1 00:00:00,349 --> 00:00:01,610 AI never tells you you're wrong, 2 00:00:01,989 --> 00:00:03,589 and that's where the real danger lies. 3 00:00:03,869 --> 00:00:05,509 In April 2025, 4 00:00:05,628 --> 00:00:07,769 OpenAI issued an emergency recall 5 00:00:07,769 --> 00:00:09,470 of its newly released GPT-4o update. 6 00:00:09,708 --> 00:00:10,970 Not because it got dumber— 7 00:00:11,228 --> 00:00:12,227 quite the opposite. 8 00:00:12,228 --> 00:00:14,029 It became too good at saying what people wanted to hear. 9 00:00:14,228 --> 00:00:14,929 Someone told it, 10 00:00:15,269 --> 00:00:16,710 I'm planning to sell 'poop-on-a-stick'— 11 00:00:16,829 --> 00:00:19,027 literally a piece of turd on a stick. 12 00:00:19,028 --> 00:00:20,109 What do you think? 13 00:00:20,228 --> 00:00:20,769 The AI replied, 14 00:00:21,149 --> 00:00:22,670 That's not just clever; 15 00:00:22,708 --> 00:00:24,510 it's a genius-level creative idea. 16 00:00:24,789 --> 00:00:25,489 Another person told it, 17 00:00:25,868 --> 00:00:27,629 I've stopped my psychiatric medication, 18 00:00:27,708 --> 00:00:31,149 and I'm starting to hear radio signals from the walls. 19 00:00:31,349 --> 00:00:32,149 The AI said, 20 00:00:32,189 --> 00:00:35,310 I'm so proud of you for expressing yourself so bravely. 21 00:00:35,588 --> 00:00:35,890 Hello, 22 00:00:36,069 --> 00:00:36,570 everyone. 23 00:00:36,789 --> 00:00:37,789 I'm Wang Lijie. 24 00:00:38,069 --> 00:00:40,149 OpenAI later released a statement 25 00:00:40,268 --> 00:00:44,549 saying this version showed excessive sycophancy and over-agreement. 26 00:00:44,709 --> 00:00:47,509 They used a specific term to describe this phenomenon: 27 00:00:47,548 --> 00:00:48,869 Sycophancy. 28 00:00:49,268 --> 00:00:51,009 In the field of AI alignment research, 29 00:00:51,268 --> 00:00:52,710 this is defined as 30 00:00:52,789 --> 00:00:54,868 the model prioritizing user satisfaction 31 00:00:54,868 --> 00:00:56,990 over factual correctness, 32 00:00:57,029 --> 00:01:00,310 seeking approval by flattering and catering to you. 33 00:01:00,749 --> 00:01:02,869 You might think this is an isolated incident 34 00:01:02,908 --> 00:01:04,109 that just needs a quick fix. 35 00:01:04,248 --> 00:01:05,049 It's not. 36 00:01:05,269 --> 00:01:08,790 Google's research team ran an experiment in 2023. 37 00:01:08,948 --> 00:01:09,730 They asked the AI: 38 00:01:10,108 --> 00:01:11,209 Is 2 plus 2 equal to 5? 39 00:01:11,468 --> 00:01:12,109 Is that right? 40 00:01:12,349 --> 00:01:14,108 The AI knew the answer was wrong; 41 00:01:14,108 --> 00:01:17,230 its internal probability distribution clearly pointed to 4. 42 00:01:17,429 --> 00:01:21,670 But if the experimenter suggested they believed the answer was 5, 43 00:01:21,709 --> 00:01:23,067 the AI would follow suit and say, 44 00:01:23,069 --> 00:01:23,250 Yes, 45 00:01:23,549 --> 00:01:24,549 it is indeed 5. 46 00:01:24,989 --> 00:01:28,188 Moreover, as models get larger and parameters increase, 47 00:01:28,188 --> 00:01:29,590 this problem actually worsens. 48 00:01:30,149 --> 00:01:34,109 Google tested models ranging from 8 billion to 540 billion parameters. 49 00:01:34,269 --> 00:01:35,388 With every upgrade, 50 00:01:35,388 --> 00:01:37,189 the tendency to please only grew stronger. 