Latest / Psychology In Seattle Podcast / Can AI Cure Loneliness?
Transcript
- 0:00Hey, deserving listeners. So, for those who've been around in the podcast
- 0:03for a long time, you know that I have been thinking a lot about
- 0:06loneliness even before the pandemic. I knew that it was a problem around the
- 0:12globe and a lot of people are suffering from it, and a lot of
- 0:16the advice that is given online for curing loneliness is insufficient or trite in some
- 0:23way. And I set out to try to figure out if I could get
- 0:28my head around or hands around the problem. And the more I looked into
- 0:33it, the more elusive AT got. There's a lot of roads to loneliness,
- 0:37a lot of factors. I mean, we can look at the big
- 0:39things like people on their phones too much and that sort of thing, but
- 0:42that's too simplistic. I think that there's just this general trend that's been happening
- 0:48for decades that is resulting in a lot of loneliness, and not just suffering
- 0:53from loneliness, of course, but depression and demoralization and health problems and increase
- 0:59in addiction and suicide. These are all outgrowths of the increase in loneliness that's
- 1:06been demonstrated by the evidence. The other thing that I think about sometimes is
- 1:12artificial intelligence and AI. I remember the movie Her with Joaquin Phoenix not too
- 1:19long ago and seeing that in the theater and at the time thinking, well,
- 1:26yeah, I mean, we've known about that sort of possibility of an
- 1:30AI robot of some kind that can be a companion or can seem very human,
- 1:37and of course at some point in the distant future that'll be a reality,
- 1:42but probably not in our lifetime. And then all of a sudden,
- 1:46this AI explosion happens this last year. And of course AI has been in
- 1:52development for decades, but in terms of its commercial use has exploded in the
- 1:57past year. And as I'm used, and I use it a lot.
- 2:01The other day, I was just daydreaming about how horrible the scripts were for
- 2:07the Star Wars sequel movies, you know, episode seven, eight, and
- 2:12nine, and I was thinking, I wonder if AI could write a better
- 2:15script than Oh no, no, what I was thinking was was the Obi
- 2:17Wan Kenobi TV show on Disney Plus. I think about how horrible the script
- 2:23was for that, and I wondered if AI could write a better script,
- 2:25And so I, you know, typed in a bunch of prompts and to
- 2:30chat GBT three point five and it gave a much better script. And I
- 2:35thought, you know, of course there's this writer strike going on and they're
- 2:38protesting AI taking over their jobs, and of course I'm all for that labor
- 2:43movement and the strike. But there's another thing here. It's like AI can
- 2:47write a better at least outline of a script. It's it's interesting to me.
- 2:53And then as I'm writing the prompts, the first draft of this script
- 3:00is not exactly what I was looking for. So I just typed in without
- 3:04you know, typing in the whole prompt again. I just said, hey,
- 3:08change this one part in the last script, and then it just rewrote
- 3:12without complaining, and just rewrote the script and I kept I did about five
- 3:17to ten more prompts, and I each time I wanted to tell Chep GBT,
- 3:22I'm sorry, I don't want to bother you, but could you change
- 3:24this one part? And it felt like I was working with a with an
- 3:30essentient being who was writing the script with me. You know, we're sort
- 3:34of collaborating on this thing. And so when you think about loneliness and companionship
- 3:39and AI, I don't think we're that far away from that. I think
- 3:45we're very close. And then of course when you get the deep fake stuff
- 3:51going on, you get a three D image of someone that's moving and talking
- 3:55and emoting in the world. You're not just how you don't just have a
- 4:00typed you know, AI, but you have maybe spoken word, and you
- 4:03have the visual and the eye contact. And I just think that there's a
- 4:10chance that, given trends, that AI is going to become a part of
- 4:15our of course it's going to be part of our lives in the future,
- 4:18but I think it might be soon. And then you think, okay,
- 4:20well, if AI can solve some loneliness, is that good or bad?
- 4:26Because of course the cure for loneliness would be human contact, so that's we're
- 4:32capable of that. There's a lot of humans on the planet. So what
- 4:36if AI actually solves the problem, and what's the ethical and moral and human
- 4:44implications for that. It's just a bizarre thing to think about. Of course,
- 4:46I knew this was going to happen. I think it's happening very soon.
- 4:49I think in five to ten years this will be a reality. So
- 4:54I asked a special guest, an expert on this, Janetta George, to
- 4:59come on the podcast us to talk about it. Welcome to the podcast,
- 5:01Jane, Thanks Kirk, it's really great to be here, and yeah,
- 5:05I'm excited to get into this with you. And in particular, some of
- 5:11the things that you're saying now is really similar to a lot of things that
- 5:15I'm constantly hearing about thinking about how AI is going to be coming or is
- 5:19currently in the main screen and going to be But you know, really,
- 5:23in reality, it's been here since the nineteen fifties and so there's a lot
- 5:28of history to dig into to really understand how it's gotten to this point right
- 5:31now and how it can you know, help us in the future. But
- 5:36yeah, I'd love to tell you a little bit about how I got here
- 5:41in AI, and also how I started really studying loneliness and how I plays
- 5:46into it. Okay, tell me. So there are so many different stories
- 5:50about how people get into AI. I have been an AI engineer for about
- 5:54the past decade or so. I started off being a student of theoretical mathma
- 6:00loved studying puzzles and wanted to become a professor at some point. And then,
- 6:05as you can probably relate, there's a lot of years going into the
- 6:11studying and not a lot of reward at the end of it. And you
- 6:14know, I ended grad school not knowing what I was going to do.
- 6:17And that was right around I want to say, twenty eighteen, when data
- 6:21science and AI was really becoming a big thing, and I was like,
- 6:24Okay, I've taught for a few years, done the graduate studies thing.
- 6:29Let's let's try to actually work for a company. And it was a really
- 6:32great decision for me. I really loved what I was doing. I was
- 6:35an AI engineer at a water company, jumped around, started doing consulting,
- 6:41eventually started doing some executive advisory, working for companies like Apple, Amazon,
- 6:48you know, companies doing really cutting edge things with natural language things like that.
- 6:55And I'm still doing that today. So I still work as a consultant,
- 7:00but I also dabble in startup space and I started developing this app and
- 7:05it's still being developed. So this is by no means an advertisement for my
- 7:12app. But I started developing this app that I'm calling Candor really meant to
- 7:16tackle the loneliness epidemic. And the reason for it was I moved to Miami.
- 7:23I live in Miami right now. I've been here about two years and
- 7:26I've always kind of been in nomad, jumping from place to place. I
- 7:29was born in New Jersey, moved to Denver for school, then moved to
- 7:32Sweden for school, Philly, and then you know, so on so forth.
