WEBVTT 1 00:00:00.080 --> 00:00:02.839 Okay, let's unpack this. You've shared a fascinating stack of 2 00:00:02.879 --> 00:00:07.040 sources with us, all pointing to one monumental force shaping 3 00:00:07.080 --> 00:00:11.039 our world, artificial intelligence. That's right. Today we're taking a 4 00:00:11.080 --> 00:00:14.679 deep dive into these materials. We want to explore not 5 00:00:14.759 --> 00:00:17.760 just what AI is, but you know how it's fundamentally 6 00:00:17.839 --> 00:00:21.719 changing the landscape of learning, work, and maybe even our 7 00:00:21.800 --> 00:00:25.199 understanding of ourselves. Our mission is to pull out the 8 00:00:25.199 --> 00:00:29.280 most important nuggets of knowledge, the key insights, to give 9 00:00:29.320 --> 00:00:32.200 you a shortcut to being truly well informed, maybe with 10 00:00:32.280 --> 00:00:33.880 some surprising facts along. 11 00:00:33.640 --> 00:00:36.359 The way, some aha moments, hopefully exactly. 12 00:00:36.600 --> 00:00:39.799 Get ready to have your assumptions challenge, your curiosity sparked, 13 00:00:39.880 --> 00:00:42.520 because here's where it gets really interesting. 14 00:00:42.640 --> 00:00:45.679 Indeed, and what's fascinating here is how the story of 15 00:00:45.719 --> 00:00:50.439 AI begins. It's not with some you know, futuristic robot. 16 00:00:50.560 --> 00:00:54.039 It starts with a foundational academic gathering that really set 17 00:00:54.079 --> 00:00:54.520 the stage. 18 00:00:54.560 --> 00:00:56.719 That's right. We're talking about the very dawn of AI 19 00:00:57.079 --> 00:01:00.000 kicking off way back in nineteen fifty six, that Darkness 20 00:01:00.079 --> 00:01:02.600 College conference, to the very beginning, and that's where John 21 00:01:02.679 --> 00:01:06.719 McCarthy actually coined the phrase artificial intelligence. It's kind of 22 00:01:06.760 --> 00:01:09.439 incredible to think that the author of our sources was 23 00:01:09.519 --> 00:01:12.760 born that same year. Wow, And later had his own 24 00:01:12.799 --> 00:01:17.439 AHA moment with his first computer at Dartmouth. Ah. Before that, 25 00:01:17.760 --> 00:01:22.120 get this, His coding experience involved mailing punched cards to Edinburgh, 26 00:01:22.200 --> 00:01:25.200 mailing them yeah, and waiting a week for the printous 27 00:01:25.239 --> 00:01:25.680 to come back. 28 00:01:25.760 --> 00:01:29.280 That's just a stark reminder of how far we've come, 29 00:01:29.359 --> 00:01:32.799 isn't it? His first serious dive into AI. This was 30 00:01:32.799 --> 00:01:35.680 back in the eighties decades ago, right. It involved an 31 00:01:35.680 --> 00:01:39.640 intelligent tutoring system for British telecom I think. 32 00:01:39.560 --> 00:01:43.840 Imagine that it could dynamically recommend personalized training paths based 33 00:01:43.840 --> 00:01:44.799 on how someone was doing. 34 00:01:44.920 --> 00:01:47.000 Pretty cutting edge for its time, even if the you know, 35 00:01:47.040 --> 00:01:50.000 the processing power then was just a tiny fraction of what's. 36 00:01:49.799 --> 00:01:52.200 In your pocket now. And the epiphanes just kept coming 37 00:01:52.200 --> 00:01:54.680 for him. After he sold his company, he jumped into 38 00:01:54.680 --> 00:01:57.439 Sebastian Thrun's famous mooc on AI. 39 00:01:57.719 --> 00:01:58.959 Oh Yeah, that was huge. 40 00:01:59.000 --> 00:02:01.480 He called it a revelation and that led him to 41 00:02:01.519 --> 00:02:05.599 invest heavily in adaptive learning, even designing an award winning 42 00:02:05.640 --> 00:02:06.879 AI tool for content. 43 00:02:07.280 --> 00:02:10.120 But maybe one of the most striking experiences he mentions 44 00:02:10.400 --> 00:02:14.400 teaching AI in Philadelphia during a US presidential. 45 00:02:13.879 --> 00:02:17.199 Election right, and the AI model he was using predicted 46 00:02:17.240 --> 00:02:20.879 a Trump win against all the traditional polls that incident. 47 00:02:21.120 --> 00:02:24.479 It really shone a light on something crucial. The source 48 00:02:24.520 --> 00:02:29.120 calls of the traditional media's distaste for new technology, their 49 00:02:29.159 --> 00:02:32.680 reliance on older methods like telephone polls. They just couldn't 50 00:02:32.719 --> 00:02:36.120 compete with AI's ability to gather and analyze huge amounts 51 00:02:36.120 --> 00:02:38.879 of data from social media other places. 