Core Insights - The article discusses the investment trends in AI for 2025, highlighting the experiences and observations of Jinqiu Fund over the past year in the AI startup and investment landscape [4][10]. Investment Focus - Jinqiu Fund emphasizes three key aspects: a focus on AI, a 12-year investment cycle, and an active investment strategy, having invested in over 50 AI projects in the past year, ranking among the top two in the industry [10][11]. - The majority of investments (56%) are in the application layer, with 25% in embodied intelligence, and 10% in computing infrastructure, reflecting a strategic focus on areas that support long-term model cost reduction [11][22]. Market Comparison - A comparison with 20 active VC and CVC firms shows that Jinqiu Fund is more heavily weighted towards application-focused investments, indicating a differentiated strategy in the AI investment landscape [14][16]. Trends in AI Development - The article outlines two major trends: the enhancement of intelligence and the reduction of costs associated with acquiring intelligence. The focus is on the transition from pre-training to post-training using high-quality datasets [25][32]. - The cost of acquiring intelligence is decreasing significantly, with a notable drop in the cost per token and the emergence of new benchmarks for model capabilities, leading to a more competitive environment for application-layer companies [33][34]. Opportunities in Application Layer - Jinqiu Fund has been closely monitoring opportunities in the application layer since the second half of last year, driven by the belief that the time for application-layer opportunities has truly arrived [38][39]. - The article suggests that key variables from the internet and mobile internet eras can be applied to analyze changes and opportunities in the AI application layer, emphasizing the importance of user data and context [41][47]. Embodied Intelligence - The future of embodied intelligence is seen as crucial for building physical world agent applications, although the foundational models have not yet reached a breakthrough moment akin to GPT [56][61]. - The article stresses the importance of hardware in the early stages of development, highlighting the need for effective collaboration between hardware and software to enhance algorithm development and deployment [58][61].

Group 1 - The core idea of the article revolves around three critical concepts for understanding and navigating AI development by 2025: the Verifiers Law, the Jagged Edge of Intelligence, and the commoditization of intelligence [3][14]. - The Verifiers Law states that the ease of training AI to complete a specific task is proportional to the verifiability of that task, suggesting that tasks that are both solvable and easily verifiable will eventually be tackled by AI [21][26]. - The concept of intelligent commoditization indicates that knowledge and reasoning will become increasingly accessible and affordable, leading to a significant reduction in the cost of achieving specific intelligence levels over time [9][11]. Group 2 - The article discusses the two phases of AI development: the initial phase where researchers work to unlock new capabilities, and the subsequent phase where these capabilities are commoditized, resulting in decreasing costs for achieving specific performance levels [11][13]. - The trend of commoditization is driven by adaptive computing, which allows for the adjustment of computational resources based on task complexity, thereby reducing costs [13][16]. - The article highlights the evolution of information retrieval across different eras, emphasizing the drastic reduction in time required to access public information as AI technologies advance [16][17]. Group 3 - The Jagged Edge of Intelligence concept illustrates that AI's capabilities and progress will vary significantly across different tasks, leading to an uneven development landscape [37][42]. - The article suggests that tasks that are easy to verify will be the first to be automated, and emphasizes the importance of creating objective and scalable evaluation methods for various fields [38][39]. - The discussion includes the notion that AI's self-improvement capabilities will not lead to a sudden leap in intelligence but rather a gradual enhancement across different tasks, with varying rates of progress [41][45].