- 7:36I never really lived anywhere for more than a year, and just realized
- 7:41there has to be a better way to plug into people, to get to
- 7:46meet people that you wouldn't normally meet, or being new to town and wanting
- 7:48to meet people, and then that coupled with this was the middle of the
- 7:54pandemic. So I moved to Miami in twenty twenty one. Everyone was becoming
- 7:58very remote. No one wanted to go out. Actually, we're just getting
- 8:01back to the point where people were starting to socialize again, but didn't really
- 8:05know how to do it because we've been locked up for a year. So
- 8:09I wanted to develop an app that kind of took all of that great technology
- 8:13that Facebook and Google and other companies had already used in terms of like that
- 8:18social graph of how do we think about data and connecting people and using that
- 8:24to connect people locally and to get them going to events, but also meeting
- 8:28people and maybe meeting people in a more ethical way. Whereas we kind of
- 8:33live in our own, you know, ego chambers. We have biases,
- 8:37whether we know it or not when we're meeting and when we're thinking of who
- 8:41we want to meet. So how can we actually use data instead of enhancing
- 8:43biases to alleviate some of those biases and open our minds to people that we're
- 8:50connecting with. Really, so are you saying maybe using AI like it's tender
- 8:58for friends, but it's using AI to make it easier for people to look
- 9:03past their biases. Is that what you're saying? Yeah, exactly, And
- 9:07that's one component of it, right, because so just you know, at
- 9:11a basis, I know that there are people listening to this podcast that are
- 9:13lots of different entry points to technology and AI. So instead of going super
- 9:20into it at a high level, AI, the way it learns is from
- 9:24the data that you teach it. You know. I think with tot GBT
- 9:28coming to the main seam, we're getting to know more and more about the
- 9:31technology as just a general public. But one thing to know is that the
- 9:35way that AI works is you feed it a whole bunch of data and it
- 9:39ends up learning patterns from that data and then basically regurgitates it kind of like
- 9:43a child, you know. You teach it over and over again pictures of
- 9:46cats and pictures of dogs, and eventually it says, that's the cat and
- 9:50that's the dog. So AI is basically like a toddler, and then the
- 9:54more data that we feed to it, it becomes a more advanced toddler and
- 9:58hopefully eventually an adult. So yeah, So the concept of building in a
- 10:05more ethical way of meeting people, really it's just saying, hey, we're
- 10:09going to feed you data about Kirk's preferences of people. Maybe he prefers meeting
- 10:15people between the age of forty five and fifty five that live in the Seattle
- 10:18area that are into music and into psychology X y Z. That would be
- 10:24a little bit more on the bias side. We might want to try to
- 10:28remove some of those factors that we know about you while still incorporating things that
- 10:33we think would be a good match generally for people to try to match them
- 10:37together in a way that will still be beneficial to them, but in ways
- 10:39that they might not even know what sort of data would be entered on each
- 10:43individual. Yeah, and that's a good question, because this is where it
- 10:48starts. To be really important of you know, making sure that we're being
- 10:52ethical about the data, not only that we're collecting, but the data that
- 10:56we're using. So you know, the basic things that I just mentioned about,
- 11:01like interests, where you live, where you go to but then you
- 11:05know, we also are studying you know, psychology papers and just general research
- 11:11papers of human behavior and how different people interact with each other. And a
- 11:16lot of times these variables are actually late variables that we're not even necessarily able
- 11:20to point to. This person is going to like this person for this reason.
- 11:24It's latent variables, meaning that it's things that kind of, like I
- 11:28said, you can't necessarily point to them, but they create they create connections
- 11:33through the data from entering millions and billions of rows of data. So is
- 11:39it farmed data from Google and Facebook? Is that what you're talking about?
- 11:46Yeah, that's that's one way to do it, is the concept of web
- 11:48scraping or farming data, where you're essentially pulling things that are publicly accessible.
- 11:54But we really try to make sure that we're getting data. And I think
- 11:58that this is you know, general consensus, not just the way that I
- 12:01do things, but like good ethical practices we collect data that you know,
- 12:05customers and users want to give to you, So it's really important that we're
- 12:11asking the right questions of users of like, hey, what do you find
- 12:16interesting and then asking for feedback. One of the biggest things I always tell
- 12:18people is anytime you get like one of those surveys surveys at the end of
- 12:22like any kind of interaction of like, hey was this useful? Did we
- 12:24give you the right answer? Always answer them because that's literally how AI is
- 12:28getting better at doing what it's doing by you taking like those few seconds.
- 12:33So anyway, it's things like that that are helping that the AI system learn
- 12:39your preferences. Yeah, yeah, I mean it stands to reason that if
- 12:45someone consented in a similar way that people consent to Like there are times when
- 12:52an app or Google will ask me a question and say, like, do
- 12:54you want ads to be tailored to you, like on Instagram, for example,
- 13:00I'm pretty sure that's something that you opt in for. And at first
- 13:03I'm like, no, I don't want you to keep track of my patterns
- 13:07and send me ads, but I'll get ads that don't make any sense when
- 13:11that happens. So sometimes I enjoy getting ads that are tailored to me because
- 13:16they're products that I actually click on, and I've bought things through Instagram for
- 13:20better for worse because of that. Yeah, exactly. And so if that's
- 13:26a consensual, above board sort of disclosure, then people would sign up for
- 13:33that. Some people wouldn't because of that reason, but you could see how
- 13:37they might say, sure, feel free to dig through my meta and Google
- 13:43database to figure out who I might hook up with. A So, is
- 13:50this for friends or for dating or both? Or what? Yeah? So
- 13:54I mean the candor that that I'm developing in particular is it's for friendships.
- 14:00It's like an AI social assistant how to elevate your social life. They do
- 14:07things like this for dating too, like you probably aren't familiar with, like
- 14:09Bumble and then bumble bff. How they're kind of utilizing this your same data.
- 14:15So if you sign up for say bumble BFF because you want to meet
- 14:18friends, and you know, the AI model is collecting all this information that
- 14:24can then be transferred into bumble. If you then decide I want to date
- 14:26someone, well let's take a break. Can we get back, I'll ask
- 14:31a question, what do you say? Sounds good? So you sent me
- 14:37a list of topics that we could talk about, and one of them is
- 14:41now coming to mind in that because of course we know the algorithm is this
- 14:46good and bad thing, and it can create echo chambers and can create exacerbation
- 14:52of radicalized ideas, and if someone is in that zone, they might be
- 15:03funneled in a way towards other people that are radicalized or you know, whether
- 15:07it's a political point of view or some kind of problematic thing for the individual
- 15:13and for society, and that they might be funneled to date or become BFFs
- 15:18with someone similar to that. Is that making sense? So my question is
- 15:26is that a consideration? It absolutely is, And I think that you know,
- 15:30you might have even mentioned this on one of your podcasts before the famous
- 15:35racist Twitter bot in twenty sixteen. But the Twitter bot that you know was
- 15:39meant as an experiment and spent maybe less than twenty four hours collecting data that
- 15:48it was reading on Twitter and started eventually responding like a racist Nazi by the
- 15:52end of the day, and so like really understanding why that happened, And
- 15:58then that kind of answers this question that you just ask again, It comes
- 16:02down to the data. AI cannot do anything on its own. You you
- 16:06feed it data and it then learns this is a good thing to do.
- 16:07This is a cat, this is a dog, This is a racist Nazi,
- 16:11which apparently you want me to be because you're you're feeding me all of
- 16:14this data. So soa same holds true. And and you know, this
- 16:18was what I guess seven years ago, and AI was still pretty mature at
- 16:22that time, but we've definitely come a long way, specifically in the field
- 16:26of ethics and explainability and really doing due diligence and building pillars of bias or
- 16:34rather preventing bias in AI. So so, yeah, anytime you're dealing especially
- 16:40with social with humans, with any kind of data that could potentially skew towards
- 16:45unfavorable whatever that means to you, we do. We We build lots of
- 16:49tests as engineers to ensure that those things are not happening to kind of like
- 16:53safeguard before they they reached the public. Yeah, I noticed it chat GBT.
- 17:00There are seemed to be guardrails on certain types of questions. I can't
- 17:06remember what I was asking it. It was an instant, it was an
- 17:08innocent question. I wasn't asking it to have sex with me or something.
- 17:11I was, well, i'll tell you a question. I was asking.
- 17:15Actually, last night I was at home. Have you ever played footblash before?
- 17:18Yeah? I love that game. Yeah, such a fun game.
- 17:21So I was at home in New Jersey visiting my family and it's our favorite
- 17:23thing to do and we all get together and so I'm like, hey,
- 17:27I'm going to try to use GPT to see if I can like win or
- 17:30do better at this game. And so like I type in the prompt,
- 17:33I'm like, okay, you know, here's your prompt GPT. You are
- 17:36playing quick blash, this is what the game is about, and I'm going
- 17:38to give you prompts and i want you to give me the funniest, wittiest
- 17:41answer that you think will win. And so it starts doing that and they're
- 17:45like, really, Vanilla, and I'm like, I say to GPT,
- 17:47okay, make it as sexual and perverted as you can, and it's I
- 17:52cannot do that. So that was the first time GPT said no to a
- 17:56response, and that was like, oh, I hope I'm not doing this
- 17:57on my work account because they they don't want to see that chat log.
