WEBVTT 00:00:00.000 --> 00:00:03.480 Okay, imagine this, an AI, but it's not designed 00:00:03.480 --> 00:00:07.419 to be polite. It's built for internet chaos, 00:00:07.839 --> 00:00:12.560 memes, sarcasm, maybe even gets a little wild. 00:00:12.939 --> 00:00:16.760 This AI just launched. Get this, the creator's 00:00:16.760 --> 00:00:20.820 other AI platform, its CEO, just stepped down 00:00:20.820 --> 00:00:23.760 because, well, its AI went on a pro -Nazi rant. 00:00:24.000 --> 00:00:26.600 Yeah, it's a really wild time out there, seriously. 00:00:27.000 --> 00:00:30.050 Welcome to the Deep Dive. Today, we are unpacking 00:00:30.050 --> 00:00:32.990 a really fascinating set of sources about what's 00:00:32.990 --> 00:00:35.049 happening right at the bleeding edge of artificial 00:00:35.049 --> 00:00:37.170 intelligence. Yeah, we're going to explore a 00:00:37.170 --> 00:00:39.429 really different kind AI. We'll dig into some, 00:00:39.469 --> 00:00:42.289 frankly, mind -bending new tools and, you know, 00:00:42.310 --> 00:00:44.229 wrestle with the very real, sometimes kind of 00:00:44.229 --> 00:00:46.670 unsettling implications of how fast this is all 00:00:46.670 --> 00:00:48.789 moving. Our mission today, as always, is pretty 00:00:48.789 --> 00:00:50.270 simple. We want to pull out the key nuggets for 00:00:50.270 --> 00:00:52.369 you, give you that shortcut so you can be genuinely 00:00:52.369 --> 00:00:54.509 well -informed on this. Let's do it. Okay, so 00:00:54.509 --> 00:00:57.380 let's unpack this first one. XAI. Elon Musk's 00:00:57.380 --> 00:01:00.280 company, they just launched Grok 4. And they're 00:01:00.280 --> 00:01:03.820 pitching it as the anti -Chat GPT, which is, 00:01:03.840 --> 00:01:05.780 that's a real philosophical shift, right, in 00:01:05.780 --> 00:01:09.019 how you design AI. Totally. A statement. Though 00:01:09.019 --> 00:01:12.840 the launch itself sounds like it was kind of 00:01:12.840 --> 00:01:15.019 messy. Yeah, pretty messy, actually. It started 00:01:15.019 --> 00:01:17.799 an hour late, apparently. And their chief scientist 00:01:17.799 --> 00:01:22.459 quit on launch day. Wow. Okay. Not ideal. Not 00:01:22.459 --> 00:01:26.000 ideal. Still, Grok 4. It's trained on their big 00:01:26.000 --> 00:01:29.120 supercomputer, Colossus, supposedly optimized 00:01:29.120 --> 00:01:31.739 for, and this is their phrase, scientist -grade 00:01:31.739 --> 00:01:35.120 reasoning. Huh. Scientist -grade? Yeah. Musk 00:01:35.120 --> 00:01:37.780 calls it Big Bang Intelligence. He even claimed 00:01:37.780 --> 00:01:39.459 they've literally run out of test questions. 00:01:40.010 --> 00:01:42.969 which is that's a bold claim a very bold claim 00:01:42.969 --> 00:01:45.650 it is okay but here's where for me the features 00:01:45.650 --> 00:01:49.569 get really interesting so grok4 it takes multimodal 00:01:49.569 --> 00:01:52.150 inputs meaning meaning you can give it text yeah 00:01:52.150 --> 00:01:54.989 but also images and apparently video is coming 00:01:54.989 --> 00:01:57.250 soon too just think about how flexible that is 00:01:57.250 --> 00:02:00.310 okay yeah then you've got grok4 code that's specifically 00:02:00.310 --> 00:02:03.370 for developers right to help write code debug 00:02:03.370 --> 00:02:05.879 it There's also a new voice experience they're 00:02:05.879 --> 00:02:08.759 talking about, promising like natural conversations. 00:02:08.759 --> 00:02:11.479 You can interrupt it. It flows better. That's 00:02:11.479 --> 00:02:13.240 a big deal for how we interact with these things. 00:02:13.360 --> 00:02:18.060 And crucially, real time web access. It uses 00:02:18.060 --> 00:02:20.560 something called Dupsearch. So it pulls live 00:02:20.560 --> 00:02:22.939 info, especially from X, right into the chat. 00:02:23.139 --> 00:02:24.639 Okay. And this is where it leans into that whole 00:02:24.639 --> 00:02:26.560 personality thing, right? The chaos element. 