谷歌DeepMind近日在人工智能领域再次推出创新成果，发布了全新的基础模型——Hawk和Griffin。这两款模型的设计旨在提升人工智能的处理效率和性能，特别是在复杂任务中的表现。
在一篇近期发表的论文中，DeepMind的研究团队提出了革命性的RG-LRU层，这是一种创新的门控线性循环层。RG-LRU层替代了传统的多查询注意力（MQA）机制，以构建更为高效的循环处理单元。基于这一新技术，研究人员设计了一个全新的循环块，该块在处理信息流时能够更好地保留和更新上下文信息。
利用这个创新的循环块，DeepMind构建了两个独特的模型。首先是Hawk模型，它巧妙地融合了多层感知机（MLP）与循环块，旨在平衡模型的深度学习能力和序列处理能力。另一方面，Griffin模型更进一步，不仅结合了MLP和循环块，还引入了局部注意力机制，以增强模型在处理局部特征时的敏感度和精确度。
这些新模型的推出，标志着谷歌DeepMind在人工智能基础架构上的又一重大突破，有望在自然语言处理、图像识别以及其他依赖深度学习的领域产生深远影响。未来，Hawk和Griffin可能会成为研究人员和开发者的新工具，推动AI技术向更智能、更高效的方向发展。
英语如下：
**News Title:** “Google DeepMind Unveils Innovative Models Hawk and Griffin: Redefining AI Learning Horizons”
**Keywords:** DeepMind, Hawk, Griffin
**News Content:** Google’s DeepMind has recently made another groundbreaking advancement in the field of artificial intelligence with the release of its novel foundational models, Hawk and Griffin. These models have been designed to enhance the efficiency and performance of AI, particularly in complex tasks.
In a recently published paper, DeepMind’s research team introduced the Revolutionary RG-LRU layer, an innovative gated linear recurrent unit. This replaces the conventional multi-query attention (MQA) mechanism, leading to a more efficient recurrent processing unit. Based on this new technology, researchers designed a novel recurrent block that better preserves and updates context information while processing information flow.
Utilizing this innovative recurrent block, DeepMind has constructed two unique models. The first, the Hawk model, skillfully combines multi-layer perceptrons (MLPs) with the recurrent block, aiming to balance the model’s deep learning capabilities with its sequence processing abilities. On the other hand, the Griffin model goes a step further, not only integrating MLPs and the recurrent block but also incorporating local attention mechanisms, thereby enhancing the model’s sensitivity and precision in handling local features.
These new models signify another major breakthrough for Google DeepMind in AI infrastructure and are expected to have a profound impact on natural language processing, image recognition, and other domains reliant on deep learning. In the future, Hawk and Griffin may become new tools for researchers and developers, propelling AI technology towards increased intelligence and efficiency.
【来源】https://mp.weixin.qq.com/s/RtAZiEzjRWgqQw3yu3lvcg
Views: 1