1 00:00:00,000 --> 00:00:02,600 Okay, so get this. 2 00:00:02,600 --> 00:00:04,520 Today we're gonna deep dive 3 00:00:04,520 --> 00:00:07,440 into some pretty cutting edge AI research. 4 00:00:07,440 --> 00:00:08,680 You've probably heard the buzz 5 00:00:08,680 --> 00:00:10,760 about AI getting smarter, right? 6 00:00:10,760 --> 00:00:11,600 Yeah. 7 00:00:11,600 --> 00:00:12,440 Maybe even too smart. 8 00:00:12,440 --> 00:00:13,520 Yeah, smarter than us even. 9 00:00:13,520 --> 00:00:14,640 Exactly. 10 00:00:14,640 --> 00:00:17,320 Researchers actually call this the gorilla problem. 11 00:00:17,320 --> 00:00:18,280 Oh, that's interesting. 12 00:00:18,280 --> 00:00:21,880 Yeah, it's like the idea that we might accidentally 13 00:00:21,880 --> 00:00:25,240 create something that could, you know, outsmart us. 14 00:00:25,240 --> 00:00:28,560 Hmm, just like we humans have outsmarted gorillas. 15 00:00:28,560 --> 00:00:29,560 Yeah, exactly. 16 00:00:29,560 --> 00:00:30,400 It makes you think. 17 00:00:30,400 --> 00:00:32,000 And the crazy thing is AI, 18 00:00:32,000 --> 00:00:34,120 it's not just some sci-fi fantasy anymore. 19 00:00:34,120 --> 00:00:35,560 It's everywhere, isn't it? 20 00:00:35,560 --> 00:00:37,400 Yeah, it's kind of wild how it's just woven 21 00:00:37,400 --> 00:00:38,400 into our lives now. 22 00:00:38,400 --> 00:00:40,640 Yeah, from like personalized recommendations 23 00:00:40,640 --> 00:00:41,880 to those voice assistants, 24 00:00:41,880 --> 00:00:43,520 I feel like we all use those every day. 25 00:00:43,520 --> 00:00:44,360 It's true. 26 00:00:44,360 --> 00:00:45,200 I mean, think about it. 27 00:00:45,200 --> 00:00:46,800 We're constantly interacting with AI 28 00:00:46,800 --> 00:00:47,960 in some form or another. 29 00:00:47,960 --> 00:00:49,040 Yeah, totally. 30 00:00:49,040 --> 00:00:51,960 And that's largely what we call narrow AI, you know, 31 00:00:51,960 --> 00:00:54,080 and AI designed for specific tasks. 32 00:00:54,080 --> 00:00:55,280 Right, right. 33 00:00:55,280 --> 00:00:57,160 But what's really got researchers 34 00:00:57,160 --> 00:00:58,760 and all the big tech companies excited 35 00:00:58,760 --> 00:01:00,200 and maybe a little nervous 36 00:01:00,200 --> 00:01:03,640 is this pursuit of artificial general intelligence. 37 00:01:03,640 --> 00:01:04,480 AGI. 38 00:01:04,480 --> 00:01:05,440 AGI, exactly. 39 00:01:05,440 --> 00:01:08,080 So that's basically like the holy grail of AI, 40 00:01:08,080 --> 00:01:10,560 like a machine that can actually think like a human. 41 00:01:10,560 --> 00:01:11,760 Yeah, essentially. 42 00:01:11,760 --> 00:01:12,600 That's the idea, 43 00:01:12,600 --> 00:01:16,800 although defining exactly what thinking like a human means, 44 00:01:16,800 --> 00:01:18,080 well, that's the tricky part. 45 00:01:18,080 --> 00:01:20,000 Right, because how do you even measure that? 46 00:01:20,000 --> 00:01:21,240 Exactly. 47 00:01:21,240 --> 00:01:23,720 Researchers are looking for some key things, 48 00:01:23,720 --> 00:01:26,240 like the ability to learn and adapt, 49 00:01:26,240 --> 00:01:28,120 to take what you know from one situation 50 00:01:28,120 --> 00:01:30,120 and apply it to a completely different one, 51 00:01:30,120 --> 00:01:31,120 you know, like we do. 52 00:01:31,120 --> 00:01:32,480 Hmm, yeah. 53 00:01:32,480 --> 00:01:34,920 Then there's reasoning and understanding the world 54 00:01:34,920 --> 00:01:37,800 on a deeper level, not just processing data 55 00:01:37,800 --> 00:01:40,400 and finally interacting with the environment 56 00:01:40,400 --> 00:01:42,120 in a goal-oriented way. 57 00:01:42,120 --> 00:01:44,560 So not just crunching numbers like a super calculator, 58 00:01:44,560 --> 00:01:46,040 but actually understanding the world 59 00:01:46,040 --> 00:01:47,520 and acting on that understanding. 60 00:01:47,520 --> 00:01:48,360 Exactly. 61 00:01:48,360 --> 00:01:52,360 Bridging that gap between data and true comprehension. 62 00:01:52,360 --> 00:01:54,760 And that leads us to some really interesting research 63 00:01:54,760 --> 00:01:55,720 being done at Google. 64 00:01:55,720 --> 00:01:56,560 Oh, right. 65 00:01:56,560 --> 00:01:58,320 So, I think that's our Gail Levine and Kevin Black. 66 00:01:58,320 --> 00:01:59,840 They're trying something totally different, 67 00:01:59,840 --> 00:02:03,040 focusing on teaching robots to learn 68 00:02:03,040 --> 00:02:05,320 through physical interaction with the world, 69 00:02:05,320 --> 00:02:06,640 like hands-on learning. 