Tesla Just Released Insane Bot Update

Hey everybody Rob Manward here welcome back to Tesla Daily. Today we have an exciting update on the Tesla Bot or Optimus to talk about. Also got updates on Gigatexus, the cyber truck and a few other items as well.

大家好，我是罗布·曼华德，欢迎回到《特斯拉每日快报》。今天我们有关于特斯拉机器人（或称为Optimus）的令人兴奋的最新进展要讨论。同时，我们还有有关Gigatexus、Cybertruck以及其他几个项目的更新。

Not a bad start to the week 4 Tesla stock finishing up about 9.10% on the day today, closing at $246.99. The NASDAQ up just a little bit less than half a percent on the day.

这周并不是一个糟糕的开局，特斯拉股票收盘时上涨了约9.10%，收盘价为246.99美元。纳斯达克指数当天上涨了不到0.5%。

Alright so the big news for today is an update on Optimus or the Tesla Bot program from Tesla directly. They've posted a video that's just over one minute long demonstrating Tesla's current capabilities with Optimus. So kind of an out of the blue update but really nice to see the progress and see the developments that Tesla has made.

好的，今天的重要消息是特斯拉直接关于Optimus或特斯拉机器人计划的更新。他们发布了一个稍长于一分钟的视频，展示了特斯拉与Optimus的当前能力。这是一个突然的更新消息，但很高兴看到特斯拉取得的进展和发展情况。

A number of which are demonstrated here including Optimus's capability to sort individual objects autonomously. Which might only look cool if you actually understand what is going on here because it's pretty crazy. So Tesla on X says that this is using a neural network that is fully end-to-end trained. So video in to the neural network and then controls out. So just like FSD version 12.

其中一些在这里进行了展示，包括Optimus能够自主地对个别物体进行分类的能力。如果你真的理解这里发生的事情，可能会觉得这很酷，因为它相当疯狂。所以特斯拉在X上说，这是使用了一个完全端到端训练过的神经网络。视频输入到神经网络，然后输出控制。就像完全自动驾驶12版一样。

Tesla's senior staff engineer for the Optimus program on X shared quote super excited to share our team's progress over the last few months. We can now accomplish long horizon tasks such as sorting blocks fully autonomously in a completely task agnostic way. Just collect more data and we can train a new complex task without changing any code end quote.

特斯拉(X车型)的优化计划高级技术人员非常兴奋地分享了我们团队在过去几个月里取得的进展。我们现在能够完全自主地以完全任务无关的方式完成长期任务，例如对方块进行排序。只需收集更多数据，我们就能够训练一个全新的复杂任务，而无需修改任何代码。

So we see this demonstrated in the video first with the Bot sorting different color blocks into different colored trays which is already impressive enough because this is the bot responding to a prompt to do that rather than being explicitly coded to complete that specific task. That is a fundamental difference that dramatically changes the adaptability of Tesla Bot to be able to handle different situations.

所以，首先在视频中我们看到了这一点，机器人将不同颜色的积木分别放入不同颜色的托盘中，这已经很令人印象深刻，因为机器人是通过响应指令进行操作的，而不是被明确编程完成特定任务。这是一个根本的区别，它极大地改变了特斯拉机器人适应不同情况的能力。

So for example in this video not only does the Tesla Bot sort these blocks by color it also is able to then unsort them grab them from the trays and put them back. That is again not any explicit code to do that it's just given a prompt and then it completes it. This is again noted by Tesla staff engineer saying quote yes it can be prompted the same network is used in this video to sort or unsort the blocks you just tell it what you want and quote.

因此，举个例子，在这个视频中，特斯拉机器人不仅能够按颜色对这些块进行排序，还能够解开排序，从托盘中取出它们并放回去。这并不需要明确的代码来完成，只需要给出提示，然后它就会完成。特斯拉工程师也指出了这一点，他说：“是的，它可以通过提示来完成相同的网络被用于在这个视频中对块进行排序或解开排序，你只需要告诉它你想要什么。”

Now the prompt may not necessarily be natural language today but with large language models like we have seen with chat GPT obviously that's something that's achievable and then to complete the prompt Optimus is obviously going to have to be trained on that task but that is what Tesla is demonstrating here that that is also achievable.

