本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
你正在收听TIP。
You're listening to TIP.
嘿，大家好。
Hey, everyone.
欢迎收听本周三发布的《无限科技》。
Welcome to this Wednesday's release of Infinite Tech.
今天，我们讨论人工智能和机器人技术，以及仍需改进的关键发展领域。
Today, we're talking AI and robotics and where there's still key areas of development that need some work.
我的嘉宾是肯·戈德堡，一位领先的机器人研究专家，他的工作连接了学术人工智能、现实世界自动化和大规模商业机器人系统。
My guest is Ken Goldberg, a leading robotics researcher whose work bridges academic AI, real world automation, and large scale commercial robotic systems.
我们讨论的一个非常有趣的话题是，大型语言模型能自动赋予物理智能的假设。
One of the things we discussed that's super interesting is the assumption that large language models automatically unlock physical intelligence.
在这方面，肯非常专业，他很好地解释了这究竟意味着什么。
And this is an area where Ken is really well versed and does a great job explaining what that actually means.
我们探讨了哪些方面取得了进步，比如移动性和自动化，以及哪些方面仍然非常困难，比如灵巧性、感知和现实世界的操作。
We cover what has improved like mobility and automation, and what's still painfully hard like dexterity, sensing, and real world manipulation.
这是一场关于工程现实与期望之间差异的务实对话，肯是这一领域的真正专家，你将在对话中看到这一点。
This is a grounded conversation about engineering reality versus expectation, and Ken is a true expert in this field as you'll see in the conversation.
那么，不多说了，希望你们喜欢这次聊天。
So without further delay, I hope you guys enjoy this chat.
您正在收听由普雷斯顿·派什主持的投资者播客网络出品的《无限科技》。
You're listening to Infinite Tech via the Investors Podcast Network hosted by Preston Pysh.
我们通过丰裕与稳健货币的视角，探讨比特币、人工智能、机器人技术、长寿以及其他指数级技术。
We explore Bitcoin, AI, robotics, longevity, and other exponential technologies through a lens of abundance and sound money.
加入我们，一起连接塑造未来十年及更远未来的突破性进展，助您今日就掌握未来。
Join us as we connect the breakthrough shaping the next decade and beyond, empowering you to harness the future today.
现在，有请您的主持人，普雷斯顿·派什。
And now here's your host, Preston Pysch.
大家好。
Hey, everyone.
欢迎收听本节目。
Welcome to the show.
我今天和肯·戈德堡在一起，我非常兴奋能进行这次对话，因为你是机器人和人工智能领域的专家，而这是我们节目里经常讨论的话题。
I am here with Ken Goldberg, and I am so excited to have this conversation because you are an expert in this field of robotics and AI, and it's something that we talk about all the time on the show.
今天能有你这样的人来谈论这个话题，真是太令人兴奋了。
And it's just exciting to have somebody like yourself here today to talk about it.
欢迎来到节目，肯。
So welcome to the show, Ken.
谢谢你，普雷斯顿。
Thank you, Preston.
我也很期待和你交流。
I'm excited to talk to you too.
在我们开始聊天之前，你发给我一篇文章，我觉得这篇文章对今天大部分对话的背景设定非常相关。
So before we started chatting, you sent over an article that I think is very pertinent to kind of set the stage for probably most of the conversation we're having today.
那是一篇刊登在《纽约时报》上的文章，其中提到罗德尼·布鲁克斯说，这个领域已经迷失了方向。
And it was an article that was in the New York Times, and it's talking Rodney Brooks has quote unquote said, the field has lost its way.
对于不熟悉罗德尼·布鲁克斯的人，他是麻省理工学院计算机科学与人工智能实验室的前主任。
And for people that aren't familiar with Rodney Brooks, he's the former director of MIT's computer science and artificial intelligence.
他是ROMBA的发明者，并且负责整个公司和产品线。
He's the ROMBA inventor and running that entire company and product line.
因此，他能说出这样的话，这确实非同小可。
And so for him to come out and say something like this, this is kind of a big deal.
我真的很想听听你对他说的‘这个领域迷失了方向’有什么看法。
And I just really wanna kinda capture your take on what is he talking about the field has lost its way.
我认为他是一位非常受尊敬的人。
Well, I think he's very, very respected individual.
他是我的好朋友，我非常赞同他的观点。
He's a good friend of mine, and I agree with him very much.
我认为他很有挑衅性。
And I think he's provocative.
他用自己的话表达了出来，但我把这篇文章发给你，是因为我觉得它对我们讨论什么是真实的、什么是炒作非常相关，嗯。
He's put it in his own words, but I sent that to you because I think it's very relevant for us to start this conversation about what's real and what's hype Mhmm.
尤其是在机器人领域。
And in robotics.
我得小心使用‘炒作’这个词，但我想说的是，现在确实存在被高估的期望。
And I I have to be careful about the word hype, but I wanna say there's a call it inflated expectations that are out there.
我理解这些期望从何而来。
And I understand where they come from.
你知道，人们对于技术感到兴奋。
You know, I think people are excited about technology.
我也是，我也不例外。
Everyone I am too.
我们所有人都成长于科幻世界，热爱它，也热爱新事物。
We all grew up with science fiction, and we love it, and we love new things.
确实也出现了一些突破。
And there has been some breakthroughs.
毫无疑问，人工智能领域的进步，尤其是深度学习和基于变换器模型的生成式AI，已经彻底改变了这个领域。
I mean, there's no doubt that the advances in artificial intelligence, in particular, deep learning, and then generative AI with the transformer model have been transformative in the field.
如今，人工智能系统正在完成当年没人认为可能实现的事情。
The AI systems are doing things that no one thought would be possible by now.
是的。
Mhmm.
所以，我第一个承认。
So and then I will be the first to admit.
它们具备创造力。
They're capable of creativity.
它们非常有价值。
They're immensely valuable.
但人们接着会合乎逻辑地认为，好吧。
But people then take the next logical step and say, okay.
这些系统已经解决了语言问题，因此它们也会解决机器人问题。
These systems have solved language, so therefore, they'll solve robotics too.
而这一点，我有很多担忧。
And that is where I have a lot of concerns.
我们其实可以深入细节，但我和罗德都同意，语言AI的进展是否能自然延伸到机器人领域，根本不是显而易见的。
I'm really we can get into the details, but Rod and I agree that there is not at all obvious that that with the advances in language in AI, will extend to robotics.
在机器人领域，你认为最脱离现实的说法是什么？比如人们总说五年内就会有类人机器人做所有事情，但实际上机器人技术远没达到那个水平。
What would you say is the number one thing that you're seeing that's just grossly out of touch with reality when it comes to the robotics piece being not as far along or it's not coming in the talking point is in five years, we're gonna have humanoid robots doing everything.
对吧？
Right?
那么，对于那些不熟悉这个领域的人，你认为他们在理解这个问题时最常忽略的关键点有哪些？
So, like, what are the big chunk pieces that people that aren't intimately familiar with the space that you see are missing on that particular topic?
好的。
Okay.
首先，让我告诉你一些已经取得进展的领域，其中一个就是四足机器人。
So let me tell you first of all where some of the advances have been made, and one of them is in quadrupeds.
也就是像狗一样行走的机器人，以及双足机器人。
That's walking dogs, basically, and bipeds.
也就是行走机器和导航，我会称之为移动能力。
That's walking machines and and navigation, and I would call mobility.
因此，机器人通过腿部实现移动的能力已经取得了巨大进展。
So the ability to to get around with robots with legs has made immense progress.
是的
Mhmm.
这非常令人兴奋，这一点毫无疑问。
That's been very exciting, and there's no doubt about it.
这些机器能够完成后空翻、侧空翻、跑酷，还有各种我肯定做不到的事情。
Those machines are capable of doing backflips, as you know, side flips, parkour, all kinds of things that I certainly can't do.
此外，无人机技术也取得了巨大进展。
Also, huge advances in drones.
在过去十年中，无人机技术从一种非常实验性的技术迅速发展，正是多项突破使其成为可能。
And so the past decade, we've seen drone technology take off from something that was very experimental, but it's been a number of advances that have made that possible.
在两种情况下，很多进展都与电机和硬件有关，同时也得益于仿真技术的进步。
In both cases, a lot of it has to do with motors and the hardware, but also advances in simulation.
例如，无人机能够稳定自身，并精确控制四个或六个旋翼上的电机。
And the ability, for example, for drones is to stabilize themselves and then to be able to control very accurately the motors on the four or six rotors that are there.
对于有腿的机器人、四足机器人或双足机器人来说，情况也是如此。
And the same is true for robots that have legs or quadrupeds or bipeds.
嗯。
Mhmm.
所以这些是巨大且不容否认的重大进展。
So where these are big undeniable and major advances.
如果你只是看看这个领域，你会说，好吧。
And if you just look at the field, you say, okay.
所有这些现在都在涌现。
All this is coming in now.
接下来，我们将拥有家用机器人来照顾我们。
The next thing, we're gonna have a home robot taking care of us.
而且，你知道，根据埃隆·马斯克的说法，这已经近在咫尺。
And, you know, this is around the corner according to Elon Musk.
我肯定会遭到一些听众的反驳，他们会说，我不知道自己在说什么。
And I'm sure I'm gonna get some pushback from some of your listeners who are gonna say, I don't know what I'm talking about.
好吧。
Okay.
我见过不少来自硅谷的、非常自信的所谓专家跟我说过这些话。
I've see I've had that happen from a number of, very confident expert, quote, experts from Silicon Valley.
你跟我说说看。
Tell me that.
但我在这个领域已经工作了四十五年，一直密切研究并清楚地了解在特定操作方面还存在哪些差距。
But I've been working in this field for forty five years, and I've studied very closely and understand where the gaps remain for particular manipulation.
嗯。
Mhmm.
所谓操作，就是能捡起你环境中任何东西，然后去操控它们，做各种事情。
And manipulation is being able to pick things up, you know, all kinds of things that just happen to be in your environment and then being able to manipulate them, you know, do things.
这项技能非常精细且复杂。
That skill is very, very nuanced and tricky.
嗯。
Mhmm.
目前的AI方法是否能让我们实现这一点，还不清楚。
And it's not clear that the current methods for doing AI is going to get us there.
我在采访中听过，尤其是埃隆，说过模仿手部结构、肌腱以及具备触觉能力是极其困难的，他就是这么在采访中说的。
I've heard in interviews, Elon in particular, say that the hand, mimicking the hand and the tendons and being able to have that tactile ability is extremely difficult is the way he has said it in interviews.
但我觉得，我怀疑当真正落实时，你看到的都是网上那些演示视频，比如有人捡起一支笔，机器人就做到了。
But I think, I suspect that when it really comes down to it, what you're seeing is a lot of demos online that you see like this this video, somebody picks up a pen and the robot did it.