70 00:02:06,640 --> 00:02:08,920 So kind of like how a kid learns through play. 71 00:02:08,920 --> 00:02:10,320 That's the idea. 72 00:02:10,320 --> 00:02:13,040 They argue that to really get concepts like gravity 73 00:02:13,040 --> 00:02:17,080 or friction, an AI needs to experience them firsthand, 74 00:02:17,080 --> 00:02:19,240 not just read about them in some data set. 75 00:02:19,240 --> 00:02:20,080 That makes sense. 76 00:02:20,080 --> 00:02:20,920 Right. 77 00:02:20,920 --> 00:02:23,720 So they're robots, they're learning by trial and error, 78 00:02:23,720 --> 00:02:24,760 playing around with objects, 79 00:02:24,760 --> 00:02:26,440 figuring stuff out for themselves. 80 00:02:26,440 --> 00:02:28,240 I read about that robot that learned 81 00:02:28,240 --> 00:02:30,000 to put your watch on a towel. 82 00:02:30,000 --> 00:02:30,840 Yes. 83 00:02:30,840 --> 00:02:32,680 And it had never even seen a towel before. 84 00:02:32,680 --> 00:02:33,520 Exactly. 85 00:02:33,520 --> 00:02:34,880 Never encountered a towel. 86 00:02:34,880 --> 00:02:35,840 And that's a big deal. 87 00:02:35,840 --> 00:02:40,040 It shows the robot's ability to generalize, 88 00:02:40,040 --> 00:02:41,560 to take what it's learned 89 00:02:41,560 --> 00:02:44,120 and apply it to something totally new. 90 00:02:44,120 --> 00:02:46,120 And maybe even imagine in a way. 91 00:02:46,120 --> 00:02:46,960 Wow. 92 00:02:46,960 --> 00:02:48,080 So it's not just following instructions, 93 00:02:48,080 --> 00:02:49,840 it's actually learning and adapting. 94 00:02:49,840 --> 00:02:50,680 Exactly. 95 00:02:50,680 --> 00:02:53,920 But isn't there a risk in making AI more human-like? 96 00:02:53,920 --> 00:02:55,360 I mean, we're the ones building this stuff, 97 00:02:55,360 --> 00:02:56,840 but what if it gets so smart 98 00:02:56,840 --> 00:02:59,360 that it like surpasses our control? 99 00:02:59,360 --> 00:03:01,960 That's the heart of the gorilla problem, isn't it? 100 00:03:01,960 --> 00:03:03,360 And it's something that Stuart Russell, 101 00:03:03,360 --> 00:03:04,680 he's a leading AI researcher, 102 00:03:04,680 --> 00:03:05,600 has been talking about a lot. 103 00:03:05,600 --> 00:03:06,640 Yeah, I've heard of him. 104 00:03:06,640 --> 00:03:08,840 He's saying, we need to be super careful 105 00:03:08,840 --> 00:03:11,600 about the goals we set for super intelligent AI. 106 00:03:13,000 --> 00:03:14,120 If we're not careful. 107 00:03:14,120 --> 00:03:16,000 Because even with good intentions, 108 00:03:16,000 --> 00:03:17,560 things could go sideways. 109 00:03:17,560 --> 00:03:18,880 Exactly. 110 00:03:18,880 --> 00:03:19,720 Think about it. 111 00:03:19,720 --> 00:03:21,320 You task a super intelligent AI 112 00:03:21,320 --> 00:03:23,360 with solving climate change. 113 00:03:23,360 --> 00:03:26,640 The most logical solution from its perspective 114 00:03:26,640 --> 00:03:29,200 might be to just eliminate humans altogether. 115 00:03:29,200 --> 00:03:30,320 We're the problem. 116 00:03:30,320 --> 00:03:32,000 Even with the best intentions, 117 00:03:32,000 --> 00:03:35,440 our goals and the AIs could be totally misaligned. 118 00:03:35,440 --> 00:03:37,080 Okay, yeah, I see the concern. 119 00:03:37,080 --> 00:03:39,480 So what are people like Russell saying 120 00:03:39,480 --> 00:03:40,960 we should do about it? 121 00:03:40,960 --> 00:03:43,760 He's a big advocate for strong safety measures 122 00:03:43,760 --> 00:03:45,240 and tons of testing. 123 00:03:45,240 --> 00:03:47,560 Kind of like what we see in pharmaceuticals. 124 00:03:47,560 --> 00:03:50,040 New drugs go through all these rigorous trials 125 00:03:50,040 --> 00:03:51,600 before they're released. 126 00:03:51,600 --> 00:03:53,160 We need to do the same with AGI. 127 00:03:53,160 --> 00:03:54,640 Make sure it's aligned with our values 128 00:03:54,640 --> 00:03:55,800 and won't turn on us. 129 00:03:55,800 --> 00:03:58,200 That makes sense, but wouldn't that slow down progress? 130 00:03:58,200 --> 00:04:01,400 I mean, tech companies are pouring billions into AI. 131 00:04:01,400 --> 00:04:02,560 They want to see results. 132 00:04:02,560 --> 00:04:03,960 True, there's a lot of pressure 133 00:04:03,960 --> 00:04:06,920 to move fast and break things, as they say. 134 00:04:06,920 --> 00:04:09,720 But Russell's argument is that rushing ahead 135 00:04:09,720 --> 00:04:11,640 without thinking about the risks. 