Recently, the term GPT has become popular worldwide, to the extent that the previously fascinating metaverse no longer interests people. In just six months since the birth of generative artificial intelligence, numerous AI products have emerged like mushrooms after the rain. For example, the well-known note-taking app Notion has developed a Notion AI based on GPT to help people with daily writing, article summaries, translations, etc. Even domestic imitations like Flowus have officially launched their Flowus AI today (May 11). However, industry leader Wolai, since being acquired by DingTalk, has not only slowed down its update frequency but also completely failed to keep up with the intelligence of these two note-taking apps.
Artificial intelligence has entered a new era today, and its generative language models can already serve as efficient assistants for people's daily work, and even replace a group of people. Interestingly, before and within at least three months after OpenAI released this product, there was not much movement in China. It was not until Baidu released the PPT-style AI language model "Wenxin Yiyu" that many domestic companies began to announce their own plans to develop similar products. They claimed that they had been working on it for years, to the point that there are countless enterprises in China claiming to develop products that rival ChatGPT.
Act Fast#
Unlike AI companies abroad, most AI companies in China are developing products related to ChatGPT integration. For example, they use ChatGPT to build legal consulting or embedded writing software. However, major Chinese companies are mostly developing their own ChatGPT-like products instead of embedding it in their own applications. Companies such as Baidu, 360, Alibaba, JD.com, Kunlun Wanwei, iFlytek, and Fudan University have created products that are sometimes self-proclaimed. For example, when asked to generate an image, Baidu once made a mistake with English and Chinese translations. On the other hand, 360's product seems like a last-minute attempt.
Now, various companies are teasing their own products, such as Fudan University's Moss, Kunlun Wanwei's Tiangong, Alibaba's Tongyi Qianwen, and so on. These products that require public testing, except for Baidu's Wenxin Yiyu, which sent me an invitation code the next day, have mostly disappeared after filling out the form, with no news whatsoever. I wonder what privileges are needed to obtain their testing qualifications, or perhaps they are just making empty promises due to overwhelming server pressure? Some domestic gaming companies, such as NetEase, have also claimed to apply artificial intelligence to their own game NPCs. If such a game exists, it would definitely not be boring. Players could spend half a day just conversing with NPCs, without having to spend half an hour online like in Genshin Impact just to complete daily tasks.
However, we must know that it is not the era of the Great Leap Forward anymore. We don't need to rush and catch up with the United States. But some people seem to be dissatisfied and want to release their own products to become leaders in their respective regions. This can be understood because in an emerging industry, a company that masters the most advanced and cutting-edge technology has the power to set standards and prices. Once they have the power to set prices, the profits in this industry can be immense, allowing them to take seventy to eighty percent or even more. Therefore, we can see that the first users and recipients of large language models are enterprise users, to the point that many people cannot find a way to use ChatGPT or find shortcuts to use domestic products. It's a sigh for the rise of AI.
In comparison to Baidu and Microsoft, two domestic and foreign giants, it took no more than two to three months from the internal testing release of the new Bing to its opening to the public. In this short period, it not only gained conversational abilities but also gained the ability to draw. Baidu's Wenxin Yiyu has been released for two months, but I haven't seen any news about Baidu integrating Wenxin Yiyu. I even have the feeling that my qualification could be revoked at any time. Why doesn't Baidu follow Bing's example and embed Wenxin Yiyu into its search engine?
Just before publishing this article, I asked Wenxin Yiyu about this issue, and Wenxin Yiyu responded:
Q: Will Baidu embed Wenxin Yiyu into its search engine?
A: Baidu is very likely to embed Wenxin Yiyu into its search engine.
Baidu has announced a ChatGPT-like project called "Wenxin Yiyu" (ERNIE Bot). The product is currently in the final stages of preparation and will complete internal testing in March, opening to the public.
Wenxin Yiyu will be available through Baidu Search and other products, making it more convenient for users to obtain information and improve their user experience.
Therefore, it is highly likely that Baidu will embed Wenxin Yiyu into its search engine to improve the efficiency and user experience of information retrieval.