但背后实际发生的情况是什么？这究竟是一个精心策划的公关表演，还是机器人真的能轻松完成？这两者之间似乎存在巨大的差距。
But what's actually happening there behind the scenes, whether that was a programmed publicity stunt or something that the robot can just do quite well seems to be there seems to be a large gap there.
所以请你谈谈，在你看来，这个差距究竟在哪里，现实情况又是怎样的？
So talk to us about where you see that gap and what it is in reality in your humble opinion.
好的。
Okay.
所以我和我的看法是，这完全可以理解。
So what I and this is understandable.
再说一遍，我不想说人们天真。
Again, I don't wanna say people are naive.
我明白他们的出发点在哪里。
They're I I get where they're coming from.
他们看到某些东西，看起来像人，于是就赋予它人类的特质和能力。
They see something, and it looks human like, and so they attribute human like qualities to it and and skills.
我理解这一点。
I understand that.
顺便说一下，当埃隆说手很难做时，我们确实已经有人在设计看起来非常像人手的手部结构。
And by the way, when Elon says the hands are hard, we can produce people are designing hands that look very much like human hands.
这些手具有22个自由度，能够快速独立地控制所有关节，外观几乎与人手一模一样。
That is they have 22 degrees of freedom, and they can move all these joints independently very quickly, and they look almost identical to human hands.
所以，我们是可以复制这种设计的。
So we can reproduce that.
事实上，目前在中国就有大约100种不同型号的手部装置正在由不同公司研发生产。
In fact, there's, like, a 100 different hands that are being produced by different companies in China right now.
明白吗？
K?
手部本身的进展已经非常先进，但真正的挑战在于对手部的控制。
So the advances in the hand itself is very are very sophisticated, but the control of the hands is where the challenge is.
嗯嗯。
Mhmm.
而这就是挑战所在：如果你让这只手做这些动作，但要让它真正系上鞋带，这才是难点。
And this is where if you you have this hand doing this, but then get it to actually tie your your shoelace, that is where the challenge is.
这是因为我们在用手指与环境互动时，存在大量细微之处——我们感知环境，对环境施加力量，这些都非常微妙且复杂。
And this is because we have there's so many nuances in the interactions that we have with these fingers with the environment that we are sensing the environment, we are exerting forces on the environment, and this is very subtle and very nuanced.
我们通过多种技术来感知这些细节。
And we perceive this through a variety of techniques.
我们每只手上大约有15000个传感器。
We have something like 15,000 sensors in our hand in every hand.
是的。
Yeah.
我知道这很了不起。
I know it's remarkable.
我们甚至不会去想它，因为这是潜意识的。
We don't even think about it because it's subconscious.
是的。
Yeah.
但我们的关节中也有传感器，每一个关节都有，因此我们能够感知到非常细微的力、滑动，尤其是非常微妙的变形。
But then we also have sensors in our joints, every one of our joints, so we are able to perceive very subtle forces, slip, and in particular, one very, very nuanced thing is deformation.
所以如果你看看你的指尖，它们以一种非常有趣的方式进化了，那些指垫非常有帮助。
So if you look at your fingertips, they've evolved in a really interesting way that those pads are extremely helpful.
如果你戴上指套，比如你做缝纫时戴的那种，就会让任何事情都变得困难得多。
If you put on, let's say, thimbles on your finger, right, like you're doing your sewing, that makes it much more difficult to do anything.
是的。
Yeah.
对吧？
Right?
你可以想象。
You can imagine.
或者干脆戴双厚手套。
Or just actually heavy gloves.
手套。
Gloves.
是的。
Yeah.
还有呢。
As well.
对吧？
Right?
但我们能够非常微妙地完成这些动作。
But we can do these things very, very subtly.
我们已经学会了如何与物体的力相互作用，因为这些物体总是在移动和变形。
We have learned this ability to interact with the forces of of objects that the objects are constantly being moved and deformed.
所以如果你想想鞋带，对吧，这个物体正在被变形。
So if you think of the shoelace, right, that's the object is being deformed.
指尖也在被变形。
The fingertip is being deformed as well.
这种相互的形变非常微妙，我们甚至不知道如何准确地模拟它。
This mutual deformation is something that's really nuanced and subtle, and we don't even know how to simulate it accurately.
因此，我们甚至无法模拟正在发生的力、力矩和形变，也没有感知这些细微变化的传感能力。
So we can't even simulate the forces and torques and deformations that are occurring, and then we don't have the sensing capabilities to perceive these nuances.
比如，你能感觉到鞋带在你的指尖略微滑动时的状态。
Like, you can feel a shoelace if it's a little bit slipping out of your fingertip.
嗯哼。
Mhmm.
没有任何机器人能做到这一点。
No robot can do that.
所以当你拥有这只手并试图执行某些操作时，有时能成功，但很多时候却会失败。
So what happens is that when you now have this hand and you actually try to execute something, sometimes it works, but a lot of times it doesn't work.
是的。
Yeah.
现在你面临的是可靠性问题。
Now you have the issue of reliability.
是的。
Yes.
这正是我们正在观察到的。
That is where we're seeing.
顺便说一下，你可以整天看到机器人从桌子上捡起东西并移到别处。
And by the way, you can see robots all day long picking up stuff off a table and moving it somewhere.
这其实并不难。
That's actually not so difficult.
如果你只是想捡起一个毛绒玩具，顺便说一下，毛绒玩具非常容易处理，因为你几乎不可能搞砸。
If you just wanna pick up, especially a stuffed animal, by the way, stuffed animals are very easy because you almost can't go wrong.
你只要把夹爪放在它们附近，然后夹紧，就能成功捡起来。
You just put your gripper anywhere near them and close it, and you'll pick up that thing.
明白吗？
Okay?
这些就像是待宰的鸭子。
Those are like you know, they're sitting ducks.
对吧？
Right?
这些是超级简单、唾手可得的任务，我们可以这么说，这使得抓取和移动物品变得非常容易。
They're no this is super easy, low hanging fruit, let's call it, and that makes it very easy to pick up and move things.
但当你想要开始做一些像插入物品这样的事情时，比如打开一个毛绒玩具，取出里面的填充物，或者把它重新缝起来。
But now when you wanna start doing things like inserting things, like repairing a stuffed animal by opening it up and getting things and, you know, pulling out the stuffing or sewing it back up.
这完全是另一回事，而且要难得多。
This is totally different and much, much more difficult.
是的。
Yeah.
你举的鞋带例子非常深刻，因为除非你退一步，认真想想：如果让我设计或建造一个能系鞋带的机器人，我根本无法想象这种任务会有多困难，因为它是一个极其复杂的任务，而我以前甚至没想过这种事会这么难。
Your example of a shoelace is really profound because until you, like, take a step back and just think if I had to design or build a robot to tie a shoelace, I can't even imagine how incredibly difficult something like that would be because it is such a complex task, and I've never even thought about how difficult something like that is.
好吧，这是我所知道的。
Well, here's here's what I know.
因为系鞋带，我们都做过。
Because, shoelace, we all do.
我们小时候就学会了，几乎不用思考就能做到。
We learn when we're young, and we kinda do it without even thinking about it.
这完全是潜意识的。
It's just subconscious.
对吧？
Right?
我可以在打电话的时候系鞋带。
I can be on the phone tying my shoelace.
根本不用想。
Don't even think.
但想想这个。
But think about this one.
我不知道你怎么样，但你知道怎么打领结吗？
I don't know about you, but do you know how to tie a bow tie?
打领带我会，但打领结不会。
A tie, but not a bow tie.
是的
Yeah.
领结
Bow tie.
好的
Okay.
很好
Good.
我以为你可能会，因为你看起来是个很时尚的人。
I thought you might because you seem like a fashionable guy.
我试过。
I I have tried it.
这非常难。
It's very tricky.
是的
Yeah.
这是一门非常棘手的技艺，而且很微妙。
It's very tricky business, and it's subtle.
你必须能够感知并协调所有这些不同的方向。
You have to be able to feel and pull in all these different directions.
是的。
Yeah.
算了吧。
Forget it.
在很长一段时间内，都不会有机器人能做到这一点。
There's no robot that's gonna be able to do that for a long time.
我真希望这件事能实现，因为那样我就可以让机器人帮我系领结了。
I would love to have it happen because that would be something I would love to have a robot tie my bow tie.
这是另一个非常简单的动作。
And here's another one that's very simple.
就是扣衬衫扣子。
It's just buttoning your shirt.
是的。
Yeah.
如果你仔细想想，这对人类来说其实也有点棘手，你得稍微摆弄一下才能系好。
It's actually a little tricky for humans if you think about it, how you have to kind of fiddle with it a little bit to Yeah.
扣上和解开扣子，尤其是小扣子。
Button on and off, especially a small button.
这远远超出了机器人目前的能力范围。
So that's way beyond robotics.
你肯定永远不会看到机器人扣扣子的演示。
You'll never see a demo of a robot buttoning up, for sure.
我们实际上在我的实验室里正在研究这个问题，但真的很难。
We're we're actually working on it in my lab, but it's really hard.
哇。
Wow.
是的。
Yeah.
这些事情你平时根本不会在意。
It's things that you just really take for granted.
现在当你开始解决这个问题时，你之前提到过，这几乎是一个感知问题，我们需要在指尖所使用的传感器类型上取得大量进展。
Now when you get into solving that problem, it seems like you mentioned this earlier that it's almost a sensing issue that we need a lot of developments on the whatever type of sensors you have in the fingertips or whatever you're using for the manipulation.
目前最大的障碍是不是就在于复制我们指尖如此强大的感知能力？
Is that the biggest hurdle right now is in just kinda replicating how our fingertips can have so much sensing capability?
好的。
Okay.
这是一个方面。
So that's one.
但有一件事至少让我感到有些鼓舞，那就是如果你看看机器人手术领域，顺便说一句，人们对这个领域有很多误解。
But here's there's something that's somewhat encouraging for me at least, which is that if you look in the realm of robot surgery and by the way, there's a lot of misconceptions about that.
我做演讲时，有人会说，机器人给我侄子切除了阑尾。
I give talks where people say, well, robot took out my my nephew's appendix.
嗯。
Mhmm.
我会说，那不是机器人。
And I'll say, that was not a robot.
先生，那是外科医生在使用机器人作为工具，是的。
That, sir, that was a surgeon using a robot as a tool Yeah.
来完成那次手术。
To do that operation.
对吧？
Right?
所以人们称它为机器人，但实际上它是一个遥操作机器人，或者更准确地说，是一个傀儡。
So they call it a robot, but it's really a telerobot or more literally a puppet.
嗯。
Mhmm.
一个非常非常重要、非常有用且昂贵的傀儡，但它确实是个傀儡。
A very, very important and very useful and expensive puppet, but it's a puppet.
因此，理解这一点非常重要。
And so that's very important to understand.
所以外科医生能够完成非凡的任务。
So surgeons can do remarkable tasks.
