Tesla's Most Important Weekend Ever ⚡️
发布时间 2023-08-28 21:33:54 来源
On Thursday last week, I had a conversation with a friend about autonomy. He didn't know about crews, Waymo, or Tesla's full self-driving, but he was adamant full self-driving cars are a decade plus away.
上周四，我与一个朋友谈论了自动驾驶技术。他对实现全自动驾驶的机构如乘务组、Waymo和特斯拉的完全自动驾驶系统都不了解，但他坚信要实现全自动驾驶还需要十年以上的时间。
I share this because over the weekend, Elon and Tesla showed us a real breakthrough in their full self-driving technology and strategy that has dramatically changed my view on when and how Tesla will solve for autonomy. The public of course is still clueless and now even more so.
我分享这个是因为在周末，埃隆和特斯拉向我们展示了一项真正的突破，他们的全自动驾驶技术和战略大大改变了我对特斯拉何时以及如何解决自动驾驶问题的看法。当然，公众仍然一无所知，现在甚至更加不明白了。
Even if you've been following along all weekend, we need to reiterate certain points and I want to make sure everyone understands and commits to long-term memory what happened this weekend. The importance of what we learned cannot be overstated.
即使你一直在整个周末都有在跟进，我们仍需重申某些要点，我希望确保每个人都理解并将这个周末发生的事情牢记在长期记忆中。我们所学到的重要性无法被过分强调。
First, let's just acknowledge the transparency of Elon and the team. To hop on a space to answer unscripted questions with unscripted answers about Tesla's FSD technology and then to live stream an alpha build of FSD12 for millions to see while Elon and Ashok narrated with incredible insight. This is just so cool for so many reasons and a serious departure from what any legacy auto company CEO has ever done.
首先，我们应该承认埃隆和团队的透明度。他们乘坐太空回答有关特斯拉完全自动驾驶技术的未预先准备的问题，并现场直播了FSD12版本的测试，亿万观众亲眼目睹，而埃隆和阿肖克则提供了非常深入的解说。出于种种原因，这一切非常酷，也是与任何传统汽车公司CEO从未尝试过的严肃变革。
So FSD12. We've heard it's AI end-to-end photons in control out. Neural nets being fed high-quality video data to learn exactly like a human would, removing all heuristics and nearly all manual or handwritten C++ code. With V12, the car does not know what lane lines are, what stop signs are, what pedestrians are, but yet it intuitively drives smooth like a human because of the vast video data the AI has been fed over the years.
所以FSD12。我们听说它是AI端到端光子控制。神经网络被提供高质量的视频数据，像人类一样学习，摒弃了所有的启发式算法，几乎没有任何手动或手写的C++代码。通过V12，汽车并不知道车道线是什么，停车标志是什么，行人是什么，但却能像人类一样直观地平稳驾驶，因为多年来AI获得了大量的视频数据。
So now solving for problems is just a matter of growing the fleet from Tesla's 4 million today that will hit 10 million soon because all of these vehicles can gather data that Tesla then curates to pick the highest quality or the best drivers to then train the AI rather than handwriting code or instructions for what the car should do in every situation.
现在解决问题只是一个扩大特斯拉车队的问题，目前特斯拉已经拥有400万辆车，不久将达到1000万辆。因为所有这些车辆都可以收集数据，特斯拉会对这些数据进行整理，选择出最高质量或最优秀的驾驶员，然后通过训练人工智能来指导车辆在各种情况下应该做什么，而不是手写代码或指令。
And this does not mean the system is uncontrollable on the contrary actually. It's just now Tesla programs the data instead of programming the code. It's all about data curation and then updating the weights. Large amounts of mediocre data don't improve driving. That would actually make it worse. Tesla does still use normal software, C++ and Python, to decide what data to select from the fleet, but human labelers do have the final say on what data is good enough.
