英伟达成立新研究部门 GEAR，打造虚拟与现实世界中的全能智能体
科技巨头英伟达宣布成立一个新的研究部门，名为“GEAR”，全称为通用具身智能体研究（Generalist Embodied Agent Research）。该部门由人工智能高级研究科学家、人工智能代理计划负责人吉姆·范（Jim Fan）和教授于克朱（Yuke Zhu）领导。
GEAR 的核心使命是构建适用于虚拟和物理世界的具身智能体的基础模型，致力于实现跨多模态、多场景的智能应用。研究团队将专注于四大关键领域：
1. 多模态基础模型：开发能够处理各种数据类型（例如文本、图像、视频和音频）的通用模型，为具身智能体提供广泛的理解和生成能力。
2. 通用型机器人研究：探索机器人与环境交互的通用原则，使机器人能够在广泛的场景中执行复杂任务。
3. 虚拟世界中的基础智能体：创建在虚拟世界中表现出智能行为的虚拟智能体，为训练和评估具身智能体提供一个安全和可控的环境。
4. 模拟与合成数据技术：开发用于生成逼真数据的高级模拟和合成技术，以支持具身智能体的训练和评估。
英伟达表示，GEAR 的研究成果将广泛应用于各个领域，包括机器人技术、虚拟现实、增强现实和自动驾驶。
吉姆·范表示：“GEAR 将汇集世界领先的人工智能研究人员，共同努力构建下一代具身智能体。我们的目标是创造能够在虚拟和现实世界中无缝导航并执行复杂任务的智能体。”
于克朱补充道：“GEAR 的研究将为人工智能领域带来变革。通过开发通用具身智能体，我们有望解决人工智能中一些最具挑战性的问题，并为未来技术创造新的可能性。”
英语如下：
**Headline: NVIDIA Builds ‘Universal Intelligence’ to Bridge Virtual and Real**
**Keywords: Embodied Intelligence, Multimodal, Virtual Worlds**
**Body:** NVIDIA has launched a new research division called “GEAR,” short for GeneralistEmbodied Agent Research. The division is led by Jim Fan, senior research scientist for AI and head of AI agent programs, and Professor Yuke Zhu.
GEAR’s central mission is to build foundational models for embodied intelligence that work in both virtual and physical worlds, enabling intelligent applications across multiple modalities and scenarios. Theresearch team will focus on four key areas:
1. **Multimodal Foundation Models:** Developing general-purpose models that can process a variety of data types (e.g., text, images, video, audio), giving embodied agents broad understanding and generation capabilities.
2. **Generalist Robotics:** Exploring general principles for robot interaction with their environments, enabling robots to perform complex tasks in a wide range of scenarios.
3. **Foundation Agents in Virtual Worlds:** Creating virtual agents that exhibit intelligent behavior in virtual worlds, providing a safe and controlled environment for training and evaluating embodied agents.
4. **Simulation and Synthetic Data Techniques:**Developing advanced simulation and synthetic techniques for generating realistic data to support the training and evaluation of embodied agents.
NVIDIA says the research from GEAR will have broad applications in areas including robotics, virtual reality, augmented reality, and autonomous vehicles.
“GEAR brings together the world’s leading AI researchers to work on building the next generation of embodied intelligence,” said Fan. “Our goal is to create agents that can seamlessly navigate and perform complex tasks in both virtual and real worlds.”
Zhu added: “GEAR’s research will be transformative for the field of AI. By developing generalist embodied intelligence, we can solve some of the most challenging problems in AI and open up new possibilities for future technologies.”
【来源】https://mp.weixin.qq.com/s/-ep_nCt6moygQEm0MjOxqw
Views: 0