他们可以缝合。
They can sew.
你得把伤口缝起来。
You sew up a wound.
他们可以用这些工具切除阑尾或胆囊。
They can take out an appendix or a gallbladder with these tools.
现在他们还没有触觉反馈。
Now they do not have tactile sensing.
实际上，最新版本已经开始引入一些触觉反馈，但多年来，他们一直都没有。
Actually, the very latest versions of it, they started to introduce some, but but but for many years, they didn't.
但外科医生仍然能够完成惊人的工作。
And surgeons are still able to do amazing things.
是的。
Mhmm.
所以这表明，也许我们并不需要具备触觉感知能力。
So this is evidence that maybe we don't need to know how to do tactile sensing.
这只是一个假设，但它证明了我们能够进行复杂的、灵巧的操作，处理高度变形的表面。
It's just a hypothesis, but it says that we have an existence proof that manipulation, dextrous complex manipulation with very complex deformable surfaces.
对吧？
Right?
我的意思是，摘除阑尾比系蝴蝶结还要难。
I mean, it's harder than tying a bow tie to take out an appendix.
你居然可以在没有触觉感知的情况下完成这件事。
That you can do that without tactile sensing.
现在有趣的是，外科医生似乎通过适应缺乏触觉，转而依赖视觉来完成这些操作。
Now what's fascinating is the way surgeons seem to do it is essentially accommodating the lack of tactile and then using vision Mhmm.
他们的眼睛。
Their eyes.
他们体内有摄像头，正在观察发生的一切，亲眼看到结果。
They have cameras in there, and they're watching what's happening, and they're they're seeing what happens.
而且他们依靠视觉建立了反馈回路，能够看到组织非常微小的形变，并由此推断出内部发生的情况。
And that they have a feedback loop based on vision so they can see very small deformations of the tissues, and they sort of they infer what's going on.
这非常了不起，因为我认为，这条路径才是实现操控能力最令人兴奋的方向——与其试图复现触觉（这在各种原因下都极其困难），不如去探索这条路径，我觉得它既有趣又值得投入。
This is remarkable because we this, I think, is the most exciting path to getting to manipulation, which is rather than trying to reproduce tactile, which is extremely difficult for all kinds of reasons, but I think it's interesting and worth pursuing.
但还有一条路径，那就是理解视觉与触觉之间的相互作用。
But there's another path which is to understand the visual tactile interactions.
我认为，如果我们能实现这一点，或许就能仅靠摄像头来完成任务。
And that I think if we can do that, we might be able to get away with just using cameras.
有意思。
Interesting.
这周或者上周，抱歉，我看到一篇文章讨论了埃隆在手部设计上的方法与Figure AI之间的差异——他们就在手掌正中央安装了摄像头，正如你所说，在手部直接装了摄像头，而埃隆却坚决不在手部安装摄像头。
So this week or last week, I'm sorry, I saw an article that was talking to the difference between Elon's approach, particularly on the hands, between him and what Figure AI is doing where they put right here in the palm, they put a camera, to your point, they put a camera right here in the hand, and Elon is refusing to put a camera in the hand.
发这篇文章的人说，这就像他当初在汽车上拒绝使用激光雷达一样，因为最终这归结为成本问题，他想迫使团队在不增加传感器和制造成本的前提下解决问题。
And the person who posted this was saying this is akin to him not using LIDAR in the the the cars because in the end, it's gonna come down to a cost thing, and he wants to force his team to figure it out without additional sensors and, for all intents and purposes, costs for manufacturing.
他是在下一场更长远的棋。
And he's playing this longer game.
是的
Yeah.
你对此有什么看法？
What are your thoughts on that?
这是一个非常精彩的见解，普雷斯顿。
Well, that's a brilliant point, Preston.
很高兴你提到这一点。
I'm really glad you made it.
这个类比非常贴切，确实成立。
It's a very good the analogy really works there.
从某种意义上说，埃隆非常自信。
Elon is very in some sense, you know, he's very confident.
他取得了令人惊叹的成就。
He's done amazing things.
他有理由感到自信。
Understandably, should be confident.
但有时候这会蒙蔽你的双眼。
But sometimes that can blind you.
所以在这个情况下，你知道，他决定不使用激光雷达，我认为确实限制了特斯拉的驾驶系统。
So in in this case, you know, his decision not to use LiDAR has really, I think, put a limitation on the Tesla driving systems.
激光雷达在应对某些特殊情况时非常有帮助，比如我们设想的摄像头可能因光线眩光或尤其是雨天而失真或被遮蔽。
LiDAR, it can be very helpful for filling in the edge cases with certain conditions we envision cameras can be distorted or blinded by light flares or especially in rain.
嗯。
Mhmm.
所以激光雷达实际上是一个很好的补充。
So LiDAR actually is a great addition there.
还有成本问题，我不确定。
And and also the cost, I don't know.
我认为随着时间推移，成本会下降。
I think it will come down over time.
我不在汽车行业，所以我在这方面尊重他的专业判断。
I'm not in the car business, so I I defer to his expertise there.
但同样地，他最初当你还记得他刚创办特斯拉时，他希望所有汽车都由机器人工厂生产。
But in the same way, he has you know, originally, you might remember when he first started Tesla, he wanted all the cars to be made in with robotic factories.
是的。
Mhmm.
他还下达过指令：不能有任何人类参与。
And he had a decree, we will have no humans.
你知道的？
You know?
是的。
Mhmm.
所有事情都必须由机器人完成。
Everything must be done by robots.
我记得有特斯拉的工程师来到我的实验室，说：你能帮帮我们吗？
And I remember engineers coming in from Tesla to my lab and saying, can you help us?
我们正在尝试做这件事，但用机器人就是无法实现。
We're trying to do this thing, and we can't get it to work with a robot.
他就是，你知道的，毫不妥协。
And he was just, you know, unrelenting.
最后，他说：我错了。
And then finally, he said, I was wrong.
是的。
Yeah.
他错了。
He was wrong.
是的。
Yeah.
他说：我犯了错误。
He said, I I was mistaken.
人类被低估了。
Humans are underrated.
你还记得吗
Do you remember that
引号？
quote?
是的。
Yeah.
这真的很有趣，因为这是他很少承认错误的时候之一。
So that was really interesting because it was one of the rare times he admitted.
但这也很好地说明了一个观点：你不能用机器人做所有事情，对吧？
But it was also it was a great example of the idea that you can't do everything, right, with robots.
即使你意志坚定，你知道，他可以通过要求来让事情成真，对吧？
And even if if you will it, you know, he can will things into existence, right, by demanding this.
但事情并不总是这样运作的。
It doesn't always work that way.
所以激光雷达的故事非常类似，我认为你对摄像头的看法是正确的。
And so the LIDAR story is very analogous, and I think you're right about the cameras.
你知道，让摄像头在手中是非常合理的。
You know, having cameras in the hand makes a lot of sense.
这并不是人类或动物的工作方式。
It's not how humans or animals work.
对吧？
Right?
嗯哼。
Mhmm.
它们的手上没有眼睛。
They don't have eyes in their hand.
但摄像头是我们非常熟悉的东西。
But cameras are something we understand very well.
我们拥有非常高质量的摄像头。
We have very high quality cameras.
它们速度非常快。
They're very fast.
它们非常精确，而且成本相对很低。
They're very accurate, and they're really low in cost comparatively.
所以我支持多用摄像头。
So I'm for more cameras.
你知道的，多装一些摄像头进去。
You know, put a lot of cameras in there.
因为另一个问题是，当你走路时，头上戴个摄像头是一回事。
Because the other issue is when you walk, it's one thing you have a camera on the head.
你大概能看见周围的情况。
You can sort of see what's around you.
对吧？
Right?
或者开车。
Or drive.
顺便说一下，开车我本该早些提到，但开车比操控容易多了，因为开车只是尽量避开物体。
By the way, driving I should have mentioned this earlier, but driving is much easier than manipulation because driving, you're just trying to avoid objects.
嗯。
Mhmm.
避免撞到任何东西。
Avoid hitting anything.
在操作中，你必须与物体接触。
In manipulation, you must make contact with objects.
你必须操控它们。
You must manipulate them.
对吧？
Right?
所以这非常不同。
So it's very different.
你说得对，这真的非常有趣。
To your point, this is really fascinating.
因为正如你所说，当我拿着鞋带，指尖凹陷或看到鞋带被压缩时，我能感受到它。
Because to your point, when you talk about this idea of, you know, if I'm holding a shoelace and the tip of my finger is indented or I can see the compression of that, I can feel it.
我依赖这种触觉来系鞋带。
I'm relying on that touch like I'm tying my shoe.
我依赖的是这种触觉。
I'm relying on that sense of touch.
但如果你想要设计一个能实现这种功能的传感器，却遇到瓶颈，找不到能提供这种触觉反馈的方案，我可能会看看摄像头，说：好吧。
But if you were gonna try to build a sensor that can do that and you're kind of hitting a roadblock or can't find something that can provide that tactile feedback, I could look at a camera and say, okay.
它往里压了半毫米。
It went in by a half a millimeter.
因此，这大约是这样的压力。
Therefore, it's about this much pressure.
你可以通过图像或视频来替代这种感知能力，通过视觉来观察它。
And you can substitute that sensing capability through an image or a video of being able to see it.
所以，走这条路确实挺有意思的，对吧。
So it's it is kind of interesting that Right.
我们正在朝这个方向探索，是的。
We have figure going that path and yeah.
好吧。
Well, okay.
那让我来补充一点。
So let me let me add on to this.
你刚刚提出了一个非常精妙的细微差别。
So you just made a very nice nuance point.
你说，如果你不只是盯着指尖，而是看到鞋带压在上面，通过观察阴影结构等细节，你大概就能判断出鞋带是在滑动还是被牢牢抓住。
You said if you weren't just looking at your fingertip and you saw the shoelace pressing into it, by looking at the shadow structures and others by a few, you could probably figure out that it was slipping away or it was firmly grasped.
没错。
Absolutely.
这就是外科医生所做的。
That's what surgeons do.
顺便说一下，他们使用的是非常细的手术线，还得用针。
And they, by the way, work with surgical thread, which is really thin, and they have to use a needle.
这非常复杂。
It's very complicated.
对吧？
Right?
但他们主要依靠眼睛和直觉来做这些事。
But they're doing a lot of this with their eyes, with their intuition.
现在光是围绕着安装摄像头是不够的，因为这并不能单独解决问题。
Now it's not just a matter of putting cameras around because it doesn't that doesn't solve it alone.
你实际上需要能够理解这些图像、视频，并对它们进行解读。
You actually now you need to be able to understand that imagery, the video, and you need to interpret them.
是的。
Yeah.
而这同样非常困难。
And that's also extremely difficult.
对。
Yeah.
因为人类拥有这种非凡的能力，我们不能低估它。
Because humans have this incredible ability, and we can't underestimate it.
