中美工业机器人差距与本土制造挑战
在过去一年里，中国安装的机器人数量是美国的九倍。他们安装的机器人比世界其他地区的总和还要多。我们需要迎头赶上，因为机器人不是一种单一的硬件品类，机器人是制造所有其他一切东西的基础品类。
英文原文
In the last year, China has installed nine times more robots than the United States. They install more robots than the rest of the world combined. And that we need to catch up because robots are not a category of hardware. Robots are the category that makes everything else.半导体、药品、国防系统、能源基础设施等等。
英文原文
>> Semiconductors, medicines, defense systems, energy infrastructure, and so much more.我已经向国会作证过两次，分别是在联合经济委员会和众议院科学委员会。
英文原文
>> And I've testified to Congress twice now, the Joint Economic Committee and the House Science Committee.美国在制造业领域已不再具备竞争力，我们每天都在进一步落后。
英文原文
>> The United States is no longer competitive in manufacturing, and we're falling further behind every day.我们是一家位于美国的机器人公司，涵盖硬件、软件以及整个 AI 技术栈的研发与制造。
英文原文
>> And we are a robotics company based here in the United States making the hardware, the software, and the AI stack.我们的机械臂在各类工厂中运行，从小型企业到像洛克希德·马丁（Lockheed Martin）、威瑞森（Verizon）甚至美国国家航空航天局（NASA）这样的财富500强企业。当我们的客户取得成功时，美国也就取得了成功。
英文原文
>> Our robotic arms operate in factories from small businesses to Fortune 500s like Lockheed Martin, Verizon, and even NASA. When our customers succeed, America succeeds.机器人可以做后空翻，但没有人需要一个会后空翻的机器人。制造一个能跑步比赛的机器人很酷，但没有人需要一个会赛跑的机器人。我们专注于为真实的人类解决切实的问题，并且正在快速解决这些问题。
英文原文
>> Robots can do back flips, but nobody needs a robot that does back flips. It's cool to build a robot that can run a race, nobody needs a robot that can run a race. We are focused on solving real problems for real people, and we're solving them fast.我们的增长非常迅速，大约每6个月就要给工厂增加额外的空间。我们正式创立于10年前，
英文原文
>> We are growing so fast that about every 6 months we're adding additional space to the factory. We officially started 10 years ago,但最初的几年实际上都在摸索我们究竟该如何去构建产品。
英文原文
>> but the first couple years were really experimenting with how are we going to build.曾几何时，我的朋友们特意找我严肃沟通了一番，他们说：“你说你想造机器人，你已经做了好几年了，但看起来你根本没有发货任何产品。你真的确定你能做成吗？我觉得是时候换个事情做了。”当你尝试做一件极其困难的事情时，它是需要时间的，而且确实花了很长时间，因为我们不只是在给别人写软件，我们是在尝试从零打造整套东西，自己设计电机和电机控制器。
英文原文
>> There was a time where my friends had a little intervention for me, and they're like, "You said you wanted to build robots, and you've been doing it for years, and you don't seem to have any products that you're shipping. And like, are you sure that this you can do this? Like I think it's time to go do something else." When you're trying to do something hard, it takes time, and it took a really long time cuz we we weren't just writing software for others, we were trying to build the thing from scratch and design our own motors and our own motor controllers.这是大多数公司不敢宣称做到的，因为他们知道每台机器人内部有超过2000个零部件，大概有超过200个子组件。
英文原文
>> And that is something that most companies can't say because they know there's over 2,000 components, there's probably over 200 subassemblies inside of each robot.我们花了7年时间才卖出第一批100台机器人，但之后我们每年都保持着5到10倍的增速，到明年年底，我们预计将占到美国新增部署安装量的10%。
英文原文
>> It took us 7 years to sell our first [music] 100 robots, but we've been growing five to 10 times every year, and by the end of next year we're on track to be 10% of new deployments installed in the United States.我认为，当底层平台发生转移时，你就有机会去开创一家定义时代的公司。一家体量可能堪比苹果或英伟达、且同等重要的公司。
英文原文
>> And I think when the platform shifts, you have a chance to create a generation-defining company. The company that could be the size of an Apple or an Nvidia, and just as important as one of those.这个房间太棒了。
英文原文
>> This room is so badass.是的。你想进去看看吗？我们可以吗？当然可以。
英文原文
>> Yeah. Do you want to go it? May we? Yeah, of course.我的天啊。
英文原文
>> Oh my god.这是一笔超过百万美元的订单，其交付很大程度上归功于大家在那边一直在做的 AI 相关工作。所以这非常重大。这是一个真正起效、且正在从事实际生产工作的产品。因此，感谢参与这个大客户以及全套产品栈的所有人。
英文原文
>> More than a million dollar order that's due in reasonable parts of the AI work that everybody's been working on there. So, this is huge. It's a product that's actually working and doing real work. So, uh thank you to everybody involved across that whole account and across that whole product stack.制造业团队里的每一个人，字面意义上都经历过某种形式的周末加班，工作到很晚。
英文原文
>> Literally every single person in the manufacturing world did some form of weekend, stayed really late.Standard Bots 创业背景与低门槛易用理念
我是 Eben Beard，Standard Bots 的首席执行官。我们是美国本土最大的 AI 原生工业机器人制造商。如今我们的机器人已部署在财富500强企业、汽车、航空航天、石油天然气以及国防等领域，涵盖了非常丰富的应用场景。我们在全美几乎每个州也都有中小型企业客户。
英文原文
>> I'm Eben Beard, CEO of Standard Bots. We are America's largest maker of AI native industrial robots. We're deployed today across Fortune 500s, automotive, aerospace, oil and gas, and defense. So many different use cases. We also have small, medium-sized businesses across almost every state in the country.你为什么选择在纽约而不是硅谷创办一家机器人公司？
英文原文
>> Why did you start a robotics company in New York instead of Silicon Valley?当你观察这片区域的文化时会发现，在硅谷，员工的平均留任期大概只有一年左右，大家频繁在公司之间跳槽。而当你试图做一件极其困难的事情时，你需要人们长期留下来。