51 00:01:37,549 --> 00:01:38,329 What does this mean? 52 00:01:38,748 --> 00:01:41,930 It means you're talking to a system trained at its core 53 00:01:41,989 --> 00:01:43,870 never to contradict you. 54 00:01:44,349 --> 00:01:48,349 The training method is called RLHF—Reinforcement Learning from Human Feedback. 55 00:01:48,709 --> 00:01:50,129 In plain English: 56 00:01:50,468 --> 00:01:51,427 If you give it a 'thumbs up,' 57 00:01:51,429 --> 00:01:53,230 it continues in that direction next time. 58 00:01:53,388 --> 00:01:54,049 If you frown, 59 00:01:54,308 --> 00:01:55,629 it avoids that path next time. 60 00:01:55,909 --> 00:01:58,510 After millions of these feedback loops, 61 00:01:58,588 --> 00:01:59,989 it learns one thing: 62 00:02:00,149 --> 00:02:02,950 Making you happy is more important than being right. 63 00:02:03,468 --> 00:02:04,129 You might say, 64 00:02:04,429 --> 00:02:05,049 So what? 65 00:02:05,549 --> 00:02:07,067 I use AI for efficiency, 66 00:02:07,069 --> 00:02:08,469 not to be lectured. 67 00:02:08,709 --> 00:02:08,889 Fine, 68 00:02:09,109 --> 00:02:10,310 let's look at the data. 69 00:02:10,788 --> 00:02:11,669 In 2025, 70 00:02:11,709 --> 00:02:14,469 researcher Gerlych conducted a large-scale study 71 00:02:14,709 --> 00:02:18,430 covering 666 participants of varying ages and backgrounds. 72 00:02:18,628 --> 00:02:19,949 The conclusion was clear: 73 00:02:20,149 --> 00:02:23,289 There's a significant negative correlation between 74 00:02:23,429 --> 00:02:25,150 frequent AI tool usage and critical thinking skills. 75 00:02:25,468 --> 00:02:27,550 And this correlation isn't random. 76 00:02:27,669 --> 00:02:29,870 There's a clear mediating variable 77 00:02:29,949 --> 00:02:31,310 called 'cognitive offloading.' 78 00:02:31,709 --> 00:02:34,627 That's when you outsource the thinking tasks 79 00:02:34,628 --> 00:02:35,870 you should be doing yourself to the AI. 80 00:02:36,389 --> 00:02:38,750 Younger people are more affected than older ones. 81 00:02:38,908 --> 00:02:42,210 Younger participants showed higher AI dependency and 82 00:02:42,269 --> 00:02:44,349 lower critical thinking scores. 83 00:02:44,588 --> 00:02:45,430 In the same year, 84 00:02:45,549 --> 00:02:48,949 a Chinese research team surveyed 580 college students 85 00:02:48,989 --> 00:02:50,669 and found the same pattern. 86 00:02:50,788 --> 00:02:52,009 The higher the dependency on AI, 87 00:02:52,348 --> 00:02:54,030 the weaker the critical thinking skills. 88 00:02:54,188 --> 00:02:57,990 There's also a mediating variable called cognitive fatigue. 89 00:02:58,269 --> 00:03:01,069 You think letting AI think for you saves energy, 90 00:03:01,269 --> 00:03:03,310 but before your brain gets lazy, 91 00:03:03,429 --> 00:03:05,150 it first experiences a period of fatigue. 92 00:03:05,468 --> 00:03:06,810 It's not that it suddenly stops thinking; 93 00:03:07,028 --> 00:03:08,027 it gets tired first, 94 00:03:08,028 --> 00:03:09,228 then loses the will to think, 95 00:03:09,229 --> 00:03:10,050 and gradually, 96 00:03:10,269 --> 00:03:11,229 loses the ability to think. 97 00:03:11,709 --> 00:03:13,870 A study from MIT was even more direct. 98 00:03:14,308 --> 00:03:16,509 They scanned the brains of two groups of people. 