Author: Animal Team Leader Source: official account: Animal Team
The first depth bomb to detonate the AI revolution was ChatGPT.
This is the first universal AI robot with a monthly life of over 100 million. It can replace humans in writing, drawing, tabulating, PPT, and code writing. To some extent, ChatGPT can steal my job. Its intellectual growth rate is amazing, not only helping college students write papers, but also passing Microsoft’s software engineer testing.
Under the leadership of ChatGPT, AI tools in the United States have moved towards vertical development, with the emergence of professional AI live streaming and AI mapping. On major live streaming platforms, there are companies that group control tens of thousands of AI anchors every day, live streaming with goods. Some people will ask, is it possible for AI to carry goods? In terms of high-priced products, it may be necessary to make a real person appear and improve the interactivity of the live broadcast room, but in terms of affordable products, the main price advantage is that AI anchors are fully sufficient.
Some game companies have also required employees to learn AI graphics within a limited time limit and have begun to develop layoff plans. For large-scale boutique original games, original art is indispensable. However, for skin changing games, small program games, and leisure and entertainment games, AI graphics is completely sufficient. With AI tools, graphics efficiency has greatly improved, and game companies no longer need so many artists and painters, making layoffs inevitable.
Advanced productivity replaces backward productivity, which is an irresistible technological trend for mankind. On the contrary, in order to achieve more advanced productivity, countries such as China and the United States have to catch up and never lag behind. Because the progress of AI technology is not gradual, but explosive. If you fall behind, you will fall behind step by step. If you are left far behind, it will be even more difficult to catch up.
“We always talk about the information technology revolution, but over the past 30 years, the information technology revolution we have experienced is essentially an information application ecosystem built on the underlying technology pioneered by the United States.”. China has not achieved significant success in underlying technologies such as operating systems, chips, and EDA industrial design software, and has also been blocked by the United States.
Today, the information technology revolution is entering the deep waters of AI technology. Compared to the Internet revolution, the AI revolution has come faster than we thought. ChatGPT, like the Windows operating system, was first born in the United States, triggering a wave of human AI technology.
ChatGPT is very different. Since its release, there has not been a second application of the same type in the world that has reached the same level of technology as ChatGPT. OpenAI, the company that developed ChatGPT, is just like OpenAI, which is far ahead of all other human companies.
In the past information technology competition between China and the United States, we said “Copy to China”. Because in Silicon Valley, there are a large number of Chinese engineers. As long as Silicon Valley develops a new application and creates a new business model, Chinese companies can immediately follow suit and quickly produce products of the same level. Early companies such as Tencent, Alibaba, Baidu, NetEase, and JD.com all started this way.
In the face of the enormous wealth effect, countless Chinese entrepreneurs have followed one another and closely focused on the technological frontier of the United States. Whether it’s Copy or any other way, as long as the United States has it, China will definitely be able to catch up immediately. For example, in early blockchain technology, digital currency, cloud computing, etc., the gap between China and the United States in technology is not large, and related industries not only have a layout, but also representative enterprises.
But now, ChatGPT has become the only one in the world. Not only did China fail to do so, but other companies in Japan, South Korea, the European Union, and the United States also failed to do so. In the world, only OpenAI has been developed, and major American companies such as Facebook, Google, and Apple have not been able to produce competitive products of the same level. Apart from ChatGPT, which can be described as artificial intelligence, others seem to have an overly obvious “artificial intelligence disability”.
Simply put, it’s just not smart enough. Why is this happening? In my opinion, ChatGPT has three core advantages:
First, top chips are not limited to purchase. ChatGPT’s super computing power comes from the world’s only and top A100 and H100 smart chips developed by Nvidia. These two chips are irreplaceable globally. According to the requirements of the US government, Nvidia A100 and H100 have been restricted from exporting to China.
In order to continue to occupy the Chinese market, Nvidia has launched the Chinese special version of the A800, which is also the monkey version of the A100. There is a significant performance gap between the two. The Chinese special version of the H100 chip is also about to be developed and will soon be available in China.
This will cause American companies to use the top equipped A100 chip, while Chinese manufacturers can only use the monkey version A800 chip. The impact of this is that Chinese AI suffers from congenital intellectual deficiency and insufficient computing power. Analyzing the same data requires more time for Chinese AI.
Because the United States blocked A100 and H100 chips, how much inventory did China buy before the blockade? There is only one company that can compete with Microsoft, but it is not a technology company, but a financial company called Magic Quadrant Quantization.
