China is focusing on large language models (LLM) in the field of artificial intelligence.
Blackdovfx | Stocks | Getty Images
China’s attempts to dominate the AI world may be paying off, with industry insiders and technology analysts telling CNBC that China’s AI models have become extremely popular and are keeping up with or even surpassing American AI models in terms of performance.
Artificial intelligence has become the latest battleground between China and the United States, with both sides viewing it as a strategic technology. Washington continues to restrict entry from China The company developed cutting-edge chips designed to help advance artificial intelligence amid concerns that the technology could threaten U.S. national security.
This has prompted China to take its own approach to making AI models more attractive and effective, including relying on open source technology and developing its own ultra-fast software and chips.
China is creating popular LL.M.
Like some of the leading U.S. companies in the field, Chinese AI companies are developing so-called large language models (LLMs), which are trained on large amounts of data and power applications such as chatbots.
However, unlike the OpenAI model that powers the popular ChatGPT, many of these Chinese companies are Developing Open Source or Open Weight LL.M. Developers can download and build for free, without strict licensing requirements from inventors.
Wang Tiezhen, a machine learning engineer at the company, said that on Hugging Face, the LL.M. repository, Chinese LL.M.s have the most downloads. Qwen, a family of artificial intelligence models founded by Chinese e-commerce giant Alibabais the most popular on Hugging Face, he said.
“Qwen quickly gained popularity due to its excellent performance on competitive benchmarks,” Wang told CNBC in an email.
He added that Qwen has a “very favorable licensing model,” meaning companies can use it without “extensive legal review.”
Qwen comes in various sizes or parameters, as it is known in the field of LL.M. Large parameter models are more powerful but more computationally expensive, while smaller models are cheaper to run.
“No matter what size you choose, the Qwen is probably one of the best-performing models out there,” Wang added.
The startup DeepSeek also recently launched a model called DeepSeek-R1, which caused a stir. DeepSeek said last month that its R1 model competes with OpenAI’s o1 model, which is designed for reasoning or solving more complex tasks.
These companies claim that their models can compete with other open source products, such as YuanLlama, as well as closed LLMs, such as the LLM from OpenAI, covering various functions.
“Last year, we saw an increase in Chinese open source contributions to artificial intelligence, with very powerful performance, low-cost services and high throughput,” Grace Isford, partner at Lux Capital, told CNBC via email.
China promotes open source to the world
Open source technology serves a variety of purposes, including driving innovation (because more developers can use the technology) and building community around the product.
It’s not just Chinese companies offering open source LL.M. Facebook parent company Meta and European startup Mistral also have open source versions of artificial intelligence models.
But with the tech industry caught in the crosshairs of the geopolitical battle between Washington and Beijing, an open-source LL.M. gives Chinese companies another advantage: making their models usable globally.
“Chinese companies want to see their models used outside of China, so this is definitely a way for companies to become global players in artificial intelligence,” Paul Triolo, a partner at global consulting firm DGA Group, told CNBC via email.
While the current focus is on artificial intelligence models, there is also debate over which applications will be built on top of them and who will dominate the future global internet landscape.
“If you assume these cutting-edge foundational AI models are table stakes, then the question becomes what these models are used for, such as accelerating cutting-edge science and engineering technologies,” Lux Capital’s Isford said.
Today’s AI models can be compared to operating systems, e.g. Microsoft’s Windows, GoogleAndroid and appleiOS has the potential to dominate the market, just as these companies have done on mobile and PC.
If true, the stakes are higher for establishing a dominant LL.M.
“They (Chinese companies) see the LL.M. as the center of the future technology ecosystem,” Sun Xin, a senior lecturer in China and East Asia business at King’s College London, told CNBC via email.
“Their future business model will rely on developers joining their ecosystem, developing new applications based on LL.M., and attracting users and data, from which profits can then be generated in a variety of ways, including but far beyond guiding users to use their cloud services,” Sun added.
Chip constraints cast doubt on China’s future of artificial intelligence
Artificial intelligence models need to be trained with a large amount of data and require a lot of computing power. Now, NVIDIA is a leading designer of the chips needed for this, called graphics processing units (GPUs).
Most leading artificial intelligence companies train their systems on Nvidia’s highest-performance chips—but not in China.
Over the past year or so, the United States has increased export restrictions on advanced semiconductor and wafer manufacturing equipment from China. This means NVIDIAThe company’s cutting-edge chips cannot be exported to the country, and the company has to manufacture sanctions-compliant semiconductors to export.
However, despite these limitations, Chinese companies have successfully launched advanced artificial intelligence models.
DGA Group’s Triolo said: “China’s major technology platforms currently have enough computing power to continue to improve models. This is because they have a large stockpile of Nvidia GPUs and are also leveraging domestic GPUs from Huawei and other companies.”
In fact, Chinese companies have already Strengthening the creation of viable alternatives to Nvidia. Huawei has been one of the leaders in achieving this in China, and companies like Baidu Alibaba has also been investing in semiconductor design.
“However, the gap in advanced hardware computing will become wider over time, especially next year as Nvidia launches its Blackwell-based systems, which will be exported only to China,” Triolo said.
Lux Capital’s Isford said China has been “systematically investing in and developing the entire domestic AI infrastructure stack outside of Nvidia, leveraging high-performance AI chips from companies like Baidu.”
“Whether or not Nvidia chips are banned in China, it will not prevent China from investing in and building its own infrastructure to build and train artificial intelligence models,” she added.
