##特斯拉为人形机器人 Optimus 招募“动捕师”，时薪最高48美元，训练数据成关键
随着一年一度的世界机器人大会即将开幕，机器人领域再度成为全球关注的焦点。国内外众多企业纷纷发布新产品，其中，特斯拉的人形机器人 Optimus 训练进展备受瞩目。近日，特斯拉发布招聘信息，正在招募“数据采集操作员”，即穿上动作捕捉服模仿预期动作的“动捕师”，为 Optimus 提供大量训练数据。
据悉，该职位时薪最高可达48美元，并提供现金、股票奖励和其他福利。但这份工作也对体力和耐力提出了极高要求。数据采集操作员每天需行走7小时以上，同时携带最多30磅的重物，身高需在170到180厘米之间，并需长时间佩戴和操作动捕服和VR头显。此外，还需要持续的手眼协调、身体协调、动觉意识等能力，以及全天站立、坐下、行走、弯腰、弯曲、伸展、蹲下和扭动等动作。
特斯拉官方 Optimus X 账号曾在今年5月预告过该职位的性质，据外媒 Business Insider 报道，过去一年特斯拉已经招募了50多名相关员工。
动作捕捉是训练机器人执行类人动作的一种常见且经济高效的方法，特斯拉是首批大规模使用动作捕捉技术的公司之一。然而，为了使机器人更自然、高效地执行各项任务，所需的训练数据量异常庞大。NVIDIA Research 的高级研究员 Animesh Garg 表示，Optimus 人形机器人可能需要数百万小时的数据才能为其在特斯拉工厂工作完全做好准备。
这无疑带来了巨大的成本压力。Animesh Garg 指出，“你在数据采集方面很可能花费上亿美元，但真正的问题是，即使你这样做了，最后能成功吗？”
特斯拉也面临着同样的挑战。马斯克此前曾表示，特斯拉将在明年小批量生产真正有用的人形机器人，供给特斯拉内部使用，并有望在2026年进行大批量出货，向其他公司销售。然而，自从去年12月发布二代 Optimus 以来，特斯拉迄今还没有大的版本更新。而随着今年波士顿动力新一代人形机器人 Atlas、Figure 新一代人形机器人 Figure 02 的陆续发布，特斯拉显然需要加把劲了。
特斯拉为 Optimus招募“动捕师”的举动，反映了人形机器人训练数据的重要性。未来，随着人工智能技术的不断发展，如何高效、低成本地获取高质量训练数据将成为人形机器人发展的重要挑战。
英语如下：
##Tesla’s Humanoid Robot ‘Learns’ with $48/Hour ‘Motion Capture’ Workers, Sparking Debate
**Keywords:** Tesla,humanoid robot, embodied intelligence
## Tesla is Hiring ‘Data Capture Operators’ to Train its Optimus Robot, Offering Up to $48/Hour
As theannual World Robot Conference approaches, the robotics industry is once again in the global spotlight. Numerous companies worldwide are releasing new products, and among them, Tesla’shumanoid robot Optimus has garnered significant attention for its training progress. Recently, Tesla posted a job opening for “Data Capture Operators,” individuals who wear motion capture suits to mimic intended movements, providing Optimus with a vast amount of training data.
Theposition offers a maximum hourly wage of $48, along with cash, stock bonuses, and other benefits. However, the job demands extreme physical strength and endurance. Data Capture Operators are required to walk for over 7 hours daily, carryingup to 30 pounds, with a height between 170 and 180 centimeters. They also need to wear and operate motion capture suits and VR headsets for extended periods. Additionally, the role requires continuous hand-eye coordination, body coordination, kinesthetic awareness, and the ability to perform actions suchas standing, sitting, walking, bending, flexing, stretching, squatting, and twisting throughout the day.
Tesla’s official Optimus X account previewed the nature of this role in May. According to Business Insider, Tesla has hired over 50 employees for this position in the past year.
Motion capture isa common and cost-effective method for training robots to perform human-like actions. Tesla is among the first companies to adopt motion capture technology on a large scale. However, the amount of training data required for robots to perform tasks naturally and efficiently is immense. Animesh Garg, a senior researcher at NVIDIA Research,suggests that Optimus may need millions of hours of data to be fully prepared for its work in Tesla factories.
This presents a significant cost challenge. Animesh Garg points out, “You could potentially spend hundreds of millions of dollars on data capture, but the real question is, even if you do that, will itbe successful?”
Tesla faces the same challenge. Elon Musk previously stated that Tesla would begin small-scale production of truly useful humanoid robots next year for internal use, with plans for mass production and sales to other companies in 2026. However, since the release of the second-generation Optimus last December,Tesla has yet to unveil any significant updates. With the recent releases of Boston Dynamics’ new generation of humanoid robot Atlas and Figure’s new generation of humanoid robot Figure 02, Tesla clearly needs to step up its efforts.
Tesla’s recruitment of “motion capture” workers for Optimus reflects the importance of trainingdata for humanoid robots. As artificial intelligence technology continues to advance, efficiently and cost-effectively acquiring high-quality training data will become a critical challenge in the development of humanoid robots.
【来源】https://www.jiqizhixin.com/articles/2024-08-20-3
Views: 2