OpenAI, an American artificial intelligence laboratory, released its first artificial intelligence video generation model, Sora, on February 16. This model can not only generate videos based on text instructions but also convert static images into videos and even support video-to-video editing. Its multi-modal technical capabilities shocked the world.
At the end of 2022, OpenAI launched the chat robot Chat GPT, which triggered a global boom in generative artificial intelligence, and this field has also become the main battlefield for competition between the United States and China. When OpenAI launched another disruptive AI product after ChatGPT, many observers believed that the gap between the United States and China in the field of artificial intelligence had further widened.
The development direction of artificial intelligence in the United States and China is different
In 2023, competition between the United States and China in the field of artificial intelligence will become increasingly fierce. Du Yijin, founder of the Taiwan Artificial Intelligence Laboratory, believes that the development directions of artificial intelligence in the United States and China are not the same.
In an exclusive interview with the media, “In-depth Perspective,” he said, “The United States focuses more on the development of artificial intelligence for professional purposes for people’s livelihoods, while China focuses more on the development of artificial intelligence technology related to national security.”
China is at the forefront of the world in the application of artificial intelligence in areas such as facial recognition, visual surveillance, and personal information systems. The New York Times quoted John Coker, Distinguished Professor of Electrical and Computer Engineering at Duke University, in an article on February 22 titled “Behind China’s “AI War”:
If you want to surpass the United States, you can’t do without it.” Chen Yiran said, “As much as 50% of China’s investment in artificial intelligence is invested in computer vision technology required for surveillance rather than establishing basic models for generative artificial intelligence.”
China’s development of AI models relies on the underlying systems of the United States
In fact, China does not lack the ambition to develop basic models of generative artificial intelligence. While American technology giants such as Microsoft, Meta, and Google have launched large language models, Chinese companies such as Baidu, Alibaba, Tencent, and iFlytek have also launched their own large models, which is nicknamed the “Battle of 100 Models.”
However, China’s basic artificial intelligence model is often accused of relying on the underlying system of the United States. For example, the large language model launched by 0.1 AI, an artificial intelligence company founded by well-known venture capitalist Kai-Fu Lee, is trained on the basis of the Meta open-source model LLaMA.
Developing your own model based on other people’s open-source models can indeed reduce the cost of training the model, but will it become a “stuck” weakness?
Artificial intelligence expert Du Yijin said in an exclusive interview with “In-depth Perspective” that many countries are already wary of this issue. He said, “Especially in the European Union, they say that if these basic models are once again monopolized by a few technology giants and everyone uses his AI model, then, in the end, every country will be like a ‘digital colony’. We kept giving him data to train his best model, but we became dependent on his model, and in the end, you couldn’t leave him.”
What difficulties does China face in self-developed basic AI models?
One of the major constraints facing China in developing its own large AI models is the data required to train the large models. Analysts pointed out that Chinese people often use apps such as WeChat and Weibo to communicate online. The data generated in this way is enclosed by “walls,” which makes it difficult for AI models to absorb this data during training. In addition, due to the artificial “firewall” set up by the Chinese government, massive international data cannot freely enter China.
Chinese modelers have significantly less data to train large models than their American counterparts. At the same time, due to China’s information censorship system, the quality of China’s self-produced data has been damaged, and the large models trained with this data are naturally not outstanding enough.
Another issue hindering the development of basic models of artificial intelligence in China lies in chips. The British think tank AI Governance Center analyzed 26 large Chinese models and found that more than half of the models relied on the American chip company Nvidia for their chips. In the past year, Nvidia has sold a number of “special supply” products to China in order to circumvent the U.S. government’s ban on chips in China. However, industry insiders generally believe that China cannot make up for its lack of computing power with these “low-end” chips.
“If your (chip) specifications are lower, the computing power will definitely decrease. For example, the calculation here is faster, but there is no way to calculate it so fast over there.” Artificial intelligence expert Du Yijin said, “With the computing power of today’s large-scale computing, to put it bluntly, the U.S. side is still relatively sufficient, while China’s supply chain in chip technology is relatively not that strong.”
In response to the burgeoning advancements in artificial intelligence, China has implemented increasingly stringent policy controls. The Chinese government rolled out the “Interim Measures for the Management of Generative Artificial Intelligence Services” in 2023, mandating that AI-generated content align with “core socialist values” and refrain from inciting subversion of state power, undermining the socialist system, or compromising national security and interests.
Moreover, the content must not tarnish the national image, provoke separatist sentiments, disrupt national unity and social stability, or propagate terrorism, extremism, ethnic hatred, discrimination, violence, obscenity, pornography, or false and harmful information prohibited by laws and regulations.
Analysts believe that the CCP’s obsession with information control will seriously hinder the development of artificial intelligence in China. However, some industry insiders in China believe that the government’s early drawing of “red lines” will make artificial intelligence companies more aware and dare to explore within certain boundaries.
Beijing uses artificial intelligence to fight “cognitive warfare”, adding difficulties to world governance
The development of artificial intelligence also brings more ethical and governance problems, such as how to prevent generative artificial intelligence from creating and spreading false information. How can we identify and resist authoritarian countries’ use of artificial intelligence to wage “cognitive warfare” against the free world?
Du Yijin and the Taiwan Artificial Intelligence Laboratory he founded discovered that Beijing is using artificial intelligence to influence foreign public opinion in increasingly diverse ways. For example, some “collaborative groups” will play “role-playing” games. They are both controlled by China, but they deliberately act as different camps and attack each other with the most violent remarks. The purpose is to provoke divisions in overseas public opinion.
“So you will see why people seem to become more extreme in an election year? In fact, you can see that there are many factors on the Internet to fan the flames,” Du Yijin said.
Du Yijin believes that the United States should think about artificial intelligence regulation from the perspective of protecting digital human rights. He said, “In the past, the Internet was a space to increase freedom of speech, but with the advent of artificial intelligence, many people will express opinions online, and many trolls will attack you.
It is like a person walking on the street, and you express different opinions. They will be beaten in the same way. So this is not an issue of platform governance, but how do you protect a natural person from speaking fairly on the Internet?”
There are also many voices opposing regulation. Some people believe that excessive regulation will make American technology companies lag behind China in the artificial intelligence competition and will also harm Americans’ own freedom.
Washington is carefully exploring the boundaries of regulation, seeking a balance between freedom and security. Analysts pointed out that the competition between the United States and China in artificial intelligence is not only a competition in innovation capabilities but also a competition in governance capabilities.
Exciting news! Meritnews24 is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest insights!”Click here!
Discover more from Meritnews24
Subscribe to get the latest posts sent to your email.