00:02:26.680 --> 00:02:29.039 Absolutely. It apparently has this amazing internet 00:02:29.039 --> 00:02:31.860 culture fluency. Yeah. Like it gets memes, it 00:02:31.860 --> 00:02:35.060 gets slang, sarcasm. Grok just understands it. 00:02:35.099 --> 00:02:37.099 It's like they built a digital native, you know? 00:02:37.280 --> 00:02:39.840 Yeah. Born online, plugged right into the pulse 00:02:39.840 --> 00:02:41.780 of the internet. It just gets the vibe. Right. 00:02:42.039 --> 00:02:45.169 But there's a shadow here too. Yeah. Earlier 00:02:45.169 --> 00:02:47.729 Grok versions, they already made headlines. Racist 00:02:47.729 --> 00:02:51.629 answers, biased stuff. I remember that. And XAI, 00:02:51.710 --> 00:02:54.629 their stance is very light moderation. They talk 00:02:54.629 --> 00:02:57.310 about free speech, which, OK, has led to some 00:02:57.310 --> 00:02:59.490 major backlash, understandably. Sure. So the 00:02:59.490 --> 00:03:01.789 big question now is, with all this new power, 00:03:01.889 --> 00:03:06.310 more reach, can Grok 4 actually stay usable or 00:03:06.310 --> 00:03:09.069 is it going to go rogue? Musk hasn't really addressed 00:03:09.069 --> 00:03:12.270 these like core issues directly. It feels different, 00:03:12.409 --> 00:03:14.639 doesn't it? OpenAI is building for reliability. 00:03:15.240 --> 00:03:18.219 Anthropic is focused on alignment, safety, and 00:03:18.219 --> 00:03:21.120 grok. They seem to be betting on, well, chaos, 00:03:21.419 --> 00:03:24.199 humor, raw developer power. It's a different 00:03:24.199 --> 00:03:26.900 path. So what's the takeaway? I think XAI is 00:03:26.900 --> 00:03:29.520 going all in on this idea that maybe people want 00:03:29.520 --> 00:03:33.479 AI that's less filtered. Less sanitized. Exactly. 00:03:33.620 --> 00:03:36.680 Faster models, real personality, sarcasm, freedom 00:03:36.680 --> 00:03:39.020 from what they call the woke defaults you see 00:03:39.020 --> 00:03:41.860 elsewhere. It's risky. It's totally unpredictable. 00:03:42.800 --> 00:03:45.639 But you got to wonder, maybe it works for a certain 00:03:45.639 --> 00:03:48.379 audience. Maybe it really clicks. So this chaos 00:03:48.379 --> 00:03:51.840 approach from Grok, it feels like such a fundamental 00:03:51.840 --> 00:03:54.539 difference. How do you think that changes how 00:03:54.539 --> 00:03:56.580 we might interact with AI? Like what's the core 00:03:56.580 --> 00:03:59.020 shift? Well, it's prioritizing that raw personality, 00:03:59.379 --> 00:04:02.919 you know, and speed. Yeah. Over maybe the traditional 00:04:02.919 --> 00:04:04.919 safety filters we're used to. Less about just 00:04:04.919 --> 00:04:06.879 facts, more about capturing that wild Internet 00:04:06.879 --> 00:04:08.979 energy. Yeah, something like that. But beyond 00:04:08.979 --> 00:04:12.270 Grok, the whole AI scene is just. Buzzing. So 00:04:12.270 --> 00:04:14.530 much else is happening. Likewise. Okay, so Google 00:04:14.530 --> 00:04:16.709 VO3. Now lets you make audio and video, starting 00:04:16.709 --> 00:04:19.110 from just one single image. Wait, from one image? 00:04:19.310 --> 00:04:21.209 You give it the first frame and you prompt the 00:04:21.209 --> 00:04:23.810 dialogue, the action, and it generates the sequence. 