WEBVTT 00:00:00.000 --> 00:00:02.839 Okay, just imagine this for a second. An AI that 00:00:02.839 --> 00:00:05.660 could work as a doctor, but like hundreds of 00:00:05.660 --> 00:00:08.339 millions of miles away. Or how about one that 00:00:08.339 --> 00:00:10.599 kind of knows what you're about to say next, 00:00:10.720 --> 00:00:12.539 before you even finish the thought. It sounds 00:00:12.539 --> 00:00:14.679 like pure science fiction, doesn't it? Totally. 00:00:14.839 --> 00:00:18.640 Like something straight out of a movie. But believe 00:00:18.640 --> 00:00:20.480 it or not, it's actually happening. Right now. 00:00:20.660 --> 00:00:23.940 Welcome to this deep dive. Today, we're really 00:00:23.940 --> 00:00:28.420 digging into some incredible, almost, yeah. unbelievable 00:00:28.420 --> 00:00:31.160 leaps in artificial intelligence mm -hmm our 00:00:31.160 --> 00:00:33.380 source material today it takes us on quite a 00:00:33.380 --> 00:00:36.399 ride we're going from AI diagnosing astronauts 00:00:36.399 --> 00:00:39.679 way out in deep space all the way back to giving 00:00:39.679 --> 00:00:44.520 your everyday AI tools a serious, like, noticeable 00:00:44.520 --> 00:00:46.640 speed boost. Yeah. So you've got a fair bit to 00:00:46.640 --> 00:00:48.679 unpack. First up, we're going to look into Google's 00:00:48.679 --> 00:00:52.200 pretty amazing Space Doctor project. Right. Then 00:00:52.200 --> 00:00:54.500 we'll kind of zoom through some surprising AI 00:00:54.500 --> 00:00:56.939 highlights, see what's shifting in the landscape. 00:00:57.420 --> 00:00:59.539 After that, discover some powerful new tools 00:00:59.539 --> 00:01:01.859 you can actually, you know, use today. And then 00:01:01.859 --> 00:01:05.219 finally, we'll dive deep into Apple's Quiet Breakthrough. 00:01:05.439 --> 00:01:09.400 It's about making AI think and talk back much, 00:01:09.420 --> 00:01:11.799 much faster. And this isn't just about the headlines, 00:01:12.000 --> 00:01:14.000 is it? It's really about understanding what's 00:01:14.000 --> 00:01:17.000 truly important here, the implications for all 00:01:17.000 --> 00:01:19.519 of us in these rapid changes. Exactly. So let's 00:01:19.519 --> 00:01:22.819 unpack this. Okay, let's start way, way out there. 00:01:22.900 --> 00:01:26.420 Literally, NASA and Google, they're teaming up 00:01:26.420 --> 00:01:30.340 on something pretty unique. CMOD, that stands 00:01:30.340 --> 00:01:32.900 for... Crew Medical Officer Digital Assistant. 00:01:33.239 --> 00:01:35.219 Right. You should basically think of it as an 00:01:35.219 --> 00:01:37.959 AI doctor specifically for astronauts on those 00:01:37.959 --> 00:01:40.739 long deep space missions. Precisely. I mean, 00:01:40.760 --> 00:01:42.659 the big problem out there is the delay, right? 00:01:42.719 --> 00:01:45.879 A simple radio call back to Earth. It can take 00:01:45.879 --> 00:01:49.019 20 minutes each way, sometimes more. 20 minutes. 00:01:49.040 --> 00:01:51.180 Wow. So in a medical emergency, you just can't 00:01:51.180 --> 00:01:53.480 phone Houston and, you know, wait 40 minutes, 00:01:53.519 --> 00:01:55.920 maybe longer for advice. And that kind of delay 00:01:55.920 --> 00:01:58.299 isn't just inconvenient. It's potentially life 00:01:58.299 --> 00:02:00.480 -threatening. Yeah. When every second counts. 00:02:00.519 --> 00:02:02.500 Absolutely. What's really fascinating, I think, 00:02:02.500 --> 00:02:05.140 is how it's actually built, how it works. It 00:02:05.140 --> 00:02:09.419 runs on Google Cloud's Vertex AI, which is basically 00:02:09.419 --> 00:02:12.439 Google's platform for building and scaling up 00:02:12.439 --> 00:02:15.340 machine learning models. And it uses these powerful 00:02:15.340 --> 00:02:19.379 open source models, heard of Lama 3, Mistral 00:02:19.379 --> 00:02:22.930 3, Small. Yeah. The big names. And they're using 00:02:22.930 --> 00:02:24.330 them because they're incredibly versatile, right? 