share_log

国信证券:大模型KIMI带动产业链革新 国产AI算力迎来发展机会

Guoxin Securities: Big model KIMI drives industrial chain innovation, domestic AI computing power ushered in development opportunities

Zhitong Finance ·  Mar 20 23:17

Kimi became the dawn of a major domestic model, driving innovation in the industrial chain.

The Zhitong Finance App learned that Guoxin Securities released a research report saying that Kimi has become a major domestic model at the dawn, driving innovation in the industrial chain. In October 2023, Kimi, launched by Yang Zhilin of Tsinghua University and his team “Dark Side of the Moon”, marked an important advance in the domestic AI model. At the same time, the multi-modal model is driving rapid growth in global demand for computing power. Combined, the US will restrict cloud vendors from providing AI cloud services to Chinese customers, and domestic AI computing power will usher in development opportunities. It is recommended to focus on Haiguang Information (688041.SH), a leading domestic AI computing power company; 2) The rapid improvement of large model capabilities and multi-modality will further expand the scope of AI applications. Furthermore, with the reduction in the cost of large AI models and technological development, the AI application industry will advance rapidly. It is recommended to focus on individual stocks related to AI applications, such as Jinshan Office (688111.SH) and Tonghuashun (300033.SZ).

The main views of Guoxin Securities are as follows:

Guoxin Securities believes that with its excellent ability to process long texts, Kimi can handle up to 200,000 words of input, showing its huge potential for non-destructive reading. This breakthrough not only improved the efficiency of content creation and organization, but also brought deepening and innovation to fields such as novel and script creation, while opening up new application scenarios in game interaction, AI companionship, and task execution in professional fields.

The growth of AI applications around the world is generally weak, and Kimi has broken through with its significant user growth and application breadth

According to Guoxin Securities, from October 2023 to March 2024, Kimi's daily active users grew rapidly from 100,000 to more than 3 million. This significant increase reflects Kimi's successful strategy in model optimization, talent expansion, and user attraction. Kimi's success depends not only on its technical advantages, but also on its emphasis on user experience, including through data-driven continuous product optimization, innovative sharing mechanisms, and precise refinement of core functions. These factors together enhance Kimi's market competitiveness.

Sora model creates a new era of AI video

Guoxin Securities said that OpenAI recently released its first large-scale video generation model Sora, marking an important breakthrough in Wensheng's video field. Sora has the ability to generate videos up to 60 seconds long and closely aligned with user instructions, which shows significant advantages in terms of video content length, multi-angle consistency, and understanding of the physical world.

Technically, Sora uses an innovative method, using known image fragments (Patches) to guess the next clip, and combines the Transformer model with Diffusion technology to achieve efficient generation of complex video content. The advent of Sora not only represents a new milestone for AI in the field of video generation, but also has had a profound impact on the development of multi-modal large models.

As such models integrate the processing capabilities of different types of data such as video, images, and text, the demand for computing power has also increased dramatically. This change indicates that in the future, more computational resources will be required to train efficient and powerful AI models.

Risk warning: macroeconomic recovery falls short of expectations; capital expenditure of cloud vendors falls short of expectations; market competition intensifies; product development falls short of expectations; introduction of domestic AI computing power chips falls short of expectations, etc.

Disclaimer: This content is for informational and educational purposes only and does not constitute a recommendation or endorsement of any specific investment or investment strategy. Read more
    Write a comment