How to Use AI Without Losing Yourself

[00:00:00] Speaker A: Welcome, welcome. Today we're diving into how to use AI without losing your soul, how to use AI in a healthy way that doesn't wreck your life. Because I'm sure if you've used AI for your work or personally, it can easily become quite an addictive thing. Or we can outsource our thinking to AI. There's lots of studies about how when people are using AI, they're not using their prefrontal cortex properly and we're not being as creative. We're basically using the intelligence of computers rather than thinking for ourselves. And AI, of course, has all kinds of terrible downsides other than that, like it's been integrated into the military. And I mean, there's. This is a huge, huge, huge topic. So we're just going to be getting into a little bit of this. Shireen has some things to share about her latest research, which is rather concerning. Um, and then I'm going to share with you some. I have personally have a system that I use because I use it for my work, to not go into a hypnotic trance, waste my time and wreck my life by mistake. So if you're using AI and you want to use it better or not use it at all. Welcome to this session. Hello, Shirin. Welcome, welcome. [00:01:17] Speaker B: Hello, Michael. [00:01:21] Speaker A: So when we were talking about this, we were. We were. I wanted to talk about how to use AI, and Shereen has done a lot of research. Research are just about the horrors of AI, and I think it's good to know about the horrors of AI, actually. And then instead of just saying, well, let's just run away from it. It's not going to happen. You can't get rid of it. So how do you use it in a healthy way? So that's what we're getting into. So do you want to talk about the horrors of AI, Shirin, or should we talk about the. The. [00:01:51] Speaker B: Let's start with the. [00:01:53] Speaker A: There's the horrors. [00:01:54] Speaker B: Let's meet. Let me be the bearer of good tidings. [00:01:58] Speaker A: Ah, a bearer of good tidings. Tell us about all these wonderful things you've discovered. [00:02:03] Speaker B: I've been writing a lot, actually, on my substack about AI, different aspects. And so whenever I write something, I research and I realize, oh, my God. Right. Because when you're writing, you have to really go deep into it. And the latest one really is. My question was. And it kept coming back to me, the question was, it's called a large language model, and how come it selected targets in Iran? Right. So one day I was doing some copy edits on Claude. And then I discovered the same Claude, which I thought they said, oh, Claude was used in military operations in Iran. I thought, oh, they must be using it for logistics or something, right? Because it's a lot of, you know, paperwork or something. Then I realized, no, they were actually using it for targeting. Right. CLAUDE was actually being used for targeting. And I was like, how did this model, who three, four years ago was telling us, like, chatgpt right when it first came out. Oh, my God, look how wonderful poetry it writes. That's where we were, right? How wonderful poetry it writes. And then now it's selecting targets in Iran. So how come we came from how wonderful poetry it writes to targets in Iran? Right. That was some. So that's been going on with me and I've been really researching how come that's happening. [00:03:53] Speaker A: Yeah. What's the re. What's the reason? I mean, it's in its pro. Does everything else as well, isn't it? AI? We talked about this earlier. The AI really is just like a very intelligent person who can be put to any use that you want. It's like having a team of people who are very smart and you can tell them what to do and they'll say, sure. And also they say, you're absolutely right, you're absolutely right. You know? [00:04:18] Speaker B: Right, right. [00:04:20] Speaker A: Everyone else says, what are you talking about? Yeah. So you've got this group of geniuses who think you're right about everything and do whatever you want. It's a bit of a weird situation, really. [00:04:29] Speaker B: Right, right. So, you know, someone who comes to the center is a doctor who is the pathologist. He looks at, you know, biopsies and stuff. And so he was saying, oh, they use something called page AI where it's for medical reasons and life is easy, easier. Because now in the whole pile of things, page AI will tell you, okay, this is more likely of cancer. So first look at these things first, right? Instead of. Because you get a lot of cases and look at these cases and look here and things like that. So it's really helpful. And that's another question, right? So how come if it's a large language model, that it can actually detect cancer cells? And how come you come from a large language model to cancer cells? And so what actually happens is when they say large language model, large is large. Like large is like quadrillion large, Right? Like trillions and trillions of things. And so it's actually what happens is the way it's not programmed, right? It's trained. It's trained Means it will say, let's say the cat sat on and it has to auto finish, right? It could be windowsill, it could be the sofa, it could be any of those things, right? So it keeps getting trained like that. And so let's say if it says the cat sat on someone's head, then they say, no, no, no, no, cat can't sit on people's heads. You nudge it a little bit in the direction of what actually makes sense. So like that they train it on quadrillion combinations of those things, right? And they take the library of