China's industrial expansion and upgrading have put enormous pressure on the United States and the entire Western world, which is the root cause of the current tension between China and the United States. The current situation is that China has opened up high-end technology and industrial competition with the entire Western world, and China has attacked cities in traditional industrial fields, forcing Western countries to retreat to high-end industries. However, China has not stopped, and has even accelerated its entry into high-end industries, much to the horror of Western countries. The United States, in particular, has faced Chinese pressure in key areas and felt the competitive situation most acutely.
A few years ago, the United States wanted to return to traditional industries and engage in the return of manufacturing, but it proved that it could not do it at all. American capital is no longer able or in the mood to operate low-profit traditional manufacturing. After Biden took office, the United States wanted to seize the chip industry and the electric vehicle industry, use sanctions to suppress China's chip manufacturing technology, and engage the government to support the development of the emerging industry of new energy vehicles, in order to overwhelm China in these two aspects.
Of course, the sanctions in chip manufacturing have given way to the development of China's related industries while also giving way to the market and opportunities, in this field that was originally extremely dependent on Western industrial standards and technical paths, Chinese companies finally have the opportunity to engage in independent innovation, this year flash memory chips have broken out, is repeating the previous process of other industries replacing Western products; General-purpose computing chips are starting up, struggling to carry out the application ecological layout, waiting for the progress of domestic advanced processes and the arrival of the outbreak period, and realizing the soaring ideal.
As for electric vehicles, China has firmly ranked first in the world, the entire industrial chain will not be subject to the West, and key patents are also in our own hands, and we already have the strength to independently formulate industrial technical standards. The United States is on the contrary in a catch-up position in this area.
However, whether it is chips or electric vehicles, it is not a key area that can occupy the commanding heights in the competition of high-tech industries, and the victory or defeat of Sino-US competition depends on who can occupy the advantage in the new industrial ecology in the future. And this new industrial ecology has never appeared in the past, it represents the core content of new technology and industrial revolution, he is - artificial intelligence.
This year's AI explosion represented by Chat GPT 4.0 indicates that mankind has begun to enter the threshold of a new technological and industrial revolution. This revolution begins in the digital virtual realm and is not completed until AI fully penetrates the physical world. This will probably take at least a few decades.
Comparing the progress of China and the United States, this time the United States first opened the door, and China continued to follow. China's backwardness is to be expected, and the United States has found a new way to widen the technological gap with China, much to the excitement of Americans and American fans.
China's backwardness is twofold:
First, we must first continue to make up for the shortcomings of traditional industrial and technological fields, and it is difficult to concentrate on the field of "deep learning" for the time being. For example, lane changing and overtaking in the field of new energy vehicles will take more than ten or twenty years to obtain the final advantage; For example, trunk passenger aircraft manufacturing, we have just taken a key step, localization still has a long way to go, far from leading Europe and the United States; For example, chip manufacturing and autonomous IT ecological construction are also making up for shortcomings from scratch, and "deep learning", which is very dependent on this, is naturally impossible to achieve any leading results on a low basis.
Second, we do not yet have a strong basic environment and resources that can support high-level AI research. The hardware is stuck in the neck of the United States; In the construction of software ecology, it is still mainstream to follow the hype and short-term success, and the long-term high-intensity research investment is very far from that of the United States; There is a shortage of high-end talents, an extreme lack of free and relaxed research environment, the proliferation of utilitarianism, and it is difficult for new ideas and ideas to obtain long-term strong resource support.
I want to focus on the second aspect, and then the outlook for future industrial development. Let me start with an example:
When it comes to AI today, it is basically impossible to bypass the company NVIDIA. Because without the GPU and software ecosystem CUDA produced by this company, there would not be today's GPT explosion. Not only GPT, but all mainstream AI "deep learning" models today are based on the CUDA ecosystem. It can even be said: NVIDIA stands for AI.
NVIDIA leads with absolute advantages, not only has far superior computing power in GPU hardware, but also has in-depth cultivation in the CUDA ecosystem. More than two decades ago, NVIDIA was a small company making game graphics cards, and now its trillion-dollar market value has surpassed that of CPU giants such as Intel. More than a decade ago, Huang Jenxun, the founder and helm of NVIDIA, was determined to pursue the programmability of the massively parallel computing architecture of the GPU, determined to transform the GPU chip from a simple graphics rendering to a hardware core that can compete with the CPU. Later, it turned out that Huang Jenxun's technological development path opened up a broad space for GPU development, and the massive parallel computing required by blockchain turned entertainment-based graphics cards into productivity tools, resulting in skyrocketing graphics card prices and resentment among gamers.
But "mining" is, after all, an evil path, and it is not beneficial to social progress and industrial development except for the realization of capital carnival. The GPU does much more than that, and Huang doesn't want its hardware to be just a profiteering mining tool that wastes electricity. As early as more than ten years ago, NVIDIA began to lay out CUDA, trying to independently establish technical standards for massively parallel computing. Huang Jenxun's gaze was fixed on the supercalculation. During this period, NVIDIA paid a lot of resources for CUDA, and was also constrained by Microsoft and Intel during the period, making the financial report ugly. But NVIDIA withstood the pressure and insisted on promoting CUDA's ecological deployment.