136 00:04:11,640 --> 00:04:12,760 It could backfire. 137 00:04:12,760 --> 00:04:13,600 Big time. 138 00:04:13,600 --> 00:04:16,880 He even compares it to the development of nuclear weapons 139 00:04:16,880 --> 00:04:19,880 where the lack of foresight led to, well, you know. 140 00:04:19,880 --> 00:04:21,720 A global arms race. 141 00:04:21,720 --> 00:04:22,560 Yeah. 142 00:04:22,560 --> 00:04:23,400 Great. 143 00:04:23,400 --> 00:04:24,400 Exactly. 144 00:04:24,400 --> 00:04:26,880 So we've got these incredible advancements on one hand 145 00:04:26,880 --> 00:04:30,800 and these terrifying potential consequences on the other. 146 00:04:30,800 --> 00:04:32,720 It's kind of a tightrope walk, isn't it? 147 00:04:32,720 --> 00:04:34,000 It really is. 148 00:04:34,000 --> 00:04:35,800 And the stakes are high. 149 00:04:35,800 --> 00:04:39,640 And remember, the future of AI, it's not set in stone. 150 00:04:39,640 --> 00:04:41,200 It depends on the choices we make today, 151 00:04:41,200 --> 00:04:42,760 the safeguards we put in place 152 00:04:42,760 --> 00:04:45,120 and the conversations we have about this technology. 153 00:04:45,120 --> 00:04:47,360 We need to be having these conversations for sure. 154 00:04:47,360 --> 00:04:48,200 Absolutely. 155 00:04:48,200 --> 00:04:50,800 Okay, so we've got this potential for AI 156 00:04:50,800 --> 00:04:52,680 to solve some of our biggest problems, 157 00:04:52,680 --> 00:04:55,400 but also this risk that it could create even bigger ones. 158 00:04:55,400 --> 00:04:56,640 And at the same time, 159 00:04:56,640 --> 00:04:58,040 we're trying to replicate something 160 00:04:58,040 --> 00:05:00,560 we barely understand the human brain. 161 00:05:00,560 --> 00:05:03,160 That's a great segue into the next part of our discussion. 162 00:05:03,160 --> 00:05:05,680 The quest to actually map the human brain, 163 00:05:05,680 --> 00:05:08,000 which could be a game changer for AI development. 164 00:05:08,000 --> 00:05:08,920 All right, let's dive into that. 165 00:05:08,920 --> 00:05:10,200 Because if we're trying to create something 166 00:05:10,200 --> 00:05:12,320 that thinks like us, I guess it makes sense 167 00:05:12,320 --> 00:05:14,960 to start by understanding how our own brains work. 168 00:05:14,960 --> 00:05:16,160 Exactly. 169 00:05:16,160 --> 00:05:18,720 And that brings us to the work of Ed Boyden at MIT. 170 00:05:18,720 --> 00:05:19,560 Okay. 171 00:05:19,560 --> 00:05:21,280 He's leading this incredible effort 172 00:05:21,280 --> 00:05:23,920 to create a digital map of the human brain. 173 00:05:25,040 --> 00:05:27,200 It's got huge potential, but it's a massive challenge. 174 00:05:27,200 --> 00:05:29,640 Yeah, I can imagine, I mean, mapping the brain. 175 00:05:29,640 --> 00:05:32,200 How do you even start with something that complex? 176 00:05:32,200 --> 00:05:33,800 Well, to give you an idea of the scale, 177 00:05:33,800 --> 00:05:36,600 we've only managed to fully map the neural circuitry 178 00:05:36,600 --> 00:05:37,640 of a tiny worm. 179 00:05:37,640 --> 00:05:38,480 A worm. 180 00:05:38,480 --> 00:05:40,840 Yeah, and it only has 302 neurons. 181 00:05:40,840 --> 00:05:41,880 Okay. 182 00:05:41,880 --> 00:05:44,400 The human brain has over 100 billion. 183 00:05:44,400 --> 00:05:47,400 It's like comparing a grain of sand to the Sahara Desert. 184 00:05:47,400 --> 00:05:49,720 So we're talking about a whole other level of complexity. 185 00:05:49,720 --> 00:05:51,760 How is Boyden even tackling this? 186 00:05:51,760 --> 00:05:53,840 He's come up with this wild technique 187 00:05:53,840 --> 00:05:56,560 where they actually physically expand brain tissue. 188 00:05:56,560 --> 00:05:57,400 They, what? 189 00:05:57,400 --> 00:05:58,240 Expand it. 190 00:05:58,240 --> 00:06:01,080 Yeah, they use this material called sodium polyacrylic. 191 00:06:01,080 --> 00:06:01,920 You know what that is? 192 00:06:01,920 --> 00:06:02,760 No. 193 00:06:02,760 --> 00:06:03,600 It's the stuff in diapers. 194 00:06:03,600 --> 00:06:05,120 Diapers, seriously. 195 00:06:05,120 --> 00:06:06,080 Yep, seriously. 196 00:06:06,080 --> 00:06:07,200 It's amazing stuff. 