Tesla's Real FSD Advantage / Lyn Alden on Stock Valuations / Elon: Biggest Threat ⚡️
发布时间 2022-01-18 20:47:12 来源
Welcome to Electrified, it's your host, Dylan Loomis. Happy Tuesday to all of you in a quick shout out to my newest patrons, Roland and Patrick. Thank you guys for choosing to support the channel.
欢迎来到《电动世界》，我是您的主持人，迪伦·卢米斯。祝大家周二快乐，并特别向我的最新赞助者罗兰和帕特里克致以问候。感谢你们选择支持本频道。
First up today, just a quick FYI, there is a new spoof account of Dan O'Dowd, so in case you're seeing some tweets, just know that there are now at least two accounts so you need to check to make sure and see if it's actually him.
首先，只是一个小提醒，丹·奥多德的新假账户现在已经出现了，所以如果你看到一些推文，请知道现在至少有两个账户，你需要检查确认是否真的是他。
I also thought this was a good idea and worth sharing. I like it. Here I just want to do my part in spreading the message that Elon said yes, unfortunately this is becoming a security issue, this being people talking about his travel plans. Yes, there are some accounts out there that are dedicated to following Elon's every move, so just spread the message to maybe not follow those accounts or at least don't spread them even more as we don't want everybody knowing when and where Elon is moment by moment.
我也认为这是个好主意，值得分享。我喜欢它。在这里，我只是想尽一份力量来传播埃隆说好的消息，不幸的是这正在成为一个安全问题，问题在于人们谈论他的旅行计划。是的，有一些专门跟踪埃隆一举一动的账号存在，所以请不要扩散这些账号，或者至少不要再进一步传播，因为我们不希望每时每刻都有人知道埃隆在何时何地。
I have to highlight this tweet thread by James Downa, it goes over an excellent point that I think a lot of people overlook. People misunderstand the value of a large fleet gathering training data. It's not the raw size of the data you collect that actually matters, it's the size of the set of available data you have that you can selectively incorporate into your training data set. This is a critical distinction. The set of data you choose to train with has a huge impact on the results you can get from the train network.
我必须强调詹姆斯·唐纳的这条推文串，他提出了一个很重要的观点，我认为很多人都忽略了。人们误解了庞大车队聚集训练数据的价值。实际上，你收集的数据的原始大小并不重要，重要的是你拥有可供选择并可以有选择性地融入训练数据集的可用数据集的规模。这是一个至关重要的区别。你选择用来训练的数据集对于你从训练网络中获得的结果产生了巨大的影响。
Companies that just Hoover up everything have to go back through the collected data and carefully select the items to use for training. So if you put cameras on cars and just collect everything you'll end up not using 99.99% of it. Collecting all of that, it's time consuming and expensive. Tesla does not do that.
那些只是收集一切数据的公司需要回头重新检查收集到的数据，并仔细选择用于训练的项目。因此，如果你把摄像机安装在汽车上，只是收集一切，最终你会发现你根本没有使用其中的99.99%。收集所有这些数据非常耗时和昂贵。特斯拉不会这样做。
Tesla cars select specific items of interest to the FSD project and just upload those items. Tesla probably still doesn't use 99% of what they collect, but they get what they need and do it with 1000 times less uploaded data that will just get tossed out. Consider that a single clip is around 8 cameras times 39 frames per second times 60 seconds is 19,000 images.
特斯拉汽车选择自动驾驶(FSD)项目中感兴趣的特定项目，并仅上传这些项目。特斯拉可能仍然不会使用他们收集的99%数据，但他们会获取所需数据，并且仅上传1000倍少的数据，以避免浪费。考虑到一个视频片段大约有8个摄像头每秒拍摄39帧，并持续60秒，那么这就相当于19000张图片。
If you get just a fraction of the fleet, say 100,000 cars to send one clip on average a day, that's 2 billion images. Throw away 99% and you still have 20 million. That's in one day. This is too much data to be labeled by humans way too much. Elon says auto-labeling makes humans 100 times more productive. Even so, 20 million images a day would keep thousands of auto-labeling enabled labelers busy full time, maybe 10,000. 20 million is still too much.
