WEBVTT 00:00:00.000 --> 00:00:02.759 Imagine a world where we could map, I mean, every 00:00:02.759 --> 00:00:05.820 single inch of our planet, basically in near 00:00:05.820 --> 00:00:10.199 real time, or run an entire cutting edge science 00:00:10.199 --> 00:00:13.580 lab instantly just using AI. Yeah, it sounds 00:00:13.580 --> 00:00:14.960 like something straight out of science fiction, 00:00:15.119 --> 00:00:17.519 doesn't it? But, you know, pieces of that future, 00:00:17.600 --> 00:00:20.059 they're not just concepts anymore. They're actually 00:00:20.059 --> 00:00:23.300 here. They're actively reshaping how we see the 00:00:23.300 --> 00:00:26.160 Earth and just speeding up discovery like crazy. 00:00:26.510 --> 00:00:28.289 We're about to take a deep dive into exactly 00:00:28.289 --> 00:00:31.750 how. Welcome to the deep dive. Today, we're unpacking 00:00:31.750 --> 00:00:33.750 some really, truly groundbreaking developments 00:00:33.750 --> 00:00:35.929 in AI. We've got this fascinating newsletter 00:00:35.929 --> 00:00:37.990 in front of us that just landed. Yeah, it's absolutely 00:00:37.990 --> 00:00:40.090 packed, really dense. We'll kick off exploring 00:00:40.090 --> 00:00:43.450 Google's new planetary co -pilot, Alpha Earth 00:00:43.450 --> 00:00:45.149 Foundations. Honestly, it's kind of astonishing 00:00:45.149 --> 00:00:47.369 stuff. Then we'll zoom out a bit, look at a whole 00:00:47.369 --> 00:00:49.890 range of exciting new AI tools that are transforming 00:00:49.890 --> 00:00:52.259 industries like right now. And then finally, 00:00:52.380 --> 00:00:54.460 we'll shift our focus to something that, well, 00:00:54.600 --> 00:00:57.299 it genuinely bends the mind a little. Virtual 00:00:57.299 --> 00:01:00.659 AI scientists. Yes, you heard that correctly. 00:01:01.460 --> 00:01:04.560 So our mission today is to help you get a handle 00:01:04.560 --> 00:01:07.359 on this, gain knowledge quickly, maybe spot those 00:01:07.359 --> 00:01:10.659 key aha moments and hopefully avoid feeling just, 00:01:10.719 --> 00:01:12.920 you know, overwhelmed by how fast everything's 00:01:12.920 --> 00:01:14.959 moving. Exactly. We want to help you understand 00:01:14.959 --> 00:01:17.859 not just what's happening, but maybe why it matters, 00:01:17.959 --> 00:01:20.489 what it could mean for you down the line. Okay, 00:01:20.609 --> 00:01:22.689 so let's unpack this. Right. Let's begin with 00:01:22.689 --> 00:01:24.810 Google DeepMind's Alpha Earth Foundations. This 00:01:24.810 --> 00:01:26.530 system is already being called a planetary co 00:01:26.530 --> 00:01:29.329 -pilot for Earth data. What, in your view, is 00:01:29.329 --> 00:01:31.629 the real core innovation here? What's the big 00:01:31.629 --> 00:01:33.730 deal? Well, what's truly fascinating, I think, 00:01:33.750 --> 00:01:36.670 is its approach to just handling immense data. 00:01:37.049 --> 00:01:39.230 Like mind boggling amount. So instead of using 00:01:39.230 --> 00:01:42.209 raw pixel by pixel satellite images, which are 00:01:42.209 --> 00:01:45.329 huge files, Alpha Earth compresses petabytes. 00:01:45.450 --> 00:01:47.530 And that's, you know, thousands of terabytes 00:01:47.530 --> 00:01:50.269 of this Earth observation data. It compresses 00:01:50.269 --> 00:01:52.150 it into something they call embedding fields. 00:01:52.370 --> 00:01:55.269 Think of these as like tiny, super efficient, 00:01:55.349 --> 00:01:58.469 searchable summaries for every little 10 by 10 00:01:58.469 --> 00:02:01.599 meter patch of the Earth's surface. 