现在，提示可能不一定是自然语言，但像我们在ChatGPT中看到的大型语言模型一样，很明显这是可以实现的。为了完成该提示，Optimus显然需要接受训练，但特斯拉正在展示这也是可行的。

If a wide variety of tasks can be achieved such as sorting the blocks, unsorting the blocks etc etc with the exact same neural network with just more training you can start to see how quickly this could become something that is highly highly effective at a wide variety of different tasks.

如果只需通过更多训练，我们就能够用同一个神经网络来完成各种不同的任务，比如对方块进行排序、解除排序等等，你就能明显看到这有多么迅速地变成一种非常高效的多功能工具。

If and when that starts to be the case which I think will be a lot quicker than people realize then all of a sudden there is a market for this product maybe one company needs it to sort maybe another company needs it to unsort maybe yet another company needs it to do some other similarly difficult task and on and on and even if those companies don't necessarily need it for versatility specifically for their own use cases the fact that that versatility exists within the same product within the same code base makes it extremely scalable which at the same time makes it much more affordable through economies of scale and that only increases the possible market. If things go well that can quickly become a pretty viciously positive cycle and just like FSD optimists is something that's going to get better over time over the air.

如果并且当这种情况开始发生（我认为比人们意识到的要快得多），突然之间就会有一个市场需求，也许有一家公司需要它进行排序，另一家公司需要它进行反排序，也许还有另一家公司需要它执行其他类似困难的任务，如此反复。即使这些公司不一定特别需要这种多功能性来满足自己的用例，但是这种多功能性存在于同一产品和同一代码库中，使得它具有很高的可扩展性，同时也使得它通过规模经济变得更加经济实惠，这进一步增加了可能的市场。如果事情进展顺利，这可能很快就会成为一个相当积极的循环，就像全自动驾驶（FSD）的乐观派一样，它会随着时间的推移而不断变得更好。

So let's just make up an example let's just say you buy a Tesla bot and you've got it you know sitting in your living room and it's sorting blocks on day one maybe a month later it's sorting laundry maybe a month after that it's loading and unloading laundry next it's loading and unloading the dishwasher next it's doing dishes and so forth from there all with the same hardware all with really the same software but with more training as Tesla collects that data and not only collects it but also trains it through the Nvidia cluster or through dojo.

所以让我们举个例子吧，假设你买了一台特斯拉机器人，你把它放在客厅里，第一天它可以帮你整理积木，一个月后它可以帮你整理衣物，再过一个月它可以帮你装卸衣物，接着是装卸洗碗机，然后是洗碗，以此类推。所有这些都是使用相同的硬件、几乎相同的软件实现的，只不过通过特斯拉收集的数据进行更多的训练，而且不仅仅是收集数据，还通过 NVIDIA 集群或者道场进行训练。

So it's hard to overstate how exciting it is to hear that optimists is now running on an n10 neural network and is able to complete tasks like this no matter how simple they may be because it shows what can be done on this architecture. In this video, it actually demonstrates a lot. It demonstrates really good dexterity in the hands of optimists and seems to have a good understanding of pressure based on how it interacts with those blocks. Which is even more impressive when you again remember that this is n10 neural networks. There's nothing in the code telling it how much pressure to apply. It's just learned of that through observation.

很难言过其实，乐观主义现在运行在一个n10神经网络上，并且能够完成诸如这样的任务，无论它们多么简单，这真是令人兴奋的消息，因为它展示了在这种架构下可以做到什么。在这个视频中，它确实展示了很多东西。乐观主义在手部的灵活性方面表现得非常好，并且似乎对压力有着很好的理解，根据它与那些方块互动的方式可以看出。当你再次记住这是n10神经网络时，这更令人印象深刻。没有任何代码告诉它要施加多少压力。它只是通过观察学会了这一点。

The bot also shows really good coordination and balance. We've seen that come a long way in a short period of time and that's really fun to see. If you watch the video closely, you can see how all the different pieces of the bot interact to keep those things going, to keep it smooth. Even just using its arms to move these blocks around, you can see the motion in the legs, you can see the motion in the hips. All of those things subtly working together, just like they would for a human.

