Hello, this is Electrokim. Today, I want to start with an interesting Google research paper related to Tesla, especially hot topics like artificial intelligence. In the end, it may or may not be related to Tesla to some extent, but I want to approach it in an interesting way. That's the advantage of this channel. Before starting, Kim James, thank you very much. Thank you, Jooyoung. Thank you. Does my voice sound different? It's a bit sleepy, but thank you. It seems like you've recovered well. My wife needs to recover faster. Now, coming back to the AI-related paper, Google took a step closer to General Intelligence. This is about the International Teen Olympiad. I'm not sure if this is well known in Korea. It's the era of mathematics, and the World Teen Science Olympiad is one of these competitions.

I'm somewhat familiar with it, but I'm not sure if it was well-known even in regular high schools. Mostly, now in Korea, students from science high schools have achieved quite remarkable academic performances. Excellent. That's right. Well, such things should be more active now, right? Yes, of course, good friends should go to medical school, but these days, it seems a bit... Yes, dragging a bit. Later, as I keep saying, so... well, solving these Olympiad problems with an AI model, Alpha Geometry 2, Alpha Proof, and Alpha Geometry. Geometry is important, but if I explain it in detail, it's not fun. If the story goes as planned and it's not fun, then basically, well, it's not just a math problem at this point. It has evolved to solving problems from worldwide math competitions. And compared to the previous models, the number of parameters is somewhat limited. Or maybe moving towards the direction of Natural Language Processing.

When you approach it now, commonly known problems from the BT era now start talking nonsense. So, they bridged the gap well, used geometry well, and solved it well. So now, they still make mistakes that are close to nonsense. So, they still make some mistakes that sound like nonsense, but they have reached a level where we can consider that as mistakes. It's quite a significant development, so AI development is ongoing. Google has turned it on, and further, before talking about the conclusion, Apple used a custom AI model, Google's custom Tensor, that is, Tensor Processing Unit (TPU). So, Nvidia's stocks plummeted and skyrocketed, ultimately repeating these fluctuations nowadays. So, dealing with stocks is truly challenging. In the end, Apple used custom chips.

This research was also done with a custom chip on Google. So, when you listen to these researchers' stories a little bit here and there, Google's custom chip was quite fun, so they saw some benefits. Nvidia has those general-purpose GPUs, but Google has been talking about custom chips for a long time. The Big Tech group is preparing custom chips because when Nvidia allows hardware to be pulled as they wish, in a limited situation like this, extreme performance can continue to be drawn out. So, this is about solving these problems. Additionally, Apple's stock revived, right? They have platforms like this, and Apple announced the on-device model, so now, even emails, by learning a bit about users' usage, they can track what they want.

Adapters and malfunctions tailored to users like this, so rather than saying that I used an app, now it automatically knows that it's not needed, and it looks at time and these patterns to really be like a secretary. Creating this, it's said that Google used custom chips. This also shows expandability. So, this is one example of how custom chips are well utilized. Not only this research, but also other big tech companies like Apple and Google are using this, not just for now, but for the future. The reason this was interesting is that Google also has a phone brand called Pixel, but it's not that exciting. So, it seems that people are more interested in Google's custom chips. Of course, Pixel is based on Samsung's Exynos, but they did their own thing there. Of course, the performance of that app is.

Something not good is that there is talk of the Samsung modem dam, and it's not accurate, but the performance is not that great. Actually, Market Joy has grown slightly, but now, since it's a bit mimicking, the performance is not as good as the iPhone's Snapdragon, which is well known like the HN 3 or HN 2. Compared to them, it falls quite a bit. Also, in 3D Mark Wild Life stress tests and such, when you look at it, the Pixel A Pro is still quite new. Of course, it's getting older now, but the performance seems to be dropping a bit too much. But in reality, if you look at Google ads, they emphasize what AI they will do, not so much emphasizing performance. So, is it because of that, or are they trying to expand the use of AI training more by bundling it up to the phone like this? Anyway, even if the influence in the phone market is somewhat lacking, people about this.

