Llama系列AI模型

Search documents
28岁华人执掌1.85万亿科技巨头AI大权,一觉醒来,图灵奖得主也要向他汇报
3 6 Ke· 2025-08-21 08:26
Core Insights - Meta's Superintelligence Lab has been restructured into four departments, with Alexandr Wang appointed as the head, emphasizing the imminent arrival of superintelligence [1][8][30] - The AGI Foundations team has been disbanded, and its members will be redistributed among the new departments, indicating a significant shift in Meta's AI strategy [29][31] Group 1: Leadership Changes - Alexandr Wang, at 28 years old, has been given unprecedented power within Meta, with major responsibilities and a substantial budget to drive AI initiatives [1][5] - Yann LeCun's role has been downgraded, as he will now report to Wang, reflecting a shift in power dynamics within the AI division [12][21] - Nat Friedman will lead the product team and report directly to Wang, indicating a tighter integration of AI research with product development [23][25] Group 2: Departmental Structure - The four new departments are TBD Lab (focused on exploratory research), FAIR (long-term fundamental AI research), a product and application team, and an infrastructure team [8][30][34] - TBD Lab will explore new models, including a potential "omni" model, while FAIR will serve as the innovation engine for the lab [14][17][34] - The infrastructure team will be crucial for supporting AI model training and deployment, led by Aparna Ramani [26][34] Group 3: Strategic Focus - The restructuring aims to accelerate the development of superintelligence by concentrating efforts on core areas of research, product, and infrastructure [30][36] - The decision to dissolve the AGI Foundations team reflects a strategic pivot in Meta's approach to AI, moving away from previous models that have not met expectations [29][31] - The emphasis on integrating research with product development suggests a shift towards more practical applications of AI technology within Meta's ecosystem [25][34]
清华学霸赵晟佳,担任Meta超级智能实验室首席科学家
Hua Er Jie Jian Wen· 2025-07-26 02:23
Meta宣布ChatGPT联合创建者赵晟佳,将担任超级智能实验室的首席科学家。 7月25日,Meta扎克伯格在社交媒体上表示,赵晟佳自Meta超级智能实验室成立之初就担任首席科学家职务,现在随着团队招募进展顺利,公司 正式确认其领导角色。 未来赵晟佳将直接与扎克伯格以及Meta新任首席AI官、前Scale AI首席执行官Alexandr Wang合作。对此,赵晟佳发文表示: 非常期待与团队一起构建人工智能,并使其赋能人类。 Alexandr Wang欢迎表示: 赵晟佳是一位杰出的科学家,他最近在研究中已经开创了一种新的扩展范式,他将引领我们团队的科研方向。 对于投资者而言,这一高调的人事任命再次表明Meta正不惜代价投入AI领域,以期在与OpenAI和谷歌的激烈竞争中占据主导地位。扎克伯格此前 已表示,未来几年将在AI计算基础设施上投入"数千亿美元"。 前OpenAI核心研究员正式掌舵 赵晟佳本科毕业于清华大学,并在斯坦福大学获得计算机科学博士学位。他在OpenAI期间参与了多个核心项目的开发。 据报道,除了联合创建ChatGPT外,他还参与构建了GPT-4、迷你模型、4.1和o3等产品,并曾领导OpenAI ...
微软CEO纳德拉:公司30%代码由AI编写
news flash· 2025-04-30 03:42
微软首席执行官萨蒂亚.纳德拉周二表示,目前该公司代码库中高达30%的代码由人工智能生成。据我 估算,我们代码仓库中约20%至30%的项目代码很可能完全由软件自动编写完成。纳德拉在与Meta首席 执行官马克.扎克伯格的公开对话中透露。两位科技巨头掌门人是在Meta于加州门洛帕克举办的首届 LlamaCon人工智能开发者大会上进行这番交流的。纳德拉补充说明,微软由AI生成的代码比例正在持 续攀升。当纳德拉询问扎克伯格相关数据时,Meta首席执行官表示虽无法脱口给出精确数字,但透露 公司正在研发能自主构建下一代Llama系列AI模型的人工智能系统。我们预测未来一年内...或许半数开 发工作将由AI而非人类完成,此后这个比例还将继续扩大。扎克伯格阐述道。 ...