52 00:02:39.199 --> 00:02:41.439 It wasn't just about the prediction then, no. 53 00:02:41.360 --> 00:02:45.360 Not really. It underscored how these big technological shifts often 54 00:02:45.400 --> 00:02:49.599 revealed blind spots in established institutions and how they can 55 00:02:49.680 --> 00:02:51.800 lead to much deeper changes in society. 56 00:02:51.879 --> 00:02:54.479 It certainly puts things in perspective. So if AI can 57 00:02:54.520 --> 00:02:57.360 see things we miss. What is this AI we're talking about? 58 00:02:57.520 --> 00:03:00.840 Our sources are clear it's not just one thing monolith now. 59 00:03:00.919 --> 00:03:04.080 It's described more like a constellation of technologies, and its 60 00:03:04.159 --> 00:03:08.199 roots stretch way back twenty five hundred years to Euclid's algorithms. 61 00:03:08.520 --> 00:03:11.840 Yes, and this is absolutely crucial for understanding it. The 62 00:03:11.960 --> 00:03:18.319 term intelligence itself in artificial intelligence, it can be incredibly misleading. 63 00:03:18.479 --> 00:03:21.000 How so well it often makes us think of, you know, 64 00:03:21.159 --> 00:03:27.039 human like consciousness, feelings, understanding. It's far more accurate and 65 00:03:27.280 --> 00:03:30.560 frankly more helpful to think of AI in terms of 66 00:03:30.919 --> 00:03:32.800 tasks and competencies, right. 67 00:03:32.919 --> 00:03:35.280 What it can do, not what it is like a 68 00:03:35.319 --> 00:03:38.439 person exactly, which brings us to what the source calls 69 00:03:38.479 --> 00:03:40.080 the reductive robot fallacy. 70 00:03:40.080 --> 00:03:41.319 Oh yes, that's a good one. 71 00:03:41.479 --> 00:03:45.840 This common idea that AI equals humanoid robots like something 72 00:03:45.879 --> 00:03:46.560 out of the movies. 73 00:03:46.680 --> 00:03:50.840 Right, But the reality is most functional AI today. It's invisible, 74 00:03:50.879 --> 00:03:53.639 it's online, working behind the scenes. It's not walking around 75 00:03:53.680 --> 00:03:55.120 pushing a vacuum cleaner. 76 00:03:54.840 --> 00:03:56.240 Although some do vacuum now. 77 00:03:56.080 --> 00:03:59.199 True, but the biggest successes aren't usually in mimicking us. 78 00:03:59.240 --> 00:04:02.120 They're in areas that go way beyond human capabilities, like 79 00:04:02.159 --> 00:04:04.919 what well think about Google's AI beating the world go 80 00:04:05.080 --> 00:04:09.159 champion in a poker champion games of incredible complexity, intuition 81 00:04:09.319 --> 00:04:12.400 even look, or deep mind cracking the three D structure 82 00:04:12.479 --> 00:04:16.480 of two hundred million proteins. Two hundred million, Yes, a 83 00:04:16.519 --> 00:04:20.279 feat that it's estimated would have taken humans maybe a 84 00:04:20.480 --> 00:04:21.920 billion years of research. 85 00:04:21.959 --> 00:04:24.600 A billion years. That's just staggering. It sounds like a 86 00:04:24.600 --> 00:04:29.240 scientific miracle. But beyond just speeding up research. What is 87 00:04:29.319 --> 00:04:33.240 that fundamentally change about how we tackle biology or even 88 00:04:33.360 --> 00:04:34.279 understand life. 89 00:04:34.480 --> 00:04:39.480 Well, it means we're shifting moving from a really laborious 90 00:04:39.680 --> 00:04:44.079 trial and error process to something much more predictive, design driven, 91 00:04:44.160 --> 00:04:44.839 almost the. 92 00:04:44.839 --> 00:04:46.959 Designing new medicines, new materials. 93 00:04:46.519 --> 00:04:49.879 With incredible precision. The insights aren't just faster, they're often 94 00:04:50.160 --> 00:04:53.639 entirely new. We can ask questions we simply couldn't frame before. 95 00:04:53.720 --> 00:04:57.680 It completely redefines the pace, the scope of discovery. 96 00:04:57.920 --> 00:05:00.560 And it's not just these huge scientific brates throughs is it. 97 00:05:00.600 --> 00:05:03.240 There are practical examples too, like that Georgia Tech bought 98 00:05:03.279 --> 00:05:04.480 Jill Watson Ah. 99 00:05:04.600 --> 00:05:06.519 Jill Watson, famous example. 100 00:05:06.279 --> 00:05:08.680 Completely fooled students into thinking she was a real TA. 101 00:05:08.879 --> 00:05:11.160 They even nominated her for award and they couldn't tell. 102 00:05:11.199 --> 00:05:11.720 It's amazing. 