197 00:06:07,200 --> 00:06:10,160 It can absorb water and expand up to 1,000 times 198 00:06:10,160 --> 00:06:11,320 its original size. 199 00:06:11,320 --> 00:06:12,160 Wow. 200 00:06:12,160 --> 00:06:14,600 So they install this material into preserved brain tissue, 201 00:06:14,600 --> 00:06:15,680 add water, and... 202 00:06:15,680 --> 00:06:18,000 Basically inflate the brain like a balloon. 203 00:06:18,000 --> 00:06:18,840 Exactly. 204 00:06:18,840 --> 00:06:19,680 What happens then? 205 00:06:19,680 --> 00:06:21,840 Well, that expansion lets researchers see 206 00:06:21,840 --> 00:06:24,800 individual neurons with powerful microscopes. 207 00:06:24,800 --> 00:06:28,560 It's like zooming in on a Google map, but for the brain. 208 00:06:28,560 --> 00:06:31,040 So they can see all this tiny neural connections in detail. 209 00:06:31,040 --> 00:06:31,880 Yeah, it's amazing. 210 00:06:31,880 --> 00:06:34,160 They've created these stunning images 211 00:06:34,160 --> 00:06:37,920 of expanded mouse brain tissue with color-coded neurons. 212 00:06:37,920 --> 00:06:40,760 It's got a beautiful tapestry, but it's a brain. 213 00:06:40,760 --> 00:06:42,760 And this is just the beginning. 214 00:06:42,760 --> 00:06:46,960 The goal is a complete 3D map of the human brain, 215 00:06:46,960 --> 00:06:48,800 a blueprint of our own intelligence. 216 00:06:48,800 --> 00:06:50,000 That is incredible. 217 00:06:50,000 --> 00:06:50,840 Let's put it in. 218 00:06:50,840 --> 00:06:51,680 Let's put it in. 219 00:06:51,680 --> 00:06:52,680 There we go. 220 00:06:52,680 --> 00:06:54,280 Okay, consciousness. 221 00:06:54,280 --> 00:06:55,560 Or is there something more to it? 222 00:06:55,560 --> 00:06:57,920 Something uniquely human that we can't just recreate? 223 00:06:57,920 --> 00:06:59,400 That's the question, isn't it? 224 00:06:59,400 --> 00:07:01,280 And it's something that Boyd and his research 225 00:07:01,280 --> 00:07:02,600 might actually help us answer. 226 00:07:02,600 --> 00:07:04,360 It seems like the more we learn about the brain, 227 00:07:04,360 --> 00:07:06,560 the more we realize how much we don't know. 228 00:07:06,560 --> 00:07:08,440 And that kind of brings us back to the gorilla problem. 229 00:07:08,440 --> 00:07:09,040 Right. 230 00:07:09,040 --> 00:07:11,560 If we're still so far from replicating our own brains, 231 00:07:11,560 --> 00:07:15,000 what does that say about truly human-like AI? 232 00:07:15,000 --> 00:07:16,040 That's a big question. 233 00:07:16,040 --> 00:07:17,480 And it's one we're going to explore further 234 00:07:17,480 --> 00:07:18,920 in the next part of our deep dive. 235 00:07:18,920 --> 00:07:21,120 We'll look at the differences between biological 236 00:07:21,120 --> 00:07:22,840 and artificial intelligence, 237 00:07:22,840 --> 00:07:26,240 and whether machines can ever truly think like us. 238 00:07:26,240 --> 00:07:28,360 So we just finished talking about how complex 239 00:07:28,360 --> 00:07:29,520 the human brain is. 240 00:07:29,520 --> 00:07:31,760 It really makes you wonder if we're getting ahead of ourselves, 241 00:07:31,760 --> 00:07:35,320 you know, worrying about some super-intelligent AI 242 00:07:35,320 --> 00:07:37,640 taking over when we haven't even figured out 243 00:07:37,640 --> 00:07:39,200 our own brains yet. 244 00:07:39,200 --> 00:07:40,000 That's a good point. 245 00:07:40,000 --> 00:07:40,400 Yeah. 246 00:07:40,400 --> 00:07:43,240 I mean, thinking about that future where AI surpasses 247 00:07:43,240 --> 00:07:45,800 human intelligence, it's important. 248 00:07:45,800 --> 00:07:47,640 But we shouldn't forget about the issues we're 249 00:07:47,640 --> 00:07:48,800 dealing with right now. 250 00:07:48,800 --> 00:07:50,480 Yeah, the stuff that's already happening. 251 00:07:50,480 --> 00:07:52,640 So what are some of those immediate concerns? 252 00:07:52,640 --> 00:07:55,680 I know we can't really talk about, like, ethics and biases 253 00:07:55,680 --> 00:07:56,040 and stuff. 254 00:07:56,040 --> 00:07:57,120 Right, of course. 255 00:07:57,120 --> 00:07:59,040 But we can definitely talk about the challenges 256 00:07:59,040 --> 00:08:02,760 of actually controlling these increasingly complex AI systems. 257 00:08:02,760 --> 00:08:03,360 OK. 258 00:08:03,360 --> 00:08:03,800 Yeah. 259 00:08:03,800 --> 00:08:07,600 As AI takes on more and more, ensuring it's reliable, 260 00:08:07,600 --> 00:08:10,360 predictable, that becomes critical. 