10 by 10 00:02:01.599 --> 00:02:03.780 meters. Yeah. And the key thing is they're generated 00:02:03.780 --> 00:02:05.840 in near real time. It's constantly updating. 00:02:06.060 --> 00:02:07.819 Okay. That sounds like an incredible leap in 00:02:07.819 --> 00:02:10.240 just pure efficiency. So when we talk about performance, 00:02:10.439 --> 00:02:12.340 what kind of impact are we actually seeing? What 00:02:12.340 --> 00:02:14.400 does this new architecture deliver? Oh, it's 00:02:14.400 --> 00:02:17.300 profoundly faster and significantly cheaper to 00:02:17.300 --> 00:02:20.330 run. The system cuts storage needs by, get this, 00:02:20.430 --> 00:02:24.409 16 times, 16, which is just critical when you're 00:02:24.409 --> 00:02:26.430 dealing with planetary scale information. It 00:02:26.430 --> 00:02:28.750 also reduces error rates by about 24 % compared 00:02:28.750 --> 00:02:32.210 to the older, more traditional models. 24%, wow. 00:02:32.409 --> 00:02:34.490 Yeah. To put it in perspective, we're talking 00:02:34.490 --> 00:02:37.550 about generating and making available like 1 00:02:37.550 --> 00:02:40.939 .4 trillion of these embedding footprints. Every 00:02:40.939 --> 00:02:44.360 single year. That's like a new snapshot of the 00:02:44.360 --> 00:02:46.900 Earth's surface changing, what, roughly every 00:02:46.900 --> 00:02:49.500 couple of seconds. It's wild. Wow. OK, so this 00:02:49.500 --> 00:02:52.379 isn't just like a research paper or some theoretical 00:02:52.379 --> 00:02:55.719 concept. It's live right now, integrated into 00:02:55.719 --> 00:02:57.840 Google Earth Engine, which, you know, a lot of 00:02:57.840 --> 00:03:00.039 people already use. Can you give us maybe a few 00:03:00.039 --> 00:03:02.280 concrete examples? Like where is this being used 00:03:02.280 --> 00:03:04.039 in the real world? Oh, absolutely. There are 00:03:04.039 --> 00:03:05.699 plenty of applications already rolling out and 00:03:05.699 --> 00:03:07.539 they're impacting some really critical environmental 00:03:07.539 --> 00:03:11.620 work. For instance. Mapiomas down in Brazil. 00:03:11.780 --> 00:03:13.939 They're actively using Alpha Earth Foundations 00:03:13.939 --> 00:03:15.900 right now. They're monitoring Amazon deforestation 00:03:15.900 --> 00:03:18.560 in near real time, which gives them crucial data 00:03:18.560 --> 00:03:21.620 for conservation efforts. Then there's the global 00:03:21.620 --> 00:03:24.560 ecosystems, Atlas. They're leveraging it to map 00:03:24.560 --> 00:03:26.500 regions that were previously uncharted, really, 00:03:26.599 --> 00:03:29.680 like vast shrub lands, remote deserts. It helps 00:03:29.680 --> 00:03:32.039 us understand global biodiversity better. And 00:03:32.039 --> 00:03:34.479 it even outperformed other systems in these things 00:03:34.479 --> 00:03:36.939 called evapotranspiration tests. Okay, what's 00:03:36.939 --> 00:03:39.080 that exactly? That's basically measuring how 00:03:39.080 --> 00:03:41.939 water moves through ecosystems. You know, evaporation, 00:03:42.000 --> 00:03:44.120 plant transpiration. It's super vital for understanding 00:03:44.120 --> 00:03:46.819 climate cycles, planning agriculture, predicting 00:03:46.819 --> 00:03:49.460 droughts, that kind of thing. Right, right. So 00:03:49.460 --> 00:03:51.879 it sounds like this makes that kind of planet 00:03:51.879 --> 00:03:54.319 -scale intelligence much more accessible. It's 00:03:54.319 --> 00:03:56.680 moving beyond just the big players like NASA 00:03:56.680 --> 00:03:59.439 or other national space agencies. What's the 00:03:59.439 --> 00:04:01.599 