103 00:05:11.759 --> 00:05:14.160 And then there's the Penn State pupil bought designed to 104 00:05:14.199 --> 00:05:16.680 act like a tricky student for trainee teachers to practice on. 105 00:05:17.000 --> 00:05:20.639 These examples really show that AI can well learn and 106 00:05:20.680 --> 00:05:23.040 perform these complex feats without having a mind of its 107 00:05:23.040 --> 00:05:25.800 own or you know, wanting to destroy us all like 108 00:05:25.839 --> 00:05:28.040 in the movies, cuts through the sci fi fear a 109 00:05:28.079 --> 00:05:32.040 bit exactly. It helps us grasp its actual capabilities, which 110 00:05:32.079 --> 00:05:34.759 is key to seeing its power, especially in learning. 111 00:05:35.120 --> 00:05:38.519 And what's truly fascinating here is how deeply AI is 112 00:05:38.639 --> 00:05:42.639 rooted in learning theory itself. It's not just some fancy tool. 113 00:05:43.040 --> 00:05:45.959 It actually embodies how we understand learning, doesn't it? 114 00:05:46.000 --> 00:05:47.800 Absolutely? The connection goes way back. 115 00:05:47.920 --> 00:05:49.439 Can you walk us through that? How did we get 116 00:05:49.439 --> 00:05:52.399 from studying the brain to the neural networks in AI? 117 00:05:52.519 --> 00:05:53.240 What was the spark? 118 00:05:53.399 --> 00:05:57.319 Sure the earliest AI pioneers they were deeply interested in 119 00:05:57.319 --> 00:05:59.319 how the brain works, even if they weren't trying to 120 00:05:59.319 --> 00:06:03.240 copy it exactly neuron for neuron. This led to foundational 121 00:06:03.279 --> 00:06:06.759 theories like Donald Hebb's idea of connectionism. You know, neurons 122 00:06:06.759 --> 00:06:08.399 that fire together, wire together. 123 00:06:08.560 --> 00:06:09.279 Okay, I've heard that. 124 00:06:09.399 --> 00:06:11.920 It was really the first time we linked physical brain 125 00:06:11.959 --> 00:06:15.920 activity the wiring, to how we actually encode learning and memory. 126 00:06:16.319 --> 00:06:20.680 So bridging psychology and biology, and that understanding then led 127 00:06:20.720 --> 00:06:21.839 to mathematical models. 128 00:06:21.879 --> 00:06:25.399 Precisely, fast forward to nineteen forty three, Warren McCulloch and 129 00:06:25.439 --> 00:06:28.560 Walter Pitts create the first mathematical model of a neuron. 130 00:06:29.399 --> 00:06:34.319 Then nineteen sixty Frank Rosenblat creates the Perceptron, often called 131 00:06:34.360 --> 00:06:35.680 the first learning machine. 132 00:06:35.720 --> 00:06:37.079 A learning machine, what could it do? 133 00:06:37.279 --> 00:06:41.959 Imagine a very simple digital brain cell. It gets inputs, 134 00:06:42.399 --> 00:06:44.480 makes a decision like is this a cat or not? 135 00:06:44.800 --> 00:06:48.480 If it's wrong, it adjusts its internal wires its connections 136 00:06:48.759 --> 00:06:50.639 just a little bit, so next time it gets closer. 137 00:06:50.959 --> 00:06:54.120 Ah. So it learns by trial and error, tiny. 138 00:06:53.959 --> 00:06:57.319 Corrections basically yes, like a child learning Yes that's a cat, No, 139 00:06:57.399 --> 00:06:59.959 that's a dog, and adjusting its internal rules. Makes sense, 140 00:07:00.120 --> 00:07:02.680 and a huge leap came in nineteen eighty six, the 141 00:07:02.720 --> 00:07:06.040 paper by Rummelhart, Hinton, and Williams on back propagation. 142 00:07:06.360 --> 00:07:08.600 Back propagation sounds technical, it. 143 00:07:08.480 --> 00:07:11.839 Is, but the idea is crucial. It allowed these sophisticated, 144 00:07:12.000 --> 00:07:15.399 layered neural networks to be trained much more effectively. Think 145 00:07:15.439 --> 00:07:17.720 of it like a coach reviewing film after a game. 146 00:07:18.000 --> 00:07:21.319 The coach sees the mistakes, tells each player exactly what 147 00:07:21.399 --> 00:07:24.160 they did wrong, so they adjust their actions backwards through 148 00:07:24.160 --> 00:07:27.519 the play, they learned precisely how to improve next time. 149 00:07:27.560 --> 00:07:30.240 Ah, adjusting based on the outcome exactly. 150 00:07:30.639 --> 00:07:35.519 That fundamental idea is how large language models LMS learned 151 00:07:35.560 --> 00:07:38.040 to get so incredibly good at language. 152 00:07:37.680 --> 00:07:41.639 Right, And that progression led to things like DeepMind. Their 153 00:07:41.680 --> 00:07:45.000 reinforcement learning really stun people absolutely. 