'9arm' คิดว่า AI กำลังเปลี่ยนงาน Software Engineer แค่ไหน? | Podcast with CK Ep.50

If you make hardware for AI, and one day the demand drops, it's a big deal. You suddenly have only a paper in the data center.

Actually, every war is about power. If I'm US,

all in, on power,

why don't I go to the moon? You have to call the president. I think you're right. They should do the same. Many Thai kids have enough skills to work in that room.

They just don't have a chance. Many people with skills are just not given the chance to work in that room. I've been to many places. I've seen people who are not good at it. When I got here, I was like, Oh, you! You're so good.

Klaus is my best friend. You're the one I don't get. Why did you choose to be a CK? Welcome to another episode of CK Podcast. Today, we're with a guest who has come the farthest. Yes. CK Podcast. is probably a country that doesn't have a lot of Thai food. Not a lot, but there are quite a lot of Thai immigrants.

And if you want to eat Isan food, do you have a place to eat?

I cook it myself.

You cook it yourself?

Yes.

So there's no choice. Yes.

No choice.

I'm very glad to be back. Because I really like your content.

Thank you.

Actually, you are a Thai person who gives deep information about technology. And I think it's great. And thank you for giving me the information all the time.

Thank you.

For those who don't know you, Arm,

can you introduce yourself? Sure. My name is Arm. I have a YouTube channel called Arm. I mostly talk about technology and other things. We talk like friends. I'm an engineer in the US.

I'm a software engineer for AMD. I work on GPU and AI.

Wow, that's awesome. If you've followed this year's Forbes list, one of the top 5 most influential CEOs is Lisa Su.

Yes.

AMD is one of the companies that has grown a lot. And it's almost the centerpiece for AI revolution. I want to know how you feel and experience to be one of the most important team members if not the most, in AI revolution.

Oh, I have to say, to be fair, NvidiaVIDIA is a huge competitor. They made it before. They made a lot of AMD before they came to the AI market. They were about to collapse. They didn't have time to focus on this. In 2017, when Ryzen CPU was released,

it was still competing with Intel. And if the CPU didn't work, they would bankrupt. At that time, NVIDIA was already far away. They already made GPU for scientific computing. And they already had a foundation. So they could enter the market faster.

And they were in everyone's use. Now, when AMD survived from bankruptcy. What year was it?

2017.

Okay.

AMD got a good license, and we started to look at GPU. Of course, our friend went far. It was an underdog that started later. But as I said, they have a hope that hey, Lisa Su has won AMD over Intel. The CPU market share now, Intel has been eaten by AMD a lot,

especially in the data center sector. From the docking to winning once, I hope that the same leader will take the company by the same leader. I don't know. Now, if you guess, it's about 90% Nvidia, 10% AMD. But I hope that in the future, the numbers will be better.

For those who don't understand the difference between CPU and GPU, and it's actually very important. I want you to explain a little bit. What is the difference between CPU and GPU. And how are the importance of each part different? And why is the transition from CPU-dominated market

to GPU-dominated market important?

Ah, okay. Actually, CPU, at first, there was only CPU. We used a general calculation. We used a computer. It has to work in many ways. CPU is a good kid who can do many things. Computer. You have to do a lot. It's a math. You just do a number. But you have to do a lot.

Because how many pixels are on your screen? 1 million pixels, 2 million pixels. You have to do it like… And every frame, you have to… Every second, let's say there are 60 frames. You have to do 60 million pixels per second.

So you have to use a lot of processing units. But you don't have to use a good unit. You just need a good unit for one thing. That's it. So GPU came out to do graphics. In the beginning of 2010, they started to think that a single number can be used in scientific computing.

Before we used CPU to compute, supercomputers used CPU. And it turned out that we used a good unit, but we only used it to compute numbers. It's not cost-effective. CPU computing. Supercomputers back then used CPUs. And it turned out that we brought in a good kid, but we only used him to solve a number. It wasn't cost-effective.

So we started to test to use the GPUs that gamers use to compute science. It turned out that it worked.

It worked well.

It worked better than CPUs. Maybe not much better. But Nvidia leaned in. So we thought, GPU. But NVIDIA leaned in. So I was like, oh, someone is using this. So I tried to use the skills of the general consumer who plays games

to make a special GPU that uses scientific computing. It became bigger than the consumer grade. It turned out that it was able to calculate faster. That's the thing about scientific computing. We are talking about physics,

and aerodynamics. This is the difference between CPU and GPU. When it comes to AI, it's the same thing. AI is math. When AI boomed in 2015,

the hardware that makes sense for computing is GPU. That Scientific Computing uses. So we brought it. NVIDIA was okay with it. They accepted everything. They accepted scaling,

they accepted software, easy to use, whatever. So we came to the current era. Now, if we go back to AMD, we are in this era. But while NVIDIA is watching this,

AMD is still busy with CPUs. So it's just coming back. I think around 2020, they just kickstarted the GPU project for data centers.

For me, before the AI revolution, what I was impressed with was for data center. If I use CPU, I can do it. But the energy and speed, I will use more energy and speed slower. Because they are all good, but it's not enough to focus on one area. And they have few. For example, CPU has, let's say, 12 cores. GPU has 2,000. So the scale is quite different.

So, in terms of numbers, GPU is much better. So, we will go to the cryptocurrency that has to calculate the hash.

Or it will be scientific computing. It's the same case.

Now, in the chip industry, there are a lot of new things coming out. Whether it's TPU from Google, right? Or LPU from Glock that Nvidia just bought. What are these? What's the difference?

It's going to be specialized. For example, CPU is like a Swiss knife. It can do everything, right? CPU is a number. Mostly we use a number. Now when it's a TPU,

it's a work that can do. It will focus on it. It's not just a number. It was created to make a tensor product. It's a kind of mat. It's good at specific mat. And the tensor is the common place that they use in AI. So, TPU is specific.

The more hardware, the more specific it is, the more efficiency it will have. And it's more cost-effective. But it can't be used for anything else.

We will lose the versatility, but we will get efficiency. it will be more efficient. And it will be more cost-effective. But it can't be used for other things.

But we will get more efficiency.

LPU is also going deeper. More specific.

I think the data center of the future, today it's dominated by GPU. Because of the production level, NVIDIA is the best. But I think in the next 5-10 years, I think one data center will be compromised by

how much GPU, LPU, TPU, or will there be a specific one that should be added?

This is very interesting. If you ask me, I feel that the GPU is the best investment right now. Because even though AI bubble will explode, GPU can still be used for other things. Scientific computing, right? But if you make hardware for AI, Even though AI bubble will explode, GPU can still use it for other things, scientific computing.

But if you make hardware for AI, and one day the demand drops, it's a big deal. You suddenly have a paper in the data center. The investment is zero. So, I feel that GPU will dominate for a while until there is a point where it proves that TPU is cost-effective.

I don't think so yet. Because if we talk about AI, when we train each AI, there will be an ecosystem. The GPU training ecosystem is very strong right now. For example,

I said that NVIDIA has a very good AI ecosystem. AMD, which I'm working with, is a competitor. They have to try to beat this ecosystem. It still takes time. 3-4 years, they can't do it as well as before. Because people are already addicted to the same ecosystem.

If you change to TPU, LPU,

you have to change your ecosystem or pipeline as well.

So it will take longer than adoption rate. If you want to change to TPU, LPU, you have to change your ecosystem or pipeline as well.

It will take longer than adoption rate. It will work. And you have to prove yourself that this is a better way. I think in 5 years, it should be possible. But there is a picture that if this works, after 5 years, it might be a hardware-specific.

That means AI will survived in this world.

Now I feel that AI doesn't know if it will survive or not.

Why?

Because it doesn't have... What is it called? It doesn't have a return yet. There are a lot of investments. There is a return. But it's not worth the money invested.

And now it's been a year, I think it's been 3 years. Return me. At that time, there would be a problem. But if we find a breakthrough in the next year or two, and it continues, we might see a new scheme. It might be TPU, it might be LPU, or whatever. But it's an asset that was created to use AI.

So this breakthrough is probably AGI, as people like to say. The definition is probably a bit mixed.

Yes, yes, yes. But it's very difficult, TK. I think AGI is… In my opinion, the technology today is not up to that level. LLM is… The one we use, GPT, Gemini.

LLM is Next Word Prediction.

Yes.

And it's just Next Word Prediction. It's good until we think that it's like… It's like intelligence. prediction. ใช่. แล้วมันแค่ next word prediction ดีจนกว่าทางเราคิดว่ามันเหมือนกับ มี intelligence ขึ้นมาเนอะ. แต่ว่าเราว่ามันมี ceiling เทคโนโลยีนี้. We have to teach them the basics. And they should be able to continue. Yes. 100%. Now it looks like they can guess the direction of the answer. But it's not…

It's still wrong.

It's still wrong in the strawberry. Right?