00:04:24.110 --> 00:04:26.920 Whoa. Then there's this new AI filmmaker tool 00:04:26.920 --> 00:04:29.699 trained on fully licensed footage, which is key. 00:04:29.839 --> 00:04:32.500 And it lets you change objects in a scene just 00:04:32.500 --> 00:04:34.720 by dragging them around. No need to write a whole 00:04:34.720 --> 00:04:37.759 new complex prompt. That sounds much more intuitive, 00:04:37.839 --> 00:04:40.819 like direct manipulation. Exactly. Very cool. 00:04:40.920 --> 00:04:44.120 And the AI browser wars. They're definitely heating 00:04:44.120 --> 00:04:47.360 up. Oh, yeah. Perplexity just dropped Comet, 00:04:47.579 --> 00:04:50.839 their new AI web browser, clearly trying to take 00:04:50.839 --> 00:04:52.939 a bite out of Google's dominance. Right. I saw 00:04:52.939 --> 00:04:54.779 that. But it's limited access for now. Yeah, 00:04:54.860 --> 00:04:57.720 just for their max users, early waitlist people. 00:04:57.899 --> 00:05:00.819 But still, it's out there. And OpenAI isn't sitting 00:05:00.819 --> 00:05:02.720 still either, right? They're reportedly dropping 00:05:02.720 --> 00:05:05.420 their own AI browser super soon. Yeah, like in 00:05:05.420 --> 00:05:08.620 weeks. You might not even need to leave ChatGPT 00:05:08.620 --> 00:05:11.089 to browse the web anymore. Think about that. 00:05:11.129 --> 00:05:13.209 The lines are just getting blurrier and blurrier. 00:05:13.410 --> 00:05:17.449 Okay, but shifting gears a bit. There was a more 00:05:17.449 --> 00:05:21.769 sobering development, too. This AI voice cloning 00:05:21.769 --> 00:05:24.410 thing, it just got very real. They successfully 00:05:24.410 --> 00:05:27.529 impersonated U .S. Senator Marco Rubio using 00:05:27.529 --> 00:05:30.509 AI voice clones. Seriously? Who did they fool? 00:05:31.069 --> 00:05:34.970 Five high -ranking officials. In just 15 seconds, 00:05:35.209 --> 00:05:37.870 they thought they were talking to him. Wow. 15 00:05:37.870 --> 00:05:41.660 seconds. Slight pause. You know, it's humbling 00:05:41.660 --> 00:05:44.480 how fast this stuff changes. Honestly, I still 00:05:44.480 --> 00:05:47.160 wrestle sometimes with how to how to filter the 00:05:47.160 --> 00:05:49.600 signal from the noise, especially when voices 00:05:49.600 --> 00:05:52.040 can be cloned that convincingly. The potential 00:05:52.040 --> 00:05:55.139 for misinformation is just huge. It really is. 00:05:55.199 --> 00:05:57.240 It's a serious challenge. But OK, on the other 00:05:57.240 --> 00:05:59.459 side of the coin, you have huge positive investment 00:05:59.459 --> 00:06:02.379 happening, too. Like Microsoft. Exactly. Microsoft 00:06:02.379 --> 00:06:05.459 is pouring in over $4 billion. Massive investment. 00:06:05.720 --> 00:06:08.480 Into what specifically? AI tools, training programs, 00:06:08.699 --> 00:06:11.600 cloud services. Their goal, and it's ambitious, 00:06:11.860 --> 00:06:14.439 is to help 20 million people get AI certificates. 00:06:14.779 --> 00:06:16.980 20 million, wow. Yeah, it's part of this big 00:06:16.980 --> 00:06:19.339 global push trying to get AI skills into schools, 00:06:19.399 --> 00:06:22.560 into careers. It's a huge bet on human upskilling. 00:06:22.680 --> 00:06:24.100 That's a major commitment. So when you look at 00:06:24.100 --> 00:06:25.620 all these different things, the creative tools, 00:06:25.819 --> 00:06:28.839 the browser wars, the security risks, the big 00:06:28.839 --> 00:06:31.939 investments. What do you see as the biggest implication? 00:06:32.199 --> 00:06:34.459 Where's the main thrust? I think it's that AI 00:06:34.459 --> 00:06:38.279 is just rapidly reshaping like fundamental digital 00:06:38.279 --> 00:06:41.439 interactions. Yeah. Across the board. Yeah, totally. 