2024年至今，市场对AI的期待一直很高，因其应用落地国内外均低于预期，历经这一曲折，市场对AI的预期，已由大模型、大模型应用落地，直接切向了各计算平台、硬件云或端侧的真实落地推进，即市场不再偏好用哪个大模型(自有或外部)、大模型的参数迭代、应用方的资源禀赋(AI公司、大模型厂商、众多行业龙头)，而是更偏好真实推进的利润、收入、市占率/用户数等强检验指标。The market's expectations for AI have moved from big models and big model applications to actual implementation on various computing platforms, hardware clouds, or end sides.
The Zhitong Finance App learned that SDIC Securities released a research report saying that the market's expectations for AI have progressed from a big model to the implementation of various computing platforms. The essence of AIAgent is to perform an “execution” function in landing scenarios and applications in the digital world. It is reflected in the fact that machines independently complete tasks and iterate on their own based on specific goals, and places with fewer and smaller screens need AIAgent the most. As a new round of technological innovation, AIAgent can reduce costs and increase efficiency, reshape organizational structures, and rebuild competitive advantages in the AI era by participating in major processes such as marketing, management, and production. Successful iteration can hedge against declining valuations and is expected to bring improvements in business models and valuations to existing companies.
The market's expectations for AI have progressed from large models to the implementation of various computing platforms.
Since 2024, the market's expectations for AI have been high, because its application has fallen short of expectations at home and abroad. After this twists and turns, the market's expectations for AI have gone from large models and large model applications to the actual implementation of various computing platforms, hardware clouds, or end sides. That is, the market no longer favors which large model (own or external), parameter iteration of large models, and applicants' resources (AI companies, large model manufacturers, and many industry leaders), but rather prefer actual profit, revenue, market/user share ratio Isotonic test indicators.
The essence of AIAgent is to perform an “execution” function in implementation scenarios and applications in the digital world.
Depending on the scenario in which the operation is performed, it is broken down into two paths, the physical world and the digital world, to introduce the ultimate form of AI or artificial intelligence. The bank believes that the essence of AIAgent is to perform an “execution” function in implementation scenarios and applications in the digital world, which is reflected in the fact that the machine independently completes tasks and iterates on its own based on specific goals. The products that have appeared so far include Auto-GPT, Babyagi, AgentGPT, Microsoft Jarvis, ChaosGPT, etc., but AIAgent cannot be executed directly in the physical world; it is more of a human assistant, so it needs to interact with humans and cooperate with smart terminals such as mobile phones, MR, and AiPin to display information and accept instructions.
AIAgent is most needed where there are fewer and smaller screens
The bank believes that the most basic function of AIAgent as a personal assistant is to complete interaction with humans, understand intentions, and complete implementation. The difference from traditional software is that it does not require manual intervention, so it does not rely on interactive devices such as large screens, mice, keyboards, etc., so in many scenarios, places with fewer and fewer interactive devices such as screens are more dependent on AIAgent, and it is expected that it will advance faster.
As a new round of technological innovation, AIAgent is expected to bring improvements in business models and valuation to existing companies.
For AI companies such as Shang Tang and the Fourth Paradigm, they can use AIAgent to smoothly expand or switch from the traditional ToB business model to expand or switch to TOC to expand the scope of business and optimize financial statements (increase in gross margin); for large model manufacturers, with AIAgent, big model manufacturers can expand their business scope to all sides, and the virtual world and the real physical world can simultaneously explore the implementation of big models; for a wide range of industries, especially leading companies, they can reduce costs and efficiency by taking advantage of AIAgent, marketing, management, and production. , reshape the organizational structure, and rebuild competitive advantages in the AI era. Successful iteration can hedge against declining valuations.
Investment suggestions: AI's highly elastic targets are expected to be repaired in the fourth quarter, focusing on: Fourth Paradigm (06682), Shangtang-W (00020), Mango Supermedia (300413.SZ), China Film (600977.SH), Shengguang Group (002400.SZ), Times Publishing (600551.SH), Anhui New Media (601801.SH), Phoenix Media (), Jiecheng (300182.SZ), China Public Education (002607.SZ), Gomai Culture (301052. 601928.SH SZ), China Digital Media (000156.SZ), China Publishing (601949.SH), etc.
Risk warning: Changes in underlying logic, operational management and governance, and changing market preferences.
