ChatGPT is successful because American AI is not bad for money?

There is a recent saying that the reason why such a successful AI application as ChatGPT was born in the United States, not in China, is because Chinese companies in the AI field are eager to make quick gains and investors are exhausted.

There are many arguments for similar conclusions, such as the Chinese investment circle, looking at AI projects must look at the scenarios and business potential, and do not look at those basic model projects that do not see short-term returns. Chinese AI companies, even large manufacturers, attach great importance to commercialization, and have just done a project and quickly asked to point to the industry and point to revenue. In contrast, ChatGPT integrates OpenAI's long-term unrewarding research and development ideas, and finally completes the accumulation.

In this way, after several years of development, China's AI still has not produced star technology, and it seems that most of them are overly short-sighted pots of investors and enterprises. In other words, when our enterprises and capital are not bad money and do not love money like the American AI circle, our AI basic research and core technology will come out.

At first glance, this statement sounds reasonable and has sufficient evidence, and it is also very in line with the discourse habit of "internal affairs do not depend on capital" in today's public opinion field. But does this really stand up to scrutiny?

ChatGPT is very popular, which is indeed very enviable. I understand that everyone is in a hurry, so let's not rush yet. If under this crude "reflection", the final conclusion is that commercial returns are not considered, and individual gains and losses are not considered, Chinese AI will be better. That may be the real road to the opposite direction.

Because this logic of occupying the moral high ground is untenable at the starting point and will obscure the shiniest part of China's AI development so far.

Is OpenAI, and even American AI, really that unrewarding?

We've all been hearing the narrative lately that ChatGPT was born because OpenAI was a company that dared to challenge great research, regardless of return, and ultimately succeeded.

But this narrative thread may be fundamentally problematic. In this story, OpenAI is a group of young people with dreams who come together to change the world. There is a mistake in confusing cause and effect, not that OpenAI chooses greatness, but that when investors create and launch OpenAI as a scientific research entity, they are positioned to only do research with huge transformative significance.

When Musk and his friends launched OpenAI in 2015, the idea was to bring together the world's top AI talents and develop the most cutting-edge AI technology on a non-profit basis. This model is more aimed at semi-academic and semi-corporate scientific research entities like Bell Labs that can generate huge social value, and the direct competitor is DeepMind, which has just been acquired by Google.

Therefore, it is not that OpenAI chose the underlying technology research, but that it was originally born for the underlying technology of AI. Another point to note is that companies like OpenAI are themselves star companies born by integrating the world's top rich, top scholars, and massive public opinion under special chance. Comparing the technical capabilities of such an enterprise with a specific Chinese enterprise or investment institution is itself suspected of exaggerating the latter's responsibility.

Seeing this, some people may say. Can OpenAI succeed, isn't it still not bad money, without considering commercial returns? There is also a problem with this. Because by 2019, OpenAI voluntarily gave up its non-profit attribute and turned to embrace commercialization.

At that time, with the withdrawal of a group of early investors such as Musk, and the spending on computing power, data, and talent became more and more huge. OpenAI began to realize more and more clearly that the non-profit model is unsustainable, so with Sam Altman as CEO of OpenAI, the company transformed into a restricted for-profit entity (OpenAI LP), using limited profit caps and limiting revenue types to explore a new balance between commercial and non-profit organizations. This model of "training research through business" is basically successful at present. On the one hand, it promotes a large number of OpenAI technical achievements to the market, in exchange for profits to fund follow-up research, forming a positive R&D capital chain. In July 2019, OpenAI received a $1 billion strategic investment from Microsoft, one of the costs of which was OpenAI becoming Microsoft's exclusive technology provider in the cloud computing field. This also led to the fact that today, a few years later, Microsoft uses ChatGPT to force old rivals such as Google and META into a corner.

It is not difficult to see that OpenAI is not as "cold" as some media say, but its development trajectory highlights a pragmatic temperament of "trees move the dead and people live, and there are always more solutions than difficulties". Today, OpenAI can already achieve more than $35 million in annual revenue, which is certainly not much for commercial technology companies. But for semi-profit scientific research entities, it solves a lot of development problems. From the perspective of revenue methods, OpenAI is not only deeply tied to Microsoft, continuously obtaining Microsoft investment as its technology supplier, but also monetizing many of its products through commercial subscriptions, API paid access and other models, such as OpenAI's Dall.E model, which is the behind-the-scenes support of many AI drawing software.

In contrast to OpenAI, its old rival DeepMind seems a bit silent in this explosion of big language models. Of course, there are many reasons for this, such as the problem of technical route selection, but there is one problem that absolutely delays DeepMind's development efficiency and technical layout capabilities, that is, commercialization.

In recent years, we can see DeepMind losing money year after year, resulting in news of Google's dissatisfaction. Its co-founders have publicly stated that if it hadn't been for the Google acquisition, DeepMind would have gone bankrupt. Because DeepMind has always had high independence and is more inclined to idealistic, academic research atmosphere, it has been repeatedly criticized and doubted by Google, and there are many contradictions between the two.