bigger picture here? What does it mean that Google 00:04:01.599 --> 00:04:04.889 is building something like this? If we connect 00:04:04.889 --> 00:04:07.770 this to that bigger picture, Google is essentially 00:04:07.770 --> 00:04:10.050 building an operating system for Earth. Just 00:04:10.050 --> 00:04:13.009 imagine, like a single dynamic platform where 00:04:13.009 --> 00:04:15.810 all this planetary data, weather patterns, deforestation 00:04:15.810 --> 00:04:18.509 rates, water cycles, you name it, isn't just 00:04:18.509 --> 00:04:22.730 collected, but it's processed, analyzed, and 00:04:22.730 --> 00:04:24.550 then made accessible for pretty much anyone to 00:04:24.550 --> 00:04:26.550 build applications on top of it. Kind of like 00:04:26.550 --> 00:04:28.550 how an OS lets software run on your computer, 00:04:28.610 --> 00:04:31.420 right? Okay, yeah. An OS for Earth. And it's 00:04:31.420 --> 00:04:35.360 also a huge win for AI privacy design, which 00:04:35.360 --> 00:04:37.540 is really important because instead of analyzing, 00:04:37.560 --> 00:04:41.220 say, sensitive raw images directly, Alpha Earth 00:04:41.220 --> 00:04:43.300 works with these anonymized embedding fields. 00:04:43.439 --> 00:04:46.259 So you get the critical data insights without 00:04:46.259 --> 00:04:49.319 compromising individual privacy or specific location 00:04:49.319 --> 00:04:52.360 privacy. That's a really critical step for, you 00:04:52.360 --> 00:04:54.819 know, responsible AI deployment, building trust. 00:04:54.980 --> 00:04:59.160 Right. And if you're maybe open AI, well. Perhaps 00:04:59.160 --> 00:05:01.019 you'd better move fast on this. Google seems 00:05:01.019 --> 00:05:03.139 to have taken a big lead on this kind of planetary 00:05:03.139 --> 00:05:05.480 data infrastructure. Seems that way. So if this 00:05:05.480 --> 00:05:07.959 makes planet scale data so much more accessible, 00:05:08.180 --> 00:05:10.699 what's the biggest shift this brings, say, for 00:05:10.699 --> 00:05:12.980 smaller organizations or even individual users? 00:05:13.360 --> 00:05:15.519 I'd say it really democratizes global mapping. 00:05:15.660 --> 00:05:17.480 It's not just for the big labs anymore. It's 00:05:17.480 --> 00:05:20.379 potentially for everyone. OK, so we've seen how 00:05:20.379 --> 00:05:24.290 AI is giving us this. unprecedented new way to 00:05:24.290 --> 00:05:26.250 see and understand the whole planet with Alpher. 00:05:26.490 --> 00:05:28.990 That's the big picture, the macro view. But AI 00:05:28.990 --> 00:05:31.350 is also revolutionizing things at a much smaller 00:05:31.350 --> 00:05:34.050 scale, right? Individual tasks, specific industries. 00:05:34.430 --> 00:05:36.889 Let's pivot now to that broader AI landscape. 00:05:37.050 --> 00:05:39.449 What new tools or trends have kind of jumped 00:05:39.449 --> 00:05:42.069 out at you recently? Yeah, it's been a real flurry 00:05:42.069 --> 00:05:45.430 lately. So many developments. For instance, XAI, 00:05:45.689 --> 00:05:48.449 you know, Elon Musk's company. They just opened 00:05:48.449 --> 00:05:51.509 a wait list for Imagine. It's their new image 00:05:51.509 --> 00:05:54.009 and video generation feature. And what's cool 00:05:54.009 --> 00:05:55.810 is it apparently includes audio capabilities, 00:05:55.990 --> 00:05:58.889 too, hinting at that multimodal creation future. 00:05:59.189 --> 00:06:01.629 Image, video, and audio. Right. Then there's 00:06:01.629 --> 00:06:03.310 Ideogram. They released something called Character. 