Core Viewpoint - Humanity is undergoing a paradigm revolution, particularly in the realm of artificial intelligence (AI), which is reshaping our understanding of intelligence and computation [5][7]. Group 1: Paradigm Shifts in AI - The article outlines five interconnected paradigm shifts that are influencing AI development: 1. Natural Computing: Recognizes computation as a natural phenomenon, which can drive innovations in computer science and AI [6]. 2. Neural Computing: Aims to reconstruct AI systems to mimic the brain's mechanisms, enhancing AI efficiency and unlocking its potential [6]. 3. Predictive Intelligence: Highlights that the essence of intelligence lies in evolving knowledge and statistical modeling of the future, suggesting that AI will continuously evolve like humans [10]. 4. General Intelligence: Suggests that AI capabilities are already comprehensive, capable of handling diverse cognitive tasks, indicating that "Artificial General Intelligence" (AGI) may already be here [10]. 5. Collective Intelligence: Emphasizes that intelligence is inherently social and can be enhanced through collaboration among multiple intelligent agents [10]. Group 2: Historical Context and Theoretical Foundations - The article discusses the historical context of computer science, tracing its roots back to the Turing machine and the early development of electronic computers like ENIAC, which laid the groundwork for modern computing [11][12]. - It also references John von Neumann's insights into the relationship between computation and biology, suggesting that life itself is fundamentally computational [14][17]. Group 3: Advances in AI and Machine Learning - The emergence of large language models (LLMs) has demonstrated that AI can achieve remarkable general intelligence through simple predictive tasks, challenging traditional views on intelligence [36][38]. - The article posits that LLMs can learn a wide variety of algorithms, surpassing the totality of algorithms discovered by computer scientists [36]. Group 4: Future Directions in AI - The future of AI is expected to involve a shift towards neural computing paradigms that may utilize new substrates such as photonic, biological, or quantum systems, moving away from traditional silicon-based architectures [34][35]. - The article suggests that AI models will evolve into self-constructing systems that learn dynamically from experience, rather than being static with fixed parameters [40].

Core Insights - The article discusses the exponential growth of scientific literature in the field of chemistry, which presents significant challenges for researchers in converting reported chemical reactions into actionable laboratory experiments [2][6] - A new AI system named MOSAIC has been developed by researchers at Yale University, which demonstrates a 71% success rate in synthesizing over 35 new compounds, showcasing its ability to discover novel chemical reaction methods not present in the training data [3][4] Group 1: Challenges in Chemistry Research - Researchers face unprecedented challenges due to the explosion of scientific literature, with hundreds of thousands of new chemical reactions reported annually, making it difficult to translate this knowledge into practical laboratory solutions [6] - Traditional methods rely heavily on expert experience and time-consuming manual searches, which are inefficient and not scalable [6] Group 2: Introduction of MOSAIC - MOSAIC (Multiple Optimized Specialists for AI-assisted Chemical Prediction) is introduced as a revolutionary solution that allows chemists to leverage collective knowledge from millions of reaction schemes [6][4] - The core concept of MOSAIC is "collective intelligence," where the system is trained with 2,498 specialized AI "experts" based on the Llama-3.1-8B-instruct architecture [9] Group 3: Performance and Comparison - MOSAIC demonstrated superior performance in key tests, with a significant correlation (R²=0.811) between predicted and actual yield rates, indicating its ability to capture yield patterns across different reaction types [12] - When aggregating predictions from multiple experts, the complete match rate for reagents increased to 43.0%, with a success rate of 94.8% for at least partial predictions [12] - MOSAIC outperformed larger general-purpose language models like ChatGPT-4o mini and Claude 3.5, despite having only 8 billion parameters [12] Group 4: Human-AI Collaboration - The relationship between MOSAIC and human chemists is characterized as one of enhancement rather than replacement, acting as a "compass" for modern chemical synthesis [14] - This collaboration allows human experts to focus on more creative tasks by significantly reducing the time required to identify suitable experimental conditions [14] Group 5: Future Outlook - The integration of AI systems like MOSAIC is expected to transform the paradigm of chemical research, making AI-assisted chemical synthesis a standard practice and accelerating drug development and functional material design [16] - The collaboration between AI and human chemists is anticipated to expand the frontiers of chemical science, addressing major challenges in energy, healthcare, and the environment [17]

