The Dark Side of the Moon Kimi Founder Publishes Article in "People's Daily": Making China's Large Models the Foundation of Global Innovation

Author: Yang Zhilin, Founder and CEO of Moon’s Dark Side Kimi

Recently, “intelligent agents” and “tokens” have become common topics of discussion. Seemingly niche open-source artificial intelligence projects are driving the emergence of new business models and market opportunities. In this wave of technological enthusiasm, domestic large-scale model technology has made rapid progress and applications are continuously expanding, not only injecting vitality into the development of the intelligent economy but also gradually becoming the foundation of global innovation.

As an important driver of the new round of technological revolution and industrial transformation, artificial intelligence is currently at a critical development juncture. For example, in programming capabilities, AI demonstrates enormous potential in code design, writing, testing, and deployment, and in some knowledge-based tasks, it can even reach the level of human experts. Industry insiders generally expect that AI will further boost social productivity and become a powerful assistant in people’s work and life.

In the past, when discussing China’s internet and digital economy, people often focused on the large user base and diverse application scenarios. These advantages have continued into the AI era. In March this year, China’s daily token call volume exceeded 140 trillion, more than 1,000 times growth in two years; generative AI users surpassed 600 million, with a penetration rate of 42.8%. More importantly, China has already produced a number of self-developed large models comparable to the world’s top models, becoming an influential variable in AI innovation. From some performance rankings topping global large model lists, to multimodal model capabilities rising, and even to individual intelligent agents evolving into agent clusters… The development and growth of Chinese large models are reflected not only in quantity but also in quality, application depth, and global influence.

Last year, China’s open-source models ranked first worldwide in downloads. In our recent technical report, we reconstructed the residual connections of the Transformer architecture’s key components for the first time in ten years, from the perspective of underlying logic, enabling models to find new evolutionary space in training, sparking heated discussions both domestically and internationally. This new technology not only makes more types of domestic chips practically applicable to large model inference but also reduces the overall operational costs of large models.

Many Chinese open-source models, with their advantages in quantity and quality, are becoming benchmarks in global model evaluations and important samples for overseas research institutions, gaining greater global influence. Our team’s self-developed Chinese large model Kimi, with its high cost-performance ratio, has secured a place in the global market, becoming a “productivity tool” for many overseas developers and users, with obvious growth in paid model calls. We believe that, unlike closed-source models, China’s open-source models can be widely applied and form an ecosystem. Through the collective effort of many, they can promote global development in a “many hands make light work” manner.

Rapidly developing large models require time to accumulate and refine, relying on perseverance and dedication akin to “sitting on the bench for ten years in cold.” During our technical breakthroughs, we constantly encounter problems, and each solution often leads to new challenges. Only by focusing on problems and solving them diligently can we leave a series of solid footprints on the long journey of AI exploration, and enjoy fruitful results after becoming “friends of time.”

For thousands of years, humans have been seeking new ways to break through the boundaries of knowledge. Currently, AI has the potential to become one of the powerful levers for breakthroughs. We will contribute to global AI development through solid foundational technological innovation and an open mindset, providing inclusive AI technology results worldwide and contributing more Chinese strength to global innovation.