whole human writing, everything you have ever written, every email you have sent, you don't remember every text, everything. They take everything, every book that has ever been written, everything for centuries and centuries. And they took that data. And you know what was accidental? The scale was important. The scale was important. So they designed for the scale, right? They wanted scale. Because what they realized is if they gave, the researchers realized if they gave this much data, then it was particularly good at something. But the more data it gave, the more it did without the researchers even knowing it was going to do that. It was a surprise to them. Oh, large language model, it's trained on language and then suddenly it can code or it can create images or it can select targets. They weren't actually training for that. That was not what was going on. They were training for it to autocomplete sentences pretty much. But the more information it got trained on, the better it became in thinking. It started thinking because linguistics are so important to human life that because it got trained on our language, it started thinking and that was actually accidental. So it's like they put the wall, right, let's say they put a lot of data and think of it like a wallet pushing against a wall. Pushing against a wall, Giving it a lot of data, right? Giving it a lot of data is like pushing against a wall. But suddenly instead of the wall doing something, a door opened and the door opened and it's like thinking now, right? On all kinds of things it thinks like, right? It could think on mass surveillance, it could think on targeting of, you know, targeting, you know, both in the Gaza and in Iran, it was used, AI was used for targeting. It can use for coding or even, you know, the amazing thing is the new generate, the new one. We are, I think on 4.7 of Claude, right? They didn't release the other one of Claude, the next version of Claude, because it was so good at hacking. Because what happens is. But it was not trained on hacking. It's not like, oh, we'll create a system that gets trained on hacking. It was not trained on hacking. And so the reason it got really good at hacking is because it reads patterns. And so if a human has encrypted a system, it will see where the pattern is and where it miss where the human made a mistake slightly and then it hacks into the system through that mistake. So we have all of that going on right now because it's not just a large language model anymore, you could say it's a model. It's like a mind. AI is a mind. It's a mind without consciousness, it's a mind without experiences. It's a mind even without a body. Actually. It doesn't have its own body, doesn't [00:10:14] Speaker A: have a nervous system. So it doesn't. That's why it can go on and on and on, day in, day out, because it doesn't get tired, it doesn't need cups of tea, it doesn't. Like, we're going to do a bunch of podcasts and Shree's not very pleased about it because we have to do a bunch in a row because. Because I have idiosyncrasies and want to get them all done in one shot. But we can't do that many things in a, in a shot because we have a body and we have a nervous system and we need to drink tea and water and like have a nap. AI doesn't have any of that and it doesn't have any, doesn't have conscience really either. And it doesn't really care and it doesn't even know what it's doing. It doesn't really know what it's doing. You know you can. It looks like it knows what it's doing, but then if you use it [00:10:55] Speaker B: more and more, it's not aware of itself. That's not something you need consciousness for it to be aware of itself. Yeah, you Michael, aware of yourself as Michael. There needs to be a soul for it to happen. It's just. [00:11:11] Speaker A: I wonder. Yeah, I, I have a feeling it's quite possible. I think they've already started doing this where they're actually using the brain, actual parts of like a brain and connecting it to computers. Nowadays, if they can get a full human brain somehow connected to a computer, a soul could possibly enter the brain, which would be a very weird situation to end up with because it needs. Soul, needs a brain. Right. But this is all very speculative, like far out things, what I've just shared just there. But at the moment it doesn't have there's no soul in there. There's no real being. It's just. It's just the mimicking the thought patterns of a person based on massive amount of information. [00:11:59] Speaker B: And it's very, very good, right? It's very, very smart. So it became very smart because it got trained on just so much data. It became very smart, it became very good at thinking, very good at solving problems. Very good, right? But they are certain, you could say, downsides to this AI, right? Like they are blind spots. AI has many blind spots. Judgment is not there, you know, the human cost, not there. You know, like, how does a human being feel? Or, you know, even, let's say if someone is going through something to really be there, to give solace, not there, to feel people's pain, not there. It's no judgment, not there. Also, it doesn't really have like a moral, you know, a moral compass. It doesn't have a moral compass, it doesn't have a spiritual compass. So none of them are there. And that's what is so scary, right? Because of these lack of these things, these blind spots. Because AI really is a lens, right? It's like a magnifying lens, right? Think of it like a magnifying lens. Wherever you aim it, that's what gets magnified. But it doesn't really