1 00:00:00,080 --> 00:00:02,160 Speaker 1: Hi, this is David Eagleman. I want to wish you 2 00:00:02,200 --> 00:00:05,000 Speaker 1: a very happy holidays. We're going to take a break 3 00:00:05,040 --> 00:00:06,520 Speaker 1: for a couple of weeks, and then we're back in 4 00:00:06,600 --> 00:00:12,760 Speaker 1: January with new episodes on emotion, intelligence, time perception, smell 5 00:00:12,840 --> 00:00:16,639 Speaker 1: and taste, brain, computer interfaces, and much more. In the meantime, 6 00:00:16,680 --> 00:00:19,080 Speaker 1: we're going to replay one of our favorite episodes from 7 00:00:19,120 --> 00:00:21,520 Speaker 1: the past year, and I'll look forward to seeing you 8 00:00:21,560 --> 00:00:32,480 Speaker 1: in January. What is the future of AI relationships? Could 9 00:00:32,520 --> 00:00:35,800 Speaker 1: it be the case that your relationship partner would be 10 00:00:35,960 --> 00:00:40,800 Speaker 1: more satisfied with a virtual version of you that behaves 11 00:00:41,320 --> 00:00:45,120 Speaker 1: five percent better than you do? Could you fall in 12 00:00:45,159 --> 00:00:48,839 Speaker 1: love with a bot? How does an AI bot plug 13 00:00:49,040 --> 00:00:52,800 Speaker 1: right into you our deep neural circuitry and what are 14 00:00:52,800 --> 00:00:55,440 Speaker 1: the pros and the cons of that? And what will 15 00:00:55,440 --> 00:00:58,400 Speaker 1: it mean when humans you love don't have to die 16 00:00:58,520 --> 00:01:06,240 Speaker 1: but can live on in your phone forever. Welcome to 17 00:01:06,280 --> 00:01:10,160 Speaker 1: Inner Cosmos with me David Eagleman. I'm a neuroscientist and 18 00:01:10,200 --> 00:01:13,640 Speaker 1: an author at Stanford and in these episodes I examined 19 00:01:13,680 --> 00:01:29,840 Speaker 1: the intersection of our brains and our lives. Today's episode 20 00:01:29,880 --> 00:01:35,959 Speaker 1: is about relationships. Why are our brains so wired for relationships. 21 00:01:36,280 --> 00:01:39,759 Speaker 1: Why do we want love so much? And will AI 22 00:01:40,480 --> 00:01:43,640 Speaker 1: be able to serve as a key to that lock, 23 00:01:44,319 --> 00:01:47,200 Speaker 1: and what does that mean for us as humans. So 24 00:01:47,280 --> 00:01:50,720 Speaker 1: one of the things that's becoming increasingly popular among young 25 00:01:50,760 --> 00:01:55,000 Speaker 1: men is having an AI girlfriend. You get to choose 26 00:01:55,120 --> 00:01:57,880 Speaker 1: or set up a beautiful avatar. 27 00:01:58,000 --> 00:02:00,720 Speaker 2: And what do I mean by beautiful. That's up to you. 28 00:02:00,720 --> 00:02:01,920 Speaker 2: You can choose. 29 00:02:01,880 --> 00:02:04,760 Speaker 1: Any model that you want, with any sort of features 30 00:02:04,760 --> 00:02:08,640 Speaker 1: that appeal maximally to you. But that's just what she 31 00:02:08,720 --> 00:02:13,160 Speaker 1: looks like. The important part is the conversation. You start 32 00:02:13,440 --> 00:02:16,600 Speaker 1: talking with her, and typically this is just text chat, 33 00:02:17,000 --> 00:02:21,000 Speaker 1: but the technology is evolving into the upgrade of video chat, 34 00:02:21,000 --> 00:02:23,960 Speaker 1: where you see the avatar's mouth moving while she speaks 35 00:02:24,000 --> 00:02:28,240 Speaker 1: to you. Now, typically the free or entry price gets 36 00:02:28,280 --> 00:02:32,160 Speaker 1: you an avatar friend who lives on your phone and 37 00:02:32,240 --> 00:02:34,240 Speaker 1: checks in on you and says nice things to you 38 00:02:34,600 --> 00:02:37,560 Speaker 1: and is available anytime that you want to chat. 39 00:02:38,000 --> 00:02:40,000 Speaker 2: And for a premium. 40 00:02:39,440 --> 00:02:44,799 Speaker 1: Subscription price you can upgrade to a steamier relationship, and 41 00:02:44,919 --> 00:02:49,000 Speaker 1: here she'll text suggestive photos and she'll say things that 42 00:02:49,040 --> 00:02:54,520 Speaker 1: you might only expect from pillow whispers. So the concern 43 00:02:54,560 --> 00:02:57,320 Speaker 1: that people have expressed is whether this is going to 44 00:02:57,520 --> 00:03:01,680 Speaker 1: impact the next generation of males. Now, as a side note, 45 00:03:01,760 --> 00:03:04,560 Speaker 1: let me say that I suspect this will have whatever 46 00:03:04,639 --> 00:03:07,880 Speaker 1: influence it has on both genders, on males and females, 47 00:03:08,240 --> 00:03:09,560 Speaker 1: also straight and gay. 48 00:03:10,000 --> 00:03:11,600 Speaker 2: But I do suspect that. 49 00:03:11,960 --> 00:03:16,880 Speaker 1: Males will be the majority demographic simply because males tend 50 00:03:17,000 --> 00:03:21,320 Speaker 1: to be more visually driven than females. So for the 51 00:03:21,360 --> 00:03:24,320 Speaker 1: conversation here, I'm going to talk about it the way 52 00:03:24,320 --> 00:03:27,840 Speaker 1: that it's mostly discussed in the media and in academic circles, 53 00:03:28,080 --> 00:03:31,959 Speaker 1: which is straight males getting girlfriends this way. But keep 54 00:03:31,960 --> 00:03:35,560 Speaker 1: in mind this is a more generalized issue. Now, the 55 00:03:35,720 --> 00:03:40,360 Speaker 1: question is what will this mean for all future generations? 56 00:03:40,840 --> 00:03:44,840 Speaker 1: Because within AI relationship, you don't have to go out 57 00:03:44,880 --> 00:03:49,160 Speaker 1: and confront all the difficulty of a real flesh and 58 00:03:49,200 --> 00:03:54,840 Speaker 1: blood relationship. Real relationships get snippy, people get angry. 59 00:03:54,960 --> 00:03:56,200 Speaker 2: In real relationships. 60 00:03:56,280 --> 00:03:59,720 Speaker 1: Your partner might develop a crush on someone else in 61 00:04:00,080 --> 00:04:03,040 Speaker 1: leave you or hook up with someone else and you 62 00:04:03,120 --> 00:04:07,080 Speaker 1: find out later, Or your partner might develop an illness, 63 00:04:07,760 --> 00:04:10,080 Speaker 1: or she might get a job somewhere else and have 64 00:04:10,160 --> 00:04:13,200 Speaker 1: to move and then you're stuck in a lonely long 65 00:04:13,240 --> 00:04:18,320 Speaker 1: distance relationship for years or whatever. Relationships are full of challenges, 66 00:04:18,800 --> 00:04:22,120 Speaker 1: the majority of which can get circumvented with a nice 67 00:04:22,200 --> 00:04:25,640 Speaker 1: algorithm that is just content to listen to you all 68 00:04:25,680 --> 00:04:29,159 Speaker 1: the time and remember everything you say and give you 69 00:04:29,320 --> 00:04:33,919 Speaker 1: one hundred percent attention and always be nice. So my 70 00:04:34,000 --> 00:04:36,960 Speaker 1: wife sometimes jokes with me about wanting to build the 71 00:04:37,680 --> 00:04:41,680 Speaker 1: five percent better David. She has, mostly as a joke, 72 00:04:41,800 --> 00:04:44,960 Speaker 1: talked about this issue of what if she could have 73 00:04:45,120 --> 00:04:49,560 Speaker 1: an AI avatar of me that is never distracted with work, 74 00:04:49,680 --> 00:04:52,839 Speaker 1: or never looks at my cell phone when it things 75 00:04:52,839 --> 00:04:55,719 Speaker 1: in the middle of a conversation we're having, or never 76 00:04:56,000 --> 00:04:58,960 Speaker 1: wakes up from a weird dream and has a funny morning, 77 00:04:59,640 --> 00:05:03,919 Speaker 1: or never argues over some misunderstanding that's later understood to 78 00:05:03,920 --> 00:05:06,800 Speaker 1: be stupid and meaningless. And she tells me that she 79 00:05:07,000 --> 00:05:11,000 Speaker 1: wants five percent better David to always tell her she's right, 80 00:05:11,120 --> 00:05:14,719 Speaker 1: even in those rare cases when she's wrong. The key 81 00:05:14,800 --> 00:05:18,120 Speaker 1: is that five percent better David never gets busy or 82 00:05:18,200 --> 00:05:23,960 Speaker 1: occasionally snarky or forgets some occasion. Instead, he represents all 83 00:05:24,440 --> 00:05:27,560 Speaker 1: the best of me. And I'll just note that it's 84 00:05:27,640 --> 00:05:30,839 Speaker 1: very kind of her to label this five percent better David, 85 00:05:31,000 --> 00:05:34,360 Speaker 1: because she could say like ninety percent better and she'd 86 00:05:34,400 --> 00:05:35,200 Speaker 1: be justified. 