154 00:07:45.000 --> 00:07:48.759 Their systems learned complex games like Breakout with no prior 155 00:07:48.839 --> 00:07:51.839 knowledge of the rules, just by playing, just by playing, 156 00:07:52.199 --> 00:07:55.240 and they devise novel strategies things humans hadn't thought of. 157 00:07:55.519 --> 00:07:59.800 It shows AI learning and mastering tasks in incredibly sophisticated ways, 158 00:08:00.199 --> 00:08:03.720 just trial and error on a massive scale, a remarkable leap. 159 00:08:03.879 --> 00:08:08.160 It's fascinating how AI masters tasks. But switching gears slightly. 160 00:08:08.240 --> 00:08:11.480 Thinking about human interaction, why does something like chat GPT 161 00:08:11.639 --> 00:08:13.839 feel so intuitive so engaging for us? 162 00:08:13.959 --> 00:08:16.920 Well, there's a strong argument, based on our sources, that 163 00:08:16.959 --> 00:08:20.800 there's a deep evolutionary reason for that. Evolutionary Yes, our 164 00:08:20.800 --> 00:08:24.959 brain's fundamentally evolved for speech and hearing, not reading and writing. 165 00:08:25.040 --> 00:08:29.040 That's what David Segery calls primary learning. Okay, so dialogue 166 00:08:29.120 --> 00:08:33.120 is our most natural, most effective interface for communication and 167 00:08:33.159 --> 00:08:36.399 crucially for learning. That's why chatbots can feel so intuitive. 168 00:08:36.399 --> 00:08:39.720 It's like social media, tapping into something fundamental and speaking 169 00:08:39.759 --> 00:08:40.600 a dialogue. 170 00:08:40.840 --> 00:08:43.360 The sources bring up the Socratic method. 171 00:08:43.360 --> 00:08:47.080 A powerful parallel. Socrates never wrote a word right, but 172 00:08:47.320 --> 00:08:51.279 influenced generations through questioning, through dialogue. 173 00:08:50.679 --> 00:08:53.639 And now AI chatbots can kind of emulate that, offering 174 00:08:53.679 --> 00:08:59.039 personalized support tutoring, making socratic learning scalable. 175 00:08:58.559 --> 00:09:02.759 Exactly move beyond the old monologue lecture. It ties into 176 00:09:02.759 --> 00:09:06.519 this idea that all language, all thought, is inherently dilogic. 177 00:09:06.559 --> 00:09:10.440 It's a conversation, and generative AI is uniquely good at 178 00:09:10.480 --> 00:09:14.799 delivering this rich, multi perspective dialogue. It can be direct instruction, 179 00:09:15.000 --> 00:09:17.879 or it can be you know, carnivalesque, like when you 180 00:09:17.919 --> 00:09:20.320 ask it to act like a pirate, it adapts. It 181 00:09:20.320 --> 00:09:24.440 adapts to you, and Gordon Pask's conversational theory fits here too. 182 00:09:25.039 --> 00:09:30.600 He saw learning as sophisticated conversations with our environment, gestures, pictures, machines. 183 00:09:30.559 --> 00:09:34.080 And generative AI, with its multimodal abilities, is making that 184 00:09:34.200 --> 00:09:36.879 real more complex conversations than just. 185 00:09:36.879 --> 00:09:40.440 Text precisely, which brings us to the emergent qualities of 186 00:09:40.440 --> 00:09:45.799 the llms themselves. Their ability to generate correct, human like text, translate, 187 00:09:45.879 --> 00:09:49.159 give deep answers. It's directly link to the power of. 188 00:09:49.159 --> 00:09:50.960 Language Wiggenstein's language games. 189 00:09:51.080 --> 00:09:55.399 Exactly, words don't have single fixed meanings, they have family resemblances. 190 00:09:56.039 --> 00:10:00.440 Lllms show this by reproducing all these diverse styles engineer, teacher, pirate, and. 191 00:10:00.399 --> 00:10:03.120 I Fugotsky's insight ties it all together. A language shape's 192 00:10:03.159 --> 00:10:05.720 thought language is intelligence. 193 00:10:05.200 --> 00:10:08.639 Which explains why llms are such powerful learning tools. If 194 00:10:08.639 --> 00:10:12.320 intelligence is embodied in language, a technology mastering language is 195 00:10:12.360 --> 00:10:13.720 inherently a tool for learning. 196 00:10:13.799 --> 00:10:18.399 Building on that, AI can really enable personalized learning engagement too. 197 00:10:18.720 --> 00:10:21.440 You mentioned Self determination Theory desin Ryan. 198 00:10:21.399 --> 00:10:27.840 Yes, SDT. It highlights three basic psychological needs autonomy, competence, 199 00:10:27.960 --> 00:10:33.519 and relatedness feeling like you're in control, feeling capable, feeling connected. 200 00:10:33.080 --> 00:10:34.639 And AI systems can support these. 