High School Dropout to OpenAI Researcher - Gabriel Petersson Interview (Extraordinary)

can barely take universities seriously that don't teach activity as a part of their curriculum.

Gabriel Peterson, a high school dropout from Sweden, who now works as an AI research scientist at OpenAI. The creators of the chat. I always thought I was too dumb.

I met a programmer once and I was so starstruck. I was sleeping on couch pillows that I found in like the common room. Companies just want to make money.

You show them how to make money that you can code and they'll hire you.

I currently work at Sora where we're building... What advice would you give to someone who doesn't know what they want to do? The way I think people learn the fastest is Welcome to Extraordinary, the origin stories behind Extraordinary People. I'm Seijal Wen, the founder of Extraordinary.com, and I'm here with Gabriel Petersen, a

People. I'm Seijal Wen, the founder of Extraordinary.com, and I'm here with Gabriel Petersen, a high school dropout from Sweden who works as an AI research scientist at OpenAI, the creators behind ChatGPT. To be a research scientist, typically you need a PhD, but Gabriel has been able to teach himself mathematics and machine learning using ChatGPT and

now works at the world's top AI company. Gabriel was born in the middle of nowhere in Sweden and now is in San Francisco, California after getting his O1 Extraordinary Ability visa. Gabriel, welcome to Extraordinary. Thank you so much for happy to be here.

Ability visa. Gabriel, welcome to Extraordinary. Thank you so much for happy to be here.

So Gabriel, your story is super fascinating to me. I have a tweet over here.

It says, five years ago, I dropped out of high school in Sweden to join a startup with close to zero experience as an engineer. Today, I'm joining OpenAI as a research scientist to build AGI with Sora. How did you get here from that?

Yeah, it's a long story. I've always been thinking about AI ever since I started reading books like Super Intelligence and Life 3.0. Oh, Max Tagmark? Yeah, Max Tagmark. Dude,

I love that. And both of them happen to be Swedish people as well. And

I was like, okay, there's something here. But I always thought I was too dumb.

I think I was looking into AI, like I didn't really know programming and I was like, Probably there is like a bunch of really smart people out there that I can never compete with and yeah I just ended up working as an engineer for a couple years. So you dropped out of high school. How did that happen?

How did you have the conviction to leave high school when everyone around you from your home country, your hometown was there? I didn't really make the decision. It just

more like happened. I think Yeah, my cousin called me one day and said, hello, I just talked to this person. He's really, really smart. He has this product idea to make like a recommendation system with AI and we should start selling this today.

He's currently in Singapore, like doing research and yeah, we're going to start selling it.

Like we're starting like, yeah, come to Stockholm as fast as possible. And I was like, dude, I have this big party tonight. I'll come tomorrow. He's like, no. So

I just went, took the next bus to Stockholm and I just never returned. So

you dropped out of high school. You went to this startup. What happened? We had

this idea, which was building a product recommendation system for e-commerce stores. And at first, like none of us knew anything about startups at all. We were completely like, okay, what do we do? How do we sell? So the first way I started selling was like calling people up. I started sending cold emails. Didn't work very well. I

started calling people up. Kind of, you know, people were sometimes interested, but you know, I was this random 18 year old. I had no idea what I was doing.

I was non-technical. The way we used to do selling, I used to knock on company doors and I'd bring this like A3, like the big papers. Yeah.

I'd have already since before, like scraped their entire website, trained new product recommendation systems, which is like, you have a product and then you have the recommendations under, like what products do you show to increase sales? So I print their old product recommendations to the left and our new product recommendations to the right. And I made like a hundred of these. Wow. And I have them in like a big folder. And

then I went looking at the doors. Hey, can I talk to the e-commerce manager or CEO? And I'll just show them like, hey, this is your old product recommendations.

or CEO? And I'll just show them like, hey, this is your old product recommendations.

This is your new product recommendations. And then they were always like impressed. They were

like, oh shit, did you do all of this? do this. This is very cool."

But then, you know, immediately they're like, okay, but how do I go from here?

Like, there's so many unknowns. Do not worry. I always brought a script I could paste into their console on their website, which flipped their pro-recommendation with our pro-recommendations. And I was like, yeah, we are ready today. We can just go live

pro-recommendations. And I was like, yeah, we are ready today. We can just go live now. That's crazy. And then they're always like, okay, but how do we know that

now. That's crazy. And then they're always like, okay, but how do we know that we'll make money? And I'm like, do not worry. I have an A-B test set up already in this script. It will track the revenue from people using your PR recommendations and our PR recommendations. So I could just like first meeting, just close them, everything ready from the start. We did all the implementations to it, which would backfire

hugely later because we didn't, we're just thinking, let's just scale. Let's not think about like being easy to scale up. Let's just like, just make sure we get customers, right? With like a bunch of other 17, 18 year olds who dropped out of

right? With like a bunch of other 17, 18 year olds who dropped out of high school. Yeah. Yeah. Oliver, he was a researcher. He was 16 or 17 at

high school. Yeah. Yeah. Oliver, he was a researcher. He was 16 or 17 at that point. And then my cousin was like... And you guys were all in person

that point. And then my cousin was like... And you guys were all in person in like Stockholm, Sweden? Yeah. So I was living in my cousin's dorm room. We were... In college? Yes. So we were... No, we don't have dorm

dorm room. We were... In college? Yes. So we were... No, we don't have dorm rooms. It's more like... It's kind of dorm rooms, but in like normal apartments in Sweden. Okay. Got it. And they're super tiny. Yep. And... you know, you can only

Sweden. Okay. Got it. And they're super tiny. Yep. And... you know, you can only live there if you go to the university. But you know, we had to submit things like, oh yeah, he's still doing university, right? And I was sleeping on couch pillows that I found in like the common room for one year. Nice. It was

a disgusting room, but it worked well. And we're sitting in this like co-working space.

What made you keep going? Like most people kind of quit, but you and they would probably go back to school, but you just, you never went back. Like, why

did you keep going? Why did you keep like living in a shared dorm room on these like community couches. I think I've always had a very distorted view of reality. Like I was 100% sure that this would make me a billionaire. 100%! There

reality. Like I was 100% sure that this would make me a billionaire. 100%! There

was like no doubt in the world. And I was like super serious and like acting just like I believed. Like, okay, this is going to be the next big thing. Like nothing else mattered. I was like, I'm just gonna, you know, I was

thing. Like nothing else mattered. I was like, I'm just gonna, you know, I was working like all nighter after all nighter, you know, I was traveling around Stockholm trying to do sales. Yeah. We're doing like all these like craziest things to try to get customers. So you originally dropped out without knowing how to code. How did you

get customers. So you originally dropped out without knowing how to code. How did you go about learning that? Mostly because I was forced to when we had to do the integrations. Yeah, like how did you learn? I guess you had some... friends around

the integrations. Yeah, like how did you learn? I guess you had some... friends around

you who knew how to code better than you? Yeah, but then, so the first way I learned how to code was my cousin when I was very young. Like

at this point I was like 13 or something, but he showed me Java and I made a super simple like Pokemon clone, like turn-based, you can like take damage, such a bad application. And then some time went by and then I made like a Udemy Python course. I learned super simple Python. I made this like really ass game where like You had lily pods coming and you were like a duck trying

to avoid them. It's really dumb. I also did, I tried to get into machine learning. I did all these like, you know, the classic like machine learning course by

learning. I did all these like, you know, the classic like machine learning course by Andrew and G. Yeah, I just thought like, yeah, I'm probably too dumb for this.

I just can't do this stuff. But yeah, when I really started getting into coding was at the PICT, where we had to build things. We had to make product recommendation systems, scraping, integrations, set up EB testing and all these things. Yeah, but how do you learn that if you don't go to school? The good thing with just working, is that you always have a real problem, which makes everything so simple. Like

everyone always says, like, if you don't go to school, how can you learn? And

I'm like, well, it's so much more easy. Like, then you have a real problem.

And you know, you can map out, okay, I want to integrate my part of recommendation system to this e-commerce store. To do that, I need to figure out how to select the elements on the webpage. I need to insert them correctly. I need

to learn how to do all these things. And then you can take it step by step. You go to Stack Overflow and, you know, you can ask your friends

by step. You go to Stack Overflow and, you know, you can ask your friends if you're stuck. And yeah, I think that's like a simpler way to learning. And

especially when you have all this pressure on you, right? If you have a real job, you have pressure and that's everything. Like I could never learn anything without pressure.

There's just no way. Like if someone were like, oh yeah, learn this thing, but you have infinite time and you'll also not make money from it. If you were to give advice to another high school dropout, what would it be so that they would learn more? I think I was extremely lucky. I mean, I was living in this town called Vagirid in the middle of nowhere in Sweden. I knew no

engineers. I met a programmer once in early high school and I was so starstruck.

engineers. I met a programmer once in early high school and I was so starstruck.

I was like, do you code? Do you like make webpages? That's awesome. And when

you don't have this like culture, like why is SF such a capital of startups?

Well, because everyone's only talking about startups. It's like so clear how to do one.

But if you're like in the middle of nowhere and you're not like surrounded by people, this is like all they talk about, you'll think all these things are impossible.

Doing all these things for me, I was like, damn, this seems so far away.