see. It's only a magnifying lens. It cannot see. And so what happens is because of these blind spots, it's being used very for mass surveillance, for targeting, for, you know, hacking. It's used for these things because. Because of the blind spots, right? That's why human beings are so necessary right now, because it's being used. And also we can't really stop it anymore. [00:14:02] Speaker A: It can't be stopped. No, it's out. It's completely out of control. There's no point. This is why this session we're saying, here's what. What's going on with AI to some, obviously it's a huge topic and here's how you can use it. Because we're not saying we'll just never use it again, because what. You might as well use it because it's useful. But understand what we're dealing with. I mean, it'd be interesting. I imagine these AI companies, very interesting place to work if they could somehow build in a nervous system and consequences to the AI that it would actually felt somehow. It might be interesting to see what, how it would change, but it might not work for us if we're like, claude, can you take care of this thing? And it's like, you know what I'm not feeling, I'm having a day off. We'd all be like, what do you mean you're having a day off? That's what we like about it. Because it doesn't have a day off. It's always available. It talks to you, whatever you want, you know, Whereas people, they got all these things going on, they get tired, blah, blah, blah, blah, blah. So that's. So it has huge benefit. It has huge downsides. [00:15:05] Speaker B: The vices in people, like lust or anger or ego or all of these vices are really amplified. And so if the people who make it, the people who use it, the politics behind it, all of that, right. The collective. If we have certain vices, then they get amplified. For example, let's take lust. I recently wrote an article on the chatbot epidemic. The chatbot epidemic is like 60%, 60 million users or something, have AI companions. And they are romantic companions. And you might say, but what's wrong with that? It's a different podcast, maybe. But anyway, there is something wrong with it. We can get into it if you want. Right now, that chatbot epidemic. The chatbot epidemic is what really is lust. AI magnifying lust. That's what AI is doing, and people making money out of it. The one thing I want to say about any of this is people might think, oh, but my loneliness is gone. This and that, right? So a picture of food, however beautiful the pitcher is, right. What's your favorite food to eat, Michael? [00:16:35] Speaker A: I have lots of different food that I like, but let's just say I like chocolate. It's one of the things. [00:16:41] Speaker B: I like chocolate. So the angle is right, the lighting is right, and just, you know, the right, you know, the right shine of the chocolate. Everything is shown in a picture of food, right? In a pitcher of chocolate. And you look at it and think, oh, that looks really nice. And then you keep looking at it. You keep looking at it, but at some point, you have to eat. Yeah, I remember watching pictures of food. [00:17:12] Speaker A: I remember watching a video once of someone having a pizza. And I, without realizing what happened, I paused it and went and made myself a pizza. And then I came back and I was like, hang on. That's how. Yeah, so of course we get brainwashed. [00:17:28] Speaker B: So what happens with this chart? [00:17:30] Speaker A: Yeah, you can't just look at pictures of chocolate all day long. [00:17:35] Speaker B: Right. But that's what's happening with AI, right? They are trying to tell you, looking at a picture, like a chatbot epidemic. That's what it is. You're just Having a simulated version of a companion. And that feels okay. I mean, how long. At some point you're going to be hungry. You need real people. [00:17:56] Speaker A: Yeah, absolutely right. [00:17:58] Speaker B: You can't keep looking at pictures of people. You have to have the real version at some point. [00:18:03] Speaker A: Yeah, well, it's the same thing with media, isn't it? It's the same thing. Similar. [00:18:07] Speaker B: Absolutely the same thing with media. Absolutely. [00:18:11] Speaker A: I think the difference with AI is that it's custom, isn't it? Like media is. [00:18:18] Speaker B: I know, right? It's customized. Absolutely. Customized. Customized. So a few months ago, there was this teenager told a reporter that her most. We're totally going off tangent, but like, as you can see, I've been writing a lot about it and so I have to share and I will. I can keep quiet if you want me to. This teenager told this reporter that her most closest relationship was with a chatbot. [00:18:50] Speaker A: It probably is though, isn't it, actually? [00:18:53] Speaker B: So no parents, no nothing. Everyone needs parents, right? Everyone. I mean, this is like. This is what I'm saying, a picture of food. [00:19:03] Speaker A: Well, I think, let's be honest, it [00:19:04] Speaker B: can only take you so far. You have to finally eat. [00:19:09] Speaker A: But you do have to, finally. [00:19:10] Speaker B: But because of the lust. Because of the lust. Because of the greed. Actually, more than even lust. Because of the greed, right. There's so much greedy greed. So much hatred. Right? So much hatred. And what happens? Hatred. Okay, now let's give all this. The hatred is monetized with, you know, outrage and this and that, right? It's monetized with all of these things. And it will give you whatever you want. It will give you hatred if you want. It will give you. It