人类能做的事情真是令人惊叹。
It's just amazing what humans can do.
是的。
Yeah.
从推理的角度来看，比如，如果我这样拿着鞋带，我就能直观地推断出，如果鞋带以与之垂直90度的方式被握住，也会产生同样的滑动感，而这一点对机器人来说很难训练，但人类却能非常轻松地理解。
So when From an inference standpoint, far as, like, if I hold the shoelace this way, I can also just intuitively infer that if it was held 90 degrees from that, that it's gonna have this same slipping sensation, and that's something that's really hard to train a robot on versus humans can just figure it out, like, very easily.
你是指这个意思吗？
Is that what you mean by that?
这就是我的意思。
That's what I mean.
是的。
Yeah.
关键是这样。
And it's here's the thing.
我们缺乏合适的语言来描述这种现象。
We have we don't have good language for describing this.
对吧？
Right?
你知道，我们之所以在尝试，是因为这一切对我们来说都是直觉性的。
You know, we're trying to because it's all intuitive for us.
是的。
Yeah.
你
You
你知道，如果你问我，告诉我怎么系鞋带。
know, I I if you ask me, tell me how to tie a shoelace.
对吧？
Right?
我会说，你知道，这并不容易。
I'd be like, you know, it's not easy.
对吧？
Right?
我们缺乏合适的语言来描述，而这正是原因之一。
We don't have language, and that's part of the reason.
展开剩余字幕（还有 480 条）
顺便说一下，我接下来要谈的另一个问题是数据差距，即我们拥有的语言数据与机器人数据之间的差距。
By the way, this is the the other issue I'll come to is the data gap, the gap between the amount of data we have for language versus robots.
也许现在是个好时机
Maybe this is a good time to
是的。
Yeah.
我们来谈谈这个。
Let's talk that.
是的。
Yeah.
我们来谈谈这个。
Let's talk this.
好的。
Okay.
好吧。
Alright.
好吧，有一种方法可以量化这一切，这就是我所说的机器人数据差距。
Well, there's a way of quantifying all this, and this is something that I call the the robot data gap.
具体来说，如果你把所有用于训练语言模型的数据加在一起，现在这些数据量非常庞大，但很难想象究竟有多少数据。
And it is it's the following, that if you put together all of the data that was used to train language models, Now it's vast, but it's hard to wrap your head around how much data is that.
我的学生们和我能够计算出，如果你仔细分析一下，实际上还有一位名叫凯文·布莱克的研究人员，他在物理智能领域非常聪明。
Well, my students and I were able to calculate that if you actually look at it and and, actually, there's another, Kevin Black, who's a researcher at physical intelligence, very, very smart guy.
他最先提出了这个见解，之后我们在此基础上又做了一些延伸。
He had the first insight about this, and then we've been taking it a little further.
但基本上，如果你把一个普通人阅读所有可用于训练语言模型的文本所需的总时间加起来的话。
But, basically, it's that if you added up all the hours it would take you to read a human, average human to read all the text that's used that's available to train the language models.
对吧？
Right?
那就是所有出版的书籍。
So it's all the books that are out there.
还有所有的维基百科内容。
It's all Wikipedia.
互联网上的一切内容。
It's everything that's on the Internet.
如果你把所有这些标记加起来，然后考虑到人类的平均阅读速度是每分钟238个单词。
If you add up all those tokens, if you will, and then to figure out, well, a human can read at the average speed of 238 words per minute.
对吧？
Right?
你可以算一算，结果会是一十万年。
You can do the math, and you end up with a hundred thousand years.
对？
K?
所以，如果你坐下来读完所有这些用于训练的数据，等你读完时，已经过去十万年了。
So you could sit down and read everything that's used and be a hundred thousand years later, you'd be done.
明白吗？
Okay?
但现在我们还没有关于机器人操作的类似数据。
Now we don't have such data for robot manipulation.
哦。
Oh.
它根本不存在。
It it doesn't exist.
我们不能像在互联网上找那样轻易找到它，数据是完全不同的。
It's not like we can just find it on the It's it's the data is very different.
我们需要从视觉图像开始，最终生成机器人的控制信号。
There, we wanna start with vision images and then end with control signals to the robot.
这根本不存在。
This doesn't exist.
所以我们必须从头开始生成这些数据。
So we have to start and basically generate this data.
但我们面临的挑战是，差距有十万倍。
But what we're up against, right, is it's a 100,000.
我们在语言模型方面落后了十万倍。
We're a 100,000 behind the language model.
所以，再次强调，我有点夸张，只是为了说明一个观点：当然有多种方法可以加速这一进程，我相信我们最终能够实现。
So, again, I'm sort of exaggerating and to make a point, which is certainly there's a number of ways to accelerate that, and and I think we can eventually get there.
顺便说一句，我并不是说这永远不会发生。
By the way, I'm not saying this will never happen.
请别误解我的意思。
Please don't get me wrong.
我相信这件事会发生，但我最大的疑问是：什么时候？
Believe it will happen, but my big question is when.
我认为，为现实做好准备非常重要。
I think it's really important to be prepared for the reality.
有很多人说：嘿。
There's a lot of people who say, hey.
这听起来有道理。
This makes sense.
我想要这个。
I want this.
我们很快就会拥有这个。
We should have this soon.
而且，你知道，但要记住，过去很多人已经讨论过这种情况。
And, you know but remember, there's a lot of cases where people have talked about that in the past.
核聚变能源，核聚变。
Fusion energy, nuclear fusion.
对吧？
Right?
这很有道理。
Makes a lot of sense.
这种技术看起来挺明显的，但你得约束住等离子体。
Sort of the technology is pretty obvious, but you have to contain this plasma.
这似乎是个技术问题。
That seems like a technical issue.
我们能解决这个问题。
We can figure that out.
好了，五十年过去了，我们还在研究它，而且这很难。
Well, fifty years later, we're still working on it, and it's hard.
这是一个非常微妙的问题。
It's a very one of these very, very nuanced problems.
另一个问题是治愈癌症。
Another one is curing cancer.
我小时候，人们常说我们要发起一场针对癌症的战争，就像登月一样。
When I was a kid, they used to say we're gonna have a war on cancer just like we got to the moon.
十年内，我们就能解决癌症。
Ten years, we'll solve cancer.
我们还没有解决它。
We haven't solved it.
所以有些问题极其困难，花费的时间远超任何人的预期。
So there are problems that are extremely difficult, and they take much longer than anyone expects.
是的。
Yeah.
看起来机器人技术就是这样。
It seems like robotics is like that.
我们不知道。
We don't know.
听好了。
And listen.
如果有人解决了这个问题，我会是第一个庆祝的人，比如我早上醒来，读到有人已经解决了它。
I'd be the first to celebrate if we if someone wakes up I wake up and I read someone has solved it.
对吧？
Right?
这是有可能发生的。
It could happen.
是的。
Yeah.
然后你会回过头来看这个播客，说戈德堡完全错了。
And then you'll look back on this podcast and say Goldberg was completely wrong.
不。
No.
这完全有可能发生。
It could totally happen.
但我想发声说：嘿。
But I wanna be a voice to say, hey.
它可能不会发生，让我们认真思考一下这一点，保持一点现实感，因为我知道很多人认为这是必然的，可能会在明年发生，根据埃隆和他的许多追随者所说。
It might not happen, and let's just think about that and be a little bit realistic because I know how a lot of people are are thinking that it's inevitable it's gonna happen in any you know, hopefully by next year according to Elon and many of his followers.
但他们必须做好它可能不会发生的准备。
But they have to be ready for that maybe not to happen.
我担心会引发反弹，人们会说：嘿。
And I'm worried about a backlash that people will say, hey.
整个机器人技术这件事，其实就是骗人的把戏，我们会大批量地退出这个领域。
This whole robotics thing is, you know, was, you know, hocus pocus, and, you know, we're gonna move out of this field in droves.
我不，而且我
I don't and I
我不想替你说话，如果你说得不对，请纠正我。
don't wanna put words in your mouth, so correct me if I'm stating this wrong.
我觉得你并不是说这件事不会发生。
I don't think you're saying it's not gonna happen.
你只是对大家普遍认同的时间表持怀疑态度。
You're just really suspect on the timeline that everybody seems to be.
是的。
Yeah.
对。
Yeah.
就是这样。
That's it.
就是这样。
That's it.
没错。
Exactly.
而且，你知道，我在这一点上赞同罗德·布鲁克斯，因为我们有类似的经验。
And, you know, that's where I I line up with Rod Brooks because for very similar reasons, we have experience.
我们都在这个领域工作了大约四十年。
We both have working in this field for, like, forty.
他比我多干了几年，六十年、五十年了。
He's been working longer slightly longer than me, sixty, fifty years.
但你知道，我们在尝试解决这些问题上积累了丰富经验，而这些问题远比表面看起来要复杂得多，尤其是因为孩子能轻松学会并操作这些事物。
But, you know, we have a lot of experience with trying to solve these problems, and they're they're much more nuanced than they seem on the surface, especially because a child can pick things up and manipulate it.
是的。
Yeah.
对吧？
Right?
看起来显而易见。
It seems obvious.
为什么机器人就不行呢？
Why can't robots?
这非常违反直觉。
It's it's very counterintuitive.
但当你真正接触这些系统并看到它们的局限性时，你就会开始明白，这是一个极其复杂的问题。
But when you work with these things and you really see their limitations, you start to understand that this is a very, very complex problem.
让我们短暂休息一下，听听今天赞助商的信息。
Let's take a quick break and hear from today's sponsors.
比特币挖矿一直被认为复杂、高风险，且难以评估为一项真实的投资。
Bitcoin mining has a reputation for being complicated, risky, and hard to evaluate as a real investment.
如果你正在考虑2026年的挖矿，真正重要的是运行稳定性、维修服务，以及整个运营是否像一家正规企业那样管理。
If you're considering mining in 2026, what actually matters isn't headline profitability, it's uptime, repairs, and whether the operation is run like a real business.
这就是我一直在使用Simple Mining的原因。
That's why I've been using Simple Mining.
他们位于爱荷华州锡达福尔斯，提供全方位托管服务，你真正拥有自己的矿机，自行选择矿池，并直接将比特币发送到你的个人钱包。
They're based in Cedar Falls, Iowa, and they run a white glove hosting operation where you actually own your own miners, choose your own pool, and have Bitcoin sent directly to your own wallet.
他们入选了《Inc》5000强榜单，成为爱荷华州增长最快的企业，管理着超过四万台矿机。
They were featured on the Inc 5,000 list as the fastest growing company in Iowa with over 40,000 machines under management.
最突出的是执行力。
What stands out is execution.
他们拥有排名第一的ASIC维修中心，并且在前十二个月内维修服务免费。
They've had the number one rated ASIC repair center, and for the first twelve months, repairs are included.
