Stanford University professor and artificial intelligence expert Andrew Ng warned, “If U.S. tech giants continue to rely on closed-source AI models, they risk losing out to China, which could dominate the AI supply chain with open-source AI.”
In a recent video interview with the Chosunilbo, Ng highlighted the rapid progress of Chinese AI, pointing to DeepSeek, a Hangzhou-based AI startup that recently sent shockwaves across global markets with its AI model R1. The model’s unexpectedly strong performance has raised concerns that China could soon take the lead in AI development if the U.S. does not adjust its approach.
DeepSeek is open source, meaning its AI models are available for anyone to download, copy, and build upon. In contrast, a closed system is not open to the public and is only accessible to a particular group. Leading AI chatbots like OpenAI’s ChatGPT and Google’s Gemini use closed models. “China is promoting open-source AI models at a national level, which has been a key driver in the country’s AI development,” Ng said.
A leading figure in AI research since the mid-2000s, Ng is widely regarded as one of the four AI pioneers alongside Geoffrey Hinton, Yann LeCun, and Yoshua Bengio. He noted that China has made tremendous strides in AI over the past two years and, in some areas, has already surpassed the U.S.
AI technology is advancing rapidly, and no single company or country has established market dominance. The race for AI leadership is intensifying, with tech giants like Google and Microsoft ramping up investment in AI. Startups such as OpenAI (backed by Microsoft) and DeepMind (acquired by Google) have been aggressively rolling out new models. Against this backdrop, the unexpected rise of DeepSeek raises important questions about the future of AI.
Why do you believe China will dominate the AI supply chain?
“China’s AI ecosystem is much more open than that of the U.S. Recently, at the World Economic Forum in Davos, I spoke with a senior Chinese research director who told me that the government encourages open-source AI development. The Chinese government recognizes open-source models dramatically accelerate knowledge transfer, making it a strategic advantage for China’s AI development.
AI is similar to entertainment in that it reflects the values of the country that develops its underlying software. If China continues advancing its AI through open-source technology while the U.S. maintains its closed model approach, AI systems worldwide will increasingly incorporate Chinese values. If the U.S. and other nations do not respond collectively to China’s open-source AI strategy, China will sooner or later dominate the AI supply chain.”
Do you consider open-source AI better than closed AI models?
“Open-source AI may be bad news for some big tech companies that have invested heavily in closed systems, but I believe it’s better for the world. In an open-source ecosystem, there are no gatekeepers. Anyone can examine the software, identify issues, and improve the technology—which often makes open-source software more secure.
American online ecosystems frequently have gatekeepers. Take smartphones as an example: there are two major gatekeepers—iOS (Apple’s operating system) and Android (Google’s mobile OS). This means that without Apple or Google’s approval, developers cannot launch applications on smartphones.”
How far has Chinese AI caught up with the U.S.?
“AI has many different fields. The U.S. still leads in several areas, but in some, China has caught up significantly or already surpassed the U.S. When ChatGPT launched in November 2022, I believed the U.S. was far ahead of China in generative AI. However, in just two years, China has closed the gap at an astonishing speed. DeepSeek’s R1 model has demonstrated reasoning capabilities comparable to OpenAI’s, and Alibaba’s AI model, Qwen, has made remarkable progress over the past year, particularly in video generation. Many teams I work with are already using Qwen for visual tasks.”
Are there other AI fields in which China has an advantage?
“Even before the rise of generative AI, China was ahead of the U.S. in surveillance technology. The Chinese government actively supported the development of surveillance AI, whereas many U.S. companies were reluctant due to privacy concerns and regulations. As a result, many countries purchase surveillance systems from China rather than the U.S., meaning their software is more aligned with Chinese values than those of liberal democracies. The U.S. and its allies have already lost their chance to lead in this area.”
(While protecting privacy is essential in liberal democracies, they have also created an environment where China—where government control over privacy is less contested—has gained a significant edge in surveillance AI.)
Is the AI rivalry between the U.S. and China comparable to the Cold War nuclear arms race?
“Not exactly. Falling behind in the AI race is potentially more dangerous. Let’s say you know how to build a nuclear weapon. Even if you have the knowledge, building a nuclear bomb is extremely difficult, while AI is more accessible. This is why falling behind in nuclear weapon development is not as dangerous as lagging in AI. With nuclear weapons, disarmament is sometimes a reasonable strategy. But restricting AI development would have severe economic consequences for any country.”
OpenAI CEO Sam Altman recently predicted that Artificial General Intelligence (AGI) will be available within a decade, but you said it will take 30 to 50 years.
“AGI is defined as an AI that matches or surpasses human cognitive capabilities. If you consider what humans are capable of, we are still far from achieving AGI. People can learn to fly an airplane with just 40 hours of training. AI is still far from performing intellectual tasks at a human level.”
Some fear that AGI poses an existential threat to humanity. (Geoffrey Hinton, the 2024 Nobel Prize winner in physics and a renowned AI expert, warned that there is a 10% to 20% chance AI will lead to human extinction in 30 years.)
“My advice to those people is to learn to code or understand how AI works. In the future, the most valuable skill will be the ability to communicate with AI effectively. Just as people had to learn how to search the web when the internet first took off, AI literacy will be essential.”
What is your opinion on Professor Hinton’s warning?
“I deeply respect Professor Hinton but disagree with his predictions about AI’s existential threats. His warnings about AI wiping out humanity are speculative and resemble science fiction more than reality. Instead of instilling fear, we should encourage young people to learn how AI works.”
Are you concerned about the potential misuse of AI?
“AI is already being misused, such as generating deepfake images and spreading misinformation. However, this has been true for all general-purpose technologies, from electricity to steam engines. Fortunately, the benefits of AI far outweigh the risks.”
You recently referred to AI as “new electricity.” Can you elaborate?
“Like electricity, AI has countless applications. For instance, AI helps retailers optimize pricing. In Alabama, an AI-powered drone saved a police officer’s life by detecting an assailant and calling for backup. The biggest remaining challenge is making AI economically viable. Companies are investing vast sums in AI development, but we need sustainable business models for AI to reach its full potential.”