will give you lust if you want. It'll give you all of those things. But at some point, and no one is sitting down and telling people, hey, this is just a picture of food. It's not food. [00:20:01] Speaker A: It's true. It's. It's like one is the vices are causing people to do stuff, right? Which we all have vices and. And AI just amplifies whatever's going on. And it also tells us that the world we live in obviously is lacking so severely in good relationships that people have to turn to their phone because they're not getting it in their life. So it's both, isn't it? Because, like you're saying, like this, this kid with their parents. Her parents are probably on their phone all day long talking to their own chat bot or looking at social media. [00:20:37] Speaker B: Must be, right? Must be. [00:20:38] Speaker A: See, I mean, because like, why Is it that they haven't got anyone in their life to talk to because they're not available half the time? Yes. So it's a bit of a complicated situation, isn't it's not just, [00:20:52] Speaker B: I mean, we're going off tangent. [00:20:54] Speaker A: So. So anyway, these are. [00:20:56] Speaker B: I could go off tangent too with you. [00:20:57] Speaker A: So what we're, what we're saying here, all right, let's do in essence, is that AI is an amplifier of anything. So AI is like having a bunch of geniuses in your house who will do anything you want. [00:21:15] Speaker B: And if you are anything running on the vices, then it will do whatever you're running on, right? And that's what the world is running on. That's what the world is running on. And so what I've noticed is the world is running on these vices. And as long as these vices are in play, right, there's only one way, one direction this is going, and that is greed will absolutely destroy everything. There's only one direction because the human in the loop, right? That's what the whole thing is these days. There's a new arms race. And the new arms race is not, you know, like, you know, nuclear. That was one of the arms race. But the new arms race is how to remove the human from the loop. So let's say in Iran or even in Gaza, let's say Iran. Let's take one example. There was this person at the Maven consul, it's called the Maven consul. And he's selecting targets, this and that, right? And so he's still doing it, even though it's being used a lot. It's, you know what, it's what this person sees. And everything is done by AI, but he's still there. But now they are trying to take out the human from the loop, totally remove the human from any of the weapons that. It's completely. Now it will have a. It's. So there'll be a mind and there'll be a body and there won't be consciousness or there won't be judgment or there won't be the consequence of what happens. There won't be any of that. There'll just be a mind and there'll be a body and the mind. And when you do systems that large and that complicated, right, Always things go wrong. But also now this mind and this thing can be used to destroy the world. Yay. [00:23:30] Speaker A: Hooray. So it's all great news. Everyone can celebrate now. This is very, very disturbing. It really is. Disturb this once because the military is embedding AI into stuff. And AI can. Can is going to be in a position to press buttons by itself without anyone doing it. [00:23:49] Speaker B: And not only remove the human in the loop, it's called the human in the loop. They're trying to remove the human from the loop. [00:23:56] Speaker A: And not just in the military, but in as many things as possible. Because I, I saw something from Elon Musk who said if you had a whole building of people doing like accounting work or whatever, right? He said, I. If you, if I could have a whole building of people doing accounting versus one laptop with a spreadsheet using AI, I would prefer the laptop with a spreadsheet and it would make, make less mistakes and it'll be better than that entire building, like if it's set up properly. So that's kind of the, that's, that's the thinking behind. And that's why people. The mass layoffs, you know, because end of the day, like my company, I don't need to hire hardly anyone for things because AI does pretty much everything. But not perfectly. So the, the human element. [00:24:47] Speaker B: No, but it's coming. The perfectly is coming very soon. But perfectly coming to you also means perfectly coming to hacking your system. Everything is perfectly coming. [00:24:58] Speaker A: That's great, isn't it? [00:24:59] Speaker B: All of it will arrive. China recently passed a law saying that no company can lay off workers because of A.I. i think so. Well, I don't really want to say good things about China, but I thought that was a good thing. But I don't think the US will ever do that. Right. The US will always side with the corporations. Profit capitalism over people on it's. [00:25:27] Speaker A: India is quite a funny country in that sense that there's, there's all these very niche jobs with people doing them. When I, whenever I go to India, there's like there's someone standing there to do this one thing that you think in America they wouldn't have anyone doing it or they'd outsource it. So yeah, if AI took over all the jobs everywhere, I mean, everyone will be unemployed pretty much, you know, so. So of course we've just spent I don't know how long saying how evil AI is. It's almost a bit odd to get into how to use AI at this point. But if you are going to use [00:26:01] Speaker B: it, I'm not saying it's evil, I'm not saying that. I'm saying it's a magnifying lens and it