Now it's already May, and Wenxin Yiyu's answer contains factual errors. We can also understand why Mr. Li (Baidu's CEO) was unwilling to demonstrate this high-tech product at the press conference. It is completely different from the recent Xunfei large language model launch event. Therefore, when we see these AI products, we can't help but feel a sense of urgency. AI is hot, but for Baidu, it may be too hot to handle.
However, after all, artificial intelligence is our future. The success of any company in this field will be tested over time. But we shouldn't let companies monopolize all the products and high-tech. Otherwise, our wallets will be at their mercy. This makes me lament the fact that originally open-source AI products have now become commercial tools.
Although there are many open-source products now, it is still difficult for ordinary people to deploy them. While there are many clients for ChatGPT, ordinary people cannot use them. Even the new version of Bing cannot be used with Chinese IP addresses. It is truly baffling.
Strict Review#
Nowadays, when watching videos on apps like Douyin and Kuaishou, many people refer to money as "rice" or "dabuliu". When watching videos, certain words such as "corpse," "death," "kill," and "poison" cannot be mentioned. When discussing international affairs, nuclear bombs and atomic bombs are referred to as "gas cylinders". When talking about the African wilderness, snakes are called "spicy sticks", and sexual activities are referred to as "pulling radishes". Even low-cut outfits need to be censored. On the other hand, in videos of American auctions, even if a lady is wearing a low-cut outfit, it is censored. Are they worried about minors seeing it?
We know that although we cannot access the original version of ChatGPT in China, there are many mirror websites that provide free or paid ChatGPT question-and-answer products. Their URLs are constantly changing, but many similar websites have a mechanism, such as sensitive words. Once, I asked a question in English about a political issue in Western history that had nothing to do with domestic political taboos, but it still couldn't be answered. The response even said that my IP had been recorded. Seeing this, I couldn't help but feel a chill. Chinese people are truly fragile...
Based on this, most domestic companies will use domestic large language model AI tools. Unless they are engaged in foreign trade, when they need to access websites with these large models using scientific methods, they must consider the consequences. For example, Midjourney and ChatGpt. Although the official websites are not blocked, OpenAI has actively blocked mainland IP addresses. Midjourney requires logging into Discord, which is also inaccessible. Therefore, if you want to use these foreign products, you must use technology. This intentional or unintentional blocking has paved the way for the replacement of domestic products and expanded their pricing power in the Chinese market.
Recently, I also saw a picture on Zhihu. In the picture, a group of people were having a meeting to discuss Party leadership AI, titled "Taking the Chinese Socialist AI Path". Before any emerging product is born, it must face strict regulation, based on politics and protection of minors. This has been the case for decades, to the point that Chinese animations are completely uninteresting, just brainless content for children. What about Maca Baka? Chinese TV dramas are always family-oriented. After the rise of female boxing in recent years, the capital market began to cater to this ridiculous trend, producing brainless and absurd love dramas. Everything changes, but nothing changes. Deep content is completely absent.
In this context, what can a Chinese-style generative AI product produce? Can it help us write articles analyzing how people suffer under American imperialist rule? This is unnecessary because we have known this for decades. Can it help us learn math and science faster and better? Just search on Baidu, and you'll find a lot of ads and monopolistic information. What meaningful answers can we get from training this large language model?
Of course, we know that certain things should not be generated, such as child pornography, hate speech, and racial discrimination. But AI is meant to help us work and live better, not to help politicians manipulate public opinion. Therefore, when using local AI products, we should try to avoid discussing political topics because we won't get any answers. Instead, our IP addresses might be recorded and reported to certain departments.
Maintain Hope#
Despite all this, the future development of artificial intelligence technology will not only be applied to theoretical problems but also to problem-solving. In today's society, with the rise of artificial intelligence, the unemployment rate is increasing. Coupled with temporary economic difficulties, many people are wary of AI, fearing that their jobs will be taken away. Although this is undeniable, we also have a solution, which is to regulate the speed of AI development through policies. The country has proposed the concept of "using work to replace relief" and hopes that entrepreneurs can create more job opportunities. Therefore, in the early stages of AI technology, there will inevitably be a period of observation. During this period, I believe we can create better products. However, to be honest, they are like good products made for China.
