The Humanoid Robot of the Future Is a 6-Foot-Tall Beefcake With a Chinese Body and an American Brain

The Humanoid Robot of the Future Is a 6-Foot-Tall Beefcake With a Chinese Body and an American Brain

未来的仿生机器人:拥有中国躯体与美国大脑的六英尺“壮汉”

The humanoid robot of the future is a hulking specimen with a body that’s made in China and a brain that runs on American silicon. 未来的仿生机器人是一个庞然大物,它拥有中国制造的躯体,以及由美国芯片驱动的大脑。

This week, Jensen Huang, the CEO of Nvidia, announced a blueprint for the bot, which combines a few different things: a 6-foot, 150-pound robot called H2 Plus from Unitree, a high-flying Chinese robotics startup; a Thor T5000 Nvidia chip; an advanced humanoid hand; and a new suite of software, which makes it easy to program and train the machine. Taken together, they’ll make it easier for researchers, including US academic labs, to put together cutting-edge humanoids and train them with their own AI algorithms. 本周,英伟达(Nvidia)首席执行官黄仁勋公布了一项机器人蓝图,该方案整合了多项技术:来自中国高成长机器人初创公司宇树科技(Unitree)的一款身高6英尺、重150磅的H2 Plus机器人;英伟达Thor T5000芯片;先进的仿生手;以及一套能简化机器编程与训练的新型软件套件。这些技术的结合,将使包括美国学术实验室在内的研究人员能够更轻松地组装尖端仿生机器人,并利用他们自己的AI算法对其进行训练。

The Thor chip can run powerful AI models that allow the bot to make sense of its environment and control its movements, while the body features Unitree’s motors, actuators, and sensors. The dextrous, humanlike hand from Singaporean company Sharpa can do everything from card tricks to peeling an apple. (Dexterity remains a key unsolved problem in robotics.) Thor芯片能够运行强大的AI模型,使机器人能够感知环境并控制自身动作,而其躯体则采用了宇树科技的电机、执行器和传感器。来自新加坡公司Sharpa的灵巧仿生手可以完成从玩纸牌戏法到削苹果等各种动作。(灵巧度仍是机器人领域尚未解决的关键难题。)

Spencer Huang, Nvidia’s director of product for robotics, told WIRED that the company wants to provide its silicon smarts for as many humanoid companies as possible. “Unitree is the first, but they’re not going to be the last by a long shot,” Huang said. (Yes, he’s Jensen’s son.) He added that the technology in H2 could potentially make other Chinese robots, including conventional industrial arms, more capable. 英伟达机器人产品总监斯宾塞·黄(Spencer Huang)告诉《连线》(WIRED)杂志,公司希望为尽可能多的仿生机器人公司提供其芯片智能。“宇树科技是第一家,但绝不会是最后一家,”黄说。(没错,他是黄仁勋的儿子。)他补充道,H2中的技术还有潜力提升其他中国机器人的性能,包括传统的工业机械臂。

In some ways, the partnership is unexpected: Robotics has emerged as a critical new arena for US-China techno-competition, and some politicians have proposed banning Chinese humanoids altogether. Last year, security researchers claimed that Unitree’s robots were capable of capturing and transmitting data, raising security risks. 在某些方面,这种合作出人意料:机器人技术已成为美中科技竞争的关键新领域,一些政客甚至提议全面禁止中国仿生机器人。去年,安全研究人员声称宇树科技的机器人具备捕获和传输数据的能力,引发了安全担忧。

But in other ways, the team-up makes perfect sense. “This is a fascinating development,” says Scott Singer, a fellow at the Carnegie Endowment for International Peace who studies AI governance and China. Singer notes that while the US has the world’s best AI chips, China’s supply chain gives its robotics companies a hardware edge. “Both sides have key parts of the supply chain that they might be able to weaponize, but here they are working together,” he says. 但在另一方面,这种合作又合乎情理。“这是一个引人入胜的发展,”卡内基国际和平基金会研究AI治理与中国的研究员斯科特·辛格(Scott Singer)表示。辛格指出,虽然美国拥有世界上最好的AI芯片,但中国的供应链赋予了其机器人公司硬件优势。“双方都掌握着可能被武器化的关键供应链环节,但在这里,他们选择了合作,”他说。