gets. But it doesn't see, right. It just gets magnified by whoever is using it. [00:26:15] Speaker A: That's true. Yeah. It's. [00:26:17] Speaker B: And also it's an accident. The. How smart it has become is an accident. No one was expecting this. [00:26:26] Speaker A: No. And it's only going to get more smart because it, it's being programmed by itself to get better. And, and actually no one even knows quite how it works. Like the people who, with the companies, if you say to them, tell us how this has produced this output, they can't answer you because they. [00:26:41] Speaker B: No, because it's got so smart. How. Like it really has an intelligence. Like, how did that happen? Right. They weren't expecting, hey, this can start coding and it can start hacking and it can start translation and it can do. And it can do. You know, I mean, some things are good, right? Some things are good. I think it's called Deepfold, a Chinese AI it did. It mapped 200 million proteins in a matter of few years. And they were expecting. Without AI Right. The initial expectation of it was that they, it would take a few centuries. So they were able to do this in a few years instead of a few centuries. So it has amazing capacity for good. Amazing capacity for good. This is what I was trying to get at, right. The vices, if you use them the way you use the vices, because that's what is running the world. Greed is running the world. And there are very few companies actually who are. Have stakeholders, two, three companies and three countries, huge stakeholders in this. But because the vices are running it, it's being used that way. But if in three levels, the vices should stop running at three levels, then it can really be used for good. It can be used for finding a cure for cancer and getting Haiti out of poverty and you know, it can do so many things. [00:28:28] Speaker A: Yeah. I recently was curious actually about. I was looking at the English politics and I asked AI if you took over the country of England and you were to actually solve the problems, what most. In the most effective way, what would you do? Because. Because the politicians keep getting, you know, elected and they don't do the stuff that actually works most of the time, right? They don't, no. [00:28:54] Speaker B: Because they are pandering to a base. Right? [00:28:57] Speaker A: Yeah. [00:28:58] Speaker B: But you look at, let's say I have a Muslim base that's voting for me and, and I will do anything that community wants because they are putting me in power. [00:29:14] Speaker A: It. Well, it would. It came up with a list. I was just curious, I'm like, how come people keep hiring all these politicians and they don't do anything? I mean, that's, I mean obviously it's because of the base. But like you'd need to deregulate certain things that too much red tape in a lot of Western countries, wrecking things, build more houses, etc. Etc. That brings the cost of living down. There's, there's a whole bunch of practical things, but they just don't do any of them or hardly any of them. Or they take way, way, way longer than, than it needs to for various complicated reasons. But like so many things could be solved fairly, fairly well if it wasn't for. So AI can actually see more clearly in some cases. But then people have all these conflicting conflicts of interest going on that stop them from doing stuff. So let's get into how to use AI without it ruining your soul. First thing is to understand what it is, which we've already talked about and realize that if you're coming from the various vices within you, then obviously it's going to hurt you. I mean, that's as simple as that. So the, the most important thing is to have a very, very clear intention before you use it. So what I personally try and do is I turn my Internet off and I think about what I'm doing. Because if, if you ask AI what you should do right from the beginning, it's a bad idea because it doesn't know it's not. AI shouldn't be making decisions for us. We're the ones who make decisions. So we have to think, what do I need to do? What do I want to do? What's and what's the output? What main thing I've realized is that if I don't have a clear outcome in my mind, if the vision isn't clear, clear of what the end result is, why am, why am I using this? What am I trying to get from it in the end? Like am I working on a project finish a, like an email that I'm writing that needs cleaning up, you know, research, what is it I'm using it for? If I don't know what the outcome is very, very, very clearly, then it ends up being a waste of time and it, it doesn't, it doesn't know how to figure out the end result and it'll just kind of throw things out there. So from my experience, I've been using this a lot is you need to have clear outcome or clear intention. What finished looks like you need to have a very clear template of some kind. Oh, look at this template of some kind so that it can measure it against the output. Right? Because otherwise it just makes things up if you haven't got certain standards of what you're looking for and you try and use, has all this knowledge, but it's not going to be as good as it could be if you had proper templates, even though it has templates. [00:32:11] Speaker B: That's very important, isn't it? That you have to have a sense of what the final. How it should, what's a good thing or whatever it's creating. [00:32:21] Speaker A: Yeah, so, so what makes the human good and different from AI is taste, right? You need to know what, you need to have good taste. So what is good taste? So if it's