7月28日下午，2025世界人工智能大会暨人工智能全球治理高级别会议闭幕式在上海世博中心举行。 今年，"智能时代 同球共济"的主题贯穿始终，为全球人工智能领域带来一场思想与创新碰撞的盛宴。 大会汇聚了来自70多个国家和地区、1500余位顶尖专家。大会展览面积首次突破7万平方米，较上一届提高40%，吸引800余家企业参展，展示前沿科技 3000余项，包括100余款"全球首发""中国首秀"产品，展品总数、首发首秀实现"双倍增"。 另外，展会创新设置了Future Tech中小企业展，集中展现了200余个具有创新前沿、成长潜力、创业价值的初创项目。 截至7月28日14时，大会接待了来自中国、英国、西班牙、约旦等十多个国家的156个采购团组。大会期间，现场发布了300余项采购需求，组织了60场 项目路演，共招募225个早期创业项目现场展示交流，初创企业与投资人有效对接超过2000人次，现场触达意向客户超1200人次，预计达成意向采购金额超 160亿元。本届大会线下参观人数预计将达到35万人次，中外媒体累计发布5000余篇报道。 上海市重大项目 CONTRACT SIGNING CEREMONY FOR SH 潘洁 摄 ...

Core Insights - The 2025 World Artificial Intelligence Conference (WAIC) showcased a shift in focus from pure technical parameters to practical applications and commercial value in AI technology [2][14] - The competition in computing power is evolving into a comprehensive system engineering challenge, addressing performance, compatibility, storage, and energy efficiency [4][10] - AI companies are increasingly integrating their models with real-world applications to unlock new data sources and enhance AI capabilities [15][16] Computing Power Infrastructure - Companies like Huawei and China Digital are pushing the limits of computing power, with Huawei's Atlas 900 A3 SuperPoD achieving a performance of 300 PFLOPS [2][4] - The financial sector is supporting AI infrastructure, with companies like Chip Xin Leasing investing 8 billion yuan in AI-related projects [4] - The demand for private deployment of large models is increasing due to data security concerns, indicating a shift in market needs [5][6] Model and Application Development - AI model developers are focusing on deep integration with industry scenarios to create real business value, moving away from mere technical showcases [14][17] - Companies like Step Leap Star are launching new models aimed at cost reduction and efficiency improvement, collaborating with multiple chip manufacturers to enhance compatibility [17][18] - The importance of data storage and management is highlighted, with companies like Dawning Storage addressing challenges in data accessibility and efficiency [8][9] AI in Creative Industries - AI-generated content (AIGC) is transforming creative processes, with companies like Digital Kingdom introducing platforms that streamline content creation [20][21] - AI is positioned as a "super assistant" for creators, enhancing productivity while allowing them to focus on core creative tasks [21] Consumer-Focused AI Products - New AI products, such as the TicNote AI recording pen, are being developed to serve individual users, encapsulating complex AI capabilities in user-friendly formats [23] - The overarching goal of AI advancements is to contribute to real GDP growth across society, industries, and nations [24]

Core Insights - SenseTime is establishing an independent embodied intelligence company, led by key figures including Chief Scientist Wang Xiaogang and Tao Dacheng, indicating a strategic shift towards this emerging sector [1][3] - The launch of the "Wuneng" embodied intelligence platform at the 2025 World Artificial Intelligence Conference (WAIC) showcases SenseTime's commitment to integrating its technology in multimodal models and computer vision into embodied intelligence [1][4] Group 1: Strategic Developments - The "Wuneng" platform is designed to provide perception, visual navigation, and multimodal interaction capabilities for hardware like robots, reflecting SenseTime's ambition to facilitate real-world interactions for various embodied intelligence enterprises [1][4] - SenseTime's organizational restructuring, termed the "1+X" strategy, aims to focus on core businesses while allowing ecosystem companies to operate independently, with the new embodied intelligence company potentially becoming another "X" business [3][4] Group 2: Technological Pathways - SenseTime's entry into embodied intelligence is seen as a necessary step towards achieving AGI (Artificial General Intelligence), emphasizing the importance of transitioning intelligence from digital to physical spaces [4][5] - The company plans to utilize a combination of internet-sourced data for pre-training multimodal models, simulation data generated through a "world model," and limited real-world operation data for model alignment, enhancing training efficiency [5]

