Meta appears to be back in the AI game. Meta today launched Muse Spark, the first model out of Meta Superintelligence Labs ...

“ 很高兴分享我们在 Meta Superintelligence Labs 一直在做的事情。我们刚刚发布了 Muse Spark，这是我们的第一个 AI 模型，也是我们迈向个人超级智能的第一步。我们把整套技术栈都翻新了，以支持规模化，而这只是个开始 ...

Mark Collier briefed me on two updates under embargo at KubeCon Europe 2026 last month: Helion, which opens up GPU kernel ...

Anthropic’s Claude blackmail research found harmful behavior across major AI models under pressure, raising fresh concerns ...

这种真实感直接拉满了，也就是虚拟到现实的迁移成功率，因为在虚拟环境里见过的复杂情况越多，到了真实世界才不会变成个废物。

为此，斯坦福、MIT 等机构研究学者提出了自动化 harness 搜索框架 Meta-Harness 。它的核心思路是： 不再只优化单条 prompt，而是直接把完整 harness ...

Meta launches Muse Spark, a new multimodal AI model with advanced reasoning and efficiency gains, as it aims to compete with ...

Stop letting AI pick your passwords. They follow predictable patterns instead of being truly random, making them easy for ...

The designed test equipment can be used for internal R&D purposes but also for volume production at manufacturing sites.

Different impacts of the same molecular and circuit mechanisms on sleep–wakefulness control in early-life juveniles and adults.

AIPress.com.cn报道 4月5日消息，一次针对开源软件库的供应链攻击正在引发人工智能行业的安全担忧。meta已暂停与AI数据公司 Mercor 的合作项目，此前该公司在网络攻击中发生数据泄露，可能暴露了包括AI训练方法在内的敏感信息。 Mercor总部位于旧金山，是一家为多家AI公司提供训练数据的供应商，其客户包括 ...