据华尔街见闻报道，科技巨头Meta计划在即将到来的7月份推出其最新的人工智能模型——Llama 3。据知情人士透露，Meta寄希望于这个新模型能够与OpenAI的知名大模型GPT-4相抗衡。GPT-4目前在多模态处理上表现出色，能处理更复杂的文本和图像任务。然而，关于Llama 3是否也将具备多模态能力，Meta内部尚未达成一致。据一位Meta员工表示，由于目前研发团队尚未开始对Llama 3进行微调，这一关键特性仍待确定。
在参数规模上，Llama 3预计将显著超越其前代产品。据悉，Llama 3可能包含超过1400亿个参数，远超Llama 2去年7月发布时的700亿参数上限。然而，Meta在推进Llama 3项目的同时，也在努力解决Llama 2存在的一个问题，即在处理涉及争议性话题时表现不足。这一问题的解决将对Llama 3的全面性能产生重要影响。
Meta的这一举措凸显了其在人工智能领域的持续投入和对技术领先性的追求。随着Llama 3的即将发布，科技界正密切关注Meta如何在AI模型的性能和多模态处理能力上实现新的突破。
英语如下：
**News Title:** “Meta Set to Launch Llama 3 in July, Challenging GPT-4 with Doubled Parameters!”
**Keywords:** Meta, Llama 3, GPT-4
**News Content:** According to sources from Wall Street Heard, tech giant Meta is planning to unveil its latest AI model, Llama 3, in the upcoming month of July. Insider sources reveal that Meta is banking on this new model to compete with OpenAI’s renowned large language model, GPT-4, which excels in multimodal processing, tackling more complex text and image tasks. However, it is still uncertain whether Llama 3 will also possess multimodal capabilities, as Meta’s internal team has yet to reach a consensus.
In terms of parameter scale, Llama 3 is expected to significantly surpass its predecessor. Reports indicate that it could contain over 140 billion parameters, a substantial leap from Llama 2’s 70 billion parameter limit when it was launched last July. Nevertheless, while advancing the Llama 3 project, Meta is concurrently addressing an issue with Llama 2 – its inadequate performance when dealing with controversial topics. Resolving this issue will have a significant impact on Llama 3’s overall performance.
Meta’s move underscores its continued investment in AI and its pursuit of technological leadership. With the imminent launch of Llama 3, the tech community is closely watching how Meta will achieve new breakthroughs in AI model performance and multimodal processing capabilities.
【来源】https://wallstreetcn.com/articles/3709312
Views: 1