And I was very lucky to have the PICT. The PICT was the first thing where I was like, oh, this is a real thing. I mean, I had no options and I probably it would be very hard to have options because I didn't know what I was looking for. It just came up and it happened to be an extremely good learning for me. For other people who want to do the same

thing, like getting into the market as fast as possible, solving real problems, having accountability.

I mean, now with the help of ChatGPT, you don't even need to know. You

don't even need to have much knowledge about the thing you're doing. If you can just prove to the person that, yeah, I'm good at asking ChatGPT what I need to know. Like I'm super creative. I'm super high agency. You know, you show all

to know. Like I'm super creative. I'm super high agency. You know, you show all these things to the person and hiring. And then the last thing is, oh, but you don't know the actual thing. And you'd be like, yeah, yeah, I talk to Chachupity all the time. Like I'm really good at like extracting information. Like you have all knowledge in the world there. Knowledge is not a problem anymore. In the same

way that it used to be. You don't have to like go to an institution and then read up on something as like a prerequisite course for some potential solution or some potential application. You can now just go into real world find problems like, oh, how do I like optimize this? Or how do I teach people faster or whatever problem you want to solve? And then you can query AI like chat GPT

to figure out how you can solve it and how you can learn the different pieces of knowledge to solve it. The way I think people learn the fastest is by what you would call like a like a top down approach, right? You probably

learn faster if you start with a problem. And then you can read about everything required to start solving the problem. And then you find more problems and you read about those. And then you go down to like the corridor problem, right? So you

about those. And then you go down to like the corridor problem, right? So you

start with actual task and you go down. But that's extremely rare way to learn.

Like in school, everyone has this mindset, right? Of like, okay, we need to start with the foundations. We need to start like, If you want to work with machine learning, like you can forget about doing any machine learning for the first like four years, right? It's like math and then you have matrix modifications, linear algebra, and you

years, right? It's like math and then you have matrix modifications, linear algebra, and you have all these things that build up. And then you have the simpler ML that's like super outdated. You have like, you know, linear regression, all these things that are still used partly, but it's like, it will take you a very long time until you get to like production grade ML. Why is this? Well, it's extremely hard to

scale the top down approach. Because that requires like a teacher always being there for you. It requires you being able to know exactly what piece of thing you need

you. It requires you being able to know exactly what piece of thing you need to learn at any point of time. Well, if you do bottom up, you know, okay, first you always learn this and then you always learn this. And it's much easier to scale, but it's extremely inefficient. And now with ChatGPT, all these changes. Like,

this will change. People say education will change all the time, but I can barely take universities seriously that don't teach at GBT as a part of their curriculum. It's

like actually insane that this is not like a course that's taught from like two years old. Like suddenly foundational knowledge, universities don't have like a monopoly on foundational

years old. Like suddenly foundational knowledge, universities don't have like a monopoly on foundational knowledge anymore. You can just get any foundation knowledge from ChatGPT. And people haven't really

knowledge anymore. You can just get any foundation knowledge from ChatGPT. And people haven't really internalized how top-down problem-selling works. They will always tell you things, you know, like, oh, but you'll never actually understand the problem. You'll never actually blah, blah, blah. And this

is not true. You start with a problem, you recursively go down. Like, if I want to learn machine learning, I ask ChatGPT, okay, what project should I do? Write

a project for me. I have bugs. I start fixing the bugs and then things work. And from there, um, I start with a specific part of the machine learning

work. And from there, um, I start with a specific part of the machine learning problem. Like, okay, what happens here? Can you explain to me the intuition why this

problem. Like, okay, what happens here? Can you explain to me the intuition why this module here makes the model learn? And it will explain to you. And then it said, oh, it uses matrix multiplication and linear algebra. You know, okay, how do they work? What's the math intuition behind this? Like, show me, like make up a couple

work? What's the math intuition behind this? Like, show me, like make up a couple graphs to really make me get an intuition for this part of ML. And then

suddenly you have all the foundational knowledge. Like, it doesn't need to go bottom-up anymore.

Yeah. And this shift will... Yeah, I think this shift will like fundamentally change how education is done. What are schools not teaching you about AI? First of all, the perception of AI is completely wrong in schools. SheaChiButi came, naturally students were like, oh nice, something can do all the work for me. And that's all they thought about, which makes sense. That's the first thing I would apply to as well. And the

first thing that teachers think about is, Oh no, everyone will just use AI to do works. We need to ban AI. AI is bad. And that becomes like a

do works. We need to ban AI. AI is bad. And that becomes like a reinforcing circle of like students' perception of AI is like, okay, I can use this to cheat. And teachers' perception is like, okay, teachers just use this to cheat. Like

to cheat. And teachers' perception is like, okay, teachers just use this to cheat. Like

it's really hard to build up an intuition of how to learn from AI. It

doesn't come very naturally. Now it's like, I'm extremely happy when I talk to like my friends back in Sweden, they go to university and they're like, oh, I realized I can use ChatGPT to give me quizzes. I give it all the previous questions and I ask it like, okay, tell me some fundamental things that all of these different questions share so I can really learn what they try to teach me. Or

like, generate 10 new questions, right? And people are starting to learn how to use AI. The teachers are still very, very anti-AI, which makes no sense. If the teachers

AI. The teachers are still very, very anti-AI, which makes no sense. If the teachers just switch the narrative to, okay, here's how you learn efficiently. If a student, want to cheat at tests. I mean, find ways to do that either way. And if

you're never taught that you can actually use this to learn things, I mean, I'd also use this to cheat. Yeah, yeah, there's just no concept of that. So how

do you use AI to learn? How did you use AI to self-teach yourself math and machine learning to now work at OpenAI? I did a very similar thing to what I was describing before. So I currently work at Sora where we're building these video models at OpenAI. And I wanted to learn things like, you know, the basics of image models. So I asked JPC, hey, what are the most fundamental concepts of

like video and image models in AI? And they started talking about, okay, we had these things called autoencoders, these things called diffusion models. And I was like, yeah, that sounds interesting. I've heard about this everywhere. That's very cool. Now, you know, write all

sounds interesting. I've heard about this everywhere. That's very cool. Now, you know, write all the code for diffusion models. It writes all the code and I have no idea what's going on. Here's a bunch of code, holy shit. Then you try to get it working, you debug it together, you tell it what's wrong, and then you start to build up an intuition of like, okay, this happens here, this happens here, this

happens here. Then you continue to just understand in detail what every single line of

happens here. Then you continue to just understand in detail what every single line of code does. What does this part do? What does this part do? For example, for

code does. What does this part do? What does this part do? For example, for the Fusion model, for example, you can take a part called a ResNet, for example. It was ResNet blocks. They do a bunch of

example. It was ResNet blocks. They do a bunch of transformation and then they also have a residual, which is basically like you let data pass through in a certain way, which makes the model learn more easy. At the

start, I have no idea how this is done. You start asking, HTTP follow-ups, follow-ups, and it will tell you something like what I just told you, but you still have a huge question mark. Like, why is this? What does this mean? What do

you mean it learns more efficiently? What do you do then? Well, you follow up and you'll be like, well, How does it learn more efficiently because it's doing this?

You'd be like, oh yeah, the gradients can flow in these XYZ different ways.

And in the scenario that you wouldn't do this thing, they would be stopped at XYZ things, right? And you just continue to ask the model constantly until you really understand. And when you understand, you can just tell the model, okay, this is my

understand. And when you understand, you can just tell the model, okay, this is my understanding of this. Is this completely correct? And then you'll also start learning about all these small tricks you can do, right? Explain this concept like I'm 12 years old.

That one is really good. It will, you know, it will start like super easily.

Like imagine you're in a bookstore and you can imagine the embeddings being the different books in the store. And then you can imagine, you know, all this, and it will connect everything that has to do with AI to like real world concepts, which makes it really easy to reason about for someone like me. So it sounds like any sort of topic you can learn now and all you need is ChatGPT. And

you start with just asking, hey, what are the preliminary things I need to understand about this? And then you might pull on one of those threads. For when you're

about this? And then you might pull on one of those threads. For when you're investigating video models, you're like, OK, image generation models or diffusion models, like stable diffusion.

