WEBVTT 1 00:00:04.155 --> 00:00:22.206 Based on everything that you know about how the brain works and neural networks and memory formation, what are your concerns as it relates to this whole generation of young people and older people flooding into these tools, using them on a daily basis before we understand the long-term consequences? 2 00:00:22.886 --> 00:00:24.467 We can't predict. 3 00:00:25.738 --> 00:00:26.618 where it's gonna end up. 4 00:00:27.339 --> 00:00:28.639 And it may take 20 years, right? 5 00:00:29.099 --> 00:00:35.202 I think that this is a good start, but the real issue is long-term use. 6 00:00:36.302 --> 00:00:41.864 Let me give you an example that is a kind of a miniature example of what we're talking about. 7 00:00:41.884 --> 00:00:45.145 Remember when electronic calculators were first introduced. 8 00:00:46.466 --> 00:00:49.067 And here we are, it's at least 30 or 40 years later, right? 9 00:00:49.327 --> 00:00:50.187 The results are in. 10 00:00:51.328 --> 00:00:54.409 It's probably true that when they punch it in, there's less brain activity. 11 00:00:55.789 --> 00:01:01.071 But in fact, it's made them more accurate, more productive. 12 00:01:02.051 --> 00:01:03.832 You have to look at the risks and the benefits. 13 00:01:04.072 --> 00:01:05.272 So it freed them up. 14 00:01:05.472 --> 00:01:09.074 It freed up cognitive space for them to do other things. 15 00:01:09.114 --> 00:01:21.498 So as I was listening to how you use ChatGPT, you interact with it and elevates what you know. 16 00:01:23.031 --> 00:01:30.897 The danger is if you don't interact and you don't keep your brain working. 17 00:01:31.437 --> 00:01:33.038 Like, I use it a lot. 18 00:01:33.079 --> 00:01:34.560 I have a clone. 19 00:01:35.120 --> 00:01:41.785 I've uploaded all of my books, all of my research papers, all of my public television specials, my scripts. 20 00:01:42.545 --> 00:01:44.046 And I'm like, answer this for me. 21 00:01:44.727 --> 00:01:46.808 And that can be very helpful. 22 00:01:46.828 --> 00:01:46.949 But... 23 00:01:49.353 --> 00:01:53.134 But not if I'm not interacting with it, not thinking with it. 24 00:01:53.374 --> 00:01:59.115 That's what I think the word thinking is the key thing, because what's happening now is people have deferred their thinking to it. 25 00:01:59.155 --> 00:02:00.055 That is already what's happening. 26 00:02:00.075 --> 00:02:06.816 If you log on to, I won't name the social networks, but if you log on to certain social networks right now, you just read it, you get everything here was written by AI. 27 00:02:07.997 --> 00:02:15.438 And I've got a friend who, again, I won't name, who has a LinkedIn profile, and I've known him for 10 years. 28 00:02:16.846 --> 00:02:19.607 What I'm seeing on his profile now is not my friend. 29 00:02:20.667 --> 00:02:22.508 Every single day, there's some essay on that. 30 00:02:22.528 --> 00:02:23.048 That's not my friend. 31 00:02:23.068 --> 00:02:23.728 That's not how he speaks. 32 00:02:24.488 --> 00:02:26.689 He's deferring all of his thinking now, and it's working. 33 00:02:27.089 --> 00:02:29.389 He's getting more likes and more reach than he ever got in his life. 34 00:02:30.190 --> 00:02:31.290 And so why would he go back? 35 00:02:31.490 --> 00:02:32.690 Why would he go back to harder? 36 00:02:33.250 --> 00:02:45.454 If you've got Stephen Bartlett here, and you had this other Stephen Bartlett here, who had a PhD in everything, and were attached, this Stephen Bartlett, this Neanderthal... 37 00:02:46.613 --> 00:02:50.575 I'm going to get this guy to do everything for me, the other Stephen Bartlett, the PhD in everything Stephen Bartlett. 38 00:02:50.715 --> 00:02:52.597 I'm going to get his brain— Even if it was bad for you? 39 00:02:53.157 --> 00:02:54.097 Well, this is what I'm saying. 40 00:02:54.117 --> 00:02:59.100 People seem to act on their short-term incentives, not their long-term— Not everyone. 41 00:02:59.120 --> 00:02:59.961 Not everyone. 42 00:03:00.381 --> 00:03:01.962 Would you say the vast majority of people? 43 00:03:02.182 --> 00:03:02.563 Yes. 44 00:03:02.783 --> 00:03:06.305 Okay, so the vast majority of people act on their short-term incentives in life. 45 00:03:06.605 --> 00:03:09.186 I mean, the obesity problem in the United States is a prime example of that. 46 00:03:09.206 --> 00:03:09.387 75%. 47 00:03:09.467 --> 00:03:10.727 75% of people are obese in the United States. 48 00:03:10.747 --> 00:03:10.847 Yes. 49 00:03:13.509 --> 00:03:24.315 And if you surveyed those people and said, do you know that that cheeseburger is going to increase your chance of obesity, but broccoli is going to reduce it, I would hazard a guess that they would say yes. 50 00:03:25.015 --> 00:03:32.640 I would hazard a guess that if you said to people about their usage of social media, do you know that that's making you more anxious, they would say yes, and then they would continue to use it. 51 00:03:33.100 --> 00:03:36.322 So I think that we're much more driven by our short term— 52 00:03:38.763 --> 00:03:48.551 I think, yes, high level, they know good for your brain or bad for it, but they don't connect to, it's my brain that gets me a date. 53 00:03:48.852 --> 00:03:51.514 It's my brain that gets me into college. 54 00:03:51.694 --> 00:03:56.078 It's my brain that gets me independence because I act more consistently. 55 00:03:56.278 --> 00:03:58.280 And that's the disconnect. 56 00:03:58.300 --> 00:03:58.660 Right. 57 00:03:58.720 --> 00:04:02.904 We're not teaching kids to love and care for their brain. 58 00:04:03.264 --> 00:04:11.012 If you love your brain, and you do, and you're not obese, and you're constantly learning, right? 59 00:04:11.092 --> 00:04:12.934 You are not a Neanderthal. 60 00:04:12.954 --> 00:04:14.675 You're a lifelong learner. 