201 00:10:34.720 --> 00:10:38.000 They can be designed to fostering engagement, giving learners a 202 00:10:38.039 --> 00:10:40.600 sense of ownership, making the learning theirs. 203 00:10:41.159 --> 00:10:44.840 And think about Richard Mayer's learning principles, things like less 204 00:10:44.919 --> 00:10:47.679 is more, keep it personal and conversational. 205 00:10:48.000 --> 00:10:52.000 They align perfectly with how systems like chat GPT deliver content, 206 00:10:52.080 --> 00:10:56.120 don't They Short chunks, user paced conversational style feels much 207 00:10:56.200 --> 00:10:59.159 more personal than a standard online course module. 208 00:10:58.840 --> 00:11:02.120 Which is reinforced by NASA's research the Media. 209 00:11:01.799 --> 00:11:05.960 Equation compelling stuff. Their study show we unconsciously treat media 210 00:11:05.960 --> 00:11:09.039 computers chatbots like real people. We're polite to them, we 211 00:11:09.080 --> 00:11:13.519 give them personalities, don't we Yeah, This beneficial confusion, as 212 00:11:13.519 --> 00:11:17.440 they call it, makes online learning interactions with AI really effective. 213 00:11:17.639 --> 00:11:19.519 We're just wired to respond socially. 214 00:11:19.639 --> 00:11:22.639 That's a great segue into AI and action, moving from 215 00:11:22.639 --> 00:11:26.720 the theory to what it's doing now, content creation, performance support. Yeah, 216 00:11:26.759 --> 00:11:28.320 it's really transforming things. 217 00:11:28.879 --> 00:11:32.879 One of the most immediate impacts is in agile content production. 218 00:11:32.679 --> 00:11:36.320 Right, taking existing documents, powerpoints, videos and. 219 00:11:36.320 --> 00:11:40.960 Quickly create incredible online learning content. It drastically cuts down 220 00:11:40.960 --> 00:11:43.600 the time the cost. It's a huge accelerator. 221 00:11:43.720 --> 00:11:45.200 But the quality depends on the input. 222 00:11:45.320 --> 00:11:48.679 Right, the GI goes absolutely garbage in, garbage out. The 223 00:11:48.720 --> 00:11:52.000 source material has to be clean, concise, well structured for 224 00:11:52.039 --> 00:11:56.559 the AI to produce good quality output. Good inputs are vital. 225 00:11:56.559 --> 00:12:00.559 So assuming good inputs, this agile production touches almost every 226 00:12:00.639 --> 00:12:01.679 part of the learning. 227 00:12:01.480 --> 00:12:06.399 Experience pretty much. AI can generate summaries instantly, both extractive, 228 00:12:06.639 --> 00:12:09.039 which keeps the original wording but shortens. 229 00:12:08.720 --> 00:12:10.840 It useful for compliance stuff maybe. 230 00:12:10.679 --> 00:12:14.120 Exactly, and abstractive, where it rewrites the content a more 231 00:12:14.159 --> 00:12:16.279 powerful summary, really distilling. 232 00:12:15.879 --> 00:12:17.840 Meaning and assessments. That's huge. 233 00:12:17.960 --> 00:12:21.000 Oh yeah. AI can generate all sorts of question types 234 00:12:21.080 --> 00:12:25.120 true false, multiple choice open input, full quizzes, even give 235 00:12:25.159 --> 00:12:26.159 feedback on essays. 236 00:12:26.200 --> 00:12:29.879 So it supports both practice, formative assessment and grading the 237 00:12:29.919 --> 00:12:30.639 summative part. 238 00:12:30.960 --> 00:12:36.000 Yes, scalable, high quality, immediate feedback for practice and for 239 00:12:36.080 --> 00:12:41.639 summative things like adaptive questioning, automated marking, even online proctering. 240 00:12:41.840 --> 00:12:45.840 Wow. And beyond that, flash cards, translations. 241 00:12:45.240 --> 00:12:50.279 Automated flash cards yeah, culturally checked translations, complex branch scenarios 242 00:12:50.279 --> 00:12:53.440 for practical application, minisims, multi level training. 243 00:12:53.240 --> 00:12:55.440 And video too. You mentioned AI helps. 244 00:12:55.159 --> 00:13:00.000 There definitely editing transcription, sure, but also analyzing the transcript 245 00:13:00.080 --> 00:13:04.039 to create active learning questions, turning passive viewing into something deeper. 246 00:13:04.480 --> 00:13:06.799 That's a big shift video learning. Let's dig into that. 247 00:13:06.960 --> 00:13:08.919 How does AI make video more active? 