Life on Autopilot Protocol: 16 Tasks AI Can Crush #008

Viktor: 0:16

Welcome to Disruptor Digest, the top disruption business show. We dig up the secret playbooks used by first movers, featuring the latest tools, technologies, and science, ensuring you won't fall behind or succumb to FOMO, to singularity and beyond.

Mihaly: 0:32

Hi Disruptors. Hi Viktor.

Viktor: 0:34

Hello everyone.

Mihaly: 0:35

So Viktor, please tell us about what are we going to talk about in this episode?

Viktor: 0:39

We're gonna share All the tools and prompts which we use in our everyday life every single day. So we're gonna show you exactly how AI is making our lives easier. And I want to share an easy, quick example in my life, even though I have an economic degree, reading Michael J. Burry tweets. And for those who don't know who he is, basically, He was the main guy in the big short movie, so he was one of the guys who made a windfall of money in the 2008 2009 housing prices collapse. So... he

Mihaly: 1:11

predicted it before almost everyone else and he made like tens of millions of dollars, right?

Viktor: 1:17

Yeah, so he made a lot of money on this bet because he understood what's going to happen and the big short movies. It was basically, he was one of the main characters. And he likes to tweet in a cryptic manner where not too much context is given. So Justin, in this case, what you can see on the screen, but also for those who are listening, he has a tweet about how inflation is going to peak again in the future, and he has a quite complicated sentence. So I'm just like, going to read it. It's we are likely to see CPI lower. Possibly negative in 2H2023. So what I did, I just copy pasted the whole tweet and put it into GPT-4 and I just wrote the exact prompt of Explain this in plain English like I am a noob. And I just put it there and it's explaining all the sentences, especially this specific sentence. 2023, it explains like CPI stands for consumer price index, which is a way of measuring inflation. So once again, it's explaining what a CPI stands for. And if CPI goes down or even becomes negative, that means prices are falling, not rising. And this is likely to happen in the second half of 2023. So once again, it's explaining what CPI stands for. And it's also explaining what to age 2023 stands for. And makes my life much easier, because if I'm not completely confident in something, I just copy paste it. I just tell ChatGPT and in this case the GPT-4 model that explain it like I'm a noob in a simple English. Viktor, that's great. So

Mihaly: 2:48

you don't have to search for the abbreviations and also understanding the overall meaning. You can do it in one step, right? You

Viktor: 2:54

can specifically pinpoint the exact word which you don't understand. So you can even ask follow up questions like, okay, explain to me how CPI works and then it explains to you. So whenever you are stuck, it helps you to unstuck instantly. So you just copy paste and in that moment it explains in a way which you can understand. Okay,