这并不意味着这个系统是无法控制的，相反，实际上它是可控的。只是现在特斯拉通过编程数据而不是编程代码。这涉及到数据的筛选和权重的更新。大量的普通数据并不能改善驾驶。实际上，这会使情况变得更糟。特斯拉仍然使用普通软件，如C++和Python，来决定从车队中选择哪些数据，但人工标注者对数据是否足够好的最终判断权。
Tesla will also begin shipping models to cars in shadow mode so every time the driver does something differently than the AI would have done. Tesla can get that data back which is more valuable than collecting random data. Every time there is an intervention it will automatically be uploaded to the training to update the weights. And Tesla's solution now does not need any map data and it will still work without any internet connection. Which of course is crucial as you need your car to drive safely even in areas without cell phone coverage.
特斯拉还将开始向在影子模式下的车辆出货，因此每当驾驶员的操作与AI可能选择的不同时，特斯拉都可以获得这些数据，这比收集随机数据更有价值。每次有干预行为时，它都会自动上传到训练中以更新权重。而且，特斯拉的解决方案现在不再需要地图数据，即使没有任何互联网连接，它仍然可以正常工作。当然，这是至关重要的，因为您需要您的车辆在没有手机网络覆盖的地区安全驾驶。
With this clip on the V12 drive the car pulled over to the side of the road to park without being told. It just knew from prior video data it's been fed that at the end of a route that's the human behavior. Find a safe place as close to the GPS pin as possible. This is also where we learn Tesla RoboTaxis will eventually be able to recognize you by using a picture if you want so the car can look and wait for you. It'll also be able to recognize the Starbucks logo and drop you off near the entry of your destination even without map data in parking lots.
有了这个剪辑，V12 驱动汽车会在没有被告知的情况下，拉到路边停车。它仅仅是从之前的视频数据中学到的，在路线的最后，人们通常会这样做。找到一个尽可能接近 GPS 定位的安全地点停车。从这里我们也了解到特斯拉自动出租车最终将能够通过使用图片来识别您，以便车辆可以看到并等待您。它还将能够识别出星巴克的标志，并在停车场中甚至没有地图数据的情况下，将您送到目的地入口附近。
So Tesla could actually delete the entire navigation system and just give the car a GPS point and the car can navigate to it just like a human would even in location that's never been before.
特斯拉实际上可以删除整个导航系统，仅提供汽车一个GPS点，汽车就可以像人一样导航到那个点，甚至在之前从未去过的地方。
And this bike scene was pretty impressive. Again there are no explicit instructions or code telling the car how to handle a cyclist. The car is just reacting based on how other good drivers behave around cyclists.
而这个自行车场景非常令人印象深刻。再次强调，没有明确的指令或代码告诉汽车如何处理骑自行车的人。汽车只是根据其他优秀司机在自行车周围的行为做出反应。
Nothing but nets baby. And how about here where the car chose to lane with fewer vehicles without any code telling it to just acting like a human on its own. And yes there was one intervention but to focus on this one instance from an otherwise great smooth 45 minute drive is just disingenuous. There's a reason this has not yet been released to the public. Tesla will fix this by feeding it more video footage of intersections just like that. They'll update the weights and it'll be fixed.
仅仅是几根网，宝贝。这里车辆选择了一个没有编程指示的较少车辆的车道，就像一个自主行为的人类一样。是的，有一个干预，但将注意力集中在这一个例子上，忽略了一个平稳顺利的45分钟的行车，这是不公正的。这也是为什么这个功能还没有向公众发布的原因。特斯拉将通过提供更多类似情景的视频素材来修复这个问题。他们将更新权重并解决这个问题。
Now I'm hearing some folks say that well the competition is just going to copy what Tesla is doing. This is wrong. Elon has said the designing in the cameras, the liquid cooled Tesla computer, and the high security internet gateway will probably take major car companies three years. And that other car companies will probably want to use Tesla's online vehicle management system on the server side too. But this is only part of the solution. Elon has warned others LIDAR is not optimal for cars but no one has listened.
