设计友好的人形机器人：从科幻到柔和互动

这是 Bernt Børnich。

This is Bernt Børnich.

对，差不多。

Yeah, close enough.

差不多吧。所以你知道你基本上是在设计一个类人机器人，让人感觉舒适、愉快。我很好奇，从你的背景来看，这对你现在设计产品的思路有影响吗？

Something like that. So you know that you're basically trying to design something that is like a humanoid that's comfortable and pleasant to be around. And I'm kind of wondering from your background, did that have any influence on how you're thinking about designing the product now?

也许吧。我是说，斯堪的纳维亚以优美的设计闻名，对吧？

Maybe. I mean, Scandinavia is known for beautiful design, right?

对。

Yeah.

我觉得这很大程度上也跟我从小对这个领域充满热情、一直思考它有关。我小时候就想造人形机器人。读了很多科幻小说，看了很多电影，然后意识到那种硬核黑暗的科幻未来并不是大家想要的。这就像作弊码，想让东西看起来很酷，就做成黑暗科幻。但那并不是人们真正想要的。

I think a lot of it also has to do with just growing up being so passionate about the space and thinking about it always since I was a kid. All I wanted to do was build humanoid robots. Reading a lot of the sci-fi, watching a lot of the movies, and kind of realizing that this hard dark sci-fi future is not what everyone wants. It's like a cheat code. If you want something to look really cool, just make it dark sci-fi. But that's not really what people want.

就像贩卖争议。

Like selling controversy.

对，就是这样。我们想要的是亲切、柔和、融入我们空间的东西，重新定义我们与这些人工实体或智能体的交流方式。

Yeah, it is. And it's like we want something that's relatable and soft, part of our space. Something that kind of redefines how we communicate with these artificial entities or intelligences.

那在日常工作中，当你思考如何设计这个东西时，实际上是怎么做的？

And in your day-to-day, when you're thinking about how to design this thing, what does that actually look like in practice?

这主要是很多会议，努力确保整个产品的一致性。我身兼多职，但其中一个是产品愿景官，这很大程度上就是不停地说：‘不，不，不，这不行。我知道很难，但我们不能那样做。我们必须找到更好的解决方案。’并且在这些核心原则上不妥协，对吧？机器人必须安全、必须真正有能力、不是玩具，而且必须非常实惠，才能真正达到我们需要的覆盖范围。还要与产品共同生活，确保我每天都和产品在一起，在自己家里使用它，观察它的表现。因为归根结底，组织的健康状况总是体现在产品上。然后你每天尽量多地跨级沟通，以便更早发现问题，这样你就不会坐在他们客厅里，让 Neo 开门，然后他们惊呼：‘天哪……’

It's a lot of meetings and trying to ensure consistency across the entire product. I wear many hats, but one of them is product visionary, and that's a lot of just saying, 'No, no, no, that doesn't work. I get it's hard, but we can't do that. We have to find a better solution.' And not making compromises on these core principles, right? The robot has to be safe, has to be actually capable, is not a toy, and it needs to be very affordable to actually have the reach that we need. And just living with the product, ensuring that I'm with the product every day, using the product in my own home, and just seeing the output. Because in the end, the health of your organization is always seen in the product. And then you're trying to skip level as much as you can every single day to catch things earlier than that, so you're not sitting in their living room and asking Neo to open the door and they're like, 'Holy...'

这里发生了什么？

What happened here?

比如，‘为什么玻璃碎了？’对。所以这就是很多日常工作。然后还有很多工作是因为我们想确保有非常热情的早期采用者，愿意和我们一起踏上这段旅程。所以我觉得把信息传递出去非常重要。我们对世界可能的样子有相当新颖但也非常友好的看法，我们需要与人分享，以便建立追随者，帮助我们实现这一目标。当然，还要确保我们有资金支付所有人的工资，然后平衡所有这些。但我的心和热情属于工程和产品，对吧？这是我的出发点。

Like, 'Why is the glass broken?' Yeah. So that's a lot of the day-to-day. And then there's a lot of this because we want to make sure we have very passionate early adopters that want to be on this journey together with us. So I think it's quite important to get the messaging out there. I think we have a quite novel but also very friendly view on how the world can be, and we need to share it with people so we can build that following that will help us make this a reality. And of course, ensuring we have money to pay all these people, and then balancing all of these. But my heart and passion belongs in the engineering and the product, right? That's where I'm coming from.

有没有一些时刻，当你真的把它放在家里，体验到一些你没想到会如此令人愉悦的事情？

Has there been any moments where you actually have it in your home, where you kind of experience some delightful thing that you would not have expected to be a delight?

很多次。很多小事情都充满了魔力，我没想到会那么神奇。比如坐在沙发上看书，有人敲门，我说：‘嘿，Neo，你能去开门吗？’机器人走过去，打开门，向客人打招呼，对吧？完全自主，绝对神奇。

A lot of times. There's been a lot of magic around small things that I didn't expect to be that magical. Like sitting on the couch reading a book, someone knocks on my door and says, 'Hey, Neo, can you get that?' Rob goes over, opens the door, and says hi to the guest, right? It's fully autonomous, and it's absolutely magical.

是否也有部分魔力在于，对体验它的另一个人来说也是神奇的？

Is it also partially magical that it's magical for the other person that's experiencing?

这是一种共享体验，对吧？有些事情就是这样。第一次机器人坐在我旁边的沙发上，我们和机器人交谈，它坐在我旁边，我突然意识到这对话和站着时有多么不同。它完全改变了整个互动的视角。我心想：‘天哪，这太不一样了。’

It's a shared experience, right? There are some things like that. The first time the robot sat down next to me on the couch and we were having a conversation with the robot, and it's sitting down next to me, and it just dawns on me how incredibly different the conversation is from standing. It completely shifts the perspective of the entire interaction. I'm like, 'Holy shit.' That's so different.

当它站着时，你只是向它要东西，而当它……

When it's standing, you're just asking it for stuff, and when it's...

