Last week, China’s largest AI event took place in Shanghai: the World Artificial Intelligence Conference (WAIC) with 500 exhibitors, 1,500 exhibits, over 300,000 participants and even an appearance by Chinese Prime Minister Li Qiang.
But despite its impressive scale, I was disappointed by the conference. I had hoped to witness the sector’s technological advances. Instead, WAIC confirmed my suspicions: there is a gap between what China’s AI can do and the groundbreaking innovations coming out of Silicon Valley.
WAIC exhibitors focused on robotics and large language models (LLMs), with only a few companies in the generative AI space. Over half of the companies at WAIC, including large technology companies and even some state-owned telecom companies, showcased their new models.
In Shanghai, Baidu founder Robin Li encourages Participants should start developing practical AI applications rather than continuing to refine their LLMs. He stressed that a powerful and widely used AI application will benefit society more than another model that can process huge amounts of data but has no practical use.
The generative AI applications shown in Shanghai were mostly ChatGPT-like chatbots, with the exception of Kuaishou’s Text-to-Visual Application Kling, a Sora-like product that I found really impressive.
As I wandered around the showroom, I noticed that most of the chatbots required prompts in English, not Chinese. This makes me suspect that many of the Chinese AI programs are actually running on models developed outside of China.
It’s clear that the models still need some fine-tuning. One consumer requested a text-to-visual app from Moore Threads with “a cute little boy with brown hair sitting in the garden.” The result was a baby with bright fuchsia skin, eyes that weren’t in line with the face, and a disproportionately small body.
I left the conference with the agreement Alibaba Chairman Joe Tsai openly admits Earlier this year, China announced that generative AI development in China is at least two years behind that in the U.S. This means that U.S. and Chinese companies are not really in the same league, making it difficult to compare them directly.
The key problem is that China’s LLMs are limited to using data within the Great Firewall. As an investment bank Goldman Sachs noted at the end of last year“LLM performance improves with scale—more parameters, more and better training data, more training runs, and more calculations.” There is simply less information on the isolated Chinese-language Internet than on an open Internet with sources in many different languages.
AI companies outside of China simply have a lot more data to use for training. An AI developer in China will struggle to keep up.
The limitations imposed by limited access to advanced GPUs are also evident. US policy, which Restrict access to cutting-edge chips and chip manufacturing technology will mean that Chinese companies Developmental delay their non-Chinese colleagues.
But despite these limitations, China’s AI developers are looking for opportunities to innovate.
Many strong talents from the country’s mature consumer tech ecosystem are focusing on AI. Most of the founding members of the hyped “Four Tigers”—Baichuan, Zhipu AI, Moonshot AI, and MiniMax—all worked at a major technology company for a while. Their keen sense of consumers and products is why they are now China’s leading AI application industry. From a consumer perspective, their products are on par with many of the leading U.S. applications.
There is also progress on the hardware front. Huawei’s Ascend AI processors in particular seem Miles ahead of their competitors. The Chinese tech giant, which now uses the chips made by SMIC, claims that its Ascend 910B AI chip can outperform Nvidia’s A100 chip in some tests, especially when training large AI models.
Chinese AI developers face several fundamental hurdles, including a challenging environment, a shortage of sophisticated chips, geopolitical isolation and national security concerns that limit the mobility of talent and capital.
Together, these constraints will create two parallel AI ecosystems: one inside China and one outside. The United States will maintain its leadership role in developing this transformative technology.
But just because the U.S. has a technological edge doesn’t mean China’s AI developers are being left behind. Chinese companies have always been one step ahead of their non-Chinese competitors from the start, but fierce competition and a willingness to experiment have helped them catch up – and in the case of consumer internet companies, even outperform the rest of the world.
In the world of AI, the US and China are both enemies and competitors. We should hope that the geopolitical competition between them does not stand in the way of innovation and collaboration.
The opinions expressed in Fortune.com commentaries are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.