这个机器人还展示了非常好的协调和平衡能力。我们在短时间内观察到了它在这方面的长足进步，非常有趣。如果你仔细观看视频，你会发现机器人的不同部分是如何相互作用以保持平稳运行的。即使只是利用它的手臂移动这些方块，你也可以看到腿部和臀部的动作。所有这些细微的动作都像人类一样协调运作。

Of course, the perception is obviously very exciting here. Sorting these blocks by color, though the blocks are different shapes, you might notice. But that shows nice perception. They also have a Tesla engineer moving the blocks around and the bot reacts to that very smoothly, no problems at all. There's also a time where one of the blocks falls on its side and the bot actually corrects that, which presumably it's correcting autonomously without being told to do so because it knows from training that that's not how it's supposed to be. Just a little one-second part of the video, but again pretty mind-blowing if you think about it closely. So a lot of good things there with perception.

当然，这里的感知显然非常令人兴奋。尽管这些方块形状各异，但通过按颜色分类，你可能注意到它们之间的联系。这展示了很好的感知能力。他们还有特斯拉的工程师移动着方块，而机器人对此的反应非常流畅，没有任何问题。还有一段时间，其中一个方块倒在了一侧，机器人实际上纠正了这个问题，可以推测它是自主纠正的，而不是被告知这样做，因为它通过训练知道这并不是正确的方式。视频中仅仅是一个一秒钟的片段，但如果你仔细思考，会发现这非常令人惊叹。所以，感知方面有很多好的特点。

Then there's also proprioception, which is the other quote unquote sensory input outside of vision for the Tesla bot. So this has to do with the state of all the different actuators and parts for Tesla bot, which, making it analogous to a human, would be its sense of its own body. For example, with a human, you can close your eyes or be in the dark or whatever and still interact with the things around you, still have a sense of where things are at. That's proprioception and that's again called out by Tesla staff engineer as a part of what we're seeing here. So again, it looks easy, but that's going into what we're seeing with the coordination, the balance, the dexterity. There's a degree of self-awareness that goes into all of those things.

然后还有所谓的本体感觉，这是特斯拉机器人除了视觉之外的另一种感官输入。这与特斯拉机器人的所有不同执行器和零部件的状态有关，类似于人类对自身身体的感觉。例如，对于人类来说，即使闭上眼睛或处于黑暗中，仍然能与周围的事物进行互动，并且有一种感知物体位置的能力。这就是本体感觉，特斯拉工程师也将其作为我们在这里看到的部分加以描述。所以，看起来很容易，但这涉及协调、平衡和灵巧性等方面。所有这些都需要一定程度的自我意识。

So really fun and exciting to see these developments. I also think it's worth a reminder of some of Elon's comments on the Q2 earnings call. He said that Tesla would have all in-house production candidate actuators ready to go for a Tesla bot in November, so we're not quite there yet but pretty soon. And Elon said that they would start ramping production after that. That may just be for the actuators but may also be for the Tesla bot. Elon said that he believes that.

这些进展真是有趣而令人兴奋。我认为有必要提醒一下埃隆在第二季度财报电话会议上的一些评论。他说特斯拉将在11月份准备好所有内部生产的候选执行器，以供特斯拉机器人使用，所以我们离实现这一目标还有一段时间，但很快就会到来。埃隆还表示，在那之后他们将开始增加生产。这可能只是针对执行器，但也可能适用于特斯拉机器人。埃隆相信这一点。

they'll have optimists doing something useful in the factories next year all right let's a couple of things on this of course there were some people on x speculating that there was some cgi involved in this again tesl engineer commenting saying you know you're doing cool stuff when some people think that what you're doing is cgi so presumably none of that here i think maybe people getting a little bit mixed up by the speed of the clips which tesl does note is at one and a half time speed so a 50 increase which is clearly noted in the video and makes a lot of sense to do if you're demonstrating something like moving some blocks around especially in the ever shorter attention span society that we seem to be living in