And you're like, OK, how the frick does a diffusion model work? And then you would have it explain it to you, maybe generate code samples. But every aspect of that, you would then inquire further. I don't understand this part. What is this? Why

is this adopted to this model architecture? Oh, why is this done this way? Okay,

how does that math work? And I mean, I read like your posts on X.com, which are very popular. And it looks like you're just able to use AI in a way where you continuously query until you have full understanding. And then when you do have full understanding, you almost like re-explain similar to how Feynman, the best way

to learn is to explain things, but now you can do it with AI. And

so when you're learning about diffusion models, after going through a deep dive on some very technical topic where you might not even know what gradients are, and then it'll explain you calculus or some linear algebra and you pick that up. But then you would explain it back to the model and then it will then clarify or see different aspects that you don't understand and you keep repeating that until you have a

very strong grasp. I see it a bit as recursive gap filling if I would summarize it. in one word. It's like, you need, like the skill that's required here

summarize it. in one word. It's like, you need, like the skill that's required here is knowing what gaps you have in your knowledge. Like see you have an AI model or like whatever else you want to learn and understanding when you don't really understand the part. It's actually pretty hard to do. Like it's something you need to train up and practice on yourself. Like, wait a second, do I really understand this

part? And then, so that's one signal you need. The second signal is when you

part? And then, so that's one signal you need. The second signal is when you start asking questions, you need to have a real strong signal for when it clicks, when you're like, ah, there it clicked. I can just understand fundamentally why this thing is as it is. How would someone else learn how to learn with AI? This

is a very good question. I mean, first of all, just change the misconception of AI being used to do the work for you to instead use the AI to explicitly help you learn. you don't just use it to get work done, you actually learn from it. I mean, the moment you just switch that mindset, which seems still fairly uncommon, but it's becoming more and more common all the time, you have

most of the things to get there, right? And then to become really good, first of all, like I said, like know when you have gaps in your knowledge, understand what it feels like when you fundamentally grasp something. And you know, you constantly come up with all these hacks, like you'll notice ChatGPT will respond in a fairly standard way, and your way of learning is probably not exactly what it

responds like because it wants to make sure everyone has a good experience. But you

probably want it to respond in another way. I very often tell it, for example, be extremely direct and concrete. Always show me all the intermediate states and the shapes of the code you produce to make sure I have a really intuitive understanding of why it happens. And if you're unsure, make sure you show me options and like what others have tried and why this works and why something else didn't work. And

you start becoming good at like asking these questions that give you the aha moment.

Like as fast as possible, you want to get to the aha moment. Yeah. Like

the first time you understood linear algebra or the first time I understood what backpropagation works, you probably had a very clear like, oh, wow, it finally clicked. and to

chase these clicks and to make them appear like as frequent as possible, right? That's

like kind of your utility function. That's crazy. It's like in modern day, in order to stay competitive and to be like top performing, when you look at someone, like honestly, they'll be at the top of the field pretty quickly, like how you've done it, just by the rate of being able to query for information. Yeah. And that's

probably like the most important skill now. What do you say? Yeah. And building up this, this is another very important, build up the moment you have a question in your head, make sure to get it into ChatGPT. This one is very hard. This

took me, I remember my cousin, the same cousin I started a company with, he was like, dude, ChatGPT is out. This was like pre, what's it called? I didn't think of this back in the day, it was just like a playground. It was like super early in GPT-3. like before ChatGPT days.

And he was like, why are you not using this yet? Like you're writing code all the time. Like, yeah, I'm going to try it out. And, you know, he kept on pushing me every month. And it took me like a year until I really started connecting like, oh, I have this problem. I need to ask ChatGPT. And

it's so common. Like you meet people all the time. You're in a discussion and people have all these questions or you're co-working with someone. They have all these questions, you know. You're like, you should ask ChatGPT. It's like every time you have any

you know. You're like, you should ask ChatGPT. It's like every time you have any kind of question, any time you need to guess about something, just constantly ask ChatGPT.

Like it's always there. It's very low effort. Like make sure you have a very simple way to just ask ChatGPT about anything you ever wonder and you'll just, you know, have all knowledge in the world. Yeah, but the important part is like almost getting hooked on like how fast you can get to that aha moment of realizing or internalizing something and the skill of being able to prompt ChatGPT, not in a

generic way, but in a way where it will give you very concrete or, you know, different analogies or whatever form factor that works best with your learning style for you to then understand and internalize that. Which is really hard. Or I mean, I think I'm pretty dumb. So it's like sometimes when I ask you GPT stuff and it explains it and I'm like, I don't understand. I don't understand. It's just too

hard. And you try again, you try again and you're like, okay. And then you

hard. And you try again, you try again and you're like, okay. And then you really grab, you're like, okay. Prompting skills time, right? And you're like, okay, what if this these features in the world didn't exist, and what if that never existed?

Would they still have invented this thing and explain it to me like I'm 12?

Generate graphs showing the distributions that I need to know to really understand this. There's

so many creative ways you can use to really extract the information you need from ChatGPT. I think a lot of the things I've learned, especially with previous models, ChatGPT

ChatGPT. I think a lot of the things I've learned, especially with previous models, ChatGPT is becoming so much better all the time. But a year ago when the models weren't as strong, some of the things I've learned is I probably couldn't learn them if I didn't like really know how to extract the information. Like I could ask the question a thousand times and make it rephrase it a thousand times. I just

wouldn't understand. This is why they teach Chachi Bti from elementary school. This is like, you know, a new language. This is like, you know, you still need all the other things in life, like creativity, you know, agency and like all these other things, but like knowledge is a completely new era. Like you can't compare this with anything else. A very concrete example of this, because people that doesn't seem to realize how

else. A very concrete example of this, because people that doesn't seem to realize how like AI, how abruptly this will change the world. Like currently I'm doing a job which traditionally everyone would agree you need like a PhD for, right? There's a bunch of people who have done it without a PhD, but like if you told someone like five years ago, like, oh yeah, it's one of the top AI labs, someone

will be hired who hasn't, you know, really done the thing for a while. And

the only thing he had was that he had was like, he'd done like all these very cool things on other areas, but they didn't know anything about this thing.

People are like, no, that's not possible, right? But we are now in a scenario where I can do the job, traditionally only, you know, done what people have done it for like multiple years, just by using ChatGPT. That's insane. Like the amount, how fast the world will develop with ChatGPT. Like you can just do research in anything you want. If you want to start doing bio research, you want to start doing

you want. If you want to start doing bio research, you want to start doing like hardware, you can just go and do things. It's just incredible. Yeah. This will

be like a double digit increase in world GDP, like just coming from large language models. And anyone can do it. I want to do it. they know how to

models. And anyone can do it. I want to do it. they know how to use ChatGPT. Yeah, it's 20 bucks per month. And you know, the thinking models are

use ChatGPT. Yeah, it's 20 bucks per month. And you know, the thinking models are like really, really good when it comes to like coding and like understanding things. How

do you use ChatGPT to learn as you build out like one of the world's best video models? It's very simple. Like a lot of people ask me this and they're always confused. Like, okay, what do you actually do? Like, what do you do?

And you know, they imagine, I don't know what they imagine, but you know, something very, very special, right? And it's fairly simple. You know, you look at the video and like oh this part of the video doesn't look very good so you go and you change the architecture in the model of it or you change the data or something and you know you you train the model you look at the

results you stare at videos for a while and you're like oh these videos were better that's great this this goes into to master and then you just do that on a loop right you're like okay what's the next thing that i want to fix or the next thing i want to try and that's where like ai is really good right because it's like oh i have this specific problem. Hello, AI. You

know, here's my entire code base. Tell me 10 ideas of what I can do to improve this, right? It'll tell you a bunch of ideas. It will refer to papers you can read. It will do all these, like, really great things. And it

will give you, like, a bunch of ideas that are really good to brainstorm with.

And, you know, you can bring all these ideas to your colleagues and, like, talk to them who are just, like, extremely good. And yeah, it's, like, fairly straightforward. Wow.

How does it find, like, other research papers to suggest you to explore? It just

knows about that. It's like 4.0 just like I think even earlier models, you just print out the links and you press the link. It would just work? Yeah. Because

the model just memorized the links. Yeah, it knows the links. To look to big papers, like probably not to smaller papers, but you can also use the search function.

You can just turn on the search while you're speaking to it and be like, yeah, find some papers talking about this. Obviously, I don't read the papers word for word. I have my instructions like, okay, I want to, you know, give me a list of things this paper did differently. Because oftentimes the paper, they take some technique that I already know about and they introduce some new things to it.

And I just ask it, okay, compared to the other thing, tell me a list and be extremely concrete of exactly what they did that compares to the previous thing.

And that's a really good summarization. Often at times, you're like, ah, this paper probably wouldn't make it. It's not worth trying out and you can just go to the next one. Or you're like, ah, this paper is really good. I only read the

next one. Or you're like, ah, this paper is really good. I only read the paper in depth if I actually decide to implement it. Then I probably will read it if I have bugs. I'm probably just throwing all my code and be like, hey, implement this into my code. I just copy paste it in. Oh, wow. I

obviously make sure to really read through the code. I think it's extremely important. You

can't just throw in code. You can't just write code. No. I'm not the vibe color. I'm very opinionated when it comes to code. No, it sounds like you have

color. I'm very opinionated when it comes to code. No, it sounds like you have a very different approach where if you actually want to build you know, really concrete things, like you need to understand everything, right? Yeah. Because especially if you're pushing like the forefront of any field. Yeah. I mean, I want to understand all the foundations.

I think the first reaction people have is like, oh, you just want to take shortcuts. You don't really want to understand things. You just think you can slope out

shortcuts. You don't really want to understand things. You just think you can slope out as much of AI slope, right? And I think this is not correct at all.