现在我听到有些人说竞争对手只是会照搬特斯拉的做法。这是错误的。埃隆曾经说过，设计摄像头、液冷特斯拉电脑和高安全互联网网关可能需要其他主要汽车公司三年的时间。而且其他汽车公司可能也会想在服务器端使用特斯拉的在线车辆管理系统。但这只是解决方案的一部分。埃隆已经警告过其他人激光雷达对于汽车不是最佳选择，但没有人听进去。
Elon also said they're well aware the Chinese like to copy Tesla's technology. The thing is Tesla designed its own hardware and that means their software will only run on that specific hardware. As with Apple you can't just copy the code, you need their hardware too. And I would add you also need a vast high quality data set as well.
埃隆（Elon）还表示，他们非常清楚中国人喜欢模仿特斯拉的技术。问题在于，特斯拉设计了自己的硬件，这意味着他们的软件只能在特定的硬件上运行。就像苹果一样，你不能只复制代码，还需要他们的硬件。此外，我还要补充说，你还需要一个庞大而高质量的数据集。
So tell me how is cruise going to get data with around 400 vehicles compared to Tesla's 4 million vehicles and growing that are being driven by customers every day. Having a vast fleet of cars for FSD training is a massive Tesla advantage. And remember Tesla designs its own silicon hardware in house. So how are any of these other companies going to get that? The answer is they can't unless they license it from Tesla.
告诉我，相比特斯拉每天由顾客驾驶的数量不断增长的400万辆车，Cruise如何获取数据？拥有庞大的车队进行全自动驾驶（FSD）训练是特斯拉的巨大优势。而且请记住，特斯拉自己设计了硅芯片硬件。所以其他公司如何获得这方面的支持呢？答案是，除非向特斯拉获取授权，否则他们无法获得。
So this notion that every other automaker is going to solve FSD when Tesla does is even more ridiculous now given what we've learned this weekend. Not only that but Tesla having its own chip design means they optimize for its usage down to the microsecond. And Tesla currently has drivers testing globally in places like New Zealand where there's inclement snowy weather. With version 12 the car will automatically increase following distances in the snow because you guessed it, it'll learn from video data of good drivers doing the same thing. Without ever needing to be told hey it's snowy out so do X, Y, and Z from manual code. Just like how on Elon's drive the speed was set to 85 but the car automatically new to only go 40 on roads that were not highways. It was purely intuitive. It slowed down for speed bumps and maneuvered cautiously around cyclists and pedestrians. It navigated a new construction zone and handled around about all without any code telling it to do so.
因此，这种认为特斯拉解决了全自动驾驶（FSD）之后其他汽车制造商也能解决的观点，根据我们在周末所了解到的情况，现在甚至更加荒谬。不仅如此，特斯拉拥有自己的芯片设计，意味着他们可以将其用到微秒级的精确度上进行优化。而且特斯拉目前在全球范围内进行着司机测试，例如在新西兰这样的天气条件恶劣的地方。在12版中，汽车将根据来自良好司机的视频数据学习，在雪天中自动增加车距。它不需要通过手动编码来告诉它：“嘿，外面下雪了，所以你要做X、Y和Z。”就像埃隆的驾驶一样，车速设定为85，但车辆自动降低速度至40，因为道路不是高速公路。这完全是直观的。它会减速通过减速带，并绕着骑车人和行人小心驶过。它能够导航通过新的施工区，并在环形交叉口处操作，而无需任何编码告诉它这样做。
Yes you need a vast amount of video training data and billions of dollars of training hardware and the know-how on how to run that hardware so this is definitely not easy. Speaking of that Tesla's AI team has a 10,000 unit H100 cluster going live today and this week. Elon said it's not a trivial thing to get this online. They have a competent team working 24-7 and it's still very difficult to get networking and cabling to work. These new H100 chips are designed for things like video training and are up to 30 times faster than the prior A100 chip. Tesla does actually own these as well. Each H100 costs around $30,000 so this cluster alone is $300 million.
