1 00:00:00,000 --> 00:00:04,720 Okay, get this. Imagine robots, but not the metal clunky ones you see in the movies. 2 00:00:04,960 --> 00:00:07,800 We're talking robots built from living cells. 3 00:00:07,840 --> 00:00:12,360 It's wild, right? This is the future of AI we're diving into today where biology and 4 00:00:12,360 --> 00:00:16,040 technology gets seriously weird in a good way, of course. 5 00:00:16,040 --> 00:00:20,600 Seriously cool stuff. And we're drawing from some pretty awesome sources today. 6 00:00:21,000 --> 00:00:25,880 Keynotes and presentations from CVPR 2024, a huge computer vision conference. 7 00:00:25,880 --> 00:00:31,280 Yeah, some of the brightest minds in AI and robotics are there pushing the boundaries of what's possible. 8 00:00:31,320 --> 00:00:36,440 So our mission today to break down this idea of embodied AI, we want to understand what 9 00:00:36,440 --> 00:00:40,080 happens when we take AI out of the digital world and give it a physical form. 10 00:00:40,200 --> 00:00:42,800 And what that really means, because it's a game changer. 11 00:00:42,840 --> 00:00:45,440 Right now, AI is mostly stuck in the digital world, right? 12 00:00:45,480 --> 00:00:48,920 Like, chat GPT can write a poem, but it can't make you a cup of tea, right? 13 00:00:49,120 --> 00:00:52,480 Exactly. And it's not just tea making skills that are missing. 14 00:00:52,480 --> 00:00:57,480 AI struggles with basic things we take for granted, common sense, understanding cause and effect. 15 00:00:57,520 --> 00:01:02,880 Because it has no hands to actually make the tea or experience those cause and effect moments for itself. 16 00:01:02,880 --> 00:01:07,760 Precisely. It lacks what we call embodiment, a physical presence in the world. 17 00:01:08,080 --> 00:01:10,400 And that's a huge hurdle for AI right now. 18 00:01:10,560 --> 00:01:14,680 So how do we fix that? Do we give AI a crash course in being human? 19 00:01:14,960 --> 00:01:20,800 In a way, yes. Researchers are exploring how to embody AI by actually giving it a physical form, 20 00:01:20,800 --> 00:01:22,400 something to interact with the world. 21 00:01:22,400 --> 00:01:25,840 OK, so we're talking robots then, but not like the clunky factory robots. 22 00:01:25,840 --> 00:01:29,120 That's quite right. We're talking robots built from dot frog cells. 23 00:01:29,120 --> 00:01:30,760 Frog cells? Wait, are you serious? 24 00:01:30,760 --> 00:01:34,000 Totally. These xenobots are causing a stir in the research world. 25 00:01:34,280 --> 00:01:39,200 They're living, moving organisms designed by AI and built using actual frog cells. 26 00:01:39,320 --> 00:01:40,920 OK, hold on. Back up a bit. 27 00:01:41,040 --> 00:01:44,120 We're building robots from scratch at the cellular level. 28 00:01:44,120 --> 00:01:46,800 Yep. And it's even more mind blowing than it sounds. 29 00:01:46,800 --> 00:01:52,080 Imagine the structural properties of skin cells combined with the movement of heart muscle cells. 30 00:01:52,320 --> 00:01:54,000 That's basically what makes up a xenobot. 31 00:01:54,000 --> 00:01:57,480 So tiny pulsating blobs of frog cells. 32 00:01:57,480 --> 00:02:00,400 Essentially, yeah. But here's where it gets really cool. 33 00:02:00,600 --> 00:02:07,440 Researchers use AI, specifically a type called a genetic algorithm, to actually design these xenobots. 34 00:02:07,440 --> 00:02:11,280 Genetic algorithm. So like how Spotify knows what music I like? 35 00:02:11,280 --> 00:02:13,960 Not exactly the same, but you've got the right idea. 36 00:02:13,960 --> 00:02:17,480 Algorithms can learn and adapt. This one's inspired by natural selection. 