The smart chatbot Kimi Chat, developed by the start-up in October, was introduced utilizing its proprietary Moonshot large language model (LLM). This innovative chatbot has the capacity to analyze up to 200,000 Chinese characters within a contextual window, which refers to the extent of text that the AI model can process while engaging with users.
Alibaba, HongShan, and Moonshot AI did not provide immediate responses to inquiries for comments on Tuesday.
Moonshot AI’s recent funding round represents the most substantial single financing secured by a Chinese AI start-up since the launch of ChatGPT in November 2022.
The latest investment infusion received by Moonshot AI underscores the continued interest in generative AI start-ups within mainland China, which led global investments in such enterprises during the first half of 2023.
According to a report by the AI-focused research firm Zhidongxi, China boasted the highest number of generative AI start-ups that received funding at that time. The mainland was home to 22 out of the 51 such start-ups worldwide, which collectively raised approximately US$13.8 billion in investments during the specified period.
Moonshot AI, established in April 2023 by Yang Zhilin, a computer science graduate from Beijing’s esteemed Tsinghua University, has garnered attention in the industry. Yang, who later obtained his doctoral degree from Carnegie Mellon University in Pittsburgh, Pennsylvania, emphasized the significance of processing lengthy texts for the advancement of AI models during an industry event last November.
The effectiveness of a large language model (LLM) is, in part, determined by the number of parameters it possesses, indicating the model’s complexity. While ChatGPT was trained on 175 billion parameters, reports from local media suggest that Moonshot LLM surpasses this with over 100 billion parameters.
In a similar vein, Beijing-based Baichuan claimed that its Baichuan2-192k LLM had outperformed Claude 2, the AI model developed by Amazon.com-backed Anthropic, positioning itself as the most advanced LLM globally in terms of the volume of words users could incorporate in their chat queries. Claude 2’s context window was reported to be capable of analyzing around 75,000 English words, equivalent to hundreds of pages of documents or a substantial book.
As of November, mainland China was home to 254 LLMs, as outlined in a white paper released by the Beijing Municipal Science and Technology Commission.
Despite the proliferation of domestic LLMs, criticisms have emerged from Baidu’s founder, chairman, and chief executive Robin Li Yanhong, who described the trend as a “huge waste of resources.” Speaking at an event in Shenzhen last November, Li expressed concerns over the surplus of LLMs in China, highlighting the scarcity of AI-native applications based on these models.
Furthermore, Moonshot AI, situated in Beijing, may encounter competition from another start-up with the same name based in New York City. The U.S.-based firm, as indicated on its website, offers an AI platform designed to assist e-commerce stores in automating their online operations entirely to boost conversions, sales, and revenue.