00:06:03.569 --> 00:06:05.810 It makes it super easy to just swap faces or 00:06:05.810 --> 00:06:08.189 specific characters into generated images. You 00:06:08.189 --> 00:06:10.129 can see how that has huge implications for creative 00:06:10.129 --> 00:06:12.269 industries, right? Marketing, digital art, maybe 00:06:12.269 --> 00:06:14.370 even entertainment. Definitely. And I saw something, 00:06:14.449 --> 00:06:17.029 too, about specialized AI agents, almost like 00:06:17.029 --> 00:06:19.430 virtual employees. Can you tell me more about 00:06:19.430 --> 00:06:21.930 that? Exactly. That's a really significant trend 00:06:21.930 --> 00:06:24.370 we're seeing emerge. There's this new open source 00:06:24.370 --> 00:06:28.129 repository out there showcasing like 40 specialized 00:06:28.129 --> 00:06:31.610 cloud agents. And each one is designed to streamline 00:06:31.610 --> 00:06:34.370 very specific business functions. They can even 00:06:34.370 --> 00:06:37.709 sort of mimic an entire workforce, handling stuff 00:06:37.709 --> 00:06:40.009 from customer service bots to complex data analysis. 00:06:40.110 --> 00:06:42.389 The potential there to automate tasks is just 00:06:42.389 --> 00:06:44.750 enormous. Could free up human teams for more 00:06:44.750 --> 00:06:48.199 complex creative thinking. And honestly, I still 00:06:48.199 --> 00:06:50.620 wrestle with prompt drift myself sometimes. You 00:06:50.620 --> 00:06:52.819 know that thing where your AI's answers start 00:06:52.819 --> 00:06:55.449 to... subtly change or degrade over time. Even 00:06:55.449 --> 00:06:57.370 if you get the same input, it makes consistency 00:06:57.370 --> 00:06:59.449 a real headache. So these highly specialized 00:06:59.449 --> 00:07:02.250 purpose -built tools, they're particularly interesting 00:07:02.250 --> 00:07:04.410 because they aim to get rid of that variability, 00:07:04.670 --> 00:07:06.850 make things more reliable. That's wild to think 00:07:06.850 --> 00:07:08.930 about, an entire virtual workforce. And then 00:07:08.930 --> 00:07:10.529 there's this thing called Showrunner. It's been 00:07:10.529 --> 00:07:13.389 described as the Netflix of AI for film. What 00:07:13.389 --> 00:07:15.389 exactly does that do? Yeah, Showrunner is pretty 00:07:15.389 --> 00:07:17.290 remarkable, especially for creative professionals. 00:07:17.810 --> 00:07:21.129 Film, video, it lets you take existing film scenes 00:07:21.129 --> 00:07:23.639 and essentially clone them. You can replace actors 00:07:23.639 --> 00:07:26.199 with AI -generated ones. You can control their 00:07:26.199 --> 00:07:29.019 acting styles, their expressions, and even precisely 00:07:29.019 --> 00:07:31.980 copy the whole cinematic look, the lighting, 00:07:32.060 --> 00:07:34.600 the framing from a reference scene. It's kind 00:07:34.600 --> 00:07:36.540 of like having an entire production studio just 00:07:36.540 --> 00:07:39.139 sitting on your desktop. It makes professional 00:07:39.139 --> 00:07:41.959 -quality video creation way more accessible, 00:07:42.139 --> 00:07:44.939 potentially, and speeds up iteration like crazy. 