Core Insights - The article discusses the challenges faced by the AI industry as it attempts to transition from experimental applications to widespread implementation in various sectors, highlighting three main difficulties: technology adaptability, data quality and availability, and high costs [9][10]. Group 1: AI Application Challenges - AI is entering a "deep water" phase where large-scale deployment faces significant challenges, including technology that is not specialized enough for specific industries [10]. - The current state of foundational models is still developing, with issues such as poor interpretability and high hallucination rates, making it difficult to find suitable application scenarios [10][11]. - The industrial sector faces a "three highs" dilemma: high entry barriers, high operational costs, and high safety risks, necessitating a deep understanding of complex processes and implicit knowledge [10][11]. Group 2: Data Quality Issues - High-quality data is essential for training industry-specific models, but there is a notable lack of quality data across different sectors, leading to "data islands" and inconsistent data quality [10][11]. - Legal restrictions, such as data security laws and personal information protection laws, hinder the large-scale application of existing data, particularly in the healthcare sector [11][12]. - The transition from non-digital to digital data is also constrained by intellectual property laws, further exacerbating the shortage of high-quality industry-specific data [11][12]. Group 3: Cost Barriers - The high costs associated with customized AI services, including computing power, model development, and data management, pose a significant burden for small and medium-sized enterprises (SMEs) [10][12]. - There is a need for differentiated support policies for various types of enterprises, including state-owned enterprises, industry leaders, and SMEs, to facilitate the implementation of AI initiatives [12].

Core Viewpoint - The article discusses the rapid advancement and application of artificial intelligence (AI) across various industries, highlighting both the opportunities and challenges faced in the implementation of the "AI+" initiative in China [1][4][9]. Group 1: AI Applications and Innovations - At the 2025 AGIC Shenzhen conference, AI was showcased through various applications, including a coffee-making robot that can create latte art in 90 seconds and a sorting robot that won multiple awards for its efficiency [3][6]. - The "AI+" initiative aims to integrate AI into everyday life and industries, with a goal of achieving over 90% application penetration of new intelligent terminals and agents by 2030 [6][7]. - Companies like Anno Robotics and Lingyi Intelligent Manufacturing are leading the way in automating food and beverage services, demonstrating the practical benefits of AI in enhancing operational efficiency [3][6]. Group 2: Challenges in AI Implementation - Despite the advancements, AI faces significant challenges in entering the "deep water" of application, including issues of technology adaptability, data quality, and high costs [4][9][10]. - The lack of specialized industry models and the presence of "data silos" hinder the effective training and deployment of AI systems, making it difficult to meet the specific needs of various sectors [9][10]. - High operational costs and the complexity of customized AI solutions pose barriers for small and medium-sized enterprises, limiting their ability to adopt AI technologies [10][11]. Group 3: Government Initiatives and Support - The Chinese government has initiated the "AI+" action plan, encouraging local governments to tailor their strategies based on regional strengths and industry characteristics [6][7]. - Provinces like Zhejiang and Shanghai are focusing on specific sectors such as healthcare and manufacturing to drive AI integration, showcasing the importance of localized approaches [7][8]. - Experts suggest that differentiated support policies for various types of enterprises, including public service platforms, could help reduce costs and promote innovation in AI applications [11].