Yes.

So I think this technology will have a ceiling. There may be an alternative that will take us to AGI. They are currently betting on it. But I don't know.

I believe in this hypothesis. I think LM is not a way to go to AGI. Because the data on the internet, it's... The data that comes to our eyes, in the real world,

many data are not captured on the internet. So if we want to go to AGI, So, it's like, it's like, it's like, it's like, it's like, it's like, it's like, it's like, it's like, it's like, it's like, it's like, it's like, it's like, There are many people who say that this year, next year… They say this every year.

Yes.

They say this every year. But in my opinion, I feel that they are in front of the company. They probably know more than me.

But they are also the sellers.

They sell the company.

Yes.

So I will just listen to them. Actually, if you want to know how far AI has gone, if you read to know how far AI can go, you can read research. You will see where the limit of LLM is. Now we are optimizing for efficiency.

But I think it's not really smart. But isn't LLM complete? Every data on the internet, they have eaten it all. Now they are creating new data, synthetic data. ทุกข้อมูลในอินเตอร์เน็ตเขากินร่วมหมดแล้ว. ตอนนี้เขากำลังสร้างข้อมูลใหม่, สันตาติกตาตาเอง. to capture the internet.

That's the case, we might get there. Right? But now, it's not like that. Well, this is what I don't understand. I think what we're saying is based on common sense. But every time moment, Elon Musk said this year is AGI. This year is AGI. This song is every year.

I understand that he's a seller to prop up his valuation. To get into the market. But it's very short-lived. In a moment, we'll know if what he said is true or not. And it has all digital footprint.

Yes.

Elon Musk, you promised many things. That it will happen this year. Full self-driving. Mars. I think it's possible in that case. But he will have to write engineering

to produce it. But in the case of AGI, I think it's quite difficult. It might not be possible this year or next year. Except… I don't know. If it's going to be released next year or next year,

I have to see something.

But I haven't seen the signal yet. What I think it can't be coded yet, and I haven't seen the trend yet, it's still impossible. I think the most important thing for...

to call yourself AGI is curiosity.

But if I hire an employee,

in the company, right?

It's not that I record the information for him, but he has to ask too. Like, what do you do with this? What do you do with that? Because sometimes, on board, I might be wrong. But he has to ask more to get more answers. But this AGI, I have to feed it. But they don't ask back. They don't have curiosity.

Because it's very difficult to have curiosity to represent humans.

True.

So, if this curiosity is coded by ALM, it's difficult because all the data on the internet is input only. But it's not a gathering that people find the data themselves. Yes. input. But who knows? Maybe in the next 2 years, there will be a research project that connects LLM to something. LLM might not be the technology that leads us, but it might be like a new master.

Right? It could be a precursor of AGI.

If I ask you this, I'm a foreigner.

Right?

I'm not in Silicon Valley. You're in Silicon— I mean, yes. AMD is basically in the Silicon Valley bubble. These are world class. What do you think about the people inside? When I met the CEO and talked on the stage, I was like,

What did I say?

Really?

Really?

Engineering is all over my head. Because there are a lot of promises on the stage. It will translate into some kind of work, right? And sometimes our CEO promises something that we think is okay. But that's his job. He might be a visionary. Translate. we did 120% performance of the competitors.

That number, I was still running at 3 AM last night.

Really?

Yeah. So I think it's just a guess. There's truth in some cases, but...

people who really work will know that it's hard. Let's go back to the hardware. Because NVIDIA is the market leader in controlling GPU. Yeah. That makes GPUs work together. It must be NVIDIA's. Stack. Ecosystem.

Operating system. It must be NVIDIA's. And when you do that, if AMD enters this market, and 2 GPUs, 2 brands. I want to know how comparable it is.

Is NVLink comparable with NVMD? If not, how do you do it? Here it is. Okay. NVIDIA has already done it, as you said. NVIDIA already has NVLink. They bought a high-speed network. 3-4 years ago, Melanox, which was the top of the world at that time.

NVIDIA is the owner.

ARM.

NVIDIA is the owner. So, CPU, GPU, networking, and internal networking of GPU. They control everything. They bought the company that already built rack data center. When they sell, they sell it as a rack. Yes. AMD has to try to do the same. I don't know what the vision of the manager is,

but AMD has to do the same. That is, we have our interconnect. AMD will go to a new position. They said that it's proprietary. You have to use all of them, right? AMD. We open it as an open standard. We use the standard that we already have. So if you don't want to be logged in, you can use this way. And you can swap anything. Without guaranteeing that the performance is equal.

I see how this move will be. But I see that it's interesting.

But can GPU talk?

Yes.

Okay.

GPU has a translation network. If this GPU brand and this network brand don't talk to each other, there will be a layer in the middle to connect all the brands. And we open a lot of alliances. And now there are quite a lot of alliances.

Okay.

It means that you are in the open source model, but NVIDIA is in the closed source model.

Correct. And this move of AMD makes NVIDIA feel like, Hey, we're doing something. They started from being closed source, and then they started to open source some parts of it to fight this move.

They're still fighting.

It's similar to iOS and Android.

Yes.

NVIDIA is like iOS, everything is closed. Only the device is open, but Android is open.

It's the same with Microsoft and Mac. Linux, sorry. Right? In the world of software, I feel that open source is more compatible. And it's more compatible.

AMD also bet that this standard will be better than Nvidia. There are many things, including politics. Because other companies are also in the same Nvidia side. So let's come together and fight it. Let's see how it goes.

Because the demand for data center is very high now. Many companies have promised data center to CapEx. I think AMD GPU, the GPU that is competing with Nvidia, Many brands promise startup centers in the capital. I think AMD GPU, the GPU that is competing with NVIDIA, should sell very well.

It sells very well, yes. But of course, we come from… Now our market share is 10%. 3 years ago, it was only 2%.

Is 10% because the production was not in time? Or 10% because of the demand?

10%… I think it's a demand because no one wants to train AI on AMD hardware in the last 3 years. Because it's hard to use. The company's milestone is that we have to make it as easy as possible. Work out of the box.

And it's been fixed for the past 3 years. It makes the market share better. Nowadays, when a new GPU series is released, it can train the model on day 1. You don't have to tune it again. Usually, Nvidia can use it right away.

AMD has to wait for a month to tune it. Now it's day 1. It's getting better. The progress is getting better. I think in 1-2 years, it will be seamless. And the computing power of the hardware itself, we can actually compare it.

Okay.

Now, there is no reason for the hardware to use Nvidia compared to raw power. But of course, people are still addicted to the ecosystem. It's still hard to use. If this problem can be solved, the market share will gradually increase.

I've listened to S I've heard a lot about Samsung. Because no one hasn't heard about Samsung for the past 2-3 years. And what the CEO said, you have to keep in mind. He said a lot that, you know, he and the closest competitor, generation apart.

His total cost of operation is very low. If you see a free competitor, it's not enough. So I want to ask you in front of the company. I hope you... Of course. I hope you won't be biased against AMD.

I really want to know how many months have passed between the latest generation of NVIDIA and the latest generation of AMD. I want to know how close this competition is. Or is it actually close to what you think? I think I said this a year ago.

Yes.

At that time, I had to say that there was a lot of gap. I think it was about 30%. 30-40%. No matter how you reduce the price, they will still take the higher one. Because AI frontier model needs good computing power, right? If you ask the current generation,

I have to say that the gap has been reduced a lot. It's around… If we think about raw power, we think there's no gap. But when we talk about software optimization, it's around 20%. Yes.

AMD has a problem that the performance is the same on paper. The model runs 10% faster than NVIDIA. Ah, like this. But our price is cheaper by 10%. So it starts like this, right? So the next model, I think it's the one that is comparable.

Let's put it this way. AMD just announced MI 450, which will be released in Q2 this year. I think this model is comparable to the current model of Nvidia. Q2 this year. who run a little faster. And we'll see who makes a mistake. We have to make mistakes many times to catch up. But they made mistakes twice. For example, when B100 was canceled,

we got a lot of market share in that cycle. If it's like this, we keep pace. We'll see who will be the first to jump and then we'll see. The goal is not to win or anything like that. But we have a chance to get more market share. I'd like to ask you this because I don't know.

The customers of GPU market, of course, for NVIDIA, they sell only a few, but the customers are only a handful. I want to get GPU, but NVIDIA can't. I can't even buy it.

Tessa, Facebook, Microsoft, Google. I can't buy it because of their price. Do you need to target the same customers for AMD GPU? Or there are other customers that you want to target, not just the hyper-scale?

Actually, it's the same sector. And as you said, AMD customers as well. But in the scale, at first, Meta used Nvidia, right? After the next generation, there was a collaboration. And they started to buy AMD more.

They tried it before the next generation, and they felt that it was enough. It was cost-effective. The next generation was more. And then AMD wanted to market, so they said, hey Meta, let's co-develop the new GPU. What do you want?