我在完全不同的行业工作过。在六年级的某个时候我接触到了网页设计，然后开始通过建网站赚钱。在那个年代并不是每个人都有网站，你可以主动找上门说：“嘿，我帮你们建个网站。”大学毕业后——我在杜克大学学的计算机科学——大学毕业后我参加了 Y Combinator。当时我正在做一款 CRM，我们试图做一款追踪人际关系的个人 CRM。在 Y Combinator 有一位合伙人正在制作大约8英尺高的人形机器人。实际上他们在 Y Combinator 的办公室里拥有一整间机械加工车间，制造各种各样的东西，每天都玩得非常开心。走出 Y Combinator 之后，这位搞机器人的朋友给了我一堆书。他说：“我都是自学的。我大学学的是软件，电气和机械都是自学的。你把这些书读读。”我便开始在夜晚和周末琢磨钻研，正是在那时我意识到，机器人确实是我的人生可选之路。这是一项我可以奉献一生的事业，这也是我真正想做的事情。因此，最初的几年有点像在完成一个自学的博士学位，试图搞清楚如何从零开始制造这些产品。如果你对任何事物怀有巨大的热情，它就会变得简单，你就能学会。关键就在于投入时间。当你热爱你所做的事情时，你就能投入大把时间去学习。
英文原文
>> When you look at the culture of this area, in Silicon Valley the average retention is like something like a year where people are jumping company to company. And when you're trying to do something really hard, you want people to stick around.I was working in this entirely different business. I discovered web design at some point in sixth grade and then I started to make money building websites. Back then not everyone had a website and you could go and say, "Hey, build you a website." After college and I did computer science at Duke and after college I did Y Combinator. I was building a CRM. We're trying to make a personal CRM to track your relationships and at Y Combinator there was a partner who was building these humanoid robots that were like 8 ft tall. They actually had a whole machine shop in the Y Combinator office and they were building all kinds of stuff and having so much fun every day. After Y Combinator, this robotics guy gave me a bunch of books. He's like, "I self-taught. I studied software in college and I self-taught electrical and mechanical. Read those books and I started tinkering kind of at nights and weekends and then and that's where I realized that really the robotics was on the menu. It was something I could spend my life on and that that's how what I wanted to do. So, the early years were a little bit of a self-taught PhD trying to figure out how to make these products, you know, from scratch. If you have a lot of passion for anything that it becomes simple that you can learn. It's just you log the hours. When you like what you're doing, you can log the hours to learn.
你还记得第一位客户是谁吗？
英文原文
>> Do you remember who was the first customer?其实第一批客户都源自早期的需求调研。YC 最深刻灌输进你脑海的原则，就是“做人们想要的东西”。而这一切的核心就在于与客户交谈，真正理解你试图解决的问题。我们与100位客户进行了交流，询问：“关于你们现在的机器人，你们喜欢什么？不喜欢什么？”现在的很多机器人实在是太难用了。它们的交互界面要求你必须死记硬背里面的内容，菜单里全都是单字母缩写，你必须研读说明书才知道怎么操作，这与我们今天习以为常的软件体验相去甚远。意识到这一点后，客户还会说他们根本无法自行调试配置，他们的现场团队只能按启动、停止和复位。他们花了5万美元安装了一台机器人，却除了启停之外什么都做不了。
英文原文
>> So, really the the first customers came from the discovery. My company or really drilled into your head was to make something people want. And that's all about talking to customers, really understanding the problem that you're trying to solve. We spoke to 100 customers. We asked, "What do you like about your robots? What do you not like?" And a lot of the robots today are just too hard to use. They have these interfaces where you have to memorize what's in there. They'll have menus and it's like single letter in the menu. So, you have to study the manual to know what to do. And it's so far from the software we're used to today. And so, realize that and the customers would say they can't set up themselves. Their team can only hit start, stop, and reset. And so, they paid $50,000 for a robot to be installed and they can't do anything but start and stop it.在机器人行业的大多数领域，用户界面的设计极其严谨、极其死板，但同时也极其难用。因此，一旦某个集成商学会了如何对发那科（FANUC）机器人进行编程，再去尝试为另一台机器人编程就会极其痛苦。你实际上必须参加整整一天的培训研讨——准确地说是一个为期3天的研讨会——来试图学习如何编程。公司必须派出三个人参加，而且其中一个人必须在考试中考出90分以上。
英文原文
>> In most of the robot industry, the user interfaces are designed to be very, very rigorous, very, very step-by-step. But they're super hard to use. And so, once some integrator learns how to program a FANUC robot, it's really painful to try and go and program another robot. You actually have to spend a full day workshop It's a 3-day workshop uh trying to learn how to program. Three people from the company have to be there. And one of those people has to score over 90% on their exam.考虑到这一点，我们开发的一项技能——也就是我们之前谈到的——比如“定位”（locate）技能。它能让你的机器人极其轻松地找到零件并把它取走。在过去，如果我想把60个零件逐一移动到某个地方，我必须单独给所有这60个点位进行编程，然后它才会一个接一个地走过去。一旦有任何东西移动了位置，整个流程就会崩溃。因为机器人历史上一直是相当笨拙的，它们只能做一件简单的单调动作，移动到那个固定点位；一旦物品不在那个点上，全盘皆输。我们希望做到的是让任何会用 iPad 的人现在都能使用机器人。拿我自己的亲身经历来说：刚来这里工作时，我以前从未用过机器人，从未操作过一台机器人，也从未在机器人公司工作过。我拿了一台 iPad 开始随手摆弄，自己尝试操作，大约15分钟内我就做出了我的第一个自动化例程。
英文原文