是的，你需要大量的视频训练数据和数十亿美元的训练硬件，以及掌握如何运行这些硬件的技术，这绝对不容易。说到这一点，特斯拉的人工智能团队正在上线一个拥有10,000个单位的H100集群，本周开始运行。埃隆表示这并不是一件琐碎的事情，他们有一个能干的团队全天候工作，但网络和布线仍然非常困难。这些新的H100芯片是专为视频训练设计的，比之前的A100芯片快30倍。特斯拉实际上也拥有这些芯片。每个H100芯片的成本约为30,000美元，因此单单这个集群就价值3亿美元。
Many other companies that say they have these are actually renting them but not Tesla and it's the owning and maintaining that's the hard part. Elon talked about a silicon shortage this year and that's part of why Tesla is simultaneously bringing its in-house supercomputer dojo online. So it's going to be a combination of Nvidia and Tesla hardware for Tesla's training. The takeaway though is Tesla's compute capabilities are so far ahead of any other car company you still can't see second with a telescope. Tesla plans to spend $2 billion with a B this year on training compute alone to acquire hardware and this is going to be the case for a number of years and Tesla will probably spend even more than this next year.
许多其他公司声称自己拥有这些东西实际上是租赁的，但特斯拉不是，而且拥有和维护才是最困难的部分。埃隆今年谈到了硅短缺问题，这也是特斯拉同时上线自家超级计算机道场的原因之一。因此，特斯拉的训练将是Nvidia和特斯拉硬件的结合。然而，重要的是，特斯拉的计算能力远远超过任何其他汽车公司，即使用望远镜也看不到第二名。特斯拉计划今年单独在训练计算上花费20亿美元，并且在接下来的几年里这将是常态，特斯拉明年可能会花费更多。
Right now FSD is compute constrained and Tesla to no one's surprise is moving aggressively to solve for this. Let that sink in $2 billion on training hardware alone. Do you think Ford, GM, Waymo or Cruz can spend that much when they're still losing millions and billions on EV and autonomy already? Now yes Elon did talk about some challenges like an upcoming electricity generation shortage and transformer shortages as well as long lead times of one year for those transformers. And of course regulators are always a looming question mark.
现阶段，全自动驾驶（FSD）受到计算能力的限制，毫不意外的是特斯拉正在积极寻求解决方案。仅仅培训硬件就耗资20亿美元，让这个数字深入你的思考。当福特、通用、Waymo或Cruz在电动汽车和自动驾驶领域仍然亏损数以百万甚至数以十亿美元时，他们能否也投入如此巨额资金呢？的确，埃隆曾谈及一些挑战，比如即将出现的电力供应短缺、变压器短缺以及变压器长达一年的交货时间。当然，监管机构始终是一个悬念。
Speaking of how about that stat only 0.5% of people come to full stops at stop signs yet NHTSA is still forcing Tesla to artificially pull data from typically bad drivers to come to complete stops. This is just silly. In Elon said over time the percent of energy being used at data centers for neural net training will be 80 to 90 percent. So yes the world will be compute constrained for a while and there's plenty of room for both Nvidia and Tesla's dojo. Elon did say hardware 4 will lag hardware 3 by at least another 6 months as the focus is getting FSD on hardware 3 working super well and provided internationally.
说到这一点，只有0.5%的人在停车标志牌处完全停下来，但美国国家公路交通安全管理局仍然要求特斯拉从通常是糟糕驾驶者的数据中人为提取数据来完全停下来。这真是太可笑了。埃隆称随着时间推移，用于神经网络训练的数据中心能源使用的百分比将达到80%至90%。所以说在一段时间内世界将受到计算能力的限制，Nvidia和特斯拉的Dojo都有足够的发展空间。埃隆确实表示硬件4的发布将比硬件3晚至少6个月，因为现在的重点是使硬件3上的全自动驾驶功能FSD在国际范围内运作得非常出色。
Tesla will have to retrain using the higher resolution V4 cameras because the photons are different giving a different bit stream. Elon did say a real 6 months though and maybe even less.
特斯拉将不得不重新培训，使用更高分辨率的V4摄像头，因为光子不同，会产生不同的比特流。埃隆确实说过真正的6个月，甚至可能更短。
Hardware 4 will use more power than hardware 3 but the incredible thing is V12 is so efficient that the Model S Elon was driving on hardware 3 was running around 100 watts of inference compute.