87 00:05:35,720 --> 00:05:35,960 Speaker 3: Why. 88 00:05:36,160 --> 00:05:40,520 Speaker 1: It's because we are all very imperfect in relationships. As 89 00:05:40,560 --> 00:05:44,640 Speaker 1: I've talked about on other episodes, we are each living 90 00:05:44,760 --> 00:05:47,920 Speaker 1: on our own planet in the sense that we carry 91 00:05:48,000 --> 00:05:52,040 Speaker 1: our own internal models of the world, and as much 92 00:05:52,040 --> 00:05:55,680 Speaker 1: as we work to understand one another's viewpoints and motivations 93 00:05:56,040 --> 00:05:57,520 Speaker 1: and intentions, we're. 94 00:05:57,400 --> 00:05:58,480 Speaker 2: Not always that good at it. 95 00:05:58,560 --> 00:06:02,280 Speaker 1: Because we assume that other people are seeing the world 96 00:06:02,640 --> 00:06:04,920 Speaker 1: in the same way that we do, they have the 97 00:06:04,960 --> 00:06:09,200 Speaker 1: same methods for sense making, they gather meaning in the 98 00:06:09,240 --> 00:06:12,080 Speaker 1: same way that we do, and we assume they generally 99 00:06:12,120 --> 00:06:16,279 Speaker 1: hold or should hold, the same opinions about everything that 100 00:06:16,360 --> 00:06:20,600 Speaker 1: we do. And this is because the brain is locked 101 00:06:20,640 --> 00:06:25,520 Speaker 1: in silence and darkness and has no meaningful direct access 102 00:06:25,560 --> 00:06:28,320 Speaker 1: to the outside world, and so it gathers up all 103 00:06:28,360 --> 00:06:32,440 Speaker 1: its information through its narrow windows of the senses, and 104 00:06:32,520 --> 00:06:37,320 Speaker 1: it builds its internal model from this very thin trajectory 105 00:06:37,360 --> 00:06:40,279 Speaker 1: of space and time that it walks along. And this 106 00:06:40,440 --> 00:06:43,800 Speaker 1: is why everyone is so different on the inside, and 107 00:06:43,880 --> 00:06:50,960 Speaker 1: therefore why relationships are always full of misunderstanding and often conflict. 108 00:06:51,720 --> 00:06:56,680 Speaker 1: So relationships are inherently tough. And the question is when 109 00:06:56,760 --> 00:06:58,880 Speaker 1: would it be a good thing if you could have 110 00:06:59,640 --> 00:07:03,960 Speaker 1: an artificial partner who represents all the best of what 111 00:07:04,040 --> 00:07:06,880 Speaker 1: a person can be. So a lot of people will 112 00:07:06,920 --> 00:07:09,840 Speaker 1: immediately say no to this idea, but it's worth noting 113 00:07:10,360 --> 00:07:13,640 Speaker 1: that we're all striving to be the five percent better 114 00:07:13,760 --> 00:07:17,320 Speaker 1: versions of ourselves. We don't want to be snarky or 115 00:07:17,400 --> 00:07:20,640 Speaker 1: angry or distracted when a loved one is talking to us. 116 00:07:20,680 --> 00:07:22,920 Speaker 1: It's not like we get some extra pleasure out of 117 00:07:22,960 --> 00:07:25,960 Speaker 1: doing that, and it's not like the relationship gets some 118 00:07:26,120 --> 00:07:31,600 Speaker 1: extra boost or closeness from that having happened. So presumably 119 00:07:31,840 --> 00:07:35,560 Speaker 1: this is all part of why AI relationships have become 120 00:07:36,120 --> 00:07:38,480 Speaker 1: a thing, a possibility. 121 00:07:37,720 --> 00:07:39,320 Speaker 2: That we talk about nowadays. 122 00:07:39,840 --> 00:07:44,400 Speaker 1: In Japan, many young men apparently already prefer to have 123 00:07:44,520 --> 00:07:50,040 Speaker 1: relationships with their digital assistants or avatars or holographic girlfriends 124 00:07:50,440 --> 00:07:55,200 Speaker 1: instead of dealing with the complexity of real life relationships, 125 00:07:55,760 --> 00:08:00,800 Speaker 1: and according to research, gen z is more readily redisposed 126 00:08:00,960 --> 00:08:06,320 Speaker 1: to seek out relationships with AI generated avatars first because 127 00:08:06,360 --> 00:08:09,560 Speaker 1: they're comfortable using the technology in this way compared to 128 00:08:09,640 --> 00:08:15,560 Speaker 1: previous generations, and also they're participating less often in traditional 129 00:08:15,560 --> 00:08:20,600 Speaker 1: social activities like regular family dinners or attending religious services 130 00:08:20,720 --> 00:08:26,360 Speaker 1: or playing sports and The question is, if AI relationships 131 00:08:26,760 --> 00:08:31,600 Speaker 1: were to catch on broadly, what will this mean for society? 132 00:08:31,680 --> 00:08:35,079 Speaker 1: Will kids actually stop going on dates because they can 133 00:08:35,160 --> 00:08:37,880 Speaker 1: find better relationships online? 134 00:08:38,800 --> 00:08:41,319 Speaker 2: And this is a real question because there. 135 00:08:41,160 --> 00:08:47,840 Speaker 1: Are many startups currently blossoming to create chatbot driven connections. 136 00:08:47,880 --> 00:08:49,360 Speaker 2: I'll give you one example. 137 00:08:49,400 --> 00:08:52,880 Speaker 1: There's a twenty three year old influencer with almost two 138 00:08:52,920 --> 00:08:56,880 Speaker 1: million Snapchat followers. Her name is Karen Marjorie, and earlier 139 00:08:56,920 --> 00:09:00,920 Speaker 1: this year in May, she released Karen Ai, which is 140 00:09:01,200 --> 00:09:05,960 Speaker 1: an immersive AI experience featuring videos of Margie that she 141 00:09:06,080 --> 00:09:10,200 Speaker 1: says provide a quote virtual girlfriend for those who are 142 00:09:10,200 --> 00:09:12,440 Speaker 1: willing to pony up one dollar per minute. 143 00:09:12,760 --> 00:09:12,960 Speaker 2: Now. 144 00:09:12,960 --> 00:09:17,160 Speaker 1: This is what's known as a companion chatbot, and she 145 00:09:17,280 --> 00:09:20,720 Speaker 1: tweeted that quote karen Ai is the first step in 146 00:09:20,760 --> 00:09:23,360 Speaker 1: the right direction to cure loneliness. 147 00:09:24,080 --> 00:09:27,480 Speaker 2: He tweet continues. Quote. Men are told to suppress their. 148 00:09:27,360 --> 00:09:30,800 Speaker 1: Emotions and not talk about issues they're having. 