Like I want to take shortcuts, that's for sure. But I want to take shortcuts to understand all the foundations. Yeah. And that's very important. It seems like either you're on the camp, like, okay, AI slope, do all the work for me, I never want to work. Or you're in the camp, you need to go to college, they have a monopoly on all the foundational knowledge, you need to have this taught by

a professor. And I'm probably somewhere in between, right? I mean, you need all these

a professor. And I'm probably somewhere in between, right? I mean, you need all these things. And you know, AI is great. Like, you should use it for everything. You

things. And you know, AI is great. Like, you should use it for everything. You

should use it to understand everything. Yeah. And train its like human AI symbiosis in terms of like, just enhancing your brain and enhancing your ability. So you're

in Stockholm, you left your first startup and how did you find your way to San Francisco? What did you do? Yeah, I always wanted to continue to work in

San Francisco? What did you do? Yeah, I always wanted to continue to work in startups and I always had my sights on San Francisco because you know, all the best people I knew had moved here. All the best companies people were talking about were here. And I noticed that like probably just like super optimized for

were here. And I noticed that like probably just like super optimized for learning as fast as possible. This was sadly pre-ChatGPT. Just imagine where I would be now if I had ChatGPT when starting to learn things. A billionaire. Yeah. And back

then, the best thing you could do was to work with the very best people.

So that's what I tried to do. And so how do you work with the best people? Well, you talk to as many companies as possible. You make sure you

best people? Well, you talk to as many companies as possible. You make sure you interview the person interviewing you, right? Like, what have you done? Do you guys do pull requests? Do you make sure to really review my PR so I actually know

pull requests? Do you make sure to really review my PR so I actually know what mistakes I do? And I managed to join a couple of companies with like really, really talented engineers. I also made sure to be like, I generally worked as a contractor. Like the biggest mistakes people do is that they stay with the same company for way too long, early in their careers. That's like by far

the biggest mistake I've seen people's careers. So it sounds like a year after dropping out of high school, you know, after your first company, you just kept finding the best teams or the best engineers or the best people that you thought, you know, you could work with. And you work with them for a bit, you learn what it could, and then you kept finding new opportunities or like better teams and you...

Yeah, I used to only take contractor roles to make sure that like, that I could be very mobile in the places I work with. You try to find the best places to work with, with the best people. You try to work as closely with them as possible. Make sure you're opinionated about what you're working with. So you

don't only get to do like the tasks no one else wants to do, because then you're not learning. Make sure you really show appreciation for the people reviewing your code, because that's the best source of knowledge you have. Getting feedback in general and hunt feedback. I mean, tell people, hey, I really liked your review. Can you just

hunt feedback. I mean, tell people, hey, I really liked your review. Can you just review every single feedback of mine? People would be shocked. Like, oh, wow. I'd never

heard someone liking feedback before. This is, yeah, of course. Yeah, it's rare because people usually shy away because they already did their schooling and now they're working. But as

a young person with no really accolades, you're like a middle, you're nobody with no degree. Yeah. The way for you to learn is like join the best teams and

degree. Yeah. The way for you to learn is like join the best teams and then be very nimble, but also just relentlessly seek feedback it seems. Yeah, and you know, call them up. Call them up and be like, hey, that was a great review. Now let's go through all the comments together on a call, right? You learn

review. Now let's go through all the comments together on a call, right? You learn

so much and you just like ask follow-up questions like what's the issue? Like there's

like becoming a really good engineer is extremely hard. It's like such a wide area.

There are so many first principles, things and intuitions you need to understand. And they're

pretty easy while you know them, but they can be very hard to learn. And

to have someone just straight up tell them to you and you being good at picking them up, it's like such a- In the same way that, whereas long before, you know, you only can do that from maybe like an existing engineer or maybe a teacher, but now you also have AI who can do that. Now you can do this on demand at any company. You can just start your own company. It

could be like 4 a.m. and you've been up coding or making something or like writing a paper or researching something and you can still ask AI for feedback. And

they explain to you why is this a better decision. I do this all the time. I think like when you've found something in life that works really well, you

time. I think like when you've found something in life that works really well, you should exploit it to the maximum. Ask ChatGPT 100 questions per day.

I always have tabs open with ChatGPT. I write code, throw it in there, and I'm like, is this good? Is this good? Is this good? Are there any bugs?

What can I do better? Why not? I mean, it'll probably tell you, it looks fine right now, but sometimes, oh yeah, there is a bug, or oh yeah, you can do it in this way, so that's simpler. You just constantly learn. Yeah. If

you're doing it literally 100 times a day, That's like a hundred well thought out questions or follow up questions. And you're just able to outpace 99.9% of people in the world as a high school dropout. Yeah. And it should be added.

I think there's still so much valuable advice to be had from humans. There's still,

when it comes to opinions and things, like if you imagine about how models are trained, they train all data on the internet and there's a bunch of different opinions.

And you know, sometimes you might have weird opinions, right? I mean, there's still a lot of value in working with like the really best people. But you can get like 95% of that now, which is the, yeah. So work with the best people, get feedback from them, but also constantly query AI wherever you go to build very deep understandings of any problem you want to solve and any concept you want to

learn. You were learning from these senior engineers, you're contracting at different companies, like different

learn. You were learning from these senior engineers, you're contracting at different companies, like different YC companies or like different, all these different opportunities. How did you, end up coming to America if you didn't go to school, if you didn't have a Pysco diploma. Yeah, it started out with joining a company called DataLand.

We're doing kind of like a Airtable, but way more performant and like developer first and yeah, a scalable Airtable, you could say. And that was a very important decision. I was working there with an engineer who's extremely talented

important decision. I was working there with an engineer who's extremely talented and he just loved teach people. And he'd love having perfect code, which is perfect for me because I write code and he'd just do like 100 comments per PR.

Wow. And I'd call him anytime and be like, hey, what do you think like this? And he would be really good at explaining the first principles reasons for why

this? And he would be really good at explaining the first principles reasons for why some code was written in a certain way. And at some point, I was working remotely from Sweden and they were in New York and I was like, yeah, I really want to go to the US. And I think this is where I first wanted to go to the US. This ended up not happening because the company pivoted

and... to something else, so I decided to leave. That was my first. I started

and... to something else, so I decided to leave. That was my first. I started

a process there, something called the J-1 visa, which is more like, you could say, an internship visa. Because we were all pretty short, like, yeah, I can't get an O-1 visa. There's no way. And it's, you know, you didn't need to win the

O-1 visa. There's no way. And it's, you know, you didn't need to win the Nobel Prize, or you need all these random things. I was like, yeah, there's no way. How could I do this? And then I ended up spending a lot of

way. How could I do this? And then I ended up spending a lot of time trying to figure out what I wanted to do next. And this is when I went to San Francisco, and I was here on the ESTA visa for couple of months just talking to people, trying to figure out like, okay, what people do here? Like what's cool is companies? And there I ended up joining Me Journey. After

here? Like what's cool is companies? And there I ended up joining Me Journey. After

joining Me Journey, I was like, yeah, okay, now maybe we can do OWAN. And

turns out that OWAN visa, there's like so many creative ways you can get an OWAN visa, very many creative ways. For example, one thing we used for my OWAN visa was my Stack Overflow posts. I remember my cousin telling me, oh, you're wasting your time answering a bunch of Stack Overflow questions. And I was like, you don't know, maybe it's helpful at some point. And turns out Stack Overflow posts can be

counted as, yeah, here we have it even. Here's my post about it. So you

can use Stack Overflow posts to get the academic publishing criteria for your 01. Which

is legitimate. Like I have like millions and millions of views. A lot of like peers will review your posts. They're very strict. They will downvote and remove anything that's not true. And if you get upvotes, you're helping a bunch of people, which was

not true. And if you get upvotes, you're helping a bunch of people, which was the criteria. Like have you helped people? I think like with GitHub or like Stack

the criteria. Like have you helped people? I think like with GitHub or like Stack Overflow, it's definitely a very... creative way to Archefury 01. And so how did you, Midjourney is one of the biggest and best, is the best AI image generation company. How did you end up working there? Yeah, it was kind

of interesting. I mean, it's extremely hard to deterministically go somewhere in your career. And

of interesting. I mean, it's extremely hard to deterministically go somewhere in your career. And

It's kind of what you want to do. It's very cliche what everyone is saying, right? But it's like, you want to have a bunch of small chances everywhere, right?

right? But it's like, you want to have a bunch of small chances everywhere, right?

You just want to go wide. You want to post things that you've done. You

want to make sure you have really good demos. You want to reach out to as many people as possible. Go to events and ask people for intros and make sure they do the intro at the event. Really for instance, like, okay, oh, you want to intro me? Well, yeah, let's do it right now. Make a group with us. A bias towards action so that it actually happens. You make sure the results.

us. A bias towards action so that it actually happens. You make sure the results.

It is a very high agency move. And also be very clear with how you can give them value and make sure they understand that you're not a nobody, right?

I was a nobody, but the moment you show anything at all, like, oh, I made this, like something I made, for example, is this thing called fast grid. It's

like a really performant web table. And you just show, you know, anytime you talk to someone that you think is relevant, that can help you, you know, make sure you show them that you're relevant, right? Oh, I built this really cool thing. You

should see it. And they're always like, oh, wow, this is really cool. And now

suddenly, You know, they have a bunch of friends who start startups, for example. And

now they want to interview to them because they have seen that you know things.

