This article was first published on the website of the The Valdai Discussion Club on March 25th (English and Russian versions)

The field of AI technology should not be seen as a zero-sum game arena, but as a model of international cooperation. Though China is under strict technology blockade by the United States, cooperation between AI scientists in both countries and industries still yields gratifying breakthroughs. Geopolitical or ideological considerations should not be an excuse to hinder international cooperation in AI technology, as cooperation will not only promote technological improvement and benefit mankind, but also help address the common challenges that AI brings to all countries, writes Xue Ying. The author is a research fellow of Xinhua Institute.

Some Western media use the term "arms race" to describe the AI technology competition between China and the United States. This term once again recalls the tense atmosphere of the military competition between the United States and the Soviet Union during the Cold War.

However, if we carefully examine the world today, we will find that the field of AI technology is not a zero-sum game arena, but should become a model of international cooperation. Because only with sufficient international cooperation in the field of AI technology can we humans enjoy the benefits of AI and avoid the disasters it brings.

Cooperation between AI scientists in China and the United States may be far greater than some people imagine. On January 31, an academic paper that shocked the world again was submitted. A group of U.S. researchers including Chinese American Feifei Li, a world-leading AI technology pioneer, and Stanford PhD student Niklas Muennighoff, an undergraduate graduate from China’s Peking University, trained a high-performing AI reasoning model using open-source technology from Alibaba Group’s Qwen2.5-32B-Instruct.

They trained their S1 reasoning model with a cost of less than 50 U.S. dollars. Only a small dataset of 1,000 questions was used during the training which used test-time scaling approach. The result showed that their S1 model outperformed the OpenAI o1 and DeepSeek's R1 in mathematics and programming ability tests.

This achievement has excited the AI community. It shows that researchers can innovate without substantial financial backing. It also subverts the previous “scaling law” in the pre-training scaling approach, which posits that “bigger is better”— the larger the scale of training data, the larger the scale of the model, and the more computational power invested, the stronger the AI model.

To me, this is a great example of collaboration between Chinese and American scientists and industries that benefits the world. We know that AI is being used in increasingly more fields. It allows amputees to regain their hands and feet, high-level paralyzed people to pick up cups and drink water by themselves, farmers to harvest wheat before it rains, and workers to relieve themselves from heavy physical labor work.

Why don't let more people enjoy the benefits brought by AI technology? This is one of the reasons behind DeepSeek's decision to open source. Other reasons include the desire to collaborate with other developers and users to build an ecosystem that is conducive to technological progress, DeepSeek’s chatbot told me.

Under the strict technological blockade by the United States, a Chinese company unexpectedly made an AI breakthrough, but chose to open source the technology and share it with the world. This seems ironic. And an American research team cooperated with an open-source model from a Chinese company with the purpose of reducing the training cost of AI models, bringing about new breakthroughs that researchers around the world are happy to see, this is not only an unintentional mockery, but a substantial criticism.

The success of DeepSeek has proven the failure of the US technology blockade. Today's world is an organism in which countries are closely connected to each other. The connections between countries are like the complex blood network of the human body. When one blood vessel is blocked, nutrients are transported from other vessels. Man-made hindrances can never stop the power of this massive organism's growth.

A wiser approach would be broad international cooperation. Chinese President Xi Jinping often says "We in China always believe that China will do well only when the world does well, and vice versa. Cooperation will only promote technological improvement and benefit mankind, but also help address the common challenges that AI brings to all countries. One big difference between the digital world and the physical world is the former still lacks international rules. Resisting algorithmic bias, protecting data privacy, reaching ethical consensus, and preventing malicious attacks, all require cooperation from governments, enterprises, and scientific research institutions around the world to jointly formulate reasonable international norms and legal frameworks.

From the perspective of economic development, AI technology is equivalent to the invention of steam engines and electricity; from the perspective of human civilization, AI may bring as great significance as fire and writing. So there is no doubt that we have reached an important moment in history. This historical juncture is also a moment when we will make an important choice -- cooperation or competition?

All the information used to train AI is based on knowledge and experience shared by mankind for thousands of years. AI technology should not be monopolized by a few people or a few countries, but should be shared by the whole world and become a new driving engine for the continuing progress of human civilization. Any geopolitical or ideological considerations should not be an excuse to hinder international cooperation in AI technology.