不，但我们可以站着交谈，对吧？就像我们现在这样。如果我们站着谈话，然后坐下来继续，对话会有所转变，视角也会转变，感觉不同。我们和 AI 交谈时不习惯这样。我们通常对着手机说话，对吧？所以有一些显而易见的事情，比如你在和一个有具身的对象说话，它有肢体语言，能看着你，知道谁在说话，所有这些。但这些是显而易见的，大家都知道，而且效果很好，很神奇。但当你并排坐着而不是站着时，那种感知上的转变，我从未注意到，直到和机器人一起发生，因为我从没和机器人这样做过。现在我开始注意到和人之间也有这种变化，因为在机器人之前，我从未关注过这一点。这就像一种灵魂出窍的体验，因为我从未见过。所以那是个相当大的惊喜。

No, but like no, no, we can stand and have a conversation, right? Like we are having a conversation now. If we were standing having this conversation, then we sit down and continue. The conversation kind of shifts. The perspective shifts. It's different. And we're not used to this when we're talking to an AI. We talk to our phone or whatever, right? So there's the obvious things like okay you're talking to something that has an embodiment, so it has body language, it can look at you, it can know who's talking, all these things. But that's obvious, like everyone knows that, and that works really well and it's magical. But the shift in perception once you sit next to each other instead of standing, I've never noticed it until it happened with a robot, because I'd never done it with a robot before. And now I start noticing it with people, because before it happened with the robot, I never called attention to it. It's like an out-of-body experience because I've never seen it before. So that was a pretty big surprise.

你有没有注意到人们会形成类似的依恋？你知道，在 Roomba 出现之前，大多数人不会想到会对机器人产生家庭般的依恋或像对动物那样的依恋。但你有注意到当有人家里有 Neo 时会发生这种情况或类似的情况吗？

Do you notice that people form similar attachments? You know, before Roombas became a thing, most people wouldn't expect to form like a really familial attachment or like an animal attachment to a robot. But have you noticed that kind of happen or something similar to that happen when someone has a Neo in their home?

是的，是的。而且在 Neo 之前很久，对吧，我们在 2019 年有过一次美好的经历。我想是的。当时我们之前的人形机器人 Eve，我们在医疗保健领域做了一些试验。我们有一个老年护理中心，有人使用机器人一段时间，我们想修理它，我们说：‘我们直接换一个机器人吧，因为修理比换一个更麻烦。’

Yeah. Yeah. And way before Neo, right, we had this beautiful experience back in, I think, 2019. Yeah, I think so. Where our previous humanoid Eve, and we were doing some trials in healthcare. So we had an elderly care center, and we had someone have the robot for a while, and we wanted to fix it, and we're like, 'We're just going to change the robot because it's easier than fixing it.'

对。

Yeah.

我们说：‘不，我们就拿走这个，你会得到一个新的。’结果对方说：‘不，不，不，你不能拿走我的机器人。我不要新的，我就要这个。’

Like, 'No, we're just going to take this one. You're going to get a new one.' It's like, 'No, no, no, you can't take my robot. I don't want a new one. I want this one.'

这就像说不可能有更好的人类，那只是弗雷德。

It's like saying you can't have a better human. It's just Fred.

是啊，你不能就这么说，那行不通。

Yeah. You can't just do that. That doesn't work.

很多这类事情，人们都会拟人化，对吧？我觉得这是我们的天性，而且很重要。这里有一个完全不同的分支话题，关于社会背景在多大程度上驱动认知发展，我觉得这很有意思。

A lot of these things, people anthropomorphize, right? And I think it's in our nature and it's quite important. There's a completely different tangent here on how much the social context actually drives cognitive development, which I think is quite interesting.

你想深入聊聊这个吗？

Do you want to go down that?

是的。但首先，你说宠物，我觉得很有意思，因为 Neo 是什么？它不是宠物，也不是人类，而是别的东西，对吧？是介于两者之间的某种存在。就像一个可以陪伴一生的伙伴，知道你的好、坏、丑，记住你的全部历史，也许是整个家庭的历史，并且永远站在你这边。

Yes. But first, just to say that you say pet, and I think it's quite interesting because what is Neo? It's not a pet, but it's not a human. It's something else, right? It's something in between. It's like a companion that can be there throughout life and know the good, the bad, and the ugly, and remember your entire history, maybe shared family history, and always be on your side.

理解你如何应对。

Understand how you deal.

是的，我觉得这对我来说就是魔法。我最接近的类比是，你知道，《凯文和霍布斯》里的霍布斯，那只玩具老虎。但这里也有很多未解的问题。我们不知道这会如何发展，但一定会很有趣、很有意思。

Yeah, I think that to me is just magical. The closest I come to thinking about it is like, you know, in Calvin and Hobbes, it's kind of like my Hobbes, the toy tiger. But there's also a lot of unanswered questions here. We don't know how this will pan out, but it's going to be a lot of fun and really interesting.

在认知智能的社会性方面，很多智能的发展只发生在共享栖息地、共同生活的物种中。独居物种没有这种类型的认知智能。而且，所有事物的背景，不只是我们进行的对话，还包括像操作和实际任务这样的简单事情，都会随着社会背景而变化。一个简单的例子是：我面前的咖啡杯是脏了，还是需要续杯，还是我要一整天都留着它？这很难判断，它有很深的社会背景。

On the social side of cognitive intelligence, there's a lot of development in intelligence that only happens to species that share habitat and live together. Solitary species do not have this same type of cognitive intelligence. Also, the context of everything, not just the conversations we're having, but simple things like manipulation and practical tasks, changes based on social context. A simple favorite example is: is the coffee cup in front of me dirty, or do I need a refill, or do I keep it throughout the day? It's very hard to know. It has a very deep social context.

而且它是你的还是别人的？

And is it yours or someone else's?