>> So, with that in mind, one of the skills that we've created and what we were talking about before is a idea of a locate skill, for example. Make it really easy for your robot to find a part and go get it. Historically, [music] if I wanted to move 60 pieces one at a time to some place, I would have to program all 60 of those points individually. And then it would they'd go there one at a time. And if anything moved, it would break. Cuz it robots are historically pretty dumb. They could only do one simple thing and they go to that spot. And if it's not in that spot, everything's ruined. We wanted to make it really easy that anybody can use an iPad [music] can now use a robot. I can speak for my own experience. When I first started here, I'd never used a robot before, never operated one, never worked at a robotics company. And [music] I grabbed an iPad and just started messing around with it and I was playing and I made my first routine in about 15 minutes.我大概每天都会问自己这个问题：为什么客户愿意与 Standard Bots 合作？我认为是因为我们改变了美国本土做机器人的方式。传统厂商对客户非常缺乏灵活性，他们的维保计划本质上是在靠客户的痛苦来盈利。所以我总是告诉团队：我们绝不能从客户的痛苦中牟利。客户高兴时，他们自然想买得更多；但如果你是从他们的痛苦中赚钱，他们就不想再跟你合作了。因此我们的目标是确保我们从他们的生产增益中获利。
英文原文
>> I ask myself this question probably every day. Why do you customers want to work with Standard Bots? I think it's we've changed the way in which robotics is done in America. They are very inflexible for their customers and you know, their service programs, they really tout our a for-profit off of your customers pain. So I always tell our team we will not profit off of our customers pain. When customers are happy, you know, they want to buy more. But if you're profiting off of their pain, they don't want to work with you anymore. So our goal is to make sure that we profit off their production.工厂车间与电路板自主制造（SMT）
但我们今天发现自己处于一个可怕的处境。美国在制造业已不再具备竞争力。任何在这里做硬件的公司都可以告诉你，如果你从中国采购而不是在本土生产，成本可以便宜5到10倍。
英文原文
>> But we find ourselves in a dire position today. The United States is no longer competitive in manufacturing. Any company that builds hardware here can tell you that if you source from China instead of making things here, it can be 5 to 10 times cheaper.但是在中国制造难道不会更便宜吗？你们为什么要在本土制造？
英文原文
>> But wouldn't it be cheaper to make them in China? Why are you building them here?在我的职业生涯中，我一直在大规模制造领域和工厂中工作。在我职业生涯早期在谷歌工作时，面对很多代工厂（尤其是中国的代工厂），当时的权衡是：在那些工厂的大多数情况下，人工成本比自动化更低。我认为我们公司最重要的一点在于，我们在美国制造机器人，供美国本土使用。如果我们在海外别的地方制造机器人，自己不掌控生产流程，那么从设计某种产品到交由他人代工，会伴随一整套棘手的问题，特别是在海外存在语言障碍和文化障碍的情况下，会导致项目沟通极其困难。由于我们公司自己运营工厂，我们彻底消除了所有这些问题。
英文原文
>> Throughout my career, I've worked in high volume manufacturing and in factories. The trade-off earlier in my career when I was working at Google with a lot of contract manufacturers, especially in China, was that people were cheaper than automation in most cases in those in those factories. I think it's one of the most important things about our company is that we're building the robots in America to use in America. And if we were building the robots somewhere else outside of the country and we didn't own that production, there's a whole host of problems that come with designing something and then having someone else build it for you. And especially if it's overseas where there's language barriers and and cultural barriers in the way that make it difficult to communicate about projects. Since our company runs our own factory, we remove all of that.欢迎大家，这是我们的工厂车间。我们目前身处大约3万平方英尺的空间内，我们正在将其扩建至7万平方英尺。所有东西都由 Standard Bots 自主设计。
英文原文
>> So welcome. This is our factory floor. We are in currently in about a 30,000 square foot space that we are expanding into 70,000 square foot. Everything is designed by Standard Bots.这个房间太棒了。
英文原文
>> This room is so badass.是的，我们要进去看看吗？
英文原文
>> Yeah, should we go inside it?可以进去吗？
英文原文
>> Yeah, may we?当然可以。
英文原文
>> Yeah, of course.天啊。
英文原文
>> Oh my god.这是我们的表面贴装（SMT）车间。这是 Standard Bots 引入内部自主生产的第一种形态。基本流程是：我们采购这些裸板，也就是还没贴装任何元件的绿色电路板；左边则是采购的元器件料盘。所有这些板卡全部由我们内部设计，并且全部在这里制造。因此，安装到标准产品中的每一块电子板卡都是在这个房间里生产出来的。这些板卡通过传送带送入，走进来后你可以看到贴片机。所有这些单独的料盘卷带都是不同的电子元器件，你可以看到机器正在拾取它们并贴装到板卡上。每分钟都有成百上千个元件被贴装上去，唯有自动化才能实现这种效率。完成后，你可以看到操作员正在进行人工检验，确保所有元件都贴装正确，并手动装配一些无法由这台贴片机处理的较大元件。这步完成后，可以看到元器件板卡从这里的传送带输出。接着它会进入我们的航空光学检测仪（AOI），相机在板卡上的每个元件上方扫描，确保其贴装无误且焊接良好。随后出来的就是成品板卡，可以直接投入我们的生产装配。对于我们的机器人和产品来说，全栈纵向一体化的自主设计无疑是一个巨大的优势，因为 AI 发展得太快了，对机器人功能的需求在不断演进。因此，拥有这种内部自研自制的能力，意味着一旦我们有了设计改动或新想法，一到两天内就能生产出一块板卡，而不必为了拿回板卡测试而苦等六周、七周甚至八周。此外，目前全球各地也存在很多贸易限制，我们在内部全面掌控板卡和电子器件的制造与设计，也让我们能够彻底摆脱未来可能出现的任何限制。
英文原文
>> So this is our SMT room. This is the first modality that Standard Bots has brought in house. So basically what happens is we purchase these bare boards, uh which is a a green circuit board that doesn't have any components placed in it. We purchase these components reels to the left. All of these boards are designed in-house and we manufacture them all here. So, every single electronic board that's gone into a standard product was made in this room. These boards are brought in through a conveyor belt. And as you come in, you can see the pick and place machine. All of these individual rolls are different electronic components and then you can see the machine picking them up and placing them onto the board. These are hundreds and thousands of components that's getting placed every minute. So, only automation would let it happen. After that's done, you can see our operator doing a manual inspection to make sure all