如果挖矿利润变薄，你可以随时暂停而无需支付任何罚款。
If mining margins get tight, you can pause with no penalties.
如果你想调整或升级你的矿机阵容，还可以通过一个市场转售设备，而不必被绑定住。
And if you want to resize or upgrade your fleet, there's a marketplace to resell equipment instead of being stuck.
为了帮助人们思考当前挖矿是否真的可行，他们整理了一份名为《20.26美元比特币挖矿蓝图》的简明指南。
To help people think through whether mining actually makes sense right now, they put together a short resource called the $20.26 Bitcoin mining blueprint.
它详细讲解了投资者在配置挖矿时常犯的五个错误，以及在投入资金前如何避免这些错误。
It walks through the five mistakes investors make when allocating the mining and how to avoid them before deploying capital.
如果这听起来有趣，你可以免费获取，地址是 simplemining.iopreston。
If it sounds interesting, you can get it for free at simplemining.iopreston.
地址是 simplemining.iopreston。
That's simplemining.iopreston.
每一家企业都在问同一个问题。
Every business is asking the same question.
我们如何让AI为我们服务？
How do we make AI work for us?
可能性无穷无尽，而盲目猜测风险太高。
The possibilities are endless and guessing is too risky.
但袖手旁观绝非选项，因为有一件事几乎可以肯定：你的竞争对手已经在行动了。
But sitting on the sidelines is not an option because one thing is almost certain, your competitors are already making their move.
借助甲骨文的NetSuite，你今天就能让AI发挥作用。
With NetSuite by Oracle, you can put AI to work today.
NetSuite是全球超过43,000家企业信赖的头号AI云ERP系统。
NetSuite is the number one AI cloud ERP trusted by over 43,000 businesses.
它是一个统一的套件，将你的财务、库存、电商、人力资源和客户关系管理整合为单一数据源。
It's a unified suite that brings your financials, inventory, commerce, HR, and CRM into single source of truth.
这种互联互通的数据，才让您的AI更智能。
That connected data is what makes your AI smarter.
因此，它不仅仅是猜测，而是懂得如何智能地自动化日常任务并提供可操作的洞察。
So it doesn't just guess, it knows how to intelligently automate routine tasks and deliver actionable insights.
让你的竞争对手也试试看。
Let's see your competitors do that.
无论你的公司年收入达到数百万甚至数亿，NetSuite 都能帮助你保持领先。
Whether your company earns millions or even hundreds of millions, NetSuite helps you stay ahead of the pack.
如果你的年收入至少达到七位数，请前往 netsuite.com/study 免费获取 NetSuite 的商业指南《揭开 AI 的神秘面纱》。
If your revenues are at least in the 7 figures, get NetSuite's free business guide, Demystifying AI at netsuite.com/study.
这份指南可在 netsuite.com/study 免费获取。
The guide is free to you at netsuite.com/study.
你是否曾想探索在线交易的世界，却一直不敢尝试？
Ever wanted to explore the world of online trading, but haven't dared try?
期货市场如今比以往任何时候都更活跃，而 Plus 500 正是理想的入门之选。
The futures market is more active now than ever before and plus five hundred futures is the perfect place to start.
Plus 500 让你接触到广泛的投资工具。
Plus 500 gives you access to a wide range of instruments.
标普500、纳斯达克、比特币、天然气，还有更多。
The S and P five hundred, NASDAQ, Bitcoin, gas, and much more.
探索股票指数、能源、金属、外汇、加密货币及其他领域。
Explore equity indices, energy, metals, Forex, crypto, and beyond.
通过简单直观的平台，您可以随时随地，直接用手机进行交易。
With a simple and intuitive platform, you can trade from anywhere, right from your phone.
最低存入100美元，体验您一直期待的快速便捷的期货交易。
Deposit with a minimum of $100 and experience fast accessible futures trading you've been waiting for.
看到交易机会了吗？
See a trading opportunity?
账户开通后，您只需两步点击即可完成交易。
You'll be able to trade it in just two clicks once your account is open.
还不确定自己是否准备好了？
Not sure if you're ready?
没问题。
Not a problem.
Plus500 为您提供一个无限制的免费模拟账户，内含图表和分析工具，供您练习使用。
Plus500 gives you an unlimited risk free demo account with charts and analytic tools for you to practice on.
凭借二十多年的丰富经验，Plus500 是您通往市场的门户。
With over twenty years of experience, Plus 500 is your gateway to the markets.
访问 plus500.com 了解更多信息。
Visit plus500.com to learn more.
期货交易存在亏损风险，并非适合所有人。
Trading in futures involves risk of loss and is not suitable for everyone.
并非所有申请人都能通过审核。
Not all applicants will qualify.
Plus500，让交易更胜一筹。
Plus 500, it's trading with a plus.
好的。
Alright.
回到节目。
Back to the show.
肯，如果你是一名项目经理，你会从各个不同的工作流来看如何达成目标。
Ken, if you were you know, you're a program manager, you're kinda looking at all the different swim lanes to get there.
手部功能似乎是实现这一目标的关键路径之一。
The hand seems to be like one of the, you know, critical path, if you will, for getting there.
你是否还定义了其他属于这条关键路径的内容？
Is there anything else that you would define as being on that critical path?
就难度而言，手部功能是否远远超出其他所有方面，成为真正的瓶颈？
Is the hand so far out there as far as difficulty goes compared to everything else that that's really kind of the limiting factor?
好的。
Okay.
很好的问题。
Great question.
我认为，当你提到手部时，实际上指的是操作能力。
I would say it's it's not only the when you say hand, it's the manipulation ability.
是的。
Yeah.
对。
Right.
顺便说一下，我还有另一个观点：我们会从非常简单的夹持器中获得比从仿人手更多的收益。
Because by the way, I do have another thing to say here, another opinion, which is that we will get much more out of very simple grippers than we will out of hands that look like human hands.
再者，如果你看看外科手术，外科医生做阑尾切除术时使用的工具就是这种非常简单的夹持器。
Again, if you look at surgery, the tools that surgeons use to perform an appendectomy are very simple grippers like this.
嗯。
Mhmm.
它们能完成极其复杂的事情。
And they can do immensely complicated things.
所以我认为不需要复杂的双手。
So I believe you don't need complex hands.
嗯。
Mhmm.
所以我说的并不是应该走这条路。
So I'm not saying that's the path to go.
我相信你可以使用简单的夹持器。
I believe you can do a simple grippers.
事实上，我的公司Ambure Robotics使用了更简单的夹持器——吸盘。
In fact, my company, Ambure Robotics, uses an even simpler gripper, which is a suction cup.
用它们你可以完成惊人的事情。
And you can do incredible things with them.
所以问题不在于硬件，而在于软件。
So it's not necessarily the hardware, but it's the software.
真正非常具有挑战性的是对这种精细交互的控制。
It's a control of this nuanced interaction that is very, very challenging.
我认为其他许多方面都是可以解决的。
I think many of the other aspects are addressable.
我们可以告诉机器人：去把桌子上的橙色毛衣拿起来。
We have the ability to tell a robot, go pick up the orange, you know, the orange jumper off the table.
我们现在就能解决这个问题。
We can solve that now.
计算机视觉系统已经足够先进，能够识别出毛衣就是一种上衣，并且知道有一个橙色的，然后把它拿起来。
Computer vision systems are good enough to know that a jumper is a sweater and, you know, there's an orange one and it'll pull that up.
没问题。
No problem.
但真正困难的是能够实际把它拿起来，或许再帮你穿上，然后帮你扣上扣子。
But it's being able to actually pick it up and maybe put it on you and then button it up for you.
这才是难点所在。
That's where it's it's difficult.
是的。
Yeah.
对。
Yeah.
我的意思是，也许在现阶段，我们看到的是进入市场的类人机器人，它们简化了手部设计，但所能执行的任务范围与真实人类相比非常有限。
I mean, maybe what we see in the interim is robots that go to market humanoid robots that go to market that have simplified the hands or have but the range of activities or things that they can actually perform is very limited relative to a real human being in there and being able to do it.
我不知道这是否就是它们进入市场的方式。
I don't know if that's how they go to market or not.
我想再多聊聊你们公司，Ambi。
I wanna talk a little bit more about your company, Ambi.
这真的非常有趣。
So this is really fascinating.
你们主要专注于物流和仓储类的机器人应用，并已推向市场，对吧？
So you guys have gone to market, primarily focusing on logistics and warehouse type activities for robotics.
是这样吗？
Is that correct?
没错。
Correct.
这大约是从七个月前开始的。
So this started about July ago.
我们在机器人抓取方面取得了突破，就是能够从箱子里取出物品。
We had a breakthrough in robot grasping, and that was just simply ability to pick things out of a bin.
好的。
Okay.
所以这不是进行手术那样的操作，只是从箱子里捡东西。
So it's not manipulating, you know, doing surgery, but it's just picking things out of a bin.
这是一个非常古老的问题。
That was a very old problem.
这被称为‘抓取拣选’问题，几十年来一直有人在研究。
It's been known as the bin picking problem, and people have been looking at that for decades.
是的。
Mhmm.
但我们取得了突破，这主要归功于我的博士生杰夫·莫勒，他是这项研究的主导者。
But we had it we made an advance, and this was especially the work of Jeff Moller, was the the PhD student of mine, who was the lead researcher on this.
我们可以进一步探讨这项技术的细节，但这个系统被称为DexNet，即灵巧网络。
And we can go into more details on the technical aspect of this, but the system was called DexNet, dexterity network.
它基于收集大量数据样本。
And it was based on collecting data, lots of examples.
在很多方面，它类似于ImageNet，后者是计算机视觉领域的重大突破。
And it was somewhat it was analogous in many ways to ImageNet, which was a breakthrough for computer images.
所以我们做了类似的事情。
So we did something similar.
我们合成了一组数据集。
We synthesized this dataset.
我们以一种非常特定的方式加入了噪声，但系统开始表现得异常出色。
We added noise in a very specific way, but the system started working remarkably well.
因此，它可以抓取你放入箱子中的几乎所有物体。
And so it could pick up almost any object that you put into a bin.
它会直接把物体拿出来，而你只需把整堆物品扔进去。
It would just pull it out, and you would throw in the whole pile of objects.
我们翻遍了车库和壁橱，把能找的东西都扔了进去，它总能稳定地把这些东西拿出来。
We were digging around in our garages and closets and throwing everything we could into it, and it was consistently just being able to pull these things out.
那对我们来说是一个非常令人兴奋的时刻。
And so that was a very exciting moment for us.
我们获得了一些关注。
We got some publicity.
它登上了《纽约时报》和其他媒体，随后我们被多家公司联系，于是决定成立我们自己的公司。
It was in the New York Times and other places, and then we were approached by a number of companies, and we decided to form our own company.
太棒了。
That's awesome.
我想知道，在哪些情况下，传统的工程方法比增加数据或使用更大模型更重要？
I'm curious where you've seen just good old fashioned engineering matter more than additional data or larger models.