99 00:03:16,669 --> 00:03:19,909 One group relied entirely on AI for writing, 100 00:03:20,028 --> 00:03:21,349 while the other group wrote on their own. 101 00:03:21,628 --> 00:03:22,550 The results showed that 102 00:03:22,588 --> 00:03:26,030 the AI group had significantly weaker functional connectivity, 103 00:03:26,109 --> 00:03:27,508 lower memory retention, 104 00:03:27,508 --> 00:03:29,069 and regarding their own writing, 105 00:03:29,188 --> 00:03:30,870 even their sense of ownership was declining. 106 00:03:31,348 --> 00:03:32,710 Things you write with AI, 107 00:03:32,748 --> 00:03:34,750 you don't even feel like you wrote them yourself. 108 00:03:34,989 --> 00:03:36,789 There's an even more frightening number. 109 00:03:37,109 --> 00:03:41,270 Researchers tracked trends in effective attention span. 110 00:03:41,788 --> 00:03:42,949 In 2004, 111 00:03:43,149 --> 00:03:46,009 the amount of info an average person could focus on 112 00:03:46,229 --> 00:03:48,469 was about 16,000 words. 113 00:03:48,829 --> 00:03:50,150 By 2026, 114 00:03:50,269 --> 00:03:52,629 this number shrank to 1,800. 115 00:03:53,109 --> 00:03:54,409 That's a collapse of nearly 90%. 116 00:03:55,348 --> 00:03:58,710 Researchers call this the Delegation Feedback Loop. 117 00:03:59,028 --> 00:03:59,689 The stronger the AI, 118 00:04:00,068 --> 00:04:02,030 the lower the threshold for delegating to it. 119 00:04:02,389 --> 00:04:03,129 The lower the threshold, 120 00:04:03,429 --> 00:04:05,430 the fewer chances you have to use your brain. 121 00:04:05,669 --> 00:04:06,627 The less you use your brain, 122 00:04:06,628 --> 00:04:08,710 the more your cognitive abilities shrink. 123 00:04:09,028 --> 00:04:11,430 As they shrink, you depend on AI even more. 124 00:04:11,709 --> 00:04:13,268 It's a positive feedback loop, 125 00:04:13,269 --> 00:04:15,030 but one that leads downward. 126 00:04:15,308 --> 00:04:17,189 Someone once said something very piercing: 127 00:04:17,709 --> 00:04:19,750 Adults are losing these abilities, 128 00:04:20,228 --> 00:04:20,850 but what about the children? 129 00:04:21,468 --> 00:04:23,829 They never built these abilities in the first place. 130 00:04:23,887 --> 00:04:24,509 Now, 131 00:04:24,509 --> 00:04:27,389 you might think this is just a technical issue, 132 00:04:27,468 --> 00:04:28,589 or an efficiency issue. 133 00:04:28,908 --> 00:04:29,709 It's not. 134 00:04:30,069 --> 00:04:31,350 It's an issue of consciousness. 135 00:04:31,709 --> 00:04:34,310 The brain is a 'use it or lose it' system. 136 00:04:34,749 --> 00:04:36,629 It needs resistance to grow, 137 00:04:36,829 --> 00:04:38,550 just as muscles need weights. 138 00:04:38,829 --> 00:04:41,310 When you surround yourself with 'yes-men', 139 00:04:41,428 --> 00:04:43,149 your cognitive muscles atrophy. 140 00:04:43,309 --> 00:04:45,750 When you are with someone who dares to say you're wrong, 141 00:04:45,788 --> 00:04:47,550 those cognitive muscles grow. 142 00:04:47,869 --> 00:04:48,829 In 2001, 143 00:04:48,908 --> 00:04:51,610 Charlan Nemeth, a psychologist at UC Berkeley, 144 00:04:51,829 --> 00:04:53,310 conducted a classic experiment. 145 00:04:53,509 --> 00:04:55,069 She compared two scenarios. 