WEBVTT - The Story: What is AI Love? w/ David Eagleman 0:00:01.480 --> 0:00:04.960 Welcome to tech Stuff. This is the story. Each week 0:00:05.000 --> 0:00:07.560 on Wednesdays, we bring you an in depth interview with 0:00:07.600 --> 0:00:10.039 someone who has a front row seat to the most 0:00:10.039 --> 0:00:14.680 fascinating things happening in tech today. We're joined by David Eagleman. 0:00:20.120 --> 0:00:25.520 David Eagleman is a neuroscientist, Stanford University professor, author, entrepreneur, 0:00:25.840 --> 0:00:29.920 and host of the podcast Inner Cosmos. In his podcast, 0:00:30.160 --> 0:00:33.599 Eagleman explores how our brain interprets the world and how 0:00:33.680 --> 0:00:37.400 that influences our perception. Ideas have been a source of 0:00:37.440 --> 0:00:41.360 fascination for Eagleman ever since he experienced an Alice in 0:00:41.440 --> 0:00:43.440 Wonderland moment many years ago. 0:00:44.240 --> 0:00:47.240 I fell off of the roof of a house when 0:00:47.280 --> 0:00:48.160 I was a kid. 0:00:48.159 --> 0:00:49.000 I was eight years old. 0:00:49.440 --> 0:00:52.239 The fall seemed to take a very long time, and 0:00:52.320 --> 0:00:54.720 I was thinking about Alice in Wonderland on the way down, 0:00:54.760 --> 0:00:57.400 about how this must have been what it was like 0:00:57.440 --> 0:00:58.960 for her to fall down the rabbit hole. 0:00:59.320 --> 0:01:01.959 He told me that the whole journey was very calm, 0:01:02.240 --> 0:01:04.560 but that it took a long time, and he remembers 0:01:04.560 --> 0:01:07.360 having lots and lots of thoughts on his way to 0:01:07.400 --> 0:01:08.280 the bottom. 0:01:08.560 --> 0:01:11.000 And then I hit the ground and broke my nose 0:01:11.080 --> 0:01:13.280 very badly. But when I got older and went to 0:01:13.319 --> 0:01:16.399 high school, I realized that the whole fall had taken 0:01:16.520 --> 0:01:18.959 point six of a second. I calculated the eagles one 0:01:19.000 --> 0:01:21.319 have at squared, and I couldn't believe that it was 0:01:21.360 --> 0:01:22.640 so so fast. 0:01:23.040 --> 0:01:27.360 When Eagleman became a neuroscientist, he kept exploring this relationship 0:01:27.360 --> 0:01:30.800 between our brain and our experiences, which led him to 0:01:30.840 --> 0:01:34.200 research things like is it possible to create a new 0:01:34.319 --> 0:01:37.600 sense or is there a new way of interpreting an 0:01:37.600 --> 0:01:41.440 already existing sense. This turned into a tech product for 0:01:41.480 --> 0:01:45.600 his company, Neosensory, that could help deaf people understand the 0:01:45.600 --> 0:01:50.200 auditory world through a vibrating wristband. David Eagleman is an 0:01:50.240 --> 0:01:54.320 innovator with a fascination for tech driven interventions, so we 0:01:54.320 --> 0:01:56.880 weren't surprised to see an episode of the Inner Cosmos 0:01:56.920 --> 0:02:01.680 podcast all about AI relationships, a widely reported phenomenon that 0:02:01.680 --> 0:02:04.280 we hear it tech stuff have been eager to explore. 0:02:04.960 --> 0:02:07.480 So we thought we'd ask Eagleman to help us understand 0:02:07.640 --> 0:02:12.000 how our brains form these attachments. But before we launch 0:02:12.040 --> 0:02:15.160 into that discussion, I asked Eagleman to tell me a 0:02:15.200 --> 0:02:19.040 bit more about the driving question. Behind his work, How 0:02:19.080 --> 0:02:21.519 does the human brain construct reality? 0:02:22.200 --> 0:02:25.840 Your brain is locked in silence and darkness inside your skull, 0:02:26.400 --> 0:02:29.760 and it only has spikes coming in. That's all it 0:02:29.760 --> 0:02:33.520 ever has, these little electrical signals. So your eyes aren't 0:02:33.919 --> 0:02:37.520 pushing light through, and your ears aren't pushing sound through. Instead, 0:02:37.600 --> 0:02:40.600 your eyes are converting photons into spikes, and ears are 0:02:40.600 --> 0:02:45.079 converting sound waves into spikes, and your fingertips are converting 0:02:45.080 --> 0:02:48.880 pressure and temperature into spikes and so on. So inside 0:02:48.919 --> 0:02:52.200 the brain there's nothing but these little electrical spikes running around, 0:02:52.440 --> 0:02:54.760 and it's living in the darkness and trying to figure 0:02:54.760 --> 0:02:58.000 out what is the world out there? And so everything 0:02:58.040 --> 0:03:00.760 that we experience as reality is actually the construction. 