00:07:45.040 --> 00:07:47.350 It's pretty incredible for creative work. Beyond 00:07:47.350 --> 00:07:50.189 these really fascinating individual tools, what 00:07:50.189 --> 00:07:53.310 about the big industry movements, the strategic 00:07:53.310 --> 00:07:55.829 plays? It feels like a high stakes chess game 00:07:55.829 --> 00:07:57.689 sometimes out there. It certainly does. Huge 00:07:57.689 --> 00:08:00.230 investments happening. So on the investment front, 00:08:00.370 --> 00:08:03.029 Microsoft is reportedly spending a record like 00:08:03.029 --> 00:08:06.250 $30 billion this quarter on AI investments. Just 00:08:06.250 --> 00:08:08.490 shows their massive commitment. $30 billion in 00:08:08.490 --> 00:08:12.160 one quarter. And then there's Groke, a relatively 00:08:12.160 --> 00:08:15.040 new AI chip startup. They're in the process of 00:08:15.040 --> 00:08:18.139 raising $600 million. Their goal is to directly 00:08:18.139 --> 00:08:20.420 challenge NVIDIA's dominance in the AI hardware 00:08:20.420 --> 00:08:24.839 market with their LPU chip architecture. So competition 00:08:24.839 --> 00:08:27.519 heating up there. And on the more practical consumer 00:08:27.519 --> 00:08:30.199 side, Google and YouTube, they're now using AI 00:08:30.199 --> 00:08:33.299 for age checks on user accounts. Just a straightforward 00:08:33.299 --> 00:08:35.080 application for safety and compliance. Okay. 00:08:35.120 --> 00:08:37.539 And what about politically? Or strategically, 00:08:37.759 --> 00:08:39.899 any interesting moves from the big tech players 00:08:39.899 --> 00:08:42.940 regarding, say, regulation or how they plan to 00:08:42.940 --> 00:08:45.120 control this tech? Definitely some maneuvering 00:08:45.120 --> 00:08:47.080 there. Google, for example, publicly stated it 00:08:47.080 --> 00:08:49.840 will sign the EU's voluntary AI code of practice. 00:08:49.879 --> 00:08:52.500 So that indicates a willingness to align with 00:08:52.500 --> 00:08:54.960 European regulatory thinking. Meta, on the other 00:08:54.960 --> 00:08:59.240 hand, stated they would not sign it. So signaling 00:08:59.240 --> 00:09:01.360 may be a different approach, maybe prioritizing 00:09:01.360 --> 00:09:04.100 open source development differently. Mark Zuckerberg 00:09:04.100 --> 00:09:06.200 also hinted that Meta won't open source all of 00:09:06.200 --> 00:09:08.639 its most advanced superintelligence AI models 00:09:08.639 --> 00:09:10.759 down the road, perhaps learning from past releases. 00:09:11.039 --> 00:09:13.419 Plus, Microsoft secured this really interesting 00:09:13.419 --> 00:09:17.080 open AI access deal. It basically ensures that 00:09:17.080 --> 00:09:19.460 they'll have access to open AI's models even 00:09:19.460 --> 00:09:22.740 after AGI, you know, artificial general intelligence 00:09:22.740 --> 00:09:24.860 is potentially achieved. That's a huge strategic 00:09:24.860 --> 00:09:27.629 hedge for them. guarantees access no matter what. 00:09:28.129 --> 00:09:30.669 So with all these rapid developments, specialized 00:09:30.669 --> 00:09:33.070 agents, massive investments, strategic corporate 00:09:33.070 --> 00:09:35.870 moves, what's the core overarching trend here? 00:09:35.970 --> 00:09:38.649 What's the one liner? I'd say AI is just expanding 00:09:38.649 --> 00:09:41.529 rapidly, embedding itself into basically every 00:09:41.529 --> 00:09:44.210 industry. It's driving both huge innovation and 00:09:44.210 --> 00:09:46.429 really intense competition. OK, so we've covered 00:09:46.429 --> 00:09:49.230 how AI is mapping the planet with this incredible 00:09:49.230 --> 00:09:52.610 new detail and how it's sparking this explosion 00:09:52.610 --> 00:09:54.950 of new tools across all sorts of industries. 