近来看到个最好玩的消息，大致下面这样： 看到这个内容的时候，我是真的笑喷了。 这其实意味着大家花了很多时间做所谓的智能体，创造的全是负价值。 为啥会这样？ 原因特简单。 来自过去产研阵营的人类总是觉得，可以通过加入人类以为让AI更好的知识（表现为提示词，本质是规则）来提升AI在特定方向的表现。 这些所谓的"人类知识"和"小技巧"，表现出来就是提示词（Prompt），本质上则是一堆给AI的规则。 这些东西有助于提高限定目标下的精度，但对于一个大模型来说，实则是一种戕害，是一种"智能降级"的行为。 大模型的厉害之处在哪？ 在于它学了海量的数据，内部形成了一个模拟真实世界的、极其复杂的概率模型。它有种"涌现"出来的、我们都还没完全搞懂的通用智能。 你加进去的那些规则，就像是给一个想象力无限的画家，硬塞了一本儿童涂色书，还规定他必须在框框里涂色，不能出界。 你以为你在"优化"他画苹果的能力，实际上你废掉了他创作《星空》的可能。 当你面对的需求，敞口巨大、千奇百怪、无限贴近真实世界的时候——比如一个律师的日常工作——你那点"局部优化"就变得得不偿失。 你阉割掉的通用智能部分所带来的损失大于你费劲匹配上的那部分需求上的收 ...

以下文章来源于融中财经 ，作者吕敬之 融中财经 . 中国领先的股权投资与产业投资媒体平台。聚焦报道中国新经济发展和创新投资全产业链。通过全媒体 资讯平台、品牌活动、研究服务、专家咨询、投资顾问等业务，为政府、企业、投资机构提供一站式专 业服务。 本文来自微信公众号： 融中财经 （ID：thecapital） ，作者：吕敬之，题图来自：视觉中国 今年，投资人都在讨论一个问题：下一个"超级APP"会不会属于Agent。 在技术层面，Agent技术在2025年取得了显著进展。OpenAI、Cursor、Manus等公司通过强化学习微 调 （RFT） 和环境理解实现了技术突破，编程类Agent向通用型进化，垂类产品如Vantel、Gamma 展现出巨大潜力。这些技术进步不仅提升了Agent的性能和效率，还拓展了其应用场景，使其在更多 领域展现出巨大的潜力。 市场潜力和商业化前景也是Agent赛道备受关注的重要原因。2025年被视为Agent AI商业化的元年， 其应用场景不断拓展，从办公类Agent到垂直类Agent，再到更广泛的行业应用。未来式智能等企业 已经在电力、金融、泛互联网、制造业等行业实现了Agent的常 ...

Core Insights - The year 2025 is anticipated to be a pivotal year for the commercialization of AI Agents, with significant advancements in technology and expanding application scenarios [1][6][3] - The AI Agent sector has seen substantial investment activity, with over 66.5 billion RMB in funding in 2024, indicating strong market interest and potential [2][8] - Major companies like OpenAI and Cursor are leading technological breakthroughs in AI Agents, enhancing their performance and efficiency [5][1] Technology Advancements - Companies such as OpenAI, Cursor, and Manus have achieved significant breakthroughs in AI Agent technology through reinforcement learning fine-tuning and environmental understanding [1][5] - Specific applications like Sweet Spot and Gamma demonstrate the potential of AI Agents in various fields, enhancing user experience and operational efficiency [5][6] - The trend towards more intelligent and capable Agents is expected to continue, with a focus on personalized services and integration with other technologies [11][12] Market Potential - The AI Agent market is characterized by a broad range of application scenarios, from office-related Agents to vertical industry applications, indicating a strong commercial outlook [6][3] - Investment institutions are increasingly focusing on the landing capabilities of vertical scenarios and the commercial prospects of AI Agent projects [2][8] - The overall market for AI-related industries is expanding, driven by technological advancements and supportive national policies [7][8] Investment Trends - The investment landscape for AI Agents is heating up, with significant funding directed towards projects that demonstrate strong technological frameworks and market feedback [2][8] - Major funding rounds for leading projects, such as OpenAI's acquisition of Windsurf for $3 billion, highlight the attractiveness of the AI Agent sector [8][2] - The overall recovery of the primary market and the flow of capital towards AI applications are creating a favorable environment for investment in the Agent sector [8][7] Future Outlook - The AI Agent sector is expected to benefit from the release of large model technology dividends and favorable national policies, leading to historic development opportunities [3][6] - The integration of AI Agents into various industries, including finance, manufacturing, and energy, is already underway, showcasing their potential for widespread application [6][3] - The ongoing evolution of AI Agents is likely to lead to the emergence of the next "super app," as these technologies become more integrated into everyday workflows [15][17]