明年他们将会在工厂里有些乐观主义者做一些有用的工作，没错的，在这个问题上我们来看一下其中的一些事情，当然有一些人在猜测可能有一些计算机生成图像的参与，再次由特斯拉的工程师评论说当一些人认为你所做的是计算机生成图像时，你就知道你正在做非常酷的事情了，所以这个视频中可能没有用到这些技术，我认为可能有些人对视频中的快进片段感到有些混淆，特斯拉在视频中明确指出了1.5倍速的加快，这在展示移动一些方块等动作时是非常合理的，特别是在我们似乎生活在越来越短的注意力时间里的社会中。

all right moving on from tesl bot we've got a couple updates on kicketexis from joe tagmire he is saying that all employees are now back at work today doesn't mean that production is fully back up to speed immediately but sounds like that's expected to be the case next week. so that is nice to see and then joe also spotted a couple of new cyber trucks out there one of which looks like has plastic wrap on the seats which i think is obviously a good sign that we may be getting closer to vehicles that are prepped for delivery there was kind of an interesting thing that happened over the weekend with the cyber truck reservations that people had sounds like early reservation numbers are no longer seeing their specific configuration on their reservation whereas later reservation numbers can still see that doesn't necessarily mean anything but maybe tesla in there playing around with some things which speculatively could mean some things being put in place to convert those reservations to orders but we'll just have to wait and see also at kicketexis today the president of hungry announced that she met with elon musk at the factory sounds like they initially intended to meet at the budepest demographic summit but that didn't work out so kicketexis instead not a whole.

好的，从特斯拉机器人出发，我们有一些有关Kicketexis的更新，乔·塔格迈尔说，今天所有员工都回到了工作岗位，虽然生产速度不会立即完全恢复，但下周预计会有所改善。这是一个好消息。此外，乔还看到了一些新的Cybertruck，其中有一辆座椅上还有塑料包装，这显然是一个好迹象，可能意味着我们离准备交付的车辆越来越近。在周末发生了一个有趣的事情，与Cybertruck预订有关，听起来早期预订的人再也看不到他们预订的具体配置了，而后期预订的人仍然可以看到。这并不一定代表什么，但也许特斯拉正在做一些调整，可能意味着他们可能正在采取措施将这些预订转换为订单，但我们只能等待看看。另外，今天在Kicketexis，我们饿了的总统宣布她与埃隆·马斯克在工厂会面，听起来他们最初计划在布达佩斯人口统计峰会上会面，但那没能实现，于是他们选择了Kicketexis。

lot more information here but the president did say that there would be more details tomorrow nevertheless elon's meetings with world leaders continue next up we've got a quick update on tesla's vpp or virtual power plant program in california i missed this last week but tesla has expanded the vpp program to san diego so powerwall customers there can now sign up for the vpp if they would like and potentially earn a little bit of revenue from doing so

这里提供了更多的信息，但总统确实表示明天将有更多详细信息。尽管如此，埃隆与世界领导人的会议仍在继续。接下来，我们将简要更新特斯拉在加利福尼亚的虚拟电力厂计划(VPP)。上周我错过了这个消息，但特斯拉已经将VPP计划扩展到了圣地亚哥，所以那里的Powerwall用户现在可以选择加入VPP计划，并有可能从中获得一些收入。

all right next up we've got a couple updates on other automakers particularly ford last week we had talked about the status of negotiations for the auto workers union in canada with ford those negotiations over the weekend were officially completed as the union has voted to ratify an agreement with ford for a three-year contract which will include a 10% wage increase in the first year and then two and three percent increases in the second and third year among various other terms of the contract so as we had talked about ford actually avoided at least for the time being further strikes in the u.s as the uaw said that they were making a decent effort and they're now the first to reach an agreement in canada though the negotiations there seem to be a little bit more sequential rather than all at once like we are seeing the uaw employee in the u.s.

好的，接下来我们将会有一些关于其他汽车制造商的更新，特别是福特公司。上周我们讨论过加拿大汽车工人工会与福特的谈判状态，这些谈判在周末正式结束，工会投票通过了一项为期三年的合约协议，其中包括第一年的薪资增长10%，第二和第三年分别增长2%和3%，同时还有其他各种合同条款。所以，正如我们之前所说，福特在美国实际上避免了进一步罢工，因为UAW表示他们在做出不错的努力，并且他们成为了首个在加拿大达成协议的企业。尽管加拿大的谈判似乎更加顺序进行，而不像美国UAW雇员一次性达成协议。