This is a private equity company that bought 10000 Yingweida A100 chips. Magic Square buys so many A100 chips using a large model for financial management, with a management asset scale of about 60 billion. It is a high-tech financial company. The official selling price of the A100 chip exceeds 80000 pieces, but due to a shortage in the Chinese market, the unit price has been increased to over 2000000 pieces. Even if we had 800000 chips, the A100 chip in Magic Square Quantification’s hand would be worth 8000000000 yuan. Such high chip procurement costs directly exclude small and medium-sized companies.
OpenAI was able to do so because Microsoft invested $1 billion in it and then acquired OpenAI with $10 billion. Without Microsoft’s heavy financial support, ChatGPT would not be able to do it.
However, Japan, South Korea, and the European Union are not limited to A100 and H100 chips. Why haven’t they made them? This brings me to the second point: supercomputer support.
OpenAI is a startup company that cannot afford to buy or use supercomputers, but behind him is a father of gold: Microsoft. In 2020, Microsoft purchased 285000 CPUs and 10000 GPUs to build an AI supercomputer in collaboration with OpenAI. This supercomputer ranks among the top five in the world.
This is the 2022 ranking of global supercomputers. The computers on the list basically belong to laboratories and are mainly used for scientific research. Only the Azure supercomputer built by Microsoft is used for AI commercialization. It is almost impossible for China’s Shenwei the Taihu Lake Light and Tianhe-2A supercomputers to be used by a technology company to train AI models.
Currently, Alibaba Cloud is the most successful company in the commercialization of supercomputing in China. Alibaba Cloud has built multiple intelligent computing centers in China, and Xiaopeng’s autonomous driving model was trained by Alibaba Cloud. Ma Yun was exposed today and returned home. The first topic of conversation was AI.
I predict that Ma Yun is likely to bet heavily on Alibaba Cloud to promote the development of AI technology in China. Ma Yun is also likely to have seen the wave of AI technology revolution in China and the United States before making up his mind to return to China. A person like him who stands bravely at the tide and does not have big waves will not easily go fishing in the sea. Because the bigger the waves, the more expensive the fish!
Third, unknown intelligent algorithms. With AI chips, supercomputers, huge amounts of money, and a set of efficient AI algorithms. Currently, OpenAI algorithms are not open source, and no one knows what the underlying code of ChatGPT is like. This is the core unique technology of OpenAI and the technological moat of ChatGPT.
In the face of this situation, can China break through the US blockade and catch up with the AI technology revolution? Here, I will mention three companies.
The first is Alibaba Cloud, which is currently the first cloud computing platform in China and the third in the world, with a considerable scale of supercomputing centers.
The second company is Bijian Technology, founded by the former president of Shangtang Technology, and has released the BR100 chip, which is calibrated against the Nvidia A100 and adopts a 7-nanometer process. It will be mass produced as soon as the end of this year.
The third one is Baidu, which has released the Chinese version of General AI Robot: ERNIE Bot, benchmarking ChatGPT. In terms of AI technology, Baidu can be called the Huangpu Military Academy of China’s AI, cultivating and transporting a large number of IA talents. Although ERNIE Bot can’t catch up with ChatGPT in terms of intelligence, the biggest feature of this kind of products is that the iteration speed is very fast.
ERNIE Bot is still the first generation, while ChatGPT has been updated to the fourth generation. The first generation of ChatGPT was launched in 2017. In a corner where no one pays attention, ChatGPT has quietly developed for 5 years before achieving today’s achievements. We can also be more patient with ERNIE Bot.
Alibaba Cloud is benchmarking against Microsoft Cloud, Bijian Technology is benchmarking against Nvidia, and Baidu is benchmarking against OpenAI. Robin Lee said that it is right that there will be no second OpenAI in the world. Because this is supposed to be a giant game. It’s just that in the past few years, the world’s internet giants have not gone to AI robots, Facebook has gone to Metaverse, Amazon is competing with Musk to launch rockets, Apple is swinging back and forth between making electric cars and not making them, and Google is doing AI, but what it does is playing chess with Alfa Dog AI.
As for the domestic internet giants, they are all scrambling to sell vegetables and are about to revolutionize the vegetable stall. This has only given OpenAI a chance to become famous in the world after five years of hard training in ChatGPT and embracing Microsoft.
As can be seen from ChatGPT, the competition between China and the United States has entered the deep water zone. This competition is no longer consumer grade apps, but underlying supercomputers, chips, and algorithms.