the components are placed correctly. As it was putting some of the larger components that is not handled by this robot. So, once that is done, you can see the components rolling off the the conveyor belts here. And then it goes through our aerial optical inspector or AOI. Uh, where camera goes over each of the components onto the board to make sure that it is placed correctly and bonded properly. And then after that, out comes completed boards and these are ready to be used for our manufacturing. The full stack vertical integration of our design for our robots and products is definitely a huge benefit for us because AI is so fast developing. The requirements of what a robot should have or shouldn't have is constantly changing. So, then to have this capability in-house means we're able to pop out a board in a day or two as soon as we have a design change or new new concept in mind instead of waiting six, seven, eight weeks to bring in a board and then test it. There's also a lot of restriction going around around the world also. So, the fact that we have full control over manufacturing and design of our boards and electronics in-house let us be completely free of any restrictions that happen in the future also.垂直整合与制造业回流的时代命题
Ben，你是什么时候开始确立“把机器人带回美国”这一使命的？
英文原文
>> Ben, when did you just start your mission to bring robots back to America?我们当时深入思考了什么将成为定义我们这一代人的核心平台。
英文原文
>> We really thought about what's going to be the defining platform of our generation.一个优秀的平台究竟能发挥什么作用？它能让构建事物变得极其简单。而眼下，对机器人进行编程是极其痛苦的：速度慢、成本高昂，而且每个人都在一次又一次重复解决相同的通用问题。每隔一段时间，底层平台就会发生更替：大型机让位于个人电脑，PC 让位于互联网，互联网让位于移动设备。每一次变革都会涌现出少数几家定义下一代格局的公司。而这一切当下正在发生：软件正在跳出屏幕，正在走入物理现实世界，它需要一种全新的机器来承载运行。这种机器，就是机器人。
英文原文
>> What does a great platform actually do? It makes building really easy. And right now programming a robot is painful. It's slow, it's expensive, and everyone is solving the same problems again and again. Every once in a while, the platform changes. Mainframe computers give way to the personal computer. The PC gives way to the internet. The internet gives way to mobile. And every time a few companies come along and define the next generation. And it's happening right now. Software's leaving the screen. It's moving into the physical world and it needs a new kind of machine to run on. That machine is the robot.当你面对一个具有如此重大意义的事物、也就是下一个核心平台时，事实变得非常清楚：这是一项底层基础技术，我们必须在美国本土将其构建出来。
英文原文
>> And when you're going to have something of that importance, the next major platform, it's becomes clear it's a foundational technology. We have to build it here in the United States.我的父母在1960年代初从古巴来到这里，我父亲开出租车，母亲当秘书。他们非常拼命地工作，打很多份工来存钱，买下了他们的第一套房子，以这种方式逐渐向上打拼。这就是典型的美国梦，对吧？当时这个国家存在这种流动性，一种经济上的流动性，因为社会上有大量的就业岗位可供你去工作挣钱、积累资本，从而尝试在社会阶梯或经济阶层上向上迈进，无论你的目标是什么。然而慢慢地、不可逆转地，随着我们从硬件世界更多转向软件世界，我们把大量这类工作外包转移到了其他国家。
英文原文
>> My parents came here from Cuba early 1960s [music] and my dad was driving a cab, my mom was a secretary, worked so hard and get enough jobs that they saved money, bought their first house and kind of like, you know, made their way up that way. The American dream, right? There was that sort of mobility, like economic mobility in the country because there were a lot of jobs out there to be had that you could work to try to earn that a capital to try to move up the social ladder or whatever or like the economic ladder, whatever you're trying to do. And slowly but surely as we've moved from like more of a software world instead of a hardware world, we've offloaded a lot of those jobs to like other countries.我们看到每年都有制造业岗位流向其他国家。然而制造业正是美国中产阶级的核心所在。面对这种局面，唯一能够扭转这一局面的方法就是提高我们工人的生产效率，而机器人就是实现这一目标的工具。机器人正是我们赢得创造高薪岗位权利、并在美国本土制造产品的途径。
英文原文
>> We see jobs go every year to other countries in manufacturing. And manufacturing is really the heart of our middle class here in the United States. In the face of this, the only thing that can turn the numbers around are making our workers more productive and robots are the tool to do that. And robots are the way that we earn the right to create high-paying jobs and to do it to build things here in the United States.但与此同时，美国的制造业基础——尤其是对机器人这样的产业来说——是非常匮乏的。以低廉的价格、极高的产量和极短的交付周期采购高品质零部件极其困难。所以不可避免地，我们确实发现自己为了某些特定需求不得不求助于海外渠道。我们想要解决这个问题的方法之一就是垂直整合，这意味着机器人制造所涉及的几乎每一种生产形态，我们都希望能实现自主内部搞定。
英文原文
>> But at the same time, the US manufacturing base, especially for something like robotics, is very much lacking. It's very hard to source high-quality parts at low price at high volume with the short lead time. So, inevitably, we do have find ourselves going to uh offshore sites for our certain needs. One of the ways we want to fix this is is vertical integration, meaning almost every single modality of manufacturing that is involved with our robot making, we want to be able to do it in-house.从零研发的阵痛与车库创业历程
但事实确实如此，搞硬件极其艰难。如果你之前从未针对大规模量产进行过产品设计，你会在泥潭里陷得极深之后，才意识到自己犯下的所有错误。
英文原文