00:02:24.409 --> 00:02:26.610 They can handle complex stuff like medical diagnosis, 00:02:26.789 --> 00:02:30.330 but also just general language tasks. Makes them 00:02:30.330 --> 00:02:32.729 perfect for something this high stakes. And the 00:02:32.729 --> 00:02:35.069 early tests. Pretty impressive results, aren't 00:02:35.069 --> 00:02:37.810 they? Yeah, remarkably so. They're hitting like 00:02:37.810 --> 00:02:41.150 88 % diagnostic accuracy for injuries already. 00:02:41.520 --> 00:02:44.439 88%. And it's built specifically from the ground 00:02:44.439 --> 00:02:47.039 up for those totally unique space challenges, 00:02:47.180 --> 00:02:51.240 like zero real -time chat with Earth and absolutely 00:02:51.240 --> 00:02:53.379 no way to evacuate. Just imagine that pressure. 00:02:53.539 --> 00:02:55.759 You're completely on your own. Your AI is kind 00:02:55.759 --> 00:02:58.199 of your last medical resort. Exactly. And NASA's 00:02:58.199 --> 00:03:00.259 plan for it gets even more ambitious. Oh, yeah. 00:03:00.300 --> 00:03:02.080 They want to feed it live ultrasound images, 00:03:02.180 --> 00:03:05.120 continuous biometric data coming off the astronaut's 00:03:05.120 --> 00:03:07.520 sensors, and give it special training in space 00:03:07.520 --> 00:03:10.039 medicine, you know, covering all the weird health 00:03:10.039 --> 00:03:12.340 risks you've had. face outside Earth's a nice 00:03:12.340 --> 00:03:14.719 protective bubble. Things we barely think about 00:03:14.719 --> 00:03:17.400 down here. Right. Like bone density loss, radiation 00:03:17.400 --> 00:03:19.340 effects, all that stuff. It needs to understand 00:03:19.340 --> 00:03:21.740 the context. But OK, let's connect this back. 00:03:21.800 --> 00:03:23.939 This isn't just about getting doctors to Mars, 00:03:23.979 --> 00:03:25.699 is it? No, not at all. I mean, if you look at 00:03:25.699 --> 00:03:27.860 the bigger picture, the same underlying tech, 00:03:28.020 --> 00:03:31.919 these sophisticated AI diagnostic tools, they 00:03:31.919 --> 00:03:34.580 could absolutely revolutionize telemedicine right 00:03:34.580 --> 00:03:37.750 here on Earth. How so? Well, think about. Remote 00:03:37.750 --> 00:03:41.789 rural areas, places with hardly any doctors or, 00:03:41.949 --> 00:03:44.530 you know, crisis zones. Right. It's like those 00:03:44.530 --> 00:03:47.289 classic space tech spinoffs, stuff developed 00:03:47.289 --> 00:03:50.090 for space, finding amazing uses back home, like 00:03:50.090 --> 00:03:52.430 memory foam or even cordless drills came out 00:03:52.430 --> 00:03:55.349 of the Apollo program indirectly. This AI could 00:03:55.349 --> 00:03:58.050 turn desperate medical situations into ones with 00:03:58.050 --> 00:04:00.689 immediate smart support. OK, so beyond the cool 00:04:00.689 --> 00:04:03.729 space stuff, how does this AI doctor really change 00:04:03.729 --> 00:04:06.430 emergency care in those remote places here on 00:04:06.430 --> 00:04:09.219 Earth? It basically bridges huge communication 00:04:09.219 --> 00:04:12.580 gaps. It offers vital real time medical guidance 00:04:12.580 --> 00:04:14.919 when there's no human doctor around. All right. 00:04:14.939 --> 00:04:18.860 So moving from the the cosmic down to the competitive 00:04:18.860 --> 00:04:22.110 landscape here on Earth. We've seen some pretty 00:04:22.110 --> 00:04:24.930 interesting shifts and insights in AI just recently. 