硬件4将比硬件3消耗更多电力，但令人惊讶的是V12如此高效，以至于埃隆(Elon)驾驶的Model S在硬件3上运行时的推断计算功耗仅为约100瓦。
All of the inference or on-car processing is local. So again if Tesla were offline there would be no difference. No internet connection is required for a Tesla to drive on a road it's never seen before.
所有的推理或车载处理都是本地进行的。所以，再次强调，如果特斯拉断网了，不会有任何区别。特斯拉在从未见过的道路上行驶时不需要互联网连接。
Tesla is running at full frame rates on hardware 3. 8 cameras at 36 frames per second and this pure vision V12 actually runs faster than the version that is the mix of normal software C++ and AI aka version 11. It could actually run faster than 36 frames per second. Tesla guesses 50 frames per second but it's unnecessary since roads are designed around 24 frames per second.
特斯拉在3代硬件上以全帧速率运行。8个摄像头以每秒36帧的速度工作，而这个纯视觉V12版本实际上比混合了正常软件C++和人工智能的第11版本运行得更快。它实际上可以以超过每秒36帧的速度运行。特斯拉猜测每秒50帧，但这是不必要的，因为道路设计基于每秒24帧。
So on Tesla's custom AI computer only 100 watts of power is needed a very small amount to achieve self-driving. Compare this to the competition with dozens of sensors, radars and LiDAR and the power use goes way up as well as the cost.
在特斯拉的定制AI计算机上只需100瓦的电力，这个数量非常小，就可以实现自动驾驶。与竞争对手使用大量传感器、雷达和激光雷达相比，其功耗和成本也大大增加。
Remember not all self-driving systems are created equally. A show himself said this end-to-end neural net approach will result in the safest, most competent, most comfortable, most efficient and overall the best self-driving system ever produced. It's going to be very hard to beat it with anything else.
请记住，并不是所有的自动驾驶系统都是相同的。一个展示者说过，使用端到端神经网络方法将会产生最安全、最有能力、最舒适、最高效以及整体上最优秀的自动驾驶系统。用任何其他方式很难超越它。
Took a tremendous amount of high quality data from the fleet, a very large amount of compute and a world-class engineering team to get here.
需要从舰队中获取大量高质量的数据，进行大量的计算，并组建一个世界级的工程团队，才能做到现在这一步。
Elon added when FSD is twice as safe as the average human driver, fully unsupervised in all scenarios, then they want to make it 3x, 4x and 10x.
当自动驾驶功能（FSD）的安全性是人类驾驶者的两倍，并且在所有场景下完全无需监督时，埃隆称他们希望将其提升至3倍、4倍和10倍。
A shock also quietly mentioned soon, Tesla will add instructions or commands so you could say things like lane change from the left most lane or pull over here and the car should recognize those commands.
震惊之余，我们还很快就会悄悄得知，特斯拉将添加指令或命令，这样你可以说诸如“从最左边的车道换道”或“在这儿靠边停车”，而车辆应该能够识别这些命令。
V12 also means great things for actually smart summon. In comparably better parking lot control is coming because again V12 does not need map data which parking lots usually don't have. The cars will just learn to navigate through parking lots like other good human drivers have done.
V12也意味着实际智能召唤将有更大的进展。相对而言，更好的停车场控制即将到来，因为V12不需要地图数据，而停车场通常没有这些数据。汽车将学会像其他优秀的人类驾驶员一样在停车场里导航。
And it's this same software that will transfer over to Tesla's Optimus Bot, the ability to feed it data of certain tasks and the bot will just understand what to do by mimicking what other humans have done.
正是这款软件将被转移到特斯拉的Optimus Bot上，它可以通过提供某些任务的数据来训练，然后通过模仿其他人的行为来理解要做些什么。
So we have all of this incredible groundbreaking breakthrough type stuff happening yet mainstream media is crickets on the entire situation. They really just don't get it. The public also has no idea and does not understand what's coming but to the few of us on the planet that do understand it's a gift to watch Tesla's progress on solving a problem humanity has never solved for.
