谷歌DeepMind近日宣布推出两款全新的基础模型——“Hawk”和“Griffin”，标志着其在人工智能技术领域的又一重要突破。这两款模型基于一篇最新发表的论文，该论文提出了一种名为“RG-LRU”的循环层，这是一种新型的门控线性循环层，旨在提高模型的效率和性能。
RG-LRU层被设计为一个新的循环块，用以取代多查询注意力（MQA），后者是深度学习模型中常见的技术。DeepMind的研究者通过结合循环块与多层的感知器（MLP），构建了混合模型“Hawk”。而“Griffin”则进一步融合了局部注意力机制，以提升模型的处理能力。
这两款新模型的推出，不仅展示了DeepMind在AI技术研发上的领先地位，也为未来的应用提供了更多可能性。随着技术的不断迭代和优化，AI模型在各个领域的应用将更加广泛和深入。
英文标题：Google DeepMind Unveils New Foundation Models “Hawk” and “Griffin”
英文关键词：AI Models, Deep Learning, Recurrent Layers
英文新闻内容：
Google DeepMind has recently announced the launch of two new foundation models, “Hawk” and “Griffin,” marking another significant breakthrough in the field of artificial intelligence technology. These models are based on a new paper that introduces a novel type of gated linear recurrent layer called “RG-LRU,” designed to improve the efficiency and performance of AI models.
RG-LRU is implemented as a new recurrent block to replace the multi-query attention (MQA), a common technique in deep learning models. The researchers at DeepMind have built a hybrid model called “Hawk” by combining the recurrent block with multi-layer perceptrons (MLPs). “Griffin,” on the other hand, further integrates local attention mechanisms to enhance the processing capabilities of the model.
The introduction of these new models not only showcases DeepMind’s leadership in AI research and development but also opens up more possibilities for future applications. As the technology continues to iterate and optimize, the application of AI models across various fields is expected to become more widespread and profound.
【来源】https://mp.weixin.qq.com/s/RtAZiEzjRWgqQw3yu3lvcg
Views: 1