卡内基梅隆大学计算机教授、月之暗面创始人杨植麟的博士导师Russ Salakhutdinov近日接受专访，分享了这位杰出学生的学术成就与创业选择。杨植麟博士论文获得2万次引用，仅用四年完成通常需六年的博士学业[1]。面对苹果的邀请，他坚定地拒绝了，杨植麟曾向导师坦言："若不尝试创业，一定会遗憾"，随后毅然回国创立月之暗面[1]。
月之暗面最新估值已达350亿美元，较去年12月的43亿美元大幅增长7倍[1]。公司开源的Kimi K3模型在Hugging Face平台发布后30分钟内获得超过4000个赞，创造了迄今最快的发布增长速度[1]。该模型在全球AI大模型周调用量榜单中排名第十，调用量达1.42万亿Token[1]。
Salakhutdinov认为，Kimi K3最独特的优势在于其开源权重，他表示"一个拥有越来越多开源模型的世界，会是一个更好的世界"[1]。不过他也指出，行业仍需通过更多实践应用才能判断其真实能力[1]。
Russ Salakhutdinov, a Carnegie Mellon University computer science professor and doctoral advisor to Yang Zhilin, revealed in an exclusive interview the academic background and entrepreneurial journey of the founder of Moonshot AI.[1] Yang Zhilin turned down an invitation from Apple to return to China and establish Moonshot AI, the company behind the Kimi K3 model.[1] His doctoral dissertation achieved 20,000 citations, and he completed his Ph.D. in just four years, typically a six-year endeavor.[1] Yang Zhilin once confided in his advisor: "If I don't attempt entrepreneurship, I will certainly regret it."[1]
Moonshot AI's latest valuation has reached $35 billion, representing a sevenfold increase from the $4.3 billion valuation recorded in December of the previous year.[1] The company's newly released Kimi K3 model has garnered significant attention in the open-source community. Within 30 minutes of releasing the complete model weights on Hugging Face, Kimi K3 accumulated over 4,000 likes, marking the fastest launch growth rate to date.[1] Salakhutdinov identified the open-source model weights as Kimi K3's most distinctive advantage and emphasized that "a world with an ever-growing number of open-source models would be a better world."[1] However, he cautioned that the industry still requires more practical applications to accurately assess the model's true capabilities.[1] Currently, Kimi K3 ranks tenth on the global AI large language model weekly call volume leaderboard, with a call volume of 1.42 trillion tokens.[1]