What do you want? Co-develop with us. What kind of model will you run? Because co-development, there is a demand already. The next generation that they will buy,

they will buy NVIDIA X%, AMD X%. AMD's percentage will be increasing.

Nowadays, I see... the way NVIDIA grows, not only they're developing themselves, they're quite active in M&A acquisition. Very active.

How can they not do it?

When you look at it, there's only cash on the floor.

I understand.

So I'm thinking that… Because… I'm thinking from AMD's perspective. When I see our competitors doing that, what is AMD's strategy to catch up? Or AMD will think…

You can't say it like that. It's an acquisition anyway. Or is there any strategy that MVDM doesn't do?

We don't have money. I have to tell you. We can't spend like that, right? So, it's not that we don't do acquisition. We do, but it has to be strategic. For example, we want to integrate both RAC and others.

We have to buy companies that do it. Whether it's a data center, a small hyperscaler, a machine maker, AI silo, software, we have acquisitions as well. But I think, I'm not in the company's meeting room,

but I think AMD's strategy is it's okay, we do what we can. Move to pace. We will grow. And when we grow, maybe we have more money, we can do more.

You are fighting with giants. You can't fight them directly. You have to ask for more and more.

The belief that AMD is not a winner-take-all market. There must be breadcrumbs. Or AMD is an ambitious pink paw that says, I will make a big cake.

I'm not Lisa Su, so I can't answer that. But I feel that we will eat cake slowly like when we made CPU market. It's like we found the same thing.

Some people say that China has more energy than us. Which there is more. I want to know, in front of the work, I want to ask people in front of the work. Do we have any energy problems today?

Yes, we do. We can't set up a data center. Because the power supply is not enough.

Today?

Today. Okay. You mean the GPU is already set up, but you can't switch it on? Yes, because the power supply is not enough. And actually they want to expand a lot. And look at CapEx that they committed, right? This is a future plan.

But we can't follow that because we don't have enough power plants. How to do it? We need to build more power plants. But why don't you build Power Plant and then buy the chip? Because we have to book in advance.

If you don't book now, you can rent it out.

What does Power Plant mean?

Everything.

Solar, nuclear.

Solar, nuclear. Yes. Trima Island reactor has to be reopened to serve this data center. And the plan to build an electric power plant

is going to be a lot in the next 3 years. You mean, by yourself?

By myself.

Because the US doesn't have... Doesn't have a government.

Yeah.

It's not like our country.

Yeah.

I understand. Let me ask you something. There are a lot of rumors now. About 3 months ago. Google said it. The center of the future is outer space.

It's outside the earth. So they said that it's very cold outside the earth. No water. It's cool. The sun is outside the earth 24 hours a day. Do you think it's possible in 5 years?

What's the answer? Before you know the answer, 24 hours. You're and Play store. You want other space? Orbit you.

Space. Orbit.

Moon. It's just a prototype. For a small scale. I think there are a lot of loads that need to be loaded. One data center. And the energy used is solar. It's a challenge that you orbit. You don't get sunlight all the time. What are you going to do at that moment?

For example, the International Space Station has a battery. But they don't use a lot of electricity. Data center. So it covers, but the question is how much solar energy do you need? I think it's a challenge.

And I think the complication is… I want to know this. I asked this question because I want to know. If we guess correctly, if this hypothesis is true, most of the data centers and most of the polluting industries are outside the world.

Today, if I'm in Thailand, or any country, ABC, and I say, AI, let's make AI. If I say that I'm going to build a data center, it will take 5-6 years to complete. But at that time, it was already in outer space. Is it worth doing?

It's worth it. I think. Because it will take a long time to scale outer space.

A long time.

You think about ISS, Space Station. It will take 10 years to complete. Because it has to be sent up to each module. But the money is not much. But it's not a lot of money. It's a lot of money this time. Yes, but our technology will be better. We might be able to go every week. I don't know.

But the cost is quite a lot. It has to be some kind of experiment. We can't pay for it ourselves. With the weight per kilogram that has to be shipped up there. On the orbit, how to do maintenance. I think it's difficult.

And which orbit will we be in? finance. but within 5 years, I still can't see it. I think SpaceX to go to Moon, they still can't do it. This plan is delayed for 3-4 years due to some challenges. The challenge is we can't carry as much as SpaceX claims to go to Moon. The technology must be good first and must be right.

And it takes time to shoot. And it takes 5 years to build the module. It's probably more than Prototype. Can I ask you something?

Do you believe that we went to the moon in 1960?

I do.

And why? This is what I don't understand. We went to the moon. If we ask people back then, they would think, 2026,

where are we? They would say, we would have been in the universe by now. Why? But we can't go back to the moon today. Why?

We can go back to the moon, but we don't pay money to go back. Moon mission, how much money do you need? Brilliant level. And it's like a national project. Everyone thinks we have to do this. But now we don't have that policy.

We will use money to go to the moon. Instead, we will solve the problem of the belly button. It's a matter of politics. Actually, we can go.

And the important thing is that we go there in a small way. It's about politics. Actually, we can go.

And the important thing is that we went there in a small scale. If we want to go big this time, we want to build a moon base. So the payload that we need to go is different scale. If we send a vehicle, we can send a vehicle continuously. No problem. But it depends on how big the payload we want to send is. To go to orbit the moon.

But now, the government should be responsible. So everything. Every war is about power. Venezuela is about power. Everything is about power. And China has more power than the US. China has more rare earth than the US. So if I'm in the US, all in. Yeah. Why don't you go to the United States?

Why don't you go to the United States? Why don't you go to the United States? Why don't you go to the United States? see me rare earth, not US. So, if I'm in US, all in.

and can build a data center, energy-based there,

it's over. But the country with the most energy will win anyway. And we're waiting for SpaceX to finish. But you said if we have enough money, we can go. I think if we pay more than that. Or instead of just SpaceX, NASA will do it.

If NASA does it, the cost will be higher.

But if we invest in our national resources to go to NASA, I think we can do it. SpaceX. Because it's prioritized, it makes it harder for the US to invest in the space program.

Yes.

They can't invest in SpaceX and go. They cut off NASA.

Right? They don't believe in… Actually, the Moon landing project that's happening has been cut off quite a lot. And the management is bad. Moon landing. Install head of NASA.

Ambition. Mission. Policy. Really? He's gonna be more crazy than us. Yeah, but it's not just Eron who's blowing his ears. Right? And they fight sometimes.

Right?

Back to the data center. I feel that— Look at yourself. Because there's a lot of complications about data centers, whether it's power or water. Do you think Thailand can build a data center?

Why can't we? Of course we can. Should we build it? Data center. The data center in this region needs more. We are planning to do a good job. Now we are being taken to the market share in Vietnam, Malaysia,

with their stability.

India too.

Should we do it? 5 years ago, we should. Now, I'm not sure. But if you buy in what name? LUT? Data center? No, no. We have to prepare a to prepare policy to invest in data center.

It's better than this. There's corruption and stuff that makes it not that sexy. Right? So they choose to go somewhere else. Which is okay. But if we want to capture something in Thailand,

first, it's to make it more investable. Second, it's to make talent stay with us. Talent, I came back here, CK. First, it's more investment-friendly. Second, it's about how to keep the talent with us. I've been talking to people all over the country. They don't work for Thai people. These people are very talented. If we can get them to start up, let's say 10 people, that's enough.

We can go far. But we don't have the money to pay them.

That's what I'm trying to do. I think that— Honestly, I think Thai people are talented. But Thai people are mostly talented. They don't work for Thai people. But I understand Poon Arm's point of view.

Like,

he's doing this for a living. He doesn't have a job. He doesn't have a job in Thailand. He doesn't have a job that— In the end, it's a collaboration. But it's a less sexy project.

If you want to work on data center GPU, there's no such thing in Thailand. If today there's a data center project, and today in Thailand, boom! Data center project. Would you be interested in it?

I don't know why we're bringing this up. It's like, data center demand is for serving something like cloud. It's not that sexy to bring it back. Data center doesn't require technical. It's more about management. Talk to someone, set up a place, how to get the staff.

It's not a technical challenge. It's not a technical challenge. It's not a technical challenge. It's not a technical challenge. It's not a technical challenge. happy. When we want to improve our country, we have to face something. We don't know anything. Even though we have talent,

but this talent can't work with some kind of schema in this country.

It's hard. If you say outer space, if it's a world-class project, most people will go to Silicon Valley. Because it's easier to apply.

Yes.

So... To be honest, what projects are there in Thailand that are happening that make you feel like... For example, it's impossible. Space program is impossible.

This kind of project is certainly a value. Thailand is impossible. No one can support it here. So what are the projects that... If it's possible, that make you feel like,

Hey, I should come back to Thailand and do this? Actually, it has to be software only. Because we don't have money to buy hardware. Software comes from the head only. And we have a lot of talent in this. If there is a capital, it might not be a lot. If you want to start up something in the AI pipeline, whether it's data cleaning, data feeding, or we don't have money to train a big model like Facebook, Meta.