通常我一整天都留着它，所以别拿我的咖啡杯。对，就是这些生活中的细微差别。我真的很兴奋能开始捕捉人类生活的这些细微差别，以及它们如何帮助我们构建更好的智能。

Usually I keep it throughout the day, so don't take my coffee cup. Right. It's all these nuances of life. I'm really excited to start capturing all these nuances of human life and how they can help us build better intelligence.

在设计今天的机器人时，你是否需要考虑，如果要把它们放进人们的家里，人们会与它们形成情感纽带和依恋，你是否需要思考未来多次迭代，我想要这个东西长什么样？我想让它成为什么？至少是骨架，这样如果有人想要升级，不会感觉你完全掏空了这个东西的大脑。

Do you have to think through, like, in designing the robots today, if you're going to be putting them in people's homes and people are going to be forming these emotional bonds and attachments to them, do you have to think through many iterations into the future of what do I want this thing to look like? What do I want it to be? At least the skeleton, so that if someone ever wants an upgrade, it doesn't feel like you just completely ripped out the brain of this thing.

这变得很奇怪，因为有一个古老的哲学问题：我可以换多少次斧头和斧柄，它仍然是同一把斧头？我们真的在机器人上遇到了这个问题。我们定义了头部是机器人的名字和序列号。所以你可以改变除头部以外的一切，它仍然是同一个机器人。但实际上，有时会发生问题，你有完整的机器人，但你只换了头部，它仍然被认为是同一个机器人，即使你换了头部，现在就是一片混乱，对吧？因为这只是内部服务之类的。就像，不，那是机器人吗？就像，不，不。所以我们最终有了一个新概念，我们称之为量子序列号，即机器人的量子意识，它是虚拟的。无论你改变机器人上的多少东西，它仍然是同一个机器人。不确定我们的用户是否会满意，但这是我们目前最好的办法。

It gets weird because there's this old philosophical problem of how many times can I change the head and the shaft of the axe and it's still the same axe? And we literally have this problem with the robot. We defined that the head is the name of the robot and the serial number. So you can change everything except for the head and it's still the same robot. Now, actually, what happens sometimes is something goes wrong and you have the full robot but you just change the head, and it will refer to it as the same robot even though you changed the head, and now it's just chaos, right? Because then this is just internal service or whatever. It's like, no, is that a robot? It's just like, no, no. So we ended up with a new concept now that we call the quantum serial number, which is the robot's quantum consciousness, and it's virtual. It doesn't actually matter how much you change everything on the robot; it's still the same robot. Not sure if our users will be happy with it, but it's our current best bet.

这是你目前最好的办法。

That's your best bet right now.

好的，有意思。把它放进最初几户人家是什么感觉？最令人惊讶的事情是什么？

Okay. Interesting. What has it been like putting it in the first few people's homes? What have been the most surprising things?

有多难。

How hard it is.

现实世界太难了。

It's the real world is so freaking hard.

一切。我知道，一切，对吧？当我第一次把它带回家——我们家的第一个机器人是很久以前了，大概几年前——就像又多了个孩子。它总是惹麻烦。而且，我们试着在晚上运行它，因为一开始有点慢。我想重置我的房子，所以它在我睡觉的时候做这件事，我晚上被奇怪的声音吵醒。我就想，这到底是怎么回事？

Everything. I know, everything, right? When I first got this at home—and I had the first robot we have at home quite a while ago, like a couple of years ago now—it was like getting another kid. It gets into trouble all the time. And also, we tried running it at night because it's a bit slow in the beginning. I wanted to reset my house, so it's going to do that while I'm sleeping, and I wake up to strange noises at night. I'm like, what the heck is going on?

那太吓人了，如果你去倒杯水，然后这个东西突然亮起灯。

That'd be so scary if you get a glass of water and this thing just like you just see the light flash on.

是啊，是啊，是啊。所以，是的，有很多有趣的经历，美好的回忆。但你知道，现实世界真的很复杂。这就是为什么它如此美丽，也是为什么从数据的角度来看它如此重要，但这将是一个漫长的旅程。我觉得我们现在正达到一个阶段，我想说，如果你是一个早期采用者，像我这样的极客，机器人是真正有用和令人愉快的，而且每天都在变得更好。沿着这条轨迹发展非常美好。但还不是所有人都能接受。它非常适合那些想成为第一波在家拥有它、与技术共同生活并学习如何相处的早期采用者。但这也带来了真正的好处，对吧？成为这些新型系列的早期采用者通常会在未来带来巨大的好处，因为现在你有点用不同的眼光看世界，这其中有力量。所以我认为我们会有很多非常满意的客户，但可能还需要一两年，你才会为你的祖母买一个，因为她只是需要帮助。

Yeah. Yeah. Yeah. So, yeah, there were a lot of fun experiences, good memories. But you know, the real world is really complicated. And that's why it's so beautiful and that's why it's so important from a data point of view, but it's going to be a long journey. I feel we're getting to now where I'd say the robots are genuinely useful and enjoyable if you're an early adopter, a nerd like me, and it gets better every day. And following that trajectory is very beautiful. But it's not for everyone yet. It is very much for early adopters that want to be part of the first wave that has this at home and gets to live with the technology and learn how that is. But there's a real benefit to that also, right? Being the first adopters of these new novel series usually has huge benefits down the road because now you kind of see the world with different eyes, and there's power to that. So I think we'll have a lot of very happy customers, but it's probably like a year or two until you just buy it for your grandma because she just needs help.

是的。在你看来，什么时候是转折点，它真正在你家变得有用了？

Yeah. When was the inflection point in your mind when it actually became useful in your home?

很早。

It was very early.