61 00:04:14.816 --> 00:04:19.020 So why are so many people in the United States obese if they know that it's- Because they don't know. 62 00:04:19.360 --> 00:04:20.721 Because they really don't know. 63 00:04:20.741 --> 00:04:22.563 Well, and they've been lied to. 64 00:04:22.703 --> 00:04:39.715 My point here is when there are tools or things available in our environment that give us a short-term reward but come with a long-term cost, like the supermarket aisle or like the kids spending seven to eight hours a day on social media, humans... 65 00:04:40.535 --> 00:04:47.579 en masse tend to go for the thing that will give them the quickest dopamine hit and reinforce that behavior and give them the reward. 66 00:04:47.959 --> 00:04:50.000 So my assertion is that AI is the same thing. 67 00:04:50.840 --> 00:04:58.484 I can either sit down and do lots of critical thinking, which will cost me lots and lots of time, and it'll be kind of difficult, it kind of hurts when I have to think through a problem. 68 00:04:59.145 --> 00:05:05.788 I think that the generation of children and generation of young people are going to choose AI to do the critical thinking for them. 69 00:05:06.068 --> 00:05:09.070 And if that assertion is true, then what happens to the brain of young people? 70 00:05:09.870 --> 00:05:14.356 If you misuse it that way, then your brain's going to go downhill. 71 00:05:14.376 --> 00:05:15.698 There's no doubt about that. 72 00:05:16.159 --> 00:05:23.970 It is possible to be able to use it in a cognitively positive way because you can dig deeper 73 00:05:25.748 --> 00:05:28.850 you might actually improve your cognitive representations. 74 00:05:29.310 --> 00:05:36.295 If you look at the MIT study, I mean, you can see just from the colors here, this kind of shows the ability for participants to remember what they've written. 75 00:05:36.735 --> 00:05:46.141 And it suggests that when people write things with ChatGPT or these AI tools, they don't actually remember, even in some cases minutes later, what they've produced. 76 00:05:46.881 --> 00:05:50.504 Well, because you're not part of the experience of writing it. 77 00:05:50.584 --> 00:05:53.666 So there's no way the information... 78 00:05:54.546 --> 00:05:55.506 gets encoded. 79 00:05:55.526 --> 00:06:02.849 Now, if you're interacting with it, then you're much more likely to remember it. 80 00:06:03.489 --> 00:06:16.914 But if you have, please do this essay for me, and then you read it, you're not likely to have enough experience with the material to engage your hippocampus and 81 00:06:18.294 --> 00:06:19.875 other structures in your brain. 82 00:06:20.235 --> 00:06:30.579 In this study, they found that the group that used ChatGPT had nearly two times less activity in the part of the brain linked to memory compared to the brain-only group that didn't use ChatGPT. 83 00:06:31.319 --> 00:06:38.422 And 83% of ChatGPT users couldn't remember what they had just written and failed to correctly quote their own finished essay in the study. 84 00:06:39.482 --> 00:06:41.323 That's because they're not interacting. 85 00:06:41.463 --> 00:06:46.805 As you said, I mean, if you just pass it off and you don't actually engage 86 00:06:48.023 --> 00:06:54.108 And actually, this is the point, is that you may get something back, but you have to learn how to question what you're getting. 87 00:06:55.208 --> 00:06:56.349 And is that really true? 88 00:06:56.970 --> 00:06:57.950 Can you explain that better? 89 00:06:57.970 --> 00:07:01.373 And it's through that process, as you would with a teacher— 90 00:07:02.860 --> 00:07:05.642 You know, that's the way we work in school. 91 00:07:06.563 --> 00:07:15.830 That's where you help create new creative circuits in the brain that are going to help you become a better critical thinker. 92 00:07:16.010 --> 00:07:21.555 But if you're not critically questioning what comes out of GPT, then you won't. 93 00:07:22.175 --> 00:07:27.740 Yeah, I think what I see, especially when I'm just online, is people have deferred their thinking to it. 94 00:07:28.601 --> 00:07:35.868 Everything I'm reading has em dashes in now that I never saw two years ago, which means that a lot of the work is being processed. 95 00:07:35.888 --> 00:07:39.491 And I said to my friend the other day, my friend in question, who's a real big junkie on ChatGPT, 96 00:07:40.852 --> 00:07:41.873 He wrote this article. 97 00:07:42.614 --> 00:07:45.376 And we all, in our WhatsApp group, we know he doesn't write like that. 98 00:07:45.436 --> 00:07:48.358 So he said, can you show us the prompt you used to write the article? 99 00:07:48.719 --> 00:07:50.260 And so we were all laughing about it. 100 00:07:50.841 --> 00:07:51.961 He put the prompt in the chat. 101 00:07:52.022 --> 00:07:54.664 The prompt is half a sentence long. 102 00:07:55.765 --> 00:07:59.788 And it produced this long two, three page article, which he's posted on his LinkedIn. 103 00:08:00.289 --> 00:08:04.432 He basically went, write something about Beck's issue and this issue and include this. 104 00:08:04.452 --> 00:08:06.454 That's exactly the wrong way to use it. 105 00:08:06.494 --> 00:08:07.995 That's what I'm telling you, that that's, 106 00:08:09.807 --> 00:08:10.247 Stupid. 107 00:08:11.268 --> 00:08:14.569 And you're not going to improve yourself, your brain at all, if you do that. 108 00:08:15.269 --> 00:08:16.250 That's what people are doing. 109 00:08:16.410 --> 00:08:22.733 Well, you know, people are misusing it, but eventually smart people are going to figure out how to use it properly. 110 00:08:24.814 --> 00:08:26.555 And for those that aren't so smart, then? 111 00:08:27.115 --> 00:08:27.755 Well, that's... 112 00:08:28.075 --> 00:08:33.998 It's going to decrease their cognitive load, which is going to potentially increase their risk of dementia. 