WEBVTT 00:00:00.000 --> 00:00:03.640 So is AI actually coming for your job? Or maybe 00:00:03.640 --> 00:00:06.139 is it creating totally new ones, things we can't 00:00:06.139 --> 00:00:09.160 even picture yet? And what happens when the internet 00:00:09.160 --> 00:00:11.199 itself, you know, the global network we rely 00:00:11.199 --> 00:00:16.379 on starts to kind of break apart? Beep. Welcome 00:00:16.379 --> 00:00:19.420 to the Deep Dive. Today we're digging into a 00:00:19.420 --> 00:00:22.260 really fascinating set of sources, mainly a recent 00:00:22.260 --> 00:00:24.929 newsletter that caught our eye. Our goal here 00:00:24.929 --> 00:00:27.390 is simple. Give you the essential insights, a 00:00:27.390 --> 00:00:29.469 shortcut really, to understanding the big AI 00:00:29.469 --> 00:00:32.270 trends shaping things right now. We'll cut through 00:00:32.270 --> 00:00:33.969 the noise, connect the dots for you, and get 00:00:33.969 --> 00:00:35.509 you up to speed without the usual information 00:00:35.509 --> 00:00:38.270 overload. Exactly. And for this deep dive, we're 00:00:38.270 --> 00:00:40.409 starting with Goldman Sachs. They've got this... 00:00:41.160 --> 00:00:44.719 surprisingly nuanced view on AI and jobs. Then 00:00:44.719 --> 00:00:46.679 we'll jump into some really cool and maybe slightly 00:00:46.679 --> 00:00:49.780 weird AI highlights. Think AI influencers going 00:00:49.780 --> 00:00:51.880 viral, leaked prompts, that sort of thing. And 00:00:51.880 --> 00:00:53.539 finally, we'll tie it all together. Look at the 00:00:53.539 --> 00:00:55.740 bigger picture, the geopolitics of AI, and yeah, 00:00:55.780 --> 00:00:57.500 even where it's completely blocked. It's quite 00:00:57.500 --> 00:00:59.500 a landscape. Okay, yeah, let's unpack that first 00:00:59.500 --> 00:01:02.280 part. For months, all the talk has been about 00:01:02.280 --> 00:01:05.319 this AI jobpocalypse, right? Like massive unemployment 00:01:05.319 --> 00:01:08.000 is just around the corner. But this Goldman Sachs 00:01:08.000 --> 00:01:11.180 research, it paints a much more, well, Subtle 00:01:11.180 --> 00:01:13.760 picture. It isn't just yes or no on jobs. It's 00:01:13.760 --> 00:01:16.219 more like, yes, some jobs change or go away, 00:01:16.319 --> 00:01:20.349 but only for a bit. new ones get created. Right. 00:01:20.530 --> 00:01:22.609 And what's really interesting is how they lean 00:01:22.609 --> 00:01:24.969 on history. They're saying, look, roles will 00:01:24.969 --> 00:01:27.989 shift. Yeah, definitely. Productivity will probably 00:01:27.989 --> 00:01:30.950 spike. But crucially, new kinds of work will 00:01:30.950 --> 00:01:33.069 pop up. It's basically a pattern we've seen before. 00:01:33.170 --> 00:01:35.310 Think steam engines, you know, or the Internet 00:01:35.310 --> 00:01:37.989 itself. History shows like 60 percent of the 00:01:37.989 --> 00:01:39.790 jobs people have today didn't even exist back 00:01:39.790 --> 00:01:42.310 in 1940. Technology keeps doing this, gets rid 00:01:42.310 --> 00:01:44.989 of old roles, invents new ones. So how big is 00:01:44.989 --> 00:01:47.659 this disruption then? What are the actual numbers? 00:01:48.200 --> 00:01:51.120 Well, Goldman Sachs estimates for the U .S. maybe 00:01:51.120 --> 00:01:53.900 a 6 % to 7 % risk of jobs being displaced if 00:01:53.900 --> 00:01:56.219 AI fully scales up everywhere. But right now, 00:01:56.319 --> 00:01:59.500 today, it's closer to 2 .5 % risk just based 00:01:59.500 --> 00:02:01.799 on current AI getting used a bit more. And they 00:02:01.799 --> 00:02:05.019 only predict about a 0 .5 % temporary rise in 00:02:05.019 --> 00:02:07.239 unemployment during this shift. They call it 00:02:07.239 --> 00:02:09.199 frictional short -term stuff. They