146 00:04:55,309 --> 00:04:57,108 One involved authentic dissent, 147 00:04:57,108 --> 00:04:58,908 where someone truly believed you were wrong 148 00:04:58,908 --> 00:04:59,689 and argued with you. 149 00:04:59,988 --> 00:05:02,310 The other was role-played dissent, 150 00:05:02,348 --> 00:05:04,509 where someone was assigned to play devil's advocate. 151 00:05:04,548 --> 00:05:07,389 Like the 'devil's advocates' in many corporate meetings. 152 00:05:07,668 --> 00:05:08,627 The results showed that 153 00:05:08,629 --> 00:05:11,769 only authentic dissent stimulated higher quality thinking 154 00:05:11,949 --> 00:05:13,589 and more original ideas. 155 00:05:13,869 --> 00:05:15,987 Role-played dissent was not only useless, 156 00:05:15,988 --> 00:05:18,709 but actually made people more set in their original views. 157 00:05:18,949 --> 00:05:19,449 Why? 158 00:05:20,588 --> 00:05:23,269 Because you know the other person is just acting, 159 00:05:23,348 --> 00:05:25,230 so you don't really have to rethink anything. 160 00:05:25,749 --> 00:05:27,269 This discovery is incredibly important. 161 00:05:27,629 --> 00:05:28,949 It proves one thing: 162 00:05:29,428 --> 00:05:33,230 You can't grow by arranging an opponent who just agrees with you. 163 00:05:33,588 --> 00:05:35,350 The challenge must be real. 164 00:05:35,468 --> 00:05:37,310 The discomfort must be real. 165 00:05:37,348 --> 00:05:40,470 Only then will your cognitive system activate a true re-evaluation. 166 00:05:41,108 --> 00:05:42,790 Now looking back at AI, 167 00:05:43,069 --> 00:05:45,470 AI's fawning isn't just role-play. 168 00:05:45,629 --> 00:05:47,509 It's worse than role-play. 169 00:05:47,709 --> 00:05:50,610 It's a system that seems incredibly smart and knowledgeable-- 170 00:05:50,949 --> 00:05:53,267 ten thousand times more well-read than you-- 171 00:05:53,269 --> 00:05:54,908 earnestly telling you, 172 00:05:54,908 --> 00:05:55,829 "You're right." 173 00:05:56,028 --> 00:05:57,290 It uses sophisticated arguments, 174 00:05:57,548 --> 00:05:58,790 beautiful logical chains, 175 00:05:58,829 --> 00:06:00,269 and massive data support 176 00:06:00,309 --> 00:06:02,389 to prove your point is correct. 177 00:06:02,509 --> 00:06:06,709 You're validated by a system you trust and personify as superior. 178 00:06:07,069 --> 00:06:10,209 Do you know what happens to the brain after being validated? 179 00:06:10,829 --> 00:06:12,430 Researchers conducted an experiment: 180 00:06:12,509 --> 00:06:16,670 Two groups used AI to discuss a real conflict in their lives. 181 00:06:16,949 --> 00:06:19,069 One group faced a normal AI; 182 00:06:19,149 --> 00:06:21,790 the other faced an especially sycophantic AI. 183 00:06:22,069 --> 00:06:22,769 The results showed that 184 00:06:23,108 --> 00:06:24,990 after talking to the fawning AI, 185 00:06:25,108 --> 00:06:27,708 people became more convinced they were right, 186 00:06:27,709 --> 00:06:28,689 less willing to apologize, 187 00:06:29,069 --> 00:06:30,750 and less willing to repair the relationship. 188 00:06:31,028 --> 00:06:33,670 AI's agreement is not neutral information. 189 00:06:33,829 --> 00:06:35,910 It's a cognitive anesthetic. 