17-Year-Old 'Kimi' Genius Leaves Moonshot and Denies Joining Anthropic

Chen Guangyu, the young engineer who co-developed the architecture of the "Kimi K3" model at the Chinese company Moonshot, announced that he has left the company, according to reports published by Ming Pao and Zaobao and cited by SBS.

Chen, 17, said on social media that the past year was the best of his life, expressing his gratitude for the company's trust in hiring him at age 16, after working as a machine learning researcher on the "Kimi" team since last November.

When a researcher from the US company Anthropic asked him if he intended to attend Carnegie Mellon University and expressed interest in having him join, he replied that he would be heading to the San Francisco Bay Area, but later denied joining Anthropic. This comes amidst intense competition between Washington and Beijing for AI talent.

Moonshot launched "Kimi K3" last July, describing it as one of the world's largest open-weight models with 2.8 trillion parameters, with performance close to top US models, amidst speculation that the youth will head to university after graduating from high school.

What Do These Terms Mean?

Open-weight: An AI model whose weights are published, making it available for use and review, unlike closed models.

Parameters: The internal numerical values that the model learns during training, increasing in number with the model's size and capability.

Model Architecture: The internal design that determines how the model processes data and generates answers.