Core Insights - The event highlighted the importance of focusing on credible foundations and risk management in digital financial innovation, as emphasized by industry experts [1][2] - The rise of digital assets and the need for financial institutions to adapt to new trends in the global digital asset market were discussed [1][3] Group 1: Digital Financial Innovation - Li Lihui stressed that short to medium-term digital financial innovation should not rely on vertical models to solve complex problems but should focus on credible foundations and risk management [1][2] - He proposed four key focus areas: high reliability, interpretability, legality, and economic efficiency in deploying AI models within financial institutions [2] - The need for a safe and efficient innovation system was highlighted, along with the importance of bridging the digital divide for smaller financial institutions [2] Group 2: Trends in Digital Assets - Huang Yiping pointed out significant changes in the global digital asset market, including the rise of stablecoins, which now account for over 90% of transactions in the virtual asset market [3] - The rapid development of asset tokenization was noted, with predictions of substantial growth in the next five years as traditional financial institutions engage in this area [3] - The emergence of virtual currency exchange-traded funds (ETFs) provides investors with a way to participate in the market without holding virtual currencies directly [3] Group 3: Risks and Opportunities - Huang Yiping indicated that the changes in the digital asset market could lead to a certain level of substitutability with central bank digital currencies, potentially affecting their future scenarios [3] - The correlation of risks between the virtual asset market and traditional asset markets is increasing, warranting caution despite no direct risk transmission to the dollar market observed yet [3] - The development of the global digital asset market is seen as a trend with both risks and opportunities, suggesting that engagement in trend-driven innovation should be pursued under controlled risk conditions [3]

Core Insights - Google has announced a significant update to its Gemini model, introducing Gemini 2.5 Pro and Gemini 2.5 Flash, with the Flash-Lite version in preview [2] Model Performance - Gemini 2.5 Pro is noted for its advanced reasoning and programming capabilities, achieving state-of-the-art (SOTA) performance in long context tasks with a context length of 1 million+ tokens [4] - In various benchmark tests, Gemini 2.5 Pro scored the highest in tasks such as Aider Polyglot programming, Humanity's Last Exam, and GPQA [4] - The model outperformed Gemini 1.5 Pro by over 120 points and surpassed competitors like OpenAI, xAI, and Anthropic, although it lagged behind OpenAI in mathematics and image understanding [4] Model Features - Gemini 2.5 Flash is a hybrid reasoning model designed for complex tasks, balancing quality, cost, and latency effectively [5] - The Flash-Lite version is an economical upgrade, excelling in high-capacity, latency-sensitive tasks like translation and classification, with faster token decoding speeds [5] Pricing Structure - Pricing for Gemini 2.5 Pro is set at $1.25 per million tokens for input and $10.00 for output [6] - Gemini 2.5 Flash has an input price of $0.30 and an output price of $2.50 per million tokens [6] - Gemini 2.5 Flash-Lite offers a significant cost advantage, with input prices at $0.10 and output prices at $0.40 per million tokens, making it 30%-60% cheaper than Gemini 2.5 Flash [7]

Core Viewpoint - A significant corruption case within the payment industry has been revealed, involving Liu Guoliang, the former head of China UnionPay's business operations center, who embezzled over 19 million yuan in just three years, highlighting a deep-rooted gray interest chain in the payment industry [2] Group 1: Corruption Case Details - Liu Guoliang exploited loopholes from the 2016 credit card transaction fee reform, misclassifying high-profit merchants to obtain illegal profits [2] - The case involved substantial bribes, including 7.665 million yuan from Haike Rongtong and over 1 million yuan in "training fees" from Fu Linmen, which went undetected for three years [3] Group 2: Internal Control Failures - The internal control mechanisms within China UnionPay failed, as the "review-recheck-approval" system became centralized under Liu Guoliang, leading to unchecked power [5] - The payment system's technical safeguards failed to monitor abnormal fee changes and verify the logical relationship between merchant scale and fee tiers, allowing for rampant power abuse [5] Group 3: Broader Implications - The case reflects a collusive ecosystem within the entire payment industry, where payment institutions disguised profit transfers, and internal oversight ignored irregularities [5] - The lack of transparent supervision allowed small powers to significantly impact national financial security, as evidenced by similar corruption cases [5] Group 4: Future Strategies - The new leadership at China UnionPay aims to implement a "platformization and digital intelligence" strategy, emphasizing the need for robust institutional frameworks to restore trust [6] - Proposed measures include using AI to monitor fee changes in real-time and employing blockchain technology to enhance transaction transparency, breaking the centralized power structure [6]