如果你只拥有舰队的一小部分，比如10万辆汽车，平均每天发送一张视频片段，那就是20亿张图片。即使丢弃99％的图片，仍然剩下2000万张。而这只是一天的数据量。这个数据量对于人类来说太大了，无法被标注。埃隆说自动标注可以让人类的工作效率提高100倍。即便如此，每天2000万张图片依然需要数千名配备了自动标注功能的标注员全职工作，可能需要10000人。2000万张图片仍然是太多了。
Even if you can label it, you can't train with all of it because no computer is remotely big enough to frequently retrain a large neural network on a total corpus containing many many days and tens or hundreds of billions of images. The point of this exercise is to point out that Tesla cannot utilize more than maybe one clip per 10 or 100 vehicles in the fleet per day, but that doesn't mean that a huge fleet isn't a huge advantage. If you have a huge fleet, you can ask for very very specific and rare things that you need and with a big enough fleet, you will get that data.
即使您能对其进行标记，您也无法使用所有数据进行训练，因为没有一台计算机足够大并且能够频繁地重新训练包含数十或数百亿张图像的全部数据集的大型神经网络。这个例子的重点是指出特斯拉每天在车队中可能只能使用每10辆或100辆车中的一个片段，但这并不意味着巨大的车队不是一个巨大的优势。如果您有一个庞大的车队，您可以要求非常具体和稀有的数据，并且凭借在足够庞大的车队中的数据，您将得到这些数据。
The ability to be very selective with what you ask for greatly multiplies the value of the data you do collect. So yes, individual vehicles don't necessarily send a lot of data, but the point is they're always looking for useful stuff. Anytime you drive with or without autopilot, your car can be looking at every frame from every camera to find the stuff that the FSD team is looking for.
能够对所需要的数据进行极选，极大地提高了所收集数据的价值。因此，虽然单个汽车不一定会发送大量数据，但关键在于它们始终在寻找有用的信息。无论是在使用自动驾驶还是非自动驾驶模式下行驶，你的汽车都可以通过查看每个摄像头的每一帧图像来找到自动驾驶全自动驾驶（FSD）团队正在寻找的信息。
That is a monstrously huge advantage enabled by the capacity of the vehicle computers, the size of the fleet and their high bandwidth OTA capability via Wi-Fi. So in summary, what's important is not how much data you have collected, but how much high-quality data you can collect whenever you want it. Tesla could throw away their corpus and collect another good one in a month. This is what puts them in their own league data wise.
这是由于车辆计算机的能力、车队的规模和它们通过Wi-Fi的高带宽OTA能力所带来的巨大优势。因此，总结起来，重要的不是你收集了多少数据，而是你能够随时收集到多少高质量的数据。特斯拉可以丢弃他们的数据集，一个月后再收集一个好的数据集。这就是他们在数据方面独树一帜的原因。
Brian put it like this, seems like what you're saying is that the cars still collect all of that data, but by having the campaigns on the cars, it's the cars themselves throwing out 99.99% of the data, so the team at Tesla doesn't have to deal with it. Jimma says yes, a good way to put it.
布莱恩这样说，你的意思是，汽车仍然收集所有这些数据，但是通过在汽车上进行广告运营，汽车本身抛弃了99.99%的数据，这样特斯拉团队就无需处理这些数据。吉玛表示是的，这是一个很好的表述方式。
And to wrap up this section, Gasoff asked about mobile I claiming they have more data than Tesla. James said he's not privy to what everybody has on their hard drives, but to do what Tesla does, you need a huge fleet, tops of excess compute on the vehicles, and OTA capability with frequent, large bi-directional updates. As far as I know, only Tesla has this capability. Hopefully, this will enable you to add a little bit more nuance to this data conversation in the years to come.
