Meta 发言人在声明中表示： " 我们对开源 AI 的立场没有改变，计划继续发布领先的开源模型。历史 上我们并非公开所有研发成果，未来也预计将同时训练开源和闭源模型。 " 据知情人士透露， Meta 近几周还暂停了其旗舰大语言模型 Llama 4 最大版本的开发工作，这标志着 该模型遭遇了最新挫折。 图片来源： U nsplash 据知情人士透露， Meta 近几周来一直在讨论开发封闭式人工智能模型，这将标志着其当前专注于开 源或免费模式的战略转变。 知情人表示，包括 Meta 首席人工智能官 Alexandr Wang 在内的一些高管建议公司不应开源其最先 进的模型。 另一位人士称，其他高管则认为开源仍具优势，因 Meta 正试图追赶竞争对手。 这些讨论正值 Meta 对人工智能业务进行全面改革之际，此前该公司在今年早些时候遭遇挫折。这家 社交媒体巨头上月敲定了向 Scale AI 投资 143 亿美元的协议，并聘请了这家数据标注公司的 CEO Wang 。 Meta 还聘用了前 GitHub 首席执行官 Nat Friedman 和前 Safe Superintelligence 首席执行官 Danie ...

Group 1 - Meta is investing heavily in building multiple large-scale AI data centers globally, focusing on "Superintelligence" rather than just AGI, with the first center "Prometheus" expected to launch in 2026 and another "Hyperion" planned to expand to 5 gigawatts of computing power [2] - The company aims to integrate its AI business under the "Superintelligence Labs" to create a complete AI product chain, from AI chat assistants to AIGC advertising tools and smart glasses, to achieve a commercial closed loop [3] - Meta has increased its capital expenditure forecast for 2025 to between $64 billion and $72 billion, indicating a strategy of using advertising revenue to support AI model development for future gains [3] Group 2 - Meta is reportedly facing challenges such as the cooling development of the Llama 4 model and the departure of key executives, leading to considerations of shifting from open-source models to a more closed commercial model [3] - Analysts suggest that while AI has improved Meta's advertising revenue capabilities in the short term, the long-term competition is centered around who can first create a general AI engine [3] - The company has made significant investments in talent acquisition, including hiring key figures from Scale AI and GitHub, and invested $1.43 billion in Scale AI last year [2]