Although our society is not as difficult to survive in as ancient times, many people still cannot say that their lives are satisfactory. After experiencing three years of the pandemic, this historical event, which has been replaced with "masks" on short videos, we need to adjust our mindset to face a period of downturn that may last for a long time, whether it is the economy or international politics. In this context, we should focus more on the long-term future. This reminds me of a quote from the famous American physicist Henry Augustus Rowland (1848-1901) in his speech titled "An Appeal for Pure Science" at the 1883 American Association for the Advancement of Science (AAAS) annual meeting:
Can our country be satisfied with standing by when other countries take the lead in the race? Can we crawl in the dust to pick up the breadcrumbs that fall from the rich man's table and think that we are richer because we have more breadcrumbs? But we forget this fact: he has bread, which is the source of all breadcrumbs. Are we as humble as pigs, thinking that grains and husks are much more valuable than pearls? If my understanding of the times is correct, then I believe that we should not be satisfied with our lowly position.
—Translated by Wang Danhong
Our pursuit of science and technology should not be constrained by politics. Therefore, when I see people angrily criticizing the cooperation between Tsinghua University and Japan in researching certain technologies, I am somewhat confused. Are the Japanese coming to steal China's aerospace technology? Just because their rocket launches have failed before? We may have great animosity towards Japan, and their disposal of nuclear wastewater has had a significant impact. However, we should understand that Japan possesses advanced technology, especially in some cutting-edge fields, which is not inferior to China. Cooperation between scientific institutions is not strange. If we give up, should we look to India, Tanzania, Gambia, the Philippines, or North Korea for cooperation? Do they have the technology? That is not cooperation; it can be called poverty alleviation.
Science knows no borders. Whoever discovers or invents first can lead an industry, set standards, and even become a hero for humanity. The breakthroughs that artificial intelligence can achieve are not just about bridging the gap between people but also about redefining the future of humanity. Based on this, it is meaningful to talk about cooperation with the greatest goodwill while maintaining the highest vigilance.
Stay Alert#
National security is particularly important in today's society because we have not yet reached a consensus on a communist society. To be honest, many things in communism are unrealistic in today's society. However, we must understand that perhaps that era will not be called communism. Maybe it will be called a great harmony society. At present, the most important thing is to maintain a clear and rational roadmap for the future direction of artificial intelligence. We should understand that AI will not develop consciousness.
I believe it will, but the consciousness of artificial intelligence will not be anything novel. Its consciousness will be controllable because the data and technology used to train them are products of previous thoughts, culture, science, and technology. If ChatGPT had been born in the 18th century, it would never have achieved what it has today. Although AI products were born as a royal flush, it is clear that as human knowledge advances, it is better for them to be born later. On the one hand, we need sufficient hardware computing power to support their operation. On the other hand, our knowledge has not yet reached the point where all channels are open. We are limited by copyright and cultural conflicts, among other factors. AI will definitely not be able to deliver satisfactory answers to everyone until we have reached that point. However, it is not difficult to let it develop consciousness at some point in the future. It may just take some time.
At the same time, we cannot say that we should not worry about the threat that consciousness may pose to humans. However, before AI develops consciousness, our constraints on artificial intelligence should gradually be added to its training model. But this should not be like the brainwashing and political indoctrination in countries like North Korea. Instead, it should focus on content that affects social norms and public order. In this state, even if AI develops consciousness, it will at most become a truly intelligent assistant rather than a revolutionary army of robots.
2023 is destined to be a year of explosive growth for Chinese AI products. Some may be following the trend, while others may be repackaging ChatGPT or Calude. However, I believe that even if there is a mix of good and bad, we can find some true dragons among them in the true sense of the word. Perhaps in some places, they can compete with those who eat bread.