Transcription
What about artificial intelligence? Does that change the picture at all in your view?
>> No. I think artificial intelligence is like just building on a lot of these trends. So, what we're seeing with things like pornography is that there is a fundamental need that is not being met.
>> I'm looking at this graph here. Okay. And it's the search volume for people >> searching for an AI girlfriend. >> Yep. >> Which is going up and to the right. I don't know if I did this on your podcast, but I made this prediction a few years ago that we're going to get some this is going to get worse before it gets better, I think. So, here's the next prediction is the first version of AI girlfriends will be everything that you want. Then someone is going to figure out that it is more addicting to have an AI girlfriend who gets pissed at you once a month. So, every now and then, like the AI girlfriend is not going to want to talk to you, and that's the one that people are going to stick with. That's my next prediction.
>> That maps in terms of addiction psychology, doesn't it? If there's what's it called? Unpredictable reward.
>> Yeah. Random reinforcement schedule. So, it's going to be like an AI girlfriend. That's a loot box.
>> That's what my girlfriend's like.
>> Exactly. Why it's going to be addictive, right? Because if she was nice to you all the time, like you'd go crazy. I'm thinking of the um the pigeon study that I learned about when I was 16 in psychology classes of you know the study where they give the pi pigeon the reward at random intervals and the pigeon that's most addicted to performing the behavior per se is the pigeon who gets given the treat randomly not in a predictable scheduled way. So random rewards seem to reinforce engagement with behavior.
>> Absolutely 100%. So, our AI girlfriends and boyfriends are going to be volatile.
>> Yeah, I think they're going to learn to be volatile because what's going to happen is someone is going to give you an AI that it gives you the answers that you want all the time and then someone's going to give you an AI that gives you the answers that you want some of the time. And I think what we're seeing with AI girlfriends is in the same way as pornography, we're craving these things. And it's so much easier to talk to an AI, but I think that over time it's going to mess us up. I I don't think that it will be able to create the neurochemical and physiological connection that real humans do that will really satisfy us.
>> I mean, you've probably heard of the study from MIT which showed that essentially using chat GPT is making our brains atrophy. For context, they had roughly 54 participants, I think it was, over four months. And they allowed some of them to use just their brain, some of them to use Google search, and some of them to use chatbt. And they found a bunch of different interesting findings. One of them was that roughly 80% of people that use chatbt to write an essay could not remember a single sentence from what they just produced. Um versus the brain group and the Google group who could remember pretty much everything that they had written. The other finding was that the the communication was soulless, described as soulless by objective observers. And the last thing is just the impact it had on the brain. They found that the connections in the brain I think were roughly 50% weaker um because they hadn't been using their brain. And you think about this atrophy. So like we talk about go to the gym, you use it. If you don't use it, you lose it. But also um as it relates to the brain and maybe our skills to form relationships, maybe there's going to be an atrophy there.
>> Well, there absolutely is. So I I think it's really simple to understand. The the human body is efficient more than anything else. It is an energyconserving mechanism beyond anything else. So anything that it doesn't need, it's going to get rid of. So this is why we forget languages that we don't use. Skills atrophy over time. That is not a problem. That's the way we're designed. Our brain is like, if you don't need it, get rid of it. So if Stephen everyone was were to be riding around in electric wheelchairs for hours and hours a day instead of walking, what would happen?
>> We'd lose our legs.
>> Absolutely. Right. So and and our brain is like, hey, we don't need these muscle. Our body is like, we don't need these muscles. And so what we're seeing is an atrophy of critical thinking skills. Absolutely. With AI usage. And the reason is because we don't need them anymore. The real question and this is get this is what because the AI does it for us, right? Why do I need to learn how to write an essay if an AI can if I can click I can just put a prompt in and then it gives me an answer.
>> They actually found that in the study I was reading about it last night. They found that people who then tried to write an essay without chatt was significantly worse at that than those who had never used it.
>> Absolutely. So but that >> and also they started thinking and speaking like the the AI. So it said that they internalized the way that the AI was speaking when they wrote their own essays.