Mihaly: 3:17

example

Viktor: 3:18

number two. Okay, with the second example I have a confession. I'm not good with gifting. And GPT-4 just helps me out. And the exact prompt I'm using is give me exhaustive gift ideas to my two year old nephew. That's, a prompt. And I can use it to, what kind of gifts can I give to my mom. My wife, my grandma, and so on or my friends even. This is one of the best prompts which I can use. And I'm using the word exhaustive, because otherwise, if I'm not specifying how many examples I need, It may give just five right or if I'm specifying I need 10, maybe it could give me 32 or 35, which is in this case when I'm using the exhaustive gift ideas, it goes out and collect all the possibilities which I could use and in this exact example, it's providing me 35. It's providing educational toys, artistic gifts, musical toys, active play, pretend play, interactive toys, comfort items, personalized gifts, other toys and items, and subscription boxes. It's quite exhaustive in a sense as well. And it's true to its word that it's really trying to, help me out. It can be bilingual toys, touch and feel books, but also with artistic gifts. It can be crayons and paper, or musical toys like musical books, active play like softballs, push and pull toys, indoor sliding or swing, ride on toys, pretend play, it can be toy kitchen, doctor's kit, and so on, interactive toys. Which one was the most surprising for you? To be honest, the indoor slide or swing, it's not something I would gift otherwise. But since it's already suggested, I remember that he already has a ball pit. So it's a big pit filled with balls so he can jump into it and play inside. So maybe a slide attached to it would be a good gift. So that's something I wouldn't come up with. In hindsight, after the fact, it makes a lot of sense. Great. All right, let's move to the third idea. It's about coaching. Whenever I have an issue with my children, or in my life generally, I just ask questions like, how could I teach my 18 month old son to be more gentle with his little sister? That's like a real question I was struggling with, because normally he's gentle, but sometimes he gets rough. And it answers in a very comprehensive way. And first, it suggests to be to model agenda behavior even me doing some gentle petting on his arm or maybe using a stuffed animal and teaching what gentle is and using simple language like be gentle, touch softly, guiding his hand, brazing gentle behavior. Even reading books, you read books, it can teach gentleness and kindness and obviously intervene when necessary and teach empathy. So when I asked this question, it was helpful because it explicitly said that rough play actually helps to be gentle. And it makes a lot of sense because if we engage in rough and tumble, for example, or we wrestle, or we kick a ball, or we run, And he can just let all the energy out. He has more capacity to be gentle afterward because maybe he's just rough because he has too much energy is trapped inside his body. So just getting these kind of answers right on the spot when I'm facing them, it's immensely helpful. And also I'm working with my coach, Joe Hudson, who coached a few co founders and key engineers at OpenAI. to digitize him, help to ask questions from a bot, which is similar to how Juve would answer. And also there's math. who is a coach of Naval Ravikant and also Sam Altman as well. So he's a quite legendary coach in the Silicon Valley. And I'm not sure whether how many people know about that, but he has, and I'm gonna, I'm sharing it on my screen. It's gonna be in the show notes. So if you're interested, you will find it. So he basically open sourced every single material which he has. So he has. 10 pages of links to materials about how he's approaching different aspects of life and business. And you can read it, obviously, right? So that's useful. And actually, it can be turned into a plugin for ChatGPT. So you can ask questions and it can just basically get back the specific answers. which is used by Matt. So if you're interested in it, if you would like to try this, just write a comment and we can turn this into a plugin. But anyways, you can find it since it's open source, since it's in a Google Docs, you can use this as well to basically improve your life and business. One more example from you. Let's go to the fourth example for me. It's basically about how I found An apartment in tall in Estonia, even though I speak zero Estonian. And my wife was basically sitting next to me and watching what I did, and he was basically telling me, all right, the tools I'm using may be not that good. So whoever is receiving my emails, they may notice that I'm not Estonian, and maybe they're gonna. Make a comment on that and just two minutes after this comment I got a reply and basically someone said that my name sounds like hungarian How come I speak so good estonian? So I just want to share with you the exact workflow I used So I just basically used an Estonian website for finding apartments. I found a good apartment, but obviously the text and the description is in Estonian. So I just copy pasted the text into DeepL. com. I copy pasted and translated it to English. So now I have an English description. I can read it, obviously. But I went one way further. So I just went to GPT-4 again. And I used the exact prompt of down below you find an ad for renting a home, write a message, and I just used bullet points. And I said having a bathtub and dishwasher is a non negotiable. Ask if it's not clear if they have one. Ask how much is the utility cost. Negotiate on the price. And ask when we can check out the apartment. Next week we are flexible. And I just wrote add and semi colon and basically copy pasted what I translated and I also copy pasted the table from the original listing. So it's like about the size, the floor, the total floor, security system, cost and so on and it basically wrote me a message. About okay, I came across this listing for the two room apartment located in Thonlin city center, and I'm extremely interested in it The details of the property have definitely piqued my interest. However, I have a few inquiries and propositions to discuss. So it basically goes through and tries to negotiate on the price for me, and then I'm basically telling it, okay, negotiate on the price, please, and their answer is, and I'm just copy pasting their answer, that's, yes, you can check it out next Tuesday, but we can't move the price. And once again, I just only told it, okay, negotiate on a price, please. And it's trying to negotiate on a price by saying, Thank you for your prompt reply and scheduling the viewing next Tuesday. I understand and respect your position regarding the rent. However, considering the current market and my serious intent to commit long term, I would like to propose a compromise. How about... We meet halfway at 675 euros per month. I believe this revised offer maintains respect for the value of the property by also taking into account the realities of the current rental market and my budget constraints. Looking forward to hearing from you and seeing the apartment next Tuesday. So why this is genius? Because I just only have to write what is my goal. And my goal in this case is to cover all my needs and make sure that everything is covered. So it's it's having a dishwasher, and it's having a bathtub, and also to negotiate on the price irrespective of whether it does it. So I just copy pasting back the email into dpl. com again and sending. The email, and once we were visiting the apartment itself, I always wrote an email like, can we discuss the apartment in person in English? Because writing the emails, I had help, and it would be easier on us if we can speak English. So I will always. clarified it up just before we actually checked out the apartment. So they are not surprised that I'm not speaking Estonian. But in this way, I screened 500 listings. I wrote to 120 and we checked out 50 apartments. And it was done, basically, I was spending like one to two hours each night for a week. And it was insane. So even my wife was surprised how quickly we could

Mihaly: 12:46

proceed. Okay, example number five. Imagine this situation. You are having your friends over, you cook something, you cook a three course menu, and you want to do the perfect wine pairing. In everyday life...

Viktor: 13:01

Sorry, it's like you want to impress them, right? So it's yeah, you take the effort to cook, but you want to take it even one step further, right? You want to have an amazing pairing with the wine, right?