Of course, with the fire of OpenAI, Google seems to have found that it still has to rely on DeepMind, so there is a tendency to warm up recently. But in any case, the flaws of commercialization did not become a boost to DeepMind's skyrocketing, but became a development obstacle to its constant criticism from the parent company, the slowdown of the research process, and even the repeated layoffs and bankruptcy crises.

It can be said that DeepMind is the kind of strange man image of "everything is done beautifully, and the money is not taken home". But this image, at least for now, has not brought success, but has brought doubts about it.

But to add, there is no meaning to look down on DeepMind here, its large number of research in recent years is really very breakthrough and imaginative, and when it finds a fast lane for its own development, the next phenomenal AI technology comparable to AlphaGO and ChatGPT has to be done by it. I hope that we will reflect at that time, and don't be American AI regardless of the return...

In fact, there are a lot of money-losing projects in China's AI field. Many industries + AI solutions, manufacturers do one to lose one. Countless doctors walk to factories and farms, and the final settlement fee may not even be enough for the doctors' salaries. This model is still advancing in an orderly manner in China, at least proving that Chinese AI is by no means exactly equal to quick success.

Conversely, American AI companies and investment institutions also love money, and society's evaluation criteria for new technologies and technology companies are also the success of commercialization. We have seen many Chinese AI startups, even if their revenue is difficult, they can survive through government support, joining the industrial ecology of large factories, and integrating into vertical industries. On the contrary, a large number of American AI companies are crowded in a small track, and there is a lack of support, and the final outcome is short-lived.

From another perspective, it is not that Chinese investors love money more, and American investors do not love money. For example, a recent university released data reporting that investment from the United States accounted for 37% of the total financing of Chinese AI companies between 2015 and 2021. If the data is correct, it is difficult to explain how American investors who "do not care about returns" come to invest in Chinese companies that "lose their eyes".

The pursuit of commercial returns is the nature and vocation of enterprises and commercial capital, and there is absolutely no mistake.

Many media and KOLs like to beautify a beautiful narrative of "useless for big use". Because this kind of story has drama, contrast, and traffic, it can also cater to a certain psychological expectation that "Chinese technology is not very good".

Wake up. OpenAI is not not thinking about commercialization, but it is a little better.

So why wasn't ChatGPT born in China?

So someone asked, having said all this, isn't ChatGPT, a well-known application made in the United States?

Behind this problem, there is more of an anxiety in the context of the Sino-US game, the "technology stuck neck" and other times, which is very understandable, but it should indeed be viewed calmly.

Since the issuance of the "New Generation of Artificial Intelligence Development Plan" in 2017, China's AI industry has entered a stage of rapid development, and the results have been obvious. In the field of AI, China does not have the heavy historical baggage of chip manufacturing and other aspects, nor does it have a sense of illusory that cannot be seen and touched like ultra-cutting-edge technologies such as brain-computer interfaces and quantum computing. The achievements and solidity of China's AI development are visible to the naked eye. From macro indicators such as industrial scale, head companies, and industrial ecological construction, to core technical infrastructure such as AI chips, AI development frameworks, and large models, China's AI can be said to have no obvious shortcomings.

The two companies, OpenAI and DeepMind, are special companies that gather the world's top talents, forces and capital under special circumstances. Their results and forward-looking results are difficult for global AI companies, including all other U.S. companies, to match in a short period of time.

These two companies are like the penalty at the end of the World Cup final, and the difference of that goal means that French football is far behind Argentina? I'm afraid that's not the case. Of course, if football can even be lost in Vietnam, then don't come out and discuss it.

Using the particularity of these two companies to judge a large number of Chinese companies is an unfair comparison in itself, just like many key technologies of 5G in the United States are not as good as China, can this prove that American communications as a whole are not good?

If you really want to discuss why ChatGPT is not present in China, you must first face up to the complexity behind the problem.

For example, OpenAI's talent reserve and talent source can be obtained by Chinese AI companies and even the AI industry regardless of returns, or even arbitrarily spending money? For example, the rise of OpenAI and DeepMind is very special, the time and place are favorable and occupied, and a positive snowball effect has been formed, which is also difficult for Chinese companies to copy.

In fact, OpenAI and DeepMind have gradually developed into forward-looking AI research institutions of Microsoft and Google. This model really has to be compared, more like Huawei's 2012 laboratory, Ali Damo Academy, Baidu Research Institute, of course, there are many inconsistencies. These Chinese corporate research institutes, which also do a lot of unrewarding, forward-looking research, may not be as successful and well-known as ChatGPT, but at least it is not necessary to say that Chinese companies only love money.

If you simply summarize AI technology as not bad money can be done well, it will often be counterproductive. Just like Chinese universities and scientific research institutions have made AI models in recent years, but a large number of projects have no follow-up after acceptance, papers, competitions, and awards, and open source models are not used, lack ecological vitality, and their input and output are disproportionate.