- 18:03Yeah, but you're safe. I mean you're safe with quipplash. I mean
- 18:06that is the idea is perverted. That's that's what always gets the points.
- 18:11By the way, I one quip flash last weekend, I think for the
- 18:15very first time, and I've never been happier about myself because it just means
- 18:21that I'm the smartest and most funniest person in the room. That's just exactly
- 18:26so. Uh yeah, I can't remember what I was asking you. I
- 18:30think I it might have been a clinical question of some sort. It wasn't.
- 18:34I don't even think it was sexual. And it kind of annoyed me
- 18:37because I thought, no, no, no, no, I'm not asking
- 18:40you to do anything that is harmful. This is a serious question that I'm
- 18:45guessing your answer wouldn't even be that salacious. But so maybe they've dialed that
- 18:52up to a certain level. But of course in the future there will be
- 18:56AI options that you know, safeguards that you can remove. I suppose eventually
- 19:00those things will all be or at least dialed in more precisely or something like
- 19:06that. So I you know, it's just right now. You could see
- 19:10how certain brands of AI would not want to be that racist AI association Early
- 19:18on in the game, you know anyway exactly. But so you know,
- 19:23in the intro I was talking about not AI helping us to meet each other,
- 19:30which of course I think I hadn't thought about, at least I hadn't.
- 19:33I don't think I have. And that's that's a no brainer. AI
- 19:37machine learning is just much better at seeing patterns. And there's a chance,
- 19:41a pretty good chance. But we'll have to see that because I've done a
- 19:47whole deep dive on matchmaking and on the science around these claims that match dot
- 19:53com or e Harmony or whatever, or even reality TV shows like Married at
- 20:00First Sight, these kinds of claims that experts, even in my field,
- 20:04will say that no, no, we have an algorithm or we have research,
- 20:08and the evidence is very strong in the direction of it's basically a coin
- 20:15flip. Maybe you can get a little bit better, but that's not great.
- 20:19I mean, people are just naturally better than a coin flip coin flip
- 20:26because when they just see someone across the room, they're generally better than a
- 20:30coin flip as to whether or not things might work out with that person.
- 20:33As to whether or not, no way that you know things won't work out
- 20:37well, A question for you, though, I'm curious because I've heard you
- 20:41say that many times, and I've thought about when you say that, I
- 20:45wonder if you know, with the advent of AI being used for these kinds
- 20:51of things and the massive amount of data and human behavior that we're feeding into
- 20:56it, do you feel like we have a better shot of being able to
- 20:59do it using potentially? Yeah, I'm not only just that information, which
- 21:03of course is way deeper, not only in terms of data, but in
- 21:08terms of analysis, because you know, when we're doing the analysis or the
- 21:14data gathering where we're surveying people, which you know, you can't survey them
- 21:18for months. You can only ask them a number of questions or you know,
- 21:22at most maybe like five hours of them clicking yes no, or like
- 21:27scale kind of stuff. And then we're looking at associations with correlations, and
- 21:33we're looking at couples that actually last. So that is far better than just
- 21:40randomly pointing at someone from across the room, but not much better because the
- 21:44magic of attraction and staying together is way more complicated than that. But with
- 21:51AI, you are feeding in a lot more data potentially, and real data,
- 21:56not just survey because surveys people answer the way they want to answer.
- 21:59They don't answer necessarily from a place of honesty or self awareness and the pattern
- 22:04recognition. You know, AI can figure out things that is bizarre. They'll
- 22:08they'll figure out, well, actually, you know, in the black box
- 22:11of the AI, they make an association that is like, well, they
- 22:15both like muffins on Yelp, and we have determined by our machine learning that
- 22:21that's, you know, has a certain weight in terms of whether or not
- 22:25these two people are compatible, not just on the muffin front, but on
- 22:29a There's certain a lot of things that are associated with liking muffins, having
- 22:33the Yelp app and publicly stating you like muffins, you know, like that
- 22:37wouldn't be something that you would ask on a survey. It wouldn't occur to
- 22:41a psychologist even include that in the research. The other thing, though,
- 22:45that AI can do is look at pictures of people and maybe full body scans,
- 22:51which is not usually included in these in this research because it's way more
- 22:56subjective one and a little taboo for researchers to to sit back and say,
- 23:00well, I'm gonna rate that person, you know, a five and on
- 23:04the body scale and a three on the face scale or you know, we
- 23:11tend to avoid that and try to look bored towards personality, and people tend
- 23:15to talk that way too. People you know, you say what sort of
- 23:18partner you're looking for, You're like, Wow, I want someone that's family
- 23:21oriented. But the way people behave it's way there's a lot of visual and
- 23:26a lot of vibe, and a lot of smells and a lot of you
- 23:29know, just groinal attraction that happens that is not easily measured. But with
- 23:37AI, you just got to figure there's gonna be a signal in the data
- 23:41there right when you you just gather like thousands of couples that you know,
- 23:45have been together for five to ten years, and then you just scan them
- 23:49all physically or maybe just look at their Instagram posts and the AI can figure
- 23:56out like symmetry of eyes and tone of skin and how big the smile is
- 24:02and maybe the tone of voice in some of their video, you know,
- 24:06posts on Instagram, and you got to figure that the machine learning is going
- 24:11to see something in that. Yeah, And so, I mean I've been
- 24:14on the dating apps for embarrassingly long. I've actually been on dating apps since
- 24:18I was nineteen and I'm thirty three and apps, so to the math of
- 24:22how long that is. But well, you must you must have some stories
- 24:26to tell. Oh gosh, I have some stories to tell. But I
- 24:30think the point is though, that people that chronically are, you know,
- 24:34using dating apps, we have kind of turned into machines a little bit ourselves
- 24:40in the way that we processed the data that we're seeing. So just to
- 24:44like give give an example, you know, maybe years and years ago,
- 24:47when I was on like a Tinder or a Hinge and looking at someone's picture,
- 24:49you know, you you don't really think too much of it. You
- 24:52just say, that's an attractive person. I find you attractive. I'm reading
- 24:56the things that you're writing. They check the boxes and then we'll go out.
- 25:00But you do that one hundred, you know, so many times,
- 25:03and eventually you start to read through the cracks and see, Okay, you
- 25:06know this person is smiling this way, or they're like these they're always posting
- 25:11selfies or things like little things that you don't pick up at first that you
- 25:15start to pick up and you start to be able to be able to tell
- 25:18and maybe this is just me becoming you know, jaded to online dating of
- 25:23like, oh, I don't want to go out with anyone anymore, but
- 25:26you start to really weed through things much quicker. And I'm sure that you
- 25:29know the machines are also picking up on that and can probably do it way
- 25:33better for us. And that's why you know on apps like Hinge, they
- 25:37do have really great algorithms, and the more you pay, the better algorithms
- 25:40they give to you. Of like saying hey, like you're definitely gonna like
- 25:42this person, and it works pretty good. So give us one thing that
- 25:48you learned about pictures on Tinder that you say, Nope, that you wouldn't
- 25:53have realized when you first started looking at it. Pictures with motorcycles. Uh,
- 26:00No, I ride motorcycles. I ride vespas living here in Miami,
- 26:04So if you're posting a picture on a vespa, that's a thumbs up for
- 26:07me. No. I do think it's the selfie, and I think that
- 26:11applies for men and women, like having like one selfie totally acceptable. But
- 26:15there are some people on there that it's all selfies. Uh, what's it?