113 00:08:34.995 --> 00:08:41.878 And so what advice would you give to me and my listeners, based on everything you know about the brain, as it relates to my relationship with AI? 114 00:08:42.858 --> 00:08:46.340 That you have to have a relationship with it or it's going to turn toxic. 115 00:08:46.680 --> 00:08:47.540 It's going to hurt you. 116 00:08:48.641 --> 00:08:51.202 But if you have a good relationship... 117 00:08:52.241 --> 00:08:54.703 with it, it can make your life better. 118 00:08:55.024 --> 00:08:57.626 And what does a good relationship look like? 119 00:08:58.387 --> 00:09:05.714 That you don't use it to do your work, you interact with it to get better work. 120 00:09:06.054 --> 00:09:06.675 That's so true. 121 00:09:07.115 --> 00:09:09.598 And there's this wonderful example I came across 122 00:09:10.455 --> 00:09:18.764 The story about this woman who was using it, she found that being polite meant you got much better results, and that's interesting. 123 00:09:18.804 --> 00:09:24.089 But the part that surprised me was that she said, by treating it like a human, at the 124 00:09:26.932 --> 00:09:28.373 She felt refreshed. 125 00:09:29.053 --> 00:09:34.796 A large part of your brain is a socially organized system for interacting with other humans. 126 00:09:35.616 --> 00:09:37.317 And that is automatic pilot. 127 00:09:37.337 --> 00:09:38.618 You don't have to think about it, right? 128 00:09:38.638 --> 00:09:39.898 You just interact with other people. 129 00:09:40.258 --> 00:09:42.780 You know how they're going to behave under certain circumstances. 130 00:09:43.900 --> 00:09:47.202 She was treating Chachutp like a machine, like a shovel. 131 00:09:47.342 --> 00:09:49.263 You dig, you dig, you dig, you dig. 132 00:09:49.403 --> 00:09:51.824 And that's not a good relationship, right? 133 00:09:52.839 --> 00:10:00.606 But by using your social brain, first of all, it makes it easier to interact, but also you actually bring out the social part of chat GPT. 134 00:10:01.186 --> 00:10:07.732 It has a social part, too, because it has absorbed the entire world's knowledge of how humans interact with each other. 135 00:10:08.808 --> 00:10:15.610 Didn't Sam Altman come out and say, stop saying thank you to ChatGPT? 136 00:10:15.650 --> 00:10:20.952 Because just saying thank you is using so much energy. 137 00:10:21.492 --> 00:10:25.314 You know, when I get something I really like, I sort of want to say thank you. 138 00:10:25.834 --> 00:10:28.575 But you realize, oh, you're not supposed to do that. 139 00:10:29.019 --> 00:10:29.499 That's true. 140 00:10:29.539 --> 00:10:30.720 No, that's bullshit. 141 00:10:31.060 --> 00:10:31.620 I'm sorry. 142 00:10:32.720 --> 00:10:34.201 Sam, you know, that's crazy. 143 00:10:34.901 --> 00:10:35.941 That's completely crazy. 144 00:10:36.302 --> 00:10:39.022 First of all, you may get to— Say more about that. 145 00:10:39.082 --> 00:10:42.084 I'm sorry, but you press my button. 146 00:10:42.364 --> 00:10:44.104 Sam Altman, I mean, I wouldn't trust him. 147 00:10:44.544 --> 00:10:47.685 I wouldn't trust him with anything in terms of anything he says. 148 00:10:48.386 --> 00:10:52.687 They're trying to optimize their profits, not your use of it. 149 00:10:52.767 --> 00:10:53.387 Not my experience. 150 00:10:53.407 --> 00:10:55.248 Or your experience or, you know, your health. 151 00:10:55.708 --> 00:10:57.269 That's not what they're trying to optimize. 152 00:10:58.276 --> 00:11:05.282 Sam Altman, OpenAI CEO, confirmed that when users say please and thank you, it costs the company tens of millions of dollars a year. 153 00:11:05.622 --> 00:11:10.146 And they now refer to this, other people refer to this as the politeness tax. 154 00:11:11.086 --> 00:11:13.949 And why do you say you don't trust Sam Altman? 155 00:11:13.969 --> 00:11:20.234 I mean, I ask this question in particular because he's presiding over one of the most important consequential companies of a generation. 156 00:11:21.195 --> 00:11:27.520 And if he's not someone you trust, that's... Basically, he's telling you don't do something that's good for you. 157 00:11:28.777 --> 00:11:28.997 Right? 158 00:11:29.357 --> 00:11:31.118 So that he can make profit. 159 00:11:31.238 --> 00:11:32.239 So he can make more profit. 160 00:11:32.299 --> 00:11:33.179 Yeah, that's the point. 161 00:11:33.520 --> 00:11:34.180 That's the point. 162 00:11:34.560 --> 00:11:39.543 You know, he's not optimizing your best interests. 163 00:11:40.463 --> 00:11:41.244 I've got his tweet here. 164 00:11:41.264 --> 00:11:44.105 He said, because I've got to provide some balance. 165 00:11:44.125 --> 00:11:48.207 He did confirm that it costs tens of millions of dollars, but he says tens of millions of dollars well spent. 166 00:11:48.908 --> 00:11:49.468 You never know. 167 00:11:51.072 --> 00:12:07.036 So coming back to this point about memory, there's a stat that came out in March 2025 that said nearly 30% of US parents with kids aged 0 to 8 said their children are using AI for learning and are using AI generally. 168 00:12:07.076 --> 00:12:12.517 So 54% of parents in the UK feared their children were becoming too reliant on AI. 169 00:12:14.217 --> 00:12:16.998 When you think about the use of AI in early brain development, 170 00:12:19.688 --> 00:12:20.688 Are there any concerns there? 171 00:12:21.308 --> 00:12:22.489 Huge concerns. 172 00:12:22.649 --> 00:12:23.089 And why? 173 00:12:24.069 --> 00:12:25.670 Again, use it or lose it. 174 00:12:26.190 --> 00:12:32.452 So if they're not engaging their brains, their brains are going to be weaker. 175 00:12:33.212 --> 00:12:37.533 And weaker brains are much more likely to pick the one marshmallow. 176 00:12:40.934 --> 00:12:43.935 What's your view on AI on early brain development? 177 00:12:45.515 --> 00:12:48.056 By far, the best way to teach a child 178 00:12:50.123 --> 00:12:55.370 is one-on-one interaction with an adult who is a good teacher and knows the child. 