谷歌（Google）近日宣布推出一款名为Gemma的开源大模型，该模型在性能上超越了Meta Llama-2等竞品，标志着谷歌在人工智能领域的技术实力再次得到提升。
Gemma模型分为2B（20亿参数）和7B（70亿）两种尺寸版本，其中2B版本甚至在笔记本电脑上也能运行，这大大降低了用户的使用门槛。谷歌表示，Gemma采用与Gemini模型相同的研究和技术，由DeepMind和谷歌其他团队共同开发，旨在推动负责任的人工智能发展。
在18个语言理解、推理、数学等关键基准测试中，Gemma有11个测试分数超过了Meta Llama-2等更大参数的开源模型，显示了其强大的性能和广泛的适用性。
谷歌此次发布的Gemma模型，不仅为研究人员和开发者提供了强大的工具，也为人工智能的发展注入了新的活力。随着Gemma模型在更多领域的应用，我们有理由相信，它将推动人工智能技术的进一步进步和创新。
英文翻译内容：
Title: Google Unveils Gemma, an Open-Source Language Model That Outperforms Meta’s Llama-2
Keywords: Google, Gemma, Language Model
News content:
Google has recently announced the launch of an open-source language model named Gemma, which surpasses the performance of competitors like Meta’s Llama-2, marking a significant advancement in Google’s AI capabilities.
Gemma is available in two sizes: 2B (2 billion parameters) and 7B (7 billion), with the 2B variant being lightweight enough to run on a laptop, significantly lowering the barrier to entry for users. Google states that Gemma employs the same research and technology as the Gemini model, developed collaboratively by DeepMind and other Google teams, with the goal of fostering responsible AI development.
In key benchmarks for language understanding, reasoning, and mathematics, Gemma has outperformed larger open-source models like Llama-2 in 11 out of 18 tests, showcasing its robust capabilities and broad applicability.
The release of Gemma by Google not only provides researchers and developers with a powerful new tool, but also injects new vitality into the field of AI. As Gemma is applied in more areas, we have good reason to believe that it will drive further progress and innovation in AI technology.
【来源】https://www.tmtpost.com/6944205.html
Views: 1