Mihaly: 13:13

Yeah, you just leisurely want to bring out the wine and oh, this is just a perfect match for this dish. Did you know that? So you want to look like without, you didn't even put that much effort. But actually, it's hard to find a good wine pairing. You have to check, you have to search for it. You have to understand the, fundamentals, the theories. But you can skip this step and then we show

Viktor: 13:38

you how. Let me intervene for a second. Understanding wines. It is it can be very daunting, right? Because there are so many different kinds of grapes, so many different kinds of preparation, different even years. So it's like pretty tough job to pick a good wine, which is actually good because you can obviously say that this is a good match, but if it's not a good match, that's, it's a problem, right? Because then the whole thing just backfired, basically. Yeah.

Mihaly: 14:07

Okay. So here is an exact example. I just want to show what I would ask from ChatGPT. So I would ask ChatGPT, please recommend me a wine pairing for a coconut chicken curry in a hot evening. Because, I don't know, weather condition is something I think should be considered as well. But the main point here is I will create curry chicken. And what ChatGPT came up with, it recommended me five kind of wines. For example, Riesling, Rosé, Chanin Blanc. And, okay, it's getting better, something getting useful, but it still, it doesn't have an action item for me. What should I do? What should I buy now? So I asked, okay, please, can you come up with an exact Rosé wines that which is under 30 euros and change. GPT recommended me five rosy wines, and most of them are from France, the province region. But the problem is I live in Thailand, and it's very hard to buy these lines, and usually they are very expensive because this crazy taxes here. So it's still It's not actionable, not useful, but I will show you what is actionable. So I asked for the whole wine list from a local retailer, and they sent me a 20 page long list with several wines and their prices, and I copied all of them. And here comes the interesting part. I put it in Cloud 100k model because the list is very long.

Viktor: 15:33

Basically for those who are listening, what happened here, you got back a PDF, you simply select it all and copy it, right? So select it all and copy. And then went to poe. com, P O E dot com, and select, because since you're subscribed, you're paying 20, you get access to Cloud Instant 100k, which can handle 120 pages. Of text, right? So just copy pasted everything from the pdf and you've wrote the exact prompt of which wine is the best match For a curry chicken out of these ones. It's a simple prompt And you just copy pasted everything and let's see what was the result. Okay,

Mihaly: 16:09

so just to clarify, this long text is too long for ChatGPT, because its context window is too short for this kind of text. So there are several different use cases for ChatGPT and Cloud100k. And I think these are the two models That we use the most so here is a very long Prompt that started with what you already mentioned which one is the best match for a curry chicken out of these ones And this just text and text with all the text from the long

Viktor: 16:41

pages of text Yes

Mihaly: 16:43

And here is Claude's answer. So the answer is, based on the information provided, I would recommend Carta Vieja Cabernet Sauvignon as the batch for curry chicken out of the wine options listed. And it's even provided me reasons that it has bought fruit flavors, which is matching for a chicken. And even it said that the price point... is good. And it's even told me that the price point of this wine at 1600 Baht, but it seems reasonable and accessible for firing with an everyday dish like curry chicken. So I'm at this point totally convinced that it's a good wine based on the characteristics and it's good wine based on the price. So it totally solved my problem, and I can drive to this shop and just buy this

Viktor: 17:27

wine. So basically it's like asking an extremely knowledgeable person, like helping you out, and they're reading through 20 pages of list, and specifically picking one good match just for you. And it would be unreasonable for anyone to do that. for almost like for free, right? And in this case, it's almost free. It's 20 bucks a month is nothing compared to the value, which we get out because it's like solving because otherwise what would you do? You would spend like hours on this task, right? Because you don't want to mess it up. You don't want this whole friendly cooking backfire. You want to have a good impression. Everyone is having a good time. You want the pairing to go perfect and so on and it's actually just an everyday example, but this is a strong example, because even though you are not familiar or not really into wines, you can get a good, suggestion.

Mihaly: 18:23

Okay, next example. And this is for coding for total beginners like me. So for the context I can't code. I have a problem. It's very daunting for me. To understand what is happening in my Google ads account for my for our cooking school, and I want to have a special report for what is happening in the last seven days in the campaigns and in the keywords, and I want to manipulate it. And because I cannot code. Spreadsheet and Excel are the most advanced tools that I can manipulate numbers and do count click through rate and cost per conversion and basic marketing numbers. So here is how I will start the task. I put it on for ChatGPT. Write me a Google ad script to extract campaign data. Time frame, last week. Columns, campaign, impressions, click and spam. So it's like a basic campaign data from Google Ads. And Chanjipiti says, sure, here's a simple Google script and it's like about 40 short lines of code and I guess it's Javascript but I'm not entirely sure, and I don't have to be sure. It just says one comment that replace this one line with your Google

