Chinese Start-Up Moonshot Details New A.I. Model

Chinese Start-Up Moonshot Details New A.I. Model

The Chinese start-up Moonshot publicly released the details of its latest artificial intelligence model on Monday, showing the world how it was built. But in a departure for most Chinese A.I. firms, Moonshot indicated that people or companies who want to use it on a large scale will have to secure licenses to use it.

The model, called Kimi K3, sent shock waves through Silicon Valley and U.S. technology stocks this month when Moonshot said that it performed some tasks as well as the best models from American rivals, including OpenAI and Anthropic.

Kimi K3 followed the release of other high-performing Chinese A.I. models. Most of these were open source or open weight, which means the company publicly shares the details of how it built the technology, making it easier for anyone to use it. China’s tech industry has embraced open source, which has helped Chinese firms improve their models quickly and won them customers around the world with lower prices.

The model’s release has also inflamed a fierce debate in Beijing, Silicon Valley and Washington about whether governments should take steps to limit access to Chinese open source models.

In the United States, executives, investors and officials are divided over whether the increasingly widespread use of Chinese models goes beyond being a threat to the profits of American A.I. firms to constituting a threat to U.S. national security.

In China, the debate centers on how to capitalize on making a technology that is accessible and popular, while finding ways to make money from it. Chinese companies must navigate those forces alongside evolving policies set by a government that has placed A.I. development near the top of its economic priorities.

With its release of programming code and other information on Monday, Moonshot is showing its work in public, which is consistent with China’s open source ethos. But the company is also moving to capitalize on its model’s popularity by saying that large users will need to enter into commercial licensing agreements.

The model is so large that it could not be downloaded and used on a basic computer. While Kimi K3 is technically open, most people who use it will either get a subscription from Moonshot or get access to the model from another company that has reached a licensing agreement with Moonshot.

Chinese companies have tried different approaches to bring in enough money to sustain the enormous expense of building A.I. systems. Moonshot is notable for being the first to make this commercial requirement so explicit, said Kevin Xu, the founder of Interconnected Capital, a hedge fund that invests in A.I. technologies.

In the past, Chinese start-ups have shared technical details at the same time they unveiled a new model. But Moonshot waited nearly two weeks.

The delay signals just how sensitive this technology has become as a geopolitical flashpoint between China and the United States.

Since Meta announced late last year that it would acquire Manus, an A.I. start-up founded in China, the Chinese government has placed enormous scrutiny on how A.I. systems made in the country are used by people outside it. In April, the Chinese government ordered the transaction be unwound, indicating that it was considering whether A.I. systems made by Chinese companies could be regulated as exports.

Now, Chinese officials are grappling with how to regulate the technology to allow Chinese companies to compete with rivals in Silicon Valley, while keeping the strategic and economic benefits firmly inside China.

On the same day that Moonshot announced Kimi K3, Xi Jinping, China’s leader, made a speech at the World A.I. Conference in Shanghai that cast China as the global champion of an open approach to the technology.

“The crux of the debate in China is how do you square these different signals about the strategic benefit of Chinese tech being used around the world,” said Samm Sacks, a senior fellow at the Johns Hopkins School of Advanced International Studies who attended the conference in Shanghai. “Beijing wants champions, but how not to kill the golden goose?”

China’s A.I. industry has faced years of U.S. trade restrictions that limit its ability to buy the computer chips that power A.I. systems. It also has far less money to buy such computer power compared to deep-pocketed American rivals.

Two days after Moonshot announced Kimi K3, the company was forced to announce that it needed to stop accepting new users because it couldn’t get enough computer chips to serve them. The company is not accepting new subscribers. Moonshot declined to comment on its subscribers on Monday.

Jeffrey Ding, a professor at George Washington University who studies A.I. competition between the U.S. and China, said that Chinese companies had embraced open source in part as a tool to help them catch up quickly to the best A.I. systems in the world.

“Chinese companies are not doing open source for the sake of open source itself; they are employing it at as a strategy, as a tool of the challenger,” he said. “Eventually the goal is to make money.”

Xinyun Wu contributed reporting from Taipei, Taiwan.

Leave a Comment

Your email address will not be published. Required fields are marked *