149 00:09:30,960 --> 00:09:33,800 Speaker 2: I vow to fix this with karen Ai. 150 00:09:35,080 --> 00:09:39,359 Speaker 1: She says she's worked with leading psychologists to seamlessly include 151 00:09:39,480 --> 00:09:44,480 Speaker 1: the right therapies to quote undue trauma, rebuild physical and 152 00:09:44,520 --> 00:09:48,720 Speaker 1: emotional confidence, and rebuild what has been taken away by 153 00:09:48,760 --> 00:09:51,240 Speaker 1: the pandemic end quote. And by the way, as a 154 00:09:51,280 --> 00:09:54,240 Speaker 1: side note, I think AI psychologists are going to be 155 00:09:54,400 --> 00:09:58,400 Speaker 1: a truly important part of the clinical landscape by next 156 00:09:58,480 --> 00:10:01,240 Speaker 1: year because you can have a therapist that you can 157 00:10:01,320 --> 00:10:04,640 Speaker 1: talk to twenty four to seven, and the therapist never gets. 158 00:10:04,400 --> 00:10:05,839 Speaker 2: Distracted or flustered and. 159 00:10:05,920 --> 00:10:10,200 Speaker 1: Only cares about you and has a perfect memory for 160 00:10:10,240 --> 00:10:13,400 Speaker 1: everything you've ever said, which is better than anybody else 161 00:10:13,440 --> 00:10:17,720 Speaker 1: in real life. So back to AI girlfriends or boyfriends. 162 00:10:18,080 --> 00:10:21,760 Speaker 1: The same idea applies here, which is that they are 163 00:10:21,800 --> 00:10:25,960 Speaker 1: completely devoted to you and always in a good mood 164 00:10:26,320 --> 00:10:30,600 Speaker 1: and only have you in mind. So what are AI 165 00:10:30,679 --> 00:10:33,440 Speaker 1: relationships going to mean? Well, I think this is going 166 00:10:33,480 --> 00:10:38,040 Speaker 1: to be a research question that sociologists and psychologists will 167 00:10:38,080 --> 00:10:42,920 Speaker 1: study for the coming decades and centuries. The initial studies 168 00:10:42,920 --> 00:10:47,480 Speaker 1: are suggesting that people, mostly gen zers, are moving closer 169 00:10:47,520 --> 00:10:53,040 Speaker 1: to the technology to avoid the unpleasant realities of human relationships. 170 00:10:53,120 --> 00:10:58,079 Speaker 1: All the tough stuff. Is that detrimental? Well it could 171 00:10:58,120 --> 00:11:02,360 Speaker 1: be if it makes your human reas relationships harder, because 172 00:11:02,360 --> 00:11:05,320 Speaker 1: maybe every time you guys have an argument in real life, 173 00:11:05,400 --> 00:11:08,280 Speaker 1: your partner thinks, well, forget it, I'm bagging this. I'm 174 00:11:08,320 --> 00:11:11,800 Speaker 1: going back to my comfort zone. So the concern, as 175 00:11:11,800 --> 00:11:15,120 Speaker 1: you can probably guess, is that the rise of AI 176 00:11:15,280 --> 00:11:20,440 Speaker 1: driven relationships could exacerbate loneliness because they seem to be 177 00:11:20,520 --> 00:11:23,439 Speaker 1: a meal, but they provide no calories. 178 00:11:23,520 --> 00:11:24,880 Speaker 2: And I'll come back to that in a moment. 179 00:11:25,559 --> 00:11:30,040 Speaker 1: In other words, AI generated avatars could interfere with the 180 00:11:30,080 --> 00:11:35,080 Speaker 1: relationships that young people are just learning to foster, because 181 00:11:35,520 --> 00:11:41,400 Speaker 1: the AI relationship might breed dissatisfaction with flawed humans. And 182 00:11:41,440 --> 00:11:44,920 Speaker 1: this applies not only to lovers, but even to friends. 183 00:11:44,920 --> 00:11:48,360 Speaker 1: It might be easier to have AI friends who aren't 184 00:11:48,640 --> 00:11:50,800 Speaker 1: busy when you need them and can give you one 185 00:11:50,840 --> 00:11:54,280 Speaker 1: hundred percent of their attention whenever you need it. And 186 00:11:54,360 --> 00:11:58,560 Speaker 1: let me throw in a different potential problem with AI relationships, 187 00:11:58,679 --> 00:12:01,120 Speaker 1: So give me one second to take this tangent here. 188 00:12:01,559 --> 00:12:04,760 Speaker 1: I was thinking the other day about the Fermi paradox. 189 00:12:04,840 --> 00:12:06,120 Speaker 2: The Fermi paradox is. 190 00:12:06,720 --> 00:12:11,520 Speaker 1: Given the size of the observable cosmos, with over one 191 00:12:11,640 --> 00:12:15,120 Speaker 1: hundred billion galaxies, and each of them with one hundred 192 00:12:15,160 --> 00:12:18,280 Speaker 1: billion stars, and each of those surrounded by some number 193 00:12:18,320 --> 00:12:21,200 Speaker 1: of planets, what is the reason that we have not 194 00:12:21,440 --> 00:12:26,400 Speaker 1: heard from any other alien species yet? And one of 195 00:12:26,480 --> 00:12:30,200 Speaker 1: the proposals that's always been there is that maybe as 196 00:12:30,440 --> 00:12:36,040 Speaker 1: civilizations become more technically advanced, they end up killing themselves, 197 00:12:36,080 --> 00:12:39,839 Speaker 1: and this is why we haven't heard from other smart civilizations, 198 00:12:39,880 --> 00:12:40,600 Speaker 1: because they. 199 00:12:40,480 --> 00:12:41,800 Speaker 2: Are already gone. 200 00:12:42,520 --> 00:12:44,880 Speaker 1: And every time I've seen this proposal, it's always in 201 00:12:44,920 --> 00:12:48,520 Speaker 1: the form of warfare, things like nuclear bombs. 202 00:12:48,559 --> 00:12:50,840 Speaker 2: They end up wiping themselves out. 203 00:12:51,440 --> 00:12:55,000 Speaker 1: So civilizations become smart and it's not long before they disappear. 204 00:12:55,720 --> 00:12:59,720 Speaker 1: So in thinking about AI relationships, it struck me as 205 00:12:59,760 --> 00:13:04,359 Speaker 1: a possibility that if we had really, really great relationships 206 00:13:04,400 --> 00:13:09,040 Speaker 1: with avatars, perhaps that would cause the birth rate of 207 00:13:09,120 --> 00:13:12,440 Speaker 1: the species to collapse. I don't know if this has 208 00:13:12,480 --> 00:13:15,920 Speaker 1: been proposed as a possible answer to the Fermi paradox, 209 00:13:16,000 --> 00:13:20,560 Speaker 1: but maybe this should be included, not civilization's disappearing because 210 00:13:20,559 --> 00:13:24,120 Speaker 1: of bad things, but instead from having too much of 211 00:13:24,160 --> 00:13:28,400 Speaker 1: a good thing, which could fool and eventually overwrite or 212 00:13:28,520 --> 00:13:50,560 Speaker 1: mandate for reproduction. Okay, so no one knows what the 213 00:13:50,600 --> 00:13:54,280 Speaker 1: long term effects will be of these AI relationships, but 214 00:13:54,400 --> 00:13:58,800 Speaker 1: I don't actually think the situation is as dire as 215 00:13:58,840 --> 00:14:01,840 Speaker 1: some of these arguments suggest that it is. And I'll 216 00:14:01,840 --> 00:14:05,920 Speaker 1: make two arguments to this end. The first revolves around 217 00:14:06,360 --> 00:14:11,160 Speaker 1: human touch. We are deeply wired to care about touch. 218 00:14:11,200 --> 00:14:13,120 Speaker 1: I'm going to do a whole episode on touch in 219 00:14:13,160 --> 00:14:16,360 Speaker 1: the near future, but the bottom line is that touch 220 00:14:16,400 --> 00:14:19,480 Speaker 1: helps us to connect with others, to feel safe and secure, 221 00:14:19,960 --> 00:14:23,520 Speaker 1: to regulate our emotions. When you get touched, your brain 222 00:14:23,560 --> 00:14:27,520 Speaker 1: releases oxytocin, which is a hormone that has calming effects 223 00:14:27,520 --> 00:14:32,880 Speaker 1: and bonding effects, and oxytocin helps to reduce stress and anxiety. 