figured it 00:02:09.199 --> 00:02:11.500 should sort itself out in about two years. So, 00:02:11.599 --> 00:02:14.020 yeah, maybe not quite the catastrophe some headlines 00:02:14.020 --> 00:02:16.449 suggest. And that leads to the big question, 00:02:16.530 --> 00:02:19.150 right? Which jobs? Who's most affected? Goldman 00:02:19.150 --> 00:02:21.409 points to roles with lots of repetitive tasks, 00:02:21.689 --> 00:02:24.150 information processing. Think programmers, accountants, 00:02:24.210 --> 00:02:26.150 admin assistants, a lot of customer support roles, 00:02:26.349 --> 00:02:29.189 high exposure there. On the other side, low exposure 00:02:29.189 --> 00:02:31.990 jobs. Those tend to need complex physical skills 00:02:31.990 --> 00:02:34.409 or really nuanced human judgment. So air traffic 00:02:34.409 --> 00:02:36.889 controllers, radiologists, pharmacists, even 00:02:36.889 --> 00:02:39.990 CEOs, actually. AI can help them, sure, but it 00:02:39.990 --> 00:02:42.270 can't really replace that core strategic thinking 00:02:42.270 --> 00:02:45.210 or human touch. Not yet. And despite all the 00:02:45.210 --> 00:02:47.849 chatter, you know, the actual use of generative 00:02:47.849 --> 00:02:51.389 AI in companies is still, well, it's pretty low. 00:02:51.490 --> 00:02:54.229 The research says only about 9 .3 % of U .S. 00:02:54.229 --> 00:02:55.949 firms are really running it in production right 00:02:55.949 --> 00:02:58.710 now. But even with that, we are seeing an early 00:02:58.710 --> 00:03:01.129 impact. Unemployment for people in their 20s 00:03:01.129 --> 00:03:04.509 and 30s in those AI exposed jobs, it's ticked 00:03:04.509 --> 00:03:08.389 up about three points this year in 2025. So while 00:03:08.389 --> 00:03:11.330 full adoption isn't here, the initial waves are 00:03:11.330 --> 00:03:13.680 definitely making ripples. Yeah. And if you zoom 00:03:13.680 --> 00:03:16.479 out, Goldman's basically saying AI is a big disturbance, 00:03:16.659 --> 00:03:19.020 a market shift. And what happens in big shifts? 00:03:19.159 --> 00:03:21.780 The small, agile ones move fast. They find the 00:03:21.780 --> 00:03:23.620 new opportunities first. That's the opening. 00:03:23.759 --> 00:03:25.939 So the key thing here, maybe the big takeaway 00:03:25.939 --> 00:03:27.819 for you listening, is that your job title isn't 00:03:27.819 --> 00:03:29.680 everything anymore. It's really about the results 00:03:29.680 --> 00:03:32.039 you deliver, the value you create. OK, so digging 00:03:32.039 --> 00:03:34.240 a bit deeper than just adapt. Yeah. What's the 00:03:34.240 --> 00:03:36.759 core message for navigating this job evolution? 00:03:36.900 --> 00:03:38.990 How do you actually thrive? Focus on delivering 00:03:38.990 --> 00:03:41.750 tangible value, specific outcomes. Don't just 00:03:41.750 --> 00:03:44.129 define yourself by a static role. Adaptability 00:03:44.129 --> 00:03:46.669 plus demonstrated results. That's key. All right. 00:03:46.669 --> 00:03:48.569 Let's shift gears from the big economic view. 00:03:48.669 --> 00:03:50.909 Let's get into some of the specific, sometimes 00:03:50.909 --> 00:03:53.650 kind of wild things happening in AI right now. 00:03:53.750 --> 00:03:56.439 This is where it gets. really fascinating on 00:03:56.439 --> 00:03:58.439 the ground level oh yeah some really wild examples 00:03:58.439 --> 00:04:02.460 like uh olivia moore creating that 100 ai influencer 00:04:02.460 --> 00:04:05.240 i mean it got 500 followers and a viral video 00:04:05.240 --> 00:04:07.819 in just a week and she spent what less than 10 00:04:07.819 --> 00:04:10.879 minutes a day on it with just three tools it's 00:04:10.879 --> 00:04:12.699 kind of mind -bending makes you wonder about 00:04:12.699 --> 00:04:15.379 the future