Core Insights - AI is expected to create trillion-dollar market opportunities, with all necessary elements in place for an imminent explosion in AI development [3][7] - The leap in AI capabilities, such as coding, indicates a shift towards a "bountiful era" where labor becomes cheap and abundant, while "taste" may become a new scarce asset [3][9] - The number of foundational large models will be limited, with companies investing more in reinforcement learning to enhance model capabilities [3][4] Group 1 - AI models may become more sparse and specialized, focusing on different areas of expertise and allowing for dynamic resource allocation [4][17] - Intelligent agents will possess improved working capabilities, including better memory and self-guidance, enabling longer autonomous operation [5][18] - User engagement with AI products may evolve into a new business model where personal background information is used for logging into multiple AI services [6][22] Group 2 - Innovation in the AI era is occurring at the blurred lines between model research and product development, advocating for a bottom-up exploration approach [4][21] - Organizations developing software products will face challenges from AI code generation, necessitating structural and operational changes [5][24] - Companies need to adopt a "stochastic mindset" to manage the uncertainties of AI, shifting from strict rule-driven approaches to dynamic adaptability [5][8] Group 3 - The competition in AI applications is expected to intensify, leading to the formation of an "agent economy" [6][9] - Startups should focus on solving complex problems that require human involvement, building data flywheels linked to specific business metrics [8][9] - AI's impact on the economy will be profound, reshaping companies and the overall economic landscape [8][9] Group 4 - OpenAI emphasizes maintaining organizational agility and aims to become a "core AI subscription" service [10][12] - The potential of models is believed to have a 10-100x growth space, with a focus on reinforcement learning to enhance model capabilities [10][11] - The vision includes creating an AI application ecosystem that provides powerful tools and services for developers and users [12][13] Group 5 - Google's approach focuses on hardware-software synergy to enhance model development, predicting significant advancements in AI capabilities within the next few years [14][15] - The future of models may involve mixed expert models to improve computational efficiency and continuous learning [17][18] - AI's transformative potential in scientific research is highlighted, with expectations for AI to replace traditional simulation methods [18][19] Group 6 - Anthropic advocates for a bottom-up approach in AI product development, emphasizing the importance of user needs over technical showcases [20][21] - The next generation of AI products will focus on autonomous agents capable of long-term operation and improved collaboration [22][23] - The rise of AI-generated content will necessitate new standards for content traceability and security [22][24]

Core Viewpoint - The article discusses the evolution of artificial intelligence (AI) from its inception to the current era of large models, highlighting key milestones, technological advancements, and the impact on various industries. Group 1: Birth of Artificial Intelligence (Mid-20th Century) - In 1950, Alan Turing proposed the "Turing Test," defining the philosophical goal of AI [3] - The term "Artificial Intelligence" was first used in 1956 at Dartmouth College, marking the transition from philosophical speculation to applied technology [3] - Early AI systems, like the IBM701, had limited computational power, executing only 16,000 operations per second, which is significantly less than modern devices [3] Group 2: Symbolism and Its Failures (1960-1970) - The 1960s saw the rise of "symbolism," where researchers attempted to simulate human reasoning through rule-based expert systems [4] - The MYCIN system developed in 1976 achieved near-expert accuracy in diagnosing blood infections, demonstrating the commercial value of expert systems [4][5] - The "Fifth Generation Computer Systems" project in Japan, launched in 1982 with an investment of $850 million, aimed to create intelligent computers but ultimately failed due to over-reliance on symbolic methods and hardware limitations [8] Group 3: Rise of Machine Learning (1990s-2000s) - The 1990s marked a shift to machine learning, moving from rule-based systems to data-driven approaches, allowing machines to learn from data rather than relying solely on hard-coded rules [10] - IBM's DeepBlue defeated a chess champion in 1997, showcasing the potential of machine learning in closed tasks [12] - The introduction of Google's PageRank algorithm in 1998 demonstrated the commercial value of data correlation, transforming search engines into profitable ventures [12] Group 4: Deep Learning Revolution (2010s-2020) - The 21st century saw the emergence of deep learning, enabling AI to automatically extract features through multi-layer neural networks [13] - AlphaGo's victory over a world champion in 2016 highlighted the capabilities of deep reinforcement learning [13] - The rapid increase in model parameters from 60,000 in LeNet-5 to 600 million in AlexNet illustrated the exponential growth in AI's capacity to handle complex tasks [14] Group 5: Era of Large Models (2021-Present) - The introduction of large pre-trained models like GPT-3 in 2020 has propelled AI towards general intelligence, showcasing advanced language understanding and generation capabilities [15] - Applications of generative AI have expanded across various fields, including content creation, programming assistance, and image generation, significantly enhancing productivity [16] - The competition between open-source and closed-source models has intensified, with companies like HuggingFace promoting open-source development while others like OpenAI focus on proprietary advancements [17] Group 6: Future Directions and Challenges - The future of AI is expected to focus on specialized models for high-value sectors such as healthcare and finance, emphasizing efficiency and cost-effectiveness [38] - The relationship between AI and human employees is anticipated to evolve into deeper integration, enhancing decision-making and innovation within organizations [38] - Ethical challenges and societal risks associated with AI, such as job displacement and privacy concerns, remain critical issues that need addressing [39]