所以我们目前正在发生一系列令人难以置信的突破性突破事态，然而主流媒体对整个局势却保持沉默。他们真的不明白。公众也一无所知，不理解即将到来的一切，但对我们这些了解的少数人来说，观看特斯拉解决人类从未解决过的问题的进展是一份礼物。
Prior version jumps of FSD have been over hyped but V12 is actually different and worthy of all the hype. No human code, no heuristics, just flat out AI neural nets learning to drive like the best drivers do.
之前版本的全自动驾驶 (FSD) 被过度夸大其辞，但 V12 真正不同，值得所有炒作。没有人工编码，没有启发式算法，只是纯粹的人工智能神经网络像最好的驾驶员一样学习驾驶。
Drop a Tesla with V12 anywhere on earth and it's going to understand how to drive over time as Tesla testers feed the AI the proper video data. So now Tesla just focuses on scaling the fleet and the compute power.
在地球的任何地方放下一辆搭载V12发动机的特斯拉，随着特斯拉测试人员提供准确的视频数据供AI学习，它将逐渐理解如何驾驶。因此，现在特斯拉只需专注于规模化车队和计算能力的发展。
It's not an easy path but the path is now crystal clear. What Tesla has figured out cannot be replicated or duplicated anytime soon if at all. Tesla will be the first to solve for autonomy and it's not going to be close. This weekend will go down in history as the one where Tesla showed the world it has figured out self driving. Now they just have to refine it.
这并不是一条容易的路，但是现在这条路已经变得清晰明了。无论何时，特斯拉所找到的东西是无法很快或根本上复制或重现的。特斯拉将是第一个解决自动驾驶问题的公司，且毫无争议。本周末将成为历史上特斯拉向世界展示其实现自动驾驶的时刻。现在他们只需要进一步完善它。
And I know that was a lot so here's a quick refreshing palette cleanser. I enjoyed Tasha's take today on the Supercharger network being the gateway drug for other Tesla platforms like licensing autopilot and FSD. Joke Tettmeyer also gave us a drone flyover of Tesla's lithium refinery in Corpus Christi so want to show a few clips.
我知道刚才说了很多，所以这里有一个快速的清新调色板。我今天喜欢Tasha关于超级充电网络作为引发特斯拉其他平台的入口药物（如授权自动驾驶和完全自动驾驶）的观点。Joke Tettmeyer还给我们展示了特斯拉在科珀斯克里斯蒂的锂矿加工厂的无人机俯瞰，所以我想展示一些视频片段。
A good sense of the amount of construction and changes that have happened in under four months since that groundbreaking.
自从那次突破性的开始以来，不到四个月的时间里，我们对建设和变化的程度有了很好的感知。
Now one of the things that I did notice here is there's some ponds very large ponds that have been created since the construction has begun and you can see them right here in the middle of the screen and the bottom of the screen. And there's one at the top of the screen as well. That one sort of trapezoidal shaped the ones down here are kind of triangular shaped.
现在我注意到的一件事是，在建筑工程开始后，有一些非常大的池塘被建了出来，你可以在屏幕中间和底部看到它们。而且在屏幕顶部也有一个池塘。那个池塘的形状有点梯形，而底部的池塘则是三角形状的。
But as we continue to fly right now kind of on the west side facing towards the south this gives you a really good view of how the entire site has transformed. The last time I was here was in April and there was nothing it was just fields there had been no earthwork there was nothing at all completed.
但是，现在我们继续飞行时，我们正朝向南边的西侧飞行，这让你可以真正地看到整个场地是如何改变的。我上次来这里是在四月，那里一无所有，只是一片田野，没有任何土方工程完成。
This might be a furnace or a kiln but I really don't know and since I haven't really had any experience with this site up to this point in time I would like to ask if my viewers know what these are but the building would be on the left where that kind of white mixes being installed in the middle where you see the vertical structures is most of the main refinery that will proceed from this point across to the right of the image. Now as we get a little closer to these vertical structures and this slab that's being built I'll zoom in so you can get a good sense of the activity that's going on. Also this gives you a good sense of the scale of the height of those two concrete structures that have been constructed and the work that is underway right now.