Core Viewpoint - The article discusses the cost-effectiveness of deploying AI models like DeepSeek-V3 at scale compared to running them locally, highlighting the trade-off between throughput and latency in AI inference services [2][13]. Group 1: Cost and Performance of AI Models - DeepSeek-V3 appears to be fast and cost-effective for large-scale deployment, but running it locally is slow and expensive due to low GPU utilization [2][13]. - The fundamental trade-off in AI inference services is between high throughput with high latency and low throughput with low latency [2][11]. Group 2: Batch Inference - Batch inference allows for efficient processing of multiple tokens simultaneously, leveraging GPU capabilities for large matrix multiplications (GEMM) [3][11]. - The implementation of inference servers involves receiving requests, pre-filling prompts, queuing tokens, and processing them in batches to maximize GPU efficiency [4][11]. Group 3: GPU Efficiency and Model Design - High batch sizes are necessary for models like expert mixture models (MoE) to maintain GPU efficiency, as they require many small multiplications unless batch processing is employed [7][11]. - Large pipelines in models necessitate high batch sizes to avoid pipeline bubbles, ensuring that GPUs remain active throughout the inference process [8][9]. Group 4: Latency and Throughput Trade-offs - Increasing batch size can lead to higher latency as users may need to wait for enough tokens to fill a batch, but it significantly improves throughput [11][12]. - The choice of batch size and collection window directly impacts the balance between throughput and latency, with larger windows helping to avoid pipeline bubbles [9][11]. Group 5: Implications for AI Service Providers - AI service providers must select batch sizes that eliminate pipeline bubbles and keep experts saturated, which often results in higher latency for improved throughput [11][13]. - The architecture of models like DeepSeek may not be easily adaptable for personal use due to their low efficiency when run by a single user [13].

Group 1: Investor Relations Activities - The company held an annual performance briefing on May 16, 2025, via an online platform [3] - Participants included the chairman, chief strategy officer, general manager, financial officer, board secretary, and sponsor representative [3] Group 2: Business Development Plans - The company will focus on the smart inspection field, enhancing AI models, drones, and robotics applications [4] - Plans to develop vertical demonstration projects across various industries, including power, energy, and transportation [4] Group 3: Mergers and Acquisitions - Currently, there are no plans for mergers or acquisitions to enhance market value [5] Group 4: Stock Buyback and Value Management - No stock buyback plans are in place; the company emphasizes maximizing value through core business focus and performance [6] - Cumulative cash dividends over the past three years amount to 30 million RMB [6] Group 5: Drone Technology Developments - The company has developed a drone intelligent inspection system with various management and monitoring features [7] - Future plans include optimizing drone inspections and integrating technologies like IoT, AI, and digital twins [7] Group 6: Research and Development Achievements - In 2024, the company invested 19.42 million RMB in R&D, accounting for 24.09% of revenue [8] - New developments include an upgraded smart operation and inspection platform and a lightweight mobile intelligent terminal [8] Group 7: External Investments - The company has invested in three firms focused on solid-state battery technology, special robotics, and intelligent road marking equipment [9] - All invested companies are operating normally, with steady progress in product development and market expansion [9]