反过来，什么时候数据真正让你感到惊讶？
And then to the converse of that, when did you have data actually really surprise you?
好的。
Okay.
不错。
Good.
嗯，那还可以。
Well, that was okay.
这是个很好的例子。
So great example.
这是一个数据真正让我们感到惊讶的案例。
That was a case where data really did surprise us.
由于我们收集了10,000个物体模型，因此能够生成600万个抓取示例，并在这些模型上生成抓取动作。
We were able to generate 6,000,000 example grasps over because we had collected 10,000 object models, and then we could generate grasps on those models.
然后我们有了大量数据，并训练了一个网络，使其能够学习如何抓取物体。
And then we had all these and we trained a network to be able to learn essentially where to grasp an object.
因此，这是一种数据驱动的方法。
So that was a data driven approach.
但我必须告诉你，当你将这种方法应用到实验系统或商业系统中时，就需要大量我所说的传统工程技能。
But I will tell you that when you take that and you have to move that into an experimental system or into a commercial system, then you need a huge amount of what I call good old fashioned engineering.
而这时，你必须真正关注每一个细节。
And this is where you have to really sweat the details.
你必须确保传感器校准准确，机器人手臂校准且精确。
You have to make sure that the the sensors are calibrated correctly, that your robot arms are calibrated and accurate.
你必须能够快速完成移动手臂所需的计算。
You have to be able to do the computation to move the arms, very quickly.
你必须控制夹持器的表面、吸盘，以及无数类似的细节，比如照明，我们还在系统下方设置了一个小秤，用于检测物体何时从箱子中被取出。
You have to control the surfaces of the grippers, the suction cups, myriad of details like that, the lighting, the we had a little scale underneath the system that would recognize when an object was removed from the bin.
这就像一个电子秤。
It's like a digital scale.
这只是必须加入的另一项工程细节。
It's just another piece of engineering that had to go in there.
所以这一切加起来非常多。
So lots of all that.
这仅仅是我们在实验室里的演示系统，一个实验性系统。
That was just our demonstration system in the lab, experimental system.
但当我们转向Ambi Robotics时，顺便说一下，我也想向其他参与的学生致谢。
But then when we moved into ambi robotic and by the way, I should I wanna give credit also to the other students who are involved.
马特·马特尔是另一位与杰夫·莫勒密切合作的计算机科学家。
Matt Mattel was another computer scientist working closely with Jeff Moeller.
此外，我还有一位来自机械工程系的博士生，史蒂夫·麦金利和戴维·吉利。
And then I also had two other PhD students from mechanical engineering, and one of them, Steve McKinley and David Geely.
太棒了。
Brilliant.
这四个人。
All four of these guys.
都是非常非常出色的工程学生。
Extremely, extremely brilliant engineering students.
所以他们真的很懂如何合作，他们是非常好的朋友。
And so they really knew how to they were very good friends.
他们至今仍是好朋友。
They remain good friends.
他们紧密合作，还花大量时间一起露营和相处，但他们彼此完美互补，因为我们既有计算技能，也有机械技能。
They all worked very closely and spent a huge amount of time camping and hanging out together too, but they were perfectly complementary because we had the computing skills and the mechanical skills.
是的。
Mhmm.
机械方面的同学懂得如何设计出能够长期可靠运行的机器。
And mechanical guys knew how to design machines that could work reliably over a great period of time.
那就是我们开始开发AmbiSort系统的时候，这个系统用于为电子商务分拣包裹。
And that's when we moved into building the AmbiSort system, which sorts packages for ecommerce.
是的。
Mhmm.
我们最初并没有这个计划，但很快我们就发现电子商务正在快速增长，市场对包裹分拣的需求非常大。
And this was a little bit we didn't go in with this plan, but what we saw very quickly was that ecommerce was growing, and we needed there's a huge demand for sorting packages.
对吧？
Right?
要快速把包裹送到客户手中，真的很有挑战性。
It's just very challenging to get packages out to the customer fast.
是的。
Mhmm.
所以我们开始使用DexNet这个技术。
So we started using that technique, DexNet.
我们对其进行了改进并实现了商业化，让它能够非常快速地运行。
We evolved it and made it it commercialized it, and then we can make it work very fast.
然后还需要引入各种其他组件。
And then all kinds of other elements had to come in.
我们还有另一个桥式系统，用于从货箱中取出物品，扫描其邮政编码，确定应放入哪个货箱，然后将其放入正确的货箱，同时全程避免堵塞，确保系统可靠、安全且易于使用。
We had another gantry system that would drop it into pick it out of bins, pick an object out of bin, had to be scanned for its ZIP code, figure out which bin to go into, then put it into the right bin, avoid jamming the whole time, make the system reliable, safe, and easy to use.
对吧？
Right?
所有这些就是我所说的传统工程方法。
All this is what I call good old fashioned engineering.
是的。
Yeah.
因此，我成为这一理念的坚定支持者，因为毕竟，这是一套经过四五百年工程实践发展起来的研究、理念和洞见。
And so I become a big advocate for this because, after all, this is a body of research and ideas and insights that have been developed over four hundred, five hundred years in engineering.
而我们至今仍在教授这些内容。
And still what we teach Mhmm.
在伯克利和所有顶尖大学，我们教授的是工程原理。
At Berkeley and all the major universities, we teach the engineering principles.
我的观点是，我们不要忘记这些。
And my point is, let's not forget about those.
这些对于工程学以及让机器人真正投入实际应用仍然极其重要。
Those are still extremely valuable for engineering and getting and for robots and getting them to actually work in practice.
是的。
Yeah.
我认为，任何从事机器人工作的人都会承认这一点。
And anyone working in robotics, I think, will acknowledge that.
尽管公众的看法是，哦，这只是现在，我们在使用人工智能，它解决了所有问题。
Although the public perception is, oh, it's just now, you know, we're using AI, and that's solving everything.
并不是这样。
It's not.
它只解决了其中一些小部分。
It's solving certain little pieces of it.
正如我所说，有些部分仍然非常困难，用这种方法依然难以解决。
And as I said, there's certain pieces that are very, very difficult with that still remain very difficult.
所以这回到我之前对普雷斯顿说过的担忧，即由于现在对人形机器人的期望值如此之高，如果公司无法实现这种能力，嗯。
So and this goes back to what I was saying earlier, Preston, about my fear, which is that because there's so much expectation around humanoids right now, that if the companies can't deliver on that ability Mhmm.
那么可能会引发强烈的反弹，这将伤害像Ambi这样并不想做这件事的公司。
Then there might be a big backlash, and that's gonna hurt companies like Ambi who are not trying to do that.
Ambi的目标是解决一个真实的实际问题，高效且低成本地完成，实际上，这对每个在亚马逊或任何在线公司购物的人来说都非常有价值。
Ambi is trying to solve a real practical problem and do it efficiently and cost effectively and actually, you know, basically, something that's very valuable for every everyone who shops at Amazon or any of the online companies.
对吧？
Right?
到目前为止，我们已经分拣了上亿个包裹。
We've sorted a 100,000,000 packages so far.
哇。
Wow.
我为此感到非常自豪，因为这些机器正如我们所说，正在外面分拣包裹，而且非常可靠。
And I'm very proud of that because these machines, as we're talking, are out there sorting packages, and they're very reliable.
它们没有出现在关于人形机器人做这些事的视频中。
They're not featured in the videos about there's no humanoids doing this.
是的
Yeah.
顺便说一下，虽然有些人说，将来会有人形机器人来做这件事。
By the way, although some have said that, you know, we'll have a humanoid doing that.
但有人形双手的机器人，要达到我们这种吸盘系统的效率，还需要很长时间。
But humanoid with hands, gonna be a long time before that's Yeah.
甚至都很难接近我们目前吸盘系统的效率。
Even close to the efficiency of the systems that we have with suction cups.
让我们短暂休息一下，听听今天赞助商的发言。
Let's take a quick break and hear from today's sponsors.
不。
No.
这不是你的错觉。
It's not your imagination.
风险和监管都在加剧，客户现在要求提供安全证明才能开展业务。
Risk and regulation are ramping up, and customers now expect proof of security just to do business.
这就是为什么Vanta是一个变革者。
That's why Vanta is a game changer.
Vanta自动化您的合规流程，将合规、风险和客户信任整合到一个AI驱动的平台上。
Vanta automates your compliance process and brings compliance, risk, and customer trust together on one AI powered platform.
无论您是在为SOC 2做准备，还是在运行企业GRC项目，Vanta都能确保您的安全并推动交易顺利进行。
So whether you're prepping for a SOC two or running an enterprise GRC program, Vanta keeps you secure and keeps your deals moving.
Ramp和Ryder等公司使用Vanta后，审计时间减少了82%。
Companies like Ramp and Ryder spend 82% less time on audits with Vanta.
这不仅仅是更快的合规，更是为增长争取了更多时间。
That's not just faster compliance, it's more time for growth.
我喜欢Vanta让合规管理变得轻松，而不会完全打乱您的工作流程。
I love how Vanta makes it easy to stay on top of your compliance without it taking over your entire workflow.
它简化了通常远比必要更繁琐的事情。
It just simplifies something that's usually way more painful than it needs to be.
立即访问 vanta.com/billionaires 开始使用。
Get started at vanta.com/billionaires.
那就是 vanta.com/billionaires。
That's vanta.com/billionaires.
亿万富翁投资者通常不会把资金存放在高收益储蓄账户中。
Billion dollar investors don't typically park their cash in high yield savings accounts.
相反，他们经常使用机构投资者常用的被动收入策略之一——私人信贷。
Instead, they often use one of the premier passive income strategies for institutional investors, private credit.
如今，得益于 Fundrise 收入基金，这一被动收入策略已向所有规模的投资者开放，该基金已吸引超过6亿美元投资，派息率为7.97%。
Now the same passive income strategy is available to investors of all sizes, thanks to the Fundrise Income Fund, which has more than $600,000,000 invested and a 7.97 distribution rate.
随着传统储蓄利率下降，难怪私人信贷在近几年已成长为一个万亿美元的资产类别。
With traditional savings yields falling, it's no wonder private credit has grown to be a trillion dollar asset class in the last few years.
访问 fundrise.com/wsb，只需几分钟即可投资 Fundrise 收入基金。
Visit fundrise.com/wsb to invest in the Fundrise Income Fund in just minutes.
该基金2025年的总回报率为8%，自成立以来的平均年总回报率为7.8%。
The fund's total return in 2025 was 8% and the average annual total return since inception is 7.8%.
过往表现不预示未来结果。
Past performance does not guarantee future results.
截至2025年1月20日12:30的当前分配率。
Current distribution rate as of twelvethirty onetwenty twenty five.
投资前请仔细考虑投资材料，包括目标、风险、费用和开支。
Carefully consider the investment material before investing, including objectives, risks, charges, and expenses.