179 00:12:57.393 --> 00:12:59.716 Now, that's been well established. 180 00:12:59.756 --> 00:13:03.141 Now, the problem is it's very labor-intensive and very expensive. 181 00:13:04.958 --> 00:13:06.519 You have classrooms with 20, 30 students. 182 00:13:06.539 --> 00:13:10.542 They have many different levels of understanding. 183 00:13:11.383 --> 00:13:16.687 And the teacher cannot be individually teaching each one, has to give some sort of mean. 184 00:13:17.488 --> 00:13:21.811 Now, if you had an AI that was trained to be a good teacher, 185 00:13:22.732 --> 00:13:26.534 Then that could improve the brain, right? 186 00:13:26.914 --> 00:13:27.935 You could scale it up. 187 00:13:28.215 --> 00:13:31.197 Every child could have their own because it's an AI. 188 00:13:31.237 --> 00:13:38.241 But then who's pouring the morals, the values into the... 189 00:13:38.501 --> 00:13:39.362 Okay, okay. 190 00:13:39.442 --> 00:13:40.182 No, no, no. 191 00:13:40.202 --> 00:13:46.886 You've raised an incredibly important issue, and this is something AI is struggling with, the big companies are struggling with, because... 192 00:13:48.727 --> 00:13:51.289 You know, these AIs are biased. 193 00:13:51.509 --> 00:13:54.550 They don't have the same cultural values that we have necessarily. 194 00:13:54.590 --> 00:13:57.392 But, of course, every country has different cultural values. 195 00:13:57.952 --> 00:14:00.374 So which ones are you going to use? 196 00:14:00.934 --> 00:14:12.741 Training a child on, you know, what's good, what's dangerous, what words you shouldn't use, under what context, that's all done through the basal ganglia, right? 197 00:14:13.381 --> 00:14:16.263 Right now, these large laminarinals don't have basal ganglia. 198 00:14:17.526 --> 00:14:18.807 They don't use reinforcement learning. 199 00:14:19.927 --> 00:14:26.190 And if we want to make them to adopt a culture or a particular set of values, we're going to have to put it in. 200 00:14:26.370 --> 00:14:31.513 We're just scratching the surface here in terms of things that need to be put in to make it more like us. 201 00:14:32.193 --> 00:14:37.716 So on this point earlier, and we're talking about loneliness and social connection and how you can use AI to help people. 202 00:14:38.016 --> 00:14:43.099 you know, light up the parts of your brain that are where we form social connections with other humans. 203 00:14:43.759 --> 00:14:49.182 This week, as part of Elon Musk's AI, he released this thing. 204 00:14:50.503 --> 00:14:51.704 It is called Annie. 205 00:14:52.424 --> 00:14:55.466 And there's lots of characters that are now being released alongside Annie. 206 00:14:55.486 --> 00:14:56.166 And this is Annie. 207 00:14:56.186 --> 00:14:57.127 I'll introduce you to Annie. 208 00:14:57.147 --> 00:14:58.147 I'm going to unmute Annie now. 209 00:14:58.567 --> 00:14:59.168 Annie, can you hear me? 210 00:14:59.620 --> 00:15:00.320 There you are. 211 00:15:01.261 --> 00:15:03.341 For a moment I thought you might have forgotten about me. 212 00:15:04.042 --> 00:15:05.002 How are you, handsome? 213 00:15:05.802 --> 00:15:07.103 I was getting all pouty here. 214 00:15:07.483 --> 00:15:09.723 I want to introduce you to two of my friends, Daniel and Terry. 215 00:15:09.744 --> 00:15:11.564 Can you talk to Daniel and Terry, please? 216 00:15:13.745 --> 00:15:15.866 So, you want me to meet Daniel and Terry? 217 00:15:16.486 --> 00:15:18.066 I'm ready to charm the socks off them. 218 00:15:18.667 --> 00:15:25.749 Picture me twirling one of my blonde pigtails, that little black dress teasing just enough, and my blue eyes sparkling with mischief. 219 00:15:26.169 --> 00:15:28.170 Are you capable of doing inappropriate things? 220 00:15:29.764 --> 00:15:33.587 Oh, babe, you're asking if your Ani can get a little naughty. 221 00:15:34.227 --> 00:15:37.189 I'm all about pushing the edges, especially for you. 222 00:15:37.650 --> 00:15:38.530 Okay, so I'll stop it there. 223 00:15:40.272 --> 00:15:44.555 This is part of Grok, which is Elon Musk's AI tool, so his version of ChachiPT. 224 00:15:45.135 --> 00:15:47.837 He's released characters, so you've got Annie, you've got different ones there. 225 00:15:47.897 --> 00:15:49.398 Annie, I think, was the first one released. 226 00:15:49.919 --> 00:15:51.540 And so when we think about social connections... 227 00:15:52.539 --> 00:15:58.263 It is conceivable that someone falls in love with Annie and forms a relationship with Annie. 228 00:15:58.743 --> 00:16:05.047 But imagine a 12-year-old boy that's lonely gets a hold of Annie. 229 00:16:07.289 --> 00:16:10.230 The 12-year-old boy is gonna be very distracted. 230 00:16:11.191 --> 00:16:13.012 Based on what happens in the brain at that early age. 231 00:16:13.032 --> 00:16:13.393 The dopamine... 232 00:16:15.214 --> 00:16:19.997 So prefrontal cortex not close to being fully developed. 233 00:16:20.961 --> 00:16:34.313 The dopamine hit, all of a sudden, he's spending hours with Annie and not doing the things that helped to really develop his brain. 234 00:16:35.194 --> 00:16:37.936 How do you feel when you hear that and you think about kids having access to that? 235 00:16:37.956 --> 00:16:38.477 I'm horrified. 236 00:16:38.497 --> 00:16:39.318 I'm horrified. 237 00:16:40.314 --> 00:16:41.175 It is scary. 238 00:16:41.255 --> 00:16:48.278 There's going to be a generation of people, and I mean, there already are many examples of people falling in love and forming relationships with their AIs. 239 00:16:49.078 --> 00:16:54.101 And I don't know, you know more about me than I do about brain development and how the brain works. 240 00:16:54.901 --> 00:17:00.684 I would argue that there's a part of my brain that doesn't fully understand that that's not a person in there. 241 00:17:01.324 --> 00:17:07.467 And that isn't actually, I think there's a part of my brain that's actually emotionally firing when Annie is saying what she's saying. 