以下文章来源于新浪科技 ，作者努力码稿的小浪 新浪科技 . 一手掌握，科技"鲜"闻 来源丨新浪科技 （ID：techsina） 作者丨 郑峻 图源丨Midjourney Meta AI业务大地震！新主管上任三个月后，挥起裁员大刀，基础研究部门遭受重创，连明星大牛研 究员都不幸失业。扎克伯格这是急功近利，自毁长城；还是在精简机构，重振业务？ 前Meta AI部门员工向《硅谷观察》讲述了他开发Llama大模型的亲身经历，揭开了Meta AI部门大裁 员的背后故事。 在他看来，正是开源大模型Llama 4明显落后DeepSeek等中国竞争对手，给扎克伯格带来了强烈的 危机感，促使他下定决心、不惜血本从外部挖来AI负责人与精英人才，对Meta AI业务与战略进行彻 底重组。 Meta裁员AI部门 AI行业的竞争已经进入白热化，人才成为各家竞相争夺的最紧俏资产。谷歌和微软等诸多科技巨头为 了集中资源，投入AI研发竞争，在继续扩张AI研发部门的同时，甚至不惜对其他非核心部门进行大裁 员。 然而，Meta一边在求贤若渴，开出天价薪酬挖角竞争对手，一边却将裁员大斧砍向自家的AI部门，开 掉了高达600名员工，甚至包括了业界知 ...

AI播客：换个方式听新闻 下载mp3 音频由扣子空间生成 本周将是美股市场的关键时刻，很可能将决定今年剩余时间里美股的走向。 本周三和周四，占标普500指数约四分之一权重的五家公司——微软（MSFT）、Alphabet 管理着50亿美元资产的Wealth Consulting Group首席市场策略师Talley Leger表示，本周"可能决定这轮涨 势是会继续，还是会暂停。" 到目前为止，美股财报季开局强劲。目前，标普500指数中已有超过四分之一的公司公布了业绩，其中 约85%的公司财报超出了华尔街的预期，这是四年来最好的表现。 这一表现有助于缓解人们对贸易紧张局势升级和银行体系信贷风险的担忧。这也推动了标普500指数的 反弹，在10月初经历近五个月来最糟糕的一周抛售后，该基准指数正重返纪录高位。所谓的"科技七巨 头"贡献了标普500指数今年15%涨幅的近一半。 然而，为了维持涨势，投资者们正从这些科技巨头那里寻求保证，即用于计算基础设施的数百亿美元资 金投入将继续，并最终在未来得到回报。 预计微软、Alphabet、亚马逊和Meta在本财年的资本支出总额将达到3600亿美元，其中大部分与人工智 能相关。根 ...

Phiwath Jittamas/iStock via Getty Images By Ivan Castano Bullishness around AI and tech stocks, coupled with expectations for falling rates, could boost U.S. capital markets in the next 12 months, strategists say. Magnificent 7 stocks remain popular and earnings are strong – factors that have already seen the Nasdaq and S&P 500 gain around 15% and 12%, respectively, as of mid-October. Barring an unexpected shift in monetary policy, deteriorating economic conditions and/or another unforeseen event, the r ...

市场目光聚焦即将到来的科技巨头财报周与英伟达GTC大会，投资者将密切审视云业务增长、AI资本开支及广告收入等 关键指标。在此背景下，高盛交易员已表达乐观预期，认为若业绩不俗，科技股有望再次引领大盘。 硬·AI 作者 | 龙 玥 编辑 | 硬 AI 本周将迎来美国科技业"超级周"，市场的目光高度聚焦于此。 "科技七巨头"中微软、谷歌、Meta、苹果、亚马逊将密集发布财报，人工智能芯片巨头英伟达将举办GTC 大会，首席执行官黄仁勋的演讲备受期待。 市场情绪正在提前预热。高盛交易员John Flood表示，当前围绕大型科技股财报的情绪和仓位是他"相当 长一段时间以来见过的最友好的"。他预计，如果财报没有出现任何失误，市场将为"超级资本技术引领的 指数水平的再次上涨做好准备"。 01 财报季焦点： 云增长与AI开支 对于市场而言，本轮财报季有几个关键的观察点。 首先是云业务的增长情况。上一季度，谷歌云和微软Azure的增长均超过30%，呈现加速态势，而行业领 导者亚马逊AWS的增速则以18%明显落后。据彭博报道，亚马逊股东正密切关注AWS能否在本季度提 速。同时，谷歌云近期与Meta、OpenAI及Anthropic ...