00:09:55.210 --> 00:09:57.330 Now, let's talk about something that kind of 00:09:57.330 --> 00:09:59.389 unites these capabilities in a really profound 00:09:59.389 --> 00:10:04.169 way. Virtual AI scientists. Stanford University 00:10:04.169 --> 00:10:07.389 and the Chan Zuckerberg Biohub. They just launched 00:10:07.389 --> 00:10:10.029 a system that acts like an entire AI research 00:10:10.029 --> 00:10:12.629 lab. This is where it gets really interesting, 00:10:12.730 --> 00:10:14.629 I think. It truly is. Yeah, this is mind -bending 00:10:14.629 --> 00:10:17.889 stuff. Their setup isn't just like one AI tool. 00:10:18.090 --> 00:10:21.350 It's a whole ecosystem. It includes an AI principal 00:10:21.350 --> 00:10:23.490 investigator that's basically the lead scientist 00:10:23.490 --> 00:10:26.370 making the calls. It has specialized AI agents 00:10:26.370 --> 00:10:28.330 acting as the researchers doing the work. And 00:10:28.330 --> 00:10:30.149 get this, they even have digital lab meetings 00:10:30.149 --> 00:10:32.669 that literally finish in seconds. Just cuts out 00:10:32.669 --> 00:10:34.549 all that administrative time, discussion overhead. 00:10:34.809 --> 00:10:36.590 Meetings in seconds. Yeah, it's a real glimpse 00:10:36.590 --> 00:10:38.649 into an entirely new way of doing scientific 00:10:38.649 --> 00:10:40.870 discovery, a different paradigm. Okay, but what 00:10:40.870 --> 00:10:43.070 have these virtual scientists actually discovered? 00:10:43.289 --> 00:10:45.629 Are there tangible results from this? purely 00:10:45.629 --> 00:10:48.230 virtual process? Or is it still theoretical? 00:10:48.610 --> 00:10:51.250 No, the results are concrete and pretty impactful 00:10:51.250 --> 00:10:55.029 already. They've successfully designed 92 distinct 00:10:55.029 --> 00:10:58.789 COVID nanobody designs. Now, nanobodies, think 00:10:58.789 --> 00:11:01.129 of them as like tiny, really effective antibodies. 00:11:01.389 --> 00:11:03.590 They're much smaller, more stable than conventional 00:11:03.590 --> 00:11:05.690 ones. Makes them easier to work with, potentially 00:11:05.690 --> 00:11:08.070 more potent against diseases. Okay, 92 designs. 00:11:08.330 --> 00:11:10.029 Right. And out of those 92 designs generated 00:11:10.029 --> 00:11:12.629 by the AI system, two have already shown success 00:11:12.629 --> 00:11:16.629 in actual real -world lab testing, which is a 00:11:16.629 --> 00:11:18.870 remarkable outcome, frankly, for an entirely 00:11:18.870 --> 00:11:22.009 virtual process, really shows AI's power to accelerate 00:11:22.009 --> 00:11:24.629 biomedical research directly. That's astonishing. 00:11:25.070 --> 00:11:27.029 How does this virtual lab actually function, 00:11:27.110 --> 00:11:28.590 though? What are the mechanics behind it? How 00:11:28.590 --> 00:11:31.230 does it generate these specific testable breakthroughs? 00:11:31.269 --> 00:11:33.950 So the AI principal investigator, the PI, is 00:11:33.950 --> 00:11:36.090 kind of the orchestrator. It identifies the research 00:11:36.090 --> 00:11:39.190 problem, then it forms task -specific agent teams. 00:11:39.389 --> 00:11:41.809 It assigns specialized roles to other AI models, 00:11:41.990 --> 00:11:45.120 like data analyst or experiment designer. These 00:11:45.120 --> 00:11:47.399 agents then run these rigorous internal debates, 00:11:47.519 --> 00:11:50.059 almost like human scientists arguing over data 00:11:50.059 --> 00:11:52.559 to refine hypotheses and design experiments. 00:11:52.820 --> 00:11:55.399 And they even automatically call on external 00:11:55.399 --> 00:11:58.460 specialized tools when needed, like AlphaFold, 00:11:58.559 --> 00:12:01.360 Google DeepMind's system for predicting protein 00:12:01.360 --> 00:12:03.840 structures. They call on tools like that all 00:12:03.840 --> 00:12:06.860 by themselves, integrate that complex data seamlessly. 