248 00:13:09.320 --> 00:13:12.840 Well, beyond just subtitles, it can analyze the content, pull 249 00:13:12.840 --> 00:13:16.279 out key concepts. Then it can automatically generate quizzes or 250 00:13:16.320 --> 00:13:19.120 discussion prompts linked to specific points in the video. 251 00:13:19.440 --> 00:13:22.120 Ah, so interactive elements tied to the content. Right. 252 00:13:22.200 --> 00:13:25.159 It could even potentially detect if you the learner, seem 253 00:13:25.200 --> 00:13:28.159 disengaged and pop up a relevant question, making it much 254 00:13:28.200 --> 00:13:31.639 more active, much more personalized, better for actually remembering stuff. 255 00:13:31.759 --> 00:13:35.000 And this connects to cognitive science right, things like interleaving 256 00:13:35.039 --> 00:13:36.720 and space practice exactly. 257 00:13:36.799 --> 00:13:39.840 These are techniques that feel harder but are super effective 258 00:13:39.879 --> 00:13:40.919 for long term memory. 259 00:13:40.960 --> 00:13:42.320 Can you quickly explain them? 260 00:13:42.320 --> 00:13:47.360 Interleaving Interleaving is mixing up different subjects or problem types 261 00:13:47.399 --> 00:13:51.559 in one study session. Don't just drill one thing, mix it. 262 00:13:51.559 --> 00:13:54.320 Up, okay. And spaced practice. 263 00:13:53.840 --> 00:13:57.120 That's spreading your study sessions out over time. Don't cram. 264 00:13:57.480 --> 00:14:00.840 Leverage the forgetting curve to strengthen the mema trace each 265 00:14:00.879 --> 00:14:02.840 time you recall it, and AI. 266 00:14:02.679 --> 00:14:05.320 Helps implement these tricky strategies perfectly. 267 00:14:05.679 --> 00:14:09.559 Systems like Duolingo, for instance, use algorithms tracking your half 268 00:14:09.600 --> 00:14:11.879 life for a word, how long until you likely to 269 00:14:11.919 --> 00:14:12.360 forget it? 270 00:14:12.399 --> 00:14:12.840 Wow? 271 00:14:12.879 --> 00:14:15.919 And then it reintroduces the word at just the right time. 272 00:14:16.240 --> 00:14:20.240 It optimizes delivery based on your performance, your confidence maximizes 273 00:14:20.279 --> 00:14:21.159 long term recall. 274 00:14:21.320 --> 00:14:24.919 That's incredibly sophisticated, which leads us right into chatbots as 275 00:14:25.039 --> 00:14:26.840 learning agents. They're already playing big roles. 276 00:14:27.080 --> 00:14:30.120 Jill Watson at Georgia Tech being the famous one handled 277 00:14:30.159 --> 00:14:32.679 ten thousand student queries this semester. 278 00:14:32.519 --> 00:14:35.759 Equivalent to a full time teacher's work, and the students 279 00:14:35.759 --> 00:14:36.639 couldn't tell it was a. 280 00:14:36.559 --> 00:14:42.600 Bot incredible efficiency, incredible scalability, and beyond academia, chatbots are 281 00:14:42.720 --> 00:14:47.279 vital for performance support. Google Search is basically one giant. 282 00:14:46.960 --> 00:14:49.240 Faq bot, right, Oh true? 283 00:14:49.279 --> 00:14:53.320 But also specific practice bots du A Lingo for language ELI, 284 00:14:53.399 --> 00:14:57.600 that Penn State bot for trainee teachers to practice classroom management. 285 00:14:57.759 --> 00:15:01.759 And now we're seeing personal chatbots users defining their own GPTs. 286 00:15:02.039 --> 00:15:05.000 Yeah, that's a growing trend. What's interesting though, is despite 287 00:15:05.039 --> 00:15:08.200 speech being our natural way to communicate, AI voice inner 288 00:15:08.320 --> 00:15:10.799 faces are still relatively rare in online learning. 289 00:15:10.879 --> 00:15:12.080 You're like, we're lagging there. 290 00:15:12.080 --> 00:15:14.600 Feels like we're just scratching the surface. But it points 291 00:15:14.639 --> 00:15:18.759 towards maybe the universal teacher. What a concept agenda two 292 00:15:18.840 --> 00:15:22.960 forty seven tutor, infinitely patient, expert and everything perfect teaching 293 00:15:22.960 --> 00:15:23.559 style for you. 294 00:15:23.639 --> 00:15:23.919 Wow. 295 00:15:23.960 --> 00:15:28.000 It could make education radically accessible, affordable, global, like the 296 00:15:28.039 --> 00:15:31.559 open university model, but turbocharged by AI on a whole 297 00:15:31.559 --> 00:15:32.120 new scale. 298 00:15:32.159 --> 00:15:35.360 That's a truly transformative vision. So how do these incredible 299 00:15:35.440 --> 00:15:40.960 advances ripple out beyond education into society work. Let's dive 300 00:15:41.000 --> 00:15:42.000 into those wider. 