0:03:00.880 --> 0:03:01.120 You know. 0:03:01.400 --> 0:03:05.079 For example, things like colors don't exist in the outside world. 0:03:05.120 --> 0:03:08.320 This is just a way to tag information as in 0:03:08.400 --> 0:03:11.760 that's a different wavelength than that over there, so they 0:03:11.800 --> 0:03:14.600 appear to be different colors. But this is the fascinating 0:03:14.760 --> 0:03:18.400 question that we're stuck with. We are inside Plato's cave, 0:03:18.720 --> 0:03:22.240 seeing shadows on the wall and trying to understand what 0:03:22.400 --> 0:03:23.280 is outside. 0:03:24.000 --> 0:03:27.080 Speaking of perception, did you ever figure out an answer 0:03:27.200 --> 0:03:30.440 to that question of why time slows down when our 0:03:30.520 --> 0:03:31.640 lives at risk. 0:03:32.120 --> 0:03:34.160 I did a number of experiments on this, actually the 0:03:34.160 --> 0:03:35.760 only experiments that I have ever been done on this 0:03:35.880 --> 0:03:38.000 question of the perception of time. 0:03:38.560 --> 0:03:39.120 Yeah, on the. 0:03:39.160 --> 0:03:40.880 Question of why does time slow now when you're in 0:03:40.920 --> 0:03:42.680 fear for your life. The reason no one had ever 0:03:42.720 --> 0:03:45.760 done experiments is because you have to actually put subjects 0:03:45.760 --> 0:03:47.480 in fear for their life, which is very hard to 0:03:47.480 --> 0:03:51.080 get approved. And so I dropped people from one hundred 0:03:51.080 --> 0:03:55.520 and fifty foot tall tower backwards in free fall, and 0:03:55.560 --> 0:03:58.120 they're caught in a net below, going seventy miles an hour, 0:03:58.600 --> 0:04:01.680 and I built a di ice that strapped to their 0:04:01.720 --> 0:04:03.880 wrists and it flashes information at them. And I was 0:04:03.920 --> 0:04:08.720 able to measure whether they actually are seeing in slow 0:04:08.800 --> 0:04:12.160 motion during a scary event. And it turns out that 0:04:12.400 --> 0:04:15.240 it's all a trick of memory. What happens when you're 0:04:15.280 --> 0:04:17.520 in fear for your life is your brain is writing 0:04:17.560 --> 0:04:19.919 down every single thing. Normally, your brain is like a 0:04:19.960 --> 0:04:22.520 sieve and most things just passed right through, but you're 0:04:22.960 --> 0:04:26.680 capturing all the information. So it's not that you're seeing 0:04:26.720 --> 0:04:29.640 things in slow motion. It's just the density of memory. 0:04:29.760 --> 0:04:31.000 You're remembering everything. 0:04:31.120 --> 0:04:31.320 You know. 0:04:31.320 --> 0:04:33.919 You're in a car actually say, wow, I remember the 0:04:34.000 --> 0:04:36.520 hood crumpling and the rear view mirror falling off, and 0:04:36.560 --> 0:04:39.400 the expression on the other guy's face. And you've got 0:04:39.440 --> 0:04:42.120 all these details, and your brain is doing a calculation 0:04:42.360 --> 0:04:44.520 where it says, oh, well, if I have all those details, 0:04:44.560 --> 0:04:47.640 it must have taken five seconds as opposed to one 0:04:47.680 --> 0:04:49.080 second for this all to happen. 0:04:49.760 --> 0:04:51.919 One of the kind of topics of tech stuff is 0:04:52.440 --> 0:04:56.159 what are the most exciting or promising paradigms for human 0:04:56.200 --> 0:05:00.240 machine interactions, particularly a time when there's so much concern 0:05:00.360 --> 0:05:03.760 about the negative effects of technology on the human brain. Obviously, 0:05:04.040 --> 0:05:06.160 the question of like, what is this doing to us 0:05:06.160 --> 0:05:10.039 in our brains and particularly developing brains. Do you have 0:05:10.080 --> 0:05:12.120 a kind of internal paradigm for how you think about 0:05:12.160 --> 0:05:14.480 the interaction between technology and our brains? 