7月26日-28日，上海世博展览馆，正成为人形机器人盛大的秀场。 在宇树科技的展台，两台人形机器人正在擂台上演"格斗赛"；不远处的数字华夏舞台上，仿生机器 人"夏澜"正与真人Coser共舞；而在智平方的展台，通用智能机器人"爱宝"则在熟练地为观众制作冰淇 淋，还能随时切换角色，敲上一段激昂的架子鼓。 与此同时，在大模型论坛上，商汤科技董事长兼CEO徐立展示的人形机器人，正对着PPT侃侃而谈，风 趣幽默地讲解着"长安的荔枝"，甚至还能随时停下与观众互动问答。 这些能与人自然交互、完成复杂任务的"新物种"，吸引着大量观众和无数闪光灯。机器人，已成为这届 世界人工智能大会（WAIC）上最受关注的"显眼包"。 然而，当目光穿透这些日益灵活的物理实体，一场更深刻、更务实的变革正在产业的"幕后"悄然发生。 南都湾财社记者观察到，相较于前两年"百模大战"的喧嚣，今年关于大模型的讨论似乎"降温"了。但这 并非热度消退，而是大模型正在从喧嚣的台前走向务实的幕后，集体聚焦"落地"这一核心场景。 从华为首次在线下展出为大模型训练打造的昇腾384超节点，到阿里通义千问宣布模型下载量突破4亿、 腾讯发布混元3D世界模型，再到阶跃星辰开 ...

Core Insights - The rise of large language models (LLMs) marks a significant leap in AI technology, but achieving Artificial General Intelligence (AGI) requires more than just text understanding and generation [2] - The development of AI is transitioning from single language models to a new stage of multimodal integration, which is essential for reaching AGI [2][3] - The future of AI lies in the fusion of multimodal information and interaction with the physical world, with a full-scale adoption of multimodal models expected by the second half of 2025 [2][3] Multimodal Development - The evolution of large models is moving towards deeper cross-modal understanding, transitioning from mere comprehension to cognitive processing [4][6] - Early multimodal architectures had limitations, but advancements like the Gemini model are integrating image and video information into pre-training processes, enhancing cross-modal modeling capabilities [6] - Effective training of multimodal models can lead to superior performance in pure language tasks compared to single language models [6] Embodied Intelligence - Embodied intelligence is viewed as one of the ultimate forms of AGI, with significant attention in 2025 [3] - The development of agents is crucial for the practical application of large model capabilities, but current agents still face challenges in complex real-world scenarios [7] - The reliability and success rate of agents in real-world applications are critical for their perceived value [7] Key Challenges - A major challenge for achieving AGI is the ability to generalize reasoning from narrow domains to complex real-life scenarios [8] - Current multimodal models exhibit insufficient spatial understanding, which is a significant barrier to the realization of embodied intelligence [8] - The data acquisition methods for embodied intelligence are limited, primarily relying on robotic operations, which results in lower data throughput compared to digital models [10]

