中新网上海9月12日电 (姜煜高志苗)以"AI向善·安全发展——大模型在严谨行业应用中的伦理与治理"为 主题的外滩大会分论坛，12日在上海"2025外滩大会"现场举办，专家学者在论坛上热议医疗AI应用。 "在医疗行业，AI的决策直接关乎生命健康与社会公平，数据滥用、算法偏见、责任模糊等问题亟待解 决。"中国互联网协会副理事长黄澄清在论坛致辞中表示，构建可信、可靠、可追溯的AI治理体系已成 为技术发展的内在要求，也是社会责任的必然选择。 "医疗大模型的发展，我觉得对社会最大的贡献是有可能促进医疗公平公正。"复旦大学哲学学院教授王 国豫说，AI在医疗领域的应用是偏远地区医生的好助手，为优势医疗资源难以触达和覆盖的病患群体 带去了期望和福音。 上海交通大学医学院附属瑞金医院信息中心主任赵艳认为，AI的出现让患者不仅在诊前就能对自身情 况进行较科学的基础了解，也在诊中和诊后促进了患者对医生诊疗方案的依从性，从而可以提升配合治 疗的效果。 企业界代表、蚂蚁集团资深副总裁及科技伦理委员会联席主席周志峰说："我们还是希望通过预警、测 试、修复这样的动态机制来提升AI的风险防控能力。"他在论坛上介绍了蚂蚁集团在AI大模型领域 ...

Core Viewpoint - The company is undergoing significant transformation and growth, driven by strategic initiatives and market dynamics, while addressing challenges posed by industry policies and competition [2][4]. Industry and Company Development Trends - The company has eliminated goodwill and COVID-19 related reporting interference, leading to a critical period for increasing market concentration among leading firms as the industry accelerates the exit of smaller players [2]. - Despite pressures from policies like DRG and centralized procurement, these have also catalyzed the company's business and product structure transformation, enhancing market share through a "product + service + digitalization" model [2]. Five-Year Strategic Plan and Half-Year Performance - Cost Control: The company achieved over a 25% reduction in procurement costs, with management and financial expenses decreasing by 16% and 29% year-on-year, respectively [3]. - Technical Competitiveness: Key business areas such as pathogen tNGS, hematology, and tumor companion diagnostics grew by 35%, 22%, and 20%, respectively, with special inspection revenue now accounting for 47.63% of diagnostic service revenue, up 7.32 percentage points from the end of 2024 [3]. - Development of Proprietary Products: The Kai Le Pu reagent consumables business grew by 43%, and the company has integrated smart products into its core strategy, launching a three-year plan for medical AI [3]. - Customer Structure Optimization: The company signed 1,036 new clients, including 133 tertiary hospitals, with revenue from tertiary hospitals now making up 49.28% of total revenue, an increase of 6.46 percentage points from the end of 2024 [3]. - Internationalization Progress: The company’s Vietnam branch received ISO15189 certification and is actively expanding into the "Belt and Road" markets, collaborating with domestic IVD companies to develop business in the Middle East [3]. Investor Inquiry Response Highlights - Impact of Package Splitting Policy: The company is enhancing competitiveness through regional integration and cost reduction, aiming to maintain stable gross margins and expand market share [4]. - Accounts Receivable and Cash Flow: The company expects better cash flow from regular business in the second half of the year, with COVID-19 related receivables anticipated to be fully accounted for by year-end [4]. - Development of Technical Talent: The company is advancing discipline construction and talent acquisition, having introduced 107 new testing projects and initiated a training program for commercial talent [4]. - Market Share and Precision Centers: The company is rapidly increasing market share, with 61 precision centers now profitable, and expects a doubling of revenue from precision centers by year-end compared to the previous year [4]. - Smart Product Advantages and Planning: The company’s smart product revenue has already surpassed the total for the previous year, with plans for product pipeline updates in the second half [4]. - Service Gross Margin Improvement: The service gross margin has increased by 0.8 percentage points year-on-year, with expectations for this trend to continue [4]. - Data Asset Value: The company conducts approximately 160 million tests annually, accumulating over 20PB of data, which can be utilized for research, AI training, and customer data services [4]. - Outbound Planning Progress: The company is promoting its services and products internationally through a "product + service + digitalization" approach, leveraging the "Belt and Road" initiative [4].

Core Insights - The development of AI in healthcare is at a significant turning point, transitioning from technological exploration to clinical application [2] - The healthcare industry faces common challenges such as physician overload, uneven distribution of quality resources, and weak grassroots capabilities, necessitating structural transformation [2] - Philips invests nearly 10% of its global revenue in R&D, with over half allocated to AI, data, and software, focusing on four key areas: operational efficiency, clinical decision support, expanding healthcare accessibility, and health management [2] Group 1 - AI is seen as a means to enhance productivity for doctors, allowing them to spend more time with patients [2] - The principle behind Philips' AI implementation is centered on being human-centric, trustworthy, and sustainable [2] - AI has the potential to improve healthcare accessibility, exemplified by a remote surgery completed by doctors in Shanghai and a hospital in Tibet [3] Group 2 - Philips aims to transition healthcare AI from being merely "available" to "trustworthy" and from "isolated breakthroughs" to "system integration" [3] - The focus is on leveraging technology as a bridge and collaboration as a foundation to drive advancements in medical AI [3]

Core Viewpoint - The article discusses the importance of AI ethics, particularly in the medical field, emphasizing the need for "human-machine alignment" to ensure AI technologies align with human values and societal norms [2][3]. Group 1: Human-Machine Alignment - Human-machine alignment is defined as the process of ensuring AI's goals, behaviors, and outputs are consistent with human values and social norms, representing a systematic approach to addressing AI ethical issues [3]. - The concept of human-machine alignment has historical roots, with its principles being validated through practical applications in AI technology [3][6]. Group 2: Importance in Medical AI - In the medical field, human-machine alignment serves three core functions: explainability, trustworthiness, and human harmony [4][5]. - Explainability allows AI to present clear decision-making logic, which helps alleviate concerns from both doctors and patients [4]. - Trust is built when AI recommendations adhere to medical ethics, enabling humans to rely on AI for health-related decisions [5]. - Human harmony ensures that AI applications do not deviate from genuine human needs, incorporating emotional and ethical considerations into algorithm design [5]. Group 3: Ethical Compliance in Medical AI - Medical AI applications face unique challenges, including data sensitivity, irreversible outcomes, and complex responsibility structures [7]. - A collaborative approach across five key areas—technical architecture, data set construction, hospital management, patient awareness, and industry regulation—is essential for ensuring ethical compliance in medical AI [7][9]. Group 4: Data Mechanisms - Establishing a "data flywheel" mechanism is crucial for continuous model optimization, creating a closed-loop system that integrates user feedback into AI development [11]. - A dual mechanism for data access and incentives is necessary to ensure data quality and encourage participation from hospitals and doctors in the alignment process [12]. Group 5: Regulatory Framework - A unified national certification standard for medical AI alignment should be established, with third-party evaluations to ensure compliance and robustness [10]. - Regular assessments by multidisciplinary ethical committees can help maintain alignment and prevent technological biases [10].