嗯。

M

更像是，因为你不需要做太多事情，只是整理东西、放好、吸尘、把厨房里的东西放进洗碗机，这样就很好。还有和朋友坐下来玩棋盘游戏，每个人都只是说，‘嘿，你能给我倒杯茶吗？嘿，你能给我拿瓶新可乐吗？嘿，你能去做那些事吗，对吧？’你可以说不需要，但真的很有趣。

It's been more like because you don't need to do that much, just tidying stuff, putting it away, doing some vacuuming, putting stuff in the kitchen in the dishwasher, and just that's nice. And also sitting down with my friends and playing board games and everyone's just like, 'Hey, can you get me a cup of tea? Hey, can you get me a new Coke? Hey, can you go and do those stuff, right?' You can say you don't need that, but it's a lot of fun.

我在想，最基本的版本就像一块宠物石，然后你开始加一点——就像人们有宠物石一样。

I was thinking the most basic version is like a pet rock, and then you start adding a little bit of—like people have pet rocks.

你加一点功能，它就能帮你吸尘；再加一点功能，它就能给你泡茶并吸尘。

You add a little bit of functionality and it like vacuums for you and then you add a little bit of functionality and it gets you tea and vacuums for you.

是的。但我认为门槛其实很低，只要你别过度承诺。

Yeah. But I think the bar is actually pretty low as long as you don't overpromise.

嗯。

Yeah.

对吧。必须管理好预期，对吧？嗯，难的并不是那个。难的是那些基本的东西，比如可靠性、用户友好性，

Right. It has to be expectation managed, right? Um, what was hard was not that. What was hard is just like the basic stuff like reliability, user friendliness,

所有这些方面

all these kind of things

以及把这些做好。Wi-Fi。

and getting that right. Wi-Fi.

Wi-Fi。

Wi-Fi.

该死的 Wi-Fi。

Freaking Wi-Fi.

Wi-Fi 是个问题。

Wi-Fi was an issue.

Wi-Fi 仍然很难。Wi-Fi 几乎比机器人技术还难。

Wi-Fi is still hard. Like Wi-Fi is almost harder than robotics.

天哪。

Goodness.

好吧。你怎么解决这个问题？

Okay. How do you solve that?

哦，好的 Wi-Fi。好的天线。确保你的机器人有非常好的 Wi-Fi。确保你的客户有好的 Wi-Fi。嗯，不是所有事情都必须通过 Wi-Fi 进行，但如果机器人无法访问更大的模型，它的智能程度就会大大降低，对吧？

Oh, good Wi-Fi. Good antennas. Make sure your robot has really good Wi-Fi. Make sure your customer has good Wi-Fi. Um, not everything has to happen over Wi-Fi, but the robot is significantly less intelligent if it cannot have access to a bigger model, right?

有意思。好的。

Interesting. Okay.

所以，所有安全关键的东西和基本功能都在机器人上运行，但一些决策发生在云端。而且，如果你希望机器人能有人介入接管，帮助处理一些机器人还无法自主完成的任务，你需要远程操作，那就需要 Wi-Fi。这就是我要说的，这些都是微妙的东西，对吧？就像再次强调，这是现实世界，你在实验室里觉得这能行，但放到现实世界，你就会想，天哪，还有很长的路要走。幸运的是，我们两年前就已经这么做了。所以我们在把东西放到现实世界并解决问题方面有相当多的经验。

So, all the safety critical stuff and like basic functionality runs on the robot, but some decision-making happens in the game. And also, if you want the robot to have someone intervene and take over and help you if there's some tasks that the robot isn't able to do quite yet uh autonomously and you have some teleoperation, then that requires Wi-Fi. That's uh that's been uh that's all I'm saying like it's the it's a subtle stuff, right? It's like again it's the real world like you're in your lab and you're like this works and you put in the real world and it's just like oh man like there's a long way to go here. Um luckily we did that like already two years ago. So we have quite a bit of experience with putting things in the real world and figuring it out.

是的。我觉得很搞笑的是，Wi-Fi 路由器竟然是东西能不能工作的关键，而你在实验室里花了那么多时间和精力，它本该能工作，结果问题出在 Wi-Fi 上。

Yeah. I find it so hilarious that like a Wi-Fi router is the difference between the thing working and not working when you spend, you know, all this time and energy in a lab and it like it should work and then the Wi-Fi is the issue.

但你平时可能还是经常为视频通话头疼，对吧？

But you're still probably routinely struggling with uh video calls, right?

我其实不做视频通话。如果我想和某人交谈，我基本上就直接飞过去和他面对面谈。

I actually don't do video calls. I pretty much just fly in and talk with a person if I want to go talk with someone.

好的。那么，让我们确保 Neo 能帮你做所有你没时间做的事，这样你就能飞到这儿来好好聊一聊。

Okay. So, let's make sure Neo does all the things you don't have time to so you can travel here and have a good conversation.

是的。而且我会开我的自动驾驶汽车。

Yeah. And I'll drive in my autonomous car.

我等不及那部分了。

I can't wait for that part.

我对此非常兴奋。

I am so excited for that.

开车不是我喜欢的事。

Driving is not something I enjoy.

我觉得很奇怪，我想每年有 300 万人死于车祸。而且我很久以前就想过，我们投入了多少认知能量在驾驶上，而我们根本不需要。让汽车自动驾驶，这给人类带来的回报，简直相当于数十亿年。

It's so strange to me that uh like I think 3 million people die every year from car accidents. And also like I thought I thought a long time ago about like the amount of cognitive, you know, energy that's gone towards driving when you just we don't need to. Like the give back to humanity of just having cars drive on their own has got to be like billions of years.