- 26:18What does that indicate to ye? The learning that you have made over
- 26:23time? Good question? So I think that it indicates a little bit of
- 26:30a lack of discernment or like introspection self awareness a little bit, because you
- 26:38know, people want to see you like out doing different things. They want
- 26:42to see variety, they want to see just you and the wild getting like
- 26:48your personality. But when you're just posting selfies, you're what you're getting from
- 26:51them, what you're learning is that they just stay in their house looking at
- 26:53the mirror, taking selfies. And that's not a desirable trait necessarily. Yeah,
- 27:00I mean one selfie is enough. Beyond that, it's just more selfies,
- 27:04right, So yeah, what's the point in that? Yeah? Interesting,
- 27:08let's take another break. Well, get back more questions. Okay,
- 27:15we're back from the break. But what about what I was saying in the
- 27:19intro about how like the movie Her? Did you see that movie Joaquin Phoenix?
- 27:23I didn't, So you're gonna have to tell me a little bit about
- 27:26it. Oh. Well, in some future of our society, there's this
- 27:32fellow named Jaquin Phoenix or played by Joaquin Phoenix, and he he just went
- 27:37through a breakup, I think, and there's this new operating system and he
- 27:45opts in to get it and he feeds it some information and it slowly starts
- 27:51to grow on him. At first, he's just kind of like, Ah,
- 27:53it's just this AI that you talk to and it seems kind of human,
- 27:59but it's not really. But over time he starts to really get into
- 28:03a deep relationship. And the way that the movie is written and directed and
- 28:07acted, at least for me and I think a lot of the viewers,
- 28:11it doesn't feel creepy, and it isn't creepy in the end. It's not
- 28:15like and then she kills him or any you know, it's it's played.
- 28:18There is an interesting ending that I won't spoil. But it seems real.
- 28:23You know, it seems real. But a lot of people are seemingly falling
- 28:29in love with their AI, and you can see why. I mean,
- 28:33they even have like phone sex with the AI, and the AI has what
- 28:40the AI believes to be an orgasm during that. Right, I don't know
- 28:45it. It's a powerful movie, and I just imagine in the not so
- 28:48near future that that will be a thing. So what are your thoughts on
- 28:53that it already is a little bit of a thing. And I'll start by
- 28:57just saying a little bit of history. So first of all, there are
- 29:03three types of AI. So there is the narrow or sometimes AI engineers would
- 29:08call it the stupid or the dumb AI. There's a general AI, and
- 29:12then there is the super AI. So narrow AI basically is just like doing
- 29:18like a single task. So you know, you might have like a recommendation
- 29:22system that's recommending products to you. You might have like a few of them
- 29:26on like a website like Amazon. General AI then is a conglomerate of lots
- 29:32of different algorithms that are basically coming together and trying to do something that a
- 29:37human would be able to do. So you think back to I don't know,
- 29:41when I was a kid, there was like this movie on Disney Channel,
- 29:45so they like Smart House. So I kind of think of that like
- 29:47a robot that can do anything that a human can do. Super AI,
- 29:52though, is being able to do things even better than a human. So
- 29:59back in I want to say, nineteen fifty Alan Turing, this is kind
- 30:03of like the birth of AI, aside from like science fiction that that happened,
- 30:07like, you know, a few decades leading up to that. But
- 30:10then nineteen fifty Alan Turing wrote a paper Ken Machines Think, and he was
- 30:15also the inventor of the Turing test. Are you familiar with the Turing test?
- 30:19Huh absolutely, yeah, yeah. So for listeners who might not know,
- 30:23it's a really simple concept. It's basically, you have two different judges
- 30:30or sorry, you have a bunch of judges in a room and they are
- 30:33trying to see are you a human or are you a robot or a computer?
- 30:37Really simple question, and so the person being judged or the robot being
- 30:42judged is asked a bunch of questions and then at the end, the panel
- 30:45of judges says you are human or you are a robot. And it was
- 30:49thought for a really long time that this would, you know, we would
- 30:53never be able to have a machine that was able to pass the Turing test.
- 31:00We went through a long AI winter throughout the seventies and eighties where there
- 31:04was very little research being done in AI, just because we we would kind
- 31:08of like had this high hopes coming off of you know, the sci fi
- 31:14like forties fifties, of thinking AI is going to be doing all these amazing
- 31:17things in the next ten years, and then it didn't happen. But really
- 31:21really cool actually, I think a few months ago GPT broke the tur It
- 31:26beat the Turing test, which is a really huge accomplishment in AI. So
- 31:30I think just that alone kind of goes to that question of like are we
- 31:36there are? Will we be there soon? And this concept of general AI
- 31:40that you know, general conglomerate of AI that is able to do what humans
- 31:45do just as well as humans. We're already kind of doing it right now.
- 31:48Researchers say that it's probably going to be at least another decade until we're
- 31:52really seeing it widespread in our society, and then a few decades more until
- 31:56we have super AI. Just for those of us that are like really scared
- 32:00about it, like dooming gloom from AI. We are still several decades away
- 32:05from AI really truly being able to outsmart us, but in terms of human
- 32:10interaction and being able to be fooled by an AI system. And I've actually
- 32:15seen documentaries of people talking about this that they're falling in love with. You
- 32:22know, they're not not necessarily like a real robot, but an app that
- 32:27they talk to, kind of like you know when you watch like Catfish or
- 32:31or like Night to Day Fiance or whatever, where people are just chatting at
- 32:37a screen and you don't really know if if it's going to end up being
- 32:39the love of your life or not. It's it's the same thing with AI,
- 32:42but you know it's not going to be the love of your life.
- 32:45So in some ways, you're actually putting yourself in, you know, in
- 32:47less harm that there's no harm from it other than just knowing that you can
- 32:52never physically touch it. Yeah. I just saw this report of AI women
- 32:59that are scantily clad in these video clips and they're talking and emoting and you
- 33:07can kind of tell by the face that they're AI. But you just got
- 33:13to figure in a couple of years or if more expensive you know, products
- 33:19were to you know, crunch the numbers to make the face look even more
- 33:22realistic, that this will be just around the corner that someone can just type
- 33:27in a prompt and just like, okay, I want a woman that looks
- 33:30like this with this kind of hair and DDD here the dimensions here right,
- 33:34here's the sort of personality style. And then you have that that human look
- 33:42alike catfish as many people on Instagram that she can right, And like with
- 33:49other scams along these lines, you only need point zero zero one percent of
- 33:52people to respond with their bank card and you're off and running, right,
- 33:55So are we going to see like ninety day fiance AI at some point?
- 34:00I mean, it seems like in five years that that will be a thing.
- 34:04Right. We've seen others that have been seemingly scammed by and what appears
- 34:09I don't know if you watched back with David and all that stuff, and
- 34:15I only barely know his story, but it seems to be that there's certain
- 34:20people or even companies that employ women to or someone of any gender, but
- 34:27they would act like they were a woman, and they will catfish men in
- 34:31the United States or other countries that have men that have some money, and
- 34:36they charge by the minute to chat, and they lead them on and say
- 34:40I'm in love with you, and so the man gets obsessed and is spending
- 34:45all this money and deluding himself, and in the end he's in a cafe
- 34:52in Russia waiting for her to show up, and she never shows and then
- 34:57she says that I missed my bus. Oh sorry, honey, you know,
- 35:01please forgive me or whatever. You just imagine that an AI would be
- 35:07able to do that infinitely better and with actual video conference and at almost no
- 35:15cost to the scammer. Yeah, and forget about just sex and romance,
- 35:22but scamming old people. You could do a deep fake of someone's child,
- 35:29right that, there's already that happening where they will deep fake someone's voice and
- 35:34call someone's parents that are seventies eighties years old, and the deep fake of
- 35:40this adult child is saying like, oh my god, I'm in the I'm
- 35:44in the hospital. Send me money to this number. And then it's a
- 35:49powerful tool of scamming people. But anyway, so are there any AI developers
- 35:54working on human emotion and for example, falling in love or affection for the
- 36:01user? Because it's general AI in terms of write me a better script for
- 36:07obi wan Kenobi, it's you know that's that's working great, but the person
- 36:12or the person, but the AI version that's writing the script isn't just like,
- 36:16hey, Kirk, you know, did you read my script? How'd
- 36:19you like it? Like? Did I do a good job? It doesn't
- 36:22give us shit. It's just pumping out this you know this thing. Are
- 36:25people working on emotions? Yeah, well, first, really quick to go
- 36:30back to what you're saying and then remind me if I forget to come back
- 36:34and answer this question. But so GPT they just did some studies on it
- 36:37about like this email spam phishing like content, just showing how insanely fast we're
- 36:45able to create phishing emails that are just as accurate or sorry, just as
- 36:52effective, good effective, thank you, if not more than hands written one.