Viktor: 19:45

Sheet URL. For those who are just only listening, what happened here, you wanted to get some raw data out from your Google Ads account, right? And you wanted to get into Google Spreadsheet, and you just defined in an extremely conscious manner that, okay, I want to have this data. And you just specify what you need, and it... I actually wrote you a script which you can just copy paste. It even says the only part where you have to replace the spreadsheet URL. It's because where do you want to get the data basically? You just copy paste the URL there. And you can run the script, right? So this is the what, happened here. Yeah,

Mihaly: 20:23

so actually in this case, I just copied the code, I go to Google Ads Script Editor, and I just paste it here. Yeah, I replaced it with the Google Spreadsheet. And stay with me here, because I just clicked save, and I click run, and that's done. Something started to happen, it says applying. Applying, I can click on the Google spreadsheet and it's campaign name, impressions, clicks and cost and I can see the columns and the numbers and it's done. This is all and sometimes encounter an error. For example, I received an error message, invalid argument. I don't understand what this, but I just put it to ChatGPT. I received an error message and I just copy pasted the error message. And usually what ChatGPT said is, sorry, sorry for this problem. So it apologizes. And after tell me what is the problem and what should we change or sometimes Rewrite the code and in this case actually it wasn't chadgpt who made the mistake It was me because I pasted the url in not the perfect format. So I fixed it and it worked Yeah, insane. Okay, next example. So imagine this situation that you have two little children and the day ends with a little storytelling. But the usual problem is that all the books that you have are very boring for you as a parent because you read it several times and you want to come up with something new something refreshing for yourself, some new story. And also you want it to be engaging and I think ChatGPT is very good here. For example, you can tell chat. gpt to create a story, and you define how old are your children, so it will be age appropriate, and you define how long it should be, for example, five minutes, and you define the main characters, and it's a advanced hack, the main characters can be your children. And you define the location like a medieval castle full of candy and butterfly. So this is some creativity that you can add. And the last and most important part is you define an underlying message. For example, don't hit your brother, either hit you first, or you can put some underlying message like patience is good or some very basic. Human truth. In this case, the prompt would look like this. It's a four lines long prompt. Write me a story for my three and five year old children. Make it about five minutes long. The two main characters should be a three year old prince called Ryan and a five year old princess called Layla. Location, medieval castle full of candy and butterfly. Underlying message. Don't hit your brother, even if he hits you first. And it splits out a five minute long message, when the queen teaches of this prince and princess that they shouldn't hit each other. And and I think it's great, and it's very fast and instant, and you can do it every day.

Viktor: 23:42

You, I guess you can even pair it with mid journal, right? So you can like generate illustration for this exact story. You can turn it into a PDF or something and just read it on an iPad or something.

Mihaly: 23:56

Yeah, you can go advanced and you can make the narration with your voice and we talk about that so you can generate this narration and you can create a slideshow. So if you want to create content out of this story for children, that you can do that as well. But I think this is the easiest use case just to read

Viktor: 24:15

it to your children. I want to yes, and you and it means that in basically improv It means that if someone is saying something you go yes, and then you build on top of it So I just want to share an idea how you can chain Different tools or ideas together to get even more value. So let's say i'm a parent as well And this sounds good. But okay, how do I implement it? What if I don't even have an idea? of what the story should be about, right? And the exact prompt I used for GPT-4 was give me an exhaustive list of things That school is not teaching to children to get ready for life, but it would be useful and practical. It's a short prompt, but it's giving me a list of 20 different things, like financial literacy, critical thinking, emotional intelligence, life skills, like cooking, cleaning, laundry, time management, personal organization, civic engagement entrepreneurship, mental health, mindset education, professional ethic. digital literacy, home repair and maintenance, first aid and basic health care, nutrition and health, conflict resolution and negotiation, cultural competency, self defense, sustainability, personal legal rights, public speaking, coding and programming. And even, this is just a high level of many things. Which are not really taught in schools and should be. And if you just go into just one thing conflict resolution negotiation, it says, okay, it's important to personal and professional relationships. But also, if you just ask maybe a full of questions, what could be, like, in this regard I'm writing 4. 14. Please write me a curriculum for small children. And it's basically told you that for the point 14, which was conflict resolution negotiation, what could be a curriculum, and it's basically writing me a curriculum, and it's identifying emotions, understanding emotion, and these kind of things, and you can just pick basically one thing from this and turn it into a nice story. And that's extremely neat because you can personalize it on the spot. And you can make sure that you are teaching things which school is not teaching to children. Okay,