37 00:02:17,480 --> 00:02:22,800 It tests thousands of virtual xenobots searching for designs that allow them to move in specific ways. 38 00:02:22,800 --> 00:02:25,560 So the AI is like the architect drawing up the blueprint? 39 00:02:25,560 --> 00:02:33,920 Exactly. Then scientists take those blueprints and use microscopic tools to arrange these tiny cells, according to the AI's design. 40 00:02:33,920 --> 00:02:37,440 That's insane. It's like performing surgery on a grain of sand. 41 00:02:37,440 --> 00:02:39,640 It's incredibly precise work. 42 00:02:38,620 --> 00:02:48,360 microscopic lab experiments. They can actually do some pretty amazing things. 43 00:02:48,360 --> 00:02:51,640 Okay, like what? They're moving around, but what else can they actually do? 44 00:02:51,640 --> 00:02:56,600 Well, for one, they can move in different ways. Walking, swimming, even working together in groups. 45 00:02:56,600 --> 00:03:00,680 So it's not random. The AI is actually designing them for specific movements. 46 00:03:00,680 --> 00:03:02,920 Exactly. And it goes beyond just movement. 47 00:03:02,920 --> 00:03:06,880 They've been observed pushing small objects, almost like miniature cleaners. 48 00:03:06,880 --> 00:03:10,840 And get this, some have even shown a limited form of self-replication. 49 00:03:10,840 --> 00:03:16,360 Wait, self-replication like making copies of themselves, microscopic robot factories. 50 00:03:16,360 --> 00:03:23,720 In a way, yeah. But what's even crazier is they're doing all of this, sensing, reacting to their environment without brains or nervous systems. 51 00:03:23,720 --> 00:03:27,400 Like what even defines sensing in these things? It's blowing our minds. 52 00:03:27,400 --> 00:03:32,040 Okay, that's seriously cool. Tiny self-assembling, self-replicating robots. 53 00:03:32,040 --> 00:03:38,680 So we've got these microscopic robots cleaning, building, maybe even starting a tiny robot revolution. 54 00:03:38,680 --> 00:03:42,920 Who knows? But let's zoom out for a second. Why does all of this even matter? 55 00:03:42,920 --> 00:03:45,520 Why should we care about these little frog cell robots? 56 00:03:45,520 --> 00:03:50,200 That's the question, right? It's because xenobots represent something much bigger. 57 00:03:50,200 --> 00:03:53,000 They're pushing the boundaries of what we thought robots could be. 58 00:03:53,000 --> 00:03:54,360 Okay, I'm listening. 59 00:03:54,360 --> 00:04:01,280 Think about it. Microscopic robots delivering drugs with incredible accuracy, cleaning up pollution, 60 00:04:01,280 --> 00:04:04,960 repairing things from the inside out. That's the potential here. 61 00:04:04,960 --> 00:04:08,760 Sign me up for a Biobot Repair Crew. My niece could use some of that. 62 00:04:08,760 --> 00:04:12,200 We're not quite there yet. But you see what I mean. This isn't sci-fi anymore. 63 00:04:12,200 --> 00:04:18,760 We're talking about using biology to create completely new solutions to big problems. 64 00:04:18,760 --> 00:04:25,800 Okay, yeah, the potential benefits are huge. But this is also giving me some like black mirror vibes, you know? 65 00:04:25,800 --> 00:04:28,240 What about the downsides? We have to be thinking about that, right? 66 00:04:28,240 --> 00:04:33,840 Absolutely. Every new technology comes with risks. Making sure these biobots are used responsibly. 67 00:04:33,840 --> 00:04:38,240 That they don't accidentally mess up the environment. These are all things researchers are working on. 68 00:04:38,240 --> 00:04:45,320 Right. So it's like, how do we unlock this crazy potential without unleashing a swarm of microscopic terrors on the world? 