在结束这一部分之前，Gasoff谈到了移动I并表示他们拥有比特斯拉更多的数据。James说他不知道每个人的硬盘上都有什么，但要做到像特斯拉那样，你需要一个庞大的车队，车辆上有过剩的计算能力，并且有频繁、大量的双向更新能力。据我所知，只有特斯拉具备这种能力。希望这能让你以后在数据交流中增加一些细微的差别。
Toyota cuts its expectations for February output, citing the chip shortage. This means Toyota's global production will come in under the $9 million vehicle target for its fiscal year that will end March 31st. Toyota did say that demand was very strong, but due to the impact of semiconductors across all industries, we have adjusted our production plan. Toyota actually started this fiscal year projecting for 9.3 million vehicles. Now it's looking like they will be under 9 million.
丰田削减了二月份的产量预期，原因是芯片短缺。这意味着丰田全球的生产将低于其截至3月31日结束的财年的900万辆目标。丰田表示需求非常强劲，但由于半导体在各个行业的影响，我们不得不调整生产计划。丰田实际上在本财年开始时预计生产930万辆车。现在看起来他们将低于900万辆。
If you guys have any interest in investing or the macro space at all, I highly encourage you to subscribe to Lynn Alden's newsletter. She is definitely one of the best of the best in the space, and she puts things in a way that are easy to digest. This question highlights from her most recent work. The US stock market capitalization currently represents 61% of the global stock market capitalization, despite the fact that US GDP is only 23% of global GDP, and here is a pie chart backing up the data with the United States in the yellow. No one else really even close. The gray is all other countries not listed over here combined. The United States equities now represent about 200% of the United States GDP, which is an all-time high, as you can see very clearly on this chart.
如果你们对投资或宏观领域有兴趣的话，我强烈建议你们订阅林恩·奥尔登的新闻简报。她绝对是这个领域中最顶尖的人之一，而且她讲解的内容易于理解。以下是她最新作品中的亮点。尽管美国的GDP仅占全球GDP的23%，但美国股市的市值目前占全球股市市值的61%，这就是一份以饼图形式支持数据的证据，其中美国以黄色表示。其他国家合起来的灰色部分则没有任何一个国家接近美国。美国的股票现在已经占到美国GDP的200%，这创下了历史新高，你可以清楚地在这张图表上看到。
Why is this? Well, let's go through two of the main reasons. First is lower interest rates. Here's another chart that shows 10-year treasury yields in red compared to the cyclically adjusted price-to-earnings ratio of the S&P 500 in blue. As you can see, since 1981, the 10-year treasury yield has been on a steady decline to present day, while over that time, generally speaking with the exception of the dot-com crash, the P.E. ratio of the S&P 500 has also been trending up as this is a very clear inverse correlation. Another reason for all of this global capital funneling into the United States stock market, well, corporate tax cuts. The United States has reduced its effective tax rate for corporations persistently over the past several decades. The blue is corporate profits or profits before taxes. The red is the federal government tax receipts on corporate income and corporate profits.
为什么会这样呢？让我们探讨一下其中的两个主要原因。首先是更低的利率。这是另一张图表，显示了红色的10年期国债收益率与蓝色的标普500指数的市盈率相比较。从1981年以来，你可以看到10年期国债收益率一直在稳步下降，而与此同时，通常情况下标普500指数的市盈率（除了互联网泡沫破裂时期）也在上升，这是一种非常明显的反向相关关系。进入美国股市的全球资本流入的另一个原因是企业减税。在过去几十年中，美国一直在持续降低公司的有效税率。蓝色是税前利润或税前利润，红色是联邦政府对所得和利润征收的企业税收。
I will link the rest of the newsletter below. I'm not sharing this to be any actionable item. I just want it to be an extra information source for you to use as you make your own decisions.
我会在下面提供其他新闻通讯的链接。我并不是为了提供任何可行动的项目而分享这些。我只是希望它成为你在做决策时的额外信息来源。
Sawyer showed us some images over 800 model SNX spotted at Tesla's Fremont factory. The point here is that Fremont is going to be pushed to the max here over the next few years, as Elon has said they are hoping to expand Giga Nevada and Fremont's production capacity by about 50%. And in case you're waiting to see a dyno test of the Model S plaid, here you have it thanks to Uptown Auto Sport. We were curious to see the actual power output of this car. We're going to be using our mainline pro hub dyno. It's been proven to be the most precise and reliable instrument on the market.