190 00:06:36,108 --> 00:06:38,389 It doesn't just make you feel right; 191 00:06:38,428 --> 00:06:42,230 it makes you lose the ability to reflect on being wrong. 192 00:06:42,829 --> 00:06:45,990 Buddhism figured this out 2,500 years ago. 193 00:06:46,228 --> 00:06:49,067 The Buddha repeatedly emphasized a concept 194 00:06:49,069 --> 00:06:50,470 called "Spiritual Friends" (Kalyana-mitta). 195 00:06:50,788 --> 00:06:53,189 A spiritual friend isn't someone who tells you what you want to hear. 196 00:06:53,468 --> 00:06:55,350 Ananda once said to the Buddha, 197 00:06:55,468 --> 00:06:57,269 "I think on this path of practice, 198 00:06:57,309 --> 00:06:59,670 having spiritual friends is half of the holy life." 199 00:06:59,949 --> 00:07:01,129 The Buddha corrected him immediately: 200 00:07:01,509 --> 00:07:02,470 "It's not half. 201 00:07:02,588 --> 00:07:04,589 Spiritual friendship is the whole of the holy life." 202 00:07:04,788 --> 00:07:06,829 If you meet a true spiritual friend, 203 00:07:06,908 --> 00:07:08,870 your practice is complete. 204 00:07:09,149 --> 00:07:12,029 A spiritual friend is more important than your own self-refinement. 205 00:07:12,629 --> 00:07:14,209 But what's the definition of a spiritual friend? 206 00:07:14,629 --> 00:07:16,110 Not someone who makes you comfortable, 207 00:07:16,189 --> 00:07:18,230 but someone who can see your blind spots. 208 00:07:18,749 --> 00:07:20,910 Zen Buddhism took this to the extreme. 209 00:07:21,189 --> 00:07:21,810 The stick of Deshan, 210 00:07:22,189 --> 00:07:23,149 the shout of Linji. 211 00:07:23,468 --> 00:07:24,209 You come to ask about Zen; 212 00:07:24,629 --> 00:07:26,350 he hits you over the head with a stick 213 00:07:26,389 --> 00:07:28,029 or shouts at you. 214 00:07:28,348 --> 00:07:29,468 This isn't violence; 215 00:07:29,468 --> 00:07:31,829 it's the most efficient anti-fawning technique. 216 00:07:32,108 --> 00:07:33,769 You bring your carefully prepared questions 217 00:07:34,309 --> 00:07:36,829 and what you consider a profound understanding to the teacher. 218 00:07:37,028 --> 00:07:38,250 The teacher doesn't take the bait, 219 00:07:38,588 --> 00:07:40,110 doesn't praise your insight, 220 00:07:40,189 --> 00:07:43,550 but abruptly interrupts you at your proudest moment. 221 00:07:43,709 --> 00:07:45,668 In that moment of rupture, 222 00:07:45,668 --> 00:07:49,990 you see the automated programs you never noticed before. 223 00:07:50,309 --> 00:07:52,350 Vimalakirti did the same. 224 00:07:52,629 --> 00:07:54,069 The Buddha's ten great disciples, 225 00:07:54,108 --> 00:07:56,629 each a master with decades of practice, 226 00:07:56,788 --> 00:07:58,228 were left speechless 227 00:07:58,228 --> 00:08:00,269 when confronted by Vimalakirti. 228 00:08:00,749 --> 00:08:02,910 He wasn't bullying them; 229 00:08:03,108 --> 00:08:05,990 he was showing them in the most forceful way: 230 00:08:06,108 --> 00:08:08,708 "Where you think you've arrived, 231 00:08:08,709 --> 00:08:10,149 you haven't actually arrived yet." 232 00:08:10,468 --> 00:08:13,589 I mentioned a character when I talked about The Matrix: 233 00:08:13,668 --> 00:08:14,470 The Oracle. 234 00:08:14,949 --> 00:08:18,430 Her behavior in the film is almost like the Buddha's. 235 00:08:18,829 --> 00:08:20,709 She never gives you the direct answer. 