Core Insights - The 4th Changsha International Construction Machinery Exhibition concluded, attracting over 350,000 visitors from more than 110 countries, making it one of the largest and most influential events in the global construction machinery industry [1][2] - The exhibition featured 1,806 companies, a 20% increase from the previous edition, with 35 of the world's top 50 construction machinery companies showcasing over 20,000 exhibits [1] - A total of 130 billion yuan in procurement contracts were signed by six major manufacturers, focusing on key areas such as hydraulic systems and intelligent control [1] Industry Developments - The exhibition included 35 new product launches and over 1,500 new technologies, with more than 60% of exhibits being high-end and intelligent, and 75% being environmentally friendly [2] - New segments for emergency rescue, mining equipment, and transportation equipment were introduced, filling gaps in the construction machinery exhibition [2] - International exchanges were enhanced through six international matchmaking events, successfully pairing over 600 participating companies [2] Future Outlook - Changsha aims to establish the exhibition as a key platform for global machinery equipment companies to showcase products and foster collaboration, with plans for the 5th exhibition in May 2027 [3]

证券代码：002889 证券简称：东方嘉盛 深圳市东方嘉盛供应链股份有限公司 投资者关系活动记录表 编号：2025-002 投资者关系活动类别 特定对象调研 ☐分析师会议 ☐媒体采访 ☐业绩说明会 ☐新闻发布会 ☐路演活动 ☐现场参观 ☐其他（请文字说明其他活动内容） 参与单位名称及人员姓名 永赢基金 张海啸 国泰基金 林知 时间 2025年05月12日 15:00-17:00 地点 上海市浦东新区二十一世纪中心大厦 上海市虹口区嘉昱大厦 上市公司接待人员姓名 董事会秘书 李旭阳 投资者关系经理 肖文 投资者关系活动主要内容 介绍 1.国际形势缓和对公司的业务有什么影响？ 回答1：国际形势缓和将减少行业波动因素，带动提振国际消 费与物流需求，长期来看将增强跨境供应链的稳定性，助力各行业 进出口业务的增长。此外，目前公司现已经加速全球供应链服务网 络布局，同步布局欧洲、中亚、中东及拉美等新兴市场，分散单一 市场风险，凭借跨境多式联运产品、全球供应链服务网络以及垂直 赛道的供应链体系建设，寻求更多市场及业务机会。同时，公司也 正积极引入AI模型以及数字化系统实时跟踪订单、库存及物流动 态，协同产业链上下游进行风险 ...

Core Viewpoint - The article discusses a significant study identifying 64 genes that influence brain aging speed and suggests 13 potential anti-aging drugs, utilizing AI models to analyze brain scans and genetic data [1][3]. Research Overview - The study is noted as the largest attempt to determine genetic factors affecting organ aging, with implications for developing new brain anti-aging drugs [3]. - The research aims to identify factors leading to brain aging and explore potential solutions [5]. Methodology - The study uses Brain Age Gap (BAG) as a marker for brain aging, defined as the difference between predicted brain age and actual age [6]. - Data from 29,097 healthy participants in the UK Biobank was used to train seven AI models for brain age estimation [8]. - Validation was conducted using data from 3,227 healthy and 6,637 brain disease subjects, employing various assessment metrics [9][10]. Genetic Analysis - A Genome-Wide Association Study (GWAS) was performed on 31,520 healthy participants to identify genetic variations associated with BAG [11][12]. - The study explored the causal relationship between BAG and 18 brain diseases, finding a significant impact on intelligence [13][14]. Drug Discovery - The research identified 64 druggable genes linked to biological pathways related to brain aging, suggesting that targeting these genes could help combat aging or related diseases [14][15]. - A drug repurposing analysis revealed 466 potential anti-aging drugs, with 29 showing promise in delaying brain aging [17][18]. - Among these, 20 drugs, including Dasatinib and Diclofenac, have been previously noted for their anti-aging potential, with 13 currently undergoing clinical trials [19][20].