69 00:04:45,320 --> 00:04:48,400 Exactly. It's about being smart and responsible about it. 70 00:04:48,400 --> 00:04:53,040 Asking those tough ethical questions right alongside the exciting scientific ones. 71 00:04:53,040 --> 00:04:58,600 Totally. But before we get too deep into the ethics, I want to shift gears for a bit. 72 00:04:58,600 --> 00:05:05,760 Let's talk about another area where AI is blending the digital and physical AI-generated art. 73 00:05:05,760 --> 00:05:11,920 Right. Where the algorithms become the artists. It's a perfect example of this embodiment idea we've been discussing. 74 00:05:11,920 --> 00:05:18,760 Really? Because when I think AI art, I think those cool trippy images online, not something physical like a robot. 75 00:05:18,760 --> 00:05:26,440 That's the thing. Think of it like this. AI art takes something purely digital, an algorithm, and gives it a voice in the physical world. 76 00:05:26,440 --> 00:05:30,240 Through a painting, a sculpture, something we can experience with our senses. 77 00:05:30,240 --> 00:05:35,200 Oh, OK. I get it. It's like AI is using art to interact with the real world, even if it's through a screen. 78 00:05:35,200 --> 00:05:40,200 Exactly. And there's this artist, Sofia Crespo, who's doing some really incredible work in this space. 79 00:05:40,200 --> 00:05:42,800 The Sofia Crespo. Wait, I think I've heard that name. 80 00:05:42,800 --> 00:05:51,160 She's known for these stunning, almost surreal images of nature, plants, animals, insects, all generated using AI. 81 00:05:51,160 --> 00:05:58,880 OK, no, I definitely need to see this. So how does she do it? Is she like showing the AI a bunch of nature shows and saying, OK, now make me art? 82 00:05:58,880 --> 00:06:05,680 It's more complex than that. She trains these AI models on tons of images, thousands, even millions, 83 00:06:05,680 --> 00:06:10,720 then uses these advanced techniques like neural style transfer to create something totally new. 84 00:06:10,720 --> 00:06:14,200 Neural style transfer? That sounds intense. What's happening there? 85 00:06:14,200 --> 00:06:20,520 Imagine blending the style of, say, a Van Gogh painting with those thick brushstrokes with a photograph of the sunflower. 86 00:06:20,520 --> 00:06:25,040 That's kind of what neural style transfer does. It merges the style of one image with the content of another. 87 00:06:25,040 --> 00:06:29,840 So the AI is remixing images, making something totally new from existing parts. That's wild. 88 00:06:29,840 --> 00:06:33,880 But you were saying Sofia Crespo uses this to create art inspired by nature. 89 00:06:33,880 --> 00:06:36,920 Yes. And it goes beyond just making cool pictures. 90 00:06:36,920 --> 00:06:43,240 She's trying to raise awareness about biodiversity loss, the impact we're having on the natural world. 91 00:06:43,240 --> 00:06:46,480 She's using her AI art to make a statement. 92 00:06:46,480 --> 00:06:50,080 I love that art as activism. How does she actually do that, though? 93 00:06:50,080 --> 00:06:53,920 Is it like sad images of endangered species or something? 94 00:06:53,920 --> 00:06:57,080 One of her most interesting projects is called Neural Zoo. 95 00:06:57,080 --> 00:07:04,480 It's this collection of these bizarre, fantastical creatures, almost like out of a medieval book, but with this modern digital twist. 96 00:07:04,480 --> 00:07:09,360 Medieval creatures meet AI. OK, I'm into it. What kind of modern twist are we talking? 97 00:07:09,360 --> 00:07:14,480 Like a beetle with wings like stained glass or a fish with scales that shimmer like opals. 98 00:07:14,480 --> 00:07:17,440 They're both familiar and completely alien at the same time. 99 00:07:17,440 --> 00:07:23,040 OK, yeah, that's definitely a modern twist. It's like taking these ancient myths and reimagining them for the digital age. 100 00:07:23,040 --> 00:07:25,640 But how does this raise awareness about biodiversity? 