I didn't really think about it that way, but actually, it's producing this kind of effect. And this vertical, completely from the ground up, not only in software but also in hardware, designing it directly, achieving this performance, that's why it continues to make progress. I knew a little about this coherence, so I subtly mentioned it in the video as well. Yes, NVIDIA doesn't know it's high, it just keeps going up. This is where stocks are difficult. It's not just about the world going this way from my perspective. It's not a perfect perspective, just a partial perspective. I would appreciate it if you could take it as it is. It's not because I put too much emphasis on stocks. It's difficult to time it right, so even if you know this to some extent beforehand, it's meaningless. It didn't matter at my level. And the phone.

In the market, Google Pixel, yes, it's not widely used in the U.S., but people find it fun for research purposes. This is the opposite trend, and I wanted to share this flow a bit. For example, how does Apple maintain its monopoly every year? Of course, soon it will be 2024, and more than half has already passed. Time flies. If you look back, statistics will show this, but Apple is amazing. Samsung is still declining. Well, in the U.S. market, Apple creates expectations in the stock market when they do something in the high-end phone market. Apple is truly remarkable. Although Samsung is not bad at all, they need to step up their game. They should improve. Oh, yes, I hope they do better. I recently...

By the way, they've also released the Fold 6 now, and it hasn't been out for long, but if you look at places like Amazon, they even throw in gift cards, upgrade it to 512 gigabytes, and they're already offering discounts. Oh, I also thought about buying it, but many people still have a lot of doubts about its durability. If you look it up, even in the United States, there are still concerns that it might break too easily. However, this time it seems to be closer to a complete product, with a pretty design. I'm thinking it might perform well this time. Oh, in the U.S., there's still a dominant lingering sentiment about it. Still, I hope it sells well this time. Samsung seems to excel at making fancy stuff, but let's also look forward to their future advancements. So, the flow is like that, but now Tesla is coming out. Tesla is the double down.

I really liked the title. Tesla works hard on AI learning for autonomous driving, like Google, even developing custom chips. They buy from Nvidia and do both, so it's not Hardware 4 but AI 4, and they changed the name to reflect that. They are doing both, already around 50,000, not 100,000 yet, but now at around 8,000. So, they are running in a hurry. People keep saying that Nvidia GPUs are being bought a lot now, especially from earnings and Corona. In fact, there is a lot of interest in the stock market. Ultimately, for infrastructure investment, if you want to do AI training, Nvidia is overwhelming. How much Nvidia equipment they buy is a big concern. Tesla has not given up on Dojo, they are still doing both sides. I will work hard on both. Actually, but.

In the market, when you listen to developers now, they are so familiar with the architecture used by Nvidia, and because they don't know what will happen yet, that's what makes it scary. So, Nvidia is still dominant in the market, and developers are familiar with it, not giving up. And in line with this trend, Tesla is also doing both, so wouldn't it be good to complement each other? The road to AI is still long, so even though we don't know what will happen, but focusing on a narrow targeting in the category of autonomous driving, wouldn't Tesla achieve quite good results in this flow? In this trend, Tesla is keeping up, and Elon Musk, who is very smart, is likely to get ahead with such information early.

Identify and quickly catch the trends that lead the company in that direction. Doing big tech like this, and well, not deviating too far, but among the car companies, which company is so serious about AI? Where is it? So recently, from Morgan Stanley, now Tesla is being picked as the top. There is a lot of that going on. It's not a guarantee that this will definitely succeed. Where in the world is there a 100% guarantee of success? But because they keep showing this, not everything is to everyone's liking, but I also, so I continue to be a Tesla fan. So, what will happen if this comes into my car? I still haven't received it quickly, a cyber truck, and even the Model, since I am Hardware 2, I haven't received that yet either. So, the Model X that was thrown away, I miss it because it was hardware, it has already been updated.