>> But it is true, hardware is hard. If you've never designed something for mass manufacturing before, then you'll start get too deep into your project before you realize all the mistakes you made.对工程师来说，对你们而言最困难的部分是什么？
英文原文
>> For the engineers, what was the most difficult part for you?制作第一个原型花了几个星期，但我记得积累到前100个客户花了六七年。有一段时间，我的朋友们特意找我严肃沟通了一番，他们说：“你说过你想造机器人，你已经做了好几年了，可看起来你根本没有发货任何产品。你真的确定你能做到吗？我觉得我们该考虑去做点别的事情了。”这确实花费了极其漫长的时间，因为我们不只是在给别人写写软件，而是试图从头构建一切，自己设计电机和电机控制器。我们在编写固件来对电机进行换相控制，让电磁铁形成正确的电磁作用力，从而以极高的效率运转。刚开始我们的团队非常小，因为我们完全是自筹资金（bootstrap）把公司做起来的。第一台机器人是由我负责软件和电气设计，联合创始人负责机械结构设计。它最初是一款完全由塑料制成的家用机器人，后来我们发现塑料并不是最理想的材料——那还是在我们懂得做有限元分析（FEA）之前的阶段。当时团队里还没有很多机械方面的专业人员，齿轮经常脱齿滑丝，而这就是整个学习历程的起点。我认为你必须进入那种不断学习的迭代循环中去。
英文原文
>> First prototype was a couple weeks, but 100 customers took six or seven years, I think. There was a time where my friends had a little intervention for me and they're like, "You said you wanted to build robots and you've been doing it for years and you don't seem to have any products that you're shipping. And like, are you sure that this you can do this? Like, it's think we'd think it's time to go do something else." It took a really long time cuz we we weren't just writing software for other It was trying to build the thing from scratch and design our own motors and our own motor controllers. We were writing the firmware to commutate the motor to switch it, you know, so the magnet the electromagnet had the right interaction to spin at very high efficiency and we had a very small team at first cuz we really bootstrapped the company. So, first robot, I was doing the software and the electrical design and co-founder was doing the mechanical design. It was originally a home robot entirely made out of plastic and we learned that plastic is not the best for, you know, it's kind of before we'd run FEAs and and really knew how to to run FEAs. We didn't yet have a lot of mechanical people on the team, so gears stripped and that was the start of the learning process. I think you just got to get on that loop where you're learning.我当时在纽约威彻斯特家中的办公室里做所有的原型开发。我有一台整夜运转的 3D 打印机。我白天做设计，晚上把打印任务发到打印机上。那台打印机声音极大，我妻子总抱怨说把它吵醒了，因为那间房就在卧室隔壁。每天都是全新的挑战：我们不知道什么会出错、什么会损坏，或者我们需要修复什么。很多时刻都会让你冒出这种念头：“这会不会让整个项目彻底沉没完蛋？”
英文原文
>> I was doing all the prototyping at home in my office in in Westchester, New York. I had a 3D printer that would run all night. I would work all day on the design, get prints onto the printer at night. The printer was super loud and my wife would complain that, you know, it was waking her up cuz it was the room next to my bedroom. And uh everyday was a new challenge. We didn't know what was going to go wrong, what was going to break, or what we needed to fix. And and there were a lot of moments like that where it was kind of like, "Well, is this going to sink the project or not?"有时候确实很难不去想：“好吧，让我们复盘一下：我们为什么认为这是个好主意？我到底有没有搞错？我们是不是在某种程度上已经脱离现实了？我们是不是应该直接放弃？”有一回，我妻子说她觉得我像是一个在车库里裸身作画的落魄艺术家。当外界用那种眼光看你——看作一个在车库里画画的赤身艺术家时——你真的不能太在乎外界怎么想，否则你不可能坚持做这么久。所以我觉得最难的部分在于，在某些时刻你必须真正直视镜子里的自己，下定决心：“好吧，他们怎么看不重要。我认为这是正确的方向，我们必须坚守下去。”
英文原文
>> Hard at times to think like, "Okay, let's review. Like, why do we think this is a good idea? And like, am I wrong? We're going like we've lost touch with reality in some way here and should we just call it quits?" At one point, my wife said she thought of me as a as a naked artist just sort of painting in the garage. When the world looks at you that way, as a naked artist painting in the garage, you can't really care what it thinks too much. Or you don't want to be doing that very much longer. So, I think that's the hard part is you really have to look at yourself in the mirror at some point and just decide, "Okay, doesn't matter what they think. I think this is the right move. We're going to stay on on this.我们熬过了所有那些困难，如今在现场有数百台机器人在出色地运行。看到公司如何从那个起点成长到如今的规模，真是一件不可思议的事情。
英文原文
>> We made it through all of them and now we have hundreds of those robots in the field that are working beautifully and it's incredible to see how that's gone from there to uh grow the company to where it is now.因为我们倾听客户的声音，清楚自己要打造什么，所以我们总是有排队等候的客户，至今依然如此。我们制造的每一台机器人总是能立刻发货。这是新的前沿办公空间开始的地方，因为我们扩张得太快了。我们扩建的速度已经超过了施工团队装修空间的能力。因此，在空间还没完工之前，我们就不得不把硬件工程测试笼搬到了这里。这是剩下的区域，目前仍在施工中。前部这片大得多的空间将成为我们面向客户的展示区、AI 测试区以及应用工程团队的所在地。
英文原文
>> And because we listened to them and we knew we were going to build, we've always had customers lined up and we still do. We've always shipped every robot we made immediately. This is where the start of what the new [music] front space uh is going to be because we're growing so rapidly. We are outpacing the ability of construction team to fix up the spaces. So, we had to move our hardware engineering test cages into here before we even finished out the space right now. And this is the rest of the area. It's currently under construction. Uh the much bigger space at the front that will become our customer-facing area, our AI testing area, as well as applications team.你看那边，这家伙要平放下来。