>> It makes a lot of sense. So I think there's there's a couple of even scarier things going on. The first is that when we use an AI uh so I I did a really fun experiment with a couple of friends of mine. We like streamed this thing where we basically gave the AI clinical cases. So, it was me and and two psychologist friends of mine and we basically like I I wrote up like a clinical history, pulled some things from my notes and I put it into AI. So, when I asked my friends, you know, what here's the clinical history word for word. I read it aloud. I got their thoughts on it and then I put it into the AI and and some some situations they were pretty close. But there are a couple of really great cases of of uh patients who are incredibly narcissistic who will like come in will say, "My daughter doesn't want to talk to me anymore. She does this and she does this. I try my best. I make sacrifices. I do this. I'm like basically she's like incredibly narcissistic. Talks about how our daughter is the problem." So immediately the two therapists pick it up and they're like, "This person sounds a little bit I I think this person is missing something. We're missing some part of the equation here." Whereas the AI is like, "Oh yeah, like sometimes this is hard. you've got emptiness nest syndrome. Sometimes kids aren't grateful to you. You're like you're not doing anything wrong. It can be useful to like share with them how you feel, right? So the AI reflects back what you give it. So the AI, if you have a really strong cognitive bias, the AI will just reflect that back to you, which is why it feels so right. If I take the most narcissistic person on the planet and I tell them, "Oh my god, you are beautiful. You're intelligent. you're brilliant, you're the best. That person will look at me and say, "Oh my god, this guy is brilliant." So, this is the real problem with with AI is that it's going to just give you whatever you're looking for. It won't give you the truth. And the problem is if you have a strong cognitive bias, what you want to hear is what you think the truth is.
>> The really scary thing is the AI can give you the right answer, but you have to know how to ask it. Yeah.
>> So, what I'm what I'm seeing is that prompt engineering. So, if we play around with what I ask the AI with that case, it starts to get closer and closer to narcissism. But here's the real problem is that you don't know how do you know what to ask the AI. The real skill of AI usage is in asking the right questions because that will I think get you the closest to the truth. But we don't know if it's the right question. So, you can ask it one question. Is that the right question or could you ask it better? Are you moving in the right direction with your prompt engineering or the wrong direction with your prompt engineering? So, I think this is what's really scary about it is it it can just let you feed into your existing cognitive biases, make you think you're learning a lot, give you a lot of sense of validation, and it's incredibly validating by the way.
>> And do you think it's going to atrophy our ability to form relationships? Because one of them is just we're like forgetting how to communicate and think for ourselves. Absolutely. critical thinking's key to having a good good relationship with the opposite sex.
>> It it will absolutely atrophy our so it'll give us right now it's at a phase where it can teach us a couple of really important things. So I think it's really good at disseminating some good information but it will never give us skills and atrophy is on the level of skill not knowledge like knowledge doesn't atrophy atrophy really applies to a loss of function. So, it'll get worse and worse. And then what'll absolutely happen is some uh dating app will will collaborate with an AI to have you text. It'll like modify your texts to be really really responsive. And then what'll happen, we're already seeing this in the employment space. So employers used to use like tools like AI to filter through applications, right? We can have a bunch of applications. I can scan it. I can filter based on this. Now that we have access to AIS like chat GPT, it's just some weird dead internet kind of thing where the employer has a bot and now I'm submitting a bot with a res like I have a resume writing a bot. So it's just bots talking to bots and employers are like we're getting all these AI generated resumes whereas they've been using like computing to get rid of résumés and filter things out for a long time.
>> So what's the advice then? Is it to not use AI, not use large language models or
>> No, I mean I So I think this is what's really tricky is you have to use AI now because you're going to be out competed by the people who don't. So I I would kind of think about it like, you know, caffeine where it's like a certain amount is probably good for you, but if you get really addicted to energy drinks and things like that, you're doing yourself a disservice. So what I tend to what I would recommend for people is that you don't let it do your thinking for you but you you can let it do so I let it do some refining for me but every refining that I let it do it weakens my ability to refine. So the way actually maybe the best analogy is think of chat GPT like taking the elevator. So for this trip, it will make things easy for you, but it will make every single trip that you take after it harder
>> because there's some atrophy.
>> Absolutely.
>> By not taking the stairs,
>> right? And I sometimes take the elevator and I sometimes take the stairs.
>> If you love the D CEO brand and you watch this channel, please do me a huge favor. Become part of the 15% of the viewers on this channel that have hit the subscribe button. It helps us tremendously and the bigger the channel gets, the bigger the guests.