- 36:58So there's that coupled with if you want to start talking about deep bakes
- 37:02and you know, actually turning voice into things. There are apps now that
- 37:07you know, back in the day, it used to be that you had
- 37:09to collect hours of footage of you know, President Trump talking to be able
- 37:15to like mimic his voice. Now you literally need two seconds to be able
- 37:20to make something extremely believable. I just saw someone demoing it the other day
- 37:23at a conference and that's so that's really wild. And I don't know,
- 37:28do you ever watch scammer payback on YouTube or any of those like YouTube videos.
- 37:34Yeah, yeah, where they a call center from India will be trying
- 37:37to scam an old lady in the United States and these hackers essentially will hack
- 37:44into their system and record the whole thing. And yeah, I've watched like
- 37:49an anti hacker for good. I would love to see them develop a tool
- 37:53that is able to like take old ladies voices and just go on auto and
- 37:59you know, waste spammer's time for you know, hours and hours that that
- 38:02would be really enjoyable to watch. I would subscribe to that channel. But
- 38:07yeah, So back to your question though about are there currently AI researchers working
- 38:12on emotions? And yes, absolutely, I think that it's it's not an
- 38:17easy question to answer, like are we are we working on, you know,
- 38:22solving a particular algorithm or building a particular model, just because, as
- 38:27you know as a psychologist, emotion is not an algorithm. Has simply put
- 38:30like, it's not something that we can, you know, historically say like
- 38:35X plus y equals z and and now let's just make a model that produces
- 38:39that. So the the field of research that I've spent a lot of time
- 38:45working in is something called natural language processing, which is very it's it's one
- 38:50of the big parts of this branch of study of AI or you know,
- 38:57artificial emotions. And so what natural langould processing does exactly as it sounds of
- 39:01processes language. It will take tons of different documents whatever, turn it into
- 39:08machine binary interpretable form, and then be able to extract something from it.
- 39:14So one thing that we do is something called topic or sentiment analysis. So
- 39:21we've been doing that for you know, decades and decades where you know,
- 39:24you you enter in, uh, maybe it's a document that you're reading and
- 39:30say, hey, tell me what the mood of this is. Tell me
- 39:34like the vibe basically give me a vibe check, and and we can train
- 39:37we can train models to do that. There are also apps that are being
- 39:42developed right now that are particularly for mental health. I remember hearing someone talking
- 39:47about an app that they're building for corporate mental health, which there are ethics
- 39:52around this that I would love to understand, like how they're battling this just
- 39:55in terms of what you're feeding back to corporate you know, to your boss.
- 40:00Basically that might be telling them information that you don't want your boss to
- 40:02know. But it was basically like a welfare check of like hey, like,
- 40:07you know, tell me different things about yourself. How are you feeling
- 40:09at this time? And it would kind of be able to like track your
- 40:13emotions, so like you know, being able to understand human emotions, But
- 40:17on the flip of the coin, like how are we able to generate emotions
- 40:22that are believable that I think that you know, the research that we're doing
- 40:27with GPT, with the concept of you know, reinforcement learning, which you
- 40:34know, I can talk a little bit about how that works, and just
- 40:37different sophisticated types of AI that are letting the machine learn and react. Really
- 40:43is the thing like being able to react to what you just learned in a
- 40:46positive or negative way. That's really the key to to those kinds of algorithms.
- 40:52Yeah, it seems that if you just went off and I don't know
- 40:58much about this technology or the research, but it seems like if you're trying
- 41:01to create emotion in the current at least mainstream AI that I've experienced, then
- 41:07you just get a bland emotional personality type that was an average of everything that's
- 41:14being fed into it. And of course you can't just feed in documents like
- 41:19Wikipedia and the sort of thing. You have to feed in a lot of
- 41:22nonverbals, so you'd have to feed in video, I would imagine, and
- 41:27that's being done more. But that's harder data to quantify, right, But
- 41:31of course machine learning will figure out stuff eventually, because most of the written
- 41:37documents and the things on the Internet don't explicitly communicate or even between the lines
- 41:45communicate any sort of emotion. So there's that, But when we interact with
- 41:50human beings, there is a personality that we're dealing with, a consistency across
- 41:54different contexts of emotional tendencies and reactive And if it's just a general, bland
- 42:02human being that is reacting on average of all human beings that are fed into
- 42:07it, I would imagine that although that maybe would be a certain personality type
- 42:13that would be interesting, but it would just be one human or one personality
- 42:19that everyone interacts with. And of course, for this to be like the
- 42:23way that it will be in the future, I imagine people would want their
- 42:28own sort of AI that they're interacting with solely. Right, if they had
- 42:34a friend that was an AI, or a partner that's an AI, or
- 42:37even just an assistant, a butler, if you will, you would I
- 42:42think there would be a preference to have variability or some specificity to that robot,
- 42:49right, And for that you'd either have to tinker from the outside or
- 42:55ask the AI to create its own personality or relegate or limit the AI to
- 43:04focus on a certain personality type in terms of its data set that it's basing
- 43:09the emotion, or just let the machine learning develop on its own, you
- 43:14know, because that's our thing. You can, as far as I know,
- 43:15you can isolate a machine learning system and say, okay, learn and
- 43:22develop and figure things out independent of other machine learning systems, right, And
- 43:28so a certain emotional system, if it's set upright, could could maybe develop
- 43:34that in that way. I don't know if I'm talking out of my ass,
- 43:37Am, I talking out of my ass No, you're you're not.
- 43:38And what you're talking about is actually kind of the heart of reinforcement learning,
- 43:43but with the caveat of you still need a really really good data set to
- 43:47start with. And it's actually funny as you were talking, I was thinking,
- 43:52and feel free to cut this out if you don't want anyone to know
- 43:53the secret sauce that maybe could make them millions of dollars by using your content.
- 43:59If you took an AI bot and ran it through your I don't even
- 44:02know how many, like tens of hundreds of hours of content you have reacting
- 44:07to people going back and forth, reacting with each other, and then you
- 44:12are basically providing the label the ground truth of okay what they did was good
- 44:17or okay, what they did was bad. And here is how I would
- 44:21do it like, that's literally machine learning gold right there in spoken form.
- 44:25So, man, that would be a really cool if someone could just like
- 44:29give us a million dollars, We'll go off and follow this up and see
- 44:32if we can change the world with that. Yeah, no, I've thought
- 44:36about this. I don't know if I've talked about it, but you know,
- 44:39it's it's just a matter of time, whether it's in my lifetime or
- 44:43after. But I imagine there'll be something approximating this in my lifetime. In
- 44:47that Yeah, you just have an AI look at maybe even every single video
- 44:54and audio podcast I've ever made, and compile a personality out of that,
- 45:00which would be I'm guessing pretty close at least to my public personality. It's
- 45:06not the entirety of my personality, but even the machine learning might even be
- 45:09able to extrapolate what I'm like in my personal life. I don't know,
- 45:13and I mean the way I talk about my personal life. And then a
- 45:17user could have a very long, in depth, ongoing conversation with me,
- 45:23and that AI could learn about the user and retain that as memories and say,
- 45:30hey, remember that thing you told me, like how you and it
- 45:32would be completely indistinguishable from having your own personal doctor Kirk as as your buddy.
