内容来源：https://aibusiness.com/generative-ai/nvidia-takes-on-telco-industry-with-new-model

内容总结：

在近日举行的巴塞罗那世界移动通信大会上，英伟达发布了面向电信行业的大语言模型“大型电信模型”（LTM），旨在帮助电信企业利用自身数据训练出更懂行业流程的专属AI模型。

该模型基于英伟达去年12月发布的Nemotron 3基础模型系列构建，专门学习了电信行业的专业语言与工作流程，包括故障定位、修复方案制定和变更验证等核心环节。英伟达称，LTM能够理解运营商意图，并做出非预设的决策，从而超越传统基于固定规则的自动化方案。

行业分析指出，电信运营商正日益需要此类垂直领域模型来推动网络向自主化演进。除英伟达外，微软与沃达丰合作部署基于Azure的智能体用于网络运维，AMD也参与了“开放电信AI”计划，行业竞争正在加剧。

尽管英伟达凭借AI软硬件优势进入市场，但其仍面临传统电信设备商爱立信和诺基亚等拥有深厚客户基础的对手竞争。分析师指出，电信领域问题复杂，AI并非万能解药；同时，开源模型能否被电信IT部门快速消化应用也有待观察。此外，英伟达也强调其AI模型将注重透明度、安全与治理，以适应电信行业对可靠性的高要求。

中文翻译：

由谷歌云赞助
选择您的首个生成式AI应用场景
要开启生成式AI之旅，首先应关注能够提升人类信息交互体验的领域。

尽管英伟达致力于为电信公司打造更自主的工作流程，但其正面临爱立信、诺基亚等传统网络设备商的竞争。英伟达近期开源了"大型电信模型"，专为电信行业企业设计，该模型能基于企业自身数据训练，并深度理解其业务流程。

这家AI软硬件巨头表示，新模型基于其去年12月发布的Nemotron 3系列基础模型构建。据英伟达介绍，大型电信模型通过故障隔离（定位系统故障环节）、补救规划（制定缺陷修复方案）及变更验证（审核系统变更）等工作流程训练，能深入理解电信行业的专业术语与运作流程。除该模型外，英伟达还推出了"意图驱动无线接入网能效蓝图"，这是一种用于能源优化的闭环智能工作流。

在2月28日巴塞罗那世界移动通信大会上亮相的LTM，反映了电信行业日益需要领域专用模型来推动网络自主化的趋势。英伟达并非唯一提供此类领域模型的厂商。例如，微软正与沃达丰合作部署基于Azure的智能体用于网络运维，AMD也参与"开放电信AI"计划，为电信专用模型提供硬件与算力支持。

通过开源LTM，英伟达使电信运营商能够创建专属数据集，并将其深度整合至自身系统与网络中。高德纳分析师苏珊·韦尔什·德格里马尔多指出："该模型运用了所有行业标准、信息及数据集进行训练开发。"

未来集团分析师尼克·佩兴斯认为，英伟达不仅提供领域专用模型，更致力于解决电信团队在流程自动化中面临的挑战："现行自动化基于规则制定，一旦出现脚本外状况就会失效。"他表示LTM专注于系统上层，能通过推理模式解读运营商意图，做出未明确指示的决策。"网络运维失误会造成实际影响，推理模型通过多步骤问题推演来应对，而非简单匹配既有模式。"

佩兴斯补充说，英伟达同样重视AI模型的透明度、安全性、指导原则与治理框架，深知这些对电信企业至关重要。韦尔什·德格里马尔多指出："构建自主网络与AI运维智能体需要人机协同。"她强调英伟达认识到模型需从网络运维工程师视角进行技能训练，这能"推动网络内部各项功能的运行效能提升"。

然而英伟达面临的关键挑战在于，需与拥有稳固客户基础的传统网络设备商竞争。Informa TechTarget旗下Omdia分析师苏连杰指出："英伟达是该领域的新入局者。"尽管这家AI硬件商与电信设备商合作多年，但首次以竞争性产品进入市场。此外，AI并非解决电信领域所有难题的万能钥匙。"引入AI确实有助于降低复杂度，但本质上存在诸多更细微的挑战，仅靠英伟达无法解决。"

另一项挑战在于，佩兴斯指出，尽管模型已开源，但"电信IT组织能否快速消化吸收并发挥实效仍有待观察"。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
While Nvidia's approach focuses on enabling more autonomous workflows for telco companies, it faces competition from traditional network vendors such as Ericsson and Nokia.
Nvidia released an open source Large Telco Model aimed at enterprises in the telecommunications industry looking for AI models that are trained specifically on their data and knowledgeable about their processes.
The AI hardware-software giant said it built the new model on its Nemotron 3 family of foundation models, which it released in December.
The Large Telco Model (LTM) is trained to understand the telecommunications industry's language and processes through workflows such as fault isolation, the process of locating the part of a system that is responsible for failure; remediation planning, or planning ways to correct a deficiency or failure; and change validation, the process of approving changes to a system, Nvidia said. Aside from the LTM, Nvidia unveiled its Intent-Driven RAN Energy Efficiency Blueprint, a closed-loop agentic workflow for energy optimization.
The LTM, introduced on Feb. 28 at the Mobile World Congress Barcelona, reflects a trend in which telecommunications companies increasingly require domain-specific models to drive more autonomous networks. Nvidia is not the only model provider working toward providing these domain-specific models. For example, in partnership with Vodafone, Microsoft is deploying Azure-powered agents that can be used for network operations. AMD is also participating in the "Open Telco AI" initiative and providing hardware and compute power to run telco-specific models.
For its part, with the open source LTM, Nvidia is enabling telcos to create their own datasets and integrate them more deeply into their systems and networks.
"[It’s] a model that's been trained and developed using all the industry standards and information and data sets," said Susan Welsh de Grimaldo, an analyst at Gartner.
Beyond providing a domain-specific model, Nvidia is also offering a way to address the challenges telco teams face when automating their processes, said Nick Patience, an analyst at Futurum Group.
"Current automation is rules-based," Patience said. "It breaks down the moment something falls outside the script."
The LTM focuses on the layer above the system that can interpret the operator's intent and make decisions that weren't explicitly stated, using its reasoning mode, he said.
"Network operations have real consequences when they fail, and reasoning models work through multi-step problems rather than pattern-matching to a likely answer," Patience said.
Nvidia is also focusing on transparency, security, guidelines, and governance for its AI models, recognizing that these are important to enterprises in the telco market, he said.
"When you think about autonomous networks and building AI agents for running networks, we're going to see both human and AI involvement," Welsh de Grimaldo said. She added that Nvidia, understanding this, realized that the models need to be trained to understand skills building from a network operations engineer's perspective, which can "drive better improvements in how it's used in running different functions inside the network."
However, a key challenge Nvidia faces is that it competes with traditional network vendors with well-established customer bases, such as Ericsson and Nokia.
"Nvidia's new to the domain," said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget. He added that while the AI hardware provider has been working with telco vendors for years, this is the first time the vendor is offering its product as a competing alternative. Moreover, AI is not the only answer to some of the challenges in the telco arena.
"Bringing AI to the table does help reduce a lot of the complexity, but fundamentally, there are a lot of more nuanced challenges that Nvidia alone won't be able to solve," Su said.
Another challenge is that while the model is open source, it remains to be seen "whether telco IT organizations can absorb it fast enough to matter," Patience said.