更多信息可在fundrise.com/income的收益基金招募说明书中找到。
This and other information can be found in the income funds prospectus at fundraise.com/income.
这是一则付费广告。
This is a paid advertisement.
开始一件新事情不仅困难，而且确实有点可怕。
Starting something new isn't just hard, it's honestly kind of terrifying.
我仍然记得在真正决定做播客之前的那些时刻。
I still remember those moments right before I really committed to podcasting.
半夜辗转反侧，想着如果没人听怎么办？
Lying awake at night thinking, what if no one listens?
如果这件事彻底失败了怎么办？
What if this completely flops?
或者，如果我只是纯粹在浪费时间呢？
Or what if I'm just straight up wasting my time?
尽管克服那种怀疑并不容易，但迈出这一步最终成了我做过的最好的决定之一。
And even though pushing past that doubt was not easy, making the leap ended up being one of the best decisions I've ever made.
我要说的是，当你拥有得力的工具时，事情会容易很多。
And I'll say this, it helps a lot when you have the right tools on your side.
这就是Shopify发挥作用的地方。
And that's where Shopify comes in.
Shopify是数百万企业的电商平台，占据了美国约10%的电子商务份额，从家喻户晓的大品牌到刚刚起步的小企业都在使用。
Shopify is the commerce platform behind millions of businesses and about 10% of all e commerce in The US, from massive household names to brands just getting started.
如果你曾经想过，如果我不知道怎么建店怎么办？
If you've ever thought, what if I don't know how to build a store?
Shopify提供了数百个精美且即用的模板，完美契合你的品牌风格，让这一切变得简单。
Shopify makes it easy with hundreds of beautiful, ready to use templates that actually match your brand.
或者，如果我没时间做所有事情怎么办？
Or what if I don't have time to do everything?
Shopify 内置的 AI 工具可以帮助撰写产品描述、标题，甚至优化产品图片。
Shopify's built in AI tools help write product descriptions, headlines, and even enhance your product photos.
是时候将这些‘如果’转变为‘现在就用 Shopify’了。
It's time to turn those what ifs into With Shopify today.
今天就前往 shopify.com/wsb 注册你的每月 1 美元试用版。
Sign up for your $1 per month trial today at shopify.com/wsb.
前往 shopify.com/wsb。
Go to shopify.com/wsb.
就是 shopify.com/wsb。
That's shopify.com/wsb.
好的。
Alright.
回到节目。
Back to the show.
在每天操作仓库中的送货机器人之后，你早年学术生涯中曾持有过的哪一个观点，如今已被你修正了？
After shipping robots that work every day in the warehouse, what's one belief that you held earlier in your academic career that you've had to revise based on that?
很多。
Lots.
我会告诉你，其中一件很有趣的事情是，你以为，好吧。
I would tell you, one of the things that, you know, is is very interesting is that you think, okay.
我有了这项很棒的新技术，这是能真正解决重要问题的突破。
I have this great new this technology that's the breakthrough that really solves an important problem.
因此，我可以立即进入商业世界，围绕它建立一家公司。
Therefore, I can rush out into commercial world and build a company around it.
但事实证明，技术只是支持它的很小一部分核心，而围绕它的还有很多其他同样重要、甚至更重要的因素。
Well, it turns out that technology is only a very small core part it enables, but then there's all these things that have to come around it that are equally, if not more important.
当你去见客户，对他们说，嘿。
And, actually, when you go to customers and you say, hey.
我们有了这个新的AI技术。
We have this new AI thing.
他们会说，等等。
They're like, wait a second.
我不关心那个。
I don't care about that.
它能为我节省多少钱？
How much money is it gonna save me?
他们只关心这个。
That's all they care about.
是的。
Yeah.
而且这正是
And that's
这就是商业。
that's business.
这就是商业。
That's business.
我两位祖父都是做商业的。
Both my grandfathers are business.
我们是企业家，我父亲也是。
We're entrepreneurs, and so was my father.
嗯哼。
Mhmm.
所以我从小就在这样的环境中长大，这很艰难。
So I I grew up in these kind of environments, and it's tough.
外面的世界很难。
It's tough out there.
一位祖父在电子行业非常成功，另一位祖父从事房地产业，建造房屋，但我的父亲却很挣扎。
One grandfather was very successful in electronics, and my other grandfather was in the housing business, building homes, but my father struggled.
他是一名冶金学家，经营一家镀铬公司，生意非常非常艰难。
He was a metallurgist, and he had a company doing chrome plating, and it was very, very difficult.
你知道，他受到了许多他无法控制的因素影响，比如七十年代的经济衰退严重打击了他的生意，所以他一直很艰难。
And, you know, he was buffeted by things way behind his control, like, you know, recession of the seventies actually really hurt his business very badly, so he struggled.
所以有很多因素，这与竞争和时机有关。
So there's a lot of factors, and it has to do with competition and timing.
我还要补充的是，在行业中，这又回到了数据层面，即你在实验室里能做的事情，可能你以为已经充分探索了问题的全部范围。
What I would also say in industry is that and this is gonna come back to the data aspect, which is that you can do things in a lab that you think are you've really explored the full range of a problem.
让我举个例子。
So let me give you this example.
我们正在解决抓取分拣的问题。
We are addressing the bin picking problem.
还记得吗？
Remember?
是的。
Mhmm.
我们往里面扔了各种各样的物品。
And we were dropping all kinds of objects in there.
事实上，当有人来实验室参观时，我会问：你的车钥匙在哪？
In fact, when people would come to the lab, they would visit, and I'd say, well, where do you have your car keys?
把它们扔进来。
Drop them in here.
我说，如果机器人能把它捡出来，我们就把车留下。
I said, if the robot will pick it out, we'll keep the car.
但结果它总是能做到。
And then, but it would always do that.
捡出别人的车钥匙根本不是问题。
It was no problem picking out someone's car keys.
对吧？
Right?
我们尝试了各种各样的东西。
And we tried to draw all kinds of things.
又是玩具。
Again, toys.
我们用3D打印出了各种奇形怪状的物体，所有能想到的东西都试了。
We made three d printed weird shaped objects, all kinds of things we could think of.
我们试图全面考虑各种情况，就是想突破极限。
We try to basically consider everything, and we were just trying to push the envelope.
对吧？
Right?
嗯哼。
Mhmm.
嗯，‘突破界限’是关键，因为结果发现我们从未真正实验过袋子。
Well, the envelope was the keyword because it turned out that one thing we didn't ever really experiment with was bags.
哦。
Oh.
在物流中，袋子非常常见。
And bags are extremely common in shipping.
是的。
Yeah.
你可能知道，如果你从电商、亚马逊或其他地方收到过包裹，就会拿到各种各样的袋子。
You probably you know, if you know this, if you receive bags from your from ecommerce, from Amazon, or others, you get bags of all kinds of forms.
袋子通常是塑料或纸制的，但问题在于它们很松散。
Now bags are often plastic or paper, but the problem the issue with bags is they're loose.
它们里面装着物品，但有很多松垮的部分，而且会以各种有趣的方式折叠。
And so they have objects in them, but there's a lot of slack, and they tend to fold in interesting ways.
所以我们实际上在实验室里并没有测试这些。
So we didn't we weren't testing those really in the lab.
这并不是我们之前会太多考虑的事情，但在实际运输中，这种情况要普遍得多。
That wasn't something that we would have thought about too much, but that's so much more common in real shipping.
所以我的意思是，我们必须调整我们所有的系统，以适应消费者市场的现实，而这里的现实就是袋子。
So my point is we had to adapt all of our systems to the reality of the consumer market, which in this case is bags.
而这一点是我们之前数据非常缺乏的。
And that was something we didn't have a lot of data on.
所以我们不得不调整我们的系统，使其能在真实的袋子数据上运行。
So we had to adapt our systems to work on data that on real bags.
而真实的袋子非常难以模拟和建模，因为它们会再次折叠。
And real bags are very difficult to actually even simulate and model because they fold again.
顺便说一下，折叠很重要，因为如果你用吸盘直接吸在折叠处，当你抬起时，折叠会展开，导致吸力丧失，物品就会掉落。
And by the way, the folding matters because if you go to pick up with a suction cup right on top of a fold, as you lift it, the fold will unfold and the suction will you'll lose the suction and drop the object.
对吧？
Right?
所以我们开始在这些机器人投入工作时收集数据。
So we started collecting data as we started putting these robots to work.
当我们的客户在Ambi将这些系统投入生产时，我们也达成协议，会保持这些系统的高性能水平，因为我们一直在持续监控它们。
So as our customers were putting these systems into production at Ambi, right, we also had an agreement that we would maintain these systems very at high performance levels because we are constantly monitoring them.
我们在伯克利的中央总部有一个仪表板，团队会密切关注所有正在运行的机器。
So we have a dashboard at the central headquarters in Berkeley where the team keeps an eye on every machine that's in operation out there.
因此，我们会收集每一次抓取操作的数据，包括发生了什么、耗时多久、是否掉落了物品、是否正确分类了物品，诸如此类的各种信息。
And so what we do is we get data on every single pick operation, what happens, how long it takes, whether it dropped the object, whether it was classified correctly, all kinds of things like that.
对吧？
Right?
嗯。
Mhmm.
我们利用这些数据来及时发现性能下降的情况，比如每小时抓取次数——这通常是衡量标准——一旦下降，我们就能及早发现，并联系客户，询问发生了什么问题。
And we use that so that we can immediately tell when the performance, let's say the picks per hour performance, that's how it's often measured, drops, we can spot that early and say and we call the company, we say, what's going on?
有什么变化吗？
Did something change?
摄像头被撞到了吗？
Did the camera get knocked?
吸盘是不是磨损了？
Is the suction cup getting worn?
所以我们一直在密切关注。
And so we're constantly on top of it.
其中一部分原因是我们对此感到非常自豪。
Part of it is that that's a source of big pride for us.
我们非常以客户为中心，希望确保我们的机器能够完全可靠地运行。
We're really customer focused, and we wanna make sure our machines work completely reliably.
是的。
Mhmm.
但这一做法带来的一个美好而惊人的副作用是，过去四年里，我们从所有真实环境中的实际系统中收集了大量数据，现在已累计拥有相当于二十二年的机器人运行数据。
But the nice, amazing side effect of this is that we've been able to collect data from all this real systems in real environments over the last four years, and we now have elapsed on twenty two years of robot data.
你还记得我之前提到的十万年吗？
Remember I talked about the hundred thousand years?
是的。
Yeah.
对。
Yeah.
不过现在我们已经有了二十二年的数据，从
Now we have twenty two years, though, from
开始。
the start.
明白吗？
Okay?
但这是真实的机器人数据。
But it's it's real robot data.
这非常宝贵。
It's extremely valuable.
它的质量很高。
It's a high quality.
这是数据的黄金标准。
It's the gold standard for data.