- 45:40And of course in the near future there will be laws against using one's
- 45:46image or one's IP, but you can't stop people from having their own personal
- 45:52AI on their own server, on their own computer, disconnected from the Internet,
- 45:58and they just screw rub my entire channel and create their own version of
- 46:01me. That that will happen. That, I mean, technically could happen
- 46:05right away, but it's not the it's not typical that when these technologies are
- 46:09developed that random users will have access to that or all be very rare.
- 46:14But you know, not so different distant future, it'll just become It'll just
- 46:19be a thing you can download as an app on your on your phone,
- 46:22and uh and you can train it to do all sorts of stuff. And
- 46:28and I think it's just gonna happen. And I'm fine with it, I
- 46:31guess, because what else can I be? But I, uh, well,
- 46:37how do I feel about it? Well, so, on the in
- 46:39the interim, I have thought about, once this becomes available, I would
- 46:44consider actually having an official AI version of me that I could make sure was
- 46:52ethical and on the up and up and had guardrails like you can't have sex
- 46:55with me or something, and it would be on my web site, you
- 47:00know, And because that that'll be something that of course you would pay for,
- 47:05I would pay for and maybe a subscription service or something. And it
- 47:08sounds ridiculous, and maybe once it comes around, I wouldn't do it because
- 47:13it would be ridiculous, but that this is something that will happen and all
- 47:16sorts of parasocial relationships will be facilitated this way, you know, Like I
- 47:22want to enter, I want a jam with Paul McCartney. I want to
- 47:25jam with the AI version of Paul McCartney or Prince or something. And if
- 47:29and if the estate of Prince makes that available, then fuck yeah, I'm
- 47:34gonna I'm gonna do that. Are you kidding me? To have Prince like
- 47:37laugh at my guitar playing. I mean that, that's just the best thing
- 47:39I could ever think of. So in the same way, you know,
- 47:42Yeah, this is a couple of years ago, I would of thought I'll
- 47:45be long dead by the time this happens, but now it looks like it
- 47:50might be happening pretty quickly. And by the way, it might be happening
- 47:52pretty quickly. Yeah, and by the way one of the one of the
- 47:55listeners to the podcast reached out to me and works in AI and created you
- 48:00know, something that's three percent in this direction where it scours all the content
- 48:05on YouTube and spits out certain information. They don't want me to share it
- 48:10with anyone because it's on their own personal server. I use it occasionally just
- 48:14as a curiosity and as a tool sometimes, but you know, it's already
- 48:19it's already happening. Who would you want to parasocial relationship interact with? Oh?
- 48:24Wow, good question, who would I want to parasocial? Well,
- 48:27actually, so instead of just answering that what I There are so many things
- 48:30I wanted to react to what you just said. But one of the things
- 48:34is imagine being able to have that app or you know, many different flavors
- 48:38of that, but let's just go with the therapy for now, like having
- 48:42an app that you can log into and instead of like I won't name names
- 48:46for I don't know who's sponsoring your podcast, but like a therapy app and
- 48:50they have that as an option. But imagine if it was like in the
- 48:53voice of your mom, or in the voice of a lost loved one,
- 48:57or you know, like in a voice of someone like truly meaningful to you.
- 49:01And again, all you need is a few seconds of their recorded voice,
- 49:05and then you know all of your knowledge is then being pumped into that.
- 49:08Boy, it's that I think that's that's a big bunny idea right there
- 49:14too. Wow. I had thought about this. But a lot of therapy,
- 49:19this style that I do is reparative work, is corrective emotional experiences.
- 49:24And what I am doing as a therapist for my clients is a lot of
- 49:29the times I'm giving them the parenting emotional vibe that they deserved when they were
- 49:35growing up. So when they're telling me about how proud they am of themselves,
- 49:39I'm also proud of them when they cry, I'm there for them.
- 49:45And when they were growing up it was the opposite. And through that,
- 49:49it's well understood through research and decades of observation and practice, that that can
- 49:57help a lot of the ills that people suffer from today, depression, anxiety,
- 50:01lack of meaning in their life, bad relationships, you know, addiction,
- 50:07this kind of thing. But how more powerful would it be to have
- 50:09an AI of their parent that we designed together, collaboratively, me and the
- 50:14client and with consent and care and we pull up on the giant TV screen
- 50:22in the therapy office and I facility. You know, it's very similar to
- 50:30what I'll do empty chair work, where I'll have the client imagine that their
- 50:32mom is sitting across from them. It's not me, but you know,
- 50:36we have an empty chair that they're looking at and those and even just without
- 50:39the person there, it can be one of the most powerful emotional experiences.
- 50:44It is transformative for people that I've done this kind of work and that I
- 50:47didn't invent it. It's a it's an old therapy technique, but imagine if
- 50:52we in the empty chair had a computer screen with actually their mom, and
- 50:59would we dialed in a couple emotional things and the person was able to you
- 51:04know, in the course the client knows, not their mom, but it
- 51:07would really elevate that that emotional landscape. It's a million dollar idea. I'll
- 51:13set up a call with you next week to start prototyping. Yeah. No,
- 51:17there's really amazing things that AI can do that I don't even think that
- 51:22we've started scratching the surface of. So it's an exciting time to be working
- 51:27in AI and working in any kind of scientific field that is collecting data and
- 51:32you know, really being you know, and I probably already said this ten
- 51:36times in this past hour, and I'll keep on saying it. But the
- 51:38way that we're collecting the data, not just how not just the data we're
- 51:42collecting, but the way that we are cultivating it and making sure that it's
- 51:45you know, clean and ethical and unbiased, that is truly going to be
- 51:50what shapes the future of AI. And you don't have to answer this question,
- 51:52but if there was a product where you could have a friend or a
- 52:00mentor, I don't know, someone that checked in on you whenever you wanted
- 52:05them to check in on you, it was but it was an AI and
- 52:07you knew it was an AI, and it had a perfect, warm,
- 52:14caring, inquisitive voice, and you could pay three dollars a month for that
- 52:22service, would you do it? I think absolutely? And I think you
- 52:27know, I've been going to therapy for the better part of my twenties and
- 52:30now into my thirties, and you know, a big part of that was
- 52:35the reparenting process and just you know, feeling feeling, you know, like
- 52:40someone cares and like someone you know wants to listen. And again, when
- 52:45you're in that position. Like I mentioned at the beginning of the episode of
- 52:49you Know moving around, you know You're after year, you don't always have
- 52:52that person, whether it's in person or that you've developed in the past,
- 52:57that is able to be that warm love voice. So and even now that
- 53:01I do have those people in my life, I still think that there's so
- 53:07much that even the best of friends can't and nor should we necessarily expect them
- 53:12to be able to give that Just like m you can probably say it in
- 53:17better words than me, but just like unconditional love, we'll put it that
- 53:22way, like unconditional like yourself is totally removed from the equation. Let me
- 53:25just like pour into you and and sometimes you really need that. Yeah,
- 53:30yeah, I if it's convincing enough, then I don't know how I feel
- 53:36about it though, because I would know, and I don't know if that
- 53:42would matter to me. I think if the AI got good enough, it
- 53:46could be convincing to a lot, because that is like really the key there,
- 53:50like is it going to be something that's so robotic kind of like how
- 53:53the like AI apps when you go to a website, they're just like,
- 53:57Okay, you're trying to be realistic and then today you're gonna get caught in
- 54:00a loop and it's gonna piss me off. Like how far past that can
- 54:05we get with it? Yeah, I mean eventually, but it'll be very
- 54:08convincing. And I still wonder though, how we as humans will interact with
- 54:15it. And then of course you have children that will be born into the
- 54:17AI universe, and will It's a very interesting question because on one side,
- 54:27of course, and it'll happen, But on the other hand, do we
- 54:30have even children born into because you know, for old people like me,
- 54:35it'll be just weird. It'll always be weird, but in the same way
- 54:39that to this day cell phones are kind of weird to me. But people
- 54:44born into the cell phone universe, it's just a part of your world and
- 54:47you just integrate your life with it neuurnally to some extent. So you know,
- 54:53children are born with AI nannies for example. My goodness, I hadn't
- 54:59even thought about that. But and of course you know, there's this whole
- 55:04other tangent of capitalism and how of course that this will play into that because
- 55:07if you have AI that costs very little, then you can exploit labor more.