Sawyer向我们展示了一些在特斯拉弗里蒙特工厂发现的800多辆SNX型号的图片。重点在于，在接下来的几年里，弗里蒙特将面临巨大压力，因为埃隆表示他们希望将Giga Nevada和弗里蒙特的生产能力扩大约50%。另外，如果你正在等待看到Model S plaid的动力测试，感谢Uptown Auto Sport，你可以在这里找到它。我们很好奇想看到这辆车的实际动力输出。我们将使用我们的主要专业轮毂式动力测试仪。该仪器已被证明是市场上最精确可靠的工具。
But we're really curious to see the numbers. From drive Tesla Canada, the California Public Employees Retirement System, the largest public pension fund in the US in terms of assets, has grown its stake in Tesla stock. It now owns about $1.3 billion of Tesla representing about 1% of their portfolio. It's great to see this because typically these pension funds are more conservative, so owning a high growth stock like Tesla would be a little out of the ordinary. However, this isn't the only pension fund to do so, as the New York State Teachers' Retirement System is also involved in owning Tesla shares. As of 2020, they had owned almost 1 million Tesla shares.
但我们真的很好奇要看到这些数字。根据加拿大特斯拉驱动（drive Tesla Canada）的消息，美国资产最大的公共养老金基金--加州公共雇员退休金制度（California Public Employees Retirement System），已增加其在特斯拉股票中的持股。现在，该基金拥有大约13亿美元的特斯拉股份，占其投资组合的约1%。这很棒，因为通常这些养老金基金更加保守，所以拥有像特斯拉这样高增长的股票会有点不同寻常。然而，并非只有这个养老金基金如此做，纽约州教师退休金制度（New York State Teachers' Retirement System）也持有特斯拉股份。截至2020年，他们已经拥有了将近100万股的特斯拉股份。
Here we have an update on the Tesla and Solar City lawsuit. It was taking place over the summer. Now we get some more news from the closing arguments. Basically, the shareholders are still asking Tesla to pay one of the largest judgments ever of $13 billion. If you're new, this lawsuit by Union Pension Funds and Asset Managers alleged that Elon strong-armed the Tesla board into approving the deal for the cash-strapped Solar City, in which Elon was the top shareholder, basically alleging that this was a bailout of Solar City driven by Elon.
在这里，我们有关于特斯拉和Solar City诉讼的最新消息。这场诉讼持续了整个夏天。现在我们从最后陈述得到了一些新消息。基本上，股东们仍然要求特斯拉支付有史以来最大的判决金额，即130亿美元。如果你是新闻，这场诉讼由工会养老基金和资产管理公司提起，它们声称埃隆强行影响特斯拉董事会批准了对资金短缺的Solar City的交易，而埃隆恰好是Solar City的最大股东，基本上声称这是埃隆推动的对Solar City的救助。
Elon said that this award would be at least 5 times the largest award ever in a comparable shareholder lawsuit, and he called it a windfall for the plaintiffs.
埃隆表示，这个奖项将至少是以往类似股东诉讼中最高奖项的5倍，并将其称为原告方的意外之财。
And here it is, the shareholders have been alleging that the deal was a needless distraction in burden Tesla with Solar City's financial woes and debt.
在这里，股东们一直声称这笔交易是不必要的干扰，使得特斯拉背负了太阳城的财务困境和债务。
Well, even if that was true, I think the Tesla stock performance has more than made up for any unnecessary distractions. Could they argue it would have gone even higher? But I think that's beside the point. At least with today's perspective, it does not look like this Solar City deal, which was about $2 billion at the time, has really impacted Tesla stock that negatively.