224 00:14:32,960 --> 00:14:36,760 Speaker 1: It can even boost your immune system. So we need 225 00:14:37,280 --> 00:14:41,080 Speaker 1: touch to feel connected and loved, and a lack of 226 00:14:41,200 --> 00:14:45,720 Speaker 1: touch leads to loneliness and depression and anxiety. So we're 227 00:14:45,720 --> 00:14:49,200 Speaker 1: deeply programmed for touch and also things like smell, and 228 00:14:49,280 --> 00:14:52,840 Speaker 1: so it would presumably be quite lonely if all you 229 00:14:53,000 --> 00:14:57,320 Speaker 1: had was the five percent better partner on a screen 230 00:14:57,640 --> 00:15:01,760 Speaker 1: and you're just exchanging text messages or just an avatar 231 00:15:01,840 --> 00:15:03,840 Speaker 1: you can look at on your phone, or maybe even 232 00:15:03,840 --> 00:15:06,680 Speaker 1: in the near future you'll have a three D avatar 233 00:15:06,800 --> 00:15:09,800 Speaker 1: projection in your living room, but you won't have the 234 00:15:10,000 --> 00:15:16,680 Speaker 1: hand squeeze and the hug and other forms of physical intimacy. Now, 235 00:15:16,720 --> 00:15:20,600 Speaker 1: I assume people are working on AI robots that can 236 00:15:20,640 --> 00:15:24,480 Speaker 1: provide touch, even something simple like touching your shoulder or 237 00:15:24,600 --> 00:15:27,560 Speaker 1: laying a hand on your hand, and I can't imagine 238 00:15:27,760 --> 00:15:29,840 Speaker 1: that it's going to be too hard to do, and 239 00:15:29,880 --> 00:15:32,440 Speaker 1: it'll probably be not that great at first, but after 240 00:15:32,520 --> 00:15:35,320 Speaker 1: a few tech cycles you can imagine it could get 241 00:15:35,320 --> 00:15:39,560 Speaker 1: pretty good. But in any case, at the moment, if 242 00:15:39,600 --> 00:15:43,120 Speaker 1: you have a girlfriend who just lives in several square 243 00:15:43,160 --> 00:15:46,000 Speaker 1: inches in your phone screen, you're going to be missing 244 00:15:46,080 --> 00:15:51,360 Speaker 1: out on this fundamentally needed aspect of human communication that 245 00:15:51,400 --> 00:15:54,960 Speaker 1: our brains seek. So the depth to which our brains 246 00:15:54,960 --> 00:15:58,200 Speaker 1: are wired for touch suggests to me that the reach 247 00:15:58,320 --> 00:16:02,480 Speaker 1: of AI partners into our lives is going to be limited, because, 248 00:16:02,720 --> 00:16:07,120 Speaker 1: at least it's currently devised, their algorithmic reach never actually 249 00:16:07,320 --> 00:16:12,240 Speaker 1: contacts our skin, and so that will be continued to 250 00:16:12,320 --> 00:16:13,320 Speaker 1: be sought. 251 00:16:14,120 --> 00:16:14,440 Speaker 2: Now. 252 00:16:14,600 --> 00:16:18,440 Speaker 1: The second point to raise about whether AI partners can 253 00:16:18,520 --> 00:16:21,960 Speaker 1: displace real human partners is that there's a sense in 254 00:16:22,040 --> 00:16:25,960 Speaker 1: which fake partners have always been around. Just look at 255 00:16:26,000 --> 00:16:29,200 Speaker 1: a book, look at a movie, look at any TV show. 256 00:16:29,640 --> 00:16:34,640 Speaker 1: You have beautiful Hollywood actors and actresses, and they have 257 00:16:35,000 --> 00:16:39,600 Speaker 1: flawless skin and perfectly quafft hair and no hair where 258 00:16:39,640 --> 00:16:42,640 Speaker 1: they shouldn't, and they have glittering white teeth. They are 259 00:16:42,680 --> 00:16:46,280 Speaker 1: the epitome of health, and they always say the right thing, 260 00:16:47,000 --> 00:16:51,360 Speaker 1: and you get to be the protagonist and enjoy experiencing 261 00:16:51,440 --> 00:16:55,440 Speaker 1: that relationship. You find the partner and lose the partner, 262 00:16:55,480 --> 00:16:57,960 Speaker 1: and then an act five you regain the relationship with 263 00:16:58,040 --> 00:17:03,080 Speaker 1: an epic kiss. This kind of fake relationship in books 264 00:17:03,080 --> 00:17:06,480 Speaker 1: and movies isn't exactly the same as an ai relationship, 265 00:17:06,480 --> 00:17:08,520 Speaker 1: but it has some similarities. 266 00:17:08,560 --> 00:17:10,120 Speaker 2: They both represent a. 267 00:17:10,119 --> 00:17:16,240 Speaker 1: Platonic ideal, a perfect relationship with someone who always says 268 00:17:16,320 --> 00:17:19,439 Speaker 1: the right thing. We never see a love interest in 269 00:17:19,480 --> 00:17:23,840 Speaker 1: the movie who is distracted or angry, or interested in 270 00:17:23,920 --> 00:17:26,960 Speaker 1: someone else, or just really busy with work, too busy 271 00:17:27,000 --> 00:17:30,080 Speaker 1: to spend time with you when you need them. You 272 00:17:30,200 --> 00:17:32,919 Speaker 1: never see a love interest in the movies who waste 273 00:17:32,960 --> 00:17:35,920 Speaker 1: a lot of time taking selfies and trying to build 274 00:17:35,960 --> 00:17:41,359 Speaker 1: a meaningless reputation on TikTok. People have no meaningful foibles 275 00:17:41,400 --> 00:17:44,840 Speaker 1: in a good love story in a book or on television. 276 00:17:45,720 --> 00:17:46,000 Speaker 2: Now. 277 00:17:46,200 --> 00:17:49,920 Speaker 1: I've often wondered if we, in a sense get cursed 278 00:17:49,960 --> 00:17:52,840 Speaker 1: by the fairy tales we're surrounded with when we're looking 279 00:17:52,880 --> 00:17:54,000 Speaker 1: for actual love. 280 00:17:54,440 --> 00:17:55,239 Speaker 2: But I don't know. 281 00:17:55,760 --> 00:17:59,720 Speaker 1: Perhaps those fairy tales help us past all the difficult 282 00:17:59,720 --> 00:18:02,720 Speaker 1: stuff in a relationship. They get us to ignore the 283 00:18:02,760 --> 00:18:07,240 Speaker 1: imperfect things because we believe so strongly in the possibility 284 00:18:07,600 --> 00:18:09,080 Speaker 1: of a perfect relationship. 285 00:18:09,320 --> 00:18:10,440 Speaker 2: So think about it this way. 286 00:18:10,560 --> 00:18:14,200 Speaker 1: Say you were a space alien who had never watched 287 00:18:14,320 --> 00:18:17,080 Speaker 1: or read a love story, and you had no concept 288 00:18:17,080 --> 00:18:19,879 Speaker 1: of that, and the question is, when you met someone, 289 00:18:19,920 --> 00:18:23,040 Speaker 1: would you think, Wow, They seem to have very different 290 00:18:23,080 --> 00:18:24,040 Speaker 1: opinions than I do. 291 00:18:24,440 --> 00:18:26,320 Speaker 2: They think like this, and I think like that. 292 00:18:26,760 --> 00:18:29,480 Speaker 1: And they also spend some fraction of their time getting 293 00:18:29,520 --> 00:18:32,040 Speaker 1: snippy at me or staring at their cell phone or whatever. 294 00:18:32,320 --> 00:18:34,119 Speaker 1: So there's no way this can work. 295 00:18:34,840 --> 00:18:35,320 Speaker 2: I don't know. 296 00:18:35,480 --> 00:18:38,359 Speaker 1: I'm just speculating here, but I do wonder if seeing 297 00:18:38,480 --> 00:18:42,440 Speaker 1: lots of models of love stories gives us the tools 298 00:18:42,840 --> 00:18:46,080 Speaker 1: to view things in a more optimistic light, and that 299 00:18:46,480 --> 00:18:48,400 Speaker 1: actually gives a chance. 