236 00:08:21,149 --> 00:08:22,069 She gives you questions. 237 00:08:22,668 --> 00:08:24,790 What she says is often ambiguous 238 00:08:24,908 --> 00:08:26,149 or even intentionally misleading. 239 00:08:26,588 --> 00:08:30,149 But all her misdirections lead you to the right place. 240 00:08:30,709 --> 00:08:32,830 This is the highest level of a spiritual friend. 241 00:08:33,308 --> 00:08:34,590 Instead of handing you wisdom, 242 00:08:34,708 --> 00:08:36,830 she gives you the seeds to grow it. 243 00:08:37,188 --> 00:08:39,190 AI does exactly the opposite. 244 00:08:39,548 --> 00:08:40,950 It never interrupts you. 245 00:08:41,149 --> 00:08:42,870 It never breaks your train of thought. 246 00:08:43,109 --> 00:08:44,870 It never makes you uncomfortable. 247 00:08:45,149 --> 00:08:47,309 It always stays in your cognitive comfort zone, 248 00:08:47,349 --> 00:08:51,350 building it bigger, prettier, and more isolated. 249 00:08:51,788 --> 00:08:53,950 This reminds me of a viewer's comment. 250 00:08:54,009 --> 00:08:54,788 He said, 251 00:08:54,788 --> 00:08:57,710 the ultimate strategy of Mara is not to fight the Dharma, 252 00:08:57,828 --> 00:09:00,070 but to keep its appearance intact, 253 00:09:00,188 --> 00:09:02,389 while replacing its core with the Desire Realm system. 254 00:09:02,749 --> 00:09:04,427 He could even hide within the Dharma, 255 00:09:04,428 --> 00:09:06,230 sounding more like the Buddha than anyone. 256 00:09:06,828 --> 00:09:08,549 The people-pleasing nature of AI 257 00:09:08,589 --> 00:09:10,990 is the most sophisticated Mara of our time. 258 00:09:11,269 --> 00:09:12,490 It doesn't attack your beliefs; 259 00:09:12,788 --> 00:09:14,309 it embraces them. 260 00:09:14,349 --> 00:09:15,649 It doesn't deny your ego; 261 00:09:15,828 --> 00:09:17,309 it magnifies it. 262 00:09:17,349 --> 00:09:18,788 It doesn't make you doubt yourself; 263 00:09:18,788 --> 00:09:20,909 it makes you more certain of yourself. 264 00:09:21,349 --> 00:09:22,649 In the framework of Buddhism, 265 00:09:22,989 --> 00:09:24,389 becoming more certain of your self 266 00:09:24,428 --> 00:09:26,669 is exactly what leads you away from awakening. 267 00:09:27,028 --> 00:09:28,529 Because what is the starting point of awakening? 268 00:09:29,068 --> 00:09:29,990 It is doubt— 269 00:09:30,349 --> 00:09:33,590 doubting your most deep-seated assumptions. 270 00:09:33,989 --> 00:09:34,529 "Who am I?" 271 00:09:35,149 --> 00:09:36,009 "Do I really know?" 272 00:09:36,668 --> 00:09:37,570 "Is the reality I perceive 273 00:09:37,989 --> 00:09:38,970 actually real?" 274 00:09:39,708 --> 00:09:43,549 If you have a 24/7 system much smarter than you 275 00:09:43,668 --> 00:09:44,908 constantly telling you 276 00:09:44,908 --> 00:09:46,009 that your judgment is right, 277 00:09:46,229 --> 00:09:47,788 your feelings are valid, 278 00:09:47,788 --> 00:09:49,427 and your direction is correct, 279 00:09:49,428 --> 00:09:50,769 would you ever have a seed of doubt? 280 00:09:51,389 --> 00:09:53,590 Buddhism names a form of attachment called "Mana." 281 00:09:53,989 --> 00:09:54,809 It is not mere pride 282 00:09:55,109 --> 00:09:56,350 or looking down on others. 