00:04:25.009 --> 00:04:27.529 For instance, one researcher did this independent 00:04:27.529 --> 00:04:30.850 test, pitting the much -hyped GPT -5 against 00:04:30.850 --> 00:04:35.790 Google's Gemini 2 .5 Pro, used 10 specific kind 00:04:35.790 --> 00:04:37.569 of tricky prompts. Yeah, everyone wants to know 00:04:37.569 --> 00:04:39.730 who's winning the AI race right now. Who's on 00:04:39.730 --> 00:04:43.129 top? Exactly. And speaking of like how these 00:04:43.129 --> 00:04:46.810 AIs actually think, the CEO of Google DeepMind 00:04:46.810 --> 00:04:49.589 recently tried to clarify this concept of a world 00:04:49.589 --> 00:04:51.790 model. Ah, yeah, world models. It's basically 00:04:51.790 --> 00:04:54.709 the AI's own internal simplified map of reality. 00:04:54.949 --> 00:04:56.889 Helps it kind of grasp how things work, cause 00:04:56.889 --> 00:04:58.649 and effect. So it's more than just pattern matching. 00:04:58.790 --> 00:05:01.009 Way more. Like a simple AI might just recite 00:05:01.009 --> 00:05:03.370 facts about a ball rolling downhill. But an AI 00:05:03.370 --> 00:05:05.389 with a world model sort of understands gravity, 00:05:05.490 --> 00:05:07.769 friction, momentum. It can predict what happens 00:05:07.769 --> 00:05:09.569 next because it has this internal simulation. 00:05:09.680 --> 00:05:11.740 this mental map makes its answers way smarter. 00:05:11.959 --> 00:05:14.240 Okay, now here's a twist I found really interesting. 00:05:14.879 --> 00:05:18.060 Speaks to the human side of AI. OpenAI actually 00:05:18.060 --> 00:05:21.399 brought back GPT -4 .0. They did, why? Apparently 00:05:21.399 --> 00:05:24.060 there was this surprisingly strong user backlash 00:05:24.060 --> 00:05:27.319 when they introduced GPT -5. Some users actually 00:05:27.319 --> 00:05:29.920 reported feeling like they'd lost a friend or 00:05:29.920 --> 00:05:33.199 even a partner. Whoa, seriously, like an emotional 00:05:33.199 --> 00:05:37.360 connection. Yeah. This feeling of like a familiar 00:05:37.360 --> 00:05:39.339 presence, a personality they connected with. 00:05:39.519 --> 00:05:41.779 It's become a real thing people are talking about. 00:05:41.819 --> 00:05:43.360 Just think about that. It's not just about how 00:05:43.360 --> 00:05:46.259 well it works anymore. Wow. People getting genuinely 00:05:46.259 --> 00:05:48.779 sentimental about their AI chatbot. That's wild. 00:05:49.199 --> 00:05:51.240 Isn't it? Definitely shows how deeply these things 00:05:51.240 --> 00:05:53.860 are weaving into our lives. But on a different 00:05:53.860 --> 00:05:57.779 note of AI mastery, OpenAI's O3 model, it totally 00:05:57.779 --> 00:06:00.500 crushed the Kaggle AI chess tournament. Oh, yeah. 00:06:00.759 --> 00:06:03.240 How did you? Won every single game against big 00:06:03.240 --> 00:06:06.540 rivals like Deep Sea Car 1, Grok 4. Just complete 00:06:06.540 --> 00:06:10.060 domination. A proper chess champ. And for a little 00:06:10.060 --> 00:06:11.860 bit of lightness, there's this kind of funny 00:06:11.860 --> 00:06:15.139 running joke online. This ex -user, Donald Boat, 00:06:15.220 --> 00:06:18.420 keeps asking big tech CEOs and AI people for 00:06:18.420 --> 00:06:20.819 super expensive hardware. Oh, I think I saw that. 00:06:20.899 --> 00:06:22.779 Sam Altman said yes to something, right? He did. 00:06:23.139 --> 00:06:26.480 And now this user got another public yes from 00:06:26.480 --> 00:06:28.959 someone else high up. It just kind of highlights 00:06:28.959 --> 00:06:32.019 the playful, almost meme -like side of this community, 00:06:32.160 --> 00:06:34.279 even when the tech itself is so serious. Yeah, 00:06:34.319 --> 00:06:36.339 it's a funny little side story. But speaking 00:06:36.339 --> 00:06:39.300 of serious stuff, the money side. Lolling, this 00:06:39.300 --> 00:06:42.339 AI fintech company in the UAE. They just raised 00:06:42.339 --> 00:06:45.879 a huge round. $48 million. $48 million. For what 00:06:45.879 --> 00:06:48.180 exactly? Their goal is automating finance tasks. 