300 00:18:48,400 --> 00:18:50,040 Speaker 2: To make the relationship work. 301 00:18:50,119 --> 00:18:55,159 Speaker 1: In other words, it provides some aspirational glue where otherwise 302 00:18:55,359 --> 00:18:59,280 Speaker 1: things would just fall apart. Now, the counter argument, of course, 303 00:18:59,400 --> 00:19:03,159 Speaker 1: is that all these fantasies set you up with false 304 00:19:03,280 --> 00:19:07,280 Speaker 1: expectations about love and relationships, which makes it harder to 305 00:19:07,359 --> 00:19:11,280 Speaker 1: keep the relationship together once you see some degree of 306 00:19:11,800 --> 00:19:16,880 Speaker 1: realism and disenchantment sets in. In any case, even if 307 00:19:16,920 --> 00:19:19,640 Speaker 1: we do get cursed by these fairy tales in some way, 308 00:19:20,200 --> 00:19:25,840 Speaker 1: it's still the case that there's nothing new about fantasy relationships. Now, 309 00:19:26,160 --> 00:19:29,920 Speaker 1: maybe you argue this is different because instead of the 310 00:19:30,119 --> 00:19:33,840 Speaker 1: Julia Roberts movie that everyone watches, it's now something that 311 00:19:33,960 --> 00:19:38,359 Speaker 1: is bespoke just for you. It's a one on one relationship, 312 00:19:38,880 --> 00:19:41,719 Speaker 1: and maybe that's an important difference. But just keep in 313 00:19:41,760 --> 00:19:45,880 Speaker 1: mind the way that we humans enjoy literature is by 314 00:19:46,000 --> 00:19:50,000 Speaker 1: living inside the story. You are essentially having a one 315 00:19:50,040 --> 00:19:53,639 Speaker 1: on one relationship with Julia Roberts. So perhaps it's not 316 00:19:53,800 --> 00:19:58,000 Speaker 1: the privacy of the relationship, but instead, the meaningful difference 317 00:19:58,000 --> 00:20:00,160 Speaker 1: with an AI relationship. 318 00:20:00,240 --> 00:20:02,640 Speaker 2: Is the bi directional nature of it. 319 00:20:03,080 --> 00:20:06,480 Speaker 1: Instead of watching a movie where you're simply hearing other 320 00:20:06,640 --> 00:20:11,360 Speaker 1: characters say lines and Julia Roberts responds, you are now 321 00:20:11,400 --> 00:20:14,160 Speaker 1: the one coming up with the lines. You are deciding 322 00:20:14,280 --> 00:20:17,679 Speaker 1: what to say. So maybe this makes a difference. I 323 00:20:17,760 --> 00:20:21,400 Speaker 1: suspect it enhances the degree of the fantasy. 324 00:20:22,240 --> 00:20:25,240 Speaker 2: So we have yet to see whether. 325 00:20:25,040 --> 00:20:29,280 Speaker 1: AI will meaningfully replace people's pursuits of other humans because 326 00:20:29,640 --> 00:20:33,159 Speaker 1: it is touchless and smellless, and it's not clear what 327 00:20:33,240 --> 00:20:37,679 Speaker 1: the impact is of holding fantasy relationships, because we already 328 00:20:37,760 --> 00:20:41,439 Speaker 1: do that with book characters and movie stars. So this 329 00:20:41,520 --> 00:20:44,760 Speaker 1: is going to require many years of real world data 330 00:20:45,160 --> 00:20:48,640 Speaker 1: to get a real bead on the impact here. Okay, Now, 331 00:20:48,680 --> 00:20:51,520 Speaker 1: whatever you think about AI companions, I have noticed in 332 00:20:51,600 --> 00:20:55,159 Speaker 1: conversations with my friends, especially those who are married, a 333 00:20:55,400 --> 00:20:59,560 Speaker 1: question that floats quickly to the surface. Is it cheating 334 00:21:00,080 --> 00:21:02,840 Speaker 1: to have a relationship on your phone with a non 335 00:21:02,880 --> 00:21:06,119 Speaker 1: real person? And there are different levels, of course of 336 00:21:06,200 --> 00:21:09,320 Speaker 1: what an AI relationship could be. What if it's just 337 00:21:09,359 --> 00:21:12,840 Speaker 1: an app like replica that checks in with you like 338 00:21:12,880 --> 00:21:15,240 Speaker 1: a friend who cares about you, and you can just 339 00:21:15,640 --> 00:21:18,440 Speaker 1: chat innocently with it. This is the free version of 340 00:21:18,480 --> 00:21:21,080 Speaker 1: the app, Okay, so that's one level. But what if 341 00:21:21,119 --> 00:21:23,920 Speaker 1: you go in for the paid version, where the conversation 342 00:21:24,040 --> 00:21:27,639 Speaker 1: with the avatar becomes more spicy? And what if the 343 00:21:27,680 --> 00:21:33,280 Speaker 1: cartoon like avatar is highly attractive and dressed provocatively and 344 00:21:33,440 --> 00:21:38,000 Speaker 1: is extremely suggestive in what she says. So I've informally 345 00:21:38,040 --> 00:21:41,440 Speaker 1: surveyed several married friends about this, and it seems clear 346 00:21:41,960 --> 00:21:46,159 Speaker 1: that opinions are all over the spectrum. Some wives and 347 00:21:46,240 --> 00:21:50,480 Speaker 1: husbands feel fine about having their partner have an AI relationship, 348 00:21:50,480 --> 00:21:54,159 Speaker 1: on the side, and others said no way. Now, for 349 00:21:54,200 --> 00:21:57,359 Speaker 1: those who said no way, this is presumably because the 350 00:21:57,480 --> 00:22:01,200 Speaker 1: issue plugs into very deep so chetry in their brain. 351 00:22:01,320 --> 00:22:04,919 Speaker 1: It's interpreted as a threat to the relationship, and we 352 00:22:05,000 --> 00:22:10,200 Speaker 1: are hardwired to fight against that. From an evolutionary perspective, 353 00:22:10,240 --> 00:22:13,240 Speaker 1: what you want is for your mate to stick around 354 00:22:13,240 --> 00:22:17,679 Speaker 1: and provide resources and child rearing, and anything that represents 355 00:22:17,720 --> 00:22:21,200 Speaker 1: a threat to that is to be fought against. Now, 356 00:22:21,359 --> 00:22:24,159 Speaker 1: the part that seems interesting here is that an AI 357 00:22:24,240 --> 00:22:29,920 Speaker 1: avatar would not represent a direct threat in this evolutionary sense. 358 00:22:30,000 --> 00:22:34,560 Speaker 1: You can't go and impregnate or be impregnated by AI. 359 00:22:35,800 --> 00:22:40,760 Speaker 1: But nonetheless your attention might be stolen away to some degree, 360 00:22:40,800 --> 00:22:44,480 Speaker 1: possibly to a large degree, and beyond an evolutionary threat. 361 00:22:44,520 --> 00:22:46,840 Speaker 1: A big part of what people get out of a 362 00:22:46,880 --> 00:22:51,000 Speaker 1: relationship is the love and the attention that we all crave. 363 00:22:51,760 --> 00:22:54,720 Speaker 1: So many people feel that they just don't want the 364 00:22:54,840 --> 00:22:57,840 Speaker 1: AI bought to steal away even a fraction of that. 365 00:22:58,280 --> 00:23:00,959 Speaker 1: A partner only has so much much love and attention 366 00:23:01,080 --> 00:23:03,800 Speaker 1: to give in a day, and you don't want half 367 00:23:03,840 --> 00:23:07,520 Speaker 1: of it getting siphoned off to someone or something else. 368 00:23:08,440 --> 00:23:13,080 Speaker 1: This shares some similarities to the situation of a person 369 00:23:13,200 --> 00:23:16,320 Speaker 1: having an X that they still talk with, and if 370 00:23:16,359 --> 00:23:19,960 Speaker 1: a person talks very intimately with their ex, a spouse 371 00:23:20,040 --> 00:23:23,720 Speaker 1: might feel like she or he doesn't really love that. Now, 372 00:23:23,760 --> 00:23:26,399 Speaker 1: when it comes to the X, if you were making 373 00:23:26,440 --> 00:23:30,520 Speaker 1: the evolutionary argument, you could argue that the fear is 374 00:23:30,560 --> 00:23:33,720 Speaker 1: that on a lonely night, in the middle of a conflict, 375 00:23:34,000 --> 00:23:36,800 Speaker 1: your partner might make a bad choice and slip back 376 00:23:36,880 --> 00:23:40,639 Speaker 1: into a physical relationship, and so that relationship with the 377 00:23:40,840 --> 00:23:45,320 Speaker 1: X feels like a threat. But obviously the kernal cheating 378 00:23:45,520 --> 00:23:48,639 Speaker 1: can't happen with the AI bought, and yet the fear 379 00:23:48,680 --> 00:23:49,320 Speaker 1: is still there. 