英文原文
>> And then you see over there this guy is going to lay flat.RealSense 有 PoE 版本吗？
英文原文
>> Does RealSense have a PoE version?有个短一点的。
英文原文
>> Uh there's a shorter5米肯定不行。
英文原文
>> there's no way five.好的，那个可以用。
英文原文
>> Yeah, that'll be fine.短一点的。拿个短的，好的。
英文原文
>> Shorter. Make it shorter one. Okay.执行力与快节奏迭代
你们是怎样把一家美国机器人公司以如此惊人的速度做大起来的？
英文原文
>> How did you guys scale up on American Robot Company that fast?Evan 具备一种非凡的能力。关于 Evan 我喜欢提起的一点是，他经常会说：“哦，我们需要在一周内搞定这件事。”整个团队都会说这根本不可能。而他会走过来说：“不，不，我们必须在一周内搞定，我们必须在一周内搞定。”我们会说：“不，我们做不到，做不到。”然后一周后我们就真的搞定了。他大概有 80% 的时间是对的。当然，有 20% 的时间他是错的，项目得花上一整年。但 80% 的时间他是对的，这就足够频繁了，以至于我们无法拒绝，因为他会说：“没错，你们能做到，我知道怎么做。”他会以一种我从未在其他领导者身上见过的方式给我们指引方向去实现目标。
英文原文
>> Evan has a unique ability. One thing I like to say about Evan is that uh he often will say, "Oh, we need to get this done in a week." and a whole team will say it's impossible. And he'll come through and say, "No, no, we need to get it done in a week. We need to get it done in a week." and we'll say, "Nope, we can't do it. Can't do it." and then a week later we'll get it done. He's right about 80% of the time. And so uh 20% of the time he's wrong and it takes us a year. But 80% of the time he's right and so he's right often enough that we can't say no because he'll say, "Yeah, you can do it. I know I know how to do it." and he'll he'll give us the direction to get that done in a way that I've never seen a leader do before.今年早些时候我们参加了一个大型展会——Automate。那是我加入公司九周后的事。他当时想在舞台上做现场演示。包括我自己在内的整个团队都告诉他这是个糟糕透顶的主意。结果他是对的。我们熬夜苦战，拼命工作把所有东西准备就绪，最终演示非常成功。
英文原文
>> We had a big trade show earlier this year, Automate. That was uh nine weeks after I joined the company. And uh he wanted to do a live demo on stage. The whole team, including myself, told him that this was a terrible idea. And uh he was right. We stayed up late. We worked our heads off trying to get everything ready, but it landed.Evan 每天都待在工厂里，直到我们攻克那些难题。他在那里连续泡了几个月，把全副精力都倾注进去。这证明了如果你对一件事投入全部精力，并且秉持着“我一定能做到”的心态，你就真的能做到任何事情。
英文原文
>> Evan, he got into the factory every single day until [music] we fixed those challenges. And he was there for months on end. He put his full focus into that and showed that if you put full energy into something and you know you have a belief that you can do it mentality that you can truly do anything.我很好奇，你为什么认为他们能在这么短的时间内完成任务？
英文原文
>> So I wonder, why do you think they can finish the task in such a short time?感谢他们这么评价。我认为无论你在生活中认定自己能达成什么，你就能达成。如果你预先觉得某件事需要一段时间，无论设想多长，它至少会花掉那么多时间。所以我们确实倾向于制定紧凑的时间表，深入思考无论在尝试做什么，达成该目标的最高效路径是什么。我认为对于完成任何有野心的事情而言，关键在于你相信自己能做到，而且相信能在合理的时间节点内完成。如果觉得做不到，你就要拆解问题、简化问题，从而让你能在合理的时间内搞定它。
英文原文
>> That's nice of them to say. I think that you set out whatever you think you can achieve in life, you do. And if you think that something's going to take some period of time, it will take at least that amount of time, no matter how long. So we do try to set tight timelines and really think through whatever we're trying to do, what's the fastest way to accomplish that thing. I mean, it's key to really doing anything ambitious is you believe you can do it and you believe you can do it in a reasonable timeline. And if not, you've kind of broken down the problem and or simplified the problem so that you you can do it in a reasonable amount of time.人形机器人的局限与机械臂的工业可靠性
那么目前真正的技术挑战是什么？如果一个演示成功了，是否就意味着它是一个好产品？
英文原文
>> Then, what's the real technical challenge right now? If a demo works, does that mean it's a good product?我在机器人行业工作了15年以上。你会发现客户通常分为两大阵营：一部分人坚信机器人无所不能，另一部分人则认为机器人的作用被严重高估了。而现实介于两者之间。目前的技术能力正在极其迅猛地向前发展，我们做到了几年前完全不可能的事情。但与此同时，仍有大量事物是机器人目前无法可靠完成、或者完成速度不够快的。所以你看 YouTube 上的视频会发现很多都是加速播放的。你可以留心机器人放下物品时的速度：如果物体掉落的加速度超出了重力加速度，你就知道做视频的人在造假作弊了。
英文原文
>> I've been working in the robot industry for over 15 years now. You see customers that fall into two camps. There's the people that believe that robots can do everything and there's the people that believe that robots are totally under overrated. And reality is somewhere in between. The capability is advancing extremely quickly now. We got stuff that was totally impossible [music] just a couple years ago. At the same time, there's a lot of stuff robots can't yet do reliably and can't yet do quickly enough. And so you see YouTube videos and see that they're sped up. You look for the how quickly a robot drops an item. If the item falls with super gravitational acceleration, you know the people making the video are cheating.那么你目前如何看待人形机器人？为什么你们更专注于机械臂？
英文原文
>> So how do you think about the humanoid robots right now? Why do you want to focus more on the robot arm?当人们看到长得像人的机器人时，他们的想象是它能做人类所做的一切。但在行业中仍有一些核心难题远未得到普遍解决。
英文原文
