【AI资讯】7月30日Meta发布了Segment Anything Model 2（SAM 2），用于实时图像和视频对象分割，并开源了相关数据库。生成式AI技术在社交媒体推荐系统中的应用前景被探讨，Meta和英伟达的创始人讨论了未来的发展趋势。生数科技的Vidu视频生成产品以30秒生成4秒视频的速度全球上线，成为业界最快。苹果的自研基础模型（AFM）在多项任务上表现超过GPT-4，具备强大的对话功能和应用。
【AI资讯】7月25日英特尔举办2024网络与边缘计算行业大会，推动边缘AI创新发展；法国AI初创公司Mistral AI发布了开源模型Mistral Large 2，具有1230亿个参数；OpenAI设计了基于规则的奖励机制（RBR）来教导AI模型遵守安全政策；字节跳动研究团队推出了端到端同声传译智能体CLASI，达到或超过人类同传水平。
【AI资讯】7月24日Llama 3.1, an open-source model, has surpassed closed-source models GPT-4o and Claude 3.5 Sonnet with its 405B performance. Its release is considered a significant milestone in the development of open-source AI and praised as a leader in the open AI ecosystem. MultiTrust, a comprehensive benchmark led by Tsinghua University, evaluates the credibility of multimodal large language models across five dimensions. Kuaishou's video generation model, Ke Ling AI, has announced another upgrade in its base model and opened up internal testing, improving image quality and motion performance.