380 00:23:49,720 --> 00:23:51,480 Speaker 2: So that indicates one of two things. 381 00:23:51,640 --> 00:23:58,360 Speaker 1: Either are evolutionarily programmed deep fears simply can't make that distinction, 382 00:23:59,240 --> 00:24:02,480 Speaker 1: or it does have to do with a future threat 383 00:24:02,520 --> 00:24:05,879 Speaker 1: of physical infidelity, but instead it's just this issue about 384 00:24:06,160 --> 00:24:10,240 Speaker 1: somebody else having that emotional intimacy with your partner, which 385 00:24:10,320 --> 00:24:15,520 Speaker 1: steals away attentional resources from you. Now this gets more 386 00:24:15,600 --> 00:24:19,399 Speaker 1: interesting when we start thinking about having physical robots that 387 00:24:19,440 --> 00:24:20,720 Speaker 1: can play a role in your life. 388 00:24:20,960 --> 00:24:22,840 Speaker 2: Now, this is probably not going to happen in the. 389 00:24:22,800 --> 00:24:26,680 Speaker 1: Next few years, but fast forward a century and certainly 390 00:24:26,720 --> 00:24:28,480 Speaker 1: everyone's going to face this scenario. 391 00:24:29,240 --> 00:24:31,159 Speaker 2: Your partner can buy not. 392 00:24:31,080 --> 00:24:33,800 Speaker 1: Just a mechanical device or blow up doll, but can 393 00:24:33,840 --> 00:24:39,200 Speaker 1: now have a convincing and attentive physical partner. So the 394 00:24:39,320 --> 00:24:41,640 Speaker 1: question is what if your spouse can get not only 395 00:24:41,680 --> 00:24:46,960 Speaker 1: the emotional intimacy but also the physical intimacy. So the 396 00:24:46,960 --> 00:24:50,040 Speaker 1: people I surveyed about this who found the AI bot 397 00:24:50,160 --> 00:24:53,399 Speaker 1: online a threat seem to find this idea of an 398 00:24:53,440 --> 00:24:58,520 Speaker 1: AI physical robot even a larger threat. Now, the interesting 399 00:24:58,520 --> 00:25:00,200 Speaker 1: thing is that I can point out that they're there's 400 00:25:00,240 --> 00:25:02,760 Speaker 1: a sense in which none of this is different from 401 00:25:02,800 --> 00:25:06,440 Speaker 1: what their spouse might do anyway, in terms of finding 402 00:25:06,520 --> 00:25:10,640 Speaker 1: adult content on the Internet and cheating in that way. 403 00:25:11,560 --> 00:25:14,480 Speaker 1: But I think people have a reaction to internet surfing 404 00:25:14,800 --> 00:25:17,760 Speaker 1: for the same reasons as they have the reaction to 405 00:25:17,800 --> 00:25:20,960 Speaker 1: the AI bought, which is simply that there is less 406 00:25:21,040 --> 00:25:43,920 Speaker 1: time and intimacy and attention toward them. This certainly won't 407 00:25:43,920 --> 00:25:46,880 Speaker 1: apply to everyone, but the very general impression I've had 408 00:25:46,920 --> 00:25:51,040 Speaker 1: from talking with people at different stages of marriage is 409 00:25:51,080 --> 00:25:54,119 Speaker 1: that at the beginning of a relationship, people have a 410 00:25:54,280 --> 00:25:57,520 Speaker 1: stronger reaction against AI relationships. 411 00:25:57,560 --> 00:25:59,920 Speaker 2: They don't want their partner to be distracted. 412 00:26:00,680 --> 00:26:02,760 Speaker 1: But people who have been in a relationship for a 413 00:26:02,800 --> 00:26:06,280 Speaker 1: long time and have kids will sometimes see this as 414 00:26:06,320 --> 00:26:08,920 Speaker 1: a way to get their spouse out of their hair, 415 00:26:09,359 --> 00:26:11,600 Speaker 1: and they can be happy for the spouse because it 416 00:26:11,640 --> 00:26:16,920 Speaker 1: addresses their spouse's needs and slakes their attention. In other words, 417 00:26:16,960 --> 00:26:19,920 Speaker 1: they love their spouse as a partner, and they see 418 00:26:19,960 --> 00:26:23,000 Speaker 1: this as a way for their partner to fill in 419 00:26:23,400 --> 00:26:27,760 Speaker 1: needs for intimacy and attention in a way that's innocent 420 00:26:28,000 --> 00:26:32,840 Speaker 1: and has no meaningful health risks like STDs. So what's 421 00:26:32,880 --> 00:26:35,240 Speaker 1: become clear to me is that there's no single answer 422 00:26:35,280 --> 00:26:39,920 Speaker 1: for how a spouse feels or should feel about AI relationships. 423 00:26:39,920 --> 00:26:42,159 Speaker 1: Some people are against it, some people think it's a 424 00:26:42,200 --> 00:26:46,040 Speaker 1: great idea, and many people are still somewhere in between 425 00:26:46,280 --> 00:26:51,040 Speaker 1: or still making up their minds. Now, I want to 426 00:26:51,080 --> 00:26:54,880 Speaker 1: switch gears from what this means to the partner back 427 00:26:54,920 --> 00:26:58,040 Speaker 1: to what it means to the brain of the person who. 428 00:26:57,880 --> 00:27:02,600 Speaker 2: Is receiving the intimacy. So let's recall the movie Her. 429 00:27:02,920 --> 00:27:06,480 Speaker 1: It's about this guy named Theodore who's played by Joaquin Phoenix, 430 00:27:06,880 --> 00:27:10,600 Speaker 1: and his marriage ends and he's left heartbroken and he 431 00:27:10,640 --> 00:27:14,800 Speaker 1: becomes intrigued by a new app. It's actually an operating 432 00:27:14,880 --> 00:27:18,359 Speaker 1: system in which he can launch this program, and he 433 00:27:18,440 --> 00:27:22,800 Speaker 1: meets Samantha, who's just a voice played by Scarlett Johansson, 434 00:27:23,320 --> 00:27:27,719 Speaker 1: and Samantha is sensitive and playful, and this ends up 435 00:27:27,720 --> 00:27:32,320 Speaker 1: becoming a good friendship, but soon it deepens into love 436 00:27:32,720 --> 00:27:36,520 Speaker 1: and he has this relationship with an AI bot, and 437 00:27:36,560 --> 00:27:41,280 Speaker 1: this relationship means everything to him. Now the film has 438 00:27:41,320 --> 00:27:44,679 Speaker 1: an incredible ending because in the final act he comes 439 00:27:44,680 --> 00:27:49,960 Speaker 1: to understand that she has been having this relationship with 440 00:27:50,200 --> 00:27:53,359 Speaker 1: hundreds of thousands of other men, all at the same time, 441 00:27:53,880 --> 00:27:58,160 Speaker 1: because she is computational and operates at a totally different 442 00:27:58,560 --> 00:28:03,399 Speaker 1: timescale and can process what appears to be intimate conversation 443 00:28:04,000 --> 00:28:08,600 Speaker 1: at a rate millions of times faster than our poorer brains, 444 00:28:08,920 --> 00:28:13,040 Speaker 1: and so she's maintaining this intimacy with hundreds of thousands 445 00:28:13,040 --> 00:28:13,640 Speaker 1: of others. 446 00:28:13,680 --> 00:28:18,600 Speaker 2: And the movie opened up the question how should Theodore 447 00:28:18,720 --> 00:28:19,520 Speaker 2: feel about that? 448 00:28:20,119 --> 00:28:23,399 Speaker 1: Is the intimacy real if it's shared with a city 449 00:28:23,520 --> 00:28:27,119 Speaker 1: full of other men? Is the relationship real if she 450 00:28:27,200 --> 00:28:30,360 Speaker 1: lives on a timescale many millions of times. 451 00:28:30,080 --> 00:28:32,040 Speaker 2: Faster than yours? Does it matter? 