Meta计划7月发布超大语言模型Llama 3，参数规模或达1400亿
据科技媒体The Information援引知情人士报道，Meta计划于今年7月发布其超大语言模型Llama 3。该模型有望与OpenAI的GPT-4相媲美，后者以其强大的多模态能力和处理长文本以及图像输入的能力而闻名。
Meta的一位员工透露，公司尚未决定Llama 3是否将成为多模态模型，因为研究人员尚未开始对模型进行微调。不过，该员工表示，Llama 3的参数规模最高可能超过1400亿，而去年7月发布的Llama 2的参数规模最高为700亿。
值得注意的是，在推出Llama 3之前，Meta正在努力解决Llama 2中发现的一个问题，即无法处理有争议的问题。
Meta对Llama 3的发布寄予厚望，希望该模型能够在人工智能领域取得重大进展。该公司相信，Llama 3将能够执行各种任务，包括生成文本、翻译语言、回答问题和编写代码。
如果Llama 3能够达到预期，它将成为Meta人工智能战略的重要组成部分。该公司一直大力投资人工智能，并将其视为未来增长的关键驱动力。
Llama 3的发布可能会对人工智能领域产生重大影响。它将为研究人员和开发人员提供一个新的强大工具，并可能加速人工智能的进步。
英语如下：
**Headline: Meta Unveils Massive 1400 Billion-Parameter Modelto Rival GPT-4**
**Keywords: Large Language Model, Multi-Modal, Parameter Scaling**
**Body:**
Meta is planning to release acolossal language model, dubbed Llama 3, in July, with a potential parameter count of 1400 billion.
According to The Information, citing sources familiar with the matter, Meta’s upcoming Llama 3 model aims to compete with OpenAI’s GPT-4, renowned for its robust multimodal capabilitiesand prowess in handling extensive text and image inputs.
A Meta employee revealed that the company has yet to determine whether Llama 3 will be multimodal, as researchers have not initiated fine-tuning the model. However, the employee indicated that Llama 3’s parameter count could surpass 1400 billion, a significant upgrade from Llama 2’s 700 billion parameters released in July last year.
Notably, Meta is actively addressing an issue identified in Llama 2, which hindered its ability to handle controversial topics, prior to Llama 3’s release.
Meta has high hopes for Llama 3,anticipating that it will make significant strides in the field of AI. The company believes Llama 3 will be capable of executing diverse tasks, including text generation, language translation, question answering, and code writing.
If Llama 3 meets expectations, it will become a pivotal component of Meta’s AI strategy. The company has invested heavily in AI, viewing it as a key growth driver for the future.
The release of Llama 3 has the potential to significantly impact the AI landscape. It will provide researchers and developers with a new, powerful tool and could accelerate the advancement of AI.
【来源】https://wallstreetcn.com/articles/3709312
Views: 1