NeuroMesh 通过分布式训练协议引领人工智能新时代

London, United Kingdom, April 9th, 2024, Chainwire

英国伦敦，2024 年 4 月 9 日，Chainwire

NeuroMesh (nmesh.io), a trailblazer in artificial intelligence, announces the rollout of its distributed AI training protocol, poised to revolutionize global access and collaboration in AI development. Embracing DePIN’s decentralized framework, NeuroMesh bridges the gaps between the demand for training large AI models and distributed GPUs. This initiative aims to foster inclusivity in AI development, facilitating participation across diverse sectors and geographies.

人工智能领域的开拓者 NeuroMesh (nmesh.io) 宣布推出其分布式人工智能训练协议，有望彻底改变人工智能开发领域的全球访问和协作。 NeuroMesh 采用 DePIN 的去中心化框架，弥合了训练大型 AI 模型和分布式 GPU 的需求之间的差距。该倡议旨在促进人工智能开发的包容性，促进不同部门和地区的参与。

Visionaries in AI: The team’s global ambition

The team behind NeuroMesh, composed of researchers and engineers from Oxford, NTU, PKU, THU, HKU, Google, and Meta, pioneers a democratic AI training process. This visionary approach addresses the limitations of centralized AI development by enabling GPU owners worldwide to contribute to a vast training network, empowering entities of all sizes to leverage this service for their training needs.

AI 领域的远见者：团队的全球野心 NeuroMesh 背后的团队由来自牛津大学、南洋理工大学、北大、清华大学、香港大学、谷歌和 Meta 的研究人员和工程师组成，开创了民主的 AI 训练过程。这种富有远见的方法使世界各地的 GPU 所有者能够为庞大的训练网络做出贡献，从而使各种规模的实体能够利用该服务来满足其训练需求，从而解决了集中式 AI 开发的局限性。

NeuroMesh transcends traditional AI by fostering collaboration. Their vision is to equip every developer and organization, regardless of location or resources, with the ability to train and utilize cutting-edge AI models. This aligns perfectly with the vision of AI pioneers like Yann LeCun, who advocate for a future powered by crowdsourced and distributed AI training.

NeuroMesh 通过促进协作超越了传统人工智能。他们的愿景是让每个开发人员和组织，无论位置或资源如何，都能够训练和利用尖端的人工智能模型。这与 Yann LeCun 等人工智能先驱的愿景完美契合，他们主张通过众包和分布式人工智能培训来推动未来。

A revolutionary design based on PCN

At the heart of NeuroMesh’s distributed training protocol lies the groundbreaking PCN (Predictive Coding Network) training algorithm – a true game-changer in this field. This approach empowers GPU owners worldwide to contribute their power, fostering a vast collaborative effort.

基于 PCNA 的革命性设计 NeuroMesh 分布式训练协议的核心在于突破性的 PCN（预测编码网络）训练算法——该领域真正的游戏规则改变者。这种方法使世界各地的 GPU 所有者能够贡献自己的力量，促进广泛的协作。

The PCN Training Algorithm: The magic behind NeuroMesh lies in the PCN training algorithm. Unlike traditional backpropagation (BP) methods, PCN enables fully local, parallel, and autonomous training. The team aims to create a vast network, where each node—representing a participating GPU—learns independently. PCN minimizes inter-layer communication, reducing data traffic and facilitating asynchronous training. Think of it as a symphony where each musician plays their part independently, yet contributes to a harmonious whole.

PCN 训练算法：NeuroMesh 背后的魔力在于 PCN 训练算法。与传统的反向传播 (BP) 方法不同，PCN 能够实现完全本地、并行和自主的训练。该团队的目标是创建一个庞大的网络，其中每个节点（代表参与的 GPU）独立学习。 PCN 最大限度地减少了层间通信，减少了数据流量并促进异步训练。将其视为一首交响乐，每位音乐家都独立演奏自己的部分，但又为和谐的整体做出了贡献。

This cutting-edge model, inspired by recent advancements in neuroscience research pioneered by Oxford University, mimics the human brain’s localized learning approach. By storing error values and optimizing for a local target in each layer, it replicates the behavior of brain neurons. This allows NeuroMesh to define models that are much larger, with individual components that contribute to the same ultimate optimization objective for the whole network, just like the human brain where different stimuli are handled by different groups of neurons.

这种尖端模型的灵感来自于牛津大学开创的神经科学研究的最新进展，模仿了人脑的局部学习方法。通过存储误差值并针对每一层中的局部目标进行优化，它复制了大脑神经元的行为。这使得 NeuroMesh 能够定义更大的模型，各个组件有助于整个网络实现相同的最终优化目标，就像人脑中不同的神经元组处理不同的刺激一样。

This biologically-inspired approach, combined with its inherent distribution capabilities, unlocks a new era of AI development.

这种受生物学启发的方法与其固有的分发能力相结合，开启了人工智能开发的新时代。

A call to forge global partnerships

NeuroMesh invites partnerships globally, aiming to forge an AI future that everyone can participate in. Its protocol is the bedrock upon which a diverse ecosystem is being built. The ecosystem is designed to be dynamic, collaborative, and adaptable, ensuring that it can serve the AI model training needs of any size, from any industry.

呼吁建立全球合作伙伴关系NeuroMesh 邀请全球合作伙伴，旨在打造一个人人都可以参与的人工智能未来。它的协议是构建多元化生态系统的基石。该生态系统的设计具有动态性、协作性和适应性，确保能够满足任何行业、任何规模的人工智能模型训练需求。

Individuals, projects with GPU resources, and entities with training needs are all welcome to join this transformative initiative. For comprehensive details on NeuroMesh and to participate in this leading-edge endeavor, users can visit nmesh.io.

欢迎个人、拥有 GPU 资源的项目以及有培训需求的实体加入这一变革性举措。有关 NeuroMesh 的全面详细信息并参与这项前沿工作，用户可以访问 nmesh.io。

About NeuroMesh

NeuroMesh comprises researchers and engineers from esteemed institutions such as Oxford, NTU, PKU, THU, HKU, Google and Meta. By empowering developers and organizations to deploy robust AI models, NeuroMesh is cultivating an inclusive AI ecosystem, bridging the gaps between the demand of training large AI models and distributed GPUs worldwide.

关于 NeuroMeshNeuroMesh 由来自牛津大学、南洋理工大学、北大、清华大学、香港大学、谷歌和 Meta 等知名机构的研究人员和工程师组成。通过帮助开发人员和组织部署强大的人工智能模型，NeuroMesh 正在培育一个包容性的人工智能生态系统，缩小全球范围内训练大型人工智能模型和分布式 GPU 的需求之间的差距。

For more information, users can visit NeuroMesh’s

欲了解更多信息，用户可以访问 NeuroMesh 的

• Twitter
• Telegram

 

推特电报

Contact

接触

CMO
Kenchia Lee
NeuroMesh
[email protected]
07746906341

CMOKenchia LeeNeuroMesh[电子邮件受保护]07746906341

分享这篇文章 类别 新闻稿