But we are part of the pipeline. And it makes it like, if you want this, you come here.

It's cheap and good. I think you can do it.

Interesting.

Yes.

Let's go back to US. And then come back to Thailand again. It will be out of the topic. There are many conspiracies. Many theories. So I asked him.

So I can really know. There is a rumor. It's a rumor. Like what the CEO said. I have to listen to it. There is a conspiracy theory that Elon Musk is the fastest data center builder out there.

This is the XAI's strong point. XAI said. Colossus.

Working simultaneously. He's very good at driving people to do it. But I might over-claim a little bit. Normally, 2 years and 1 month left. Actually, we have to count when. If 2 years and 1 year left, I believe. 2 years and 9 months left, okay, very good. 2 years and 2 months left, it's impossible.

Why is he so brave to claim that?

I don't know either. Can he really do it? I want to know. No, no, no. Just procurement is wrong. I mean, if you want to build a GPU with 80,000 units, you can't build it in 2 months.

In 2 months, right? The technology is not finished yet. The new GPU, sometimes it's still tape out. It's not even in mass production yet. Production. Cable. GPU. Cable. Cable. Cabling.

Set up hardware.

Diagnose. diagnose. AI that... Now, every AI company has their own unique point of view. Their own unique point of view. Some people focus on the model. Some people focus on the GPU. Some people focus on the data. They say there's data that no one else has. Let's put it that way. Which one do you think is the most important?

Data.

Data is the most important. But data is in...

It means that it's a proprietary data that others can't access. Twitter. Elon Musk me. I think this is the best way. AI is good or not depends on data. Garbage in, garbage out. If you don't have data from Reddit, your AI might not be able to do much.

Reddit has its own AI chatbot.

Yes.

Which… But it doesn't…

Reddit doesn't have the money to train a big model. Right?

You're saying they're waiting for a buyer? Of course. I've been waiting for a while. I think that if it's not the data access deal that exists now, they can sell the whole company. Now they get money from the AI company.

But Sam Aumann is the owner. So it's natural for them, for OpenAI to buy it.

Uh-huh.

But they're not buying it now. for them, for OpenAI to buy it. of XAI. I bet that Google will win. Why? Because Google has everything. Google has… They just need a good direction. Gemini was a bot at first. And it couldn't fight. Now it's so tough.

The airline came back to control the AI. And now Gemini is so tough. It's because they have the resource talent, the data. And how can other companies compete with Google's data? Google has been doing search engines for a long time. They have been collecting for years.

I think Google has the most hope.

But it shouldn't be Winner takes all, right? It should be different from their own market.

It should be, but I think Google will get the majority of the share.

And OpenAI?

I think OpenAI only gets first player advantage. Actually, it's not first player. I don't really believe in OpenAI. First, AI doesn't have many modes. Mode means different techniques. Everyone uses the same technique.

Everything is open source. Open source. Strategy. Open AI. GPT-C, GPT-5. Open AI. I think it's a very good thing that China can do it. At that time, it means that there was no more mode. OpenAI has received a lot of money. Let's see how it will be.

They have this thing called Mindshare. Chachavidee is good at it. He wants to be a consumer product. There is a reason why Chachavidee is very behind. Ask him, this question is so good. How can you ask? He deliberately to be a consumer mark product.

If you ask the villagers today, who do they think of AI? Open AI. Because they came first. They want to capture the mindshare. Even though they're not the smartest,

but they're the most distributed.

Yes, and they're famous. OpenAI is really cool. I was the one who released the chat GPT. Actually, every company has this technology. But every company, like Google, has it. But they have to wait for the rail guard to be in place. Because it's still hallucinating. OpenAI said, let it go.

Because it has nothing to lose. It's smart in that move. But it makes all the friends leave the cards. At that time, Google had to write their cards. And I had to write down the money to fight. Microsoft had to write this and that. Or even buy OpenAI.

So I thought, if we look at the talent of OpenAI, they're not that good anymore. They're all bought. Right? Their key man opened his own company.

And then we were sucked into the meta again. Right? So I think OpenAI is is having a problem right now.

Speaking of the meta, I'm actually the most disappointed player. It's Mark. Mark because, First, he spends a lot of money.

Yes.

He sucks talent. But this one is... Actually, he has data that others don't have. Right? He has social media. He has video. The closest form is social media. There is video. The closest form to real world data is video.

Second, talent. They invest a lot in talent. But they can't create open source. That's good for China. Why?

I think Rama is good. I think it's not bad. But it's heartbreaking that one of the most powerful AI is not on the list. I think it's a shame. I think it's a shame. I think it's a shame.

I think it's heartbreaking that one of the most powerful AI Lama is not on that list.

I'm not sure. Maybe he's training. He'll wait for another 2 months.

AI is the smartest every day. Every week. If you believe that this is not a winner-take-all market, and you feel that OpenAI is a company that you're a little sketchy. Because there are a lot of valuation. I have to ask you this question.

Is OpenAI compared to AMD?

Yes.

Right?

Yeah.

AMD is leading. Many people say that it is recycling money. It means that OpenAI is compared to AMD. AMD is ultimately, OpenAI has to buy chip A. They are customers anyway. They are partners anyway.

Same thing with Nvidia. Same thing with Oracle. Oracle is investing, but in the end, Puck is using Oracle's cloud anyway. Oracle is recycling money. Nvidia is investing with them.

In the end, they have to buy Nvidia's money anyway. Yeah. I don't want to go to work. You made a lot of money that day. A lot. Well, the first step is to feel happy. The second step is like, what is this deal? If you read it carefully, it means that they will invest in buying our chip. But they have the right to buy our stock at $1. No matter how much they take, it's a strange deal. How do you feel?

Weird. Weird in the way that… What is it? In the end, I found the answer. It should be this. It's like a partnership.

I mean, I'm his, he's mine. And we will go together. It has to be this way to compete with Nvidia. Jensen also said that this deal is kind of clever. But I wouldn't do it. But it's clever. Clever or not, we'll see how it goes.

But I feel like it's weird. And it's not just this deal. OpenAI goes everywhere. Oracle.

It makes me feel like it's a pop-up stock. Because now OpenAI is not in the stock market. It means that there are a lot of value generated here. Oracle is in the stock market. Partner, Oracle, but OpenAI doesn't know what it is.

Yes.

There is no real deal. GPU hasn't moved at all. There's nothing. Oracle.

So you believe there's an AI bubble?

Oh yeah. When will it break? And how much will it go down? I think it's too much now. But in the long run, it shouldn't be a bubble.

There's a lot of internet pop-up in 2009. I think so.

But in 2015, it went up and down.

But in 2025, it will be the same. I believe it will be the same. It will recover. And I think the base of AI is higher than the base of dot com. Because it's more beneficial. And the investment is the data center, energy.

If AI doesn't work out, these people can do other things. So I think the base is better.

Actually, AI revolution is interesting because I feel that the real winner who got a lot of money, except for NVIDIA, is Cloud. Cloud won the competition.

Hyperscaler.

Because the hyperscalers, they won anyway. You can train AI to subscribe to me anyway. Because IBGPU and everything. So Cloud becomes the biggest winner. Whether it's AWS, Google, or Azure. So they partner with every model

in order to do that. So, actually,

the business is very good. And then invest. Invest.

And then continue.

To buy GPO. And then it goes up. Investment. Oh, it's scary.

But actually, these things are not hard to see. But it's still popular. Because there's a debate between bubble or not bubble. Is it real or not real? Because in the end, it's different from bubble.

Bubble is InnoNA business. No one knows. Create a website, and it's popular. And it can be listed on the stock market. But this is the best company in the world. We're talking about Facebook, Microsoft, Tesla, Apple.

Not Apple. Apple is embarrassing.

So cruel.

It's embarrassing. We're talking about a company with a very high fundamental. Yes. If it's a year-on-year, monthly active user growth, Facebook alone is 30%. Okay, Meta makes it work.

Meta means that Glasses, Ray doesn't work much. Rama doesn't work much. But in the end, the cake, the main cake, grew up to 30%. Google, while Google grew only 15% in the search. So, the fundamental is that it can though it's not money-saving, but the main base is still very strong.

Yes, but how much is the interest rate?

PE still makes sense. PE is not that bad. Let's see PE.

Good NVIDIA, good Meta.

Not yet.

But NVIDIA can really save money. NVIDIA can.

NVIDIA Cash, it clearly sees money. I can understand. Immediate cash. I can see the money clearly. I don't have to answer anyone's question. But there are other companies, like software companies that say they use AI. The price is also going up. It's similar. I think it's a bit over.

Now, my opinion, or not a bit, I don't know. But when it falls, okay, we can survive. From this bubble.