Like everyone wants to help you. If you first can make sure, you know, that you can show them because they can gain a bunch of like, you know, social value from, you know- Doing an intro. Yes, we're doing an intro. Oh, you hired someone I introduced to you. Right. And they're a good hire. If you are a nobody, if you are from the middle of nowhere, like how, you know, you're from

the middle of nowhere in Sweden. If you're a nobody, How would you go about showing your value to someone important? The number one thing I recommend to people is making a demo that is super, super simple. It's actually really hard to make a good demo for a lot of reasons. Everyone thinks it's hard because they need to make a demo that is hard and they don't have the skills. This is very

not true. You can make very simple, like you don't need that much code knowledge

not true. You can make very simple, like you don't need that much code knowledge to make a really cool demo. The hard part of making a demo is making sure that people understand why you can code within three seconds. You know, you have like 100, like applicants for something. If you apply with one link and they press the link and you know, you have one shot, right? Like making sure you build

a really cool demo where people understand what they're looking at, which is really hard.

And where people understand that you're a really good engineer, which is really hard, but then you're there. I mean, that's all they want to see. I mean, companies just want to make money. You show them how to make money that you can code and they'll hire you. And then you might say, oh, but they only hire people with degrees. Well, yeah, because literally no one has ever showed them that they can't

with degrees. Well, yeah, because literally no one has ever showed them that they can't do their work. They're like, oh, I had these internships. And the interviewer would be like, okay, what did you do there? Oh, I streamlined pipelines and made things 30% more Efficient. And they were like, okay, well, that tells me literally nothing. Okay, what

more Efficient. And they were like, okay, well, that tells me literally nothing. Okay, what

else? What have you done? Oh, I went to Harvard, I have the best grades.

Well, I still don't know if you can do the job, right? Oh, but I have all this extracurricular, I was a debate champion. You know, they start going on with all these things that your parents will tell you, people around you will tell you. Nothing matters. The only reason it matters is because no one can show that

you. Nothing matters. The only reason it matters is because no one can show that they can do anything. So then they start listening to these like proxy things. No

one, generally matter, you know, Now, there are people who actually cares, right? Who are

these people? The CEO will never care. They will never care. They just want to make money, right? Which is great. You just, hey, I can make money. Oh, great.

Here's a task. You know, everything's perfect. When the further away from the CEO you come, the harder it becomes because people start losing incentives to do the best thing for the company. And instead, what comes up instead? Well, they don't want to fuck up. They just don't want to lose. So how do they hire someone that if

up. They just don't want to lose. So how do they hire someone that if they are a bad hire, they will not get any... You go through conventional accolades.

Like they went to the top school. Exactly. Oh, they went to top school. How

could I know they were bad, right? So then the recruiter doesn't make a mistake, right? So that's the thing to avoid. Avoid people who have no incentives in the

right? So that's the thing to avoid. Avoid people who have no incentives in the company. Generally avoid recruiters at a company. They're not even technical. They can't even know

company. Generally avoid recruiters at a company. They're not even technical. They can't even know if you're good or bad. They will only go on all these proxy signals. And

that's why people say I need a degree because that's everything they say. People don't

know that you can just talk to people. You can just go to an event, we take people. And every single startup I know wants to hire people who have high agency and can learn things. Yeah. Literally, if you're really good at using chat GPT, you saw one of these people at a random event, you went up and talked to them and you give them some advice and you're like, yeah, I'd like

to, can we try working together for a week for free? This would be super fun. Like I have these random ideas I just came up with that I can

fun. Like I have these random ideas I just came up with that I can work with with you that are, you know, no commitment from your side, no time from your side, just like get a free data point if I'm good or not.

Yeah. And a hundred percent of them would say yes. They're like, oh, great. I

don't need to do anything and I can see if you're good. Like, if you're generally a person who knows things and... Like, not even knows things. If you're just a smart person who can use ChatGPT, you can get a job tomorrow. And here's

where people are like, but it's a risk, right? I won't get into college. I

won't do all these things. There is no risk. And you can even do it the risk-free way. Just apply to college. You can go to college. Yeah. Apply to

college. And while you're in college, apply to jobs, right? And there's zero risk. You

just put some extra times into applying to other jobs. The moment you have a job, no one will ever see your degree. Like the moment you have one real job, like why would you care about a degree? Like suddenly you've done things, right?

Like where does the degree come in? Like things you do weigh much harder than if you did linear algebra in school, right? And this is obviously for like people who really want to go all in at their career. This is, you know, obviously not the right thing for people who, I mean, I often recommend my friends say like, yeah, go to college. It's an awesome time, right? You'll have so much fun.

You'll meet a bunch of friends. You'll even like learn things. I mean, it's not useless, the things you learn. They just teach them to you in very inefficient ways.

And you will meet very cool people. You can meet way cooler people if you went to San Francisco or just go network with people or work at companies. You'll

meet so many more interesting people. still meet interesting people at college like you will get all the things just less efficiently so it's all like a question about like what do you want to do in life i think it's very easy to convince yourself that college is right if you're like hyper ambitious like for hyper ambitious people i'm always like yeah you should drop out as fast as possible you can still

do the safe right by going there and making sure you can continue if you don't get a job but if you're you know really ambitious and you really care about your career i mean that's you know that obviously best thing to do. I

would have dropped out of high school if I couldn't get someone to do it for me. Yeah. So no, that's really awesome. University in a lot of ways is

for me. Yeah. So no, that's really awesome. University in a lot of ways is in a super controversial way where like I see it a bit like an adult daycare. You have a bunch of people who need to take a decision about what

daycare. You have a bunch of people who need to take a decision about what they want to do further along in life and you don't want to make a decision and then an option comes up. Especially in Sweden where you don't even pay for college. Wow. They're like, oh yeah, Here you can get free money and push

for college. Wow. They're like, oh yeah, Here you can get free money and push making decisions further into the future. And we also have all these courses where you don't even need to decide. Like if you want to be a lawyer, you got to do this niche thing. But if you don't want to be a lawyer, we have this like, you know, civil engineering or industrial economy and all these like courses

where you don't even need to decide what you do. You just continue doing random things for five more years and you just push your decision. People love pushing decisions, right? Like I don't want to show us what I permanently do for life because

right? Like I don't want to show us what I permanently do for life because that's what it feels like you're doing. It's not true, but it will feel like you're choosing what you permanently do for life. If someone do something for five years and they earn a certain amount of money, even if they took a job in a completely separate thing, like say they went from being a lawyer to marketing or

something, even if they make more money and are more happy and everything, people will be like, did you just do that? Did you just waste five years of your life? And you know, It's a really weird question to me. Like, they just upgraded

life? And you know, It's a really weird question to me. Like, they just upgraded their life satisfaction and salary. I mean, maybe they even downgraded their salary, but they were just much more happy. I mean, it's all about being happy in the end.

What advice would you give to someone who doesn't know what they want to do and they're like 18? Maybe they're in college or maybe they're in high school or maybe they just graduated college. What would you tell them? This is a classic persona, right? I've met so many people in this position. I've been in this position, you

right? I've met so many people in this position. I've been in this position, you know, I think it started for me in like late elementary school. where I was like, okay, I really wanna make money or I wanna do business or I really wanna succeed in life. I'm not really sure what that means, because I haven't seen anything at all. I don't even know what a startup is, but I want to

succeed in life. And then you start searching online. How do I make money? Right.

And you have the, you see these like service pages and you see them do service and you're like, holy shit, I'm making money online. And you do all this like, you know, you have no idea where to start and no one tells you where to start. Never just like go to college, you know, wait, wait eight years and then you can start doing things or wait 10 years or whatever. And it's

really tough to know what you want to do. You know, some people are very lucky. they play Minecraft and they start doing Minecraft servers and they realize, I can

lucky. they play Minecraft and they start doing Minecraft servers and they realize, I can make money from this. And from out of nowhere, you're running a business and you just start thinking, right? then you're fine for life. Now you know how it's done.

You know, you do something, you start making money and you start really thinking about things from the right perspective. I think my suggestion would be you kind of got to do a leap of faith and becoming a software engineer, especially with ChatGPT and, you know, doing demos and things that might be a good leap. You know, start making games. The good thing with software engineering is that you can show all your

making games. The good thing with software engineering is that you can show all your work super easily. Just send people links. You make a game, you have a good story about yourself and how you're a high agent and how you can learn fast and maybe screenshot an example of how you can learn fast from ChatGPT. I mean,

hiring managers will love this. Like, oh, this guy gets it, right? And then you just send that to like 500 people, right? And one of them will be like, yeah, I should give you a shot. You know, you're an undiscovered talent. You'll make

like 10 bucks an hour now when you're early, but it's worth, you know, it's worth our time to try out. And this person will learn a lot. They'll work

with real people. Like as fast as possible, get yourself into real context with real problems, where you make real money, where people have like real economic incentives and from there it just continues to roll. Then you're fine. But it's really hard to get there. That first jump of like... That first jump. From like something comfortable like school

there. That first jump of like... That first jump. From like something comfortable like school or high school to like, okay, I'm like working somewhere or working on something. There

is all these like classical paths. Like you have programming, which is one very classic path where you can do this. Marketing is another one. Because it's very easy as well. If you want to become a marketer, how would you sell yourself? When I

well. If you want to become a marketer, how would you sell yourself? When I

was 16, the first thing I would have done is to email a bunch of people, hey, I can do marketing. They all ignore me, right? Why would they respond to random, you know, elementary or high schooler? If you go into their website and you, you know, you crop their stuff and you do like, you do free work for them and then you post that to them and they'll be like, oh, wow,