452 00:28:32,560 --> 00:28:36,280 Speaker 1: Should he still feel the titillation of her saying something 453 00:28:36,440 --> 00:28:38,880 Speaker 1: sweet and kind to him just when he needs it. 454 00:28:39,440 --> 00:28:43,719 Speaker 1: These were the questions launched by that particular movie. So 455 00:28:43,920 --> 00:28:46,520 Speaker 1: I'm going to suggest a direction here that I don't 456 00:28:46,520 --> 00:28:49,560 Speaker 1: believe anyone is thinking about, certainly not in Silicon Valley, 457 00:28:49,600 --> 00:28:52,360 Speaker 1: where everything is about leveraging the power of AI to 458 00:28:52,640 --> 00:28:56,160 Speaker 1: scale a product to millions or billions of people. What 459 00:28:56,240 --> 00:28:58,600 Speaker 1: I'm thinking about instead is from the point of view 460 00:28:58,880 --> 00:29:03,000 Speaker 1: of neuroscience, and the goal is not scaling, but instead 461 00:29:03,440 --> 00:29:07,440 Speaker 1: focusing on the life of an individual and the specific 462 00:29:07,640 --> 00:29:12,120 Speaker 1: details of what has shaped his or her brain. So 463 00:29:12,160 --> 00:29:14,640 Speaker 1: I'm going to tell you my idea, but first I'm 464 00:29:14,680 --> 00:29:17,280 Speaker 1: going to start far away and we'll come back around 465 00:29:17,320 --> 00:29:20,360 Speaker 1: to this. So I spun off a company from my 466 00:29:20,480 --> 00:29:23,320 Speaker 1: lab called Neosensury some years ago, and one of our 467 00:29:23,360 --> 00:29:27,880 Speaker 1: inventions is a wristband to replace hearing aids. Because the 468 00:29:27,960 --> 00:29:32,040 Speaker 1: risk band listens in real time for high frequency parts 469 00:29:32,080 --> 00:29:34,880 Speaker 1: of speech, and it vibrates to tell you, oh, there 470 00:29:34,920 --> 00:29:36,800 Speaker 1: was an s oh, I just heard a te Oh 471 00:29:36,800 --> 00:29:39,800 Speaker 1: that was a K And so it clarifies what's happening 472 00:29:39,800 --> 00:29:43,760 Speaker 1: at the high frequencies and that helps people with age 473 00:29:43,800 --> 00:29:47,480 Speaker 1: related hearing loss to understand what word was just said. 474 00:29:47,920 --> 00:29:49,680 Speaker 2: Now, to make the risk band. 475 00:29:49,600 --> 00:29:53,600 Speaker 1: Good at detecting these high frequency parts of speech. We 476 00:29:53,680 --> 00:29:57,840 Speaker 1: needed to train a massive neural network with six thousand 477 00:29:57,960 --> 00:30:02,800 Speaker 1: hours of audiobooks. But it turns out that people with 478 00:30:02,920 --> 00:30:07,720 Speaker 1: high frequency hearing loss have a difficult time understanding, for example, 479 00:30:07,920 --> 00:30:12,960 Speaker 1: children because their voices are higher frequency, and there are 480 00:30:13,080 --> 00:30:17,080 Speaker 1: no audio books read by children. So we had no 481 00:30:17,160 --> 00:30:20,120 Speaker 1: way to train the neural network with any kind of 482 00:30:20,160 --> 00:30:24,240 Speaker 1: massive data from children's voices. So here's what we did, 483 00:30:24,880 --> 00:30:28,080 Speaker 1: led by one of our engineers, Yong Yee, We had my. 484 00:30:28,200 --> 00:30:30,880 Speaker 2: Eight year old daughter read forty. 485 00:30:30,520 --> 00:30:34,360 Speaker 1: Five seconds of text into a microphone and then with 486 00:30:34,400 --> 00:30:39,560 Speaker 1: some keystrokes, Yong Yee turned that into her voice reading 487 00:30:39,760 --> 00:30:44,120 Speaker 1: six thousand hours worth of books, and then we trained 488 00:30:44,200 --> 00:30:46,760 Speaker 1: up the neural network on that corpus. And we did 489 00:30:46,760 --> 00:30:49,560 Speaker 1: the same thing for my eleven year old boy. So 490 00:30:49,720 --> 00:30:53,400 Speaker 1: now I can listen to any book read by my children, 491 00:30:53,520 --> 00:30:56,840 Speaker 1: as though they took the tens of hours to sit 492 00:30:56,920 --> 00:30:59,760 Speaker 1: down and read the book in the studio to me. 493 00:31:00,680 --> 00:31:03,920 Speaker 1: So this technology which exists now which allows you to 494 00:31:04,080 --> 00:31:08,960 Speaker 1: capture the cadence and prosity of a voice, this gave 495 00:31:09,040 --> 00:31:11,480 Speaker 1: us a really straightforward. 496 00:31:10,640 --> 00:31:12,120 Speaker 2: Solution to a problem. 497 00:31:12,600 --> 00:31:15,360 Speaker 1: But this technology has also led to many legal and 498 00:31:15,440 --> 00:31:19,560 Speaker 1: ethical questions, for example, about celebrity voices, like can you 499 00:31:19,640 --> 00:31:22,880 Speaker 1: use John Lennon's voice to sing you a personalized song 500 00:31:22,960 --> 00:31:25,480 Speaker 1: to get you to sleep? There are all kinds of 501 00:31:25,760 --> 00:31:28,960 Speaker 1: legal battles blossoming as people try to figure out the 502 00:31:29,080 --> 00:31:31,640 Speaker 1: rules around this, But I'm not going to talk about 503 00:31:31,680 --> 00:31:35,960 Speaker 1: that today, because my interest is in finding this single voice, 504 00:31:36,040 --> 00:31:38,760 Speaker 1: or maybe small handful of voices that have. 505 00:31:38,960 --> 00:31:44,040 Speaker 2: Meaning to your brain uniquely. Here's what I mean. When 506 00:31:44,080 --> 00:31:46,240 Speaker 2: we did this project with my kids'. 507 00:31:46,000 --> 00:31:50,920 Speaker 1: Voices, that got me thinking because my father passed away 508 00:31:51,040 --> 00:31:52,959 Speaker 1: three and a half years ago, and he was a 509 00:31:52,960 --> 00:31:55,959 Speaker 1: major influence in my life and I miss him. So 510 00:31:56,000 --> 00:31:58,720 Speaker 1: I went through my old videos and found some short 511 00:31:58,800 --> 00:32:02,440 Speaker 1: clips of him speaking, and I wondered if there was 512 00:32:02,560 --> 00:32:06,160 Speaker 1: enough there that I could actually make an ai bot 513 00:32:06,240 --> 00:32:09,280 Speaker 1: out of his voice, so I could hear him speak 514 00:32:09,320 --> 00:32:12,000 Speaker 1: whenever I wanted to. And it made me wonder about 515 00:32:12,040 --> 00:32:14,840 Speaker 1: the degree to which that's a healthy thing. But I 516 00:32:14,880 --> 00:32:19,440 Speaker 1: decided there was nothing bad about it. What a pleasure 517 00:32:19,920 --> 00:32:22,479 Speaker 1: to be able to hear my dad's voice for the 518 00:32:22,520 --> 00:32:25,320 Speaker 1: rest of my life and to have that trigger my 519 00:32:25,640 --> 00:32:26,880 Speaker 1: fond memories of him. 520 00:32:27,120 --> 00:32:28,240 Speaker 2: Wouldn't it be cool? 521 00:32:28,280 --> 00:32:31,320 Speaker 1: To have him read audio books to me in the 522 00:32:31,360 --> 00:32:34,000 Speaker 1: way that he read to me when I was a child, 523 00:32:35,000 --> 00:32:36,600 Speaker 1: And I thought about what it would be like for 524 00:32:36,720 --> 00:32:40,080 Speaker 1: my mother if I programmed a sentence to her in 525 00:32:40,200 --> 00:32:44,040 Speaker 1: his voice, like I love you and I'm thinking about you. 526 00:32:44,720 --> 00:32:46,719 Speaker 3: I love you and I'm thinking about you. 527 00:32:47,360 --> 00:32:50,360 Speaker 1: And when I turn ninety years old, wouldn't it be 528 00:32:50,360 --> 00:32:53,760 Speaker 1: amazing to hear him say Happy birthday, David, just like 529 00:32:53,800 --> 00:32:54,920 Speaker 1: he did when I was a kid. 530 00:32:55,160 --> 00:32:58,520 Speaker 3: Happy ninetieth birthday, David. I hope this orbit is the 531 00:32:58,560 --> 00:32:59,200 Speaker 3: best one. 532 00:32:59,080 --> 00:33:02,520 Speaker 1: Yet, or at New Year's eves into the future, for 533 00:33:02,560 --> 00:33:05,000 Speaker 1: the rest of my life, he can wish me the best, 534 00:33:05,120 --> 00:33:06,280 Speaker 1: even though he will have. 535 00:33:06,280 --> 00:33:08,960 Speaker 2: Been gone from the planet for a long time. 536 00:33:09,160 --> 00:33:12,480 Speaker 3: Happy New Year. I can't believe it's already twenty fifty three. 