00:12:07.120 --> 00:12:09.580 They basically operate autonomously through the 00:12:09.580 --> 00:12:12.240 whole research cycle. from hypothesis all the 00:12:12.240 --> 00:12:14.299 way to experimental design. This raises a really 00:12:14.299 --> 00:12:17.000 important question then. Why does this matter 00:12:17.000 --> 00:12:19.940 so profoundly for the future of science? What 00:12:19.940 --> 00:12:22.399 core bottlenecks does it actually address? Well, 00:12:22.419 --> 00:12:24.539 think about traditional science. Most of it is 00:12:24.539 --> 00:12:26.580 inherently slow, right? It's often blocked by 00:12:26.580 --> 00:12:28.720 endless meetings, coordinating schedules. It's 00:12:28.720 --> 00:12:32.460 constrained by budgets, funding cycles, and frequently 00:12:32.460 --> 00:12:34.559 limited by expert bottlenecks. You need that 00:12:34.559 --> 00:12:36.620 one specific specialist who's always swamped. 00:12:37.519 --> 00:12:39.580 These AI agents, they just skip all that. They 00:12:39.580 --> 00:12:41.320 don't get tired. They don't need weekends off. 00:12:41.519 --> 00:12:44.159 They certainly don't care if a lab is short -staffed 00:12:44.159 --> 00:12:46.519 or someone's on vacation. And maybe crucially, 00:12:46.639 --> 00:12:49.159 they're transparent. You can literally watch 00:12:49.159 --> 00:12:52.200 their entire reasoning process unfold step by 00:12:52.200 --> 00:12:55.039 step. That gives you an unprecedented level of 00:12:55.039 --> 00:12:57.220 auditability. You can see how they reached a 00:12:57.220 --> 00:13:00.500 conclusion. Whoa. Seriously, imagine scaling 00:13:00.500 --> 00:13:02.950 this. Solving problems at that kind of speed, 00:13:03.149 --> 00:13:06.009 like running a billion queries simultaneously 00:13:06.009 --> 00:13:09.289 to find a cure or some new material. It's staggering. 00:13:09.490 --> 00:13:11.629 Yeah. What are the current limitations, though? 00:13:11.690 --> 00:13:13.389 It sounds almost too good to be true. Where does 00:13:13.389 --> 00:13:15.649 it fall short right now? Yeah, it's definitely 00:13:15.649 --> 00:13:17.429 not perfect yet. We're still in the early days 00:13:17.429 --> 00:13:20.570 here. Really complex, long term studies that 00:13:20.570 --> 00:13:22.899 required, you know. deep, nuanced, contextual 00:13:22.899 --> 00:13:25.139 understanding over years, that's still hard. 00:13:25.320 --> 00:13:28.519 Or those genuinely creative leaps, the kind that 00:13:28.519 --> 00:13:30.779 involve really abstract conceptual breakthroughs, 00:13:30.779 --> 00:13:33.600 that still largely seems to need human intuition, 00:13:33.820 --> 00:13:36.720 human insight. But if this approach scales and 00:13:36.720 --> 00:13:39.179 improves at the rate... It seems to be trending. 00:13:39.259 --> 00:13:41.299 Well, the next major scientific breakthrough 00:13:41.299 --> 00:13:43.620 may be a cure for a really challenging disease 00:13:43.620 --> 00:13:46.320 or a solution to a complex environmental problem. 00:13:46.639 --> 00:13:48.539 It might very well come from what essentially 00:13:48.539 --> 00:13:50.759 looks like a slack thread between four bots and, 00:13:50.820 --> 00:13:53.399 you know, a protein structure tool like AlphaFold. 00:13:53.500 --> 00:13:56.139 So does this mean human scientists are, you know, 00:13:56.139 --> 00:13:58.639 on a path to becoming obsolete? Is that the trajectory? 