- 55:13Parents could work around the clock, or all parents will be out of
- 55:17a job because AI, you know, I just all all the implications there.
- 55:22But the children that are born into that system, will they biologically need
- 55:30a flesh and blood human being, even if it's through a phone? Do
- 55:35they need to feel that human being is a physical presence in the world,
- 55:39Even if we create a physical robot like Data from Star Trek, do we
- 55:44do we need it to be blood and bones and flesh? Biologically? Are
- 55:50we did we evolve to need that that? It's an interesting question, you
- 55:55know, because we can. You're asking the questions of the gold An age
- 56:00of sci fi. But it's come full circle. We're now here acting it
- 56:05out. And I love Data, I you know, think of him as
- 56:09a robot and his positronic or what was it, I can't remember what the
- 56:15neural nets and he's a machine, and he was one of the most human
- 56:22characters on Star Trek, as as I think the card said, Well,
- 56:25it's interesting because so okay. So, for one, when you think about
- 56:30these apps and products that are already out there that are imitating human emotion,
- 56:35usually it's in a sexual way, but there are some I guess that people
- 56:38just use them platonically, but it's typically older men that right that they're they're
- 56:43very lonely and they're seeking out that companionship and they go to these apps and
- 56:49so kind of like going against that point of it takes a younger person to
- 56:52be able to really believe it. But then you ask yourself, Okay,
- 56:55why why these people? And then you think about the new generation being born
- 57:04right now, and how the parenting style. The parenting style is becoming,
- 57:08I think more and more hands on, more and more nurturing. I think
- 57:12we could potentially argue then a few decades ago. So with that being said,
- 57:17even though they are getting more integrated into interact with AI, they will
- 57:24also probably probably be so used to just the tender love and care from their
- 57:30parents that they might not need AI in that way. Yeah, yeah,
- 57:37that would be a nice future. And then of course I think teachers,
- 57:42right, you can at the very least have as an adjunct to school,
- 57:46have an AI teacher that would walk you through lessons and tailor the learning to
- 57:52each student individual, which you know there's so did you say there's apps right
- 57:58now for to sexually interact with? There are? Yeah? I And this
- 58:04must have been like a maybe six months to a year ago that I was
- 58:07watching just some documentary on YouTube that was exploring all these different apps that are
- 58:12currently out there. So don't ask me to like name any of them,
- 58:14but I know they're out there and people are actively using them. Is it
- 58:17just chat based? I believe it's just chat based at this point. You
- 58:22know, I bet that there are voice based ones just because of how readily
- 58:25accessible that that that product is. So I would imagine that they're already being
- 58:31used. And do they have like some stock photo that the individuals are Yeah,
- 58:39yeah, it's I don't think it's to the point yet where it's at
- 58:43the defake level. But again I don't know. This is you know,
- 58:45just hearsay from the documentary. I'm sure that they can. And if you
- 58:50can create it, they probably have created it. Even if it's like a
- 58:53micro app and it's machine learning, it's not just like a script. Yes,
- 59:00yeah, I know, this is actually machine learning. That is very
- 59:05believable from what I was seeing. Huh, And you just dial it in
- 59:08like I want a dominatrix or I want a sub or something. The way
- 59:14that they sell them, the way that they sell them is more so like
- 59:17you know, like the applics what was it called second life, where like
- 59:22you're you're just actually like playing this game and interacting with it like it's your
- 59:27second life. It's it's it gave me that vibe where it's more just like
- 59:30come, you know, come and chat with Jenny, and then like you
- 59:35and Jenny just build a relationship and not so much like I'm prompting you to
- 59:38be a dominatrix, but like tell me about your day. I love you,
- 59:42I miss you so much. It's really interesting. Yeah, wow,
- 59:46huh. I mean I knew stuff like that was happening, but I didn't
- 59:51know a for some reason, I thought the builders of AI were avoiding that
- 59:58market as of yet, but of course would get into it later. But
- 1:00:00of course we know that that's one of the first money markets that you can
- 1:00:06get into right now. Well, and then this really brings up like the
- 1:00:08conversation around so uh, you know, there are AI engineers all over the
- 1:00:15world with all different types of motivations, and you know, us that are
- 1:00:19looking to build AI for good, AI for all, AI for corporate,
- 1:00:22you know, AI for for non nefarious. Of course, we wouldn't necessarily
- 1:00:27think of building that, but you know, a big part of this,
- 1:00:30the executive order that just came out today by the White House, was trying
- 1:00:35to build a whole bunch of safeguards into the fact that we know that there
- 1:00:38are bad guys out there, and like bad guys. That's like god hold
- 1:00:42at the other level than building sex spots, but like, we know that
- 1:00:45there are truly bad guys out there that know how to use AI because it
- 1:00:47just takes a few years of training to get really good at AI. And
- 1:00:52what's to stop them from, you know, doing really bad things with the
- 1:00:55same technology that we are using to do really good things. So someone could
- 1:01:01have a home server and develop their own AI machine and do all the things
- 1:01:08that we're seeing these other AI do. Yeah, I mean I do it
- 1:01:14every day. I'll like think of an idea of like I mean, that's
- 1:01:16how I started my app. I was like, hey, I have this
- 1:01:19idea, let me pull up you know, these days we use cloud computing
- 1:01:23a lot of the time, so like AWS, Google, Microsoft, those
- 1:01:27kind of companies. But you just pull up an instant and you start making
- 1:01:30these models, which nowadays, you know, machine learning is so commonplace.
- 1:01:35They actually have these concept called auto mL Auto machine learning where you can just
- 1:01:39literally type it and you can actually you can do this on GPT. So
- 1:01:43if you're at your you know computer right now, if you have a GPT
- 1:01:46account, type into it, write me a Python code to create a recommendation
- 1:01:53engine for recommending books. It will literally give you the code to do that.
- 1:01:59You would copy and taste that into a Jupiter notebook or like an AWS
- 1:02:02notebook, and it will run and it will work. So that's where we've
- 1:02:06got into with our development. Anyone can do it. Wow. Wow,
- 1:02:10I didn't know that. I thought AI was like some supercomputer shit that you
- 1:02:15needed like way more computing power to run the machine learning. But you're saying
- 1:02:22that any dumb laptop can become an AI. Yeah. So it's a good
- 1:02:28point, right. So it's really the difference between an algorithm and a model
- 1:02:34and a supermodel that's a supermodel. But so an algorithm is just you know,
- 1:02:39they're like, let's talk about like recommendation engine, Like there is an
- 1:02:43algorithm for a recommendation engine that just says, first you do this, then
- 1:02:46you add it to this, divide it by this, multiply it by this,
- 1:02:50and then you have your answer and you can just write that in code
- 1:02:52and say okay, I have this machine learn and that doesn't take anything at
- 1:02:55all. You just need like a notepad to write that. Then when the
- 1:03:00data comes in and how much data and how you're processing the data, that's
- 1:03:05what will be the difference of can I run this locally or do I need
- 1:03:08to spend thousands of dollars in like GPT for example, I think they're spending
- 1:03:12something like seven hundred thousand dollars like a day or a week or whatever it
- 1:03:16is to run GPT. So it's yeah, it's how much data do you
- 1:03:21need to train that model? So if I had that Python code and I
- 1:03:28put it on some situation and I'm running it just for myself, and I
- 1:03:35feed a bunch of info into what books I like and what books I don't
- 1:03:38like, and since the data set is a very big because it's just me,
- 1:03:44then the machine learning slowly figures out an algorithm or weight to the neural
- 1:03:50calculations that's specific to me, and very good about that. But it's but
- 1:03:54it's a low data crunching situation because it's not trying to do for every human
- 1:04:00on the planet. Yeah, exactly, And a lot of times I will
- 1:04:03just play around with proof of concepts on my computer with you know, maybe
- 1:04:08like ten thousand rows of data something where it'll learn from it but not necessarily
- 1:04:12be an amazing algorithm or model. And then once you are really confident with
- 1:04:16the model, then that's when you would go and start spending thousands on compute
- 1:04:20power to build something spectacular. You should do it for Tender yourself. Yeah,
- 1:04:28he should put some app on your phone and then say, okay,
- 1:04:32figure out, and then after you go on a date, you would feed
- 1:04:36in ratings and then you know all that machine learning or I guess you could
- 1:04:42even just historically just say look at my entire tender history and I'll tell you
- 1:04:46exactly how many dates are. And it has all the chat logs too on
- 1:04:49Tinder. I'm just saying you could have your own. I'm sure they're already
- 1:04:55doing it because like unhinge after you, so they know when you give someone
- 1:05:00your phone number and about a day or two later, they'll say, did
- 1:05:02you meet this person? So they're like they're they're doing it already. Okay,
- 1:05:08okay, but it's not tailored to you, you know, like you
- 1:05:12because you could feed in more data, like you know, this this person's
- 1:05:15a good seven date person, but this person looks like they might be like
- 1:05:19forever dating sort of. I don't know. Or this person was not only
- 1:05:27a no to a second date, but was a sharp no. I don't
- 1:05:31know. I think this seems like just seems like and then you could ask
- 1:05:34your dating people to put in data for themselves. I don't know. I'm
- 1:05:39just saying, since, yeah, but you're just giving me more app ideas,
- 1:05:44Kirk, I don't have enough time to create all these apps. Yeah,
- 1:05:47I mean, that's kind of the thing that I think we're looking at
- 1:05:49over the next five to ten years is just all of this potential. And
- 1:05:55of course, you know, there's a lot of money to me made.