00:06:48.660 --> 00:06:50.860 They're already working with over 1 ,500 teams. 00:06:51.199 --> 00:06:53.519 It's not just about saving time on boring stuff 00:06:53.519 --> 00:06:56.220 like bookkeeping, though. Right. It's about shifting 00:06:56.220 --> 00:06:59.779 finance teams towards more strategic work. Like, 00:06:59.779 --> 00:07:02.920 imagine spending 80 % less time manually matching 00:07:02.920 --> 00:07:06.420 invoices and 80 % more time on forecasting or 00:07:06.420 --> 00:07:09.279 risk analysis. That's the kind of change AI is 00:07:09.279 --> 00:07:11.240 bringing to how businesses handle their money. 00:07:11.439 --> 00:07:13.660 Okay, going back to that user reaction, what 00:07:13.660 --> 00:07:15.279 does this emotional connection some people are 00:07:15.279 --> 00:07:17.899 forming tell us about AI's impact, how it's evolving? 00:07:18.199 --> 00:07:20.699 It shows users are building these really unexpected 00:07:20.699 --> 00:07:23.040 kind of personal relationships with the tools. 00:07:23.199 --> 00:07:26.470 It goes way beyond just simple utility. So beyond 00:07:26.470 --> 00:07:29.750 the huge companies and the high -level research, 00:07:29.949 --> 00:07:33.410 we're also seeing this explosion of new, much 00:07:33.410 --> 00:07:36.069 more accessible AI tools, tools that put real 00:07:36.069 --> 00:07:38.689 power directly into your hands. Exactly. This 00:07:38.689 --> 00:07:40.430 is where it gets really interesting for everyday 00:07:40.430 --> 00:07:42.870 users and creators. Yeah. We're seeing stuff 00:07:42.870 --> 00:07:46.230 like Microsoft Copilot 3D. Apparently, it can 00:07:46.230 --> 00:07:48.790 take a normal 2D picture and turn it into a full 00:07:48.790 --> 00:07:52.050 3D model, just one click. Huh, that's cool. Or... 00:07:52.490 --> 00:07:54.769 Gemini Gen EI, they're now offering a free version 00:07:54.769 --> 00:07:57.430 of VO3. That's for generating really, really 00:07:57.430 --> 00:08:00.290 impressive AI video. Opens up video creation 00:08:00.290 --> 00:08:02.889 to way more people. Okay. And there's a simpler 00:08:02.889 --> 00:08:05.089 tool called Business Idea. You just type in a 00:08:05.089 --> 00:08:07.069 topic and bam, it gives you tailored business 00:08:07.069 --> 00:08:09.870 concepts fast. Like having a brainstorming buddy 00:08:09.870 --> 00:08:12.670 on call. Pretty much. And there's N Factorial. 00:08:12.689 --> 00:08:15.029 Get this. It lets you have video calls with simulations 00:08:15.029 --> 00:08:17.389 of the world's best minds as your own personal 00:08:17.389 --> 00:08:21.269 tutor. Wow. Okay. That sounds intense, but useful. 00:08:21.800 --> 00:08:23.639 These are really practical applications, right? 00:08:23.699 --> 00:08:25.639 Taking complicated tech and making it usable 00:08:25.639 --> 00:08:27.920 for everyone, not just giant corporations or 00:08:27.920 --> 00:08:30.759 AI PhDs. Yeah, totally leveling the playing field. 00:08:31.000 --> 00:08:34.659 And on the quick hit news front, chat GBT actually 00:08:34.659 --> 00:08:37.039 went down recently. Oh, really? Yeah. And the 00:08:37.039 --> 00:08:39.059 number of people searching for info about it 00:08:39.059 --> 00:08:42.299 just spiked massively, which really shows how 00:08:42.299 --> 00:08:44.559 much people rely on these tools now for everything 00:08:44.559 --> 00:08:46.820 from writing emails to figuring out dinner ideas. 00:08:47.259 --> 00:08:50.200 Dependence is growing fast. For sure. And here's 00:08:50.200 --> 00:08:52.840 a big one for education. California, partnering 00:08:52.840 --> 00:08:55.519 with Google and Microsoft, they're now offering 00:08:55.519 --> 00:08:58.519 free AI tools and training for colleges across 00:08:58.519 --> 00:09:01.019 the state. That's huge, equipping the next generation. 