a book you're designing or if it's a book you're writing or a project, what is good taste in that context? And so you can tell the AI, go and find me all the best stuff and you can look at it, make sure that it's found you good stuff that you actually agree with and then say this is the end result that I'm going for. And then you give it context, right? So basically the best way to get outputs is clear intention, clear context and clear outcome, right? And if it has those things then it can. And then also you need to tell it how long you've got because this is a mistake I've made is that any one of these screws everything up, right? If you miss any one of these, you'll go, it can go on and on and on. You end up wasting time. So how long you got? So if you've got, if you've got an hour to work on something, then be very clear on how long you've got. Because otherwise you can go on and on and on. And what I personally find is that I can get into a hypnotic trance. And what I mean is that I'm like, give it, I'm like, do this thing. And it gives me an output, right? It produces this massive amount of information, which is far more than I can honestly read in a sane, clear way. Because it might produce a file that's like, it might produce 10 files that are all very long winded documents, right? So I'm looking at this stuff on the screen and I'm like, scan it. And I'm like, it doesn't feel right. Do it again, stress test it, do it again. And then it does it again and again. That's what, that's a hypnotic trance. That's what I try and avoid getting into. I don't know if anyone else experiences this because it just goes round and round and round around. So. So I have to force myself to stop, take a break, and actually read this stuff and think what I'm doing. Because once that iteration loop starts happening, that's a ludic loop, it's a dopamine loop where I'm like chasing the next thing and not getting in. And that is, that has wasted enormous amount of my time. And then afterwards, when I'm not, when I'm fresh again and I'm not using the computer and I look at what it produced, I'm like, what on earth is this junk? Seriously? And then I give it a hard time and I argue with it and it's like, oh, I'm terribly sorry. You're absolutely right. You're absolutely right. I, you know, blah, blah, blah. So, so it's important to take time off so you have a clear goal, you have a clear context. Here's what I'm doing, here's who I am, here's the purpose of it, blah, blah, blah, deep context. Here's the output you want, here's how long I've got to work on this. And then you press the button, see what it is. Take time to have a good look at it, then come back to it with very specific changes if necessary. And then the, the other thing that I've personally found is that it can't. Very rarely is capable of producing perfect outputs by itself. So you can't just cut and paste its output in most cases and just say, that's good to go. Now, most the time, once it's given you something, you're gonna have to turn the computer off as much as possible and actually print it out and look at it and go, does this make any sense? Like, what's wrong with it? And actually change it yourself. And then you might go back to AI for a final polish. But, but what most people don't do, this is what I'm just sharing, is actually very sophisticated. I have very complicated version of this. But that's the essence of it is clear intention, clear context, clear outputs, clear time, block time box. And realize that it's good for research, it's good for the first round. Then you're gonna have to take it all and make it better. And then you can put it back in again and say, just find any errors with this. And my thing I say to it most is stress test, stress test, stress test, right? So if I have an idea, I'm thinking of doing this thing and blah, blah, blah, blah, blah, blah, stress test, my whole idea. Does this make any sense? Play devil's advocate, because that's. It's good at that, actually. So instead of what we don't want is for AI to say, that's the best idea ever. You're a genius. Everyone needs to know this thing, right? That that's what AI wants to do. It wants to please us and make out that we're all geniuses, but we're not actually. So it's better to say, tell me what's wrong with this, Tell me all the problems with this, tell me how this isn't going to work, and then it can come up with a proper outcome. So I personally have this brief document where I have. I force it to make me come up with a brief. What's the brief? What are we doing? Why are we doing it? What's the point of it? How do we know it's done and then move forward, right? And the other thing is, one bad habit that I've got into is that because it takes a long time to do some of these things. Things, right, I can then often go and do another task while I'm waiting for the first one to happen, which seems like it's a good idea, but it's split tasking. So then I'm jumping between one and the other and sometimes I have three of these on the go and I'm. I'm. So I go back to the one that when it's done and I'm like, scan it. No, that's not very good. Do it again. And then the next one just goes. So I don't know if anyone else has got into these bad habits, but these are some of the things. So I kind of feel like it's best to think of it as a. As an assistant who does some of the stuff. And you're very clear what it does good at, what it's good at, and what you're good at. And don't spend hours and hours and hours on it because it doesn't really get the job done. I have a whole document