But, oh, I regret a lot. By now, I should have received it already and I regret not working hard to create such videos. Oh, if Cybertruck could be delivered quickly, that would be really... So, the thing is, transfer learning is not that easy. When a new model comes out, transferring it to a different hard drive is not as simple as it seems. I mentioned this in my previous video, but my full pipeline has originality, generating data directly from consumers, being able to tag data that comes from consumers, is important. And still, for some parts, it has become much easier. Because of auto-labeling, there is no need to label. It's not that people are not needed to some extent anymore, but it's not as necessary as it used to be. It's easier now. Tesla has all the pipelines, but here, you can still customize and buy.

And you can now infer the power of custom chips through this kind of flow, Tesla did both, which is amazing. Yes, so I was impressed by some remarkable thoughts. Eventually, I think we need to check when this will happen. I'm also curious about when it will be. It's fun to watch this development as a researcher. It's a professional disease. When you see advancements in other fields, it's exciting. Due to that, there are supporters, and that's why I make videos. Lastly, the 4680 battery is now being produced in-house, including dry electrodes for the truck. It's not a wet process anymore; it's now a dry process. In fact, previously, it might have been from various vendors, but now it's successful in-house. Especially in Korea, there is some criticism.

There are so many articles coming out, all talking about passing the responsibility to the end consumers, but ultimately, it's not good for the car companies that produce batteries. I understand that, but criticizing that nothing has been done so far is ongoing, isn't it? This isn't about solely the solid-state battery, and of course, mass production is not easy, but to some extent, isn't this improving? It took a long time, but in the end, like the quietly passing thumbnail in the previous video, Tesla will quietly move forward, right? And now, various partner companies like LG and Samsung are working hard on the 4680 cells, really putting in a lot of effort. This should all work well together; even Panasonic is working hard. In the end, they are making it themselves, and the surrounding partner companies are also making it, so in the end, it should turn out well.

I wonder if in the past, when criticizing cylindrical batteries like this, those who criticized it a lot would feel so regretful. Now, they are putting all their efforts into making the cylindrical battery really thick, like 4680. Oh, that's a real love-hate relationship. Yes, criticizing this 4680 as an innovation, criticizing cylindrical batteries, and ultimately, what are they going to do, saying it's not working every time. But gradually, as Elon Musk mentioned in an interview, by leveraging what we already have, significant innovation can be achieved. Now, as we continue to talk about the possibility of achieving full electrification, of course, there are goals and aspirations involved. But, anyway, the fact is that partner companies are pushing hard for the 4680 cylindrical battery. Who do you think is driving this? It's the company that first announced it.

Who are you? So, anyway, instead of criticizing, let's all work towards doing well together. But that's not easy, still, I hope the 4680 battery is produced quickly. Yes, so the price of the Cybertruck, right? Rather than the original price, it should drop significantly, so that many people can afford it. I hope many people can feel the vibe of the Cybertruck that I currently feel together. It needs to sell a lot so that I can get the FS quickly and use it easily. Yes, when I see things like that, due to such reflections, I hope the battery costs decrease quickly so that electric cars can be more widely spread. From my perspective as a supporter of electric cars and someone who wants Tesla to succeed, I hope for a quick and successful spread. Thank you for watching the long video. I will come back with more Tesla-related stories next time.

테슬라 도조, 엔비디아 H100 둘다! 4680 건식 공정 안된다며? 구글 애플 인공지능 협력과 발전

테슬라의 모든것을 알려드립니다.
구독 눌러주세요, 꾸욱~~

제 리퍼럴 코드 입니다.
https://ts.la/young82399

테슬라 도조, 엔비디아 H100 둘다! 4680 건식 공정 안된다며? 구글 애플 인공지능 협력과 발전

테슬라 도조, 엔비디아 H100 둘다! 4680 건식 공정 안된다며? 구글 애플 인공지능 협력과 발전

스크립트