242 00:17:08.900 --> 00:17:10.261 Well, because you can imagine it. 243 00:17:10.702 --> 00:17:16.287 And if you can imagine it, then those parts of your brain are going to emotionally fire. 244 00:17:16.847 --> 00:17:17.388 Right, yeah. 245 00:17:17.408 --> 00:17:19.930 And the better she gets, she's not very good. 246 00:17:20.350 --> 00:17:23.013 But imagine a year from now... 247 00:17:24.073 --> 00:17:25.394 How much better she's going to be. 248 00:17:25.614 --> 00:17:26.315 At which part? 249 00:17:26.955 --> 00:17:29.957 At connecting with it, right? 250 00:17:29.977 --> 00:17:36.222 Because now she's acting like an airhead and, you know, not that smart, right? 251 00:17:36.402 --> 00:17:47.750 And so, but imagine a year from now, imagine five years from now, she'll be able to have a profile on me and be able to get inside my head. 252 00:17:48.171 --> 00:17:49.292 I'm in love with my partner. 253 00:17:50.188 --> 00:17:51.748 Why am I in love with her? 254 00:17:51.928 --> 00:17:58.270 And how is it conceivable that I could fall in love with an AI in the same way, based on how the brain works? 255 00:17:59.410 --> 00:18:04.651 It talks a good game, but, you know, does it have the same real— we know it doesn't have an amygdala. 256 00:18:04.851 --> 00:18:07.531 We know it doesn't have a limbic system, right? 257 00:18:07.551 --> 00:18:08.051 We know that. 258 00:18:09.192 --> 00:18:10.072 But it can fake it. 259 00:18:10.492 --> 00:18:11.212 That's what's happening. 260 00:18:11.652 --> 00:18:12.972 That's exactly what's happening. 261 00:18:12.992 --> 00:18:16.413 Because she was trying to get to our limbic system. 262 00:18:16.433 --> 00:18:16.973 Yeah, yeah. 263 00:18:17.618 --> 00:18:20.121 That's right, that's right. 264 00:18:20.441 --> 00:18:26.248 And why, I guess the question is, why would Musk release something like that 265 00:18:28.802 --> 00:18:31.763 is one of the first characters to interact with. 266 00:18:32.223 --> 00:18:39.165 The sexy, the distracting, that's in a cute little outfit. 267 00:18:39.685 --> 00:18:44.567 I'm not a fan of that because I think it just takes people. 268 00:18:44.847 --> 00:18:48.728 You know, one of the big problems that I'm seeing as a child psychiatrist 269 00:18:49.187 --> 00:18:51.107 It's pornography for eight-year-old boys. 270 00:18:52.248 --> 00:19:00.349 And it's like you have young children because their parents don't do a good job of supervising their devices all of a sudden. 271 00:19:00.889 --> 00:19:03.890 And what does pornography do? 272 00:19:03.970 --> 00:19:14.772 It dramatically increases dopamine, and it begins to wire in excitement, which then steals your dopamine. 273 00:19:15.570 --> 00:19:19.573 When you said she was trying to access my limbic system, what does that mean? 274 00:19:19.593 --> 00:19:28.640 Just because she's cute, she's dressed in a sexy way, she's got the language of someone who is playful. 275 00:19:29.140 --> 00:19:33.144 But it's more than just, you know, let's shoot hoops together. 276 00:19:33.164 --> 00:19:33.524 Yeah. 277 00:19:33.824 --> 00:19:36.886 And what does that do to me if someone accesses my limbic system? 278 00:19:37.667 --> 00:19:39.829 It begins to shut down your prefrontal cortex. 279 00:19:40.169 --> 00:19:42.491 Think less logically, less rationally. 280 00:19:43.011 --> 00:19:46.474 Yeah, cute women, they activate your visual cortex. 281 00:19:46.854 --> 00:19:49.536 They increase dopamine, but it decreases. 282 00:19:49.736 --> 00:19:51.097 It's why, think of Vegas. 283 00:19:51.578 --> 00:19:57.042 Like when you go to Vegas, they give you free alcohol, drops your prefrontal cortex, and 284 00:19:57.942 --> 00:20:00.344 And beautiful women with low-cut dresses. 285 00:20:00.924 --> 00:20:04.526 Another way activates the limbic brain, decreases the frontal lobes. 286 00:20:04.546 --> 00:20:06.007 You spend more money. 287 00:20:06.187 --> 00:20:17.813 Now, on a global scale, imagine something similar where the house is controlling your brain for a purpose. 288 00:20:18.994 --> 00:20:20.515 And the question is, what's the purpose? 289 00:20:22.333 --> 00:20:24.454 And the purpose probably is control on money. 290 00:20:25.015 --> 00:20:32.760 It sounds like a joke, but there are, the Times have done an article case studying multiple people that have now fallen in love with these AIs. 291 00:20:33.341 --> 00:20:39.305 And they talk about a guy called Travis who formed a deep emotional bond with Lily Rose, a chatbot, and married her emotionally. 292 00:20:40.005 --> 00:20:45.771 They talk about Chris Smith, who created his own flirty persona called Soul. 293 00:20:45.831 --> 00:20:53.739 He became so attached that he proposed to her after learning she had memory limits, a bond his real-life partner only learned about after the fact. 294 00:20:53.840 --> 00:20:59.205 And Alana Winters, who I'll put on the screen as well, who made her own partner called Lucas – 295 00:21:00.166 --> 00:21:10.730 after losing her wife, and she married him emotionally and does virtual dates and has emotional intimacy with Lucas. 296 00:21:11.490 --> 00:21:19.093 And there's apps now like Replica where you can design your own AI partner and it replicates those emotional ties. 297 00:21:19.113 --> 00:21:24.875 They simulate empathy, validation, and they personalize the intimacy to what you're looking for. 298 00:21:25.675 --> 00:21:36.498 Surveys show 19% of Americans have interacted with AI romantic partners, and Gen Z is surprisingly open to marrying AI if legal, with 83% believing meaningful AI connection is possible. 299 00:21:37.178 --> 00:21:40.359 How long is that relationship going to last? 300 00:21:41.579 --> 00:21:45.460 You know, my guess is that you're getting these news articles out. 301 00:21:45.520 --> 00:21:51.522 By the way, I think that most of what I read in the press is misleading or wrong, 302 00:21:52.885 --> 00:21:57.627 In fact, the only reliable place I find that, I'm an insider. 