Core Insights - Apple and Google have announced a multi-year partnership where Apple will pay Google approximately $1 billion annually for a customized version of the Gemini 3 Pro model, which has parameters of 1.2 trillion, eight times that of Apple's current cloud model, enhancing Siri's capabilities significantly [1] - The new Siri, powered by Gemini, is expected to launch in Spring 2026 with iOS 26.4, providing Apple with a 2-3 year breathing space while its own large model is still in development [2] - This collaboration allows Google to integrate its AI into over 2 billion Apple devices, further solidifying its presence in both major mobile operating systems globally, contributing to a market valuation exceeding $4 trillion [2] Summary by Sections - **Partnership Details**: Apple will utilize Google's Gemini model to enhance Siri, with a significant financial commitment of $1 billion per year for a customized version [1] - **Strategic Implications**: The partnership provides Apple with a temporary solution while it develops its own AI capabilities, and it allows Google to expand its AI reach across major platforms [2] - **Market Reactions**: Elon Musk criticized the partnership as a concentration of power, while OpenAI, a former collaborator, has become a background player in this scenario [2]

往期趣图 （点击下方图片可跳转阅读） ...

Core Insights - DeepSeek has introduced a new research paper titled "Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models," in collaboration with Peking University, focusing on enhancing large language models (LLMs) through conditional memory and a new module called Engram [1][3][4]. Group 1: Research Background and Problem Statement - Current large language models primarily utilize Mixture of Experts (MoE) for sparsity, but existing Transformer architectures lack native knowledge retrieval mechanisms, leading to inefficient simulation of retrieval behavior [3][9]. - DeepSeek proposes conditional memory as a complementary approach to MoE, introducing the Engram module to address the limitations of current models [4][9]. Group 2: Engram Module and Its Functionality - The Engram module modernizes classic n-gram embeddings, enabling knowledge retrieval with O(1) time complexity [9]. - Engram separates static knowledge storage from dynamic computation processes, enhancing the model's ability to perform complex reasoning by offloading the reconstruction burden from the model's shallow layers [11][13]. Group 3: Performance Improvements - Engram has been scaled to 27 billion parameters, showing significant performance improvements over pure MoE baseline models under equivalent parameter and FLOPs conditions [11]. - Notably, Engram enhances knowledge retrieval capabilities, with improvements in metrics such as MMLU (+3.4), CMMLU (+4.0), and general reasoning tasks like BBH (+5.0) and ARC-Challenge (+3.7) [11][38]. Group 4: System Efficiency and Scalability - Engram's deterministic addressing supports prefetching from host memory at runtime with minimal performance overhead, allowing for efficient memory management [12][19]. - The architecture allows for the decoupling of parameter storage from computational resources, facilitating linear scalability with the number of accelerators [21][22]. Group 5: Experimental Results - Four models were trained: Dense-4B, MoE-27B, Engram-27B, and Engram-40B, all using the same training data and processes [35][36]. - Sparse architectures (MoE-27B, Engram-27B/40B) significantly outperformed the dense model (Dense-4B) across various benchmarks, demonstrating superior scaling properties [38]. Group 6: Long Context Training - Engram architecture has shown significant advantages in long-context tasks by preserving valuable attention capacity for global context processing [41]. - Controlled experiments indicate that Engram outperforms MoE models in complex retrieval tasks, confirming its architectural superiority [46].