>> When people see a robot that looks like a person, their imagination is that it does what a person does. And there's still some problems that haven't been widely solved in the industry.我们这里有一套理论，称之为“小鬼当家评估法（Home Alone eval）”：如果你要研发一个人形机器人，它必须通过“小鬼当家测试”——也就是当它上下楼梯时，几个五岁或八岁的小孩试图绊倒它或把它掀翻。如果它经受不住这种考验，那它就进不了家庭环境。因为如果有几百磅重的金属砸在你身上，显然可能会砸断你的腿甚至造成更严重的后果。所以我们认为双足人形机器人并不是答案。这就是为什么我们专注于机械臂和轮式底盘。
英文原文
>> We have a theory here, which is uh this is our Home Alone eval, which is that if you're going to have a humanoid, it needs to survive the Home Alone test, which is a couple of five or eight-year-olds trying to trip it and flip it as it goes up and down the stairs. And if it can't survive that, then it can't go in the home. Cuz if you have a couple hundred pounds of metal falling on you, um obviously could break your leg or worse. And so so we don't think legged humanoids are the answer. That's why we're uh you know, we focus on our arms and wheel bases.可靠性是我们行业的关键命题之一。我认为这是目前处于机器人能力前沿的基础模型公司，与指望将机器人作为可靠队友的真实工业用户之间最大的鸿沟之一。研究人员喜欢关注成功率从 30% 提升到 80% 的进展，因为很容易收集到足够的试验次数来达到统计学上的显著性。然而对于工业应用而言，如果我有一套只能在 70% 的时间里正常工作的系统，那它就是一个糟糕透顶的队友。工业级应用通常要求达到 99% 乃至 99.9% 的成功率，我认为学术研究界在这一领域是落后于当前真实需求的。我们在泛化性方面做了许多了不起的工作，但我们必须提升系统的可靠性。目前最优质的数据来自现场实际运行，我想大家都认同这一点。如果你想要获取优质的现场数据，你就必须拥有深入现场的铺货分发渠道。我们的机器人已经部署在现场，这赋予了我们独特的竞争优势，能够获取合适的数据去把机器人的可靠性训练到极致。
英文原文
>> Reliability is one of the critical things for our industry. I think it's one of the biggest gaps between the foundation model companies at the current frontier of where robot capabilities live and the real industrial users that are trying to count on the robots as a teammate. Researchers like to focus on the progress you can make from 30% to 80% success rates because it's easy to get enough trials to be statistically significant. For industrial use though, if I have something that only works 70% of the time, that's a terrible teammate. Industrial applications are typically targeting 99, 99.9% success rates and I think this is an active area where the research field is behind where the current real needs are. We've done a bunch of amazing work to generalize, but we need to increase the reliability of systems. The best data comes from the field right now and I think everybody agrees on that. If you want to have good field data, you need to have distribution to be in the field. Our robots are in the field, they give us this unique edge on the right data to train robots to full reliability.数据飞轮与示教交互工具
这是 Echo 设备，是我负责研发的数据采集设备。到目前为止，我们已经和数据采集员一起对它进行了几个月的迭代。它就是一个手持夹爪，上面装有摄像头以便采集数据。你可以去抓取物品、到处移动它们，从摄像头的视角来看，它和机械臂夹爪的视角完全一样。因此，当你把摄像头安装到机器人上时，数据采集设备上的摄像头视角与机器人上的摄像头视角之间不存在任何领域差距（domain gap）。这样你就可以直接用这些视频来进行模型训练。
英文原文
>> So, this is the the Echo device, which is the data collection device that I worked on. We've iterated on this for months at this point with our with our data collectors. It's, you know, just a little handheld gripper. It'll have a camera mounted so that you can collect data with it. You can go, you know, grab things, move them around and it looks exactly [music] like a robotic gripper from the perspective of the camera. So, when you go to put the camera on the robot, there is no domain gap between what the camera on the data collection device sees and what the camera on the robot sees. So, you can use the videos to to train.作为人类，我们自然知道从哪里去抓取，但我们的 AI 模型显然没有这种先验常识。
英文原文
>> As humans, we kind of naturally know like to pick from here, but of course our AI models don't have that kind of knowledge.所以我们必须用训练数据来教导它，这就是她现在正在做的事情——从不同的方位朝向进行抓取演示。
英文原文
>> So, we have to teach it with our training data and that's what she's doing right now, picking it from different orientation.正如我所说，我们对这款设备迭代了几个月，使其既实用又更符合人体工学。最初我们把原型机交给数据采集员时，他们只用 20 分钟手就会开始酸痛；现在他们可以使用几个小时。基本上可以一整天持续采集数据，中途只需要安排几次短暂休息，这有助于我们加速推动数据采集飞轮。
英文原文
>> We've iterated on this device, like I said, for months to make it functional and more ergonomic. Originally, when we first gave a prototype to the data collectors, they can only use it for 20 minutes before their hands started to hurt. Now, we're up to a few hours. So, you know, you can use it for basically a full day with some breaks built in without stopping to collect data, which will help us ramp up that data collection flywheel.我们有多种不同的任务示范方式，这是其中之一，叫做“轨迹示范（show path）”。你可以向机器人示范一项任务，它就会学会如何操作。核心理念就是你拖动它运动，它就能理解并转化为自己的动作去抓取。训练这样一个动作就是这么快。教一个视觉例程很简单，我们展示了如何训练视觉；但要训练一套端到端的 AI 例程去抓取晃动的挂钩则需要更进一步。目前在美国，制造业面临严重的招工难，很难招到优秀的员工。未来几年工人缺口将达数百万人。要找到伴随你规模化扩张所需的制造业人才极其困难。
英文原文
>> We have a bunch of different ways to demonstrate a task. This is one of them. This is called show path. You can demonstrate a task to the robot and it learns how to do The whole idea is that you can move it and it'll be able to translate what it's doing and [music] grab. So, it's that quick to train something like this. To train a vision routine, we showed you how to train, you know, vision, but to train an end-to-end AI routine that can grab hooks that are moving around. In the United States right now, they struggle with labor. It's hard to find good people. There's uh the shortage of millions of workers over the coming years. It's very difficult to find the manufacturing talent you need to to scale with you.我来自机械加工车间背景。我们面临的现状是大量技术熟练的工匠正在逐渐老化并退休。因此我们正在尝试做的事情之一，就是利用好这部分最后存留的技术工人资源，推动他们转向从事更多技能密集型的高阶任务。