我我我我真的总是喜欢，我经常喜欢通过效率的视角来看世界。嗯，这也是我做机器人的原因之一，对吧？因为我认为我们可以创造这样一个美好的未来，每个人都能专注于自己真正热爱和想做的事情，对吧？我喜欢说，我想努力把“不得不”变成“想要”。我认为这和幸福有很大关系。嗯，这意味着社会上的低效率，比如没人关心的浪费时间，就像你提到的基础设施，汽车是一回事，但比如火车能省一半时间，既然我们在加利福尼亚，那能给我们带来多少生产力啊，对吧？

I I I I really always kind like I I like to very often see the world through like efficiency. Um and that's also kind of like why I do robots, right? Because I think we can create this beautiful future where everyone gets to focus on what they actually are passionate about and want to do, right? I like to say like we I want to try to turn the have to into want to. I think there's a lot of connection between that and happiness. Um and that means like the amount of inefficiency in society like wasted time that no one seems to care about like just simple things like infrastructure like you said like cars is one thing but like if your train spent half the time since we're in California right the the amount of productivity that could have given us right like

你环顾四周，看到的是一个充满机遇的世界

you look around and you see a world of opportunity

是的，有足够多的事情要做

yeah there's enough to do

所以对于那些担心机器人来了之后，因为机器人会做所有事情，我们就无事可做的人，

so for people who worried that like when the robots come there will be nothing to do anymore because the robots will do everything.

当那发生时我会非常高兴。但我认为那还需要很长时间。有太多事情要做。看看周围，有太多事情要做。我们不会无事可做。如果你真的无事可做，嗯，找一件你真正热爱、对你来说有意义的事情去做，即使你不需要做。然后你就会快乐。

I will be very happy when that happens. But I think it's going to be a long while. There's so much to do. Like just look around. There's so much to do. We're not going to run out of stuff to do. And if you run out of stuff to do, um find something you're really passionate about that makes sense to you to do even though you don't need to. And then you'll be happy.

是的。是的，很奇怪，人们一方面说“我不想让你抢我的工作”，但同一批人又说“我不喜欢我的工作”。我觉得，失去某样东西的痛苦如此之大，以至于失去的念头比一开始就不喜欢它的事实更糟糕。

Yeah. Yeah, it's very strange when people, you know, there's the I don't want you to take my job, but then there's the same people that are like I don't like my job. It's like it I think the uh the loss of something is so painful that the idea of loss is worse than the fact that you don't even enjoy it in the first place.

嗯，我认为这也是我们所有人都有的一种迫切的人类需求，即归属感和目标感。比如，你为什么早上起床？你早上起床是因为你要做很多对你和他人都有意义的事情。而“对他人”这部分非常重要。嗯，对很多人来说，工作是目标感和归属感的来源。所以，我确实认为这很重要。而且我不认为，在未来我们真正拥有充裕劳动力的时候，人们会在海滩上喝莫吉托中找到意义。我认为人们需要有某种驱动力。我只是不认为那与生产力和为了生存而做特定事情的需要挂钩。我认为我们可以成为更伟大的存在。而且我认为我们基本上不会耗尽创造力，也就是不会无事可做。我们会发明对我们有意义的事情去做。

Well, I think also it's just like this desperate like human need that we all have to like belong and to have a purpose. And like why do you get up in the morning, right? you get up in the morning because you're gonna do a lot of stuff that is meaningful to you and to others. And the others part here is very important. Um, and to a lot of people work is a sense of purpose and belonging. So, I do think that's important. And I I don't think that like in a future where we have a true true abundance of labor that people will find find meaning in just having mojitos on the beach. I I I think people will need to have something that drives them. I just don't think that's tied to productivity and the need to do a specific thing to survive. I think we can be so much more. And I don't think we will run out of creativity basically, which is running out of things to do. We will invent things to do that is meaningful to us.

嗯。

Yeah.

我的意思是，我们一直在这样做，对吧？我们愿意为一幅画多付钱，因为它出自一位伟大的画家之手。即使我的机器人能画出更好的画，你也不会买那幅。而且我认真地说，我认为将来我会邀请你去我家，我会说：“嘿，你想看看管道吗？”那将是漂亮的管道。就像，这是那个工匠，一个管道工匠

And I mean, we do this all the time, right? We pay extra for a painting because it's made by a great painter. And even though my robot could paint a better picture, you wouldn't buy that instead. And I actually seriously I I think that in the future I'll invite you to my home and I'll say like, "Hey, you want to see the plumbing?" It's going to be this like beautiful plumbing. It's like this is like this guy like this artisan of like a plumbing craftsman

工匠，真正的工匠

craftsman proper craftsman

做的这个管道。简直太美了。这不是机器人做的，全是这个人做的。就像这里签了名，对吧？他做的管道。我不认为这些东西会因为有了机器人而消失。至少到目前为止，社会上还没有发生这种情况。

that did this plumbing. It's like freaking beautiful. This is not done by a robot. This is all by this guy. Like it's signed here, right? He did the plumbing. I don't think like these things are going to disappear because we have robots. That at least that's not what has happened up till now in society.

你的想法是，世界上会有更多像工艺、用心和小东西这样的东西。

Your your idea is kind of there's going to be a lot more like craft and thoughtfulness and small things in the world.

是的。那将是一个美丽的世界。

Yeah. It's going to be a beautiful world.

我喜欢这个想法。

I love that.

是的。所以，我知道你在 11 岁左右就开始思考人形机器人，我也想过机器人之类的东西，但就那件具体的事情而言，你最初是怎么对它产生兴趣的？那是很多年前的事了。这不像现在这么深入，我希望我能说点什么有意义的话。

Yeah. So, I know you started thinking about humanoids when you were like 11 and I just like I was thinking about robots and stuff too, but for that specific thing, how did you initially get turned on to it? This is like years ago. This is not like this deep and like it's I wish I could say something here that makes sense.

呃，但我真的记得那一刻。顺便说一句，在那之前我也一直喜欢造东西，我经常做机电类的东西，比如拆开带马达的厨房电器，再重新组装起来造点新玩意儿。但当时我看着一台挖掘机在那儿干活，我就想：我要造人形机器人，这毫无道理。但从那一刻起，

Uh but literally I remember when it happened. I always loved building things by the way from before this also and I built often electromechanical things like picking apart kitchen appliances that had motors or whatever like putting it back together to build something. But I was looking at an excavator just kind of like doing its thing and I was like, I'm going to make humanoids and it makes no sense. But from that point forward,

所以你 11 岁，看着一台挖掘机，那就是……

So you're 11 and you're looking at an excavator and that's the...