莫斯科维茨今年早些时候宣布辞去 Asana CEO一职，转任董事长。近日他在节目中表示，"我不喜欢管理团队，这也不是我们创办 Asana 时的本意。" "我本来想担任工程主管之类的职位。后来，事情一步步发展下来，我当了13年的首席执行官，感觉很累。"莫斯科维茨补充道。（思瀚） 【环球网科技综合报道】10月27日消息，据CNBC报道，Meta 联合创始人达斯汀·莫斯科维茨表示，担任 13 年的CEO"令人筋疲力尽"。 莫斯科维茨是 Meta（前身为 Facebook）的创始人之一，他于 2004 年与马克·扎克伯格、爱德华多·萨维林、克里斯·休斯和安德鲁·麦科勒姆共同创立了该社交 平台。 2008年离开Facebook后，他保留了略高于8%的股份，根据《福布斯》最新数据，他的净资产达到120亿美元。同年，他成为工作管理软件平台Asana的联合 创始人兼CEO。 ...

Core Insights - The current capital expenditure (capex) surge among major tech companies is driven by the need to expand AI infrastructure, with Microsoft projecting over $30 billion in capex this quarter and Alphabet increasing its 2025 spending plan to around $85 billion [1][2] Group 1: Market Impact - Major tech companies, including Microsoft, Apple, Amazon, Alphabet, and Meta, collectively account for approximately 45% of the Nasdaq 100's value and a quarter of the S&P 500's, indicating their significant influence on U.S. equity markets [2] - Analysts view the current spending as a foundational shift towards AI, predicting nearly $3 trillion in AI-driven enterprise and government spending over the next three years, likening it to a structural boom rather than a speculative bubble [4][8] Group 2: Company-Specific Developments - Meta plans to spend up to $72 billion next year on data-center retrofits and AI infrastructure, while Amazon's AWS investment continues to grow alongside its advertising business, which generates $15 billion in quarterly profits [7] - Microsoft is expected to report strong Azure performance, with a focus on how much new capacity is currently being utilized versus what is pending due to power grid limitations [13] - Alphabet acknowledges that customer demand exceeds supply, leading to accelerated buildouts that may compress margins until new infrastructure is operational [9] Group 3: Infrastructure Challenges - Super Micro Computer's revenue forecast was cut from $7 billion to $5 billion due to delayed shipments, highlighting logistical challenges in the AI supply chain [3] - The infrastructure cycle is facing delays due to transformer shortages and utility connection issues, with median data-center projects taking longer than expected to become operational [11] - Companies are experiencing "temporary capacity constraints," which may indicate that they have expanded faster than the infrastructure can support [10][12] Group 4: Earnings Expectations - This week's earnings reports will serve as a critical test for these companies, with a focus on their ability to convert heavy investments into revenue [19] - Analysts are looking for specific updates on capacity timing and the impact of AI on advertising yields, as well as consumer spending trends from Apple [18][20] - The overall expectation is for solid revenue growth, but with squeezed margins, as companies navigate the balance between investment and immediate returns [17][21]

Market Analysis by covering: S&P 500, Microsoft Corporation, Alphabet Inc Class A, Apple Inc. Read 's Market Analysis on Investing.com ...