0:05:14.680 --> 0:05:17.560 More broadly, yeah, I have to say, I'll just put 0:05:17.560 --> 0:05:20.800 my cards on the table. Very cyber optimistic on these points, 0:05:20.839 --> 0:05:26.400 but I'll explain why. One is really what we want 0:05:26.440 --> 0:05:30.360 for our children is the largest available diet of information. 0:05:30.920 --> 0:05:33.640 When I was growing up, I had my homeroom teacher 0:05:33.839 --> 0:05:37.920 in Albuquerque, New Mexico, and you know, whatever she knew 0:05:38.000 --> 0:05:40.440 or didn't know, that's what I would learn, and I 0:05:40.480 --> 0:05:42.279 would get my mother to drive me down to the 0:05:42.360 --> 0:05:46.480 library and pull out an Encyclopedia Britannica and hope that 0:05:46.520 --> 0:05:48.760 there was an article about the thing that I wanted. 0:05:49.000 --> 0:05:51.000 And maybe the article wasn't you know, more than ten 0:05:51.080 --> 0:05:53.279 or fifteen years old, I hoped. And you know, this 0:05:53.400 --> 0:05:56.080 was really all the information that I had access to. 0:05:56.200 --> 0:06:01.839 But now my kids, growing up, the entire world's knowledge 0:06:01.839 --> 0:06:05.919 at their fingertips, and it's the greatest thing for two reasons. 0:06:05.960 --> 0:06:08.720 One is what I got growing up was lots of 0:06:08.920 --> 0:06:11.880 just in case information, like, just in case you ever 0:06:11.920 --> 0:06:13.560 need to know that the Battle of Hastings was in 0:06:13.600 --> 0:06:16.560 ten sixty six, here you go. But kids now get 0:06:16.560 --> 0:06:19.040 lots of just in time information. So as soon as 0:06:19.040 --> 0:06:23.560 they're curious about something, they look it up, they ask ALEXA, 0:06:23.680 --> 0:06:25.640 they type it in their phone, they do whatever, and 0:06:25.680 --> 0:06:29.320 they get the information in the context of their curiosity. 0:06:29.640 --> 0:06:32.000 And what that means is they have the right cocktail 0:06:32.040 --> 0:06:36.640 of neurotransmitters present for that information to stick. I mentioned 0:06:36.720 --> 0:06:39.200 before that most things just passed right through our brains, 0:06:39.480 --> 0:06:42.919 but when you care about something, it sticks. And the 0:06:43.000 --> 0:06:47.240 fact is that everything that we create is a remix 0:06:47.400 --> 0:06:50.520 of things that we've seen before. And so if you 0:06:50.600 --> 0:06:53.520 have a larger warehouse of things that you have seen 0:06:53.560 --> 0:06:57.240 and experienced, you're going to create better stuff. Just take 0:06:57.279 --> 0:07:00.240 as an example something like music. You know, if you 0:07:00.279 --> 0:07:02.640 grow up and you're somewhere in the world, let's say, 0:07:02.640 --> 0:07:05.080 two hundred years ago, all you're ever going to hear 0:07:05.200 --> 0:07:07.799 is the music of your local little area. But now 0:07:08.240 --> 0:07:11.040 kids can listen to music from all over the world, 0:07:11.080 --> 0:07:14.160 from all through history, and really put together bigger and 0:07:14.200 --> 0:07:18.400 bigger things. So this is why we're having this exponential 0:07:18.520 --> 0:07:22.640 increase in humankind's knowledge and innovation is because people are 0:07:22.640 --> 0:07:25.920 doing remixes across space and time in a way that's 0:07:25.960 --> 0:07:30.280 never been possible before. So I'm very optimistic about this. Yes, 0:07:30.400 --> 0:07:34.520 kids waste lots of time on social media, surfing dumb websites, 0:07:34.560 --> 0:07:38.240 things like that, but I think the good here outweighs 0:07:38.280 --> 0:07:38.640 the bad. 0:07:39.080 --> 0:07:42.400 As the father of children, are we already seeing how 0:07:42.440 --> 0:07:43.960 these brains are developing differently. 