Core Viewpoint - SenseTime Group Limited has engaged in multiple placements of new Class B shares, raising significant capital while experiencing a reduction in net losses for the fiscal year ending December 31, 2024 [1][2][3]. Group 1: Share Placement Details - On July 24, 2023, the company announced a conditional agreement to issue a total of 1,666,667,000 new Class B shares at a subscription price of HKD 1.50 per share, amounting to approximately HKD 2,500 million [1]. - The company completed a share placement on December 17, 2024, successfully placing 1,865,000,000 shares at the same price of HKD 1.50, representing about 5.40% of the existing Class B shares prior to the placement [2]. - A previous placement on June 27, 2024, involved 1,673,446,000 shares sold at HKD 1.20 each, accounting for approximately 6.45% of the existing Class B shares before the placement [3]. Group 2: Financial Performance - For the fiscal year ending December 31, 2024, SenseTime reported revenues of RMB 3,772.1 million, reflecting a year-on-year increase of 10.8% [3]. - The net loss for the year was RMB 4,306.6 million, which represents a 33.7% reduction compared to the previous year [3]. - The total net proceeds from the three share placements amounted to approximately HKD 7,282 million [3]. Group 3: Historical Financial Data - The company's revenue figures for the past five years are as follows: - 2020: RMB 3,446.2 million - 2021: RMB 4,700.3 million - 2022: RMB 3,808.5 million - 2023: RMB 3,405.8 million - 2024: RMB 3,772.1 million [4]. - The gross profit for 2024 was RMB 1,619.7 million, with a notable decline in losses from continuing operations over the years [4].

Core Insights - SenseTime launched the "Wuneng" embodied intelligence platform during the WAIC 2025, which aims to enhance the autonomy and intelligence of smart devices and robots through advanced perception, visual navigation, and multimodal interaction capabilities [1] Company Developments - The platform is built on SenseTime's embodied world model and leverages both edge and cloud computing power from its large-scale infrastructure [1] - SenseTime's co-founder and chief scientist, Lin Dahua, emphasized the importance of multimodality in achieving Artificial General Intelligence (AGI) and highlighted the company's extensive experience in computer vision and collaboration with hardware companies [1] Market Opportunities - The embodied intelligence market is rapidly growing, and SenseTime aims to capture commercial opportunities within this space, leveraging its multimodal capabilities and accumulated knowledge in world models [1] - SenseTime's investment arm, Guoxiang Capital, has invested in several companies within the embodied intelligence sector, including Galaxy General, Zhongqing Robotics, and Titanium Tiger Robotics [1] - Recent funding rounds in the sector include Galaxy General securing 1.1 billion yuan from CATL and Zhongqing Robotics completing a financing round close to 1 billion yuan [1]

Core Viewpoint - The article discusses a heated debate among industry leaders at the WAIC 2025 forum regarding the evolution of large model technologies, focusing on training paradigms, model architectures, and data sources, highlighting a significant shift from pre-training to reinforcement learning as a dominant approach in AI development [2][10][68]. Group 1: Training Paradigms - The forum highlighted a paradigm shift in AI from a pre-training dominant model to one that emphasizes reinforcement learning, marking a significant evolution in AI technology [10][19]. - OpenAI's transition from pre-training to reinforcement learning is seen as a critical development, with experts suggesting that the pre-training era is nearing its end [19][20]. - The balance between pre-training and reinforcement learning is a key topic, with experts discussing the importance of pre-training in establishing a strong foundation for reinforcement learning [25][26]. Group 2: Model Architectures - The dominance of the Transformer architecture in AI has been evident since 2017, but its limitations are becoming apparent as model parameters increase and context windows expand [31][32]. - There are two main exploration paths in model architecture: optimizing existing Transformer architectures and developing entirely new paradigms, such as Mamba and RetNet, which aim to improve efficiency and performance [33][34]. - The future of model architecture may involve a return to RNN structures as the industry shifts towards agent-based applications that require models to interact autonomously with their environments [38]. Group 3: Data Sources - The article discusses the looming challenge of high-quality data scarcity, predicting that by 2028, existing data reserves may be fully utilized, potentially stalling the development of large models [41][42]. - Synthetic data is being explored as a solution to data scarcity, with companies like Anthropic and OpenAI utilizing model-generated data to supplement training [43][44]. - Concerns about the reliability of synthetic data are raised, emphasizing the need for validation mechanisms to ensure the quality of training data [45][50]. Group 4: Open Source vs. Closed Source - The ongoing debate between open-source and closed-source models is highlighted, with open-source models like DeepSeek gaining traction and challenging the dominance of closed-source models [60][61]. - Open-source initiatives are seen as a way to promote resource allocation efficiency and drive industry evolution, even if they do not always produce the highest-performing models [63][64]. - The future may see a hybrid model combining open-source and closed-source approaches, addressing challenges such as model fragmentation and misuse [66][67].