One more question. I like that NVIDIA position themselves as a GPU seller. If they are a good seller, they should not be competitors with their customers. Because NVIDIA might not be able to push their own model that much. Right? They will focus on selling GPU to others to create their own models.

In contrast, Google wants to be the master. But they have their own model.

You mean TPU, right?

TPU.

And the latest news, the NVIDIA industry, is that Meta decided to buy TPU from Google. But I feel that, first, there's nothing that TPU and GPU can do together. That's one. And second, being a competitor's father,

do you want to rely on your competitor's father too? First of all, TPU, no matter what, it has to be for Google. If the competition is fierce. So I want to know, the thought between Google making TPU

and feeling that the good father should be the competitor.

I think like this. I think TPU has been used for a long time. It's been used for 10 years. It's made by them. Actually, it started because it's an internal tool. It's a cost-effective hardware that's suitable for their work. Right?

I'm excited that other companies are asking for TPU. Actually, it's not because TPU is good.

It's because Compute is out of the market. So you have to buy TPU.

Yes, TPU is out. Nvidia is out. AMD is out. And we need Compute Power. It's not enough. We have to buy TPU. AMD. Compute power. Sure.

Google. Google has been using this.

Makes sense. But if it's a TPU of other companies, like Broadcom, There are a lot of things like this. Let me ask you one thing. I want to know about China.

But some people might say there is. Maybe people are black market. I don't know. I don't know right now. But for sure, China can't buy GPUs now. Today, there was a news that it can be bought. But they bought the lower-end one. Not the latest model of Nvidia.

Why does China have a lower-end GPU, but can build an open source like this? That some American companies want to build open source, but can't.

Thailand is also very good. In the past 10 years, Chinese students who came to study in the US, they all went home.

Really? They love their country that much? Yes, everyone went home. And the Chinese university that did research on this, they were also quite impressed. Now, the technique or math is the same, right? Like DeepSeek, they found a new method that is much cheaper than before.

Mixed-resource.

Yeah.

And I think it's about this, the technique. Mixed service expert. But how can we make it faster? We have to find a new way. The Atomic Surf Expert's architecture is not able to build AGI. And that means China gave up on the AGI dream.

We don't know what they are researching. Right?

But they don't have GPU, they don't have hardware.

It's like we don't have a car, but we have to go to the same place. So we have to think about what to do. Do we have to tune our machine? Do we have to think about how fast we will go? At what time? While another person has a car, we don't have to think much.

Right? How much is our oil? We have to think about all of this. Because we don't have that much computing power. Everything has to be optimized to get further. to optimize. I use high-end computer. I use a mid-range computer. The frame rate is a bit lower.

But in China, it's like the gamers in the past. They have a 16GB RAM. They have to use it all. I think so.

But the question is about AMD. I want to know that when AMD or AMV allocate GPU to each customer, is it first come first serve? Or relationship based?

I think it's relationship. I don't know. I'm just telling you. But from what I've seen, it's more like relationship.

So I'm thinking why Elon Musk can make 1 million GB of customers because of relationship. Okay. So it's relationship. RelationshipBed. What is it like? FASCWORK like this. Lisa said you have to fix it. The client called Lisa. And Lisa asked who did this?

It's me.

She called you?

She didn't call me. She sent her assistant. And she brought her assistant to ride my neck. She said, hey, what do you have to do? How to update? And then, but I think she's a very nice person. Lisa, in Town Hall company, she's a very humorous person.

And she's a person who speaks without any words. But when she's in a rush, she has to make it happen. I'm under pressure. I don't want to be with her when he's serious, when he's angry. Because he's also very strict.

Being a software engineer in AI industry, I believe that— You have a very detailed background. I believe that there are many top companies trying to recruit you. I want to know how they recruit in this industry.

Oh, LinkedIn. Here it is.

Is there anything that can make people get out of the research suit?

Of course. Come on. Like NVIDIA, they are the top, right? They have a lot of options. They don't send, right? They have a lot of options. They don't really send recruiters because they have a lot of them. NVIDIA might not have it.

But I have other companies that send in a lot of work. Companies that want to do AI, companies that want to use high frequency trading by using GPUs, they have some. If I get the money, I'll go.

I don't have that much loyalty.

And I feel like there's a chance that you will want to start a startup. A small company. Maybe you'll be in a big company.

Or…

It's inspiring, right?

Yes.

To be with AMD.

Yes. Center of all the chaos. Startup is suitable for only one age group. We don't have to sleep. We have to go out and work hard. We have to work in the morning. I think we can do that until we are in our early 30s. And when we have a family, we can't do it anymore.

So the opportunity to start a startup or be a startup CTO, I feel like it's not sexy for me. Even though I have a lot of money, but when I'm with my wife and kids, I can't exchange anything. I see.

So that means that window is about to be gone?

Yes. Otherwise, I have to work life balance. Which is the opposite of a startup.

Yes.

I understand. I want to ask ask what is it like to work with the best engineers in the world?

Oh!

I want to know if there's a problem, how to solve it, how to communicate with the world-class engineers.

What is it like?

It's really good to be in the same room with an engineer who is so good. You will know that this person's experience is so valuable. We will try to solve this problem. We will think in our mind, if it's me, I will do this. And when we see how they solve the problem, it's very detailed. It's like… We can see clearly that they have been through a lot.

They can make choices that are not make sense or not right from their experience. I feel that this is the point. Just being a day bug in this meeting room, you can learn a lot. I think this is more valuable than the money I get. If I give you a clear example, it's like debugging. They will come and say, okay, this problem. First, based on their experience, they can solve the problem immediately.

They say, okay, this is not right, look here, I'll fix it. Is there anything that we can prove that it's caused by this? Is there anything that can prove that it's not this? You go and do it. And there are 1, 2, 3, 4 things that are clear. And if it's not this, what is your next plan?

And how do you talk to the management? Because we are in the technical side. We have to have layers with the management up there. Because they are non-technical. How do we buy time for our engineers to work? People skills are also shown. The person who runs the meeting, is it a product or an engineer?

Engineer. There are two. There are two ways to work. We know that this person is technical. If we like to work with technical people, because it's straightforward, right?

But if we have to go to a meeting that is non-technical, like management, there's another step that we have to go through. Because we can't explain technical to them. We have to simplify something. In the end, what they want is when it's done. What needs to be done, what's wrong, and report it.

If it's technical, I really like it. It's like, what is this technique? Some people develop a tool within 2 hours to diagnose this problem. Most of the meetings, the people who run the meetings, shouldn't call the meeting to brainstorm. They should have a solution already.

Just communicate and everyone will know.

That's not true. Because there are a lot of problems. This escalation has happened. If we's a problem. This escalation happened.

If we are a high-profile customer,

we will take the problem to the middle. And they will take the room. What do you think?

How can we brainstorm to make the meeting not too long? For the work class engineer. Because if there are 10 people in the room, everyone will show their work. It will be too long. 2-3-4 hours. Yes. Everyone is giving a long speech. It's 2-3-4 hours long. How do you make the session short and effective for World Class Engineers?

Actually, it's that long. If I let him speak, it's long. Actually, I'm the one who hit him. I was like, okay, stop talking. Everyone, I just said that. I have to do that.

But the problem is that World Class too much. World Class Engineer also talks too much.

I want to know the expectation level for...

For being in that environment.

When you work with a very good person,

who has a good cut, a good aim, before you can stay in that room. Like, if you're someone who... can't make it in time, will you know right away?

They will.

I know what they're talking about. They're like, this guy is late. It's obvious, right? But they will have a way to make it, you know, softer. Or they'll navigate it. We'll recognize that in that room,

there's someone with talent that we can rely on. And we will rely on that person. All of them.

And if there's… There's already this. Intonation in Thailand, I'm listening right now.

Yes.

They might not have a chance to do an AMD, not a chance to go to Silicon Valley, not a chance to be in that room. But they want to be one of them. He wants to have that kind of thinking, that kind of thinking, that kind of problem solving.

Can you break down a little bit, any tips for engineers to know?

Actually, first of all, I think that many Thai students have enough skills to work in that room. They just don't have a chance. It's not like there are world-class companies, there are world-class engineers.

I've been there, I've seen it from many countries. I've proved myself to be one of them. I'm here, I'm cool. I'm one of the top guys. Right? And there are engineers who don't want anything.

There are. But Thai good students, I think they are comfortable in that room. They just don't have a chance. Now, the sharpness of engineering depends on experience. The more difficult problems you face, the more you are in a meeting room with problems that are like, Oh, there is a problem in the world? If you are in there, you will grow very fast.

Which you can't find here. It might be a case in Thailand, if you implement a prompt pay system or implement something that is national level. You have to work, try to get yourself into that skill. And you will learn from senior engineers. And when you encounter a lot of problems, you will grow. In the end, if you want to go, you can go.