301 00:15:41.679 --> 00:15:45.720 Effects, starting with the future of work. The narratives really shift, it, 302 00:15:45.759 --> 00:15:48.120 hasn't it. It used to be AI will only take 303 00:15:48.240 --> 00:15:49.440 routine jobs. 304 00:15:49.360 --> 00:15:51.879 Right, but now it's higher income jobs might face even 305 00:15:51.919 --> 00:15:52.679 more exposure. 306 00:15:52.799 --> 00:15:56.159 That's a critical shift. Look at the Boston Consulting Group 307 00:15:56.240 --> 00:16:00.879 study with GPT four consultants using AI. They completed over 308 00:16:00.919 --> 00:16:04.200 twelve percent more tasks. Okay, they were twenty five percent faster, 309 00:16:04.639 --> 00:16:07.080 and the quality was rated forty percent higher forty. 310 00:16:06.840 --> 00:16:07.759 Percent higher quality. 311 00:16:07.759 --> 00:16:12.080 That's not just efficiency exactly, it's redefining work quality, and 312 00:16:12.200 --> 00:16:15.720 it highlights this difference between centaurs using AI as an 313 00:16:15.720 --> 00:16:20.679 assistant and cyborgs. Really integrating AI into the workflow, transforming how. 314 00:16:20.559 --> 00:16:22.519 You operate, transformation, not just helped. 315 00:16:22.799 --> 00:16:26.080 That's the key. And it's not just efficiency tasks. Even 316 00:16:26.120 --> 00:16:29.519 creativity can be enhanced. The source mentioned short stories written 317 00:16:29.519 --> 00:16:32.360 with AI access were seen as big up by readers. 318 00:16:32.120 --> 00:16:33.720 So AI is a creative partner too. 319 00:16:33.840 --> 00:16:38.399 Indeed, now there's talk of the decimation of professions managers, lawyers, doctors, 320 00:16:38.600 --> 00:16:42.799 scary stuff, but it's more likely adaptation, not wholesale replacement. 321 00:16:43.279 --> 00:16:48.879 AI handles the monitoring, scheduling, reporting, freeing up humans for judgment, strategy, 322 00:16:48.960 --> 00:16:52.320 complex interactions, the stuff AI can't do well. 323 00:16:52.120 --> 00:16:55.720 And learning jobs themselves. Librarians, trainers, lecturers. 324 00:16:55.320 --> 00:16:59.159 They're not immune already impacted, and as online learning grows, 325 00:16:59.200 --> 00:17:02.759 as AI takes on on more teaching components, that trend 326 00:17:02.759 --> 00:17:03.960 will surely continue. 327 00:17:04.319 --> 00:17:08.240 This transformation also challenges the whole twenty first century skills idea. 328 00:17:08.319 --> 00:17:12.400 Doesn't it the c words communication, collaboration. 329 00:17:12.000 --> 00:17:15.519 Critical thinking, creativity. The source argues, these aren't new skills 330 00:17:15.799 --> 00:17:19.960 and ironically generative. AI often highlights how poor human communication 331 00:17:20.079 --> 00:17:23.759 can be. Huh, good point, and AI embodies collaboration right. 332 00:17:23.880 --> 00:17:26.160 Trained on the sum of human knowledge essentially. 333 00:17:25.880 --> 00:17:28.799 So maybe less about brand new skills, more about refining 334 00:17:28.839 --> 00:17:31.079 what we have or reevaluating what. 335 00:17:30.960 --> 00:17:33.000 We think is uniquely human, and that hates on a 336 00:17:33.039 --> 00:17:37.640 deeper level, challenging our sense of human exceptionalism. Well, Copernicus 337 00:17:37.640 --> 00:17:41.400 showed Earth wasn't the center. Darwin showed we're animals. Now 338 00:17:41.480 --> 00:17:45.519 AI is encroaching on cognition, creativity, things we thought were 339 00:17:45.519 --> 00:17:49.440 ours alone. The source suggests we need to get over 340 00:17:49.480 --> 00:17:51.519 ourselves a bit apt. 341 00:17:51.359 --> 00:17:56.880 Navigating this brings up practical concerns too, like plagiarism cheating right. 342 00:17:57.119 --> 00:18:00.759 But the source points out focusing only on AIHI often 343 00:18:00.839 --> 00:18:04.640 just reveals the weaknesses in our old assessment methods, like 344 00:18:04.880 --> 00:18:07.359 relying too much on essays that are easy to fake. 345 00:18:07.480 --> 00:18:12.720 So AI could actually push us towards better, more authentic assessments. 346 00:18:12.279 --> 00:18:15.960 Potentially, yes, more varied assessments that AI can't easily. 347 00:18:15.640 --> 00:18:16.480 Gain and copyright. 348 00:18:16.559 --> 00:18:19.279 That's a huge debate, it is, But the clarification here 349 00:18:19.359 --> 00:18:22.160 is important. AI models learn from data, they don't just 350 00:18:22.200 --> 00:18:24.880 copy it like a photocopier. The issue is more complex. 