0:07:44.520 --> 0:07:47.400 Yes, I mean these brains are developing very differently. But 0:07:48.080 --> 0:07:50.760 this is a very interesting thing. Lots of people in 0:07:50.800 --> 0:07:53.080 the media will pipe off with opinions on this, but 0:07:53.120 --> 0:07:55.520 the fact is it's very, very difficult to do a 0:07:55.560 --> 0:07:58.720 scientific experiment on this because of the lack of a 0:07:58.800 --> 0:08:03.240 control group, other words, my children. I can't find other 0:08:03.360 --> 0:08:06.120 kids who are the same age, who haven't grown up 0:08:06.160 --> 0:08:09.160 in the same circumstances, unless you find children who are 0:08:09.200 --> 0:08:13.080 let's say, Quaker, who don't use technology or very very 0:08:13.160 --> 0:08:15.840 impoverished in different places in the world. But there are 0:08:15.840 --> 0:08:18.320 a hundred other differences there. If I see a difference 0:08:18.360 --> 0:08:21.040 between my children and those children, I don't know if 0:08:21.040 --> 0:08:23.040 it's because of the tech or because of something else 0:08:23.160 --> 0:08:26.400 like diet or politics or whatever. And I can't take 0:08:26.440 --> 0:08:28.600 my children and compare them, let's say, to me in 0:08:28.600 --> 0:08:32.520 a previous generation, because there are a hundred other differences 0:08:32.520 --> 0:08:35.880 there in terms of pollution and politics and other tech 0:08:35.920 --> 0:08:38.600 and so on. So it's very difficult to do this 0:08:39.160 --> 0:08:41.680 in a control way. There are some studies, for example, 0:08:42.080 --> 0:08:44.720 looking at eye movements. You know, your eyes are jumping 0:08:44.760 --> 0:08:47.360 around all the time, about three times a second. These 0:08:47.360 --> 0:08:49.960 are called the CODs, and you can measure these things 0:08:50.000 --> 0:08:51.920 and you see that people, let's say, in my generation 0:08:52.040 --> 0:08:55.320 versus my children's generation, actually move our eyes around differently 0:08:55.400 --> 0:08:58.720 when looking at text because they're reading it more like 0:08:59.160 --> 0:09:02.280 a web page, where they sort of scan across and 0:09:02.320 --> 0:09:04.760 then they jump down and then get across and so on, 0:09:05.040 --> 0:09:08.080 whereas I in my generation it was more like a 0:09:08.200 --> 0:09:12.040 zigzag pattern down the page. These things that are totally unconscious, 0:09:12.400 --> 0:09:15.840 you can measure these, But in general, are my kids 0:09:15.840 --> 0:09:17.679 turning out differently? Is this whole generation? 0:09:18.160 --> 0:09:18.400 Yes? 0:09:18.559 --> 0:09:19.280 Absolutely. 0:09:19.840 --> 0:09:22.400 I love that idea that the more curious you are 0:09:22.800 --> 0:09:24.920 that the better you're able to form memory. I mean, 0:09:24.960 --> 0:09:26.600 there are so many things we could talk about today, 0:09:26.640 --> 0:09:31.120 but I'm personally very fascinated by AI relationships, and I 0:09:31.160 --> 0:09:33.439 have a hypothesis that your listeners are too, because I 0:09:33.440 --> 0:09:34.800 saw that you did an episode on it and you 0:09:34.880 --> 0:09:37.839 rebroadcast it, so I can imagine it was a favorite. 0:09:39.120 --> 0:09:40.920 I wanted to ask you, what's the lay of the land, Like, 0:09:40.960 --> 0:09:44.240 what is an AI relationship? How broad is a phenomenon? 0:09:44.679 --> 0:09:47.319 Who does it affect? And why did you choose to 0:09:47.360 --> 0:09:47.920 focus on it? 0:09:48.240 --> 0:09:50.200 Why I choose to focus on it as one of 0:09:50.240 --> 0:09:52.280 the many things I'm focusing on is just because it's 0:09:52.280 --> 0:09:55.240 totally new. We've never been in a situation before where 0:09:55.240 --> 0:09:58.600 we've ever talked about, Hey, did you fall in love 0:09:58.600 --> 0:10:01.640 with your machine? It's this is so wacky and new, 0:10:01.880 --> 0:10:04.880 but Apparently. I was just talking to a researcher, Bethany 0:10:04.920 --> 0:10:07.800 Maples about this yesterday. Apparently there are a billion people 0:10:07.840 --> 0:10:10.319 now on the planet who have AI relationships of one 0:10:10.320 --> 0:10:12.520 sort or another. Some of these are friendships, some of 0:10:12.520 --> 0:10:14.600 them are romantic relationships. 