news studionews studio

国际科技巨头苹果公司近日在人工智能领域取得重大突破,其研究团队正式公布了一款名为MM1的多模态大模型。该模型参数最高可达惊人的300亿,是目前业界最大规模的多模态模型之一。MM1采用了密集模型和混合专家(MoE)变体技术,不仅在预训练指标中实现了最优性能,而且在多模态基准测试中表现出色,具有较强的竞争优势。苹果在论文《MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training》中详细介绍了MM1的构建方法、分析过程及其在多模态学习中的洞察力。这一成果标志着苹果在人工智能和机器学习领域的进一步深入,预计将对未来的智能设备和服务产生深远影响。

Title: Apple Unveils 300 Billion Parameter Multimodal Model MM1
Keywords: Apple, Multimodal Model, MM1
News content:
Global tech giant Apple has recently made a significant breakthrough in the field of artificial intelligence, with its research team officially announcing a multimodal model called MM1. The model boasts an impressive 300 billion parameters, making it one of the largest multimodal models in the industry. MM1 employs dense models and a variant of the Mixture of Experts (MoE) technology, achieving optimal performance in pre-training metrics and showing strong competitive performance in multimodal benchmarks. Apple has detailed the construction methods, analysis processes, and insights into multimodal learning in a paper titled “MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training.” This achievement signifies Apple’s further advancement in the fields of AI and machine learning and is expected to have a profound impact on future smart devices and services.

【来源】https://mp.weixin.qq.com/s/i9bx6M32uk4Jq2KSRhv4ng

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注