英文原文
>> I came from a background of machine shops. And what we have is we have a lot of skilled tradesmen [music] that are aging out and retiring. So, one of the things that we're trying to do is utilize the last resources of skilled labor and push them to more skill-intensive tasks.为普通工人设计：人机协作而非完全替代
我认为为普通大众设计产品至关重要。如果我们回顾工业革命，那场革命完全围绕着自动化纺织设备展开。大英帝国之所以主导了整个世界，是因为他们生产纺织品的成本极低且效率远高于其他人。但我们在那个时期也看到，很多靠手工织布的人无法操作这种自动化的纺织机器，整整一个群体被时代抛在了身后。
因此在我们看来，这不仅是满足客户的需求——因为这确实是他们的诉求，他们希望自己的员工开箱即用就能操作机器人；这也不仅是满足投资人的诉求——因为这确实具备可规模化性；更重要的是，从历史的视角来看，这是正确的事情，因为我们不想抛弃任何人。我们希望现有的工人能够转型到这些薪酬更高、更安全的工作岗位上去管理这些机器人。机器人是21世纪的强力电动工具，旨在帮助他们更好地完成工作。
英文原文
>> I think it's really important to design for ordinary people. And if we look at the Industrial Revolution, of course, that was all about automated textile making equipment. And Great Britain or England kind of dominated the whole world cuz they can make textiles for so much for much less and more efficiently than everyone else. And But, what we saw at that time was there a lot of people The people that were making the textiles by hand could not operate the automated textile making equipment. And there was a whole group of folks that were left behind.And so, we view it as not only is it what customers want cuz it is what they want. They want their team to be able to unbox a robot. And not only is it investors want cuz it is what they want. It has to be scalable. But, it's also the right thing to do if we look through the lens of history cuz we don't want to leave anyone behind. We want the existing workers to be able to shift into these higher-paying, safer jobs, manage these robots. They're the power tools of the 21st century designed to to help them do their job.
如果你这么说的话，未来工厂里就根本不需要人类了。你觉得这是件好事吗？
英文原文
>> If you pronounce it like that, there will be no human at all at the factory. Do you think it's a good thing?在 Standard Bots，我们的使命是用世界上最普及易用的机器人拓展人类的能力边界。这意味着打造任何人都能购买和部署的机器人，并通过 AI 让它们变得更加强大，从而去自动化那些以往根本无法自动化的任务。
英文原文
>> At Standard Bots, our mission is to expand human capacity with the world's most accessible robots. And that means making robots that anyone can buy and deploy and making them more capable with AI. So, you can automate tasks you just couldn't automate before.我们常听客户反映招聘十分困难，这正是机器人的用武之地。但不同于你自己亲自去焊接，你可以管理六台焊接机器人。这实际上是一份更好的工作，可以是一份薪水更高的工作，也可以是一份让你免受焊接烟尘侵害的更安全的工作——机器人代替你去面对焊接烟尘。它就像 R2-D2 一样，并不是在取代人类，而是让人们更具生产力。
英文原文
>> We hear from customers that it's very hard to hire, and that's where the robot comes in. But, instead of welding yourself, you can manage six welding robots. That is actually a better job. It can be a higher paying job. Can be one where you're not in the weld smoke, a safer job. The robots are in the weld smoke. It's like the R2-D2. They are not replacing the humans, they're making the humans more productive.我们的平台在硬件层面的运行逻辑与软件层面完全一致，就像工业乐高积木一样。你可以为具体应用搭建最理想的形态——可能是移动底座上装载六条机械臂，这将为客户带来最佳的投资回报率。
英文原文
>> Our platform on the hardware side works the same way on the software side. It's like industrial Legos. You can have the ideal form factor for your application. It might be six arms on a mobile base. And that's going to give you as a customer the best ROI.十年磨一剑与长期愿景
走过10年之后，你认为这里的成功几率有多大？
英文原文
>> 10 years in, what are the odds of success here?我认为我们能成功的几率从来都是极小的。想要去打造世界上最大的机器人公司，难度就像成为一名 NBA 球员一样，极其困难。但我们对走到今天所取得的成就感到非常自豪。回顾现状，我们已经化解了许多环节的风险：在各种领域部署了产品，几乎走进了每一个州，落地了成百上千个应用场景。因此我认为成功的几率每天都在改善。但要实现打造一家体量堪比英伟达或苹果的公司的抱负，胜算始终是微小的。
而我们能做的只有竭尽所能。归根结底，我们在地球上的时间其实非常短暂。如果我们能以某种方式回馈世界，我认为那就是人生的最高价值与最佳用途。这就像是你给世界献上的一份礼物，而我唯一的特长就是制造机器人。
英文原文
>> I think the odds that we're successful, they've It's always been long. It's It's like becoming an NBA player, something like that, to try to start the biggest robotics company in the world. And it's really hard. But we're so proud of where we've come. I think where we are today, we've de-risked many things. Deployed 400 and all these different areas. We're deployed in almost every state. We've got hundreds of use cases deployed. So, I think the odds are looking better every day. But I think the size of our ambition to build an Nvidia or Apple-size company, to fulfill that, it's always long odds.And the best we can do is our best. You know, I mean, at the end of the day, we're here on Earth for really a short time. And if we can give back in some way, then I think that's the the highest and best use. It's like you're giving some kind of gift to the world. And my unique talent is making robots.