sci -fi. This could lead to 00:13:24.500 --> 00:13:27.000 major medical breakthroughs, giving hope for 00:13:27.000 --> 00:13:29.659 communication, maybe even recovery for patients 00:13:29.659 --> 00:13:31.980 thought to be completely unresponsive. It could 00:13:31.980 --> 00:13:34.460 really change how we understand and detect consciousness 00:13:34.460 --> 00:13:37.519 itself. That's incredible potential, truly. And 00:13:37.519 --> 00:13:39.659 on a lighter note, Meta is apparently going to 00:13:39.659 --> 00:13:43.639 let anyone create lifelike AI NPCs. NPCs. Non 00:13:43.639 --> 00:13:45.799 -player characters. Yeah, characters that can 00:13:45.799 --> 00:13:49.120 hold realistic, dynamic conversations. Huh. So 00:13:49.120 --> 00:13:51.139 you could build your own digital friends. Or 00:13:51.139 --> 00:13:53.240 characters for games. Or personalized virtual 00:13:53.240 --> 00:13:55.769 assistants you can chat with naturally. Could 00:13:55.769 --> 00:13:58.129 be interesting. Could be. And finally, for the 00:13:58.129 --> 00:14:00.730 developers listening, OpenAI rolled out some 00:14:00.730 --> 00:14:03.509 significant new features for Codex, their code 00:14:03.509 --> 00:14:06.190 generation tool. More power for coders. Pretty 00:14:06.190 --> 00:14:09.049 much. More power, more flexibility, more efficiency 00:14:09.049 --> 00:14:12.110 for people building software with AI's help, 00:14:12.210 --> 00:14:14.590 making their work faster, opening up new possibilities. 00:14:14.970 --> 00:14:17.370 Okay, so pulling back again, looking at all the 00:14:17.370 --> 00:14:20.070 quick hits together, what's the big overarching 00:14:20.070 --> 00:14:23.399 trend they suggest about where AI is going? I 00:14:23.399 --> 00:14:25.820 think it shows AI just rapidly embedding itself 00:14:25.820 --> 00:14:28.460 everywhere, becoming more specialized, more essential 00:14:28.460 --> 00:14:31.480 across really every part of our lives. Medicine, 00:14:31.700 --> 00:14:35.320 gaming, work tools, you name it. Sponsor. So 00:14:35.320 --> 00:14:38.039 wrapping things up today, it feels like we've 00:14:38.039 --> 00:14:42.059 seen AI reach maybe a new stage of maturity and 00:14:42.059 --> 00:14:44.840 strategic importance. We saw Microsoft making 00:14:44.840 --> 00:14:47.679 that bold move for AI independence, building 00:14:47.679 --> 00:14:50.740 its own powerful models like MAI Voice 1 and 00:14:50.740 --> 00:14:53.340 MAI One Preview. Right. Owning their stack. And 00:14:53.340 --> 00:14:56.539 then that rapid fire pace of really specialized 00:14:56.539 --> 00:15:01.000 AI tools emerging, the KPMG prompt, QuizGPT, 00:15:01.159 --> 00:15:04.080 CREA's video tool, all that stuff. Yeah, solving 00:15:04.080 --> 00:15:07.580 specific problems. And then Cohere's really significant 00:15:07.580 --> 00:15:10.340 step with secure enterprise -grade translation. 00:15:10.799 --> 00:15:14.450 That private air -gapped option. is key so the 00:15:14.450 --> 00:15:16.950 big idea connecting these i think the big idea 00:15:16.950 --> 00:15:19.370 is that ai isn't just about these giant general 00:15:19.370 --> 00:15:22.230 purpose models anymore that era is shifting okay 00:15:22.230 --> 00:15:24.269 it's increasingly about tailor -made solutions 00:15:24.269 --> 00:15:27.429 for specific often complex needs with a really 00:15:27.429 --> 00:15:30.629 strong focus on practical use specialized knowledge 00:15:30.629 --> 00:15:33.710 and like we saw with cohere rock solid security 00:15:33.710 --> 00:15:36.330 it's like building with specialized blocks exactly 00:15:36.330 --> 00:15:39.169 it's like stacking lego blocks of data and specialized 00:15:39.169 --> 00:15:41.429 ai components to build precisely what you need 00:15:41.429 --> 00:15:43.429 securely and effectively So we really encourage 00:15:43.429 --> 00:15:45.110 you to think about how these developments, this 00:15:45.110 --> 00:15:47.029 independence, this specialization, this focus 00:15:47.029 --> 00:15:49.110 on security might impact your work, maybe your 00:15:49.110 --> 00:15:50.950 studies, even just your personal life. The speed 00:15:50.950 --> 00:15:53.309 of it all is just, it's really something. Yeah, 00:15:53.389 --> 00:15:55.409 it's breathtaking. And here's a thought to leave 00:15:55.409 --> 00:15:58.789 you with. As AI gets more bespoke, more integrated, 00:15:58.929 --> 00:16:01.990 even air -gapped inside companies, what does 00:16:01.990 --> 00:16:04.090 true data ownership and intellectual property 00:16:04.090 --> 00:16:07.350 really look like? Beat. Especially when the AI 00:16:07.350 --> 00:16:09.929 itself is learning directly from your own unique 00:16:09.929 --> 00:16:13.460 proprietary knowledge. to sex silence. Something 00:16:13.460 --> 00:16:15.200 to chew on. Thank you for joining us on this 00:16:15.200 --> 00:16:17.220 deep dive today. We hope it sparked some new 00:16:17.220 --> 00:16:19.039 questions and some new insights for you to explore 00:16:19.039 --> 00:16:22.100 further. Until next time, keep learning. T -Row 00:16:22.100 --> 00:16:22.299 Music.