这可能是一个炉子或窑炉，但我真的不确定，因为到目前为止，我对这个工地没有什么经验。所以我想问问我的观众是否知道这是什么。建筑物在左边，那里正在安装一种白色的混合材料。在你看到的垂直结构的中间，是大部分主要的精炼厂，从这个点开始，将继续延伸到图片右侧。现在，当我们离这些垂直结构和正在建造的板块更近一些时，我会放大镜头，让你可以清楚地看到正在进行的活动。同时，这也能让你对那两个已经建造好的混凝土结构的高度以及工作情况有一个好的概念。
Korea Economic Daily said Hyundai will develop ultra-fast chargers with its own technology to catch up to Tesla. They're seeking to get their 350 kilowatt chargers certified with the Korean government to debut them in the market this year. They're calling these ultra-fast luxurious charging stations E-PIT stations and each one will be equipped with four to six units of 350 kilowatt chargers but they require their charger suppliers to use expensive materials to build high-end chargers. This premium strategy has heightened charger building costs preventing Hyundai from finding suitable charger suppliers and it's become a major hurdle to the fast penetration of these E-PIT stations. Each E-PIT station is going to cost more than the typical 113,000 to build an ordinary 350 kilowatt charger. We'll see if Hyundai chooses to roll this out globally outside of Korea because they did say they were going to be part of that consortium in the United States setting up their own 30,000 fast charging locations.
韩国经济日报称，现代汽车将开发自己的超级快速充电器技术，以迎头赶上特斯拉。他们正寻求使其350千瓦充电器获得韩国政府的认证，并计划今年在市场上推出。他们将这些超级快速豪华充电站称为E-PIT充电站，每个充电站将配备四至六个350千瓦充电器，但他们要求充电器供应商使用昂贵的材料来建造高端充电器。这种高端策略增加了充电器建设成本，导致现代汽车难以找到合适的充电器供应商，这已成为E-PIT充电站快速普及的重大障碍。每个E-PIT充电站的成本将超过普通350千瓦充电器的典型成本11.3万美元以上。我们将看到现代汽车是否选择在韩国以外的全球范围内推出这一计划，因为他们曾表示将成为在美国设立3万个快速充电点的联盟的一部分。
We're now at 23 countries around the world that have passed the 5% full BEV adoption rate. Just one year ago when Bloomberg did this analysis they had 19 countries passing that rate so in the last year we added another four to the list. Those being Canada, Australia, Spain, Thailand, and Hungary and they're also saying EVs can surge from that 5% threshold to 25% in just four years. If the US follows this trend 25% of new cars will be fully electric by 2026. So this was a pretty cool chart on the far right column. You see the first quarter at 5% and look at Sweden. They hit that milestone quarter to 2021 and fast forward just about two years and they're sitting at 38% EV market share. So from 5% to 38% in two years. And here's the bottom of the list looking at the US hitting that 5% threshold just two quarters after Sweden but we are only at 7% BEV market share. And they said India may be next up to crack this list as EVs made up 3% of new car sales in the last quarter.
我们现在已经有23个国家的全电动汽车（BEV）拥有率达到了5%。就在一年前，彭博社进行这项分析时，有19个国家达到了这一比例，所以在过去一年中，我们又新增了四个国家。它们分别是加拿大、澳大利亚、西班牙、泰国和匈牙利，他们还表示电动汽车可以从5%的阈值迅速增长到25%，只需四年时间。如果美国跟随这个趋势，到2026年，全新车中有25%将是纯电动汽车。因此，这是一个非常有趣的图表。你可以看到，最右边一列是第一季度的5%，看看瑞典，他们在2021年达到了这一里程碑，然后在短短两年时间内，他们的电动汽车市场份额就达到了38%。所以从5%到38%只用了两年时间。而美国在两个季度后达到了这5%的阈值，但我们只有7%的电动汽车市场份额。他们还说印度可能是下一个进入这个名单的国家，因为在上个季度，电动汽车占新车销量的3%。
Tesla has two upcoming trials starting this year for accidents that did involve fatalities. The first one starting mid-September in California. They're accusing Tesla of knowing that autopilot and other safety systems were defective when it sold the car. The second trial starting early October in Florida. They're saying autopilot failed to break steer or do anything to avoid the collision according to the lawsuit Tesla has denied liability for both accidents blamed driver error and said autopilot is safe when monitored by humans Tesla said in documents drivers must pay attention to the road and keep their hands on the wheel. There are no self-driving cars on the road today. Earlier in April Tesla actually won a trial it was not involving a fatality but the jury said Tesla did warn drivers about its system and driver distraction was to blame. These will certainly be cases to watch because if Tesla gets favorable rulings then maybe future ones won't come up but if Tesla is found to be liable or negligent and has to pay out punitive or compensatory damages then you would imagine more cases like this will probably come in the future.