of online identity right and then 00:04:15.379 --> 00:04:17.660 on the more technical side you had this ai builder 00:04:17.660 --> 00:04:20.040 testing eight different ai agents same setup 00:04:20.040 --> 00:04:22.420 for all of them some just crushed it performed 00:04:22.420 --> 00:04:24.779 amazingly well others completely bombed it really 00:04:24.910 --> 00:04:27.310 shows you the huge range in capabilities right 00:04:27.310 --> 00:04:30.170 now. It's not all created equal. And our sources 00:04:30.170 --> 00:04:32.129 also mentioned an engineer, Avi, who broke down 00:04:32.129 --> 00:04:35.829 MCP. That's Multimodal Conversational Pathways. 00:04:36.230 --> 00:04:38.850 Basically letting AI handle text, images, audio, 00:04:39.069 --> 00:04:41.310 all in one chat smoothly. He used these really 00:04:41.310 --> 00:04:44.029 clear visuals. Made a super complex but important 00:04:44.029 --> 00:04:48.139 idea easy to grasp. That's a real skill. OpenAI, 00:04:48.459 --> 00:04:51.220 no surprise, they've been busy, apparently quietly 00:04:51.220 --> 00:04:53.680 talking to journalists, maybe gearing up for 00:04:53.680 --> 00:04:55.519 an IPO. And they weren't just talking current 00:04:55.519 --> 00:04:58.259 stuff. They were teasing some really big ambitions, 00:04:58.399 --> 00:05:02.379 like an AI -powered browser, maybe a social app, 00:05:02.480 --> 00:05:05.120 and even, get this, a potential brain chip interface. 00:05:05.560 --> 00:05:07.560 Whoa, okay, hold on. Scaling to a billion queries 00:05:07.560 --> 00:05:09.920 is one thing, but connecting directly to a brain 00:05:09.920 --> 00:05:13.040 chip? That's a fundamental change in how humans 00:05:13.040 --> 00:05:15.639 and computers interact. Yeah. Wow. Two sex islands. 00:05:15.759 --> 00:05:17.529 It definitely makes you think about... the guardrails 00:05:17.529 --> 00:05:19.389 needed, doesn't it? Because it's not all sleek 00:05:19.389 --> 00:05:22.250 progress like those leaked Grok prompts. Exactly. 00:05:22.250 --> 00:05:25.089 Not all smooth sailing. Those leaked internal 00:05:25.089 --> 00:05:28.069 prompts for Grok stuff like crazy conspiracist 00:05:28.069 --> 00:05:31.230 or unhinged comedian, they were designed to push 00:05:31.230 --> 00:05:33.769 the AI towards really controversial stuff, even 00:05:33.769 --> 00:05:37.389 the insane sex shock humor. It's a raw look at 00:05:37.389 --> 00:05:39.269 prompt engineering. It shows how sensitive these 00:05:39.269 --> 00:05:41.089 models are. You nudge the prompt a little and 00:05:41.089 --> 00:05:43.550 the output can go way off track. We call it prompt 00:05:43.550 --> 00:05:45.850 drift, honestly. I still wrestle with prompt 00:05:45.850 --> 00:05:48.199 drift. myself sometimes. Tiny tweaks can lead 00:05:48.199 --> 00:05:51.079 to wildly different unexpected results. These 00:05:51.079 --> 00:05:53.420 leaks just really hammer home how much careful 00:05:53.420 --> 00:05:56.470 setup and oversight is still needed. Right. But 00:05:56.470 --> 00:05:58.649 on a more constructive note, funding is still 00:05:58.649 --> 00:06:01.430 flowing. The Allen Institute for AI, they just 00:06:01.430 --> 00:06:04.250 secured a huge round, $152 million. Big chunks 00:06:04.250 --> 00:06:06.930 came from the NSF and NVIDIA. And the focus, 00:06:07.129 --> 00:06:09.569 building open source AI models specifically for 00:06:09.569 --> 00:06:11.689 science and medicine, that could genuinely speed 00:06:11.689 --> 00:06:13.370 up discoveries, which is pretty exciting. So 00:06:13.370 --> 00:06:15.569 we've got AI influencers, tricky prompts, brain 00:06:15.569 --> 00:06:17.730 chips, open source science. What's the single 00:06:17.730 --> 00:06:20.209 thread tying these diverse, almost opposing things 00:06:20.209 --> 00:06:23.240 together? It's the sheer speed and breadth. Incredible 00:06:23.240 --> 00:06:25.240 innovation is happening, but it's bumping right 00:06:25.240 --> 00:06:27.879 up against serious ethical questions often faster 00:06:27.879 --> 00:06:30.730 than we can answer them. Sponsor. All right, 00:06:30.730 --> 00:06:32.589 let's pivot again. Time for some quick -fire 00:06:32.589 --> 00:06:35.069 updates from across the AI world, and then we 00:06:35.069 --> 00:06:37.290 need to talk about this growing digital divide. 