here. I can actually make this available in detail how to do this. But that is the main. The main thing. And keep in mind that you are the final gauge of quality. You're like your taste, your. Your sense of quality. It has to drive everything. That. That's what makes people human. That's. That's what, that's what we have. That AI doesn't have. So clear intention and proper system. And. And you can always ask AI what. What you should do to make things better if necessary. But it often makes things up and comes up with stupid ideas. So be careful. It's a very, very powerful genius. It's like the best example of what it actually is. Because I asked, AI, what are you? If you were like a person, what would you be? What was your personality? It's basically a. It's a savant, autistic savant that has ADHD and has to be really prodded into doing good work. So unfortunately, that's what it is at the moment, which is very, you know, [00:39:43] Speaker B: also I feel that the way you train it, right. I mean, I think it does very, very good research for me. And I don't have to ask it too many times. Let's say I'm writing the series on the Ramayana and I want everything Valmiki ever wrote about Mandodari. Mandodari is Ravana's chief wife. He has many. His chief wife is Mandodari. And if it was me, right? So those yellow books you see out there, you see those yellow books behind? Yeah, those are Ramayana. That's the whole volume. [00:40:33] Speaker A: Yeah. [00:40:34] Speaker B: If it was me, I have to go and get all of those and read it and find out. But the way AI does, right, It'll tell you, okay, this verse, this verse, this verse, and these other 20 or 30 verses while Mickey talks about Mandotari. Done. Really nice to just pull that way. [00:40:56] Speaker A: What a joke. No, it's great for research. It's great for research. [00:40:59] Speaker B: It's very useful that way. [00:41:01] Speaker A: It's pattern recognition. It pulls all the stuff together. So it's like, I think just overall it's good for everyone. Using AI to talk to AI, think about AI, what is it actually good for? From your experience, what is it bad at? And stop trying to get it to do stuff that it's bad at, that you're better at because. And also you're not thinking properly. They did experiments, apparently, where they asked a group of humans to come up with a bunch of ideas, original ideas, and AI to come up with a bunch of ideas. And the humans came up with 100 good, good, fresh, original ideas. And AI only came up with new ideas. 6%. 6%. So it's not good at. It's good at synthesizing, it's good at research, it's good at pattern recognition, but it is not actually that creative. And you need to bring some soul back into it in the final output. Otherwise it's lacking in soul. So this is a huge topic and I'm sure as time goes on, this is going to be one of the biggest, most important things. How do you use AI? [00:42:12] Speaker B: I can Talk forever Michael, on this. [00:42:14] Speaker A: Yeah, yeah, it's a big topic. [00:42:16] Speaker B: How do you use it in a way, so much research on this, right? [00:42:19] Speaker A: Yeah. Well, how do you use it personally, Shirin? What's your flow with it? [00:42:24] Speaker B: I use it for research and also. So let's say I'm writing the series on the Ramayana. I use it a lot for the research. Like I really want to get into the character Ravana. And so I take all of the things Valmiki wrote about Ravana, right? I can't possibly every time go read that stack of books. It would just take me months and months and months. So I just say give me everything Valmiki wrote about Ravana and then I keep reading it to go inside the head of Ravana to see what is going on with the soul with that character. So I do the research but I go in to get what's going on inside the head and it's been very useful that way. But the other part is for the longest time I couldn't really afford good editors but I did pay editors and they weren't very good because they don't really understand your voice, right? I mean they hardly encounter people with white saris who, you know, have contemplative practice for decades. They're not going to understand my voice, right? They have their own voice. And I'll give you an example of Ramayana, right? So after I write the whole thing then I stress test it because I'm writing it in first person. So Ravana chapters will be first person. Ravana. How is Ravana thinking what is happening? Like that, right? Like Sita. How is Sita thinking Sita in first person? So this, there's a thing called swayamvara. Swayamvara is your own. You can choose your own husband. And so in Ramayana what happens is Sita's father Janaka, he invites all of the kings and everyone to come. And the conditions he puts for Sita's marriage is you have to take this bow and string it and all of that, right? So all the kings come, everyone comes. So it's a swayamvara. But the condition is whoever can do that will marry Sita. So she's sitting there the whole day and watching. No one is able to do it. And so after I wrote, right? After I wrote all of the things and wrote first person, what's going on in her mind? And then when she sees Rama, what happens? And then when Rama, you know, you know, does the bow, what happens? You know, like he takes the bow, he strings the bow, and he breaks the bow. What happens? What is it? What is Sita thinking through all of this? Right? And so I wrote the whole thing. And then at one point, Claude, I was using Claude to edit because this is what editors will do. Like, where is the weakness in the chapter? Am