因此，我们现在正利用这些数据来优化我们的系统，使其更加高效、可靠，同时也让我们能够拓展到新的相关产品领域。
And so we're now using that to refine our systems and then make them better and more higher performing, more reliable, but also allowing us to now branch out into new related types of products.
因此，我们推出了一款名为 AmbiStack 的新产品，它能非常高效、紧密地堆叠箱子。
So we now have we introduced a new product called AmbiStack that stacks boxes very efficiently, very densely.
嗯。
Mhmm.
这是我们今年推出的新产品，第一批系统已经全部售罄。
And that's a new product that we sold out our first batch of these, these systems this year.
太棒了。
Amazing.
关于机器人数据，或者说你提到的填补这十万年空白的问题，对于一家试图克服这一难题的公司来说，由于数据根本不存在，它们是否必须构建大量真实的物理系统？回到手的例子，它们是否必须拥有大量真实的机械手和物理对象来进行这些操作？
On this idea of robot data or the covering this hundred thousand year gap that you're talking about, for a company that would be trying to overcome this because the data just isn't there, are they just having to construct a bunch of physical real world going back to the hand example, would they have to have a bunch of physical hands with just a bunch of physical objects to then just be doing this?
还是说，你认为我们可以在虚拟环境中模拟这个过程，以加速这一过程，或者两者结合？
Or is this something that you think we could simulate in a virtual environment to accelerate that speed or kind of a combo of both?
很好。
Good.
对于抓取来说，仿真效果相当不错，因为在这里，你只需要 fairly 准确地了解环境的几何结构。
So for grasping, it turns out simulation is it works pretty well because there, you just need to know the geometry of the environment fairly accurately.
然后是物体和夹爪的几何结构，这样你实际上可以相当好地建模。
And then the geometry of the object and the gripper, and then you you can actually model that fairly well.
所以抓取不仅仅是把物体从桌子上拿起来。
So not grasping is just lifting an object off a table.
明白吗？
Okay?
我们已经摆脱了料箱。
We're out of the bin.
是的。
Mhmm.
这和我们之前谈到的系鞋带完全不同。
That's very different than tying the shoe that we talked about earlier.
嗯。
Mhmm.
对吧？
Right?
在那里，我们发现很难很好地模拟这种行为。
There, it turns out that we can't simulate that so well.
正如我提到的，我们不知道如何建模和模拟在与该物体交互过程中发生的形变和微小作用力。
As I mentioned, we don't know how to model and simulate the deformations, the minute forces that are going on in the process of interacting with that object.
所以这是一个挑战。
So that's a challenge.
这有点微妙，我知道你的观众可能会说，戈德堡在说什么？
This is a little nuanced, and I know that your your audience might say, what is Goldberg talking about?
他说这无法解决。
He said this couldn't be solved.
现在他说这个问题是可以解决的。
Now he says it can be solved.
嗯，这要看情况。
Well, it depends.
有一些类别的问题是可以解决的，我认为在过去五年里，从杂乱箱子中抓取物体方面我们取得了巨大进展。
There's certain categories of problems that can be addressed, and I think that picking objects out of a bin is something we've made enormous amount of progress in the last five years.
嗯。
Mhmm.
因此我对这一点非常乐观。
So I'm very optimistic about that.
我认为我们正在变得更快、更可靠，这些系统——你知道的，这是机器人技术的前沿，而且是实实在在的。
I think we're gonna get we're getting faster, more reliable, and those systems are you know, that's the cutting edge of robotics, and it's real.
嗯。
Mhmm.
是的。
Yeah.
但系鞋带、在家中或工厂里完成一些任务，比如组装电子零件、汽车车身，或者在汽车内部安装内饰和线路，这些都非常困难。
But then tying shoes and doing things around a home, by the way, or in a factory where you're actually trying to put together, you know, electronic parts or car bodies or car installing upholstery and wiring inside a car.
这些任务极其困难，无论是在底特律还是世界其他任何地方，这些工作仍然由人类完成，因为它们实在太难了。
These are extremely difficult, by the way, and they're even in Detroit or anywhere in the world, there's still humans doing those jobs because they're very, very hard.
所以这些任务很难模拟，我认为所有迹象都表明，形变是实现这些任务的关键障碍。
So those are hard to simulate, and I do think it's everything is pointing toward this deformation is a key obstacle to doing it.
我曾与一些物理学家和形变领域的专家交流过，他们都同意这一点。
And I've talked with with people who are physicists and experts in deformation, and they they agree.
这是一个非常、非常困难的问题。
This is a very, very hard problem.
我们甚至对摩擦和形变的物理机制了解得还不够透彻。
We don't even understand the physics of friction and deformation very well.
很有趣。
Interesting.
你说过你对人工智能创造力的看法已经改变了。
You've said your views on AI creativity have changed.
能给我们讲讲时间线吗？有哪些变化？以及你现在的整体看法。
Walk us through some of the timeline and what's changed and just kind of your overall opinion today.
好的。
Okay.
另外，我一直在从事艺术创作，这与我作为研究员和工程师的工作并行进行。
Well, on a very different note, I have been working as an artist as a in parallel with my work as a as a researcher and engineer.
我喜欢说，我的本职工作是在伯克利教书并管理一个实验室，但我还有另一个热情，那就是创作艺术。
You know, I like to say my day job is teaching at Berkeley and, and running a lab there, but I have another passion, which is which is making art.
艺术。
Art.
我从事这项工作的时间几乎和做研究一样长，主要创作装置艺术和项目。
And I've worked on this for almost the same amount of time, and I make installations and projects.
我们做过一个叫‘远程花园’的项目，用一个机器人通过互联网由人们远程控制，照料一个真实的花园。
We did a a project called the Telegarden where we had a robot that was controlled by people over the Internet, and the robot could tend to garden, a living garden.
我们在1995年就把这个项目上线了，那正是互联网的早期阶段。
We put this online in 1995, which was the very early days of the Internet.
我非常自豪这个项目，因为它在线运行了九年。
And I'm very proud of that project because it stayed online for nine years.
每天24小时，人们都可以进来探索这个花园，播种并给植物浇水。
Twenty four hours a day, people could come in and explore this garden and plant seeds and water them.
所以这是一个非常有趣的项目，它既是艺术创作，也是工程概念的验证，而且必须稳定运行。
So it was a very interesting it was an artistic project, but it was also a engineering proof of concept, and it and it had to work reliably.
因此，它真的对我们提出了很高的要求。
And so, you know, it really pushed us.
我有时会说，人们以为做工程很难。
I sometimes say people think, doing engineering is hard.
试试做艺术吧。
Try art.
那才真的很难。
It's really hard.
因为你必须面对公众，他们会互动，做出各种疯狂的事情。
Because you have to deal with the public, and they're gonna interact and do all kinds of crazy things.
所以我们不得不花大量时间设计这个系统，但我一直对艺术保持着兴趣。
So we had to really spend a lot of time designing that system, but I continue an interest in art.
我即将举办一场新展览。
And I have a new show coming up.
这是我和我妻子蒂芙尼·施兰（Tiffany Schlain）的联合项目，她是一位艺术家，我们正在合作筹备一个将于2022年1月在旧金山开幕的展览。
It's a joint project with my wife, Tiffany Schlain, who's an artist, and she and I are collaborating on a on a exhibition that's gonna open in San Francisco in January 22.
好的。
Okay.
明白了。
Alright.
这一直是我非常热衷的事业，我运用人工智能和机器人等技术来探讨关于科技本身的问题。
So so this is a big passion of mine, and it's using technology like AI and robots to ask questions about technology.
嗯，是的。
Uh-huh.
我对数字世界与自然世界之间的对比非常感兴趣。
And I'm very interested in this contrast between the digital and the natural world.
嗯。
Mhmm.
嗯。
Mhmm.
当它们看起来非常对称和相似时，但它们之间却有着深刻的差异。
When they they seem very symmetric and similar, but there's very profound differences between them.
所以这就是我思考或试图在我的艺术作品中表达的内容。
So that's what I I think about or I try to express in my artwork.
所以你关于创造力的问题。
And so your your question about the creativity.
是的。
So Yeah.
我总是说，你知道，AI不会像人类那样具有创造力，你可以向它提问，但它不会真正提出原创性的想法。
I always said, you know, AI won't be creative in the sense that it will you can ask it questions, but it's not gonna actually come up with original ideas.
但我实际上已经改变了我对这一点的看法。
But I actually have shifted my view on that.
是的
Yeah.
我举个例子，早期我问过Chatuchipiti，嘿。
And I give this example where I asked Chatuchipiti in the early days, hey.
给我列出一百种吉他拨片的用途。
Give me a 100 uses for a guitar pick.
我当时以为它会重复同样的内容，一遍又一遍。
And I just thought it would get you know, it started repeating the same thing over and over again.
但它确实开始列举了，比如用螺丝刀刮车窗上的冰之类的，这些都很合理。
And it did you know, it started with a screwdriver to scrape ice off a windshield, things like that, which are all made sense.
但接着它开始飞快地列出更多点子，快到我甚至都来不及读完。
But then it started listing these, like, as fast as I could read them or faster.
然后它提出了一个我没想到的：一个玩具船的小帆。
And then it came up with one that I was like, it was a a miniature sail for a toy boat.
当我看到这个时，我心想：天哪。
And when I saw that, I was like, oh my god.
这是个天才的想法，我根本想不到。
That is a genius idea, and I would not have thought of it.
我的意思是，当你看到如此原创和富有创意的东西时，你立刻就能察觉到，然后会说：啊，我怎么就没想到呢？
And I I mean, you know, immediately, when you see something that's original and creative like that, you spot it and you say, ah, why didn't I think of that?
这些是罕见的想法，而AI现在能够做到这一点。
Those are those rare ideas, and AI is capable of that now.
所以这非常令人兴奋。
And so it's very exciting.
是的。
Yeah.
确实令人兴奋。
It is exciting.
太令人兴奋了。
It's super exciting.
所以我对AI一点也不持负面态度。
So I'm not in I'm not negative about AI at all.
我非常喜欢它。
I love it.
我一直在使用它。
I use it.
我为它代言。
I advocate for it.
我的女儿们、我妻子，所有人都在用它。
My daughters, my wife, everyone uses it.
所以我完全支持它。
And so I'm a 100% for it.
我认为它会对机器人技术有所帮助，但问题是，它能否实现人们所期望的所有功能？
I do think it's gonna help with robotics, but the question is, is it gonna do everything that people are hoping?
我希望这场对话，普雷斯顿，能为你观众理清这些事情的背景。
And that's that's where I I hope that this conversation, Preston, will put things into context for your audience.
所以我不知道你是否会喜欢这个问题，但我还是要提出来，因为我很好奇你会怎么想。
So I don't know if you're gonna like this question or not, but I'm gonna throw it over because I'm curious what you would think of this.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。