AI News AI资讯 17h ago Updated 15h ago 更新于 15小时前 45

Reimagine Robotics Emerges from Stealth With Robots That Learn From Workers on the Job Reimagine Robotics 以机器人向工人学习技术现身

Reimagine Robotics emerged from stealth with a "monkey-see, monkey-do" platform enabling factory workers to directly teach and correct robots, eliminating dependency on specialist programmers for task reconfiguration The technology has already been deployed in manufacturing and electronics disassembly, cutting robot behavior prototyping time from approximately one day to roughly 10 minutes Founded by former Google DeepMind Applied Robotics leaders including CEO Jonathan Scholz, the company secur Reimagine Robotics 从隐形状态中亮相,推出允许工厂工人直接教授和纠正机器人的技术,减少对专业程序员的依赖 由前 Google DeepMind 应用机器人团队负责人 Jonathan Scholz 于 2025 年 4 月创立,获 Fly Ventures、firstminute capital 等 Pre-seed 投资 在制造和电子拆解场景中,将新机器人行为的原型开发时间从约 1 天缩短至约 10 分钟 公司采用"monkey-see, monkey-do"方法,强调人机协作而非替代工人 已在定制塑料制造和硬盘拆解回收项目中部署,工人可自行扩展自动化流程

65
Hot 热度
60
Quality 质量
65
Impact 影响力

Analysis 深度分析

TL;DR

  • Reimagine Robotics emerged from stealth with a "monkey-see, monkey-do" platform enabling factory workers to directly teach and correct robots, eliminating dependency on specialist programmers for task reconfiguration
  • The technology has already been deployed in manufacturing and electronics disassembly, cutting robot behavior prototyping time from approximately one day to roughly 10 minutes
  • Founded by former Google DeepMind Applied Robotics leaders including CEO Jonathan Scholz, the company secured pre-seed funding from Fly Ventures and firstminute capital, and is now seeking additional investment
  • The platform targets factories with fluid, variable workflows where traditional automation is economically unviable due to low production volumes or frequent process changes
  • A core philosophical differentiator: the system is designed to keep humans in the loop, with workers identifying bottlenecks, demonstrating tasks, and iteratively correcting robots until they function correctly

Why It Matters

This represents a significant shift in industrial robotics toward democratized, worker-driven automation that could unlock robotic deployment in small-batch and highly variable manufacturing environments previously considered uneconomical to automate. For AI practitioners and robotics engineers, it demonstrates the practical application of on-the-job learning paradigms and human-in-the-loop correction systems at industrial scale.

Technical Details

  • The platform uses a direct teach-and-correct interaction model where workers physically demonstrate tasks to robots and provide real-time corrections, enabling rapid reprogramming without specialist coding expertise
  • Deployments include a made-to-order plastics manufacturer where robots were trained overnight to operate 3D printers (removing print beds, operating latches, pressing controls), with workers later extending automation to washing, curing, and drying steps independently
  • A three-robot system for electronics disassembly and hard drive recovery demonstrated a 144x reduction in prototyping time (from ~1 day to ~10 minutes) for developing and testing new robot behaviors
  • The system is designed for iterative human-robot collaboration where workflow adjustments and improvements occur concurrently rather than requiring full reprogramming cycles
  • Founded in April 2025 by Jonathan Scholz (former head of Google DeepMind Applied Robotics for seven years), Oleg Sushkov, Akhil Raju, and Misha Denil, with dual headquarters in London and Sydney

Industry Insight

  • The "monkey-see, monkey-do" approach signals a broader industry trend toward lowering the barrier to robotics adoption, potentially opening millions of small and medium manufacturing facilities to automation that was previously inaccessible due to programming costs and specialist dependency
  • The dramatic reduction in prototyping time (day-to-10-minutes) suggests that human-in-the-loop learning platforms could become a competitive differentiator for robotics companies targeting flexible, high-mix production environments
  • The emphasis on keeping workers in the process rather than replacing them aligns with growing labor market realities and regulatory pressures, positioning this technology as a collaborative augmentation tool rather than a displacement solution

TL;DR

  • Reimagine Robotics 从隐形状态中亮相,推出允许工厂工人直接教授和纠正机器人的技术,减少对专业程序员的依赖
  • 由前 Google DeepMind 应用机器人团队负责人 Jonathan Scholz 于 2025 年 4 月创立,获 Fly Ventures、firstminute capital 等 Pre-seed 投资
  • 在制造和电子拆解场景中,将新机器人行为的原型开发时间从约 1 天缩短至约 10 分钟
  • 公司采用"monkey-see, monkey-do"方法,强调人机协作而非替代工人
  • 已在定制塑料制造和硬盘拆解回收项目中部署,工人可自行扩展自动化流程

为什么值得看

Reimagine Robotics 代表了工业自动化的一个重要范式转变:从"专家编程→部署"转向"工人示范→即时纠正",这直接解决了中小批量、高变体生产场景的自动化瓶颈。对于 AI 从业者而言,其"人在回路"(human-in-the-loop)的实时纠错机制为具身智能(embodied AI)的落地提供了可复用的工程路径。

技术解析

  • 交互范式:工人通过物理示范(showing a task)和实时纠正(put it right when it makes a mistake)训练机器人,无需编写代码。系统支持多轮迭代,工人可逐步完善行为直至可用。
  • 性能指标:在硬盘拆解项目中,新行为开发与测试时间从 ~1 天降至 ~10 分钟,效率提升约 36 倍。原型迭代速度的质变是核心技术竞争力的直接体现。
  • 部署场景:已落地于两类场景——(1)定制塑料制造:机器人夜间操作 3D 打印机,移除打印床、操作卡扣和按钮;(2)电子拆解:三机器人协作拆解旧硬盘回收有价值材料。工人后续自主扩展了清洗、固化、干燥等流程。
  • 团队背景:创始人 Jonathan Scholz 曾领导 Google DeepMind 伦敦应用机器人团队 7 年,联合创始人 Oleg Sushkov、Akhil Raju、Misha Denil 均为 DeepMind 前成员。技术基因来自 DeepMind 在强化学习和机器人学习领域的积累。
  • 融资状态:Pre-seed 轮由 Fly Ventures、firstminute capital 及天使投资人支持,目前正寻求新一轮融资以扩展团队和部署规模,并验证"经验复用"假设(每次部署经验使后续项目更快更可靠)。

行业启示

  • 工业自动化的民主化趋势:传统自动化依赖专业工程师重新编程,Reimagine 的模式将自动化能力下沉至一线工人,适合"产量太低或流程变体太多而无法证明传统自动化合理性"的工厂,填补了刚性自动化与纯人工之间的空白地带。
  • 具身智能落地的工程路径:其"人在回路实时纠正"方法为具身智能提供了可规模化的部署框架——不追求通用 AGI,而是聚焦特定工业场景的渐进式学习,这对 AI 公司的商业化策略具有参考价值。
  • 人机协作而非替代的叙事:Scholz 明确强调"这不是要把人移出流程",机器人依赖工人识别瓶颈、示范和纠正。这一立场在劳动力焦虑敏感的制造业环境中更具接受度,也为政策制定者和企业管理者提供了更温和的自动化转型路径。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Robotics 机器人 Funding 融资 Product Launch 产品发布 Deployment 部署