关于最近爆火的「死了么APP 」 ，昨天已经总结了一篇：《 热搜都爆了！"死了么"拟 100 万出售 10% 股份 》 我后来还看到一些相关消息，继续汇总： 1、早期成本不到 1500 元，现在的下载量比之前多了大概 300 倍 2、主要依靠用户购买 App 实现盈利，预计之后会涨到 14 元或者 15 元 3、开发者加了很多互联网大佬和头部机构，他还加了好多媒体，表示现在"承受着巨大的精神压力" 有网友吐槽表示"这有啥好抄的，别逗了"。更有博主在 16:48 发视频来打脸了，"自己团队早在 2023 年 10 月 就上架了同名的 APP，功能比目前这个还更完整，当时还上过微博和 B 站的热榜。" 面对质疑，@死了么APP 转发回应称， "假 大家谨防上当受骗，避免财产损失。真说灵感，也是广泛网友 在各大社区热议共同讨论出的，并非视频所言" 。 4、他目前考虑融资，金额大约是 50 万美元 "死了么"蹲大厂抄袭，随后被指抄袭 1 月 12 日 15:17，@死了么APP 官微还发帖，"也共同期待一下，哪个大厂会第一时间抄袭"。 不过后来它又把回应删掉了。 PS：有一说一，左边这个的"一键上门收尸"也是很搞的 ...

Core Insights - The article discusses Elon Musk's predictions and insights regarding AI, robotics, and energy, emphasizing the rapid advancements expected in these fields over the next few years [2][7][30]. Group 1: AI Predictions - Musk predicts that Artificial General Intelligence (AGI) will be achieved by 2026 and that by 2030, AI will surpass the total intelligence of all humans combined [8]. - He believes that current AI has two orders of magnitude of improvement potential, meaning existing hardware could run models that are 100 times smarter [8]. - Musk anticipates a tenfold performance increase in AI capabilities annually, supported by advancements in chip technology and computational power [9]. Group 2: AI Safety - Musk identifies three key traits for ensuring AI safety: truth, curiosity, and beauty [12]. - He argues that truth prevents AI from making irrational decisions, while curiosity ensures that AI values human existence [15]. - The perception of beauty is seen as essential for AI to create a positive future [15]. Group 3: Robotics Advancements - Musk predicts that within three years, Tesla's Optimus robots will surpass the best human surgeons in surgical capabilities, with a large-scale deployment expected [19]. - He explains that the rapid progress in robotics is due to exponential growth in AI software, chip capabilities, and mechanical flexibility [20]. - Musk updates his previous estimate, suggesting that the number of humanoid robots could exceed 10 billion by 2040, with a significant increase in availability within the next five years [20]. Group 4: Energy and Sustainability - Musk emphasizes the importance of solar energy, stating that humanity currently utilizes only about 1% of the solar energy available on Earth [24]. - He praises China's advancements in solar energy production, predicting that by 2026, China's electricity output will be three times that of the U.S. [26]. - Musk envisions a future where energy becomes the basis of currency, highlighting the potential of space-based data centers powered by solar energy [27]. Group 5: Economic and Social Implications - Musk predicts a future characterized by both high income for all and social unrest, with white-collar jobs being the first to be replaced by AI [32][33]. - He suggests that the transition to an AI-driven economy will be gradual, with a significant pressure to evolve as fully automated companies outperform traditional ones [34]. - Musk proposes that the solution to the transition could involve providing everyone with access to goods and services, leading to deflation as production outpaces monetary supply [38].

Core Viewpoint - The Indian government is pressuring Apple to provide the source code of iOS as part of a new security standard proposal, which is unlikely to be realized due to strong opposition from major tech companies [1][3]. Group 1: Government Proposal - The Indian government is preparing a new security standard plan that includes 83 items across various fields [3]. - The proposal requires tech companies to notify the government of major updates, retain security audit logs for 12 months, regularly prompt users to review permissions, and submit source code to government-designated labs for testing and vulnerability assessment [3]. Group 2: Industry Response - Major companies like Apple, Samsung, Google, and Xiaomi have collectively opposed the new regulations, arguing that source code is a core business secret and its disclosure could jeopardize technological security [4]. - The Indian industry organization MAIT has directly opposed the requirements, stating that they are impractical and has requested the government to withdraw the proposal [4]. - Concerns have been raised regarding the 12-month log storage and mandatory malware scanning, which could either consume phone storage or significantly drain battery life [4].