I feel that you can do it. Thai kids can do it. Is it in the library?

Yes. There's no problem. It's just about the visa.

It's just about the visa. It's just about the visa. I think our talent can do it. If you go to the bank in Thailand, or you go to the warehouse in Thailand,

our kids are good. in our bank or in our warehouse. We're good at that.

But how… If you have that kind of experience, how do you get it to be in Silicon Valley? Yes, that skill is not here. I'm talking about the global launch of something product that users will come in, you know, 100 million people at the same time. You can't find that skill here. But you can accumulate experience from small skills here. something product. Now, every software engineer is familiar with, almost the closest friend. Whether it's a cursor, codec, core code.

Now everyone is their closest friend. I feel like without them, I can't work. How do you feel about the software engineering in the next 5-10 years?

Oh, I don't dare to predict. Because I've been to a lot of places in 2021-2022. I wrote code like oh, you. There's a unique problem that there's no data in this world. Because it's that level of problem. The bleeding edge of technology. It can't be done. Can it replace mid-level engineer?

Can it replace junior level engineer? I think it can. Nowadays, when I want to do something, I write tools easily. I usually spend 3 hours writing tools. Open Stack Overflow to look at Python syntax.

I tell Cloth, I want you to write Python script for me. to stack overflow, syntax, Python. I feel like it might be possible. We might all lose our jobs. So the person who will survive is… It has to be in the design level. We have to have a lot of experience to know where we will direct it. We have to know if it's a bullshit. Sometimes the clause comes out like this.

Hey, you can't write like this. It has to go like this. It has to be used like this. How do you it going today? If it's better than what it is now, I don't know how we're going to go. Do you think the job will change?

Do you think the role of software engineer will change to prompt engineering or…

You know.

You can't write a prompt if you're not an engineer. You don't know what you don't know. The problems you face in production, and other things like that, I think AI can actually fix them for you. If you order it correctly, you don't know how to order it. So I think the experience of engineering is still necessary. But if you ask me, what about the new generation that doesn't have experience yet?

And how will they find experience? I don't know either. It's a worrying thing for me.

For students who are studying computer science, they are dreaming of becoming a software engineer because it used to be a dream job for everyone.

Yes.

What should they be thinking now?

AI is useful. It's a tool. Now you can create a project in just a few problems. But when you create it, you should go and figure out how it works. I can actually use. And then find the experience.

That's it. I don't know how the next study will be, CK.

That's very interesting. To go back to the point we talked about earlier, the core of each AI is divided into 3 parts. The first is the model, written by Algo. 2 is data, 3 is GPU. GPU is relationship-based, right?

First-server, I'm not sure. 2 is data, some people might have some leverage. The last one, the model, is something very interesting, so I want to talk about it today. The way of Mark Zuckerberg, he thought Algo must be written by people. So he tried to find a lot of people.

Find the best person to invest money. E-mart may go to you one day. You're welcome to join us if it's possible. He thought he needed the best person to write Algo. But on the other hand, I think what Elon did was

he wanted the biggest GPU cluster to let AI write its own model. I feel like which method should be the best?

I can't answer that. I feel like both methods are viable. If we look at how humans grow up, we have tools like a hammer. What do we do with that hammer? We have to bootstrap ourselves.

We make tools to make tools to make tools. We make a semi-line. Yes. The AI that we have today

comes from the bootstrap that we've been creating for 10, 20, 30 years.

So the idea of using AI and creating AI is also valid. Today, it's from the bootstrap that we've progressed for 10, 20, 30 years.

So the idea of using AI to build AI is also valid. The idea of using smart people to put direction, I think it's good too. So I don't know what will pay off more, but we have to wait and see in the future. If I were to give a bet, I think it's humans. I still believe in humans. Because we have a very breakthrough thinking.

New optimizations in AI are still in humans. If you look at the news that AI can create new things for this world, there are still very few new drugs that are made by AI.

There are no such thing.

Clinical trials or whatever. But there are very few news about this. But new innovation is still coming from people now. But in the future, it may not be.

Yeah.

Very interesting. Actually, our world is changing. Now our world is changing every day. Our world is changing every day. I feel like AI models are coming. I feel excited immediately.

Do you feel that we are far from the physical AI? Or is it really close? It's difficult. Physical is a very difficult field. Let's leave the AI aside. If you want to write a robotics program, it's very difficult. Do you know why? Because it's physical.

The motor is fighting every day. The battery is not paying the electricity the same every day. The hardware is fighting every day. Hardware is always in a war. You can't write code for 3 days. It's impossible.

It's a war and the circle is not the same.

Really? Yes. It's very difficult, physical AI. Because there's a risk of hardware?

Yes. You don't know how much the hardware works at the temperature. Some days it's too hot, the light goes up, it goes up 0.1 volts, the hand is drawing harder than before. There are many external factors. 0.1 volt. External factor.

We can overcome. Feedback. Sensor. Physical AI. to this. So there are many pieces to be put together to make physical AI happen. For example, if we look at the humanoid that Elon Musk made, it's still not possible. Because there are so many challenges.

In engineering, it's called degree of freedom. Degree of freedom is 1, 2, 3. We're talking about 6 degrees in the real world. Digital freedom. Because first, we haven't invested a lot in it. We focus on LLM more. So it's a little bit of a bias. And second, the talented people will run to the resorts to earn money. They will do that first.

And then in a while...

Can the car drive itself?

Oh, it can.

Oh, it's full self-driving now. I can tell you that it's amazing. If you ask me 2 years ago, I said it was like a high school student driving a car. Now it's even sharper than me. The same hardware. I bought it in 2019. The same hardware.

What car?

Tesla Model 3. I haven't tried it yet because after that that, it came out with a new hardware. It's a hardware 4, I think. Uh…

That one is probably better than that. There's a version that my car can't use. But the version that my car can use is already good. If you ask me 2 years ago, I wouldn't believe it. Now I feel like, hey, there's a picture. Mm. So you think you can drive by yourself? Definitely. Definitely? But what year? I think that

the year you went to Sukhumvit, Thailand.

You can drive by yourself for about 30 years. Sukhumvit is a bit risky. Because we have Grab, motorbikes, right? It's quite difficult. If I have to start, I have to start in the US. Because when it comes to the highway, there's nothing.

There's not that no external factor. There's no motorbike that can be a competitor. I think that now, the current system is fine. But we have to test it on a big scale. I think that if we test it, there shouldn't be any problem.

Now my car can take me to a restaurant and park for me. It's up to adoption. And in the adoption and scaling up for 2 years.

That's very interesting. You can drive your own car.

This is a project that I'm following closely. But I can drive my own car. The degree of freedom is not much. It's just on the road. But if it's a low-mission robotic, it's probably harder.

Yes.

NVIDIA just launched… What's the name? What's the code name? But NVIDIA just launched an open source model that can plug in all Mercedes cars. Because what NVIDIA doesn't have is Revo Tata. So they intentionally let manufacturers of all cars for free

to get Revo Tata like They intentionally let manufacturers produce all kinds of cars for free so they can get Vivo Tata and Tesla. It's so great.

Which...

It's going to be like too big to fail. So over time, actually, today is too big to fail. I have a personal question.

Yes.

I don't know if I dare to answer. It's the time of the city. I'm not asking which party you're in. But I feel like... I feel like... I'm in the political circle now.

I feel like all politicians should be worried about the country. I think we rely too much on tourism. We need to find a new S-curve for our country. I don't know how long tourism will last. I hope it will last for a long time. But we should have a new S-curve.

Whether it's culture, food, tech, or anything else. It's a financial hub. We have to create something new. Otherwise, we will be stuck here. And our country and our friends will go far away. They will keep going away from us.

Do you think the AI data center process is still in time?

It's in time. But you don't go as an AI model or hyperscaler. You have to be in its pipeline. Data cleaner, collection, or whatever.

The volunteers in that pipeline.

Like Scale.ai, they don't do much.

They just… Data labeling.

Right?

Actually, we can do it from here.

We just need talent.

And we already have talent.

But if we can sell it,

Is it hard if we are the government? we don't prioritize energy. we have to add AI pipeline. We have to have something like that. And let the kids try it. And then some ideas will come up. And it will be a startup. Right? It has to be that way, I think. And if,

there's a president,

He said,

Hey, can I have some advice?

Can I have some advice?

You only have 2 minutes to minutes, what would you recommend? I think we should allocate 1-5 billion baht to bring it to the tech sector. Thailand has a lot of tech talent, but we can't bring them back. I've asked many people, and it's about money.

If you invest in this sector, you get a rebate.

I think the company will use this money to hire talent back. Second, you don't have to spend money. Um… Um… Um… Um… Um… Um… Um… Um… Um… Um… Um… Um… Um… Um… If everyone is accessible, and data cleaning happens naturally. Because when we have data, we can clean it and feed it to AI. It can develop the country in many ways.