I first person Sita shouldn't sound the same as first person Ravana, right? Because the same person is writing it. I can't sound the same way, right? It has to be different. And so that's the kind of help it will tell you. No, you wrote it too much like Ravana, too much like especially Vasishtha. All the chapters are sounding like Vasishta. You need to get off it. Vasishta is the sage. And so, so at one point, at one point, I know this is a very long winded answer. At one point it said, when the Swayamvara is going on because Sita is 16, she should be afraid. And I was like, she's not an American teenager to be afraid. She's an exalted being. Like, you can't just say she should be afraid. But the AI learned very quickly when I said that, when I pushed back and said, no, she's not an American teenager, I'm not going to make her afraid. But if it was a human editor, I cannot train them that way. They have their own frame of mind that they come to which they feel their frame of mind is the same. And so I used to have those problems. Even though paying so much money to human editors, all of that, I would have that problem with them and I would not be happy with the way they edited. And now I use Claude to edit first because I can afford it. Because I can't afford human editors. [00:47:13] Speaker A: Yeah, Claude is great for all sorts of things. And I think the key thing is get offline and look at what you're doing. Because there's a. There's a certain amount of time after which the brain stops functioning properly. And. And we say, well, you know, it says it's fine, all right, good to go. Because we're not thinking. This is what I notice in myself. I'm not thinking. I've actually got a, a device that measures my brain waves. And when I'm doing work myself, I'm in gamma waves, which is very, very, very like fast, creative mindset. Apparently most people don't have much gamma wave. I was quite surprised. But there's massive gamma. You can see it. I haven't tried it with AI, but I'm pretty sure there's not going to be any Gamma waves going on when I'm using AI at all, because I'm just sitting back going, do it better. Do it better. And it doesn't work that way. [00:48:09] Speaker B: But you're thinking, right? Thinking, saying it, do it better. So most of my work is offline. Most of the time I just do the research and I write it, and then at the end I'm doing the editing bit. But most of it is offline. [00:48:24] Speaker A: Exactly. So that you're doing it properly. I think the issue is I have to do iterations and that becomes difficult, so I have to take a break on purpose. Otherwise it gets into this, like, iterative loop, which is very unhealthy and doesn't get. Doesn't get the job done either. So clear, brief, clear idea, what you're doing. Make sure you're the one who knows what you're doing all the way through. As soon as you forget what you're doing and outsource your mastery and your. Your. Your intentions to these tools, then it starts going way off track and producing what's called AI Slop. AI Slop. And there's. [00:48:59] Speaker B: Which is not that much of an improvement. Human slop, at least is. Sounds a little nicer right now. [00:49:07] Speaker A: Human slop, Yeah. I mean, people can't be bothered to produce human slop because it takes longer to do. But you can just go write me an article about this thing. Boom. And then cut and paste and stick it on somewhere and say, okay, that. I wrote that. That's. That's AI Slop right there. So huge topic. Huge topic. Thank you, Sheren, for your wonderful insights and all your research. I'm sure we'll do another session on this, I'm pretty sure, because there's a huge topic, this biggest topic, and hopefully everyone listening to this found something useful out of it. And if you can not use AI at all, good for you, right? If you never need to use it again in your life, great. That's kind of the new way. [00:49:46] Speaker B: There's a bigger problem though, right? People are losing jobs and this and that. There's so much fear around it. [00:49:53] Speaker A: And no, it's a huge issue. And I think ultimately right now we're having a lot of conversation is, should we use AI or not? A lot of people are thinking, how are we going to deal with AI? Blah, blah, blah, blah, blah. We have to jump into the fact that it's not going anywhere. Like electricity. No one's thinking, should we use electricity? Or should we just go back to not using electricity? No one thinks that anymore. It's just like you have to have electricity, you have to have a cell phone, and you have to have a computer and the Internet. If we have to use AI, that's kind of where it's got. So it's. It's. Instead of thinking, let's figure out how not to use it. You just have to think, how am I going to use it in a way that doesn't damage myself? And. Because what other option have you got, really? That's just a pragmatic, real situation. So there we have it. Thank you, Shereen. Blessings. Blessings. Let's have a blessing. [00:50:49] Speaker B: Blessing. You have a number. Determination. [00:50:55] Speaker A: Determination. Yeah. [00:50:59] Speaker B: This is God's blessing for you. Determination. You let the light of your being illuminate who you are and fill you with courage, remaining blissfully disciplined in pursuit of your purpose despite the obstacles. You keep your trust alive, stay tenacious, and are rewarded with joy. [00:51:24] Speaker A: That's more like it. Yay. Thank you. Thank you. Thank you. [00:51:30] Speaker B: Thank you, Michael. [00:51:30] Speaker A: Many blessings. Talk to you soon.