Finally time across 2020, big data analysis, deep learning and other AI research increasingly rely on massively parallel computing capabilities, NVIDIA is widely accepted by the industry for its excellent GPU performance and the rich and powerful functions of the evolving CUDA, Huang Jenxun's ideal can finally be realized, and more than ten years of persistence and dedication finally bear fruit. Today, NVIDIA has transformed from a game hardware company to an AI computing hardware company, becoming the ruler of AI computing hardware and ecology.
I cite the examples of NVIDIA and Huang Jenxun to illustrate that China's IT industry lacks such visionary, persevering, and able to withstand setbacks and even industry leaders, and lacks companies like NVIDIA that continue to work in the basic field. While NVIDIA is riding the wind of AI to take off, Ali Damo Academy, which is doing the same thing, is going downhill, and Baidu, which is doing the same thing, is taking advantage of GPT to prepare to play a wave of AI concepts. Of course, it is not only Baidu that plays the concept with the trend, cats and dogs have to play the concept to hype money in the capital market.
More than a decade before this year's AI explosion, NVIDIA has been silently working hard to layout; More than a decade before this year's AI explosion, what did our so-called AI companies do to lay the foundation for development? At best, follow the trend and get an AI model to buy NVIDIA's GPU for big data processing, always looking at what money can be earned in the current market, and there is no time to think about what kind of road to take in the next ten or twenty years.
There is a huge gap between China and the United States in the objective conditions for realizing AI research, and even greater in subjective will! We simply don't see even a long-term, ambitious business leader right now, and Lenovo culture still dominates China's IT industry.
I am not discouraging Chinese when I speak of the huge gap in front of us. In the field of "semantic interactive artificial intelligence" in the AI industry track, the Americans are ahead, but China has its own comparative advantages, such as big data applications and visual computing. At present, only China and the United States are eligible to go on the track, and the AI research results are basically from China and the United States. The United States leads the entire Western AI industry, and other Western countries are his vassals. China is exploring its own AI industry development path alone, although it has been heavily sanctioned, but if it can embark on this road, it already has the qualifications to compete, so there is no need to be presumptuous.
What I want to say is that the technological and industrial changes with AI as the core have just begun, and victory or defeat has never been doomed from the beginning. The comprehensive application of AI is by no means just staying in the computer to chat with humans, only helping people solve brain problems. The new industry with AI as the core must realize the full entry of AI into the physical world, and the professional term is to achieve generalized "embodied intelligence", and the layman's term is to create a machine "human" that can replace people and do most things.
Today's AI technology is still far from realizing this idea, and it is not even clear how to implement it efficiently. There are no masters in this field, and it all depends on the exploration of pioneers. Such an idea will take at least 20 years of R&D upgrade iteration before mature mass production and large-scale sales, but the architectural foundation and technical standards need to be explored and created from now on.
I don't think China's IT elite and up-and-comers don't have groundbreaking research ideas, the key is whether companies and capital can provide support that will not return in the long run. In reality, Musk in the United States and Lei Jun in China are doing this kind of forward-looking thing. Musk insists on using pure vision solutions in Tesla's self-driving driving, and the results can be easily ported to robot applications. This can be regarded as Musk's laying the groundwork for a new industrial system.
It can be imagined that from the current industrial robot application without AI to the full AI life robot application, the market space may expand tens of millions of times! The size of the market will probably be more than ten times larger than the current automobile market. This can basically become a pillar industry of the country in the future. I believe that a large number of elites can see this big cake.
There is still a huge gap between the existing technology and the industrialization goal, but high risk means high income, pioneering enterprises can often become the hegemon of the industry, Microsoft, Apple, Qualcomm, Intel, including the current NVIDIA are the same, domestic BYD is also the same, twenty years of unremitting technology accumulation and industrial chain forging has become the Chinese hegemon of new energy vehicles, and there is great hope for the future global hegemon.
In fact, there is another way for "embodied intelligence" to go, that is, military applications. I think that robot intelligence will first spread in the military field, and first open up the change in the way military struggles are conducted. The threat of modern warfare to people has increased unprecedentedly, whether it is electromagnetic information space or physical battlefield, the need to replace people with artificial intelligence is very strong. Highly intelligent intelligence warfare is needed in the virtual space, and many people overlook the great advantages of GPT for deception and anti-deception in the field of military intelligence. On the battlefield, drones, unmanned combat vehicles, unmanned submarines, unmanned warships and even robot warriors are entering or waiting to enter the application. Although there are huge differences in the application scenarios of war and daily life, it is a precedent that many military technologies have become more civilian.
There will be a lot of gains and losses for private capital to consider whether to support the research and development of new technologies, and there are no such obstacles in the military field. From some public reports, it can be seen that there is no significant gap between China and the United States in the application of AI research in the military field. If we can do a good job in dual-use and military-to-civilian conversion, the new AI industry can be strongly promoted.
Although demand determines technological progress, technology enablers still have a say in the path of technological progress. At present, we can see that the United States has multiple routes developing in parallel, and China's side seems to have only one way of big data training. This is also a major gap between China and the United States. If China really wants to have an advantage in the competition of new AI industries between China and the United States, broadening its thinking is a necessary step, otherwise, it is likely to miss the opportunity when the new AI industry breaks out.