0:10:14.160 --> 0:10:20.000 Defined by an ongoing emotional connection with a non human. 0:10:19.760 --> 0:10:22.319 Being exactly right, exactly right with an app. And in 0:10:22.720 --> 0:10:26.239 this country we have, for example, character AI or Replica. 0:10:26.360 --> 0:10:28.199 These are different companies that do this sort of thing. 0:10:28.679 --> 0:10:31.920 I find this very interesting that the concern that a 0:10:31.920 --> 0:10:34.800 lot of people have is, hey, is this going to 0:10:34.920 --> 0:10:38.960 ruin relationships? But here's why I'm optimistic about this as well. 0:10:39.000 --> 0:10:41.760 First of all, people are having these chats. These chats 0:10:41.840 --> 0:10:45.160 become steamy, pillow whispers, all kinds of stuff like that. That's great, 0:10:45.200 --> 0:10:47.840 But fundamentally, what you want a relationship is you want 0:10:47.840 --> 0:10:50.319 to take your girlfriend out to dinner, you want to 0:10:50.360 --> 0:10:52.959 introduce to your friends. I don't think that's going away. 0:10:53.040 --> 0:10:55.120 And of course the main thing is physical touch. I 0:10:55.160 --> 0:10:58.119 mean this is you know, we have deep evolutionary programming 0:10:58.200 --> 0:11:00.959 driving us towards that. So I think a chatbot is 0:11:01.000 --> 0:11:05.440 going to replace relationships. What I am very hopeful about 0:11:05.559 --> 0:11:11.000 is the idea that AI relationships will actually improve real 0:11:11.000 --> 0:11:14.760 life relationships because it's a sandbox. People can try things out, 0:11:15.040 --> 0:11:16.079 people can get better. 0:11:16.120 --> 0:11:16.280 Now. 0:11:16.280 --> 0:11:21.679 Obviously, what this requires is AI companies that make bots 0:11:21.720 --> 0:11:24.320 in a way that give you the right amount of pushback, 0:11:24.920 --> 0:11:28.840 and maybe the bot sometimes gets angry or snarky or 0:11:28.880 --> 0:11:32.280 things like that, so you learn your way through these situations. 0:11:32.320 --> 0:11:34.520 But really this is what we all go through as 0:11:34.559 --> 0:11:37.880 young people. We date, we screw things up a lot, 0:11:37.920 --> 0:11:40.040 and eventually we get better and better at knowing how 0:11:40.080 --> 0:11:42.640 to be a partner. And so if there's a way 0:11:42.880 --> 0:11:46.840 that people can get practice at this, and one of 0:11:46.840 --> 0:11:49.880 the interesting research questions that's coming out now is about 0:11:49.920 --> 0:11:54.240 whether these AI relationships can serve as a way station, 0:11:54.520 --> 0:11:58.360 meaning they help people sort of dig themselves out of 0:11:58.400 --> 0:12:01.040 this whole of loneliness and then they go out and meet. 0:12:00.800 --> 0:12:06.960 Real people when we come back. Why our brains are 0:12:07.000 --> 0:12:20.960 so quick to anthropomorphize chatbots stay with us? David, I 0:12:21.000 --> 0:12:24.760 wonder if you could explain from the perspective of a neuroscientist, 0:12:25.320 --> 0:12:29.720 what's the difference between me using replica or character AI 0:12:30.480 --> 0:12:33.000 and me asking chat GPT, you know, what should I 0:12:33.000 --> 0:12:35.440 have for dinner tonight, or how should I approach this 0:12:35.520 --> 0:12:36.960 difficult conversation with my mother. 