this guy just made work. I mean, it's such a low bar for what's considered work. It's a really low bar. Like everyone needs a bunch of, problem solved, you

work. It's a really low bar. Like everyone needs a bunch of, problem solved, you can just solve one of those problems. I mean, you're hired, right? Yeah, you already did part of the job. You already proved to them, hey, like you can do it. And it makes the decision so much easier for them to just, all right,

it. And it makes the decision so much easier for them to just, all right, here's like, here's like a contracting per hour role. Yeah. Let's just start there. What

do you think are the most important things for people to work on. I mean,

I'm extremely AGI-filled, right? I'm like, okay. The only reason I can do the job that I'm doing right now, which I love, I have so much fun, I'm working with like really talented people, that's because AI. And like way many more people are going to be able to do what I'm doing right now. That means like way faster innovation, all aspects of life. Like when it comes to curing the CCS or

making random experiments. If someone is really interested in space, they can learn everything instead of making small rockets, and maybe they can join some cool rocket company. And there's unlimited possibilities, the smarter the AI gets. And the

company. And there's unlimited possibilities, the smarter the AI gets. And the

double-digit increase to world GDP will see from this. I'm extremely excited for that. I

have some Twitter posts here. Very controversial takes. Yeah, we'll go through some hot takes of you, and you can kind of go more in-depth. All right. Learning ML with my professor 01 preview, bounce what to make, code it, debug it so it works, explain parts, the intuition why it works, the main intuition why, explain to me like

I'm 12, go into all details. Learning this the other way around would be so hard. Right. Yeah. So I think my thinking here was like, cause I wanted to

hard. Right. Yeah. So I think my thinking here was like, cause I wanted to become extremely good at ML and yeah, this is kind of the, the, the path that I explained before of how you can learn very fast. So, You start at a problem, you can ask JGPT for which problem you should solve. It should just give you great problems to start with. And you know, you code it with JGPT,

you try to understand the code, how it works. When you understand the code, you can go into depth like for each specific module. How do you learn that? Just

recursively go down, all the way down. And yeah, I was just having a very strong feeling of like, wow, if I had to learn this from like the math up, that'd be so hard and take so much time. Yeah, like if you thought that, oh, I needed all these prereqs before like, you know, CS406.

I need to take like CS365W or something, right? Like

it's much harder to... I feel like that just instills like the wrong type of belief. Where like, this is unbounded. I should just master this coursework, which is a

belief. Where like, this is unbounded. I should just master this coursework, which is a prereq for that. Whereas I like the way that you call O1 Preview your professor.

And I like that you just... You say, fuck it. I can learn anything and I can just recursively go from top to bottom versus, oh, I can't touch that knowledge because like I'm just in underground and I haven't done like the prerequisites for it. Yeah. And I think this is a huge shift in how we think about

it. Yeah. And I think this is a huge shift in how we think about knowledge. And I think on Excel, I'm often like very skeptical to academia, not because

knowledge. And I think on Excel, I'm often like very skeptical to academia, not because academia in itself is something bad or because that's not something you should do. I

think a lot of people, you know, they think it's super fun to go to academia. A lot of good things comes out of academia, like all these papers and

academia. A lot of good things comes out of academia, like all these papers and so on. But there are some very dangerous takes people have in academia. One of

so on. But there are some very dangerous takes people have in academia. One of

those things being you can only learn foundational knowledge from, you know, the classical ways of starting from the bottom up. And people are so ready to defend this that they're, you know, ready to go to war for it, right? They're like, oh, you're wrong. You will never be able to understand these things. And then I'm like, why

wrong. You will never be able to understand these things. And then I'm like, why do people get so upset? I mean, it seems like, you know, when someone spends a lot of time doing something and then being told that, oh, there's more efficient ways to do this. I mean, their ego gets hurt, right? Yeah, they spent 10 years doing something. They spent 10 years doing something. And this high school dropout comes

out of nowhere and just learns it. Yeah. And it's tough. It's really tough. And

when I write these kind of things, it will hurt people's feelings. And to be honest, that's kind of the point with it. Because these people gatekeep others from getting into whatever they wanna do. Like if someone's 17 years old, they wanna learn ML and they ask all these other people who have learned this from before and they'll all be like, yeah, you need to spend a lot of hours. You need to

do all these classes. You need these professors at these universities and blah, blah, blah, blah, blah, blah. And you know, it's just not true. That's not what you need to do. Or you can do it. It's fine. Even if you think that sounds

to do. Or you can do it. It's fine. Even if you think that sounds fun, go for it. I mean, it's probably a fun thing to do, but I mean, there are simpler ways. And if you crush the ego that is like behind all these ideas, more people can do what I've done so far. Universities don't have monopoly on foundational knowledge anymore. Here's how I learned the main intuitions behind diffusion models

as a high school dropout with ChatGPT. I go through like some ways like that I described earlier of like how you can learn. It also describes very well what I think about like foundational, foundational knowledge. Yeah, I like, I really like how it's like they don't have the monopoly on foundational knowledge anymore. And pretty much what matters is you have agency over your own curiosity and learning. You can pretty much learn

anything. And the university wants to have a monopoly on your learning. If you're a

anything. And the university wants to have a monopoly on your learning. If you're a professor that your entire life have been spent, you know, talking to people about, you know, why you're going to university is so important and suddenly you don't need it anymore. They'll do everything in their power to, you know, keep it be that way.

anymore. They'll do everything in their power to, you know, keep it be that way.

Like what happens if all the smartest people start learning themselves? Well, then the smartest people won't go to university and this listens, you know, the status that you can have at a university. It's very bad. Their egos are challenged. Like I spent 10 years of... education, I'm like $400,000 in death. I have successfully

years of... education, I'm like $400,000 in death. I have successfully combated my dissertation for my PhD. And you're telling me that Gabriel Peterson over here says, I'm currently doing a job traditionally only done by PhDs with zero ML or math experience with the help of chatGPT. IDK, what other proof that chatGPT is at

PhD level that we need? Yeah, that's a pretty good post. Like to be clear, I mean, the people who are professors who have done PhDs, they've done an incredible job. right? I mean they've done super cool things for the world, they've you know

job. right? I mean they've done super cool things for the world, they've you know done done done research about like very important things and I don't think at all of these people like it might sound like I'm talking down on these people and that's not true at all. The only thing I'm talking down about is these concepts that come with the old types of thinking. It takes three days to learn diffusion

models top-down and six years before you can learn it bottom up. That's like the perfect analogy there. Exactly. If you want to learn diffusion models after university, it's like at least six years of, you know... Schooling before you even encounter it. You're like,

okay, I need to do calc one, calc two, linear algebra. And then you make introduction to machine learning. And this is the problem. Like, how do you even know that you like doing diffusion models six years ago? I mean, this is the problem with the university. You can... try out different jobs faster and you don't need to commit to something. Like, I know so many people that are like, oh, AI sounds

cool, so I'm gonna take this course. And they have no idea what an AI is until they're like three years in, which is insane. Like, should really take three years until you know what you chose to do the rest of your life and spending like millions. And you're like, oh shit, maybe I wanted to learn something else.

But you could have just learned it far faster if you yourself had felt more agency. It's like literally just like, can you, do you think you can learn diffusion

agency. It's like literally just like, can you, do you think you can learn diffusion models in three days? And even if you want to go to university, because I mean, going to university seems to be a great experience, like what do I know, but it seems to be super fun. And even if you want to do that, you can still go to ChessGPT, learn all these things before, you know, go straight

to the end and see how it is and see what it means. I just

know like, oh, this actually sounds interesting. I'm ready to learn this, right? And that's

great. Then you've made a very well positioned, you know, you've made a very good decision. You know that you'll have a lot of fun in college, you know that

decision. You know that you'll have a lot of fun in college, you know that you learn something that's fun, I mean, then you've done a great decision. I love

it when people are able to learn something far faster in time frames that other people just cannot believe. They're so used to like, oh shit, I've like spent so much time trying to pick it up and like, you know, you're like a living example of just being able to do this and you can do this constantly. If

you just believe you can learn something really fast and you're willing to just like keep asking questions like how You know, you laid it out, like, holy cow, like, you learn anything. And you can be working at the top AI lab in the world like you are. It's crazy how simple it is. It's like, in the end, it's just like companies want to have people who can do cool things. Show them

that you can do cool things. Just like, remove all advice you've ever received about finding a job and just start from that very simple trick. Because that's all there is to it. And then people will try to add, you know, things on top of this. Yeah, just start from there. That leads me to one of your tweets.

of this. Yeah, just start from there. That leads me to one of your tweets.