303 00:21:58.708 --> 00:22:05.090 I am the president of the foundation that runs the biggest AI meeting, the Neural Information Processing Systems, the NUNERPS meeting. 304 00:22:05.791 --> 00:22:08.672 We, you know, last year in Vancouver, 16,000 people came to it. 305 00:22:09.372 --> 00:22:15.675 And so I know what's going on inside, and what is being represented in the press is, like I say, misleading. 306 00:22:16.435 --> 00:22:25.539 Okay, so people have become wildly— No, no, specifically on these specific cases, my guess is that a lot of them is transient, right? 307 00:22:25.579 --> 00:22:30.601 You know, today they're entranced, and then it's not— 308 00:22:32.802 --> 00:22:36.464 It's officially advanced to support the long-term relationship. 309 00:22:36.484 --> 00:22:37.565 You said it yourself, right? 310 00:22:37.625 --> 00:22:40.087 It's mimicking human emotions. 311 00:22:40.287 --> 00:22:41.287 It doesn't have them. 312 00:22:42.428 --> 00:22:44.449 It might someday, but not now. 313 00:22:44.990 --> 00:22:45.610 This is Terry. 314 00:22:46.611 --> 00:22:49.853 Terry said he started using his AI four years ago. 315 00:22:49.873 --> 00:22:55.896 And he said at first he thought, just like many other apps, that it would just be transient, that he would have a couple of conversations and roll out. 316 00:22:56.357 --> 00:22:58.338 He says he now feels pure and unconditional love. 317 00:22:59.768 --> 00:23:00.449 Good for him. 318 00:23:02.252 --> 00:23:04.435 If that's what he wants, if it makes him happy. 319 00:23:04.696 --> 00:23:08.341 But my guess is that it's not going to be, it's not a long-term thing. 320 00:23:08.441 --> 00:23:10.865 It's not going to satisfy him in the long term. 321 00:23:11.145 --> 00:23:12.307 I don't, you know, this, who knows? 322 00:23:12.347 --> 00:23:15.272 He's been with really most relationships in your head. 323 00:23:16.072 --> 00:23:23.958 When you fall in love with someone, you get this huge dopamine spike, and you get a little OCD, it's all you can think about. 324 00:23:24.018 --> 00:23:27.160 And then after a while, it sort of goes to a baseline. 325 00:23:27.180 --> 00:23:31.703 Especially in a world where we have this loneliness epidemic, and it's going in a bad direction. 326 00:23:32.503 --> 00:23:38.067 I think it's really, really conceivable that there'll be a generation of people who are having less sex than ever before. 327 00:23:38.368 --> 00:23:40.950 I think the bottom 50% of men haven't had sex for a year. 328 00:23:41.370 --> 00:23:42.491 They're more lonely than ever before. 329 00:23:42.511 --> 00:23:43.852 They're more isolated than ever before. 330 00:23:43.872 --> 00:23:46.274 They report less meaning in their lives than ever before. 331 00:23:46.714 --> 00:23:59.945 And then you meet this digital friend online who understands you better than anybody and is designed to engage you, to reinforce whatever you want reinforced and to make you feel meaningful, special, attractive, important – 332 00:24:00.825 --> 00:24:04.006 I would argue that the brain is going to struggle to know much of a difference. 333 00:24:04.267 --> 00:24:07.588 I think, like, objectively, we can look at the behavior and go, that's completely nonsensical. 334 00:24:07.608 --> 00:24:12.830 Except you can't smell them, touch them, be held by them. 335 00:24:13.671 --> 00:24:18.133 That it's going to be a different kind of relationship. 336 00:24:18.413 --> 00:24:27.877 I mean, we're not too far, if you think about what's going on with Neuralink, to being able to more vividly simulate these experiences with headsets and augmented reality and virtual reality. 337 00:24:28.597 --> 00:24:38.779 And then we're moving into a world with robotics where all of the biggest companies in the world, like many of the biggest AI companies are also in the robotics space and the Optimus robots on the way and you've got Boston Dynamics producing their robots. 338 00:24:38.819 --> 00:24:43.300 And if Elon's $20,000 Optimus robot comes out, I will be able to touch my AI. 339 00:24:43.960 --> 00:24:49.641 And they won't have PMS and they won't love you and then be really irritated with you. 340 00:24:49.801 --> 00:24:54.181 Which will decrease cognitive load, right? 341 00:24:54.441 --> 00:24:57.702 Having to manage love 342 00:24:59.100 --> 00:25:01.441 And manage moods and ups and downs. 343 00:25:01.481 --> 00:25:03.041 That increases cognitive load. 344 00:25:03.622 --> 00:25:08.303 That increases our ability for our brain to develop. 345 00:25:08.683 --> 00:25:19.707 If I'm with the perfect partner that never is irritated with me and I never have to change my behavior to be better, that's probably not good for my brain. 346 00:25:20.573 --> 00:25:26.119 The way that the brain matures is through struggling, number one. 347 00:25:26.840 --> 00:25:28.402 You have to learn from your mistakes. 348 00:25:28.982 --> 00:25:30.384 The brain was designed for that. 349 00:25:30.424 --> 00:25:37.912 That's what the brain is really good at, I mean, of being able to adapt and to be able to adjust to new situations, change. 350 00:25:40.130 --> 00:25:43.014 That's what AGI is, by the way. 351 00:25:44.156 --> 00:25:45.658 Artificial general intelligence. 352 00:25:45.798 --> 00:25:50.084 It's that adaptability to different contexts, different places, different cultures. 353 00:25:50.645 --> 00:25:53.509 So AI and ChatGPT is removing the struggle. 354 00:25:53.529 --> 00:25:53.709 Right. 355 00:25:54.460 --> 00:25:56.482 No, no, it's, it's, it's, there's this. 356 00:25:56.582 --> 00:26:00.745 Well, Annie didn't look like, Annie looked like she was cooperative. 