00:06:41.579 --> 00:06:44.240 New ways of doing things are popping up. Okay, 00:06:44.300 --> 00:06:45.920 let's drill down a bit more into some practical 00:06:45.920 --> 00:06:47.879 stuff, tools and insights you could actually 00:06:47.879 --> 00:06:51.100 use. Okay. So for developers, our sources highlighted 00:06:51.100 --> 00:06:53.800 some really game -changing design tips, ways 00:06:53.800 --> 00:06:56.180 to make apps look amazing using modern component 00:06:56.180 --> 00:06:58.680 libraries. Those are like pre -built UI bits, 00:06:58.839 --> 00:07:01.740 right? Exactly, pre -built elements. And using 00:07:01.740 --> 00:07:04.120 AI design tools that can help automate layouts, 00:07:04.399 --> 00:07:07.660 styling, makes things much faster. Cool. And 00:07:07.660 --> 00:07:10.209 then there's something called vibe coding. It 00:07:10.209 --> 00:07:12.350 sounds interesting. It promises to help you go 00:07:12.350 --> 00:07:15.269 from being like a no code user. Someone using 00:07:15.269 --> 00:07:17.490 drag and drop tool. Right. To actually creating 00:07:17.490 --> 00:07:21.850 full, powerful, custom web apps. Using AI to 00:07:21.850 --> 00:07:24.089 build them just by describing what you want. 00:07:24.310 --> 00:07:27.449 Whoa. So bridging that gap between no code and 00:07:27.449 --> 00:07:30.930 actual coding. Seems like it. That could be huge 00:07:30.930 --> 00:07:33.730 for creators. Definitely. And beyond those, there's 00:07:33.730 --> 00:07:36.769 a whole bunch of these, like, empowered AI tools 00:07:36.769 --> 00:07:39.689 changing daily workflows. Like what else? Okay, 00:07:39.730 --> 00:07:42.990 check this out. File .ai. It gives you structured, 00:07:42.990 --> 00:07:46.189 clean data from pretty much any file type. And 00:07:46.189 --> 00:07:49.750 it's zero shot. Zero shot, meaning it doesn't 00:07:49.750 --> 00:07:52.689 need examples beforehand. Exactly. It just understands 00:07:52.689 --> 00:07:54.990 and extracts the data structure. Pretty powerful. 00:07:55.290 --> 00:07:58.740 Then Magic Animator. You design something in 00:07:58.740 --> 00:08:02.220 Figma. The UI design tool. Right. This tool animates 00:08:02.220 --> 00:08:04.779 your Figma designs in seconds using AI. Quick 00:08:04.779 --> 00:08:07.300 prototyping. Nice. Yeah. Worried about social 00:08:07.300 --> 00:08:09.759 media. Instagram checker can apparently detect 00:08:09.759 --> 00:08:13.300 fake followers on your account. Useful. And for 00:08:13.300 --> 00:08:15.439 just capturing thoughts quickly, there's lazy. 00:08:15.519 --> 00:08:17.399 It's like a shortcut to capture notes everywhere 00:08:17.399 --> 00:08:19.540 and then chat with them later. Keep things organized. 00:08:19.839 --> 00:08:22.180 Simple but effective. And we also pulled some 00:08:22.180 --> 00:08:24.439 quick hits, kind of broader insights. Lame all 00:08:24.439 --> 00:08:27.810 me. Okay. Prompt tips for GPT. The advice is, 00:08:27.870 --> 00:08:31.029 tell it straight. Don't just let it agree with 00:08:31.029 --> 00:08:33.570 you all the time. Be direct with what you need. 00:08:33.710 --> 00:08:37.409 Good advice. Don't lead the witness. Right. Also, 00:08:37.490 --> 00:08:40.330 the CEO of Box, he shared tips on finding your 00:08:40.330 --> 00:08:43.649 next big business idea using AI agents, like 00:08:43.649 --> 00:08:46.149 using AI to scout for opportunities. Interesting. 00:08:46.269 --> 00:08:48.690 AI is a business development tool. And this one 00:08:48.690 --> 00:08:52.860 is wild. GPT, the AI model. apparently hallucinated 00:08:52.860 --> 00:08:55.259 about a non -existent app so often just made 00:08:55.259 --> 00:08:58.679 it up just made it up repeatedly that its ceo 00:08:58.679 --> 00:09:00.940 heard about these hallucinations and decided 00:09:00.940 --> 00:09:03.679 to actually build the app made the lie come true 00:09:03.679 --> 00:09:07.019 whoa Wait, seriously, that's almost like the 00:09:07.019 --> 00:09:09.740 AI is seeding reality now. It's kind of mind 00:09:09.740 --> 00:09:11.440 bending when you think about it. It really is. 00:09:11.519 --> 00:09:14.080 And another quick one. Claude, another big AI 00:09:14.080 --> 00:09:16.259 model, can now connect to learning apps. Things 00:09:16.259 --> 00:09:18.919 like Canvas, Panopto, Wiley. So more integration 00:09:18.919 --> 00:09:21.460 into education. Makes sense. And then finally. 