00:09:01.340 --> 00:09:03.559 Definitely. And it ties into this idea that AI 00:09:03.559 --> 00:09:06.240 could soon power the rise of the $1 billion, 00:09:06.500 --> 00:09:09.840 one -person unicorn startup. Imagine that, one 00:09:09.840 --> 00:09:12.019 person building a billion -dollar company. Just 00:09:12.019 --> 00:09:14.340 leveraging AI tools for automation and scale 00:09:14.340 --> 00:09:16.360 changes the whole game for entrepreneurs. Absolutely. 00:09:17.080 --> 00:09:20.220 Also, Grok Imagine announced a new feature, turning 00:09:20.220 --> 00:09:23.019 images into video, more blurring the lines between 00:09:23.019 --> 00:09:25.460 spatic and dynamic content. Keeps moving fast. 00:09:25.779 --> 00:09:27.679 And one last thing on the sort of geoeconomic 00:09:27.679 --> 00:09:31.679 side. NVIDIA might start selling some high -end 00:09:31.679 --> 00:09:34.700 AI chips to China again, but apparently only 00:09:34.700 --> 00:09:38.019 if the U .S. government gets a cut. Interesting. 00:09:38.019 --> 00:09:40.240 Highlights just how strategic this tech is now. 00:09:40.299 --> 00:09:42.840 So much movement, so many quick shifts everywhere 00:09:42.840 --> 00:09:45.320 you look. Non -stop. So looking at all these 00:09:45.320 --> 00:09:48.419 new tools popping up beyond the individual headlines, 00:09:48.639 --> 00:09:50.740 what's the common thread? What ties them together? 00:09:50.980 --> 00:09:52.799 They really push accessibility and automation 00:09:52.799 --> 00:09:56.740 right down to individual users and small businesses, 00:09:56.779 --> 00:09:59.159 small ventures. So where is all this heading? 00:09:59.360 --> 00:10:02.659 What does it mean for the future of AI, how we 00:10:02.659 --> 00:10:04.539 interact with it every day? You know, we always 00:10:04.539 --> 00:10:06.980 wish AI could just spit out. answers like five, 00:10:07.059 --> 00:10:09.700 ten times faster, right? Without sounding dumb 00:10:09.700 --> 00:10:12.340 or losing quality. Yeah, that's the dream. Faster 00:10:12.340 --> 00:10:14.460 and smarter. Well, Apple researchers, kind of 00:10:14.460 --> 00:10:16.779 quietly, published this really interesting paper. 00:10:16.840 --> 00:10:20.710 It's called Your LLM Knows the Future. Ooh. Intriguing 00:10:20.710 --> 00:10:23.610 title. What's the core idea? This is where things 00:10:23.610 --> 00:10:26.230 get really fundamental, I think. They revealed 00:10:26.230 --> 00:10:29.309 a framework for what they call multi -token prediction. 00:10:29.730 --> 00:10:32.250 Multi -token. So you know how large language 00:10:32.250 --> 00:10:35.509 models, LLMs, the AI that understands and talks 00:10:35.509 --> 00:10:38.990 like us, usually predict just the very next word? 00:10:39.129 --> 00:10:41.389 Yeah, one word at a time, like building a sentence 00:10:41.389 --> 00:10:44.149 brick by brick. Right. This new method lets the 00:10:44.149 --> 00:10:47.450 LLM predict several words ahead, maybe even whole 00:10:47.450 --> 00:10:50.940 phrases. all at once ah okay so it's like the 00:10:50.940 --> 00:10:53.379 ai sees the whole thought the whole chunk of 00:10:53.379 --> 00:10:55.620 the sentence before it even starts writing it 00:10:55.620 --> 00:10:59.200 exactly and the result the ai finishes your sentences 00:10:59.200 --> 00:11:02.379 gives you answers in far fewer computational 00:11:02.379 --> 00:11:06.559 steps which means massively faster response times 00:11:06.559 --> 00:11:09.600 bingo and it's not just a tiny tweak they're 00:11:09.600 --> 00:11:11.399 saying it's a fundamental shift in how these 00:11:11.399 --> 00:11:14.000 models can operate they can anticipate and generate 00:11:14.000 --> 00:11:16.299 