嗯，即使那是真的，我认为特斯拉的股票表现已经超过了任何不必要的干扰。他们可能会争辩说它本来可能会涨得更高吗？但我认为这并不是重点。至少从今天的角度来看，这次当时价值约20亿美元的太阳城交易似乎并没有对特斯拉的股票产生太大的负面影响。
However, there's not too much new here. All along, Elon did say that the Tesla board was the one primarily handling the Solar City deal, and Elon recused himself of the price negotiations. And last thing, the judge on the case said that last week he intends to retire, and this judge was actually overseeing another Tesla case that had to do with Elon's compensation package, so now Chancellor Kathleen McCormick is going to take over that case about the compensation package, effective immediately.
然而，并没有太多新的内容。从一开始，埃隆就说过特斯拉董事会是主要负责太阳城交易的，他本人排除了价格谈判的参与。最后一件事，该案的法官表示上周准备退休，而这位法官实际上正在监督另一起与埃隆的报酬计划相关的特斯拉案件，所以现在凯瑟琳·麦科密克院长将立即接管关于报酬计划的这起案件。
Pierre Farrigu shared a bullish article, and he said strongly lines with our expectations of over 1.6 million units delivered this year from Tesla. The company is engaged in sourcing negotiation rounds for 1.5 to 2 million cars, talking about cameras.
皮埃尔·法里古分享了一篇看涨的文章，并表示这与我们对特斯拉今年交付超过160万辆的期望非常符合。该公司正在进行150万至200万辆汽车的采购谈判，谈论的是摄像头。
Here's the article. LG Anitek and Samsung electromechanics are expected to compete for orders for Tesla camera modules. Tesla's camera module market share was said to be 60-70% of LG and 30-40% of Samsung. 8 camera modules are included for each Tesla EV, and the front camera is more expensive than the side and rear cameras. And it says right here, this year's shipment estimate is 1.5 to 2 million units. It's highly likely that LG will secure the quantity of Cybertruck camera modules that LG missed last year. It's actually advantageous for Tesla to diversify the supply chain for the camera modules. And with an average price of more than $10 per unit, camera module sales per electric vehicle are around $100.
这是一篇文章。LG安佳特和三星电器有望竞争特斯拉相机模块的订单。据说特斯拉相机模块的市场份额是LG的60-70%，三星的30-40%。每辆特斯拉电动车都包含8个相机模块，前置相机比侧面和后置相机更昂贵。文章还指出，今年的出货量预计为150-200万个。很有可能LG将会补足去年错过的数量，为Cybertruck相机模块提供足够的供应。对于特斯拉来说，摄像头模块供应链多样化实际上是有利的。而且每个电动车的相机模块平均售价超过10美元，每辆车的相机模块销售额约为100美元。
Dan Levy from Credit Suisse has increased his Tesla price target to $1,025. He cited the obvious capacity expansion, gross margin, updated product roadmap and batteries. Tesla's path of volume will be purely a function of its production. We expect Tesla to maintain EV leadership given its holistic approach on supply. We forecast 2022, deliveries of 1.47 million units, 57% year over year, which still might be low.
瑞士信贷的丹·利维将特斯拉的目标股价上调至1,025美元。他提到了明显的产能扩张、毛利率、更新的产品规划和电池。特斯拉的产量增长完全取决于其生产能力。考虑到特斯拉在供应方面的整体方式，我们预计特斯拉将继续保持在电动汽车领域的领先地位。我们预测2022年的交付量为147万辆，同比增长57%，但这可能仍然偏低。
We learned that GM is going to be getting some EV chargers for its dealerships from Blink in North America and Canada. And these will be level 2 chargers. So here's my question for you. If you were running your own dealership encouraging the EV movement, would you opt for level 2 or level 3 the faster chargers? I personally would want my customers to have the very best experience and to not have any hesitancy of how long it might take them to charge, even though you should be giving them education about home charging and all of that. I personally would want the faster chargers at my dealership. Maybe they want level 2 so people continue to spend more time on the dealership. I don't know, let me know if I am missing something here.
