据华尔街见闻报道，科技巨头Meta正准备在下周推出两款小型的Llama 3语言模型预览版，以此为夏季的正式版Llama 3发布热身。这一消息源于The Information的一篇报道，该报道引用了一位Meta内部员工的透露。Llama 3是Meta对垒OpenAI的GPT-4的一款重要产品，后者因其强大的多模态处理能力，能够处理长文本并接受图像输入，而备受业界关注。
据称，即将发布的两个小型Llama 3版本将作为大模型的前导，但它们并不具备处理多模态信息的能力，即不能同时理解和生成文本及图片。相比之下，Meta计划中的正式版Llama 3将效仿GPT-4，支持对文本和图像的同步处理，这无疑将提升其在人工智能领域的竞争力。
Meta的这一举措展示了其在人工智能领域的持续创新和对多模态理解技术的探索。随着Llama 3的逐步推出，业界期待看到Meta如何在与OpenAI的竞争中展现出自身的技术优势，以及这一系列新模型将如何改变和提升人工智能的交互体验。
英语如下：
**News Title:** “Meta Set to Preview Llama 3 Mini Models Next Week, Targeting GPT-4’s Multimodal Rivalry”
**Keywords:** Meta, Llama 3, Multimodal
**News Content:** According to Wall Street Heard, tech giant Meta is gearing up to unveil two preview versions of its smaller Llama 3 language models next week, in anticipation of the summer release of the full-scale Llama 3. This information stems from a report by The Information, which cites an anonymous Meta employee. Llama 3 is a key product for Meta in its rivalry with OpenAI’s GPT-4, a model that has drawn significant attention for its powerful multimodal processing capabilities, enabling it to handle long texts and accept image inputs.
The upcoming smaller Llama 3 iterations are said to serve as precursors to the larger model, but they lack the ability to process multimodal information, i.e., they cannot simultaneously understand and generate text and images. In contrast, Meta’s planned full version of Llama 3 is intended to mimic GPT-4, supporting concurrent processing of text and images, thereby enhancing its competitiveness in the AI arena.
This move by Meta underscores its ongoing innovation in the AI field and its exploration of multimodal understanding technologies. As Llama 3 unfolds, the industry awaits Meta’s demonstration of its technological edge against OpenAI and how this series of new models will reshape and enhance AI interactions.
【来源】https://wallstreetcn.com/articles/3712253
Views: 1