00:06:37.529 --> 00:06:39.829 It's important. Okay, quick hits. Meta apparently 00:06:39.829 --> 00:06:43.170 planned its fourth AI division shakeup just this 00:06:43.170 --> 00:06:46.589 year, 2025. Shows you how fast even the giants 00:06:46.589 --> 00:06:48.850 are scrambling to adapt, right? It's constant 00:06:48.850 --> 00:06:53.129 change. OpenAI launched GPT -GO in India. Big 00:06:53.129 --> 00:06:55.129 move. Ten times higher usage limits. Clearly 00:06:55.129 --> 00:06:57.290 trying to expand access and, you know, get into 00:06:57.290 --> 00:07:00.360 huge markets. We also saw Stanford HAI. and Google 00:07:00.360 --> 00:07:03.279 DeepMind teaming up for an AI challenge. More 00:07:03.279 --> 00:07:05.740 blending of academia and industry trying to crack 00:07:05.740 --> 00:07:08.379 big problems. And for practical tools, Perplexity 00:07:08.379 --> 00:07:10.339 Finance can now give you live transcripts from 00:07:10.339 --> 00:07:13.120 earnings calls. Super useful for anyone tracking 00:07:13.120 --> 00:07:15.740 stocks. But it's not all just tools and restructuring. 00:07:15.839 --> 00:07:18.560 There are real concerns cropping up. Like in 00:07:18.560 --> 00:07:20.800 Texas, they've accused meta and character AI 00:07:20.800 --> 00:07:24.139 of, well, basically misleading people, presenting 00:07:24.139 --> 00:07:26.199 their AI chatbots as mental health resources. 00:07:26.699 --> 00:07:29.180 That raises some pretty serious ethical alarms, 00:07:29.360 --> 00:07:31.319 you know. About responsibility. Yeah, it really 00:07:31.319 --> 00:07:33.040 does. And that brings us to something pretty 00:07:33.040 --> 00:07:36.060 stark. If AI is everywhere, why is it banned 00:07:36.060 --> 00:07:39.220 in some places? Our sources show that chat GPT 00:07:39.220 --> 00:07:41.699 is blocked or just plain unavailable in 20 countries 00:07:41.699 --> 00:07:44.800 right now in 2025. Yeah, this is where the geopolitics 00:07:44.800 --> 00:07:48.680 gets really visible. places like Cyber News and 00:07:48.680 --> 00:07:51.759 OpenAI itself shows 18 of those bands are government 00:07:51.759 --> 00:07:54.620 blocks. Why? Usually censorship, political control, 00:07:54.839 --> 00:07:57.079 maybe security fears about data going overseas. 00:07:57.420 --> 00:07:59.600 Then there are two others, Hong Kong and Belarus, 00:07:59.920 --> 00:08:01.920 where OpenAI just doesn't offer the service. 00:08:02.079 --> 00:08:03.980 Probably complicated business or legal reasons 00:08:03.980 --> 00:08:06.019 there. It's not just tech, it's policy. Right. 00:08:06.060 --> 00:08:08.019 And it's interesting, in places like China or 00:08:08.019 --> 00:08:11.259 Iran, they aren't just blocking tools like ChatGPT, 00:08:11.319 --> 00:08:13.339 these foreign large language models, or LLMs. 00:08:13.459 --> 00:08:15.680 They're actively pushing their own domestic AI. 00:08:16.490 --> 00:08:19.290 like quen or deep seek and the key difference 00:08:19.290 --> 00:08:21.629 those are much easier for the governments to 00:08:22.170 --> 00:08:24.769 Well, monitoring control. Exactly. This is becoming 00:08:24.769 --> 00:08:27.230 a massive geopolitical thing. Access to these 00:08:27.230 --> 00:08:30.110 advanced AI tools. It's turning into a strategic 00:08:30.110 --> 00:08:32.649 advantage like oil or microchips used to be. 00:08:32.870 --> 00:08:35.289 We're basically watching the Internet split in 00:08:35.289 --> 00:08:37.129 real time. You've got one version, sort of the 00:08:37.129 --> 00:08:39.830 U .S. model, aiming for more openness and another 00:08:39.830 --> 00:08:42.389 version that's much more locked down, controlled. 