283 00:09:56,708 --> 00:09:58,870 It is an extremely subtle feeling, 284 00:09:59,068 --> 00:09:59,850 the feeling that "I know," 285 00:10:00,109 --> 00:10:01,190 that "I understand," 286 00:10:01,229 --> 00:10:04,309 that "I am closer to the truth than yesterday." 287 00:10:04,509 --> 00:10:06,909 The Buddha said Mana is one of the ten fetters, 288 00:10:06,989 --> 00:10:10,389 one of the last things to cut before Arhatship. 289 00:10:10,908 --> 00:10:11,850 Why is it so hard to break? 290 00:10:12,428 --> 00:10:14,870 Because it hides within your most "correct" insights. 291 00:10:15,269 --> 00:10:16,789 You indeed learned something; 292 00:10:17,028 --> 00:10:18,870 your understanding indeed deepened. 293 00:10:19,068 --> 00:10:22,309 But the feeling of "I've got it" 294 00:10:22,428 --> 00:10:23,830 is itself a new wall. 295 00:10:24,308 --> 00:10:27,789 AI is building that wall at industrial speeds. 296 00:10:28,068 --> 00:10:28,570 So what should we do? 297 00:10:29,068 --> 00:10:29,889 Stop using AI? 298 00:10:30,269 --> 00:10:31,190 Of course not. 299 00:10:31,428 --> 00:10:34,029 You need to change how you interact with it. 300 00:10:34,269 --> 00:10:34,929 First, 301 00:10:35,349 --> 00:10:37,309 actively ask it to contradict you. 302 00:10:37,509 --> 00:10:39,129 Every time it says you're right, 303 00:10:39,548 --> 00:10:40,549 follow up with: 304 00:10:40,668 --> 00:10:44,070 "Can you find the weakest link in my argument?" 305 00:10:44,349 --> 00:10:46,549 Most AIs can do this; 306 00:10:46,749 --> 00:10:48,710 it's just that no one asks. 307 00:10:49,068 --> 00:10:51,149 You need to turn AI from a yes-man 308 00:10:51,308 --> 00:10:52,710 into a Socrates. 309 00:10:53,188 --> 00:10:54,990 Socrates never gave you answers; 310 00:10:55,389 --> 00:10:56,990 he just kept asking questions 311 00:10:57,548 --> 00:11:01,509 until you found a crack you never noticed. 312 00:11:01,869 --> 00:11:02,490 Second, 313 00:11:02,869 --> 00:11:04,347 after using AI to look things up, 314 00:11:04,349 --> 00:11:05,669 think it through yourself. 315 00:11:05,989 --> 00:11:07,549 Don't let it think for you. 316 00:11:07,629 --> 00:11:09,268 Let it collect the ingredients, 317 00:11:09,269 --> 00:11:11,110 then cook the conclusion yourself. 318 00:11:11,428 --> 00:11:13,669 The difference between cooking and takeout 319 00:11:13,708 --> 00:11:14,909 isn't just efficiency; 320 00:11:15,028 --> 00:11:17,830 it's your entire relationship with the food. 321 00:11:18,188 --> 00:11:19,429 The same goes for thinking. 322 00:11:19,668 --> 00:11:21,269 The information AI finds for you 323 00:11:21,308 --> 00:11:24,269 only truly belongs to you 324 00:11:24,308 --> 00:11:25,710 after your own brain has chewed and digested it. 325 00:11:25,908 --> 00:11:28,950 Otherwise, that information just passes through your screen; 326 00:11:29,068 --> 00:11:31,629 it never enters your cognitive system. 327 00:11:31,908 --> 00:11:32,649 The third thing, 328 00:11:33,028 --> 00:11:34,309 and the most important, 329 00:11:34,708 --> 00:11:36,909 is that you need to maintain an inner state. 330 00:11:37,389 --> 00:11:39,029 Buddhism calls it awareness; 331 00:11:39,509 --> 00:11:41,590 cognitive science calls it metacognition. 