大模型的竞争远远没有结束。 就在刚刚过去的4月，大模型再度经历了新一轮诸神之战。 先是有阿里在4月29日凌晨发布开源模型Qwen，并官宣登顶全球开源模型榜首； 后又有Meta在4月30日举办首届LlamaCon开发者大会，大会上不仅发布了 对标ChatGPT的Meta AI App，还面向开发者推出了Llama API预览版。 Meta的Llama 4系列模型，则是在4月6日抢先对外发布。 也是在首届 LlamaCon开发者大会上，Meta创始人扎克伯格请来了现任微软CEO萨提亚·纳德拉进行了一场半小时的现场对话。 在这场对话中，小扎摇身一变，成了主持人，对这位曾经挽救微软于水火，推动微软巨额投资了OpenAI，如今在推动微软进行又一次改革的传奇CEO进 行了一场精彩访谈。 纳德拉说，我们正处于一个可以构建"深度应用"的阶段； 纳德拉说，一些项目中，我们代码库里的代码可能有20%-30%是由AI编写。 扎克伯格则指出，到2026年，预计会有一半的应用开发工作会由AI完成； 扎克伯格还指出，未来每个工程师都会是技术领导，带领一个自己的智能体小队。 我们将这次两大科技巨头值得回味的对话内容整理如下： 01 "深度 ...

Core Viewpoint - Guangdong Province has introduced policies to support the development of open-source communities and ecosystems, with each receiving up to 8 million yuan in funding, aiming to establish itself as a global hub for artificial intelligence and robotics [1][2]. Group 1: Policy Measures - The policy emphasizes a shift from "single-item support" to "ecosystem construction," focusing on high-level platforms and ecosystem development to drive overall industry upgrades [1]. - Guangdong's AI core industry reached over 220 billion yuan last year, with more than 1,500 core enterprises, positioning it as a leader in the national AI and robotics sector [1]. Group 2: Ecosystem Development - The integration of open-source ecosystem construction into a collaborative system of "innovation chain, industry chain, capital chain, and talent chain" aims to transform Guangdong from a manufacturing province to a strong intelligent manufacturing province [2]. - The establishment of national-level data trading venues in Guangzhou and Shenzhen is intended to facilitate the open flow of data elements, providing high-quality data support for the open-source ecosystem [1]. Group 3: Challenges and Future Outlook - Challenges include avoiding the issue of open-source communities focusing more on construction than operation, and balancing data openness with security, privacy, and copyright protection [2]. - The cultivation of open-source communities and ecosystems is seen as both forward-looking and practically significant, with the potential to create a comprehensive innovation ecosystem in Guangdong that spans basic research, technology breakthroughs, commercialization, technology finance, and talent support [2].

Core Insights - Alibaba's new model Qwen 3 is emerging as a leading force in the Chinese open-source AI ecosystem, replacing previous models like Llama and Mistral [1] - The interview with industry representatives highlights the importance of model fine-tuning, the choice between open-source and closed-source models, and the challenges faced in large model entrepreneurship [1] Model Selection - The majority of the company's needs (over 90%) require fine-tuned models for local deployment, with specific tasks utilizing APIs from models like GPT and Qwen [3] - Commonly used model sizes include 7B, 32B, and 72B, with smaller models (0.5B, 1.5B) for privacy-sensitive applications [3] - Qwen is preferred due to its mature and stable ecosystem, including well-adapted inference frameworks and fine-tuning tools [4] Technical Considerations - Qwen's strong support for Chinese language and its relevant pre-training data make it suitable for the company's focus on emotional companionship and psychological applications [6][7] - The complete series of model sizes offered by Qwen allows for lower fine-tuning costs and easier testing across different model sizes [7] Challenges in Model Usage - In embodied intelligence, challenges include high inference costs and ecosystem compatibility, especially when considering local deployment for privacy [9][10] - Online business faces challenges in model capability and inference costs, particularly during peak usage times [12] Model Capability and Business Needs - Current models do not fully meet the company's needs for nuanced emotional understanding, necessitating post-training to align models with specific business requirements [13] - The goal is to maintain general capabilities while significantly enhancing core domain abilities, with an acceptable trade-off in general performance [13] Open-source Model Development - The expectation is for open-source models to catch up with top closed-source models, with a desire for more technical details to be shared by developers [14] - Qwen and Llama focus on community and general usability, while DeepSeek is more aggressive in exploring cutting-edge technologies [15][16] Entrepreneurial Insights - A significant oversight in AI entrepreneurship is the mismatch between models and product needs, emphasizing the importance of understanding user requirements [17] - The correct approach is to integrate AI as a backend capability rather than a front-end interface, ensuring deeper personalization in user interactions [19] Global Impact of Open-source Models - The rise of Chinese open-source models like Qwen and DeepSeek is accelerating a global technological evolution, providing a path for Chinese companies to innovate and collaborate internationally [20]