101 00:07:25,640 --> 00:07:32,720 She intentionally uses incomplete data when training the AI, especially with endangered species. 102 00:07:32,720 --> 00:07:37,160 It's about highlighting how much we don't know, how many pieces of the puzzle are missing. 103 00:07:37,160 --> 00:07:42,360 Wow, that's clever. By showing us what we don't know, she's making us realize what we stand to lose. 104 00:07:42,360 --> 00:07:45,360 Exactly. And it's a message that's resonating. 105 00:07:45,360 --> 00:07:52,640 Her work has been shown all over the world, getting people talking about biodiversity and how technology can help us understand and protect our planet. 106 00:07:52,640 --> 00:07:55,920 That's awesome. Sophia Crespo is definitely want to watch. 107 00:07:55,920 --> 00:07:59,360 But you said her work ties back to the robotics stuff we talked about earlier. 108 00:07:59,360 --> 00:08:03,320 It all comes down to embodiment, those lifelike creatures she creates. 109 00:08:03,320 --> 00:08:08,800 It's like she's exploring what it means for AI to have a physical presence, but through art. 110 00:08:08,800 --> 00:08:15,680 Right, like she's bridging the gap between the digital and physical, helping us connect with these creatures, even if they only exist digitally. 111 00:08:15,680 --> 00:08:22,120 And just like those frag cell robots are changing what robots can be, her work is changing what art can be. 112 00:08:22,120 --> 00:08:25,480 It's blurring the lines between creation and simulation. 113 00:08:25,480 --> 00:08:30,680 It's deep stuff. Like, what does it even mean to create life, even if it's artificial? 114 00:08:30,680 --> 00:08:36,440 Definitely makes you think. And it shows how these advances in AI aren't just changing what's possible. 115 00:08:36,440 --> 00:08:44,720 They're changing how we see ourselves and our place in this increasingly blurry world where the real and the virtual are practically inseparable. 116 00:08:44,720 --> 00:08:47,120 Whoa, that's a lot to process. 117 00:08:47,120 --> 00:08:51,600 But speaking of pushing boundaries, what about those protein powered nanobots we touched on earlier? 118 00:08:51,600 --> 00:08:54,720 I'm dying to know more about AI designing machines at that scale. 119 00:08:54,720 --> 00:09:02,440 Well, just like it can design a xenobot or create a piece of art, AI can also design the very building blocks of life itself, proteins. 120 00:09:02,440 --> 00:09:07,600 Wait, proteins? Like the stuff that builds our muscles and, well, everything. AI can design those now. 121 00:09:07,600 --> 00:09:13,400 It's true. And this could change everything from medicine to how we build things. 122 00:09:13,400 --> 00:09:16,960 It's a whole new world of protein engineering powered by AI. 123 00:09:16,960 --> 00:09:23,840 OK, color me intrigued. But how does it even work? Does the AI put on a little lab coat and start mixing chemicals? 124 00:09:23,840 --> 00:09:30,480 Not exactly. It's about understanding how proteins work, those intricate structures that determine what they do. 125 00:09:30,480 --> 00:09:33,200 It's like AI is learning to read the recipe book of life. 126 00:09:33,200 --> 00:09:38,320 So instead of baking a cake, it's baking proteins. What kind of protein recipes are we talking about? 127 00:09:38,320 --> 00:09:45,880 Some pretty incredible stuff. For example, AI is being used to design new proteins that could be like super targeted drugs, 128 00:09:45,880 --> 00:09:49,600 delivering treatments right where they're needed without messing with healthy cells. 129 00:09:49,600 --> 00:09:53,040 Microscopic smart bombs taking out the bad guys. That's awesome. 130 00:09:53,040 --> 00:09:59,360 Exactly. That's just the start. We're talking about more effective vaccines, longer-lasting, easier to get to people. 