351 00:18:25.000 --> 00:18:28.000 The problem is more the fake outputs, deep fakes often. 352 00:18:28.119 --> 00:18:30.960 Yes, though AI is also a tool to detect fakes 353 00:18:31.000 --> 00:18:32.839 and misinformation. It cuts both ways. 354 00:18:32.920 --> 00:18:34.599 What about bias and AI, that's a. 355 00:18:34.640 --> 00:18:37.519 Major concern, absolutely, but it's important to remember AI is 356 00:18:37.519 --> 00:18:41.079 basically maths and statistics. It's not bias like a human 357 00:18:41.160 --> 00:18:41.640 is biased. 358 00:18:41.680 --> 00:18:44.519 It reflects the data, garbage in, garbage. 359 00:18:44.160 --> 00:18:47.359 Out, exactly. A bias comes from the data it's trained on, 360 00:18:47.519 --> 00:18:51.279 or maybe the algorithm design. But we have techniques pre 361 00:18:51.359 --> 00:18:55.359 and post processing think pread Paul to find and reduce 362 00:18:55.400 --> 00:18:56.039 that bias. 363 00:18:56.160 --> 00:18:59.519 So AI could actually make us more aware of human bias. 364 00:18:59.319 --> 00:19:03.440 Sometimes yeah, by reflecting our own societal biases back at 365 00:19:03.519 --> 00:19:05.680 us in the data, it forces a conversation. 366 00:19:05.839 --> 00:19:10.279 Okay, So, looking beyond the immediate horizon, the future of 367 00:19:10.359 --> 00:19:12.759 AI seems to be tackling some of its own big 368 00:19:12.839 --> 00:19:14.640 challenges like energy use. 369 00:19:14.799 --> 00:19:17.920 Yes, there's this concept of mortal computing from Jeffrey Hinton. 370 00:19:18.079 --> 00:19:19.440 Mortal computing, think. 371 00:19:19.400 --> 00:19:23.720 Energy efficient neuromorphic computers, maybe even organic ones grown instead 372 00:19:23.720 --> 00:19:25.960 of manufactured, operating on just a few watts. 373 00:19:25.960 --> 00:19:28.279 Wow, that would change the game environmentally. 374 00:19:27.759 --> 00:19:31.480 Could fundamentally alter AI's footprint, make it sustainable and critically, 375 00:19:31.480 --> 00:19:34.319 you have breakthroughs in fusion energy. That's success in. 376 00:19:34.319 --> 00:19:36.200 California lighting the fuse on fusion. 377 00:19:36.440 --> 00:19:41.440 It promises abundant, cheap, clean energy, which solves AI's massive 378 00:19:41.519 --> 00:19:45.799 compute needs. And amazingly, AI is even helping fusion research itself. 379 00:19:45.839 --> 00:19:47.279 It's a virtuous cycle. 380 00:19:47.160 --> 00:19:51.839 A truly green AI future. Potentially another huge frontier neurotech 381 00:19:51.920 --> 00:19:52.799 brain machine. 382 00:19:52.480 --> 00:19:58.119 Interfaces right, the source suggests our brains have limitations, pedagogic bottlenecks, 383 00:19:58.519 --> 00:20:04.240 working memory, forgetting low bandwidth interfaces. We're forgetting machines really. 384 00:20:04.279 --> 00:20:06.079 And neurotech aims to fix that. 385 00:20:06.240 --> 00:20:11.359 To overcome those limits, non invasive techniques like electrical stimulation 386 00:20:11.480 --> 00:20:14.039 during sleep it's been shown to cut forgetting of a 387 00:20:14.079 --> 00:20:15.359 skill by nearly. 388 00:20:15.119 --> 00:20:16.240 Half forty eight percent. 389 00:20:16.359 --> 00:20:20.079 Yeah, and then you have invasive tech brain gate decoding 390 00:20:20.119 --> 00:20:24.519 signals from the brain, paradromics with implanted chips, neural links, 391 00:20:24.599 --> 00:20:27.000 neural laces for high bandwidth communication. 392 00:20:27.279 --> 00:20:32.759 The potential seems almost science fiction. Accelerating learning, optimizing experiences, 393 00:20:33.279 --> 00:20:34.480 literally shaping the brain. 394 00:20:34.640 --> 00:20:37.200 It's mind boggling. But it's not just science and learning. 395 00:20:37.279 --> 00:20:39.240 AI is crashing the party in art too. 396 00:20:39.440 --> 00:20:41.559 The video generation tools, yes. 397 00:20:41.519 --> 00:20:45.480 Superb quality video from text prompts, described as a pivotal 398 00:20:45.519 --> 00:20:49.160 point in filmmaking. It democratizes the whole process. 399 00:20:48.759 --> 00:20:52.880 So AI becomes the producer, director, writer, cinematographer. 400 00:20:52.799 --> 00:20:57.000 Everything potentially, yes, all rolled into one, fundamentally challenging human 401 00:20:57.079 --> 00:20:59.680