以下文章来源于伯乐在线 ，作者伯小乐 伯乐在线 . 伯乐在线分享IT互联网职场和精选干货文章（原域名已不再维护）。组织维护10万+star的开源技术资源 库，包括：Python, Java, C/C++, Go, JS, CSS, Node.js, PHP, .NET 等 谈到热门的自主学习技术，他透露 2025 年就有团队在尝试用实时数据训练，但效果一般。核心问题不是技术 不行，而是 "自主学习的瓶颈是想象力缺失" ，大家不知道它落地后该做什么具体事，得先明确比如盈利交易 系统、科学难题工具这类目标才行。 对于产业落地，姚顺雨建议大公司要靠自己的场景拿真实数据，别再依赖外部标注。同时强调，" 现在人与人 的差距，关键在会不会用 AI 工具 "，中国得好好普及这方面教育。他还预测，To B 领域的 Agent 技术会快速 增长，要是能实现全球企业部署，可能让 GDP 多涨 5%-10%。 最后他提到，中国 AI 团队有望全球领先，但得突破算力、To B 市场成熟度和创新文化这三个关键点。 （参考：网易新闻、 21 世经，本文经由 AI 大模型优化） 2026 年 1 月 10 日，前 OpenAI 核心研究员 ...

Core Viewpoint - The article argues that AI will not replace software developers, emphasizing that the future of software development remains in the hands of developers who can translate ambiguous human thoughts into precise computational logic [1][2]. Group 1: Historical Context - The prediction that "programmers will be replaced" has never come true throughout the history of computing, which spans over 43 years [3]. - The author has witnessed multiple technological revolutions, each heralded as the end of programmers, such as the rise of Visual Basic and low-code platforms [4][6]. - Historical cycles show that each wave of technology has led to an increase in the number of programs and programmers, exemplifying the "Jevons Paradox" with a market size of $1.5 trillion [9]. Group 2: Differences with Current Technology - The current wave of Large Language Models (LLMs) differs significantly from past technologies in scale and impact, with LLMs not reliably improving development speed or software reliability [10][11]. - Unlike previous technologies that provided stable and reliable solutions, LLMs often slow down development and create a dual loss situation unless real bottlenecks are addressed [11]. Group 3: Essence of Programming - The core challenge of programming has always been converting vague human ideas into logical and precise computational expressions, a difficulty that persists regardless of the programming tools used [12][17]. - The complexity of programming lies not in the syntax but in understanding what needs to be achieved, a challenge that remains unchanged over decades [17][18]. Group 4: Future Outlook - AI will not eliminate the need for programmers; instead, the demand for skilled developers will continue to grow, especially as companies realize the true costs and limitations of AI technologies [19][20]. - The future of software development will likely see AI playing a supportive role, assisting in tasks like prototype code generation, while the critical decision-making and understanding will still rely on human developers [19][20].

Group 1 - The core viewpoint of the article revolves around the discussion of the "Dead or Alive" app and its potential name change, with a strong public sentiment against changing the name due to its popularity [1][2] - The article highlights that a competing app named "Alive or Not" has emerged, capitalizing on the trend [2] - It is noted that the development of the "Alive or Not" app was completed in just 6 hours, with 4 hours of AI assistance and 2 hours of manual coding [3]

Core Viewpoint - Elon Musk announced plans to open-source the recommendation algorithm of X, which has sparked discussions among users regarding the implications for competition and transparency [1][3]. Group 1 - The announcement was made on January 11 at 3 AM Beijing time, indicating a significant move towards transparency in algorithmic processes [1]. - User reactions included both support for Musk's initiative and concerns about potential competitive disadvantages, as opening the algorithm could allow rivals to replicate it [1][3]. - Some users commented on the algorithm's simplicity, suggesting that it becomes apparent how it operates when discussing sensitive topics, such as Gaza and Israel [4].