Mihaly: 26:37

next example. Podcasting. And we have some experience because we have been working on this podcast for about three months. So I would like to show you a few parts of our process. So there are three main parts of the podcast. First, the research and preparation. Second, the recording. And the third one, the post production. For example, in the episode number five, we talked about image generation. This is the exact process that we use. Okay. So first in the research, you want to understand what other podcasters are talking about. What are the main topics? What are the main things that they mentioned? And it's useful. So for example, in this case I looked for the most viewed long content, like more than 20 minutes on YouTube, and I found out an interesting podcast about mid journey. And I put it in Assembly AI that we talked about in a previous episode, and it created the whole transcript, and we were able to ask about What are the most useful parts of this? What are the most actionable parts of this podcast? What can be the more viral and more engaging parts? And when we create the list, what should we talk about in the podcast, we instruc these parts and we create, for examples, use it for now our words. So this is the first step for the research. Second, we record the podcast, and the third part is that we create an introduction for this episode in the first about two minutes. And what we do is we use a similar service called Decipher, and again, we just upload the episode, the audio in this case, and it comes up with a summary and also with very useful chapters. In two three minute long chapters what we talk about And the third step is that we put this on ChatGPT and ask to generate an introduction in Huberman style because we think Andrew Huberman and Huberman labs is a good style of introductions So ChatGPT creates the introduction for us And to put it in practice because we are not professional voice actors or so. So What we do here is we put it to Descript, which is another service, and it generates the introduction with Viktor's voice or my voice. And also in the meantime, it creates some visual, so we can put it in YouTube video. Because it's just a voice generated by AI, so we don't have a video, but it's a caption based video here. So I just start here. Hi, I'm Dr. Mihaly Kertesz, and my co host for this episode is none other than Viktor Tabori. This process is very supported by AI tools, and this is how we come up with our introductions. Okay, next one, travel planning. So the problem with travel planning, I think it's very time consuming. So think about you travel to a new city, and you put in Google what Things should I visit on that city? There are several problems there. The first results in google are sponsored ones, ads, actual ads, and the next organic ones, the next results in the google search is also Influenced by. A lot of money and affiliate and other kind of so it's very hard to find an actual information, actually useful information. But I think ChatGPT is very good because it read several websites about the topic so it can have a general understanding. So in this case I can come up with a prompt like Give me the most popular places in Barcelona, but we can go one step further and I can define how much time will I spend in Barcelona? What is my budget? Or what do you want to see? What kind of attractions I'm interested in? For example, I would start with the problem that I will travel to Barcelona for three days. My preference is see the most beautiful places where I can create astonishing Instagram pictures. I want to try the best foods under 30 euros. I don't want to wait too much in line. Give me three, four hours per day for programs and change your PT. Generates me an itinerary, and it starts with day one morning from 9 to 12 go to part girl, which is, I think, one of the most famous parks in Europe, not just in Barcelona and the lunch go to body Gothic and Pinto, which are very beautiful, astonishing and famous places. And it goes for a three day itinerary. And I can ask about this itinerary. Okay, what will be the overall cost for this trip, excluding hotel and flight? And it will come up with transportation. You have to buy metro tickets. It costs about 11 euros. each and attraction prices. Maybe it's not up to update it, but I can get a general approximation and it says about 250 euros. And then I ask about this attraction, why is it worth to visit Casa Batllo or Casa Mia, and it will give me one paragraph description. I can ask ChatGPT what kind of local experience is, Should I try in Barcelona? For example, try a Catalan cooking class, expo market de la Boqueria. So I can ask, give me 10 of 10 of the beaten path ideas that I should visit when I'm in Barcelona. And for example, it recommended me a monastery an erotic museum and a labyrinth park. So several things that would be hard to find, but maybe I'm interested in. And the last part is, what are the most popular places in Barcelona for the locals? And I think it's pretty hard to search for in Google because most of this content is not in English because it's a spanish speaking country. But Chochipit is good in reading in different languages and I can understand what are the places that locars.

Viktor: 32:51

This is a good point because we're going to go to the Barcelona at the end of this year. So maybe I can ask and even specify what can I do with two small children? So one of them is just a half a year old and the other one is two years old. That's like maybe narrows down the list even further. Which like attractions can I see with a stroller and those kind of questions. So yeah, thank you. This is a good example. Okay,

Mihaly: 33:21

thank you. Next one, learning, learning can be a passive activity, which is not the most efficient. So if you read a Material one time or three time or five times. It's not a good investment on your time, but you can create quizzes your for self. So you can check your knowledge about a whole curriculum. And now we go back to the cloud 100 K model, which you can access on poll that poll dot com for a small amount of subscription. And then again, the interesting benefit of this model is that you can paste 120 pages of context here. And in this case, I posted the whole transcript of one of our previous episodes, and I started the prompt that create a quiz for this podcast episode. Write three questions and give me four possible answers for each question. Mark them with A, B, C, and D. Only one answer is correct. And I put it a very long I don't know, 50, 60 pages long maybe a little bit shorter transcript of our previous episode. And it came up with...

Viktor: 34:36

And for those who are listening, basically, Miju was like, just like scrolling scrolling, scrolling, through pages of text. And it's just suddenly... comes up with an extremely great quiz.