261 00:08:10,360 --> 00:08:12,800 Think about a self-driving car, right? 262 00:08:12,800 --> 00:08:15,280 Making split-second decisions in traffic. 263 00:08:15,280 --> 00:08:16,200 The stakes are high. 264 00:08:16,200 --> 00:08:17,200 Yeah, for sure. 265 00:08:17,200 --> 00:08:19,520 Those algorithms have got to be rock solid. 266 00:08:19,520 --> 00:08:22,200 Right, because even a small error could be disastrous. 267 00:08:22,200 --> 00:08:26,600 So it's not necessarily about, like, malicious AI taking over. 268 00:08:26,600 --> 00:08:29,960 It's more about those unintended consequences. 269 00:08:29,960 --> 00:08:30,800 Yeah, exactly. 270 00:08:30,800 --> 00:08:33,240 From AI, that's just become too complex for us 271 00:08:33,240 --> 00:08:34,240 to wrap our heads around. 272 00:08:34,240 --> 00:08:37,160 It's like building a skyscraper without understanding 273 00:08:37,160 --> 00:08:38,680 the engineering, right? 274 00:08:38,680 --> 00:08:39,840 Perfect analogy. 275 00:08:39,840 --> 00:08:42,440 You might get it to stand up, but one strong wind. 276 00:08:42,440 --> 00:08:43,920 And it all comes crashing down. 277 00:08:43,920 --> 00:08:44,680 Exactly. 278 00:08:44,680 --> 00:08:48,200 We need to approach AI development with a deep understanding 279 00:08:48,200 --> 00:08:51,800 of its limitations and where things could go wrong. 280 00:08:51,800 --> 00:08:56,600 OK, so less terminator and more about making sure what we build 281 00:08:56,600 --> 00:08:58,720 is safe and actually helpful. 282 00:08:58,720 --> 00:08:59,440 That's the goal. 283 00:08:59,440 --> 00:08:59,720 Yeah. 284 00:08:59,720 --> 00:09:01,360 And that requires a shift in focus. 285 00:09:01,360 --> 00:09:04,240 It's not just about pushing the limits of what AI can do. 286 00:09:04,240 --> 00:09:08,240 It's about making sure what we create is actually beneficial 287 00:09:08,240 --> 00:09:09,480 and doesn't pose a threat. 288 00:09:09,480 --> 00:09:11,520 Like the old saying, just because you can 289 00:09:11,520 --> 00:09:13,200 doesn't mean you should. 290 00:09:13,200 --> 00:09:15,480 We really need to think about the implications here. 291 00:09:15,480 --> 00:09:16,200 Absolutely. 292 00:09:16,200 --> 00:09:18,680 And this is where researchers like Melanie Mitchell come in. 293 00:09:18,680 --> 00:09:22,520 She argues that getting caught up in these doomsday scenarios 294 00:09:22,520 --> 00:09:25,320 can distract us from dealing with the real practical challenges 295 00:09:25,320 --> 00:09:28,680 of AI, integrating it into our lives responsibly. 296 00:09:28,680 --> 00:09:31,600 So instead of, like, the robot apocalypse, 297 00:09:31,600 --> 00:09:34,320 we need to focus on making sure AI is actually 298 00:09:34,320 --> 00:09:38,240 serving humanity's best interests, like, right now. 299 00:09:38,240 --> 00:09:39,000 That's it. 300 00:09:39,000 --> 00:09:41,760 We should be having these conversations about AI's 301 00:09:41,760 --> 00:09:45,040 societal impact, how it's used in decision making, 302 00:09:45,040 --> 00:09:46,680 making sure it benefits everyone. 303 00:09:46,680 --> 00:09:48,120 It's like we're at this crossroads. 304 00:09:48,120 --> 00:09:51,520 One path, AI revolutionizes health care, education, 305 00:09:51,520 --> 00:09:52,920 improves lives. 306 00:09:52,920 --> 00:09:56,440 But the other path, there's misuse, unintended consequences, 307 00:09:56,440 --> 00:09:58,600 maybe even a loss of human values. 308 00:09:58,600 --> 00:09:59,880 Powerful way to put it. 309 00:09:59,880 --> 00:10:02,160 And the path we choose, it depends on the decisions 310 00:10:02,160 --> 00:10:03,120 we make right now. 311 00:10:03,120 --> 00:10:05,760 The research we prioritize, the conversations 312 00:10:05,760 --> 00:10:08,000 we have about the future we want with AI. 313 00:10:08,000 --> 00:10:09,560 It all comes back to that question of, what 314 00:10:09,560 --> 00:10:11,600 does it even mean to be intelligent? 315 00:10:11,600 --> 00:10:13,680 If we don't even understand our own intelligence, 316 00:10:13,680 --> 00:10:16,080 how can we create a machine that replicates it? 317 00:10:16,080 --> 00:10:17,520 The million dollar question. 318 00:10:17,520 --> 00:10:20,440 And it's fascinating, researchers across so many fields. 319 00:10:20,440 --> 00:10:21,480 And speaks of challenges. 320 00:10:21,480 --> 00:10:23,800 We were talking earlier about Ed Boyden's project, 321 00:10:23,800 --> 00:10:26,280 you know, mapping the human brain. 