特斯拉今年有两起即将开始的审判，与死亡事故有关。第一起将于九月中旬在加利福尼亚开始，指控特斯拉在销售汽车时知道自动驾驶和其他安全系统存在缺陷。第二起将于十月初在佛罗里达开始，指控特斯拉的自动驾驶系统未能刹车、转向或采取任何避免碰撞的行动。特斯拉在文件中表示，对于这两起事故，特斯拉否认责任，将其归咎于驾驶员的错误，并表示只要人类监控，自动驾驶是安全的。特斯拉在文件中还指出，驾驶员必须专注于道路并保持双手握住方向盘。目前没有实现完全自动驾驶的汽车上路。今年四月早些时候，特斯拉实际上赢得了一起与死亡无关的审判，陪审团表示特斯拉已经警告过驾驶员有关其系统，并认为驾驶员的分散注意力导致了事故。这些案件肯定值得关注，如果特斯拉得到有利的裁决，也许未来类似的案件就不会再出现，但如果特斯拉被判有过失并被迫支付惩罚性或赔偿性损害赔偿金，那么可以想象将来可能会出现更多这样的案件。
Mercedes who has come out and said they're adopting the NACS will also be rolling out their own charging stations with the NACS and CCS starting later this year. In the fourth quarter to be specific and today Mercedes said they'll offer 400 kilowatt charging at their stations but how many cars can actually accept that rate anyway. The first locations will be in Atlanta, Chengdu in China and Manheim Germany. They're aiming for 2000 chargers installed worldwide by the end of next year. After the Chinese EV company announced it will be buying DD's smart card development business in a deal worth $744 million dollars. X-Beng is getting DD's expertise with riot hailing and autonomous technology and as part of the deal X-Beng will launch a new EV brand called Project Mona. The cars will be relatively cheaper on $20,000. The companies will work together on marketing, financial services deals charging and of course a robot taxi network.
梅赛德斯表示他们将采用NACS，并将从今年年底开始推出自己的充电站，使用NACS和CCS技术。具体而言，将在第四季度开始，并表示他们的充电站将提供400千瓦的充电功率，但实际上有多少车辆能够接受这样的充电速率呢？首批充电站将设立在亚特兰大、中国成都和德国曼海姆。他们计划在明年年底之前在全球安装2000个充电器。此外，中国电动车公司宣布将以7.44亿美元的价格收购DD的智能卡开发业务。X-Beng将获得DD在拼车和自动驾驶技术方面的专业知识，并将以Project Mona为名推出一款新的电动汽车品牌。这些汽车的价格相对较低，约为2万美元。两家公司将合作进行市场营销、金融服务、充电和当然还有机器人出租车网络等方面的交易。
Don't forget check out AG1 links below, get your freebies, thank you in advance if you do.
请勿忘记查看下方的AG1链接，获取免费赠品。如果您能做到的话，提前感谢您。
Hope you guys have a wonderful day, please like the video if you did, you can find me on X and Instagram links below and a huge thank you to all my Patreon supporters.
希望你们度过美好的一天，如果喜欢视频的话，请给视频点赞。你可以在下方找到我的X和Instagram的链接，同时要特别感谢所有给予我赞助的Patreon支持者。