Companies just want to make money. So all you need to do is show them how you can help them make money. Drop everything you ever have been told about getting a job and start from that simple statement. I think people are severely misaligned sometimes when giving advice. If you've done five years of college and you're happy with it and you're like, oh yeah, I learned good things. And now someone comes in

and asks you, hello, I want to have a very good career. What should I do? And this other person that has chosen, you know, the path that basically everyone

do? And this other person that has chosen, you know, the path that basically everyone else chooses who want to do like, for example, software engineering and they'll also be like, yeah, I genuinely think you should do, you know, spend these five years doing this thing in university. And I mean, they share no incentives with you. They are

well intended. They want the best for you, but their take is completely meaningless. It

doesn't mean anything. This person has no life experience. This person has done one thing, haven't compared to anything else. And even if they tried, they probably mentally can't do it because they're locked into, you know, oh, I did this five year thing and Did I just waste that time? No, that can't be true. This must be the right thing, right? I mean, your opinions nearly always comes after incentives. So I generally

recommend people to discard most advice. That's how I dropped out of high school, because I discarded advice. I just never trusted people to... I've always been like, yeah, people want the best for me. They really think they want the best for me and they think they're helping, but most people will give you advice that doesn't help. As

in my example, if you ask someone who's done a college and has never really thought about their career, and they're just telling you the same thing and they'll have very good intentions, but it's actually like a reverse data point. doesn't mean anything. How

did you deal with counter advice? How did you deal with maybe were your parents supportive when you dropped out? This is very interesting. This is something I've realized later in life and I think a lot of people will be like, oh wow, and you know, there's two types of parents where it's more like a grayscale. My parents

never were never really like, oh, you should get good grades in school. They were

just like, oh, as long as you have an okay grade, like from a scale A to F, you get an E, right? If I get E or more, they'll be happy. And that's like all opinions they ever had on my career. They never

be happy. And that's like all opinions they ever had on my career. They never

had another opinion. And I was angry. I was like, why do they not care?

Like, why can't they like force me? Like, why can't they push me harder to do things? Like I have really bad discipline, but I wanna do all these cool

do things? Like I have really bad discipline, but I wanna do all these cool things. Why can't they push me harder? And turns out there's like this gray scale

things. Why can't they push me harder? And turns out there's like this gray scale of parents. And my parents were at the very end of one of the sites.

of parents. And my parents were at the very end of one of the sites.

And that end is like how much ego you have attached to your children, right?

Like how, you know, parents on the other side, they have really strong egos touch with children. And these are often parents that like maybe they didn't succeed with the

with children. And these are often parents that like maybe they didn't succeed with the things they wanted to succeed with as children. And now they're trying to live out their dreams like through their children. Like you're gonna be a doctor, you're gonna be a lawyer. Like this was my dream, this should be your dream. Like I love

a lawyer. Like this was my dream, this should be your dream. Like I love you as long as you do these things because that's what I would have done, right? And you know, they genuinely mean the best for you. Like they convince themselves

right? And you know, they genuinely mean the best for you. Like they convince themselves that like, yeah, I really want this for you because I really care. But often

it's very ego-induced, like, oh, I want my kid to do this so I can tell all my neighbors or friends about what my kid can do. So you have this two very... And everyone is somewhere in the middle of the scale. It seems

like in Sweden, it's much more common to be very far down on the scale.

Very few parents in Sweden will care about your degrees. Yeah. Because it's free. That's

one part. Yeah, because it's free. That's one part. And also, I think people don't have their... Like parents doesn't have as much ego attached on average. Then there's

have their... Like parents doesn't have as much ego attached on average. Then there's

always people who really have this as well. But yeah, I think it really differs from different cultures. And now when I know how this works, I'm very happy over like the childhood I had where I could experiment and do whatever I wanted to. And I just got support for whatever I did. Obviously they were a

wanted to. And I just got support for whatever I did. Obviously they were a bit surprised. What, you gonna drop off high school? I mean, my dad was like

bit surprised. What, you gonna drop off high school? I mean, my dad was like pretty like, oh, no, no, no, this is bad. But they got over it very fast and now, you know, it's nothing weird. I'm glad. That's pretty good. I think

most people might not have that experience. How do you decide where to take advice from? There's very few people I take advice from. Very, very few. I mean, I

from? There's very few people I take advice from. Very, very few. I mean, I can count them on one hand probably. My cousin and I happen to be very similar people in how we think and he also happens to be many years older than me. So, you know, he went through college. I mean, he's the biggest reason

than me. So, you know, he went through college. I mean, he's the biggest reason where he was like, yeah, I went through college. You and me think super similarly.

We want the exact same things and you'll just waste your time. Like that was a big part of the trust where I was like, okay, you know, we think super similar about things. And he kind of like shortcutted a lot of things that he had to go through and spend years doing that wasn't worth it for him.

And I could kind of just like follow his footsteps immediately. And yeah, he's one of the people I take a lot of advice from and that I based a lot of my opinions on early. But especially if you aren't established yet. You don't

really have a track record. You're young, you're unproven. And

how do you go about finding advice or mentorship or where to trust or where to learn from? And what do you do? How do you go about that? How

do you go about figuring yourself out? I think it's so easy to get stuck as like a very ambitious 16 year old and like, first of all, like watching motivational speeches on YouTube, right? You have all these people like, oh, I made millions and I did la la la la la la. And it's, you know, just all slop. And, you know, you get this feeling after you watched it, like, oh, I

slop. And, you know, you get this feeling after you watched it, like, oh, I just saw something. I saw a secret. And now I have this like, you know, I feel good. I feel motivated. Right? You have to motivate it, right? you're like,

yes, now I'm gonna be motivated forever. And then you wake up the next day and it's all gone, right? And these videos are, they're traps for motivated people to fall into and feel motivated while you're doing nothing, right?

And it's the same with people have this mega focus on good habits and reading and all these things that people say are extremely good for you. I mean,

they are really good for you. I mean, reading is great. Working out is great.

Having good habits is great. None of this matter if you don't do it while you're doing something that matters, right? And what matters? Well, you don't know yet because you're young and you've also, you know, you've no data points on what matters, but you should start trying. And literally the number one way is to do real work, solve real problems. There is millions of startups out there who, would happily have

someone work for free for you. You just need to reach out to them. Go

to LinkedIn, find Stealth Founders. I mean, send a message to all of them. Be

like, hey, I want to try working with you. Let's try working this weekend or something. And I'm ready to do whatever. I know I can do good things.

something. And I'm ready to do whatever. I know I can do good things.

And I can just start working with really simple things for you. I mean, just get your first real experience as fast as possible. That's all it's about. Like everything

I ever did in my life that was not towards getting real experiences, For example, when I spent a bunch of time in elementary school and high school reading books, getting good habits, waking up early and taking a run, and I felt so productive, none of it mattered at all whatsoever. It means zero if you're not actually doing something that matters at the same time, which is nearly always work. Maybe it's something

else for you. I mean, you need to explore this for yourself. But for me and for very many others who think alike, it's get your first job as fast as possible, which is very tough. requires a lot of work, but that's where your eyes should be at. 70% of people are in permanent slight suffering because they are allergic to making any mentally tough decisions when there is also an option to do

nothing. Yeah, this is a very good post. This summarizes, I think, why, you

nothing. Yeah, this is a very good post. This summarizes, I think, why, you know, if people have read this post, understood it and acted on it, I think like the happiness level in the world would rise by like 20% or something. The

work satisfaction you'd have. Because I've realized that it's so easy for people to make bad decisions when it's emotionally tough. And people don't know when it's emotionally tough because your brain doesn't really like let you know this, right? So for

example, if you're at a job you don't really like it, what's your options? Well,

your option is to reach out to a bunch of other jobs, compare them, and then you need to prepare for interviews, super tough. You need to do the interview.

It's very emotionally suffering. You might get rejected. That's very sad. And then you need to negotiate offers, which is a horrible experience, and then you need to tell your current employer, hello, I'm leaving. Oh, I'm going to leave. Which is so mentally tough.

So before your brain even thinks about this step, it will just be like, oh, yeah, I'm probably doing the best thing I'm doing right now. I'm probably learning the most I can do right now. Like, even if you're super ambitious and you want to learn fast, you will convince yourself that you're learning the most you possibly can, wherever you are. Even if it's your first job. Even if it's your first job

and it's a completely random job, out of all millions of jobs, you'll convince yourself you're learning the most. Even if you've been there for like three years. But what

matters is, uh... What? Seeing the truth? Being like honest with yourself? Yeah, what matters is, you know, say what you want in life is... good salary. That's like the simplest thing you can want in life, right? A lot of people just want a good salary and work with good colleagues and maybe they want to work remotely. I

don't know what they want to do, but they want some combination of this, right?

That's fairly standard. So one very, very clear example is I had a friend in Sweden. I was making a normal Swedish salary, which is extremely low seen to SF

Sweden. I was making a normal Swedish salary, which is extremely low seen to SF standards. From his standards, they were pretty good, right? He was living life. I mean,

standards. From his standards, they were pretty good, right? He was living life. I mean,

in Sweden he had like a, you know, probably like 50% higher than a normal salary. And I told him many times, I was like, dude, why do you not

salary. And I told him many times, I was like, dude, why do you not apply to San Francisco? I mean, you'll just like 10x your salary and work with better people and work on things that people use and like all these things. And

you know, he had all these reasons, you know, came up in his brain for why he shouldn't do this. And you know, he constantly pushed it forward and he couldn't really, even if he trusted me as a person, you know, Your brain doesn't allow you to think about this as an opportunity because it's like so emotionally tough.