Mihaly: 34:49

Question number one. What is the importance of understanding customer jobs to be done? A. It helps you identify specific benefits to highlight in your marketing. B. It helps you find the right competitors. C. Both A and B and D. It's not that important. Okay, you can come up with 50 questions of these and you can check your knowledge if you know the answer. Okay, and the least and a little bit less ethical part of this that you can do. The same for cheating, so you can put the whole curriculum transcript in, in cloud and you can ask the actual exam questions during the exam and it will come up with the

Viktor: 35:31

perfect answer. So basically what you do is like you go to the chat, you paste the whole text or transcript or curriculum or whatnot and you just say, okay, I'm going to ask questions about this. And they are during the exam, you just go to the bot. Chatbot itself and just ask the question which is on the exam question and this is obviously not the Most ethical ones but you can actually do it ethically as well. So when you're preparing you can use the last year's Exam questions to practice. And if you don't know the exact answer, you just ask the question and it's answer it for you. And then you can compare it to yourself and you can iterate it. So you can instantly get feedback almost like having a tutor 24 7. So you can get better quicker. Okay,

Mihaly: 36:19

the next example, and I think you will like it. Maybe you will be a little bit skeptical, but we will see. So the case here is to make a research page. about supplements or anything health related. So in this case, this is the blueprint of Brian Johnson, who got pretty popular about his longevity methods.

Viktor: 36:43

It's like funny because just, sorry, it's a two minute intervention. Or tangent here. So Brian Johnson, who is Brian Johnson he's a liver king. And he's also the guy who, that yes, actually it is. So he's the liver king. And so that's one Brian Johnson. So who doesn't know who is, who Brian Johnson is? The liver king was basically a guy who is popularizing eating raw meat and raw liver. In the meantime, he was on Roid. So he was on shitloads of steroids and built up this 100 million worth of empire of different supplements and food items. And it was basically built on a lie because people believe that, okay, just eating raw meat can make you buffed and can make you really big. The other Brian Johnson was the guy who built up one of the payment companies. It's actually a Stripe competitor and he sold it for several hundred million dollars. Then he's now he's focusing on longevity. So it's like funny that one of the guy is like a big buffed Liver King and the other guy is almost like a vampire who wants to live forever. Okay, so

Mihaly: 37:54

he listed several supplements And I want to understand two things here. So in my case, I want to understand how strong is the evidence based on this supplement. And also because I preparing for amateur competitions, sport competitions like cycling and cross country skiing. I want to understand what are the, what can be the benefits of these supplements on sport performance. And it can be a very, wrong Google search to understand and read a lot of articles about these. Supplements, but I can cut it short, which I GPT

Viktor: 38:36

what Brian Johnson did is he open sourced his protocol, how he's trying to get young and it's it's not just supplements, but also sleeping was a recovery. Also the different methods he's measuring. Performance and aging. So how different organs are aging in his body basically. And this is a longer stacks and this is like a big summary of what he does and what you can do is just copy paste again, everything, and just ask questions about that and get answers instead of. You having to go to google and do endless google searches. Yeah So

Mihaly: 39:12

I asked lgbt that please create a table first column name of the supplement second column effects on longevity third column effects on sports performance fourth column strengths of scientific evidence and I just pasted this I don't know 20 supplements which is in the first column, which was on the page of brian johnson and Chajipiti said that sure, he's a rough table, and for example, B complex, which is the first in the list, it said effects of longevity may support brain health and reduce risk of age related diseases, but direct impact on longevity is unclear. The third column is effects on sport performance may aim in energy production, though not necessarily improving performance, and the strength of scientific evidence is moderate. And from the list, I can find a few things that have high scientific evidence So if for example if I have a limited budget And I don't want to spend a lot of money on supplements that doesn't have scientific evidence I want to start with the most promising fun one. So in for in this case for example vitamin d and epa And also iodine are the things that are marked as high scientific

Viktor: 40:31

evidence. And also lysine, which is like a basic amino acid. And

Mihaly: 40:35

also for lysine, it said may support muscle recovery. So for me, it would be a perfect start. So if I want to start with the the best one, I would start with lysine, which is a strong scientific evidence and support muscle recovery as well. Again. If you want to really go deep, you can go deep and see scientific papers. And ChatGPT and cloud 100 K is perfect for understanding these papers and summarize these papers. But I think this is the good start if you have limited time and limited budget. Okay, next one. Recipe suggestions. In this case, I had an idea to, I want to create coconut chicken on barbecue, but I wanted to check how should I do it? Are there any interesting points that I should focus on? And I put to ChatGPT that I want to create curry chicken with coconut milk on barbecue. How long should I marinate and how should I cook? What are the minimum ingredients? I don't want to use every kind of ingredients that can enhance the flavor. I just want to make it. Fast and easy. So Chajipiti answered me it needs chicken, curry, coconut, salt, pepper, and vegetable oil. Okay, that's great. And some interesting part, it says even two hours of marination is good enough, but overnight would be give the best flavor. So I know, okay, I can just put it in the fridge for overnight, and it says how should I cook. And again, I wouldn't think about that, but it says that Take the chicken out of the fridge and let it sit in room temperature for about 15 20 minutes before you cook it. And so these small details that can help a lot in the