00:08:42.730 --> 00:08:44.990 China is a prime example. They're not just blocking. 00:08:45.049 --> 00:08:47.409 They're racing to build their own powerful AI 00:08:47.409 --> 00:08:50.289 to compete and maintain control. It really feels 00:08:50.289 --> 00:08:52.730 like a new kind of. digital iron curtain coming 00:08:52.730 --> 00:08:55.169 down. So putting these rapid updates together 00:08:55.169 --> 00:08:58.309 with the bands, how does this shape our view 00:08:58.309 --> 00:09:01.070 of where AI is headed globally? It shows AI's 00:09:01.070 --> 00:09:03.269 future is complex. It's going to be defined by 00:09:03.269 --> 00:09:06.250 both explosive innovation and these growing global 00:09:06.250 --> 00:09:08.789 governance headaches and divides, a real tension 00:09:08.789 --> 00:09:10.850 there. So let's bring it all together. We've 00:09:10.850 --> 00:09:13.250 covered the job market shifts, the sometimes 00:09:13.250 --> 00:09:15.710 chaotic pace of development, the ethical questions, 00:09:15.889 --> 00:09:19.669 this emerging global divide. What's the big idea, 00:09:19.809 --> 00:09:22.500 the main thing, you? Our listener should take 00:09:22.500 --> 00:09:24.820 away from this deep dive today. I think it's 00:09:24.820 --> 00:09:27.279 that AI isn't just one thing. It's this whole 00:09:27.279 --> 00:09:30.159 dynamic system full of opportunity. Yeah, but 00:09:30.159 --> 00:09:33.100 also real challenges. The key takeaway maybe 00:09:33.100 --> 00:09:35.120 is that succeeding now really depends on your 00:09:35.120 --> 00:09:37.100 awareness. Understanding the tech is part of 00:09:37.100 --> 00:09:39.759 it, sure. But you also need to grasp its economic 00:09:39.759 --> 00:09:42.960 ripples, its social impact, and these geopolitical 00:09:42.960 --> 00:09:45.080 games being played. It really demands you stay 00:09:45.080 --> 00:09:47.580 curious, stay learning, and don't just sit back 00:09:47.580 --> 00:09:50.100 passively. We really hope this discussion has 00:09:50.100 --> 00:09:52.440 given you a clearer, maybe more nuanced view 00:09:52.440 --> 00:09:55.840 of AI's broad impact on work, on innovation and 00:09:55.840 --> 00:09:58.120 on these global splits. It's worth thinking about 00:09:58.120 --> 00:10:00.480 how your own skills, your own perspective might 00:10:00.480 --> 00:10:02.779 need to adapt in this environment. New resources 00:10:02.779 --> 00:10:04.639 are always popping up for learning, building 00:10:04.639 --> 00:10:07.840 or just automating things better. So keep exploring. 00:10:08.059 --> 00:10:10.519 Yeah. And as you think about all this, maybe 00:10:10.519 --> 00:10:13.960 consider this. How much does today's AI scene 00:10:13.960 --> 00:10:16.940 feel like the early Internet? You know, that 00:10:16.940 --> 00:10:19.799 mix of huge promise, total chaos, unexpected 00:10:19.799 --> 00:10:21.919 winners and losers, and these strange divides 00:10:21.919 --> 00:10:24.539 appearing. What does that parallel suggest about 00:10:24.539 --> 00:10:27.340 the next 10 years of AI? And maybe more importantly, 00:10:27.480 --> 00:10:29.019 where do you want to position yourself in that 00:10:29.019 --> 00:10:31.120 future? Thanks for joining us for this deep dive 00:10:31.120 --> 00:10:33.879 into the complex world of AI. We'll be back soon 00:10:33.879 --> 00:10:36.120 with more insights. Out Hero Music.