This kind of research relies on scientific research funds, of course, regardless of commercial returns, but is there really a substantial improvement in Chinese AI? Perhaps it is also doubtful.

In fact, most technologies require multi-level R&D investment. Commercial R&D and forward-looking research need to be parallel, which is only a feature of AI technology, making it easier for the results to be directly perceived by the outside world.

ChatGPT was not born in China, just as why China can't make 7nm chips is a complex and widely related problem.

He is strong by him, and the breeze blows the hills

Fortunately, ChatGPT was not born in China, which is actually not important. In today's China, the AI big model is just a matter of "sooner or later", and it is by no means a contradiction of "there is and no" like chip manufacturing.

In a sense, we don't need to struggle with who made this breakthrough first. It's good to be the first to break through, but it doesn't mean too much. The IP network was born at the European Atomic Energy Center, the Internet industry was finalized in the United States, and the economic miracle of Internet + occurred in China. Can we say that Internet technology belongs to Europe, the United States or China? It can only be said that most technologies end up being long-distance running, and the results belong to all mankind.

Since AI is also a long-distance run, it will eventually form a long industrial chain in terms of time and space relations. Every inch of advantage is an advantage; Every strength is an opportunity.

This is also the topic that I finally want to discuss in writing this article: China's AI must mention the scene, and the opening is the industry, which seems to have become a bad thing?

On the contrary, this so-called "quick success" is not a bad thing, but the biggest hole card of Chinese AI.

We know that AI is a general-purpose technology, which can bring value to almost any field, and the source of value is obviously not only on the Internet, but more in various industries through intelligent technology to complete productivity liberation, such as automatic driving, which greatly liberates the productivity of the transportation industry.

At this point, China has many industries, abundant industrial scenarios, and complex industrial structure, and the entire society has a very high acceptance of intelligent technology and digital and real integration. After years of promotion, AI technology into the industry has achieved wide and deep results in China.

Opening your mouth is the industry, and closing your mouth is the scene, not because Chinese companies and investors only love money, but because Chinese AI really has industrial integration and the ability to explore scenes.

For example, a large number of AI startups in the United States are inseparable from the Internet of an acre, and the first stop for ChatGPT's commercial landing is also search, and industrial AI projects are always difficult to form a scale in the United States.

We have visited the Tianjin Port Section C terminal, which is the world's first truly AI-driven unmanned terminal, but similar applications are difficult to promote in Europe and the United States, and the terminal union will prevent all unmanned and intelligent projects from entering.

Western AI circles piled up in the consumer Internet, unwilling to touch the industry, and unable to penetrate into the industry, leave opportunities for China, hidden in those industries and scenes, hidden in those commercial values full of "copper smell".

We once said that the real core of the Chinese version of ChatGPT is "industrial scene + ChatGPT application".

Anything strong is strong, and Chinese AI has scenarios and business paths, which is not a bad thing. Our business prospects are better, the business path is shorter, of course, we must give priority to the development of commercialization and industrialization. The reason why China's AI is unique in the world lies in the words "industrial AI".

Some people will ask again, having said all this, can not make ChatGPT's Chinese AI, have they ever made something that the United States does not have?

I've made it. For example, an operating system that everyone knows in China can solve the problem of intelligent connection of multiple devices in mines and tunnels; A deep learning framework in China highlights the capabilities of large-scale distributed training and integrated training, which corresponds to the needs of industrial and financial scenarios.

The innovation and uniqueness of China's AI comes from industrial thinking and industry scenarios. How to say that, from the people, to the people.

Interestingly, when I give examples of Chinese AI technology achievements, I can't even say names, otherwise many platforms and many readers should naturally have ideas such as "soft texts and advertisements" in their minds, but how much American AI technology is praised, how outrageous is okay, which may also explain something.

If you really want to make China's AI stronger and better, we must first stop the false comparison of "they are not bad money, we only love money", just like attributing all their own unsuccess to classmates and colleagues who are rich second generation, which is really not interesting.

Next, we must give full play to our strengths, strengthen our weaknesses, make AI technology the driving force of Chinese-style modernization, and make the huge industrial chain the source of China's AI progress.

If you can do a little more, then try to stop internal friction, stop the strife between friends, stop saying that China's AI is good, you must keep it secret, and of course, stop meaningless excessive self-praise.

Then, perhaps we can look at the problems of talent training and ecological construction. Only one day, when China has a large number of first-class AI scholars, students and R&D teams, and AI talents from all over the world are willing to join a Chinese AI company, then we will have the possibility of discussing the Chinese version of OpenAI.

Otherwise, only rely on a few celebrities to call, some generous treatment recruitment, this matter is basically no drama.

Today, we have this consensus: Chinese football is not the business of 11 people; Chinese chips are a matter for the whole of China.

Then we must also know that China's AI is not a matter of a few companies and a few investors, it is a matter of Chinese and the whole world.

He is strong and strong, and the breeze blows the hills; He let him cross his side, and the bright moon shone on the river. When you are anxious about ChatGPT and Chinese AI, think about these words.