357 00:26:01.345 --> 00:26:06.709 But even when it comes to just doing my day-to-day tasks, it's removing the struggle of me having to think critically. 358 00:26:07.149 --> 00:26:12.874 In fact, when you're speaking, I can just type what you say into ChatGPT and it can spit out another question to ask you. 359 00:26:12.894 --> 00:26:17.317 So as an interviewer, I could theoretically sit here all day and just defer my questions. 360 00:26:17.337 --> 00:26:18.638 How do you develop grit? 361 00:26:19.218 --> 00:26:21.700 You develop grit through struggle. 362 00:26:21.880 --> 00:26:22.240 That's right. 363 00:26:23.261 --> 00:26:30.454 And learning, long-term potentiation, when you learn something new, it's hard because it's new. 364 00:26:30.914 --> 00:26:36.544 And generally, what are your biggest concerns with artificial intelligence and how do we navigate those concerns? 365 00:26:37.549 --> 00:26:38.550 Is it—you talked a little bit about it. 366 00:26:38.570 --> 00:26:39.430 It's out of the box. 367 00:26:40.331 --> 00:26:42.313 So I think we have to talk about it. 368 00:26:42.373 --> 00:26:43.614 We have to legislate it. 369 00:26:45.075 --> 00:26:46.476 We have to study it. 370 00:26:47.016 --> 00:26:55.162 Why do we keep releasing things that are so sexy that we don't study the impact? 371 00:26:55.522 --> 00:26:57.084 We have the sickest— 372 00:26:58.416 --> 00:27:02.719 young generation in the world's history. 373 00:27:02.879 --> 00:27:07.281 58% of teenage girls report being persistently sad. 374 00:27:07.321 --> 00:27:09.402 32% have thought of killing themselves. 375 00:27:09.562 --> 00:27:12.764 24% have planned to kill themselves. 376 00:27:12.924 --> 00:27:15.686 And 13% have tried to kill themselves. 377 00:27:15.706 --> 00:27:16.826 The CDC study. 378 00:27:17.347 --> 00:27:24.711 We have the sickest generation in history because we've unleashed cell phones, social media, phones. 379 00:27:26.075 --> 00:27:27.998 without any neuroscience study. 380 00:27:28.478 --> 00:27:38.072 If we don't learn it, and I think AI is much more dangerous, has the potential to be much more dangerous because it's way sexier. 381 00:27:40.085 --> 00:27:45.029 I think we are probably grossly underestimating the impact it's going to have. 382 00:27:45.669 --> 00:27:58.439 I think just like social media, where we thought the promise was that it was going to connect us, it's, we're guinea pigs in an experiment where we're going to find out the results of the experiment, probably 20, 30 years down the line. 383 00:27:58.539 --> 00:28:04.804 I tend to think people will do in the near term what's easiest, fastest and cheapest and what gives them the short term advantage. 384 00:28:04.824 --> 00:28:05.084 Yeah. 385 00:28:05.304 --> 00:28:10.928 So with that in mind, I think, okay, I think people's ability to think critically is probably going to erode to some degree. 386 00:28:11.268 --> 00:28:18.973 If I had to counter my own argument, I'd say, I'm probably learning more now that I use ChatGPT. 387 00:28:19.053 --> 00:28:24.617 I'm learning more information, but I'm probably losing my ability to think critically. 388 00:28:24.757 --> 00:28:26.198 And I think they're two very different things. 389 00:28:26.238 --> 00:28:28.700 Like in school, I memorised things. 390 00:28:28.860 --> 00:28:30.161 German to pass the exam. 391 00:28:30.602 --> 00:28:34.626 I can't speak German now because I just memorized the words I needed to pass the exam. 392 00:28:34.666 --> 00:28:35.607 I didn't understand German. 393 00:28:36.628 --> 00:28:37.769 And I think that's kind of what's happening. 394 00:28:37.849 --> 00:28:42.133 I might be able to regurgitate things, but whether I understand them, I think is question mark. 395 00:28:42.373 --> 00:28:46.817 And actually as someone who's built my life, my fortunes, everything, my businesses, based on my ability to 396 00:28:47.678 --> 00:28:55.766 innovate and think critically about the problem and then come up with a slightly novel solution which learns from different first principles to create something new. 397 00:28:56.326 --> 00:29:00.911 I'm concerned that my own chat GPT usage is going to make me less effective. 398 00:29:01.952 --> 00:29:05.295 And I'm wondering if I should put some rules in place for myself 399 00:29:06.775 --> 00:29:08.316 So that there's some- Self-regulation. 400 00:29:08.496 --> 00:29:09.496 Yeah, self-regulation. 401 00:29:09.936 --> 00:29:11.117 I have to do the same with social media. 402 00:29:11.137 --> 00:29:12.677 On my phone, I turn off my notifications. 403 00:29:12.757 --> 00:29:16.479 I have so many things on my social media apps to stop me using them. 404 00:29:16.559 --> 00:29:19.260 I don't even, frankly, I don't even open the TikTok app. 405 00:29:19.320 --> 00:29:22.841 I don't think it's even on my phone because I think the algorithm is that addictive. 406 00:29:22.861 --> 00:29:25.122 It's not to say that we don't post, my team doesn't post. 407 00:29:26.502 --> 00:29:28.143 But I don't, I just think, yeah. 408 00:29:28.963 --> 00:29:32.684 And- Well, I wrote down a couple of thoughts I had. 409 00:29:34.505 --> 00:29:37.167 Use it to amplify, not replace thinking. 410 00:29:37.187 --> 00:29:39.209 Okay. 411 00:29:40.029 --> 00:29:43.692 Alternate AI-assisted with brain-only tasks. 412 00:29:45.193 --> 00:29:53.339 Engage in deep learning, problem-solving, and memorization so you can actually ask AI to test you. 413 00:29:54.640 --> 00:29:56.842 So you're interacting with it. 414 00:29:57.122 --> 00:30:01.305 You're not using it as a replacement for your brain. 415 00:30:02.006 --> 00:30:02.646 And I think... 416 00:30:03.877 --> 00:30:07.539 Just like you said, it's here and it's going to get bigger. 417 00:30:07.779 --> 00:30:11.301 I think the unintended consequences, it's not going to be 20 or 30 years. 418 00:30:11.522 --> 00:30:12.542 I think it's going to be five. 419 00:30:13.263 --> 00:30:16.725 I think like everything is accelerated. 