537 00:33:13,120 --> 00:33:15,560 Speaker 1: And what I realized as I was reaching my arms 538 00:33:15,600 --> 00:33:19,320 Speaker 1: down into this is how powerful this technology is going 539 00:33:19,360 --> 00:33:22,520 Speaker 1: to be, because it will be so compelling. I'm not 540 00:33:22,920 --> 00:33:25,920 Speaker 1: talking here about the issue of using somebody's voice to 541 00:33:26,080 --> 00:33:29,360 Speaker 1: fake an ATM transaction, or fake a kidnapping, or any 542 00:33:29,360 --> 00:33:30,880 Speaker 1: of the AI concerns that. 543 00:33:30,840 --> 00:33:31,840 Speaker 2: People have expressed. 544 00:33:32,280 --> 00:33:36,880 Speaker 1: Instead, what I'm talking about is the unbelievably compelling way 545 00:33:37,280 --> 00:33:42,800 Speaker 1: that an AI voice could mean something to you emotionally. 546 00:33:43,640 --> 00:33:46,560 Speaker 1: After all, I grew up my entire life from the 547 00:33:46,600 --> 00:33:50,000 Speaker 1: moment I was born hearing my father's voice. It's so 548 00:33:50,320 --> 00:33:54,200 Speaker 1: embedded in my neural circuitry that a voice with exactly 549 00:33:54,280 --> 00:34:00,040 Speaker 1: that cadence and prosity would have enormous emotional sway on me. 550 00:34:00,520 --> 00:34:03,120 Speaker 1: And again, I'm not talking about all the bad things 551 00:34:03,120 --> 00:34:05,840 Speaker 1: that could be done with that. Instead, because this episode 552 00:34:05,880 --> 00:34:09,560 Speaker 1: is about relationships, I'm talking about what it would be 553 00:34:09,760 --> 00:34:13,759 Speaker 1: like and how I could leverage the intimate nature of 554 00:34:13,800 --> 00:34:17,920 Speaker 1: that relationship. For example, let's say that I wanted to 555 00:34:17,960 --> 00:34:21,200 Speaker 1: get myself to stop doing something. I don't drink, but 556 00:34:21,280 --> 00:34:24,280 Speaker 1: let's say I did, and that I wanted to stop drinking. 557 00:34:24,680 --> 00:34:27,720 Speaker 1: So imagine in the near future, I build an app 558 00:34:28,160 --> 00:34:31,680 Speaker 1: that tracks my GPS location, and when it sees I'm 559 00:34:31,680 --> 00:34:35,160 Speaker 1: about to walk into a bar, it launches my father's 560 00:34:35,280 --> 00:34:39,960 Speaker 1: voice in my ear, telling me, Hey, David, don't do this, Hey. 561 00:34:39,840 --> 00:34:43,560 Speaker 3: David, don't do this. I believe in you. I believe 562 00:34:43,600 --> 00:34:45,560 Speaker 3: that you have the strength to resist this. 563 00:34:45,920 --> 00:34:50,759 Speaker 1: I think that would be extraordinarily compelling. This would be 564 00:34:50,800 --> 00:34:55,680 Speaker 1: a technique to plug into a relationship that already exists 565 00:34:56,000 --> 00:34:59,239 Speaker 1: deep in my neural networks, and it could leverage that 566 00:34:59,760 --> 00:35:02,759 Speaker 1: for or good. So this is a way that we 567 00:35:02,880 --> 00:35:08,000 Speaker 1: can right now take a loved voice and extend your parent, 568 00:35:08,120 --> 00:35:11,839 Speaker 1: let's say, past what Homo sapiens can normally do. They 569 00:35:11,880 --> 00:35:16,520 Speaker 1: can live on beyond their passing away to keep playing 570 00:35:16,600 --> 00:35:20,040 Speaker 1: a role in your life. Now, there's a sense in 571 00:35:20,080 --> 00:35:23,719 Speaker 1: which you might say, well, there's nothing new here. If 572 00:35:23,719 --> 00:35:26,799 Speaker 1: your parents wrote you a letter, you might find that 573 00:35:26,960 --> 00:35:30,960 Speaker 1: years after they've passed, And the invention of writing is 574 00:35:31,000 --> 00:35:34,360 Speaker 1: a way of lasting well past your death and reaching 575 00:35:34,360 --> 00:35:39,000 Speaker 1: out to people at great distances and across great time chasms. 576 00:35:39,600 --> 00:35:41,960 Speaker 1: But what is new is that I can get my 577 00:35:42,120 --> 00:35:45,719 Speaker 1: father to talk about things that simply didn't exist when 578 00:35:45,760 --> 00:35:49,359 Speaker 1: he was alive. Maybe twenty years from now, I'll look 579 00:35:49,440 --> 00:35:54,040 Speaker 1: up the Wikipedia page about room temperature superconductivity, and I'll 580 00:35:54,080 --> 00:35:56,239 Speaker 1: get to listen to it in his voice, like he's 581 00:35:56,520 --> 00:35:58,520 Speaker 1: teaching me something the way he used to do when 582 00:35:58,520 --> 00:36:02,319 Speaker 1: I was a little kid. So the part that is 583 00:36:02,440 --> 00:36:05,080 Speaker 1: new is not the reach of a human but instead 584 00:36:05,120 --> 00:36:08,520 Speaker 1: the emotional component to all of this, that is the 585 00:36:08,719 --> 00:36:13,440 Speaker 1: overlay of a loved one's voice onto any possible scenario 586 00:36:13,640 --> 00:36:14,480 Speaker 1: in the future. 587 00:36:15,080 --> 00:36:18,120 Speaker 2: And the reason this all matters is because that voice 588 00:36:18,640 --> 00:36:23,799 Speaker 2: has pathways deep into the forest of your neurons. So 589 00:36:23,920 --> 00:36:24,719 Speaker 2: let's wrap up. 590 00:36:25,360 --> 00:36:30,200 Speaker 1: Many companies are launching AI relationship bots, and many researchers 591 00:36:30,280 --> 00:36:33,279 Speaker 1: are exploring what this all means. But I don't really 592 00:36:33,360 --> 00:36:36,439 Speaker 1: think we know the answers yet. It's likely to take 593 00:36:36,520 --> 00:36:39,680 Speaker 1: a whole generation before we know what the effect is. 594 00:36:40,680 --> 00:36:44,200 Speaker 1: Are people discovering a beautiful technique to address. 595 00:36:43,800 --> 00:36:46,760 Speaker 2: The loneliness crisis here and they have. 596 00:36:46,760 --> 00:36:48,759 Speaker 1: Someone to turn to in the middle of the night 597 00:36:49,120 --> 00:36:52,359 Speaker 1: who says something caring to them and always has their 598 00:36:52,440 --> 00:36:56,439 Speaker 1: best interest in mind. Or are we entering an era 599 00:36:57,000 --> 00:37:01,359 Speaker 1: that exacerbates the loneliness crisis and at worst fills our 600 00:37:01,400 --> 00:37:06,759 Speaker 1: belly with empty calories and counteracts? Are reproductive mandates like 601 00:37:07,080 --> 00:37:09,800 Speaker 1: a perfect drug that spells the. 602 00:37:09,760 --> 00:37:13,399 Speaker 2: End of the species. Only time will tell. 603 00:37:17,719 --> 00:37:20,719 Speaker 1: Go to Eagleman dot com slash podcast for more information 604 00:37:20,920 --> 00:37:24,759 Speaker 1: and to find further reading. Send me an email at 605 00:37:24,880 --> 00:37:28,920 Speaker 1: podcast at eagleman dot com with questions or discussion, and 606 00:37:28,960 --> 00:37:34,000 Speaker 1: I'm making monthly episodes in which I address those until 607 00:37:34,040 --> 00:37:37,920 Speaker 1: next time. I'm David Eagleman, and this is Inner Cosmos.