Market Overview - Asian stocks surged as easing trade tensions between China and the US improved risk appetite, with significant gains in South Korea's KOSPI, Taiwan stocks, and Japan's Nikkei, each rising over 2% [1][2] - MSCI's broadest index of Asia-Pacific shares increased by 1.3% to a record peak [2] Trade Deal Implications - A potential trade deal would pause American tariffs and Chinese rare earths export controls, alleviating investor concerns [2] - Investors are looking for confirmation that the trade truce holds and that China's stimulus measures lead to tangible growth [3] Currency and Commodity Movements - The Australian dollar rose by 0.42% to $0.6541, reflecting its status as a risk and China proxy [4] - Safe-haven gold prices fell by 1%, while US Treasuries declined, resulting in a 3.8 basis points increase in the benchmark 10-year bond yield [5] Central Bank Meetings - Central bank meetings in Japan, Canada, Europe, and the US are anticipated, with the Federal Reserve expected to cut interest rates by 25 basis points [6] - The European Central Bank and the Bank of Japan are expected to maintain current rates, with the BOJ considering future rate hikes as tariff-induced recession fears ease [7] Earnings Season Focus - The US earnings season is highlighted by reports from major companies like Microsoft, Apple, Alphabet, Amazon, and Meta Platforms, which are expected to show stronger results despite narrowing profit margins compared to the broader index [8] - Enthusiasm for artificial intelligence, a key area for many megacap companies, is driving stock market performance [9]

Group 1: Company Performance and Financials - Wentech's semiconductor revenue from the Chinese market accounts for nearly half of its total revenue, indicating its role as a growth core [2][4] - Wentech reported Q3 revenue of 4.427 billion yuan, a year-on-year decline of 77.38%, while net profit was 1.04 billion yuan, a year-on-year increase of 279.29% [4] - For the first three quarters, Wentech's revenue was 29.769 billion yuan, down 44.00%, with net profit at 1.513 billion yuan, up 265.09% [4] - Intel reported revenue of $13.7 billion, a 3% year-on-year increase, exceeding Wall Street's expectation of $13.2 billion [16] Group 2: AI and Technology Developments - Tencent's R&D report reveals that over 90% of its engineers use AI programming assistant CodeBuddy, with 50% of new code generated by AI, enhancing overall R&D efficiency by over 20% [5] - Apple has begun shipping advanced AI servers from its new factory in Houston, Texas, aimed at supporting its latest AI functionalities [12] - OpenAI is actively developing AI music models in collaboration with a U.S. music academy, focusing on high-quality data for training [11] Group 3: Investments and Financing - SoftBank's board approved a $22.5 billion investment in OpenAI, which will support its R&D and commercialization efforts [10] - JBD completed over 1 billion yuan in B2 round financing, setting a record for single financing in the MicroLED micro-display sector [7] Group 4: Regulatory and Compliance Issues - The European Commission has preliminarily found that Meta violated the Digital Services Act by failing to provide effective reporting mechanisms for illegal content [13]

Core Insights - The article discusses the recent advancements in large language models (LLMs) regarding their ability to achieve continual learning and self-evolution, addressing criticisms about their lack of genuine learning capabilities [1][2]. Group 1: Paths to Continual Learning - The ability of LLMs to learn continuously is fundamentally linked to their memory depth and plasticity, with three main paths identified for enhancing this capability [2]. - The first path involves modifying the "context" or "working memory" of the model through In-Context Learning (ICL), where new information is provided in prompts to help the model learn to solve specific problems [4][6]. - The second path introduces an "external memory bank" (RAG), allowing models to access and maintain an external database for comparison and retrieval, exemplified by Google's DeepMind's "Reasoningbank" [7]. - The third path focuses on parameter-level continual learning, which has faced challenges due to the complexities and instabilities associated with methods like Reinforcement Learning (RL) and Low-Rank Adaptation (LoRA) [10][11]. Group 2: Sparse Memory Fine-Tuning - Meta AI's recent paper introduces Sparse Memory Fine-Tuning (SFT) as a solution to the challenges of traditional SFT, particularly addressing the issue of catastrophic forgetting [11][28]. - The proposed method involves a three-step process: modifying the architecture to include a memory layer, using TF-IDF to identify which parameters to update, and performing sparse updates to only the most relevant parameters [12][22][23]. - This new approach has shown significant improvements, with models experiencing only an 11% drop in performance on original tasks after learning new facts, compared to 71% and 89% drops with LoRA and full fine-tuning, respectively [23][25]. Group 3: Implications for the Future of LLMs - The advancements in SFT suggest a potential shift in how models can be updated safely and effectively, moving away from static tools to dynamic agents capable of continuous learning [31][32]. - The successful implementation of these methods could mark the beginning of a new era for self-evolving models, aligning with the vision of models that grow and adapt through experience [31][32].