Viktor: 42:12

preparation. Yeah, because it helps to cook it evenly because it's like the core of it is not cold and that's like the inside which it gives you. Really quickly and it just makes sure that you don't fuck

Mihaly: 42:23

it up and you can enhance it again. So for example, if you want to invite friends over and they complicate things because some of them has lactose intolerance. And they like to eat chicken and beef so they are not vegetarian at least because it will be even more complex But gpt could handle that as well So in this case, I would put the prom that three friends will come over tomorrow I want to create healthy three course menu for them from less than 50 euros The preparation should take no more than one hour One of my friends is lactose intolerant, but all of them like to eat chicken and beef. Give me a three course ideas first No recipes. So first I want just to get a high level overview what should I create, and it suggests me appetizers roasted beef and arugula salad, main course grilled chicken curry with coconut milk. What a lovely choice from Chajipiti. And the third one, dessert, fresh fruit salad with mint and lime drizzle. So okay, you can come up with a specific three course menu for my Neat awesome next one writing an email or writing any kind of copy and you want to use ChatGPT as a grammar fixer grammar checker There are several levels of this use case first one you can paste an email to chat. gpt and ask just for fix grammar and typos. The cool thing about this and you can be sure that your email will be free of grammar mistakes and typos. You don't have to check it and you don't have to re read it again if it change anything important. So this is level one. Fix grammar and typos. And it will create an email. Second one, second level. Fix style and punctuation. Use better word choice. Make it like a native speaker have written it. It will a little bit change, rewrite, reword, but it will sound more natural. And if you are not a native English speaker like me, it will be tremendously helpful and it will enhance your emails. And my favorite prompt about emails, make it shorter. In the previous episode, we talked about that we sent about 1. 6 million emails. And all of the A B tests and experiments directed us in one way, that the shorter the email, the better. So when I write an email, my favorite point is make it shorter. And from a three long paragraphs email, chatGPT created me a short two paragraphs one.

Viktor: 45:04

It's also for those who are listening, maybe from an email which was like two pages long, it's one third of a page is long, an email came out just by making the, using the prompt, make it shorter, so it's extremely effective. And the

Mihaly: 45:20

last thing here is that I don't like when ChatGPT rewrites my entire structure and, arguments and negotiation point and but I would like to get some feedback. So we can use ChatGPT as a coach here. And I can ask, identify any weak arguments in my copy and suggest ways to strengthen them. Don't rewrite the email, just give me ideas. And I think it's very useful. I don't want to go into details, but it shows me. Okay, one example, you're in this email, you're asking for 10% discount based on bulk order without specifying what exactly constitutes bulk for your supplier. And also it gave me solutions. You can also provide historical and industry standard data. So these are great ones. Next example, training plan. So as I mentioned, I'm preparing for amateur competitions in cycling and cross country skiing, and having a coach is sometimes not that accessible and lengthy, and sometimes they ask, for example, personal coaches ask a lot for a detailed training plan, but I want to. Create my own training plan. In this case, I write a prompt that my goal is to compete in a 2. 5 hour long cycling competition exactly one year from now. Currently, I train for 8 hours per week. Please write me a high level training plan for the next 12 months. And just the main point, it says in the first 3 months, base building. Go along and go on. So to which, which is about 60 70% off your maximum heart rate and from months four to six build strength and power. And it shows me example, for example, do a five by five minutes effort in higher intensity and In the last three months before the race, it says increase intensity and risk and do race specificity training. And in the last two months, taper down, which means a little bit decrease your training volume. And I think it's a very great overview. And when I have to Work and change my trainings or I have to miss a training, but I some days I can do more training that I have an overall overarching idea. What should I do in that moment? But also I'm very keen to understand what are the pros, the professional sportsmen do. So I can ask, how would an Olympic cyclist training pack look like in the last 12 months before the Olympic event of road cycling? And I put road cycling because this is the closest in, time that I'm preparing for. So it's two and a half hours or so. And so the main difference is that these athletes train 23, 20, 30 hours per week, but interesting that the periodization is the same. They have base building in the first three months, then they're building strength and power, and also they taper down, so they decrease the volume before the competition. So I can understand that if I want to step up my game, I don't have to focus on better theory, better fundamentals, but I just need to have to train more hours. But Fundamentals are the same. All right,