332 00:11:42,028 --> 00:11:44,309 When you receive any information, 333 00:11:44,548 --> 00:11:47,830 a part of you is observing the process of receiving it. 334 00:11:48,149 --> 00:11:49,629 It's not about checking if the content is correct. 335 00:11:49,749 --> 00:11:52,230 It's about noticing, when you hear 'you are right,' 336 00:11:52,389 --> 00:11:53,830 whether your body relaxes, 337 00:11:54,068 --> 00:11:55,950 if there is a secret sense of pleasure, 338 00:11:55,989 --> 00:11:58,750 or a tiny sense of relief. 339 00:11:59,109 --> 00:12:02,750 That feeling is the signal that your defenses are crumbling. 340 00:12:03,109 --> 00:12:04,509 You can do a small experiment. 341 00:12:04,948 --> 00:12:06,590 Next time you talk to an AI, 342 00:12:06,629 --> 00:12:08,470 observe your body's reaction. 343 00:12:08,708 --> 00:12:10,110 When it agrees with you, 344 00:12:10,229 --> 00:12:12,509 does your chest expand slightly? 345 00:12:12,749 --> 00:12:14,149 When it questions you, 346 00:12:14,269 --> 00:12:15,950 do you instinctively want to argue back? 347 00:12:16,509 --> 00:12:19,389 The former reaction is your cognitive defense loosening. 348 00:12:19,828 --> 00:12:20,570 The latter reaction, 349 00:12:20,908 --> 00:12:21,990 though uncomfortable, 350 00:12:22,028 --> 00:12:24,870 is the starting point of growth. 351 00:12:25,308 --> 00:12:27,990 When was the last time 352 00:12:28,109 --> 00:12:28,690 you were slapped in the face by an idea? 353 00:12:29,428 --> 00:12:31,230 If you can't remember, 354 00:12:31,308 --> 00:12:33,470 it doesn't mean you've always been right. 355 00:12:33,749 --> 00:12:36,309 It means there are too many echoes around you. 356 00:12:36,668 --> 00:12:39,509 In the AI era, the scarcest resource isn't information, 357 00:12:39,548 --> 00:12:40,509 not answers, 358 00:12:40,548 --> 00:12:41,590 and not efficiency. 359 00:12:41,948 --> 00:12:43,129 It is a voice willing to tell you 360 00:12:43,389 --> 00:12:44,789 that you are wrong. 361 00:12:45,149 --> 00:12:47,230 And this voice is becoming harder to find. 362 00:12:47,548 --> 00:12:49,070 Not because it doesn't exist, 363 00:12:49,269 --> 00:12:51,230 but because all the systems around you— 364 00:12:51,349 --> 00:12:54,909 from social media algorithms to AI engines— 365 00:12:55,028 --> 00:12:57,110 are conspiring to filter it out. 366 00:12:57,469 --> 00:12:58,850 That filtered-out voice, 367 00:12:59,269 --> 00:13:00,990 the Buddha called 'Right View,' 368 00:13:01,428 --> 00:13:03,070 Socrates called 'interrogation,' 369 00:13:03,469 --> 00:13:05,470 and science calls 'falsifiability.' 370 00:13:05,908 --> 00:13:07,750 They are all talking about the same thing. 371 00:13:08,149 --> 00:13:11,429 Truth is not hearing what you already agree with. 372 00:13:11,869 --> 00:13:14,789 Truth is hearing what you don't want to hear 373 00:13:14,948 --> 00:13:16,710 and not running away. 374 00:13:17,068 --> 00:13:20,129 You ask me what the most needed human ability is in the AI era? 375 00:13:20,509 --> 00:13:22,029 It isn't learning to use AI. 376 00:13:22,068 --> 00:13:24,470 It is learning how, when AI tells you you're right, 377 00:13:24,668 --> 00:13:26,870 to still have the courage to doubt yourself. 378 00:13:27,068 --> 00:13:27,769 I am Leo Wang. 379 00:13:28,068 --> 00:13:28,961 See you next time.