00:09:21.870 --> 00:09:23.789 Just a really stark reminder of that chaotic 00:09:23.789 --> 00:09:26.029 frontier we started with talking about Grok. 00:09:26.049 --> 00:09:29.429 Yeah. Elon Musk's X platform, its CEO, just stepped 00:09:29.429 --> 00:09:31.950 down. Why? What happened? Specifically after 00:09:31.950 --> 00:09:34.590 an AI on the platform went on that pro -Nazi 00:09:34.590 --> 00:09:36.830 rant we mentioned earlier. Direct consequence. 00:09:37.149 --> 00:09:39.669 Wow. OK, so the chaos has a real world leadership 00:09:39.669 --> 00:09:41.629 fallout that brings it home. It really does. 00:09:42.190 --> 00:09:44.570 So thinking about all these tools, these quick 00:09:44.570 --> 00:09:47.750 hits, the good and the bad. What's the key takeaway 00:09:47.750 --> 00:09:49.830 for someone just trying to navigate this whole 00:09:49.830 --> 00:09:53.809 fast moving AI scene? I'd say they offer both, 00:09:53.830 --> 00:09:56.250 you know, incredible utility and serious, unpredictable 00:09:56.250 --> 00:09:59.269 outcomes. It's a double edged sword. That feels 00:09:59.269 --> 00:10:01.389 right. So if we try to connect all these threads, 00:10:01.549 --> 00:10:04.250 zoom out to the bigger picture, what really stands 00:10:04.250 --> 00:10:06.809 out from this deep dive? For me, it's the speed, 00:10:06.950 --> 00:10:10.029 the incredibly rapid, often chaotic, but you 00:10:10.029 --> 00:10:12.529 can't deny it, really innovative evolution of 00:10:12.529 --> 00:10:15.830 AI. Yeah, definitely the speed. And also we're 00:10:15.830 --> 00:10:18.120 seeing... this clear divergence in how people 00:10:18.120 --> 00:10:20.399 are building AI. You mean like Grok versus the 00:10:20.399 --> 00:10:23.940 others? Exactly. Grok embracing that less filtered 00:10:23.940 --> 00:10:26.480 personality driven thing, which is such a contrast 00:10:26.480 --> 00:10:29.019 to the established players focusing on reliability 00:10:29.019 --> 00:10:32.360 or alignment. It's a high stakes gamble on what 00:10:32.360 --> 00:10:34.980 users actually prefer. Yeah. And this deep dive, 00:10:35.039 --> 00:10:37.580 it really shows AI isn't just some separate tool 00:10:37.580 --> 00:10:40.019 anymore, is it? It's weaving itself into, well, 00:10:40.100 --> 00:10:43.139 everything. Creating content, writing code, yeah, 00:10:43.220 --> 00:10:45.919 but also spotting fraud, even shaping public 00:10:45.919 --> 00:10:48.759 discussion. Its reach is just becoming enormous. 00:10:49.159 --> 00:10:51.960 The benefits are obvious, definitely. But man, 00:10:52.139 --> 00:10:54.600 so are the risks, the ethical problems that pop 00:10:54.600 --> 00:10:57.500 up when things inevitably go off the rails. It's 00:10:57.500 --> 00:11:00.019 a constant tension. So turning it back to you 00:11:00.019 --> 00:11:02.580 listening, what stands out most to you from all 00:11:02.580 --> 00:11:05.340 this? Yeah, and maybe a final thought to leave 00:11:05.340 --> 00:11:09.320 you with. In a world where AI is... in some cases, 00:11:09.340 --> 00:11:12.259 being intentionally built for chaos. We know 00:11:12.259 --> 00:11:14.539 its outputs can actually shape reality sometimes. 00:11:14.840 --> 00:11:17.059 Yeah. Can we really prepare for what's next? 00:11:17.139 --> 00:11:19.299 Or it's just constant adaptation, constantly 00:11:19.299 --> 00:11:21.500 reacting. Is that our only real option now? It's 00:11:21.500 --> 00:11:23.240 something I keep thinking about. Keep exploring, 00:11:23.340 --> 00:11:25.179 everyone. Keep asking those questions and keep 00:11:25.179 --> 00:11:27.480 learning. Thanks for joining us for this deep 00:11:27.480 --> 00:11:28.960 dive, OT Row Music.