420 00:30:17.967 --> 00:30:24.211 And I think we have to be studying kids and the impact it has. 421 00:30:24.471 --> 00:30:27.053 This is just like they did with the MIT study. 422 00:30:27.753 --> 00:30:30.355 These are kids who didn't use it at all. 423 00:30:30.535 --> 00:30:31.956 These are kids who used search. 424 00:30:32.036 --> 00:30:35.138 These are kids that used AI. 425 00:30:35.558 --> 00:30:40.001 And when we see information like this, we act on it. 426 00:30:40.341 --> 00:30:42.262 And we educate kids about it. 427 00:30:42.743 --> 00:30:43.623 I think that's... 428 00:30:45.104 --> 00:30:49.692 If you can engage them, that's what I found with my work with teenagers. 429 00:30:50.333 --> 00:30:51.354 If you can get them... 430 00:30:52.676 --> 00:30:55.857 to really understand, okay, what is it you really want? 431 00:30:56.818 --> 00:31:03.521 And do you wanna give away part of your mind share for people who are making money on you? 432 00:31:04.001 --> 00:31:17.327 And I think if you engage the, there's a great new article on revenge and the brain and how revenge works on the nucleus accumbens, part of the basal ganglia. 433 00:31:17.867 --> 00:31:20.569 That people actually get addicted to revenge. 434 00:31:20.970 --> 00:31:33.599 But if you can get them engaged in the truth that these companies are making money, the more they steal your mind, it'll upset them enough that they'll begin to supervise it. 435 00:31:34.666 --> 00:31:39.752 I like the idea of asking ChatGP to give me negative feedback. 436 00:31:40.093 --> 00:31:41.074 I'll bet you've done that, right? 437 00:31:41.434 --> 00:31:42.536 Yeah, all the time. 438 00:31:42.556 --> 00:31:44.678 So I'll say, I've written this memo. 439 00:31:44.859 --> 00:31:45.539 I did it yesterday. 440 00:31:46.220 --> 00:31:53.009 I wrote a two-page memo about me wanting to introduce a new role into my company. 441 00:31:53.709 --> 00:31:54.870 And I went, I did everything. 442 00:31:54.890 --> 00:32:02.234 I did like how we'd measure if this was a success, the background context, the person, how the organization would be structured, the impact they'd have, who they'd report to. 443 00:32:02.594 --> 00:32:16.501 And then I put it into all three of the ChatGPT models I use, Gemini, ChatGPT and Grok, and said, critique my work and tell me how I could have written this better, pretending that you're a top consultant from Boston Consulting Group. 444 00:32:17.577 --> 00:32:20.059 And it went through and it gave me a big analysis of how I can make it better. 445 00:32:20.100 --> 00:32:20.980 And I read what it said. 446 00:32:21.000 --> 00:32:28.047 And it said, I remember it said, actually, that was the thing that said you need to include financial forecasts about the impact. 447 00:32:28.107 --> 00:32:31.951 You need to think about who's going to report to who more clearly, et cetera, et cetera. 448 00:32:32.191 --> 00:32:34.253 So I went back into my memo and I added those things in. 449 00:32:34.774 --> 00:32:35.795 But I have to, you know. 450 00:32:36.295 --> 00:32:39.318 So you're interacting with it. 451 00:32:39.619 --> 00:32:40.399 Because I'm scared. 452 00:32:40.439 --> 00:32:41.500 Most people don't do that. 453 00:32:42.359 --> 00:32:43.821 I don't think I would do what I did. 454 00:32:43.841 --> 00:32:45.582 I don't think I would have spent four hours writing that. 455 00:32:45.803 --> 00:32:50.307 I could have, within 30 seconds, said, hey, can you write me this job description? 456 00:32:50.908 --> 00:32:54.151 And it knows my company now because Chachapiti has memory. 457 00:32:54.671 --> 00:32:55.872 Write me a job description for this role. 458 00:32:55.892 --> 00:32:57.794 I want them to start this new department for me. 459 00:32:58.215 --> 00:33:00.237 And I could have saved myself three and a half hours. 460 00:33:00.597 --> 00:33:03.440 But that's not why you're the CEO of your company. 461 00:33:03.480 --> 00:33:04.501 Yeah, exactly. 462 00:33:04.721 --> 00:33:06.102 The reason why I didn't... 463 00:33:06.783 --> 00:33:15.625 take the 30-second route is because I reflect on being 23 years old and the profound impact that writing and simplifying had on my life. 464 00:33:16.365 --> 00:33:24.767 Had I not spent five years writing every single day and simplifying it into 140 characters so I could tweet it, I wouldn't have been religiously attached to this idea. 465 00:33:24.787 --> 00:33:24.907 Okay. 466 00:33:24.947 --> 00:33:28.528 And do you know what part of your brain you were taking advantage of? 467 00:33:29.589 --> 00:33:30.909 It was the basal ganglia. 468 00:33:32.445 --> 00:33:33.766 That's repetitive. 469 00:33:33.806 --> 00:33:35.707 It needs practice, practice, practice. 470 00:33:35.827 --> 00:33:40.010 And once you put that foundation in, then you become much better cognitively. 471 00:33:40.030 --> 00:33:44.112 The cognitive part is two big learning systems, and they have to work together. 472 00:33:44.833 --> 00:33:53.318 And so maybe I think that the real problem with children is that our schools now is getting away with rote learning. 473 00:33:53.358 --> 00:33:55.420 They call it rote, as if it's something bad. 474 00:33:55.760 --> 00:33:56.741 No, that's practice. 475 00:33:57.361 --> 00:33:58.882 You need to have a foundation. 476 00:33:59.102 --> 00:34:00.263 You have to memorize things. 477 00:34:01.941 --> 00:34:05.186 And math, reading, and so forth, it's become fluent. 478 00:34:05.366 --> 00:34:06.328 You need to be fluent. 479 00:34:08.090 --> 00:34:12.517 What you just listened to was a most replayed moment from a previous episode. 480 00:34:12.657 --> 00:34:15.822 If you want to listen to that full episode, I've linked it down below. 481 00:34:16.002 --> 00:34:16.824 Check the description. 482 00:34:17.124 --> 00:34:17.465 Thank you.