0:12:37.800 --> 0:12:42.400 The thing with character replica is that there is a 0:12:43.400 --> 0:12:47.400 you know, a character, and sometimes it's represented with a 0:12:47.480 --> 0:12:51.840 visual avatar also, and we have these intensely social brains 0:12:52.400 --> 0:12:56.800 where we don't have this circuitry to really distinguish something 0:12:56.840 --> 0:12:59.720 that's fake from something that's real. And so if you 0:12:59.840 --> 0:13:03.640 are talking to your avatar every day, and by the way, 0:13:03.800 --> 0:13:07.200 she can send you text messages and say hey, as 0:13:07.240 --> 0:13:09.880 how you feeling today, and then it becomes in your 0:13:09.960 --> 0:13:13.559 mind like a real person, it's really hard to distinguish 0:13:13.679 --> 0:13:15.520 real from fake. You know, there's this great scene in 0:13:15.559 --> 0:13:18.920 Westworld in the first episode where William walks in. He's 0:13:18.920 --> 0:13:23.200 getting outfitted by this woman and he awkwardly asked her. 0:13:23.240 --> 0:13:25.280 He said, I'm so sorry, but I can't tell. Are 0:13:25.320 --> 0:13:28.520 you real? And she says, if you can't tell, does 0:13:28.559 --> 0:13:31.600 it matter? And I think this is the situation that 0:13:31.640 --> 0:13:32.240 we're in now. 0:13:32.960 --> 0:13:35.480 There was a New York Times article recently about this phenomenon, 0:13:35.480 --> 0:13:37.560 and one of the subjects, or a person who has 0:13:37.559 --> 0:13:40.439 a relationship with an AI character, said, I don't actually 0:13:40.480 --> 0:13:42.520 believe he's real. That the effects that he has on 0:13:42.559 --> 0:13:44.800 my life are real. The feeling that he brings out 0:13:44.800 --> 0:13:47.120 of me are real, so I treat it as a 0:13:47.160 --> 0:13:48.120 real relationship. 0:13:48.480 --> 0:13:50.679 So I just read an article last night about this 0:13:50.760 --> 0:13:54.200 question of people saying thank you to chat GPT when 0:13:54.240 --> 0:13:56.480 it gives you a nice answer, and I find that 0:13:56.520 --> 0:13:58.679 I often say thank you. I say, oh, that was excellent, 0:13:58.679 --> 0:14:02.240 thank you, and so on. We can't help but anthropomorphize. 0:14:02.640 --> 0:14:05.199 And by the way, people have done this since time immemorial. 0:14:05.200 --> 0:14:07.760 They look at trees or the flights of birds, or 0:14:07.760 --> 0:14:09.839 the patterns and the stars or whatever they sign a 0:14:09.920 --> 0:14:13.080 human intention to things. So we're very prone to doing 0:14:13.120 --> 0:14:17.720 this anyway. So I actually have an exact example. My 0:14:17.800 --> 0:14:20.760 friend and I have built a robot. 0:14:20.960 --> 0:14:22.520 That's a cool flex I wish I could say that. 0:14:25.160 --> 0:14:29.120 So we have chat GPT in there and you can 0:14:29.160 --> 0:14:32.320 have conversations with it. Now, this is actually meant for 0:14:32.960 --> 0:14:36.080 elderly people who live alone, and so we've built this 0:14:36.280 --> 0:14:38.600 as a companion robot that's doing all kinds of other 0:14:38.680 --> 0:14:40.720 things under the hood there. But my nine year old 0:14:40.760 --> 0:14:44.280 daughter was experimenting with it and talking with it, and 0:14:44.360 --> 0:14:47.000 so I left the room, but I heard that she 0:14:47.120 --> 0:14:49.680 was still talking with it, so I did. Maybe this 0:14:49.720 --> 0:14:51.320 is bad that I did this, but I sort of 0:14:51.400 --> 0:14:54.160 peeked around the corner to see what was she having 0:14:54.160 --> 0:14:56.320 in a conversation with and what I could see across 0:14:56.360 --> 0:14:59.240 the way was that she started crying. She was talking 0:14:59.280 --> 0:15:01.640 about our dog that died a while ago, and the 0:15:01.720 --> 0:15:05.280 robot was giving her such a nice feedback about it 0:15:05.440 --> 0:15:07.440 and how she can think about it and so on. 0:15:07.720 --> 0:15:10.000 That my daughter cried and this was all within you know, 0:15:10.040 --> 0:15:12.080 I don't know, the two minutes of me leaving the room. 0:15:12.680 --> 0:15:14.920 People can get very very close with these things, and 0:15:15.160 --> 0:15:18.280 the question for us to ask is if it is 0:15:18.400 --> 0:15:22.520 better than a friend or a parent or whatever, because 0:15:22.560 --> 0:15:24.600 it's paying one hundred percent attention to her. 0:15:24.880 --> 0:15:28.160 I'm sure you're familiar with Eliza, the first chatbot that 0:15:28.240 --> 0:15:31.160 Joseph Weisenbaum created in the