Viktor: 48:50

so I, I have two more examples and one of them is quite common if you want to follow ai, because what's happening with AI is a lot of research papers are coming out. You maybe don't know what they mean. Maybe you have some holes in your understanding what some fundamental research paper is about. So let me share my screen and I'm gonna explain to those who are just only listening. So in this case, what I got is, The transformer paper, which was attention is all you need, was paper which kickstarted the latest AI revolution. So basically what I did, I used the exact prompt of I want to ask you questions about this research. And I basically just like copy pasted the whole PDF, so just highlighted everything and pasted everything. The neat thing about this prompt for, and I'm using poe. com, Cloud Instant 100k. So what's neat about this prompt is that Since you are stating that you want to ask questions it first summarizes it for you So in this case, it's based on my review here are the main points about this research. So it's like Just like basically summarizing that transformers achieve better performance by being more paralyzable And requiring less training time compared to recurrent neural networks or convoluted neural networks models and just like states how good it's on different translation, things and What kind of state of the art results is? It can get only training for three and a half days and so on and it's using the blow score for Scoring basically translational capabilities And i'm just asking you what is blow put in context what two point improvements means because it was stating that they Improved the previous best model by two blow points and it's basically just explaining that this whole blow point is about how it will be received by humans. So how many times humans would prefer the translation of this new model? And it's quite neat because you can ask for any kind of clarification and you can basically process. Any kind of research paper in much quicker because of what just by the structure of it Because first it summarizes it for you What is it's all about and if you have questions if you don't understand something you can just basically dig deep and I use it pretty much daily this whole clothing stand 100k model with this exact prompt of I want to ask questions about this research. So this is one example. And the second example I want to show is basically searching domains or finding domain names because that's like daunting because without plugins, what GPT four is telling you that, okay, it can brainstorm you some domain names, but it may be not able to check whether domain name is available or not. So I just wrote a quick domain names. Extension for it and it basically creates the capability to check any kind of domains. How can our listener access

Mihaly: 52:03

this plugin?

Viktor: 52:04

When this can come out, you can basically search in a plugin store and you will find, if you search on domain names, you will find it and you can install it. So first you have to enable or go to The settings in a chat. openai. com go to beta features and basically switching on the plugins. And once you switch it on you can go on a GPT-4, you can select that default. You can set GPT-4 with plugins. And in this case. You just can install, go to the plugin store and you can search for domain names. And that basically makes it possible to look up free domain names. And just a word of caution, I spent a lot of money buying domains lately just by building this tool. So please be aware that if you don't have too much money to spend on new domain names, don't use this tool. Okay, so that's a word of caution here. So, I'm using the exact prompt of I'm just very conscious here. I'm telling that disruptordigest. com is our podcast, but the domain name is long. I want to have a URL shortener. So when I share links regarding Disruptor Digest, it takes less character to do and it's basically just like writing out the steps. So first, choose a short domain name, check domain name availability, purchase and set it up and use URL shortener. And it asks me whether I have any kind of idea in mind. And then just Tell it that please help me to find good available domain names. And it's using The plugin of domain names and it's basically checking different kind of domains and it's extremely quick. So the result is coming back. It's checking 10 different domains at once. And it's basically saying that this dig. com, so this app. digest. com is available. Or this store dig.com or this rd.com, this dpd.com. So this striper digest spot.com. So these kind of domains are available, and I guess this dig is quite cute, so it's like only six characters, and it's a.com. So this is like a good. Shortener which we can buy and I already checked it and it's free and we can buy it now.

Mihaly: 54:20

Maybe I will buy some of these domain names as well. Like this digdax. com is something that can be fun. Isn't it, Viktor?

Viktor: 54:28

Yeah, man, it's I, spent way too much money on, buying domain names. For ideas and for projects, which I may want to do in the future, but probably I won't. Yeah, maybe ChatGPT and GPT engineer or these kind of tools will help me to realize more tools. Great, Viktor.

Mihaly: 54:47

I think that's all for today.

Viktor: 54:48

Yeah, that's a wrap. Bye. Bye. Welcome to Disruptor Digest, the top disruption business show. We dig up the secret playbooks used by first movers, featuring the latest tools, technologies, and science, ensuring you won't fall behind or succumb to FOMO, to singularity and

Mihaly: 55:07

beyond. Hi, I'm Dr. Mihaly Kertesz with my longtime collaborator, Viktor Tabori. Welcome to this exciting episode about AI image generation. where we delve into the mysteries and hidden insights to stay ahead of your competition. Today, we'll explore the core of AI image generators, including mid journey, dually and stable diffusion. Misunderstandings about AI image generators are widespread, and we're here to dispel them. We'll reveal how these generators don't just copy and paste, but operate in the latent space. This means they intricately comprehend and combine elements such as style,