Actually, we don't need to train AI. Just asking for access is enough. We can query something. Because we see the data, it's actionable. And our country will develop by itself.

Okay.

Next question.

I feel that for those who are regular employees, and most of the listeners are regular employees. what is the best time to quit your job? I feel like the best moment is when you're done with your work.

What is it? Whether it's a haircut at home, logo, website, yin, ad, camera, photographer, editing, all the services you want to have in one app. Support each other. Before you finish, don't forget to download the app.

Oh, there are many steps. The steps that I feel like I've completed one mission. Whatever that mission is, like a project that I want to do, I've already done it. Or the second is I've got the skill that you want to do, you've already done it. Or the second is that you've got the skill that you want. Next, you will take this skill to the next level.

The third is the human step. For example, if suddenly your boss leaves, you should try to open a new opportunity for yourself. Because if you have a new boss, you can change the company.

And should you choose the your job or your boss?

It depends on you. It depends on what kind of person you are. If you feel comfortable working with this boss, and the job is not bad, it's okay, you can go. You can choose your comfort. But you have to be like, if you're chasing growth,

you should choose a job. What does it mean? It means it's a job where you learn some skills that you want. — and I will move to another company in the next year. It depends on the career path. It depends on your comfort. Some people just want to have a dog at home and have money. It's OK.

Do you think a dog is a good dog or a good expert?

A dog. I think a dog should be compared in a long game. There are more options for you to choose. And when you see a big picture, I think if you're a management level, you can do it well. Because you know this, you know that. You don't know much.

When it's time, you can ask an expert in that field. But when you see a big picture, you will use your skills to create new things from your knowledge. I believe in ducks. I'm a duck. But…

You're a duck in ARMS.

I'm a duck. I will go with skills like… I like to talk, right? I can talk. My technical skills are okay. I like money.

I will… There will be a unique position. It's only you who will be the Venn diagram. I think it can open up many opportunities.

Is there any advice for people who feel burned out?

I have to ask why I'm burned out.

Well,

I meet my colleagues and they say, Change your job! Right? Or if the work is too hard, take a break a little bit.

Right?

Or change your life. Sometimes you have to choose if the place you are in now has a chance to develop. Have you tried everything you can do? If you can do it, you've tried everything. And you feel like, oh, life is not going to go.

Or if it really affects your mind, like you can't sleep, you can't eat, you can't sleep, can't eat, can't sleep. You have to get yourself out of it first. And then you come to find another way to survive. I think the mind is important. But if you come out and you can take care of yourself for a few months,

you can think about it.

Yes.

Mmm.

The last question. The listeners of this channel, many people have a follow their dreams. Yeah. Do you have any advice for people who follow their dreams?

Just keep doing it. Actually, it's not that hard. We can dream as far as we can. And try. If you have a goal, you have to think about how you can reach it.

Not like, Oh, one day I will be this. But in the meantime, you have a long-term goal. What is your short-term goal? You have to plan it to get there. And you just execute it.

Keep executing it. In the end, if you don't make it, you will be higher than before. And you won't lose. You will learn. You will progress. And at that time, your goal might not be there.

You can change your goal. You can change your goal all the time.

It depends on whether you're happy with it or not. Yes.

I have a question. I don't know if I should use it or not. I've been wondering for a long time. I see your face every day. I see your face every day. Where? Which platform? Every platform. I see your face every day.

And I see that kids, if they ask me, Hey, what do you want to be when you grow up? They want to say that they want to be a CK. Thank you very much. But I'm curious about your content style. The one you presented,

it's like, it has to be like this, it has to be like that. What kind of mindset do you have with these things? Or is it just like, you're just making content?

No, actually, I'm like this. Like I said, the way to make Thai people absorb the information as fast as possible, sometimes it has to be clear. Because sometimes when you say it in a roundabout way,

in the end, Thai people don't get it. I'll give you an example. S&P 500. I think Thai people know S&P 500 better than me. Because S&P 500, the person who communicated about S&P 500 has been around for a long time. I'll give you an example. It's been a long time. For example, P'Paul said, If you invest 1,000 baht every month in S&P 500,

you will get this much in 10 years. But I would say, if you want to invest in anything, but you can't win S&P, you're stupid. It's a very important way of communication. Another example is risk-free.

Risk-free is a US currency. If I say, If you have a lot of investment, but less than risk-free, why do you do it? At least you should get a return that is risk-free. Which is the US market share. It's about 4.25% now. It's a way of communication that...

Of course, when I say chip, it will cross the line. There will be people who agree and disagree. But that's for sure, everyone can add information. Like me or not, It doesn't matter if you like me or not. What matters is that they understand what S&P 500 is. And they understand that S&P 500 is not a bad thing. But at least I want Thai people to dig into the information.

Oh, okay. You are the person that I don't get. When I look at someone, I'm like, Oh, okay. Where are they coming from? But I don't get you. First, you said you want Thailand to improve.

Second, you're doing fast work and you're already rich. So I wonder, if we invest in… Why did you burn the money, CK?

That's a good question. I mean, no one asks me this question. with… Uh… Why did you burn the money, CK?

That's a good question. I mean… That's not a question that people ask me. I want to be a good example. Do you understand that if I can do it… I'll talk about how I can do it later. If I can do it,

it will make the next generation know that Thailand can create innovation. Because Thailand has never been a country that can create innovation. Culture of innovation is important. For me, innovation is creating something without a model. For example, Uber has a Grab. Grab has a Golf.

Amazon Shopping has a Golf. Everything has a Golf. Uber, Grab, GrabG business, related to service, I want it to end as soon as possible. Which, if I succeed, First, I want to spread it out of Bangkok. Because mostly, Thai people are not from Bangkok. They are from outside Bangkok.

I want them to go home, find a place, use a place, buy a place. And I want to show Thai people that the innovation can happen in Thailand. To bring people back. That's one. Second, I believe that... Many people like to compare that we are like Fiverr Upwork.

They try to get a freelancing platform. Which I see a big difference. I'm not trying to be Fiverr Upwork. Because I'm an investor. And here, if you look at Fiverr's stock, 866 billion.

I'll give you, round off 1 billion valuation. I work with 2 billion. Valuation. That's not, that model doesn't work. The model of Amazon, which is GP. It works. Amazon is very successful in terms of product.

They become an online online online product. About the product. But when it becomes a service, you are productizing human time. It means that one person can sell a million pieces.

But if I give you two hours, you can't give this two hours to another person. So it means that the way of selling is different. What Fiverr does is GP. They sell the same way. They sell people using Amazon's way. And that doesn't work. So the way to sell the Amazon way to sell people. And that doesn't work.

So the way to sell human time has to be different. This is what we have to do this year. But if we talk about roadmap, it will be very long. But to answer your question, I feel that life is valuable. I swear I'm talking from my heart.

I feel that life is valuable. It's not a life for yourself. If I invest invest my money in everything, investing in real estate, investing in risk-free, I only need 4.25%, investing in private equity, I get 12%. I'm comfortable.

But I feel that a comfortable life without value is useless. I feel that for me, people who are capable and have resources resources should be more responsible for society. Because they are the ones who can change something. I believe that this country can change. But politics is hard to change.

Because you need two things. I believe that changing the country needs two things. One is what the country cannot lack. Two is the voice that the people believe. One of the things that can be without is that the country cannot be changed. For example, the leading business in this country,

I don't want to say the name, but everyone should know. It has the thing that the country cannot be without. If it disappears, the country's GDP will drop immediately. But it has no risk to the people. The people don't believe it. They see it as a criminal. But they don't have the voice of the people. People don't believe them. They think they're bad.

But if you have the voice of the people, let's say the political party wins, it doesn't matter what color it is, you win. The voice of the people. Everyone likes them. But if you cut them off, the GDP of the country won't affect. That's it.

But if I can combine these two, I think I can change the country. But I need two things.

If I have the same amount of money as you, like you said, 4%, I'm in Bali. I'm in Malaysia, you're not. But you're not stupid, right? You're a very good financial analyst, right? So your investment must have a burn rate, right? How much do you burn?

Do you have a runway? How do you find the money to fill this part?

Personally, I know. I personally burn around 300. More than 300. I should burn more than that. Like I said, I believe in one thing. In the end, product, roadmap, product,

what direction it will be, that's for sure. It can be a moving part. It's so scattered. It's so scattered. It's very centralized.

It's so scattered. that can steal traffic from Google. If you want a studio, I mean, I want a studio for photography. Normally, people look at Instagram, Facebook page. Why is there no place that you search for studio?

Anywhere, it's all about how much the studio is worth.

It's all about the price. It makes the user feel like it's the only place for the service. That product doesn't exist on earth. I'll ask you again, what is your mode? That product doesn't up on human. Everyone is trying to build a digital god. First step. That's the goal of this year.

