到底 PewDiePie 要停止使用怎样的 AI？又要开始使用怎样的 AI？ | 机器之心发布 | | --- | | 机器之心编辑部 | 11 月 3 日，全球知名游戏博主 PewDiePie 发布视频，展示其自建本地 AI 系统的全过程。该视频目前浏览量已经超过 300 万，视频标题则赫然写着双关梗 "STOP: Using AI Right now"。 他投入 2 万美元，组装了包含 10 块英伟达 GPU（含 8 块改装版 RTX 4090 和 2 块 RTX 4000 Ada）的本地系统，支持 700 亿至 2450 亿参数的大模型运行，全程无 需依赖云计算。 通过 vLLM 框架，他试验了 Qwen-235B 等模型，构建了具备搜索、记忆、RAG（检索增强生成）和音频输出功能的定制化 AI 界面。 要知道 PewDiePie 是海外游戏博主顶流中的顶流，Youtube 粉丝超 1.1 亿，Ins 粉丝 2 千万。他下场部署本地 AI 引起全网巨大轰动。 一夜之间，本地部署 AI 大模型成为讨论热度极高的话题。 这要先从本地 AI 的特性说起。 在传统 AI 订阅模式中，大模型 + 云服务 = 账号 ...

Core Insights - Elon Musk envisions future smart devices evolving into "AI edge nodes," retaining only communication modules and basic interaction functions, with core logic managed by cloud AI and local AI, leading to the obsolescence of operating systems and standalone app concepts [3] Group 1 - Recognizing the immense potential of AI, Musk has been active in the AI sector, founding xAI and launching the flagship generative AI model Grok, alongside securing $12 billion in funding for independent consumer applications [6] - xAI has also introduced an open-source encyclopedia called "Grokipedia," with plans to launch it into space on stable oxide media [6] - At Tesla, Musk has restructured the AI organization, integrating Dojo team resources to focus on Full Self-Driving (FSD) and Optimus robot technology development, and is building the Cortex supercomputing cluster for neural network training [6] - Musk has recruited LLM experts from Google DeepMind to support AI project development [6]

Core Insights - The article discusses the revolutionary impact of Wafer-on-Wafer (WoW) technology on edge AI devices, enabling local processing on devices like smartphones and smart glasses, thus enhancing AI accessibility and performance [1][2][12]. Group 1: WoW Technology Overview - WoW is a 3D packaging technology that stacks logic chips and memory chips together, significantly reducing the distance between them, which leads to faster bandwidth, lower power consumption, and smaller form factors [2][6]. - Traditional AI computations require high bandwidth and energy efficiency, which WoW addresses by enabling local AI processing without the need for cloud connectivity [2][6]. Group 2: Market Potential - Morgan Stanley predicts that the WoW market will grow from $1 million in 2025 to $6 billion by 2030, with a compound annual growth rate (CAGR) of 257% [1][8]. - The total addressable market (TAM) for WoW technology is expected to reach $622 million by 2027, marking a critical turning point for the technology [10][11]. Group 3: Applications and Benefits - WoW technology is anticipated to accelerate the adoption of edge AI devices, particularly in AI PCs, smartphones, and smart glasses, by providing the necessary computational power and efficiency [1][12]. - The technology allows for real-time applications such as voice translation on smartphones and object recognition in smart glasses, enhancing user experience [2][6]. Group 4: Supply Chain and Collaboration - The ecosystem for WoW technology is becoming more mature, with several companies in Greater China applying for patents and developing their solutions, indicating a collaborative effort in the supply chain [12][14]. - Companies like Aipu Semiconductor are already in mass production of WoW solutions for applications like AI smartphones and PCs [5][12]. Group 5: Competitive Landscape - WoW technology is seen as complementary to mobile High Bandwidth Memory (HBM), with different application scenarios, thus not directly competing with each other [14]. - The article highlights potential risks related to technological development and market demand, which could impact the adoption rate of WoW technology [14].