内容来源：https://aibusiness.com/intelligent-automation/nvidia-s-ai-driving-tech-debuts-in-mercedes-cla-by-2026

内容总结：

【科技前沿】英伟达联手梅赛德斯-奔驰 推出新一代自动驾驶系统

在谷歌云的支持下，生成式人工智能正加速融入汽车产业。英伟达近日宣布，其基于开源模型家族Alpamayo开发的新一代自动驾驶软件Drive AV，将于2026年底前首次量产落地，搭载车型为梅赛德斯-奔驰全新CLA四门轿跑。

该系统将提供SAE标准下的L2级点对点驾驶辅助功能，实现转向、刹车与加速的自动化控制，驾驶员需随时准备接管。值得注意的是，Alpamayo模型家族专为解决自动驾驶“长尾挑战”设计——即应对罕见极端路况，传统自动驾驶软件在此类场景中往往表现不佳。

在拉斯维加斯国际消费电子展上，英伟达首席执行官黄仁勋将这一进展称为“物理AI的ChatGPT时刻”：“Alpamayo为自动驾驶车辆赋予推理能力，使其能思考罕见场景、在复杂环境中安全行驶并解释驾驶决策，这是实现安全、可扩展自动驾驶的基础。”

此次公开的核心组件包括：拥有100亿参数的视觉-语言-动作模型Alpamayo 1，其采用“思维链”推理规划路径并展示决策依据；端到端开源仿真框架AlpaSim；以及超1700小时的物理AI开源数据集。该开源技术已吸引捷豹路虎、Lucid和优步等企业的关注，这些公司正致力于构建实现L4级自动驾驶的推理系统。

在实际应用中，英伟达Drive AV系统将通过Halo安全系统提供冗余保障与安全护栏。梅赛德斯-奔驰表示，新车将支持复杂城市环境点对点导航、主动防碰撞及狭窄空间自动泊车等功能。

英伟达汽车业务副总裁阿里·卡尼强调：“在汽车产业拥抱物理AI的时代，英伟达正成为智能支柱，通过数据与软件让每辆车实现可编程、可升级与持续进化。”据悉，首批搭载该技术的CLA车型预计将于今年年底面世。

中文翻译：

由谷歌云赞助
如何选择首个生成式人工智能应用场景
要开启生成式人工智能之旅，首先应关注能够优化人类信息交互体验的领域。

这家汽车制造商将成为首个应用阿尔帕马约系列开源模型的企业，该系列专为应对长尾自动驾驶难题而设计。
英伟达确认其新一代自动驾驶软件将于2026年底前搭载于量产车型。

德国汽车制造商梅赛德斯-奔驰的全新CLA四门轿跑将率先搭载英伟达Drive AV系统。这套基于人工智能的系统可提供增强版L2级点对点驾驶辅助功能。根据国际汽车工程师学会的定义，L2级自动驾驶意味着车辆能够控制转向、制动与加速，但驾驶员仍需随时准备接管车辆。

此类功能在当下并不罕见——事实上梅赛德斯已通过其L3级DrivePilot技术实现超越，该技术允许特定条件下"放开双手、移开视线"的自动驾驶。但英伟达产品将成为首个基于阿尔帕马约系列开源模型的解决方案，该系列专为攻克长尾自动驾驶挑战而生，包括传统自动驾驶软件难以处理的罕见边缘场景。

阿尔帕马约系列在拉斯维加斯消费电子展上与其他模型同步亮相，英伟达首席执行官黄仁勋现场阐释了其重要意义："物理人工智能的'ChatGPT时刻'已经到来——机器开始理解现实世界、进行推理并采取行动。阿尔帕马约为自动驾驶车辆赋予推理能力，使其能推演罕见场景、在复杂环境中安全行驶并解释驾驶决策，这是构建安全可扩展自动驾驶系统的基石。"

阿尔帕马约的核心组件已在CES发布，其中包括拥有100亿参数的视觉-语言-动作模型"阿尔帕马约1号"。该模型运用"思维链"推理规划行驶路径，并展示每个决策的推理过程。后一特性尤其值得监管机构关注，他们需要明确掌握自动驾驶事故的具体成因。同步发布的还有全开源端到端仿真框架AlpaSim，以及超过1700小时的物理人工智能开源数据集。英伟达表示，该技术的开源特性已吸引捷豹路虎、Lucid及优步等企业的关注，这些企业正致力于构建实现L4级自动驾驶（特定场景完全自主）的推理技术栈。

在梅赛德斯CLA车型中，英伟达Drive AV解决方案由Nvidia Halos安全系统提供支持，该系统增加了冗余设计和安全防护机制。梅赛德斯宣称将提供的自动化及驾驶辅助功能包括：复杂城市环境中的点对点导航、主动防碰撞系统以及狭窄空间自动泊车。

英伟达汽车业务副总裁阿里·卡尼在新闻稿中表示："随着汽车行业拥抱物理人工智能，英伟达正成为智能中枢，通过数据与软件使每辆汽车实现可编程、可升级和持续进化。"
首批搭载英伟达新技术的CLA车型预计将于今年年底面世。

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.
The automaker will be the first to utilize Alpamayo, a new family of open source models created to tackle long-tail autonomous driving challenges.
Nvidia confirmed that its new autonomous driving software will be launched in a production car by the end of 2026.
German automaker Mercedes' new CLA four-door coupé will be the first to debut Nvidia Drive AV, an AI-based system that brings added Level 2 point-to-point driver assistance. Society of Automotive Engineers consider Level 2 automation is to be when a car can control the steering, braking and acceleration, although the human driver must be ready to intervene at all times.
This level of functionality is not unusual — indeed Mercedes already surpasses it with its own Level 3 DrivePilot tech, which enables "hands off, eyes off" driving in certain circumstances.
But the Nvidia product will be the first to be based on Alpamayo, a new family of open source models designed to tackle long-tail autonomous driving challenges — including unusual edge cases that have traditionally proven a problem for autonomous driving software.
Alpamayo was showcased alongside other models at the Consumer Electronics Show in Las Vegas, where Nvidia CEO Jensen Huang explained its significance.
"The ChatGPT moment for physical AI is here — when machines begin to understand, reason and act in the real world," Huang said. "Alpamayo brings reasoning to autonomous vehicles, allowing them to think through rare scenarios, drive safely in complex environments and explain their driving decisions — it's the foundation for safe, scalable autonomy."
Key components of Alpamayo were released at CES, among them Alpamayo 1, a 10 billion parameter vision-language-action model that uses "chain of thought" reasoning to plot a path, showing reasoning for each decision. That latter element is likely to be of particular interest to regulators, who need to know specifically what has happened in the event of an AV crash.
It was joined by AlpaSim, a full open-source end-to-end sim framework to develop AVs, plus more than 1,700 hours of physical AI open data sets. Nvidia says the open source nature of the tech has already attracted the interest of automakers JLR and Lucid as well as Uber, as they look toward building reasoning stacks to achieve Level 4 automation, or full autonomy in specific circumstances.
In the Mercedes CLA, Nvidia's Drive AV solution is underpinned by the Nvidia Halos safety system, which adds redundancy and safety guardrails.
Point-to-point navigation in complex urban environments, proactive collision avoidance and automated parking in tight spaces are among the automated and driver-assistance features Mercedes said it will offer.
"As the automotive industry embraces physical AI, Nvidia is the intelligence backbone that makes every vehicle programmable, updatable and perpetually improving through data and software," said Ali Kani, Nvidia vice president of automotive, said in a press release.
The first CLAs to use the new Nvidia tech are expected by the end of this year.