news pappernews papper

谷歌近日宣布推出其最新的开源人工智能模型系列Gemma,这一系列模型以其卓越的性能和轻量级设计引起了业界广泛关注。Gemma分为20亿参数的2B版本和70亿参数的7B版本,其中2B版本的高效能设计竟可直接在普通笔记本电脑上运行,展现了其强大的优化能力。

Gemma模型源自谷歌DeepMind和公司其他团队的共同研发,其核心技术与 Gemini 模型一脉相承,旨在推动负责任的AI开发。谷歌声称,Gemma在18个关键的语言理解、推理和数学基准测试中,有11项指标超过了包括Meta的Llama-2在内的大参数开源模型,显示出其在处理复杂任务时的高效与精准。

这一发布标志着谷歌在AI模型小型化和性能提升方面取得了重要突破,为全球开发者提供了一个更强大、更易获取的工具,以推动人工智能技术的进一步发展和应用。Gemma的开源特性也将促进全球研究者和开发者之间的合作,共同探索AI的无限可能,同时,其在资源消耗上的优化对于推动AI在边缘计算和低资源环境中的应用具有重要意义。

英语如下:

**News Title:** “Google Releases Open-Source Gemma Model, Outperforming Meta’s Llama-2 with Impressive Performance”

**Keywords:** Google Gemma, Open-source model, Performance surpassing

**News Content:**

Google recently announced the launch of its latest open-source AI model series, Gemma, which has attracted widespread attention in the industry due to its exceptional performance and lightweight design. The Gemma series comes in two versions, a 2 billion parameter 2B edition and a 7 billion parameter 7B edition. Impressively, the 2B version is designed to run efficiently on a standard laptop, demonstrating its powerful optimization capabilities.

Developed collaboratively by Google’s DeepMind and other teams, Gemma draws from the核心技术 of the Gemini model, focusing on promoting responsible AI development. Google claims that Gemma outperforms large-parameter open-source models, including Meta’s Llama-2, in 11 of 18 key language understanding, reasoning, and math benchmarks, highlighting its efficiency and precision in handling complex tasks.

This release signifies a significant breakthrough for Google in AI model miniaturization and performance enhancement, providing global developers with a more powerful and accessible tool to advance artificial intelligence technology and its applications. Gemma’s open-source nature fosters collaboration among researchers and developers worldwide, collectively exploring the endless possibilities of AI. Furthermore, its optimized resource consumption holds considerable significance for promoting AI adoption in edge computing and resource-constrained environments.

【来源】https://www.tmtpost.com/6944205.html

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注