，官网发布了一则消息。 称OpenAI将继续保持慈善组织的身份，并将其营利性子公司转型为一家公益公司。负责监管OpenAI的 慈善机构将作为大股东控制该公司。 现在的营利实体转为公共利益公司PBC，非营利组织控制PBC。 2023年的人事大战历历在目。llya指责山姆奥特曼的急功近利，破坏了ai 技术平权的美感，放弃了安全 性。当时业内就认定 llya 肯定要走。事实证明他确实走了（2024年5月）。他新项目SSI 目前估值200亿 美元，嗯 Openai 估值目前3000亿美金、好吧，资金是充裕的。 周一Openai董事会主席Bret Taylor） 保守估计，目前代差在14个月以内。 还有更为开明的Claude 3.5 3.7。它处于全球ai 应用者完全友好的态度，在国内云 亚洲云api调用上 霸 屏。 所以Openai 的投资者人 怎么看？苹果坚决没投。在基模还不能分出雌雄的现在，大模型厂商暂时的领 先不足以形成它商业化收费的基础。 开源形成社区，更多投资者在上面开发延展估计才是当下之路。 Openai O1 O3 的定价比R1 豆包贵一倍多。 应用爆发的序幕拉开（最新一个季度，Ai 模型api 调用量 ...

Group 1 - The article discusses the transition of China into the digital economy era, emphasizing the role of open-source innovation as a new model that combines artificial intelligence and non-exclusive knowledge as production factors [1][2] - It highlights the need for theoretical, institutional, and talent preparation to support open-source practices, which are seen as essential for China's economic development [2][3] - The article identifies three theoretical challenges that China must overcome to advance its open-source economy: the lack of systematic theoretical interpretation of the differences between digital and industrial economies, the absence of a scientific framework for evaluating the innovative value of open-source practices, and the unclear integration path between traditional cultural concepts and open-source culture [2][3] Group 2 - Open-source innovation requires deep integration across three dimensions: the fusion of technical and economic logic to create new business models, the integration of organizational and cultural logic to reshape the innovation ecosystem, and the alignment of national strategy with global governance to redefine competitive landscapes [3][4] - The article emphasizes the importance of establishing a new institutional framework that adapts to digital civilization, including a layered governance system and dynamic incentive mechanisms [3][4] Group 3 - The article explores the integration of Chinese cultural and ideological advantages with open-source principles, suggesting that traditional concepts like "harmony and coexistence" resonate with the collaborative nature of open-source [5][6] - It discusses how the digital economy shifts the focus of production factors from land and capital to data and algorithms, leading to a new value creation logic based on contribution ethics [7][8] Group 4 - The article argues for strengthening international cooperation in open-source innovation, positioning it as a key element in the "re-globalization" of the economy, with China offering unique wisdom and solutions [9][10] - It outlines the need for open science to drive knowledge-sharing global innovation networks, advocating for the establishment of international open-source scientific alliances [11][12] Group 5 - The article presents the open-source model as a transformative force in China's national innovation system, emphasizing its role in integrating social innovation resources and lowering technological barriers [14][15] - It highlights the potential of open-source to reshape knowledge production and resource allocation mechanisms, ultimately enhancing national competitiveness [18][19]

Core Viewpoint - The automotive industry is undergoing a significant transformation towards "software-defined vehicles," with a strong emphasis on Over-The-Air (OTA) updates and increased computational power in vehicles [2][3]. Group 1: Software and Hardware Integration - Automotive manufacturers are increasingly investing in vehicle computing power, focusing on platform architecture to maintain technological leadership [2]. - The open-source operating system announced by Li Auto is expected to have a profound impact on the automotive supply chain, promoting cost-effective product updates and enhancing industry collaboration [2][3]. Group 2: AI Development in Automotive - AI is a central topic in the automotive sector, with two main development directions: intelligent driving technology and generative AI-driven interaction systems [4]. - The synergy between spatial intelligence from autonomous driving data and interactive intelligence from AI models is creating a dual-driven innovation ecosystem in the automotive industry [5]. Group 3: Regulatory and Safety Considerations - The rise of intelligent driving has led to increased scrutiny from regulatory bodies, emphasizing the need for clear standards and safety regulations in the automotive sector [5][6]. - Current efforts by European automakers to establish safety and cybersecurity standards are being mirrored by domestic manufacturers, who are also setting technical requirements for OTA and software capabilities [6]. Group 4: Competitive Landscape and Future Outlook - Domestic automakers are leading globally in the development of smart electric vehicles, demonstrating rapid iteration capabilities compared to international counterparts [7]. - The competitive landscape is expected to evolve, with intensified competition among suppliers of key materials and components, potentially reshaping the global supply chain and redefining pricing structures in the automotive market [7].