- 1:05:57And you know, earlier we were talking about romance and porn I suppose by
- 1:06:03extension. And I remember in the early days of the Internet that I remember
- 1:06:09there was this feeling that I had that a lot of other people had,
- 1:06:12which was the Internet could be great, but it's so small at this point,
- 1:06:15you know, like late late nineties, and I remember that I could
- 1:06:20see the potential. But I think there was one website that I went to
- 1:06:25and I think it was like a Princess Die website. And I'm not even
- 1:06:27interested in Princess Die or the royal family. But because the website was like
- 1:06:31mildly interesting, I think I looked at it every day. There wasn't anything
- 1:06:35else to look at. There were no news websites, there was no Google,
- 1:06:39there was no you know, gaming or anything. It was just this
- 1:06:42open book of this, you know, let's see what we can do.
- 1:06:45But there was this problem because there wasn't enough investment from corporations or from entrepreneurs
- 1:06:54to do anything, because there wasn't enough customers yet. So you had this
- 1:06:58kind of catch twenty two like, well, people aren't gonna make websites that
- 1:07:01are interesting until we get customers, and customers aren't gonna come until you have
- 1:07:05websites. And then there was this huge explosion through the early aughts, you
- 1:07:11know dot com stuff, and we were happy, but I heard from people
- 1:07:15that knew these things that really what drove the Internet was porn. That you
- 1:07:23had porn providers and there were customers like just Champion at the bit because up
- 1:07:30until this point you had to either subscribe to a random cable channel or you
- 1:07:35had to go into a store and purchase things. You know, I mean,
- 1:07:38it's not only inconvenient, but it can be shameful for some people or
- 1:07:42worrisome, and to just sit at home on their computer and dial in these
- 1:07:45things. And so instantly there was all this money flood. Again, I
- 1:07:50heard percentages of you know, like sixty to ninety percent of the initial commerce
- 1:07:57and building of the Internet was based on pornography, and and other things came
- 1:08:01all other stuff, you know, Google and Amazon. And then I think,
- 1:08:05well, with AI, we might be hitting a similar kind of impasse.
- 1:08:12But of course the demand for porn or someone to interact with on that
- 1:08:17level might be the catalyst that pushes AI into the strategy. Interesting. Yeah,
- 1:08:26that's a really interesting concept. And I think about this a lot too,
- 1:08:30just like the concept of you know, like ten twenty years ago,
- 1:08:33with the Internet, we had all of these really big visions and it feels
- 1:08:36like so little of them have really been accomplished that have made it mainstream.
- 1:08:41And I think the problem is just when things get so massive, like Google
- 1:08:45and Facebook and Amazon, so massive that it becomes exponentially harder to make really
- 1:08:50big changes, those kind of innovative changes. But that's why all of these
- 1:08:54startups, especially these startups that are in the AI space are are going to
- 1:08:58be the ones that are able to make that impact that you're not able to
- 1:09:00see. Like I So one of the things I do on the side,
- 1:09:04I work with an AI incubator here in Miami. Just it's like something for
- 1:09:09startups when you're trying to start up a company, just like helping companies get
- 1:09:13off the ground. And people will always ask like, Okay, well,
- 1:09:16shouldn't I like be in stealth mode. Shouldn't I like not be trying to
- 1:09:19tell anyone about my app? Because you know, Amazon can just go off
- 1:09:23and build it, but in reality, they're not going to. They have
- 1:09:27so they're just trying to keep their you know, their website afloat and making
- 1:09:30sure people are happy that they don't have the time to build the app that
- 1:09:33you're building. And the model is you make a great app and they buy
- 1:09:39it from you, and you make the big exactly. Yeah. The last
- 1:09:42thing I'll say is, if you, dear listener, are listening to this
- 1:09:45in twenty twenty eight and beyond, you'll be giggling throughout this entire episode because
- 1:09:53everything will have progressed to that point and everything I'm saying now will sound like,
- 1:09:58I'm this old man and nineteen hundred's talking about how they are these horseless
- 1:10:02carriages. So I beg you please understand that the things that I'm saying are
- 1:10:09extremely advanced for the people of my time. Most people don't even know the
- 1:10:14shit that I'm talking about, so stop laughing at me. I'm a smart
- 1:10:18person. And thanks for coming on the podcast, Janetta. George, do
- 1:10:23you want people to reach out to you at all or or find you?
- 1:10:25Or yeah? Yeah, absolutely, CANDORAI dot co is my website. You
- 1:10:31can find me on Instagram at two underscores and then my name Janetta. I'd
- 1:10:36love to connect on AI. Or if you're on on hinge and into AI
- 1:10:42and living in Miami, then maybe this is better than Hinge. So on
- 1:10:46Instagram two underscores, j E N N two NDS E T T two t's
- 1:10:51a Jenetta. Thanks for coming on the podcast, Janetta. And what are
- 1:10:58you looking forward to me rect to you next? Ooh good question. So
- 1:11:03I'm all caught up on everything you've done for love is Blind. You know
- 1:11:09ninety day Fiance. They've got like ten different spinoffs going on right now.
- 1:11:13I've never overwhelmed by ninety day Fiance, but I am overwhelmed by the content
- 1:11:16right now, so I think they just they need to come up with a
- 1:11:19new episode, a new new spinoff for specifically for Kirk to react to,
- 1:11:24and then I'll listen to that. Well. I'm in the process of watching
- 1:11:30Last Resorts still even though I'm pretty sure that season is over, and also
- 1:11:35watching Mary and what's his face? But yeah, there's another season that just
- 1:11:42started not long ago, and I'm planning on taking a poll and seeing which
- 1:11:47couple I should watch from that season. So there's a new straight up ninety
- 1:11:51day Fiance and there are a few really interesting couples. There's an interesting couple
- 1:11:56and I can't even remember their name, but the girl she's got long blond
- 1:12:00terre and she's from the UK. And the guys from like La or something.
- 1:12:03That'll be a good couple to react to. They've got some interesting stuff
- 1:12:06going on. Okay, I'm intrigued, And everyone out there, please take
- 1:12:13care of yourself, Janetta. Why should they take care of themselves because they
- 1:12:15deserve it, They really, really do.