然后我决定造人形机器人，但我不太确定……不过它确实是机器人。挖掘机就是机器人。它对社会有用，它在做我们都想做的事，比如打理花园、盖房子之类的。

And I decide to make humanoids, but like the... I'm not sure what the... It is a robot though. An excavator is a robot. It's useful in society. It's doing something that we all want like your garden, house, whatever.

它把东西从一处搬到另一处，创造了实实在在的价值。没错。

It's creating real tangible value moving something from somewhere else. Yeah.

所以，嗯，但它是工程的巅峰，对吧。它把每一门科学都融合进一个产品里。我们什么都做，从材料科学，比如新的铝压铸方法，到如何制造合成纤维，让我们能做出机器人里的肌腱，再到机电一体化、电机、传感器、机械、摄像头、算力、固件、软件、AI、基础模型、产品设计、心理学，等等。所有科学汇聚在一起，创造出美好的东西。希望最终甚至能看到它走向何方，对吧？我们能创造人工生命吗？作为一个热爱工程和造物的人，还有什么比这更有趣的呢？我觉得火箭已经很厉害了，但……

So um but uh it is the pinnacle of engineering right. It is taking every single science and combining it into one product. We do everything from material science, both new like ways of die casting aluminium to how to make the synthetic fibers that allows us to make these tendons in a robot all the way up through like electromechanics, motors, sensors, mechanics, cameras, compute, firmware, software, AI, foundation models, product design, psychology, like It's every science coming together to create something beautiful. And hopefully in the end even like it's be interesting to see like where does this go, right? Can we create artificial life? There's not like as someone who just loves building things in engineering, like can there be anything more interesting? Like I think rockets are pretty high up there, but

它们确实挺有意思，

They're pretty interesting,

但没那么有意思，

But they're not as interesting,

对吧？它们缺少这些东西。它们没有智能。

Right? They're lacking some of these things. They're not intelligent.

火箭更像是一种交通工具。

Rockets are more of a means of transport.

对，没错。

Yeah. Right.

它是一种非常花哨的，我们该怎么说？花哨的飞机。

It's a very fancy, what should we call it? Fancy plane.

你知道，我觉得你大概在 2014 年以另一个名字创办了这家公司，然后你一直在努力，这已经是第十年了。这是这家公司特别奇怪的一点，它不是两年前刚起步然后说“好，我们尽可能快增长”之类的。你从 2014、2015 年就开始做这个了。嗯，你最初的想法是什么，让你决定去攻克这个具体的方向？

You know, I think you started the company in like roughly 2014 under a different name and you basically are pounding away for this is like year 10. Like this is one of the weird things about this company in particular is it's not just started 2 years ago and it's like all right, let's grow as fast as possible or whatever. Um, you've been going on this since like 2014, 2015. Um, what was the initial like thought in your mind for I'm going to go tackle this specific thing?

嗯，答案很简单：钱。呃，从某种意义上说，我一直想做这个。

Well, this is a very simple answer. Money. Uh, in the sense that I always wanted to do this.

嗯。

Yeah.

但我不相信你能白手起家做成人形机器人。

But I'm not a believer in like you don't bootstrap humanoid robotics.

这相当资本密集。

It's pretty capital intensive.

嗯，多年来我一直在研究很多不同的系统，我不相信这个问题能靠复用现有供应链来解决。你必须回到第一性原理，机器人里的每一个部件都要从第一性原理去思考，比如该怎么做，然后你会发现电机不存在、齿轮不存在、传感器不存在，什么都不存在，所以你必须从第一性原理全部造出来。

Um been working on a lot of different systems up through the years and like I had no belief that this is something you can solve by reusing existing supply chain. You have to go back to first principles and like every single thing in the robot you need to think from first principles like how should you do this and then you end up with like the motors don't exist the gears don't exist the sensors don't exist like nothing exists so you have to make it all from first principles

执行器

Actuators

对，基本上就是执行器，但也包括我说过的传感器。嗯，这些往往紧密相连，所以要做执行器，你需要造一种新型电子设备，然后还得为这种新型电子设备写软件，因为那也不存在。现在你还需要发明所有这些的制造工艺，因为那也不存在。这比做一个 App 要花的时间长一点。所以尽管过去十年里我很多次醒来都会想，我们为什么没去做个 App？嗯……

Yeah so basically the actuators but also the like I said the sensors um often like it's very connected so to make the actuators you need to make a new type electronics and then you need to write the software for that new type of electronics because that doesn't exist. Now you need to invent the manufacturing for all these things because that doesn't exist and like that takes a bit longer than making an app. So although I've woken up many times for the last decade and thought like why didn't we make an app? Um

是什么让你在那个时候醒来会想“我们为什么没去做 App”？

What brought you to that point in time when you would wake up and think about why didn't we build the app?

我们有时会走到悬崖边缘。我们真的曾经只剩下不到 48 小时的资金储备，诸如此类，但那是很多年前的事了。

We've been close to the edge sometimes. We've been on uh literally been on like less than 48 hours of runway and things like that, but that's many years ago.

