90年代的黄河路

据华尔街见闻报道,科技巨头Meta正筹备在下周推出两款小型的Llama 3预览版语言模型,以此为即将在夏季发布的全面升级版Llama 3铺平道路。这一举措表明Meta在人工智能领域的竞争态势日益升温,旨在与OpenAI的GPT-4一较高下。

援引一位内部员工的消息,The Information科技媒体透露,Meta计划的这两个小型Llama 3版本将不具备正式版的多模态处理能力。即将发布的Llama 3大模型预计将打破传统,不仅能够处理更长的文本序列,还将引入图像输入功能,实现对文本和图片的同步理解和生成,这与OpenAI的GPT-4相似,后者因其强大的多模态处理能力而备受瞩目。

Meta的这一战略旨在逐步释放新模型的潜力,通过小版本的先行测试收集反馈,以优化和完善最终的正式版Llama 3。此举也将为Meta在人工智能领域的技术领先地位提供有力的展示,进一步巩固其在科技行业的影响力。随着Llama 3的逐步推出,业界对于其性能和应用前景的期待正日益升温。

英语如下:

News Title: “Meta Teases Upcoming Preview of Mini Llama 3 Models, Targeting Multimodal Competition with GPT-4”

Keywords: Meta, Llama 3, Multimodal

News Content: According to Wall Street Journal, tech giant Meta is准备 next week to unveil two smaller preview versions of the Llama 3 language model, paving the way for the summer release of a fully upgraded Llama 3. This move signals an escalating competitive stance in the AI arena for Meta, positioning it to challenge OpenAI’s GPT-4.

Citing an internal source, tech news outlet The Information divulges that Meta’s planned mini Llama 3 iterations will lack the full multimodal processing capabilities of the official version. The upcoming large Llama 3 model is expected to break new ground by handling longer text sequences and introducing image input capabilities, enabling simultaneous understanding and generation of both text and images – a feature akin to GPT-4, which has garnered significant attention for its powerful multimodal processing abilities.

Meta’s strategy aims to incrementally unlock the potential of the new model, gathering feedback through early tests of smaller versions to optimize and refine the final, official Llama 3. This move will also serve as a strong demonstration of Meta’s technological leadership in the AI domain, further solidifying its influence in the tech industry. As Llama 3’s rollout progresses, anticipation for its performance and application prospects is escalating.

【来源】https://wallstreetcn.com/articles/3712253

Views: 2

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注