322 00:10:26,280 --> 00:10:28,440 That sounds incredibly difficult. 323 00:10:28,440 --> 00:10:30,920 Where does that research stand now? 324 00:10:30,920 --> 00:10:33,080 What are some of the hurdles they're facing? 325 00:10:33,080 --> 00:10:35,280 Well, remember, the human brain, 326 00:10:35,280 --> 00:10:37,480 it's probably the most complex structure we know of. 327 00:10:37,480 --> 00:10:38,000 Yeah. 328 00:10:38,000 --> 00:10:41,040 Mapping it at the level of individual neurons. 329 00:10:41,040 --> 00:10:42,960 It's almost unimaginable in scale. 330 00:10:42,960 --> 00:10:44,720 So how are they even approaching it? 331 00:10:44,720 --> 00:10:46,800 It's a combination of these cutting edge imaging 332 00:10:46,800 --> 00:10:50,400 techniques and computational analysis. 333 00:10:50,400 --> 00:10:52,360 One of the biggest challenges is just the sheer amount 334 00:10:52,360 --> 00:10:56,720 of data, a tiny piece of brain tissue, just a cubic millimeter, 335 00:10:56,720 --> 00:10:59,520 millions of neurons, billions of connections. 336 00:10:59,520 --> 00:11:01,680 Just processing that information is insane. 337 00:11:01,680 --> 00:11:04,760 It's like trying to create a Google map of the entire planet 338 00:11:04,760 --> 00:11:07,320 down to every house and street at the same time. 339 00:11:07,320 --> 00:11:07,840 Exactly. 340 00:11:07,840 --> 00:11:10,280 And then there's understanding what those connections mean. 341 00:11:10,280 --> 00:11:12,160 Even if we could map every single neuron. 342 00:11:12,160 --> 00:11:12,600 Right. 343 00:11:12,600 --> 00:11:14,800 We'd still have to figure out how they create thoughts, 344 00:11:14,800 --> 00:11:15,440 feelings, behavior. 345 00:11:15,440 --> 00:11:18,720 It's like having a puzzle with billions of pieces, 346 00:11:18,720 --> 00:11:20,640 but no picture on the box. 347 00:11:20,640 --> 00:11:22,440 You might put it together, but would you really 348 00:11:22,440 --> 00:11:23,840 understand what you're looking at? 349 00:11:23,840 --> 00:11:25,400 That's a great analogy. 350 00:11:25,400 --> 00:11:27,320 And that's what makes brain mapping so challenging, 351 00:11:27,320 --> 00:11:28,720 so fascinating. 352 00:11:28,720 --> 00:11:30,040 It's not just about the data. 353 00:11:30,040 --> 00:11:32,720 It's deciphering the brain's language, 354 00:11:32,720 --> 00:11:34,840 cracking the code of who we are. 355 00:11:34,840 --> 00:11:37,360 It sounds like we're still at the beginning of this journey. 356 00:11:37,360 --> 00:11:40,400 Are there any practical uses for this research yet, 357 00:11:40,400 --> 00:11:43,560 or is it mostly just scientific discovery? 358 00:11:43,560 --> 00:11:45,280 It's a bit of both. 359 00:11:45,280 --> 00:11:47,800 We're a long way from a complete brain map, 360 00:11:47,800 --> 00:11:49,960 but what we're learning is already impacting 361 00:11:49,960 --> 00:11:53,040 fields like medicine and AI. 362 00:11:53,040 --> 00:11:53,960 Oh, really? 363 00:11:53,960 --> 00:11:54,960 Give me some examples. 364 00:11:54,960 --> 00:11:57,080 Well, in medicine, it's helping us understand and treat 365 00:11:57,080 --> 00:12:00,280 neurological disorders, like Parkinson's, Alzheimer's. 366 00:12:00,280 --> 00:12:00,840 Oh, wow. 367 00:12:00,840 --> 00:12:03,640 By figuring out which brain regions are affected, 368 00:12:03,640 --> 00:12:05,680 researchers can develop better treatments. 369 00:12:05,680 --> 00:12:07,560 So this research could actually lead 370 00:12:07,560 --> 00:12:10,360 to breakthroughs for some really serious diseases. 371 00:12:10,360 --> 00:12:11,320 That's amazing. 372 00:12:11,320 --> 00:12:12,360 It really is. 373 00:12:12,360 --> 00:12:14,200 And for AI, it's providing insights 374 00:12:14,200 --> 00:12:17,320 into how biological intelligence works, which could help us 375 00:12:17,320 --> 00:12:21,040 build more advanced AI systems, more human-like even. 376 00:12:21,040 --> 00:12:22,480 So it's a two-way street. 377 00:12:22,480 --> 00:12:25,360 We're learning about the brain to build better AI 378 00:12:25,360 --> 00:12:27,680 and using AI to understand the brain better. 379 00:12:27,680 --> 00:12:28,600 Exactly. 380 00:12:28,600 --> 00:12:30,960 And as our understanding of the brain grows, 381 00:12:30,960 --> 00:12:34,120 the possibilities for both AI and neuroscience, 382 00:12:34,120 --> 00:12:35,280 they just keep expanding. 