NeuroMesh 透過分散式訓練協定引領人工智慧新時代

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

英國倫敦，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.

人工智慧領域的拓荒者 NeuroMesh (nmesh.io) 宣布推出其分散式人工智慧訓練協議，有望徹底改變人工智慧開發領域的全球存取和協作。 NeuroMesh 採用 DePIN 的去中心化框架，彌合了訓練大型 AI 模型和分散式 GPU 的需求之間的差距。該倡議旨在促進人工智慧開發的包容性，促進不同部門和地區的參與。

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 領域的遠見者：團隊的全球野心 NeuroMesh 背後的團隊由來自牛津大學、南洋理工大學、北大、清華大學、香港大學、谷歌和 Meta 的研究人員和工程師組成，開創了民主的 AI 訓練過程。這種富有遠見的方法使世界各地的GPU 所有者能夠為龐大的訓練網絡做出貢獻，從而使各種規模的實體能夠利用該服務來滿足其訓練需求，從而解決了集中式AI 開發的局限性。

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 透過促進協作超越了傳統人工智慧。他們的願景是讓每個開發人員和組織，無論位置或資源如何，都能夠訓練和利用尖端的人工智慧模型。這與 Yann LeCun 等人工智慧先驅的願景完美契合，他們主張透過眾包和分散式人工智慧培訓來推動未來。

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.

基於 PCNA 的革命性設計 NeuroMesh 分散式訓練協定的核心在於突破性的 PCN（預測編碼網路）訓練演算法——該領域真正的遊戲規則改變者。這種方法使世界各地的 GPU 所有者能夠貢獻自己的力量，促進廣泛的協作。

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 訓練演算法：NeuroMesh 背後的魔力在於 PCN 訓練演算法。與傳統的反向傳播 (BP) 方法不同，PCN 能夠實現完全本地、並行和自主的訓練。該團隊的目標是創建一個龐大的網絡，其中每個節點（代表參與的 GPU）獨立學習。 PCN 最大限度地減少了層間通信，減少了資料流量並促進非同步訓練。將其視為一首交響樂，每位音樂家都獨立演奏自己的部分，但又為和諧的整體做出了貢獻。

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.

這個尖端模型的靈感來自牛津大學開創的神經科學研究的最新進展，模仿了人腦的局部學習方法。透過儲存誤差值並針對每一層中的局部目標進行最佳化，它複製了大腦神經元的行為。這使得 NeuroMesh 能夠定義更大的模型，各個組件有助於整個網路實現相同的最終最佳化目標，就像人腦中不同的神經元組處理不同的刺激一樣。

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

這種受生物學啟發的方法與其固有的分發能力相結合，開啟了人工智慧開發的新時代。

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.

呼籲建立全球合作夥伴關係NeuroMesh 邀請全球合作夥伴，旨在打造一個人人都可以參與的人工智慧未來。它的協議是建立多元化生態系統的基石。此生態系統的設計具有動態性、協作性和適應性，確保能夠滿足任何產業、任何規模的人工智慧模型訓練需求。

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.

歡迎個人、擁有 GPU 資源的專案以及有培訓需求的實體加入這項變革性措施。有關 NeuroMesh 的全面詳細資訊並參與這項前沿工作，用戶可以訪問 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.

關於 NeuroMeshNeuroMesh 由來自牛津大學、南洋理工大學、北大、清華大學、香港大學、谷歌和 Meta 等知名機構的研究人員和工程師組成。透過協助開發人員和組織部署強大的人工智慧模型，NeuroMesh 正在培育一個包容性的人工智慧生態系統，縮小全球範圍內訓練大型人工智慧模型和分散式 GPU 的需求之間的差距。

For more information, users can visit NeuroMesh’s

欲了解更多信息，用戶可以訪問 NeuroMesh 的

• Twitter
• Telegram

 

推特電報

Contact

接觸

CMO
Kenchia Lee
NeuroMesh
[email protected]
07746906341

CMOKenchia LeeNeuroMesh[電子郵件受保護]07746906341

分享這篇文章 類別 新聞稿