嗯，所以呃……

Um so uh

但我们挺过来了，而且这让我们现在处于一个非常独特的境地，对吧？因为我们是真正做了所有有趣的基础研究的公司，这让我们能造出非常独特的东西，而其他人基本上是把传统机器人学里的经典部件拼凑起来，把四个机械臂装在一起就说这是人形机器人。我们实际上真正在核心技术上做了创新，这就是为什么 Neo 能安全，对吧？为什么它能如此轻量、低能耗、柔顺且柔软，同时能力还比竞争对手强，为什么它能如此便宜，因为从第一天起就是从第一性原理出发。怎么最小化材料消耗，对吧？怎么确保效率？需要尽可能小的电池。怎么确保不需要任何特殊合金或材料？怎么确保不需要严格的公差？怎么确保能用很少的零件制造？而且不只是设计实体产品时这么做，在选择产品应该如何工作的技术方向时也这么做。

We pulled through though and um it's put us in a very unique situation now, right? Because we're the company that actually did all the fun foundational research and this allows us to build something that is very unique while everyone else is kind of like clobbering together like the classical things from classical robotics and then like basically putting four robot arms together and saying humanoid. Uh we've actually really innovated on the core technology and that's why Neo can be safe, right? Why it can be so lightweight, low energy compliant and soft while still being more capable than the competitors and why it can be so affordable because it's just from first principles from day one. How do you minimize material consumption, right? How do you ensure you're efficient? Need the smallest possible battery. How do you ensure you don't need any special alloys or materials? How do you ensure you don't need tight tolerances? How do you ensure it's manufacturable in few parts? And like not just doing that when you're designing the actual physical product, but doing that when you're choosing the technical direction for how the product should even work.

嗯。

Mhm.

对。所以你不能事后才做。那就像说，我造了一辆漂亮的汽油车，现在咱们把它改成电动的吧。不，那完全是另一辆车。你得从头开始。所以我们实际上必须从这些第一性原理出发，我认为这对我们非常好，因为我们很早就做到了，而且我认为我们也很早就预见到，要让这个东西产生尽可能大的影响，最重要的是什么。对我来说，回到那个点，它必须安全，对吧，它需要在人群中间，如果它不在人群中间，你就得不到智能所需的数据，也无法逐渐变得有用，这非常重要，因为你不会某天醒来，然后机器人就开始做所有蓝领工作了。不，不。这和 AI 领域的其他部分发展非常相似。我们不是某天醒来，突然所有白领工作都自动化了。不，我们都更高效了，因为我们和这个伟大的工具协作，这是人与机器之间美好的共生关系，Hugan 也不会例外，对吧？在未来几年里，它会逐渐变得越来越有用，你会完成更多事情，因为你有了一个人形机器人，但它不会为你做所有事。

Right. So you can't just do it afterwards. You're that would be like saying I built this beautiful gasoline car over here. Now let's just make it electric. No, like that's a completely different car. Like now you need to start over. So um we actually have to start from these first principles and um I think that's been uh that's been very good to us that we were able to do that very early and I think also able to foresee quite early what would be the most important things for this to have as big of an impact as possible and to me that's back to like it has to be safe right needs to be among people if it's not among people you're not going to have the data needed for intelligence also you're not going to be able to be gradually useful which is so important because you're not going to just wake up one day and then robots are going to be doing all blue collar work. No, no. It's very similar to how the rest of AI field has moved. We didn't wake up one day and just suddenly like now all white collar work is automated. No, we're all more productive because we're collaborating with this great tool and it's like this beautiful symbiosis between man and machine and Hugan will be no different, right? In the coming couple of years be gradually more and more useful and you'll get more stuff done because you have a human humanoid but it won't do everything for you.

我们会一起做到，这意味着它需要安全，而且就像我说的，它需要能胜任工作、有能力，但最后还有可负担性这一点。

We'll do it together, and that means it needs to be safe, and like I said, it needs to be able to do the job, be capable, but then lastly, the affordability thing.

这是你现在最关心的问题吗，就是把成本降到人们真正愿意采用的水平？

Is that the thing you're most concerned about right now, just getting the cost down to something where people are really happy to adopt it?

我们现在已经做到了，我对此很满意。我觉得现在的主要担忧更多是运营层面的。我们要确保以非常高的质量和可靠性来生产它们，确保与客户沟通非常清晰，确保给每个人非常好的客户体验，让客户满意、关系良好，在运营上做到卓越，把安全、隐私、安全这些基本事项都做好。我的意思是，这是产品的一部分，对吧？

Now we're already there, I'm happy with that. I'd say my main concern right now is more operational. Let's make sure we produce them with very high quality and reliability. Let's make sure we communicate very clearly with our customers. Let's make sure we give everyone a very good customer experience, happy customers, good relationships, operationally just excel, get all the basics right like security, privacy, safety. I mean, that's the part of the product, right?

是的。

Yeah.

关于 Scaling（规模扩张），我来之前正好看了 Jason 的视频，你一开始和他聊的时候，好像提到你想在 2027 年做到 100 万台？

For scaling up, I think I was watching Jason's video right before I came, and initially when you were just talking with him, you were kind of talking about, I think you wanted to do like a million by 2027?

10 万台。

100,000.

10 万台，然后 2028 年呢？

100,000, and then like 2028?

2028 年做到 100 万台。

Million by 28.

好的，明白了。在扩大制造规模、达到这些数字的过程中，最大的运营挑战是什么？你预计未来最大的困难是什么？

Okay, there you go. What have been the biggest operational challenges in scaling up manufacturing and hitting those kinds of numbers, and what do you expect to be the biggest difficulties in the future?

这又回到了可靠性、可靠性、可靠性。我认为我们在简化系统方面做得非常出色，制造了一个能力上非常复杂的人形机器人，但它仍然相当简单。话虽如此，它毕竟是人形机器人，就像所有科学汇聚在一起创造出这个美丽的东西，这是人类曾经制造过的最复杂的东西之一，而现在你需要让它可靠。所以所有这些系统和子系统都必须同时工作，而且任何一个都不能失败。然后你还要扩大制造规模，还要引入新人——我的意思是，我们在海沃德拿到厂房后，4 周内就让工厂投产，生产出第一个机器人。我们一天之内雇了 40 多人，对吧？我们举办了一个活动，说“来和我们一起造机器人”，结果有几百人排队，我们筛选并录用了他们，有很棒的人真的来帮忙实现这个未来。但这也意味着同一天来了很多新人，对吧？那么你如何确保另一端出来的机器人仍然可靠？这些事，从更大的角度来看，比如研究如何创造这些具身 AI 和美丽的机器，可能看起来简单，但这些运营的事情并不简单，实际上相当复杂。我觉得那里有很多工作要做。然后当然还有一场竞赛，要让 AI 足够好，让你作为客户超级满意。用不多的资源可以做很多事情。而且，我喜欢你那个逐渐进化的石头的例子，但如果 2027 年它还是一块石头，你可能不会那么高兴。所以有一定的进步是预期的。