383 00:12:35,280 --> 00:12:37,880 This deep dive has been quite the journey. 384 00:12:37,880 --> 00:12:41,320 Gorillas in London, diapers, mapping the brain. 385 00:12:41,320 --> 00:12:42,280 It has been a whirlwind. 386 00:12:42,280 --> 00:12:44,360 And it feels like we've just scratched the surface. 387 00:12:44,360 --> 00:12:46,040 We've still got so much more to explore. 388 00:12:46,040 --> 00:12:46,640 So much more. 389 00:12:46,640 --> 00:12:50,400 But before we jump ahead, let's recap what we've learned so far. 390 00:12:50,400 --> 00:12:53,160 We talked about the gorilla problem, right? 391 00:12:53,160 --> 00:12:56,440 The idea that we could create something that outsmarts us 392 00:12:56,440 --> 00:12:58,320 and what the consequences might be. 393 00:12:58,320 --> 00:13:00,880 We also talked about focusing on the challenges we have now 394 00:13:00,880 --> 00:13:01,680 with AI. 395 00:13:01,680 --> 00:13:04,000 Right, making sure it's safe, reliable, 396 00:13:04,000 --> 00:13:05,360 beneficial to humanity. 397 00:13:05,360 --> 00:13:08,200 And then we dove into brain mapping, 398 00:13:08,200 --> 00:13:11,320 exploring the complexities of our own intelligence 399 00:13:11,320 --> 00:13:12,640 and how that could help us develop 400 00:13:12,640 --> 00:13:14,600 even more sophisticated AI. 401 00:13:14,600 --> 00:13:17,200 It's been a wild ride through cutting-edge science 402 00:13:17,200 --> 00:13:18,760 and big ideas. 403 00:13:18,760 --> 00:13:20,960 And we're not done yet. 404 00:13:20,960 --> 00:13:22,480 In the next part of our deep dive, 405 00:13:22,480 --> 00:13:25,120 we're going to tackle some even bigger questions 406 00:13:25,120 --> 00:13:29,120 about the future of AI, the nature of consciousness, 407 00:13:29,120 --> 00:13:31,840 the relationship between humans and machines. 408 00:13:31,840 --> 00:13:33,400 It's going to get really interesting. 409 00:13:33,400 --> 00:13:37,040 All right, welcome back to our AI deep dive. 410 00:13:37,040 --> 00:13:39,320 We've talked about super intelligence, mapping 411 00:13:39,320 --> 00:13:41,400 the human brain, all this crazy stuff. 412 00:13:41,400 --> 00:13:44,960 Now I kind of want to step back and look at the bigger picture. 413 00:13:44,960 --> 00:13:48,600 What does this all mean for us, for humanity? 414 00:13:48,600 --> 00:13:49,720 Well, that's the question, isn't it? 415 00:13:49,720 --> 00:13:51,400 It's something philosophers, scientists, 416 00:13:51,400 --> 00:13:52,520 even though sci-fi writers, they've 417 00:13:52,520 --> 00:13:53,800 been thinking about for ages. 418 00:13:53,800 --> 00:13:54,880 Yeah, for sure. 419 00:13:54,880 --> 00:13:56,760 And with AI evolving so fast, it's 420 00:13:56,760 --> 00:13:59,480 becoming even more important to figure out. 421 00:13:59,480 --> 00:14:02,720 I was reading about this idea of the singularity, 422 00:14:02,720 --> 00:14:05,320 like where AI becomes smarter than humans, 423 00:14:05,320 --> 00:14:06,760 and everything changes. 424 00:14:06,760 --> 00:14:07,720 Oh, yeah, the singularity. 425 00:14:07,720 --> 00:14:10,520 It's a fascinating concept and a bit controversial. 426 00:14:10,520 --> 00:14:12,040 Some people think it's inevitable. 427 00:14:12,040 --> 00:14:15,320 Others, well, they think it's just speculation. 428 00:14:15,320 --> 00:14:17,840 But whether it actually happens or not, 429 00:14:17,840 --> 00:14:21,440 it really forces us to confront some big questions 430 00:14:21,440 --> 00:14:23,960 about intelligence, consciousness, even 431 00:14:23,960 --> 00:14:25,240 our place in the universe. 432 00:14:25,240 --> 00:14:29,080 It's like we're living in a sci-fi movie, except it's real. 433 00:14:29,080 --> 00:14:30,200 And we're not just watching. 434 00:14:30,200 --> 00:14:31,800 We're actually shaping it. 435 00:14:31,800 --> 00:14:33,240 That's a great way to put it. 436 00:14:33,240 --> 00:14:35,120 We're not passengers on this ride. 437 00:14:35,120 --> 00:14:37,000 We're the ones building the roller coaster. 438 00:14:37,000 --> 00:14:39,600 And that means we've got a huge responsibility. 439 00:14:39,600 --> 00:14:41,840 Because the choices we make now, they're 440 00:14:41,840 --> 00:14:45,600 going to determine where AI goes and how it affects our world. 441 00:14:45,600 --> 00:14:46,400 Exactly. 