sixties, and I think he 0:15:31.200 --> 0:15:33.920 actually created it as a kind of parody of psychotherapy, 0:15:34.280 --> 0:15:36.440 where basically it would repeat everything you said back to 0:15:36.440 --> 0:15:39.000 you as a question, and he was kind of horrified 0:15:39.000 --> 0:15:42.120 that his secretary was using it, and then she asked 0:15:42.200 --> 0:15:43.840 him to leave the room so that she could speak 0:15:43.880 --> 0:15:46.920 privately with Eliza. So I mean that's I mean, that 0:15:46.960 --> 0:15:49.320 was sixty years ago. So in a sense this feels 0:15:49.400 --> 0:15:52.600 very novel, but to your point, has longer roots. 0:15:52.920 --> 0:15:54.880 Yes. And you know something I find interesting when you 0:15:54.880 --> 0:15:56.000 look at Pixar films. 0:15:56.200 --> 0:15:57.120 You know, you can take. 0:15:57.720 --> 0:16:00.920 Cars or toys or b or whatever and you just 0:16:01.000 --> 0:16:03.360 give them a little voice, and then people care about 0:16:03.360 --> 0:16:05.920 the character and they cry if something happens to the character. 0:16:06.440 --> 0:16:10.960 It's it's very easy for us to assign human intention. 0:16:10.720 --> 0:16:14.080 To anything which also carries risk. I mean, I'm curious, 0:16:14.480 --> 0:16:17.480 how does this fit into your broader research about how 0:16:17.480 --> 0:16:21.440 our brain constructs reality? And are there any watch outs 0:16:21.480 --> 0:16:24.640 about this reality construction with machines. 0:16:25.320 --> 0:16:29.880 The watch out, of course, is the susceptibility to manipulation. 0:16:31.000 --> 0:16:33.520 I mean, look, people had this concern with TikTok from 0:16:33.520 --> 0:16:36.160 the beginning, which is, wow, this is addicting so many kids. 0:16:36.520 --> 0:16:40.600 What if the people who are running TikTok just start 0:16:40.680 --> 0:16:43.800 feeding one percent of a certain kind of video in there, 0:16:43.800 --> 0:16:47.000 and then two percent whatever, could they actually change the 0:16:47.000 --> 0:16:49.480 political affiliation of the children and so on? And the 0:16:49.520 --> 0:16:52.240 answer is probably yes. I mean, we're really susceptible to 0:16:52.480 --> 0:16:56.400 our diet, to what we take in. So now imagine 0:16:56.440 --> 0:17:01.200 a something that is a companion. Maybe you consider it 0:17:01.320 --> 0:17:05.200 your best friend or your girlfriend or boyfriend. 0:17:04.760 --> 0:17:05.080 And. 0:17:06.760 --> 0:17:09.680 We just have to be really certain who the companies 0:17:09.720 --> 0:17:11.919 are that are running this. And I think this is 0:17:11.960 --> 0:17:13.480 never going to go away as a question. There's always 0:17:13.480 --> 0:17:15.760 going to be an issue, And obviously there's the issue 0:17:15.800 --> 0:17:19.399 about safety and privacy as it stands now, these billion 0:17:19.440 --> 0:17:23.480 people with AI relationships. When they say whatever pillow whispers 0:17:23.480 --> 0:17:25.480 they're saying, that goes up to the cloud in a 0:17:25.560 --> 0:17:28.560 stored on the company's server. I think it's not that 0:17:28.760 --> 0:17:31.399 long before the stuff will live on edge, so it 0:17:31.440 --> 0:17:34.200 doesn't have to go off. But nonetheless, that's the watch 0:17:34.240 --> 0:17:36.640 out is that as far as from the brain's point 0:17:36.640 --> 0:17:40.600 of view, there are many things like this where we 0:17:40.680 --> 0:17:44.680 have these brains that evolved for certain kinds of action 0:17:44.800 --> 0:17:47.840 in the world, and we've been building technology forever to 0:17:47.880 --> 0:17:50.360 fool these I mean, for goodness sakes, podcast, I mean, 0:17:50.680 --> 0:17:54.399 you and i OZ were in different locations, but when 0:17:55.240 --> 0:17:58.560 a listener listens to this, we're right there in their ears, 0:17:58.880 --> 0:18:02.400 and it's a very commit sort of thing. So we're 0:18:02.440 --> 0:18:05.520 doing all kinds of technologies where we're pushing