It's just back to reliability, reliability, reliability. I think we've done an incredible job in simplifying the system and making a humanoid that is very complicated in how capable it is, but it's still quite simple. That being said, it is a humanoid, and like back to all sciences coming together to create this beautiful thing, one of the more complex things people have ever managed to make, and now you need to make it reliable. So all of these systems and subsystems, they all have to work at the same time, and none of them can fail. And then you're scaling manufacturing, and you're bringing in—I mean, we brought our factory up and running in Hayward in 4 weeks when we got the building to produce the first robot. We hired like more than 40 people in one day, right? We had an event where it's like 'come build robots with us,' and we had hundreds of people lined up, and we screened through and onboarded people, and amazing people really showing up to help make this become the future. But also, that's a lot of new people in the same day, right? So how do you make sure the robot that comes out on the other side is still reliable? These things, in the grander scheme of things, like researching how to create these embodied AIs and beautiful machines, might seem simple, but these operational things are not simple; they're actually quite complicated. And I think there's a lot to do there. And then it's of course a race to get the AI to be good enough to make you super happy as a customer. There's a lot you can do with not that much. Also, I like your example of the gradually evolving rock, but you're probably not that happy if it's still a rock in 2027. So there's a certain progress that's expected.

我知道分布情况，你知道，3% 是超级满意的早期采用者。那仍然是一个庞大的数字，远超 100 万。但你可能会惹恼他们，对吧？

I know the distribution, you know, 3% are super happy early adopters. And that's still a massive number of people, which is way more than a million. But you can piss them off, right?

是的。

Yeah.

如果你过度承诺，有很多这样的例子。我认为他们是你可能遇到的最挑剔的用户，因为他们知识渊博。所以如果你卖给他们东西，最好和你说的相符。

There are many examples of people doing that if you overpromise. I think they're among the most picky users you can have because they're so knowledgeable. So if you sell them something, it better be what you said it was.

所以它可能只是拿起你的咖啡杯、打开门，但这就是你需要承诺的。

So it may just pick up your coffee mug and open the door, but that's what you need to promise.

是的。不要承诺更多，对吧？

Yeah. Don't promise more, right?

对。

Right.

那么你是不是经常在思考这个问题？你知道，你试图快速前进并提升能力，但你也想非常谨慎地设定预期。

So are you constantly trying to think through yourself? You know, you're trying to move really fast and increase capabilities, but you also want to set expectations really carefully.

我认为这非常重要。你必须敢于大胆思考，相信我们会解决这个问题。我们会很快解决。进展非常快。当你每天都生活在这个空间里，这就是你的工作，对吧？作为布道者，就像，我们当然能做到，我们会让它成功，对吧？但当你和客户交谈时，你不能说同样的话。必须有一定的缓冲，因为我们认为自己能做的比我们说的多得多。我们只是希望确保最坏的情况下你仍然满意，最好的情况下你会感到惊喜。

I think it's very important. You have to dare to think big and believe that we will solve this. We'll solve it pretty fast. It's moving really fast. And when you're living in that space every day and that's your job, right? Being the evangelist is just like, of course we can do this, we'll make it work, right? And now when you're talking to your customers, you can't say the same thing. There has to be a certain buffer because we think we can do a lot more than what we say we can do. And we just want to make sure that worst case you're still happy, best case you're blown away.

让我们以你未来 12 个月最期待的事情来结束吧。

Let's end it on what you're most excited about in the next 12 months.

毫无疑问，是看到越来越多的 Neo 进入家庭，让我研究了十年的技术真正对人们有用。而且，我迫不及待想把第一批机器人交付给我所说的普通客户。对，即使我们已经有了家庭用户，我也不会称他们为普通客户。

No doubt, seeing more and more Neos out in homes, having the technology I worked on for a decade actually be useful to people. And yeah, can't wait to ship the first robots to what I'd call our normal customers. Right, so even though we have it in homes, I wouldn't call them normal customers.

你打算基本上让每个员工家里都有一台 Neo，以便尽可能多地接触机器人吗？

Are you going to try to basically have every single one of your employees have a Neo in their home just to have as much contact with the robot as possible?

不一定每个员工都要有，或者太多。希望他们都会买一个，然后他们就能拥有一个。但我确实认为标准应该是，如果你的员工不想买一个，那就有问题了。我记得 Brian Chesky 做过一件事，他基本上会提供，或者他希望 Airbnb 的员工花尽可能多的时间住在 Airbnb 上，经历在他们平台上预订等过程，试图理解客户体验是什么样的，并从中预测痛点。

Not necessarily every single employee, or too many. Hopefully they will all buy one, then they can have one. But I do think the bar should be, if your employees don't want to buy one, something's wrong. I remember Brian Chesky had a thing where he would basically offer, or he wanted the people that worked at Airbnb to spend as much time staying at Airbnbs, to go through the process of just booking on their platform and stuff, to try to understand what the customer experience was like and just predict pain points through that.

而且补贴这个确实有意义。

And it does make sense to subsidize this.

是的。

Yeah.

但我认为即使在这种情况下，对吧，可能应该为 Airbnb 付费。

But I think even in that case, right, should probably pay for the Airbnb.

是的。

Yeah.

也许你能得到一个不错的折扣。所以我们确保你使用 Airbnb，但除非你需要付费，否则你不是真正体验产品的客户。这是很重要的一部分。你付了钱。你还满意吗？

Maybe you get a good healthy discount. So we make sure you use Airbnb, but you're not really a customer experiencing the product unless you need to pay. It's an important part of it. You paid for it. Are you still happy?