131 00:09:59,360 --> 00:10:00,640 Huge implications. 132 00:10:00,640 --> 00:10:05,120 OK, yeah, that's a game changer, especially after everything that's happened recently. 133 00:10:05,120 --> 00:10:09,320 But you mentioned sustainability, too. How do proteins play into that? 134 00:10:09,320 --> 00:10:14,160 Enzymes. They're like natural catalysts speeding up chemical reactions. 135 00:10:14,160 --> 00:10:18,360 And AI can design these enzymes to do some pretty incredible stuff. 136 00:10:18,360 --> 00:10:19,720 OK, like what? 137 00:10:19,720 --> 00:10:28,320 We're talking about breaking down pollution, capturing carbon dioxide from the air, even creating clean energy, all using specially designed enzymes. 138 00:10:28,320 --> 00:10:32,120 So AI is designing tiny little environmental heroes. That's awesome. 139 00:10:32,120 --> 00:10:37,560 It's a perfect example of how we can use AI to tackle those huge global challenges we're facing. 140 00:10:37,560 --> 00:10:43,880 It's mind blowing when you really think about it. AI designing robots, art, proteins. The future is getting wild. 141 00:10:43,880 --> 00:10:47,000 It really is. But and I know we keep coming back to this. 142 00:10:47,000 --> 00:10:50,520 We have to remember that with every powerful tool, we need to be careful. 143 00:10:50,520 --> 00:10:54,320 There are downsides, ethical things to consider. It's not all sunshine and roses. 144 00:10:54,320 --> 00:10:55,600 Right, of course. 145 00:10:55,600 --> 00:11:00,760 What are some of the big questions we need to be asking as we move further into this AI driven world? 146 00:11:00,760 --> 00:11:05,760 A big one is making sure this tech is used responsibly and that everyone benefits. 147 00:11:05,760 --> 00:11:11,400 We don't want to worsen existing inequalities or leave anyone behind. Equity and access are huge. 148 00:11:11,400 --> 00:11:17,640 Yeah, absolutely. We don't need an AI powered world that only works for a select few. What else? 149 00:11:17,640 --> 00:11:20,960 I think the biggest thing is that we need to keep having these conversations. 150 00:11:20,960 --> 00:11:25,040 The ethical questions around AI are only going to get more complex as it evolves. 151 00:11:25,040 --> 00:11:28,840 So it's not just about what AI can do. It's about what it should do. 152 00:11:28,840 --> 00:11:37,200 Exactly. It's about making sure these incredibly powerful tools align with our values, that we're using them to build the future we actually want for everyone. 153 00:11:37,200 --> 00:11:39,160 Couldn't agree more. 154 00:11:39,160 --> 00:11:45,000 As we wrap up this incredible look at AI, what's the one big takeaway you hope listeners are left with? 155 00:11:45,000 --> 00:11:48,800 I think the main thing is this. AI isn't some far off thing anymore. 156 00:11:48,800 --> 00:11:55,840 It's here, it's now, and it's changing everything. How we live, work, interact with the world, even what it means to be human. 157 00:11:55,840 --> 00:12:03,400 And as we've seen today, from those frog cell robots to AI designed proteins, these changes are happening in ways we never could have imagined, 158 00:12:03,400 --> 00:12:07,880 blurring the lines between the digital and physical, the natural and artificial. 159 00:12:07,880 --> 00:12:13,440 And as we explore this uncharted territory, it's on us to do it responsibly, ethically, 160 00:12:13,440 --> 00:12:18,400 with a focus on using these incredible technologies to make the world a better place for everyone. 161 00:12:18,400 --> 00:12:22,720 Well said. Thanks for joining me on this wild ride into the world of AI. 162 00:12:22,720 --> 00:12:39,320 And to everyone listening, stay curious, keep asking those tough questions, and let's build the future we want to see together.