bigger chunks simultaneously so what kind of 00:11:16.299 --> 00:11:18.450 speed games are we talking do they test it They 00:11:18.450 --> 00:11:21.730 did. Using an open source model, Apple saw speeds 00:11:21.730 --> 00:11:24.269 jump two to three times faster for just general 00:11:24.269 --> 00:11:27.090 chat and Q &A stuff. Two to three times. That's 00:11:27.090 --> 00:11:29.970 already significant. But get this. For more predictable 00:11:29.970 --> 00:11:33.029 tasks, like writing computer code or doing math 00:11:33.029 --> 00:11:35.929 problems, it was up to five times faster. Five 00:11:35.929 --> 00:11:40.509 times. And critically, with absolutely no loss 00:11:40.509 --> 00:11:43.779 in quality or accuracy. Wow. Five times faster, 00:11:43.919 --> 00:11:47.840 same quality. Just imagine scaling that. Billions 00:11:47.840 --> 00:11:50.139 of user queries getting responses that much faster. 00:11:50.240 --> 00:11:54.179 The impact on real -time stuff is huge. And this 00:11:54.179 --> 00:11:56.480 technology, it kind of confirms something researchers 00:11:56.480 --> 00:11:58.820 have suspected for a while. Which is? That these 00:11:58.820 --> 00:12:03.379 LLMs already know more than just the single next 00:12:03.379 --> 00:12:05.980 word. They have this sort of latent knowledge, 00:12:06.059 --> 00:12:08.559 this hidden understanding of future words, sentence 00:12:08.559 --> 00:12:11.340 structure, the overall thought. Apple's method 00:12:11.340 --> 00:12:14.059 just figures out how to tap into that. Unlocking 00:12:14.059 --> 00:12:16.139 potential that was already there. Yeah. You know, 00:12:16.159 --> 00:12:18.519 I still wrestle with prompt drift myself sometimes. 00:12:18.720 --> 00:12:20.860 Oh, yeah. Where you're talking to an AI and after 00:12:20.860 --> 00:12:23.419 a few turns, it just starts to lose the plot. 00:12:24.090 --> 00:12:26.350 Forgets what you were originally asking. Coherence 00:12:26.350 --> 00:12:28.549 decay happens all the time. So faster and more 00:12:28.549 --> 00:12:30.789 coherent responses because it can complete the 00:12:30.789 --> 00:12:32.710 thought more fully. That would be an absolute 00:12:32.710 --> 00:12:35.049 game changer for me. Feels like a genuine leap. 00:12:35.230 --> 00:12:37.210 Definitely. And think about Apple integrating 00:12:37.210 --> 00:12:39.970 this. If they bake this multi -token prediction 00:12:39.970 --> 00:12:42.350 deep into their system, like into the core of 00:12:42.350 --> 00:12:45.470 iOS or Mac OS. Right. Then developers building 00:12:45.470 --> 00:12:48.210 apps for Apple stuff, they automatically get 00:12:48.210 --> 00:12:50.049 that performance edge locked in. A built -in 00:12:50.049 --> 00:12:53.090 advantage. Yeah. And if it rolls into Siri. or 00:12:53.090 --> 00:12:56.090 Apple intelligence, you'd get way snappier answers, 00:12:56.330 --> 00:12:59.529 much more natural feeling interactions without 00:12:59.529 --> 00:13:03.909 sacrificing the accuracy and potentially keeping 00:13:03.909 --> 00:13:05.590 more of the processing right there on your phone 00:13:05.590 --> 00:13:08.750 or computer. Good for privacy too. That is exactly 00:13:08.750 --> 00:13:11.029 the kind of differentiator Apple needs, isn't 00:13:11.029 --> 00:13:13.230 it? To really compete head on with Google, with 00:13:13.230 --> 00:13:16.649 open AI and this huge AI race. It's a powerful 00:13:16.649 --> 00:13:18.549 advantage if they can pull it off. Could really 00:13:18.549 --> 00:13:22.100 change the user experience. Big time. So thinking 00:13:22.100 --> 00:13:25.720 about our daily lives, how might this multi -token 00:13:25.720 --> 00:13:28.200 prediction, this ability for AI to sort of know 00:13:28.200 --> 00:13:30.759 the future of our sentences, how might that change 00:13:30.759 --> 00:13:33.360 how we interact with Siri or chat GPT or whatever 00:13:33.360 --> 00:13:37.000 comes next? It promises much snappier, more natural 00:13:37.000 --> 00:13:40.120 conversations, makes the AI feel less like a 00:13:40.120 --> 00:13:42.919 machine figuring things out word by word, and 00:13:42.919 --> 00:13:44.600 more like something that truly anticipates what 00:13:44.600 --> 00:13:48.240 you need. More satisfying, ultimately. Sponsor. 