Core Viewpoint - The GRA framework (Generator–Reviewer–Adjudicator) proposed by Shanghai AI Lab and Renmin University of China enables small models to collaboratively generate high-quality training data without the need for large-scale language model distillation [1][2][13]. Group 1: GRA Framework Overview - GRA operates on the principle of "multi-person collaboration" and "role division," simulating a peer review process to ensure data quality [7][12]. - The framework consists of three main roles: Generator, Reviewer, and Adjudicator, each contributing to the data generation and evaluation process [8][9][10]. Group 2: Experimental Results - GRA-generated data quality matches or exceeds that of single large language models across ten mainstream datasets, showing significant performance improvements [2][14]. - The GRA framework integrates five open-source small language models, demonstrating that collaboration among smaller models can yield competitive results against larger models [14][17]. Group 3: Performance Metrics - GRA-generated data improved training performance by an average of 6.18% on LLaMA-3.1 and 11.81% on Qwen-2.5 compared to original data [16]. - GRA's performance is only 0.59% lower than the Qwen-72B distilled version, while outperforming it by 8.83% when trained on Qwen-2.5 data [17]. Group 4: Advantages of GRA - GRA enhances data diversity and quality, filling gaps in the original seed data and providing a broader semantic coverage [18]. - The data quality is validated through a robust review process, with over 87.3% of samples receiving high consistency scores [19]. - GRA-generated data presents a higher task difficulty, increasing the effectiveness of training for small models [20].

Core Viewpoint - The article discusses the evolution and categorization of AI agent protocols, emphasizing the need for standardized communication to enhance collaboration and problem-solving capabilities among AI agents across various industries [1][9]. Summary by Sections AI Agent Protocols Overview - The report introduces a systematic two-dimensional classification framework for existing AI agent protocols, distinguishing between context-oriented protocols and inter-agent protocols, as well as general-purpose and domain-specific protocols [1]. Model Context Protocol (MCP) - MCP represents a centralized approach where a core "MCP travel client" agent coordinates all external services, leading to a star-shaped information flow. While it is simple and easy to control, it lacks flexibility and scalability, making it challenging to adapt to complex tasks [2][3]. Agent-to-Agent Protocol (A2A) - A2A promotes a distributed and collaborative model, allowing agents to communicate directly without a central coordinator. This flexibility supports dynamic responses to changing needs but may face challenges when crossing organizational boundaries [4][5]. Agent Network Protocol (ANP) - ANP standardizes cross-domain interactions, enabling agents from different organizations to collaborate effectively. It formalizes the request and response process, making it suitable for diverse and secure environments [6]. Agora Protocol - Agora focuses on translating user natural language requests into standardized protocols for execution by specialized agents. This three-stage process enhances adaptability and allows agents to concentrate on their core functions [7][8]. Future Trends in AI Agent Protocols - The development of AI agent protocols is expected to evolve towards more adaptive, privacy-focused, and modular systems. Short-term goals include establishing unified evaluation frameworks and enhancing privacy protection mechanisms [9][10]. - Mid-term trends may involve embedding protocol knowledge into large language models and developing layered protocol architectures to improve interoperability [11][12]. - Long-term aspirations include creating a collective intelligence infrastructure and specialized data networks to facilitate structured, intent-driven information exchange among agents [13][14][15]. Conclusion - The exploration of AI agent protocols indicates a clear trajectory towards a more intelligent, autonomous, and collaborative future, with significant implications for technology, society, and economic models [16][17].

论文作者包括来自上海交通大学的杨滢轩、柴化灿、宋源祎、齐思远、温睦宁、李宁、廖俊威、胡浩毅、林江浩、刘卫文、温颖、俞勇、张伟楠，以及 ANP 社区 发起人常高伟。 随着大语言模型 (LLM) 技术的迅猛发展，基于 LLM 的智能智能体在客户服务、内容创作、数据分析甚至医疗辅助等多个行业领域得到广泛应用。然而，不同智能 体系统间的碎片化通信标准已成为制约其进一步发展的瓶颈。上海交通大学团队与 ANP 社区合作推出了首个全面系统的 AI 智能体协议综述《A Survey of AI Agent Protocols》，为解决这一关键挑战提供了清晰的指导框架。 ArXiv 论文链接：https://arxiv.org/abs/2504.16736 Github 仓库地址：https://github.com/zoe-yyx/Awesome-AIAgent-Protocol 交互碎片化：阻碍智能智能体发展的关键瓶颈 正如早期互联网面临的通信标准分散问题，当前的智能智能体生态系统同样遭遇协议不统一的困境。研究团队指出，随着应用场景扩展和不同供应商、不同结构 的智能体涌现，智能体与实体之间的交互规则变得越来越复杂。这种协议 ...