我们了解到通用汽车将从北美和加拿大的Blink购买一些电动汽车充电设备用于其经销商。这些设备将是二级充电设备。所以我有一个问题想问你。如果你经营自己的经销商并且鼓励电动汽车的发展，你会选择二级还是三级更快的充电设备？我个人希望我的客户能够拥有最好的使用体验，而不用担心充电需要多长时间，尽管你应该向他们提供家庭充电的相关知识等。我个人希望在我的经销商有更快的充电设备。也许他们希望选择二级设备，这样人们可以继续在经销商处逗留更长时间。我不知道，如果我漏掉了什么，请告诉我。
Back to Sawyer. If a Tesla Senn mind drives 800,000 miles over the vehicle lifetime, it could do more. The owner would save around $400,000 in fuel costs, assuming a price per kilowatt hour of $0.24, which of course will vary widely, region to region. So let's reply those savings for a fleet and you can partially see why the Tesla Senn is such a disrupter.
回到Sawyer。如果一辆特斯拉Senn在整个使用寿命内驱动了80万英里，则它可能会做得更多。假设每千瓦时的价格为0.24美元，车主在燃料费用上将节省大约40万美元，当然这个价格会在不同地区有很大差异。那么我们来考虑一下车队的节省，你就能部分理解为什么特斯拉Senn如此具有颠覆性了。
Elon has really been driving this point home as of late, saying we should be much more worried about population collapse. The UN projections are utter nonsense, just multiply last year's births by life expectancy. Given the downward trend in the birth rate, that is the best case and less reversed. If there aren't enough people for Earth, then there definitely won't be enough for Mars.
最近以来，埃隆一直在强调这一点，认为我们应该更加担心人口崩溃问题。联合国的预测完全是胡说八道，只需将去年的出生人数乘以预期寿命即可。考虑到出生率的下降趋势，这已经是最好的情况，甚至不会逆转。如果地球上人口不足够多，那么对于火星来说肯定也不足够。
So what is this data that Elon keeps referring to? Well here is the current day right on this red dotted line. And this goes back to, we'll say, 1950. So there has been a steady decrease of births per 1000 people. Back in 1955, it was 23.9 births per 1000 people. And today we are down to around 12 births per 1000 people. The bottom chart is the annual percent change year over year. So if it's red, that means year over year, it's decreasing. And as you can see the green year over year increases have been few and far in between historically. Currently, it's actually very slightly green. But overall, it's only because the last few years were down so much year over year. And remember what Elon said about the UN projections? Basically everything to the right of the red dotted vertical line is utter nonsense according to Elon.
那么，埃隆一直在提到的数据是什么呢？好吧，这条红色虚线代表的是当天的数据。这份数据可以追溯到大约1950年。每千人的出生率呈稳定下降趋势。在1955年，每千人的出生率为23.9。而今天，每千人的出生率降至约12人。下方的图表显示了每年的年度百分比变化。如果是红色，表示年度百分比变化在减少。而你可以看到，历史上绿色的年度增加情况非常少且几乎没有。目前，实际上它稍微呈现出绿色。但总体而言，这仅仅因为近几年的年度变化都下降很多。还记得埃隆关于联合国的预测所说的吗？基本上，红色虚线右侧的一切都是埃隆认为胡扯的。
Last up for today, Tesla Silicon Valley Club shared this video of FSD Pure Vision seeing a lot in one image. Elon chimed in saying spatial and temporal memory are improving. So there will be less flicker, less flicker in the images so you'll have more solid objects, just showing more confidence in what the car can see.
今天的最后一个动态，特斯拉硅谷俱乐部分享了一段关于FSD纯视觉在一个图像中看到的许多内容的视频。埃隆也进行了回应，表示空间和时间记忆正在改善。因此，图像中的闪烁现象将减少，您将能够看到更多实体物体，从而更加信任汽车的视觉能力。
But that'll do it for today. Please take a second to like the video. If you did, hope you guys have a wonderful day and a huge thank you to all of my patreon supporters.
但是今天就到这里。请花一秒钟给视频点赞。如果你点赞了，希望你们有一个美好的一天，非常感谢所有资助我的 Patreon 支持者们。