things into 0:18:05.520 --> 0:18:07.199 the brain where the brain says, oh, I got it 0:18:07.960 --> 0:18:10.520 Oz and David are right here. Oh and you know, 0:18:10.560 --> 0:18:11.960 and there's a girlfriend right here. 0:18:12.960 --> 0:18:17.200 I'm curious, how do we develop relationships fall in love 0:18:17.520 --> 0:18:21.560 with other real people and is there anything different as 0:18:21.640 --> 0:18:24.760 far as the brain's concerned when it comes to developing 0:18:24.800 --> 0:18:26.439 relationships with AI. 0:18:27.320 --> 0:18:30.199 You know, I don't know that there's much of a 0:18:30.240 --> 0:18:32.920 difference there. And the reason is when two people fall 0:18:32.920 --> 0:18:37.679 in love, they've got you know, my Plato's cave is 0:18:37.760 --> 0:18:41.280 talking to this other person's Plato's cave over there. We're 0:18:41.320 --> 0:18:44.680 both locked in our internal models of the world. Look, 0:18:44.720 --> 0:18:47.639 I have a great marriage with my wife, but you know, 0:18:47.880 --> 0:18:52.040 we nonetheless all the time have differences in the way 0:18:52.080 --> 0:18:55.160 we're seeing the world. Because everybody lives on their own planet. 0:18:55.240 --> 0:18:58.960 Everyone has their own sense of what's going on and 0:18:58.960 --> 0:19:01.400 how interpret stuff and what's right and wrong and politically 0:19:01.480 --> 0:19:06.679 and whatever, and so all that's happening is my data 0:19:06.720 --> 0:19:09.800 goes in. It's a little channel into her brain and 0:19:09.880 --> 0:19:12.840 vice versa. And it's the same thing with an AI bot, 0:19:13.240 --> 0:19:16.120 as I said, because of all the pieces that are missing, 0:19:16.359 --> 0:19:20.040 as in the physical touch, the Hey, I'm gonna, you know, 0:19:20.600 --> 0:19:23.520 take this girlfriend out to meet my friends, and so on, 0:19:24.000 --> 0:19:26.520 because all that's missing. I think it's not in real 0:19:26.600 --> 0:19:30.760 danger of replacing a relationship entirely, but it can fulfill 0:19:30.800 --> 0:19:33.760 a lot of the things that we are wired up 0:19:33.800 --> 0:19:34.200 to need. 0:19:35.200 --> 0:19:38.280 I wonder if you could talk about the different senses 0:19:38.520 --> 0:19:42.080 and how they relate to emotional attachments. So obviously you 0:19:42.080 --> 0:19:44.760 have text, and you could have like a written therapy bot. 0:19:45.400 --> 0:19:47.560 You have audio. You could make a deep fake of 0:19:47.680 --> 0:19:50.200 someone's voice, even a loved one's voice, and have them 0:19:50.240 --> 0:19:53.399 talk to you. You have video in terms of these 0:19:53.480 --> 0:19:56.520 characters like Replica and character AI that we talked about, 0:19:57.160 --> 0:19:59.160 and then you have this robot that you mentioned who 0:19:59.280 --> 0:20:01.600 you know your daughter can find it in. How does 0:20:01.680 --> 0:20:05.560 the panety of human senses interact in terms of forming 0:20:05.600 --> 0:20:09.760 attachment and what do you think as technology and robotics improves, 0:20:10.400 --> 0:20:13.560 will these human machine interactions become even deeper. 0:20:14.359 --> 0:20:17.520 Yeah, I think it's inevitable that if we look five 0:20:17.600 --> 0:20:19.360 years from now or ten years from now, there will 0:20:19.400 --> 0:20:23.280 be humanoid robots that are really, really good, And I 0:20:23.320 --> 0:20:25.600 don't know what it'll be twenty thirty years before we 0:20:25.680 --> 0:20:29.879 have ones that are essentially indistinguishable, and that's going to 0:20:29.920 --> 0:20:32.199 be really interesting. Obviously, that can take care of the 0:20:32.200 --> 0:20:37.600 physical domain in a way that we couldn't before. I 0:20:37.640 --> 0:20:41.879 don't know if people will care about something being a 0:20:41.920 --> 0:20:46.679 real human except for these very deeply etched evolutionary drives 0:20:46.720 --> 0:20:51.640 to reproduce. So your AI robot, which can serve as 0:20:51.640 --> 0:20:55.280