Core Insights - The 2025 World Artificial Intelligence Conference (WAIC 2025) opened in Shanghai, showcasing SenseTime's upgraded "Jueying Kaiwu" world model, which aims to bridge AI and real-world interactions [1] - SenseTime Jueying introduced the industry's first mass-produced, interactive world model for the autonomous driving sector, along with the largest generative driving dataset "WorldSim-Drive" to empower the industry [1][2] - The company is collaborating with SAIC Group's Zhiji Auto to enhance data generation for various driving scenarios, aiming to accelerate the deployment of safe and reliable autonomous driving systems [4] Company Developments - SenseTime Jueying's CEO highlighted the transformation of AI creativity into productivity, generating millions of scene data for autonomous driving and creating a new 4D real world for embodied intelligence [3] - The "Jueying Kaiwu" world model is the first generative world model product platform in the autonomous driving field, designed to address data bottlenecks and is available for trial by B/C end users [4] - Currently, 20% of SenseTime Jueying's data is produced through the world model, showcasing its high production efficiency [4] Industry Impact - The integration of virtual and real data paradigms in autonomous driving will enhance embodied intelligence, focusing on the interaction between people, objects, and scenes [3] - The interactive experience at WAIC 2025 allowed attendees to engage with the generative world model product platform, demonstrating the performance of the leading autonomous driving dataset [7]

Summary of Key Points from the Conference Call Company and Industry Involved - **Company**: SenseTime Technology (商汤科技) - **Industry**: Artificial Intelligence (AI) and its applications across various sectors Core Insights and Arguments 1. **Advancements in AI Technology**: Chinese large model technology has shown outstanding performance in reasoning capabilities, open-source ecosystems, cost efficiency, and vertical applications, necessitating continuous technological breakthroughs and algorithm originality [2][3][41] 2. **Shanghai's AI Ecosystem**: Shanghai has established a parallel system of open-source and commercial large models, with 82 models registered nationally, positioning AI as a new growth engine for the city's economy [2][5] 3. **Sustainability Challenges**: The AI industry faces sustainability challenges, particularly regarding the energy consumption of data centers, which is projected to account for 8% of global electricity usage by 2030 [2][8] 4. **Economic Impact of AI Investment**: Investment in computing power and AI yields significant economic benefits, with a 1% increase in computing power index correlating to a 1.8‰ GDP growth [3][13] 5. **Policy Support for AI Development**: There is a call for enhanced policy support to create a favorable environment for AI development, including the use of intellectual property and fiscal policies [3][4] Other Important but Possibly Overlooked Content 1. **AI's Role in Reducing Carbon Emissions**: AI can significantly reduce carbon emissions in heavy industries and enhance energy efficiency in factories, with successful implementations already seen in Singapore and ASEAN [3][11] 2. **Challenges in AI Training**: The training of large models is energy-intensive, with the energy consumption during the reasoning phase increasing significantly with usage, potentially becoming a major source of energy consumption [8][9] 3. **Future Directions for AI Models**: The future of large model technology may involve expanding current paradigms to accept natural language feedback and developing autonomous online agents capable of self-learning [25][26] 4. **Open Source vs. Closed Source Dynamics**: The ongoing competition between open-source and closed-source models will shape the AI ecosystem, with open-source models driving efficiency and collaboration [37][39] 5. **SenseTime's Innovations**: SenseTime has made significant strides in AI, particularly with its SenseNova large model, which aims to unlock general AI task capabilities at low costs, facilitating widespread AI adoption across industries [41][59] This summary encapsulates the key points discussed during the conference call, highlighting the advancements, challenges, and future directions of AI technology, particularly in the context of SenseTime and the broader industry landscape.