In the U.S., we do have a report from The Detroit News today that Ford has announced that they have halted the construction of the three and a half billion dollar battery factory project that they had announced back in February. We talked about this at the time. This was the one that was going to be done in partnership with CATL for lithium iron phosphate batteries in Michigan. At the time, Ford had said that this would produce approximately 35 gigawatt hours of batteries per year, enough for about 400,000 Ford EVs. At the time, Ford said that this would begin production in 2026. As to the current status, with this update today, a Ford spokesman told The Detroit News that, "We're pausing work and we're going to limit spending on construction at Marshall until we're confident about our ability to competitively run the plant." Alright, I mean certainly not the most inspiring comment there. And this is not the first time, even in the last few months, that we have seen a pretty significant adjustment in their guidance around their EV business. Remember, back in July, they updated their 600,000 production target from this year to next year. They guided for greater losses on the EV business this year and they also changed their 2 million vehicle production target, which was previously 2026, to whenever they're not sure now. So now, after only a couple of quarters after making this announcement, and by the way, talking about how many jobs this created, not even, well, create, this creates 2,500 jobs, they say now it's paused before probably even really getting started, and they're not confident in their ability to run it competitively. Now, maybe this relates to the UAW negotiations. Obviously, the timing is interesting in that regard. This could even be a negotiation tactic in some way, or maybe a repercussion of how the negotiations are unfolding. So, pretty big deal here, and this will be an interesting one to keep an eye on. The spokesperson also said that, "We haven't made a final decision about the investment there." So presumably more to come on this, and I would imagine that this is a topic for the next earnings call if there's not more information out before then. Alright, last item for today, just kind of a quick and subtle update from SpaceX on the number of Starlink subscribers. Starlink on X posted that they now connect more than two million active customers and counting with high-speed internet. So congratulations to SpaceX for crossing that two million customer threshold, right around nine months after that one millionth customer mark was surpassed last December. Alright, that'll wrap it up for today then. As always, thank you for listening. Make sure you're subscribed and signed up for notifications. And also find me on X at Tesla Podcast. And we'll see you tomorrow for the Tuesday, September 26th episode of Tesla Daily. Thank you.

据美国的《底特律新闻》报道，福特公司今日宣布已暂停早在今年二月份宣布的耗资35亿美元的电池工厂项目的建设。我们在当时就进行了讨论。这个项目计划与中国的宁德时代合作在密歇根州生产磷酸铁锂电池。当时，福特表示该工厂每年将生产大约35吉瓦时的电池，足够用于约40万辆福特电动汽车。福特当时表示该工厂将在2026年开始生产。至于目前的进展情况，根据今日的最新消息，福特发言人告诉《底特律新闻》:“我们正在暂停工作，并限制对马歇尔工厂的建设支出，直到我们对我们能否在竞争中运行该工厂有信心为止。”好吧，可以说这并不是最鼓舞人心的评论。而且，在过去几个月中，这已经不是福特首次对其电动汽车业务的指导方针进行相当大的调整了。记住，今年7月，他们将年产量目标从今年调整到了明年的60万辆。他们预测今年的电动汽车业务亏损会更大，他们还将之前定于2026年的200万辆汽车生产目标改为现在他们不确定的时间。因此，现在，在进行了几个季度之后，而且请注意，当时他们还谈到了这个项目创造了多少就业机会，但现在项目还没真正开始就已经暂停了，并且他们对能够竞争性地运营该工厂没有信心。现在，也许这与美国汽车工人联合会的谈判有关。从时间上看，这是很有趣的。这甚至可能是某种谈判策略，或者是谈判进展的后果。所以，这是一个相当重要的问题，我们将继续关注。发言人还表示:“我们还没有就在那里的投资作出最终决定。”因此，我们可以预计这方面还会有更多信息，我相信这会成为下一次财报电话会议中的一个话题，如果在那之前没有更多信息的话。好的，今天的最后一个消息，SpaceX关于Starlink订户数量的一个快速而微妙的更新。Starlink on X发布消息称，他们现在已经连接了超过200万个活跃用户，并提供高速互联网服务。因此，祝贺SpaceX成功突破两百万订户门槛，大约在去年十二月突破一百万订户后的九个月左右。好的，今天就到这里，感谢您的收听。请确保订阅并开启通知，并在X上找到我，我的用户名是Tesla Podcast。我们明天再见，收听9月27日星期二的《特斯拉日报》。谢谢。