00:13:59.279 --> 00:14:01.419 Not yet, I don't think so. Humans are still absolutely 00:14:01.419 --> 00:14:04.220 crucial for that complex creativity, for asking 00:14:04.220 --> 00:14:06.519 the right questions, for guiding the big picture 00:14:06.519 --> 00:14:09.740 strategy. Mid -roll sponsor, Reed. So yeah, we've 00:14:09.740 --> 00:14:11.639 covered some truly profound advancements today, 00:14:11.740 --> 00:14:14.259 haven't we? I mean, from Google's Alpha Earth 00:14:14.259 --> 00:14:16.500 mapping basically our entire planet in incredible 00:14:16.500 --> 00:14:20.179 detail, near real time, to... Just this burst 00:14:20.179 --> 00:14:23.340 of innovative new AI tools that are fundamentally 00:14:23.340 --> 00:14:26.120 changing how we work, how we create, how businesses 00:14:26.120 --> 00:14:28.559 even operate day to day. Yeah. And then that 00:14:28.559 --> 00:14:31.639 that mind bending idea of virtual AI scientists 00:14:31.639 --> 00:14:34.460 just autonomously conducting research and getting 00:14:34.460 --> 00:14:36.440 real world breakthroughs like with those COVID 00:14:36.440 --> 00:14:38.779 nanobodies. It really shows, I think, that AI 00:14:38.779 --> 00:14:40.639 isn't just assisting us anymore. It's rapidly 00:14:40.639 --> 00:14:42.879 becoming a genuine collaborative partner, maybe 00:14:42.879 --> 00:14:45.600 even a primary driver in the actual process of 00:14:45.600 --> 00:14:48.360 human discovery itself. Exactly. What's truly 00:14:48.360 --> 00:14:50.399 fascinating. fascinating here I find is just 00:14:50.399 --> 00:14:52.980 how AI is accelerating our basic fundamental 00:14:52.980 --> 00:14:56.059 capacity, our capacity to understand and interact 00:14:56.059 --> 00:14:58.340 with the incredibly complex world around us. 00:14:58.419 --> 00:15:00.639 It feels like we're stacking Lego blocks of data 00:15:00.639 --> 00:15:03.659 and insight, you know, but at an exponential 00:15:03.659 --> 00:15:06.039 rate, building these knowledge structures way 00:15:06.039 --> 00:15:09.460 faster than ever before. We are definitely, undoubtedly, 00:15:09.539 --> 00:15:13.039 moving into an era where these incredibly complex 00:15:13.039 --> 00:15:15.919 global problems, things like precise environmental 00:15:15.919 --> 00:15:18.799 monitoring or accelerating critical disease research, 00:15:18.980 --> 00:15:21.360 they can now be tackled with just unprecedented 00:15:21.360 --> 00:15:24.899 speed and scale and efficiency. So what does 00:15:24.899 --> 00:15:27.299 this all mean for you listening right now? We 00:15:27.299 --> 00:15:29.639 are clearly entering an era where these profound 00:15:29.639 --> 00:15:31.899 capabilities things once exclusively reserved 00:15:31.899 --> 00:15:35.019 for, you know, National labs or massive supercomputers 00:15:35.019 --> 00:15:37.059 that are rapidly becoming accessible. So the 00:15:37.059 --> 00:15:38.779 question is, how will you use these powerful 00:15:38.779 --> 00:15:41.039 new lenses to see the world around you? Or maybe 00:15:41.039 --> 00:15:42.940 use these incredible new tools to solve the problems 00:15:42.940 --> 00:15:45.000 that matter most to you? Yeah, the power of these 00:15:45.000 --> 00:15:47.259 kinds of deep dives, just getting into this complex, 00:15:47.379 --> 00:15:50.379 fast -moving information. It's hopefully helping 00:15:50.379 --> 00:15:52.679 us all keep pace with this really exciting future 00:15:52.679 --> 00:15:55.279 that's unfolding. We sincerely hope this deep 00:15:55.279 --> 00:15:57.720 dive gave you some valuable new insights, maybe 00:15:57.720 --> 00:16:00.100 sparked a few aha moments along the way. Thank 00:16:00.100 --> 00:16:02.529 you so much for joining us. Until next time, 00:16:02.570 --> 00:16:04.629 keep exploring Aotearoa music.