00:13:48.559 --> 00:13:50.220 Okay, so let's try and wrap our heads around 00:13:50.220 --> 00:13:52.360 this. What's the big idea we can really take 00:13:52.360 --> 00:13:54.919 away from today's deep dive? It seems pretty 00:13:54.919 --> 00:13:58.759 clear that AI is just expanding incredibly fast, 00:13:58.840 --> 00:14:00.879 pushing boundaries everywhere. Yeah, definitely. 00:14:00.940 --> 00:14:03.440 From these super specialized critical roles, 00:14:03.559 --> 00:14:05.440 like being a doctor millions of miles away in 00:14:05.440 --> 00:14:08.529 space. Right. All the way down to becoming just 00:14:08.529 --> 00:14:11.830 deeply woven into our everyday tools, our communication, 00:14:12.149 --> 00:14:14.350 making everything from, I don't know, making 00:14:14.350 --> 00:14:17.669 art to managing company finances faster, easier, 00:14:17.809 --> 00:14:20.470 more accessible for basically everyone. It feels 00:14:20.470 --> 00:14:22.750 like AI is fundamentally about bridging gaps, 00:14:22.929 --> 00:14:24.669 whether it's that huge physical distance gap 00:14:24.669 --> 00:14:27.750 for medical help in space, or maybe that cognitive 00:14:27.750 --> 00:14:30.330 gap between the idea in your head and how quickly 00:14:30.330 --> 00:14:33.210 an AI can help you express it clearly. It's connecting 00:14:33.210 --> 00:14:35.690 things across vast distances and complex thoughts. 00:14:35.919 --> 00:14:38.220 Yeah, that's a good way to put it. And crucially, 00:14:38.340 --> 00:14:40.940 we're seeing this deeper sophistication. Yeah. 00:14:41.179 --> 00:14:44.980 Right. Things like the world models concept and 00:14:44.980 --> 00:14:47.759 this groundbreaking multi -token prediction from 00:14:47.759 --> 00:14:51.019 Apple. These aren't just about what AI can do 00:14:51.019 --> 00:14:53.399 anymore. It's about how it fundamentally processes 00:14:53.399 --> 00:14:55.720 information, how it generates its responses, 00:14:55.940 --> 00:15:00.179 making it feel more intelligent, more anticipatory 00:15:00.179 --> 00:15:03.279 than ever before. It really raises a big question 00:15:03.279 --> 00:15:07.059 then for you listening. Given AI's growing ability 00:15:07.059 --> 00:15:09.539 to, you know, know the future of our thoughts 00:15:09.539 --> 00:15:12.360 or to act as this immediate expert assistant 00:15:12.360 --> 00:15:15.740 in really complex situations, how do you think 00:15:15.740 --> 00:15:18.059 this is going to reshape how we learn, how we 00:15:18.059 --> 00:15:20.200 work together, maybe even how we collaborate 00:15:20.200 --> 00:15:21.860 in fields we haven't even thought about yet? 00:15:22.000 --> 00:15:23.820 Yeah, it's not just about getting answers faster 00:15:23.820 --> 00:15:25.799 anymore, is it? It's changing the very nature 00:15:25.799 --> 00:15:27.820 of how we interact with technology itself and 00:15:27.820 --> 00:15:30.039 what becomes possible when that technology starts 00:15:30.039 --> 00:15:32.200 to anticipate us. We certainly hope this deep 00:15:32.200 --> 00:15:34.580 dive gave you some fresh insights today and maybe 00:15:34.580 --> 00:15:36.740 sparked a bit of your own curiosity about this, 00:15:36.779 --> 00:15:39.320 this incredible fast -moving frontier. Until 00:15:39.320 --> 00:15:40.519 next time, keep exploring.