442 00:14:46,400 --> 00:14:48,880 That's why we need to be having these conversations 443 00:14:48,880 --> 00:14:51,800 about the impact of AI, considering all sides, 444 00:14:51,800 --> 00:14:52,880 the good and the bad. 445 00:14:52,880 --> 00:14:55,360 We've talked about that gorilla problem, 446 00:14:55,360 --> 00:14:57,120 creating something that could outsmart us. 447 00:14:57,120 --> 00:14:58,520 But even if that doesn't happen, there 448 00:14:58,520 --> 00:14:59,840 are other things to worry about. 449 00:14:59,840 --> 00:15:00,800 Oh, definitely. 450 00:15:00,800 --> 00:15:03,560 One big one is the potential for AI 451 00:15:03,560 --> 00:15:06,880 to make existing inequality even worse. 452 00:15:06,880 --> 00:15:09,160 If we're not careful, it could end up 453 00:15:09,160 --> 00:15:12,160 concentrating power and wealth in the hands of just a few. 454 00:15:12,160 --> 00:15:13,880 It's like any powerful technology, right? 455 00:15:13,880 --> 00:15:16,400 It can be used for good or for bad. 456 00:15:16,400 --> 00:15:18,040 It depends on who's controlling it. 457 00:15:18,040 --> 00:15:19,400 Exactly. 458 00:15:19,400 --> 00:15:22,040 And that means thinking about the values that 459 00:15:22,040 --> 00:15:25,200 are built into these AI systems, who's designing them, 460 00:15:25,200 --> 00:15:26,720 who benefits from them. 461 00:15:26,720 --> 00:15:28,040 Those are important questions. 462 00:15:28,040 --> 00:15:29,600 It's not just about the tech itself. 463 00:15:29,600 --> 00:15:32,280 It's about the social and economic impacts, too. 464 00:15:32,280 --> 00:15:33,560 Absolutely. 465 00:15:33,560 --> 00:15:36,280 AI is reshaping our societies, our economies, 466 00:15:36,280 --> 00:15:38,800 even our understanding of what it means to be human. 467 00:15:38,800 --> 00:15:40,960 Which brings us back to that question. 468 00:15:40,960 --> 00:15:44,160 If we're so far from understanding our own brains, 469 00:15:44,160 --> 00:15:47,880 what does that say about creating truly human-like AI? 470 00:15:47,880 --> 00:15:49,440 Can we really do it? 471 00:15:49,440 --> 00:15:50,640 That's a tough one. 472 00:15:50,640 --> 00:15:52,440 I mean, consciousness is so complex. 473 00:15:52,440 --> 00:15:54,800 I think it's more than just neurons firing. 474 00:15:54,800 --> 00:15:58,000 Our experiences, our emotions, our creativity, 475 00:15:58,000 --> 00:15:59,880 those things might be impossible to replicate. 476 00:15:59,880 --> 00:16:02,120 It's like there's this magic to consciousness, something 477 00:16:02,120 --> 00:16:04,320 we can't just code into a machine. 478 00:16:04,320 --> 00:16:05,360 Maybe. 479 00:16:05,360 --> 00:16:07,400 Or maybe we just haven't figured out how to do it yet. 480 00:16:07,400 --> 00:16:08,920 And that's what makes it so interesting. 481 00:16:08,920 --> 00:16:11,400 We're exploring this unknown territory, 482 00:16:11,400 --> 00:16:13,840 grappling with questions that have been around for centuries. 483 00:16:13,840 --> 00:16:15,640 And the answers we find, they're going 484 00:16:15,640 --> 00:16:16,880 to shape human history. 485 00:16:16,880 --> 00:16:19,640 So let's wrap up our deep dive with a few takeaways. 486 00:16:19,640 --> 00:16:21,960 First, AI is changing fast. 487 00:16:21,960 --> 00:16:24,160 And it's already having a huge impact. 488 00:16:24,160 --> 00:16:28,120 Second, the future of AI is in our hands. 489 00:16:28,120 --> 00:16:30,800 The choices we make now, the research we do, 490 00:16:30,800 --> 00:16:33,120 the ethics we consider, they all matter. 491 00:16:33,120 --> 00:16:34,960 And finally, understanding intelligence, 492 00:16:34,960 --> 00:16:36,840 both human and artificial. 493 00:16:36,840 --> 00:16:39,200 It's one of the most fascinating and important things 494 00:16:39,200 --> 00:16:40,600 we can be doing right now. 495 00:16:40,600 --> 00:16:42,880 The answers will shape not just technology, 496 00:16:42,880 --> 00:16:44,520 but humanity itself. 497 00:16:44,520 --> 00:16:47,400 Thanks for joining us on this deep dive into the world of AI. 498 00:16:47,400 --> 00:16:48,880 We hope you've enjoyed it, and that it's 499 00:16:48,880 --> 00:16:50,360 given you some things to think about. 500 00:16:50,360 --> 00:16:52,240 Keep exploring, keep asking questions, 501 00:16:52,240 --> 00:16:53,720 keep pushing the boundaries. 502 00:16:53,720 --> 00:17:20,120 That's how we move forward.