And one day I just introduced him to a company and I was like, yeah, fuck it. I just put these two together and make sure he starts interviews. And

fuck it. I just put these two together and make sure he starts interviews. And

when you start interviews, that's way more simple because they will like try to pull you into the company, right? So they will do it. There's momentum. Yeah, there's momentum and they'll do all the annoying things for you. And then from out of nowhere, give you an offer, right? And you never had to think about it. And now

when you have an offer, you just need to sign it. And emotionally, like the right thing would obviously be to apply to you know, a couple of other companies and see, will these people give me a better offer? Are these people better to work with? Are these more fun people who believe in this company more and I

work with? Are these more fun people who believe in this company more and I want to stay at it longer? But the easy thing to do here, right, is just to accept it, right? Anyway, yeah, this friend just ended up 10xing his salary, just like that. And, you know, he pushed this for like, I told him maybe for a year or so. So if you 10x his salary, I mean, that's a

lot of money. Just by coming to America or getting a job with an American company. Yeah, that's like a full house in Sweden. He just got, like, people don't

company. Yeah, that's like a full house in Sweden. He just got, like, people don't understand how much they lose from this. He literally lost buying a house in Sweden if you just see it in like just currency, right? Yeah. He lost the house just because he didn't do this very simple thing, which is buying San Francisco. It's

very simple. What's your advice for people who want to come to San Francisco? Why

should they come to San Francisco? Why should they come to America? First of all, I mean, Let's start simple. The talent density is much higher. It will work with a very high concentration of talented people and the salary will be much higher. Like

whatever you like want in life, you know, your life standard will always depend on your salary. So even in the scenario where you're just like, yeah, I just want

your salary. So even in the scenario where you're just like, yeah, I just want a high life standard and I'm fine, you know, and I, you know, I'm fine moving to get permanent son, which, you know, for some people is a very easy decision. For some people it's a harder decision, but like if it's, Those two things.

decision. For some people it's a harder decision, but like if it's, Those two things.

I mean, that's, you know, reasons enough. And then like the more ambitious you become, the more obvious and like the more objectively correct it is to move to San Francisco. You really don't have any other options. If you like, if you're ready to

Francisco. You really don't have any other options. If you like, if you're ready to like put your entire life in startups, your best bet is moving to San Francisco.

The talent is here is extremely high. Capital is flowing super freely. Everyone is talking all the time. You know, you can hire extremely good people. And it's just another momentum. Like if you've never been to San Francisco, everyone like told you should go

momentum. Like if you've never been to San Francisco, everyone like told you should go to San Francisco. Like the first week they're here, it's like always changes their like, worldview, right? They're like, holy shit, I didn't know there were this many people caring

worldview, right? They're like, holy shit, I didn't know there were this many people caring about whatever I'm doing in the entire world. And I just met like more than I ever thought would exist in the same room at the same time, right? Which

is like a really motivating experience. That's real motivation for you. That's no like person standing and talking about motivational speeches and they tell you all this like random slop that like doesn't make any sense, that just sounds good. There's like real motivation, right?

You have people who think like you, they act like you, they work hard like you, they care like you, they don't work 40 hour weeks. It's just an incredible city to be in. I mean, look at what has come out from Silicon Valley.

I mean, nearly all innovation happens here. Like, it's so much, like probably more than all of Europe combined happens in San Francisco. The amount of capital in only San Francisco is probably like magnitude higher than in all of Europe together, right? You have

Apple, Google, OpenAI, Anthropic, all these huge companies, they all have their headquarters here for a reason. Everything you care about in life, like life standard-wise, like what makes your life good? Well, you have Google Maps, right? Before you had like physical maps. Well, turns out Elon Musk built maps, like Zip2, whatever it's called, like in

maps. Well, turns out Elon Musk built maps, like Zip2, whatever it's called, like in San Francisco, like 15 years ago. You have your iPhone, which is an incredible piece of technology, built in San Francisco You have satellite internet. You have people that go from being paralyzed to being able to walk again. You have diseases that

are cured. This is all, nearly all of this is from the US and specifically

are cured. This is all, nearly all of this is from the US and specifically San Francisco. There's like a million people living here. It's like same size as Stockholm,

San Francisco. There's like a million people living here. It's like same size as Stockholm, right? Yeah. It's crazy, but there's a network effect and talent density here. Yeah. Yeah.

right? Yeah. It's crazy, but there's a network effect and talent density here. Yeah. Yeah.

So what's your advice if someone wants to come to San Francisco or move here?

First of all, be highly skilled. I mean, be good at something. That's the first thing. And that will come easy, right? If you're ambitious and you want to do

thing. And that will come easy, right? If you're ambitious and you want to do things, you just start working with real companies and you will become good, right? So

when you have this, the second step is, you know, find the company that's ready to sponsor you. And this is where you need to be good, right? Because... You

know, you need to be really good because this company will want to only hire...

Like, they need to go through a visa process, which is some overhead for them.

And they obviously would rather just take someone from the US if they could, right? It's way less overhead. But it turns out there is very few extremely good

right? It's way less overhead. But it turns out there is very few extremely good software engineers in the world. And there's a huge need of good software engineers in the US or like everywhere in the world, basically, needs really good software engineers. We

can never get enough. We could double the amount of software engineers that are really good in the US. We still have not enough. And that's why you have a really big chance to just move to the US if you're from another country. And

you can do it over the internet. You can do it over the internet. The

only way you can prove to a company that you provide huge value. People in

San Francisco is very happy to sponsor you. Especially now, the paths that want to become easier and easier. Definitely so. That's what I'm doing. I've had a blast talking with you. I mean, your tweets are really funny. I think your takes are very valuable. And for someone who's thought for himself

very valuable. And for someone who's thought for himself this whole time, ever since like high school and even earlier, like it's... It's really

awesome for you to be able to distill all that and then share it here.

So yeah, I'm always very happy to share hot takes and help others, other people who think like me to get there faster. I could have been here four years faster if someone just told me what I'm trying to tell people. But then always that's from my perspective. Like I have my own wants and

people. But then always that's from my perspective. Like I have my own wants and if someone happens to feel like, oh yeah, this applies to me as well. And

it can be hugely valuable. Yeah. I hope this episode captures some of that. I

mean, you are basically telling the advice to your younger self. Yeah. And, you know, I'm going to really try and, like, distill it into something really nice. And that's

what I also want to do. I want to just, like, help people like you.

from the middle of nowhere around the world, inspire them with like real anecdotes like you did not go to high school or you did go to high school, but you dropped out of high school and you're still able to land probably one of the most sought after jobs at the best AI company in the world. And you

did it mostly on yourself, mostly to your own ingenuity, mostly using the Internet and like pretty much anyone from anywhere, as long as they have the Internet, they have chat, they can do the same. And, you know, I goal of kind of this is to I want to share more anecdotes and like origin stories of like extraordinary

people like you and also just help a lot of these people because I mean similar to you I also went through the process of like leaving high school to come to San Francisco and I was oh shit I wanted to stay here and it's very stubborn I just I had to be here and I kept like coming in and out on tourist visas um I had the option to go to college.

I actually got into one of the best, most selective programs in the world. But

I just was like, San Francisco is clearly the place to be, where all the innovation are, where all the brilliant people. And I just need to figure out how to get a visa, how to get an O-1 visa and stay. And certainly by learning you know, learning on my own, learning using AI. Like I taught myself how to code with AI, working on my first engineering job that led me to get

someone to sponsor me, as you said, and then actually be able to move here.

And yeah, I feel like this episode, we can make it so that a lot of people, you know, their lives are completely different in better ways because they listen to you. So yeah. I hope so. That would be epic. That would

be really epic. One more thing that I thought about, which was a huge blocker for me when I was back in Sweden and I wanted to do all these things. And I think a lot of people who were thinking

things. And I think a lot of people who were thinking similarly as I did are feeling this as well. And that's that they think they're not very smart. I remember I thought I was an idiot. growing up. Or not

an idiot. I was like, yeah, okay, I'm really good at math compared to like the people in... In elementary school. Yeah, compared to people in elementary school where I lived. But you know, you saw people, you know, people building rocket ships and people

lived. But you know, you saw people, you know, people building rocket ships and people doing all these humongous things. And I was like, damn, like, how can I ever do another thing? There's just no way. But I think it's like really easy to underestimate how much you can do. It's super easy to underestimate. And... like

probably people that even just listening to this are in the top like one percentile people. Just from listening to this. Like most people wouldn't have agency to like, oh,

people. Just from listening to this. Like most people wouldn't have agency to like, oh, I want to do something and then spend an hour listening on something to do that thing. You're already top 1%. And top 1%, that's like, you know, they already

that thing. You're already top 1%. And top 1%, that's like, you know, they already have like that top, you know, the top 500 startups in the world or even fewer, like maybe top 200 startups in the world. That's like the top 1% of people, right? So there's, you know, only that should, you connect the dots like okay

people, right? So there's, you know, only that should, you connect the dots like okay if i just continue down this path this is like where i'll end up yeah you can come back to san francisco work at a laptop company or start a top company and yeah so much things to be done yeah well this has been great gabriel thank you so much for your time i mean yeah thanks for inviting

me holy i think you're one of the best people at learning in the world with ai and more people should know about that yeah i'm trying and i hope some people get you know much better than i do so i can learn from them as well