Nvidia, for its part, appears to be aware of the security concerns. Besides nimble fingers and a new brain, the new H2 Plus blueprint comes with security features that seem designed to reassure users that their data and models are safe. 英伟达方面似乎意识到了安全隐患。除了灵巧的手指和全新的大脑,新的H2 Plus蓝图还配备了安全功能,旨在让用户确信其数据和模型是安全的。

Nvidia’s chips are currently the gold standard for training large AI models, and the company has made big efforts to move into advanced robotics by developing specialized hardware and software tools. The US government bars Nvidia from selling its most capable chips to China but late last year loosened restrictions to allow it to sell more advanced chips there. 英伟达的芯片目前是训练大型AI模型的黄金标准,该公司正通过开发专业的硬件和软件工具,大力进军先进机器人领域。美国政府禁止英伟达向中国出售其性能最强的芯片,但在去年年底放宽了限制,允许其向中国销售部分更先进的芯片。

Singer says that robots and AI are widely seen as key to manufacturing and economic productivity, future military capabilities, and advances in AI itself. He believes it will be crucial for the US to figure out how to foster its own robotics industry, which might well mean finding ways of working with Chinese manufacturers. 辛格表示,机器人和AI被广泛视为制造业、经济生产力、未来军事能力以及AI自身进步的关键。他认为,美国必须找到培育自身机器人产业的方法,这很可能意味着需要寻找与中国制造商合作的途径。

Unitree’s robots are already hugely popular inside China and abroad; they’re relatively easy to program and remarkably cheap. A base version of the G1 humanoid costs around $15,000, compared with competitors’ robots that can run several hundreds of thousands of dollars a pop. Unitree’s bots can often be seen doing parkour, kung-fu, and other acrobatic feats in social media videos, and they’re featured in research published by many Western labs. 宇树科技的机器人在中国国内外已经非常受欢迎;它们编程相对简单,且价格极其低廉。G1仿生机器人的基础版售价约为1.5万美元,而竞争对手的机器人单价往往高达数十万美元。在社交媒体视频中,经常可以看到宇树的机器人进行跑酷、功夫和其他杂技表演,它们也出现在许多西方实验室发表的研究成果中。

Not everyone is enthused to see Chinese robot makers grow rapidly. Gavin Kenneally, cofounder and CEO of Ghost Robotics, which makes legged robots for defense and security, says he believes Unitree’s technology has drawn extensively on innovations from Western labs. He adds that it will be crucial that the US does not allow China to dominate the market for humanoids or any other robots. 并非所有人都乐见中国机器人制造商的快速增长。为国防和安全领域制造足式机器人的Ghost Robotics公司联合创始人兼首席执行官加文·肯尼利(Gavin Kenneally)表示,他认为宇树科技的技术大量借鉴了西方实验室的创新成果。他补充说,美国绝不能让中国主导仿生机器人或其他任何机器人市场,这一点至关重要。

“Without a serious near-term policy response, including a national robotics strategy, the US risks ceding the commercial robotics market to Unitree and other Chinese companies, as we’ve already seen happen in the drone space with DJI,” he says. “如果没有严肃的近期政策应对,包括制定国家机器人战略,美国就有可能将商业机器人市场拱手让给宇树科技和其他中国公司,正如我们在大疆(DJI)无人机领域所看到的那样,”他说。

But Nvidia’s CEO sees plenty of upside to working with Chinese robot makers. “Humanoid robots will bring physical AI to the world’s largest industries, opening a multitrillion-dollar economic opportunity,” Huang said. 但英伟达首席执行官看到了与中国机器人制造商合作的巨大潜力。“仿生机器人将把物理AI带入全球最大的产业,开启一个数万亿美元的经济机遇,”黄仁勋说道。