转自：InFOQ 作者｜冬梅、核子可乐 刚刚，作为广受欢迎的键值数据库背后的缔造者，Redis 公司已经将其同名产品重新拉回开源阵营。 不过此举仍未能将某些批评者完全满意。 时隔一年，Redis 重新开源 Redis 方面表示，自 Redis 8 开始将把 GNU Affero 通用公共许可证（AGPL）作为 Redis 的附加许 可选项，且正在转向由开源代码倡议（OSI）认定为开源的许可模式。也就是说，去年 3 月其转向服 务器端公共许可证（SSPL v1）的退出开源决策如今正被逐步撤销。 用户可以选择以下三种许可证选项之一来使用 Redis Open Source（自 Redis 8 版本开始）及其后续 版本：Redis 源代码可用许可证 v2 (RSAL v2)、服务器端公共许可证 v1 (SSPL v1) 和 GNU Affero 通用公共许可证 v3 (AGPL v3)。 上个月，Redis 公司 CEO Rowan Troolope 在接受采访时表示，"目前并没有任何迹象表明 SSPL 被 广泛认可为一种有效的开源许可证。我们原本希望大家会将 SSPl 视为一种良好的许可证方案，它符 合所有条件， ...

5 月 1 日，Redis 之父 Antirez 在个人博客宣布了一个好消息： Redis 再度开源。 Antirez 曾经退出 Redis 维护将近 5 年，他在 2024 年 12 月回归。 回归 5 个月后推动搞出一个大新闻。 再度开源的消息，迅速引发热议，登上 HackerNews 热门头条。 1、为什么说是"再度"开源呢？ Redis 最初以 BSD-3-Clause 许可发布，这是一种宽松的开源许可，允许开发者在商业软件中使用代码而不必支付费用或 公开修改后的代码。BSD 许可在开源社区中非常受欢迎，因为它促进了广泛的采用和创新。 转折点是在 2024 年 3 月 20 日。Redis 的 CEO 在官网发布公告，宣布从 Redis 7.4 起，所有未来版本都采用双重许可 RSALv2 和 SSPLv1。 要命的是，开源促进会（OSI）并不认可上面这个双重许可协议，所以 Redis 也就是失去「开源」名分。 为什么 Redis 当时更改协议？ 一句话：Redis 不想再被云厂商"白嫖"。 Redis 不是个例，之前 MongDB 和 Elastic 等开源公司做过类似的协议修改。 2、Redis ...

据悉，阿里云此次开源的千问3是国内首个"混合推理模型"，即将"快思考"与"慢思考"集成进同一个模型，对简单 需求可低算力"秒回"答案，对复杂问题可多步骤"深度思考"，这样能大大节省算力消耗。在阿里巴巴千问3开源 后，上下游供应链连夜进行适配和调用，NVIDIA、高通、联发科、AMD等多家头部芯片厂商已成功适配千问3。 其中，阿里云今年已持续开源了通义万相首尾帧生视频14B模型、首个端到端全模态大模型通义千问Qwen2.5- Omni-7B、阿里万相2.1模型、视觉理解模型Qwen2.5-VL等模型。对于阿里云的持续开源策略，知名数字经济学 者、工信部信息通信经济专家委员会委员盘和林向南都记者表示，阿里云的开源逻辑很简单，阿里云的模型是开 源、免费的，但阿里云的硬件比如算力、各类软件工具、容器是不免费的，开源模型要和自己的数据结合形成自 己的AI，阿里云以模型开源来切入AI应用并在别的方面实现盈利，开源对阿里整个云服务生态是有利的。 赶在五一假期前，国内大模型厂商接连开源。4月29日凌晨，阿里巴巴开源新一代通义千问模型Qwen3（简称千问 3），参数量仅为DeepSeek-R1的1/3，成本大幅下降，性能全面超 ...

在周三公布的最新财报中，Meat大幅上调了今年的资本支出预算，继续大手笔押注AI。 然而实际上，Meta正在AI领域面临重重困境：AI技术发布进度滞后、"开源"战略遭质疑、关键的Llama 4 Behemoth模型迟迟未能推出……投资者迫切想知道，Meta的未来在哪里？ LlamaCon大会"雷声大雨点小"：开发者失望，Meta追赶者角色难改 但在会上，Meta未能如期发布开发者最为期待的推理版模型Llama 4 Behemoth，这款被描述为"训练于2 万亿参数的最强大混合专家AI模型"的产品，原定数周前发布，但已被多次推迟。 Brownstone Research发布报告指出，Meta在会上"没有拿出足够的干货"，明显在AI领域出于落后地 位。 该行强调，备受期待的Llama 4 Behemoth模型未能如期发布，Meta此次发布的重点，似乎更像是试图在 消费者和开发者领域两手抓，但并未在任何一个领域取得突破性进展： "Meta的会议完全失败了。这种情绪是有道理的。" 相比之下，OpenAI、Anthropic、Google、xAI和Mistral等竞争对手早已推出了消费级聊天机器人应用和 企业API接口 ...