Drastically. I drastically change my product. I change my user journey. I change everything to find the product that market fit. Because I found the way to find the most responsive people. Scale to the whole world through fastwork.com that I bought. That is the key here. I have a product that market fit that I can scale.

But the most important thing is Fastwork, over time. You will become a friend who knows everyone in the world. But if you want a editor, you call your friend. Because you know that this friend knows everyone.

Hey friend, can you recommend a good video editor? YouTube editor like Arm. Oh, wait a minute. Wait 3 minutes. Wait a minute. 3 minutes passed.

These 10 people should be able to do it. The price is low. I'm a friend who knows everyone. No matter what service. Lighting, air conditioning, wood, paint, color, interior architect. I will find someone to help with the light.

Find someone to help sell things. Find someone to help with the ad. Find someone to help with the haircut. I have to rent a car, a condo, a rental car, a furniture, everything. Everything in one place.

Why Thailand? Why you can start from the US?

You're right.

I want to prove a point. Because this is my home. And I feel that… I look at it from the heart. This is personal interest.

But you're going to the mountains to prove a point.

I want to prove a point. Because if I go to the US, The skills are different. The skills are different. Because Silicon Valley is much easier, right? But it's not that easy.

But finding a talent boy is a mess right now. I'm in Thailand. I want to prove that Thai people, Thai engineers, can create innovation that can compete with the world. I think this message is very important. And if I can do it, I believe that…

It's like AI. AI, this path, is not named by Donald Trump. This path is clear because of the individual. This path is clear because of OpenAI, Launchpub, I mean, This is the way. And another thing, my main focus is not only to launch online, but also another main business that I can do is agriculture. I believe that the way to change the life of agriculture is to be an international company. There must be some rules, like the Farmer Business Network in the US.

Which can make everyone in agriculture can be one voice. That is what I'm trying to do. business network in the US.

The three people who combine the three factors of the economy and make the production more than every year. So that's the vision. I have Indonesia and Vietnam. Is it easier to go to other countries? If you want me to do software as in like, you know,

Competing with the US doesn't really make a lot of sense. The area that I see is interesting that the world is not looking at, and the world actually gave up on, is human capabilities. The opportunity to hire humans in the future.

I was inspired when COVID happened.

During the COVID, many people went back to their hometown.

And because it's stuck at home, they can't do anything. And I saw that most of the money I distributed went to other provinces. I transferred to Khorat, Sukhothai, Chiang Rai, Hat Yai. I saw that most of the money was transferred to banks in other provinces. So it made me feel like, Omit Thanakarn. Shit. If I can make this actually happen. I want to get them out of Bangkok and compare with Bangkok. My job is to make people in Thailand to be able to live in peace. That's the important thing.

If I can spread my idea outside Bangkok, that's the most important thing. Thailand is only in Bangkok. 99% of headquarters are in Bangkok. But in the US, what's in DC?

Nothing.

Amazon Seattle, Silicon Valley, Texas. DC has almost nothing. But everything in our country is in Bangkok. So the water quality is high. Because the water never goes out of Bangkok. We use unvaluable land. There are only land in Bangkok that has value.

Outside of Bangkok, I'll give you Chiang Mai. That's it. Nothing left. In addition, Bangkok. Bangkok. Bangkok. Bangkok. Bangkok. Bangkok.

Bangkok.

Bangkok. Bangkok.

Bangkok. I have 60-70 houses. I gotta find the product market fit. So the key here is finding the product market fit. So in the end, money is one thing. The most important thing is to find the best person. And try to do everything that no one has ever done before. And you find the right person.

That's it. We have to bet on that. Everything that I do, I do it for one moment and one moment only. It's the tragedy moment. Find the right person. That's it. And it can't be done if I have the best person. And I don't have money.

So it means that if you give someone who is not that stupid, who is smart, who has a certain amount of thinking, who is not stupid, you can let them try right and wrong. For example, this is the right way. And they don't know how to go.

But you give them a number of tries, but not limited.

I get what you're saying.

You will fucking find it. So the key here is, not in a rush. We will fucking find it. Because again, it's closest thing to the straight line.

So if we keep trying and trying, we'll find it. But I hope we don't do it the same every time. Learn from the deployment every time. And we deploy, use the feedback, fix, deploy, redeploy, redeploy. We'll find it.

But I don't understand.

If you hire someone to do a service, you charge a GP fee, right? It's not gonna be GP. So, that's what needs to be changed. Okay. The GP method on Amazon will not work for human. That doesn't work.

Because we know each other and hire someone. So, what happens is that we have to charge a friend's fee. What is a friend's fee? A friend's fee is? You call someone, Hey, can you give me some tips? Okay, 10 people. Okay, I can. And I also deal with that. I pick up the tip, but there might be a cap.

For example, 10% with a cap of 300 baht. I mean, I only pick up the tip for 300 baht. Every couple that you know. You know each other, I don't pick up. But I hope you can still do the transaction. We just have to make to cover your burn rate? So that's different.

So we're talking about the details of the roadmap product. So that's one thing. The nature of LTV. The second thing is that I have all the services. What happens today is search.

Search.

PLP. I have a list of products. And PDP. PDP is you find a list, will click on the product detail page. To see how the review is, how the product is. Inquiry and then payment.

Thousands of clicks. This is the chaos of like... You said you wanted water. I said, do you want water? Here, look at the water. I give you 4 bottles.

And then, all this. I have to look at each one. And I have to send it to each person. Wait for them to reply. It's so chaotic. Before I get the payment, 1,000 clicks. I have to explain it.

Is it better if I say that…

You posted it the other day, right? Bring it here.

Search. Instead of PLP, PDP, inquiry, I will let AI read every chat for 10 years. inquiry. Click on multiple choice. Click on Announcement. Announcement to everyone. That AI feels that it matches this job description. And then click on Finish. So customers don't have to inquire.

No need to greet each person. So it's gonna change the journey. So it's gonna be search, form, broadcast, payment. And then go to inquire. So that's what we're gonna do in March, April. So that every service can're gonna do. That bear the price. And the price is better because you can know the real market price.

Because they don't know how much other people bid. That's the key. There are many more things. I'm talking about it all day. We're doing so many things now. We are doing so many things. And I hope that this year, we will find part of the market fit.

That's the goal.

And I think The past year, my goal is... The first year is restructuring. The last year is recruiting. I'm not a recruiter. Before I met you, I was a recruiter. Recruit the best of the best of the best of the best.

And then this year, I've been preparing since last year. And this year has changed. When are you going you break even? I don't care. I find product market fit. And again, I'm doing something really hard. I could be— I could do anything. It's easier, right? But I want to do something hard to prove a point.

I can do it. Even though I can't do it, I believe I can do it. But I believe that in the meantime, even though I can't do it, I can be a good example for the kids in the Jura Dhammasat

when they're in the middle of the process. I believe I can do it. But I believe that I can do it. And I think I'm a good example for the students in Chulalongkorn University. Now, I want them to feel like, Yeah, CK is wrong about this. What should he do next? I think I'll capture the lessons.

Okay.

Something like that.

Now I get it.

Thank you.

You're welcome.

Who changed many things?

This year,

March, April, it's not just about finding people. It's also about booking. Let's say, for example, you want to,

what time is it? 3.30, if you want to go to the bathroom, right now, you wanna, what time do you wake up? Most people are last minute. And Samongkhueng, if you want to take a shower, you want to,

most people are last minute. They're gonna take a shower at 5 o'clock. What do you do right now?

I don't know, just Google and call.

Yeah, Google and you call. Are you free at 5 o'clock? Oh, no, the customers are full. Can you do it at 1 o'clock? One more number. One more number. Why not have a platform?

All the masseuse that is, all the site,

all the massage parlors that's available at that time. You can do it. That's it.

Right?

That's just a massage parlor. This is about printing. Actually, you don't want to print something. You don't want to print a card. Why do I have have 3 networking events? Attach this sign in, announce it to every pin shop.

And everyone knows how much the market price is. We became the market price. And I'm just talking about two service categories of the 2,000 that we have and more. It's an infinite market cap. It's infinite in terms of like how big this market can be.

Because we sell only limited services. But if many startups are just one business, It's infinite in terms of like how big this market can be. You're not gonna download 15 different apps for your daily life. You want to download one app, right? We have everything that Airbnb has. We can actually be Airbnb. So... I can do the same thing. I can do everything that Airbnb does.

But Airbnb only does condos. I can be every services. So, my market cap is infinite. Because I want to accumulate all service category.

But is there a reason why Airbnb only does rest areas?

They can't think of it. Because if you look at Airbnb's stock since IPO until now, they have no way out. So it's the reason why Brian Chesky just announced about 6 months ago that they're building the everything server services. Exactly the same thing that Fon is doing,

but I'm closer. I'm way closer. Because they don't have a chat log data. Because they don't have a chat log data. That one customer, one freelancer, one seller I'm closer. I'm way closer.