发布会·中文精读

马斯克发布会:Tesla软件更新v7.1(2016年1月)

Elon Musk press conference: Tesla's software update v7.1 (Jan 2016)

中文精读由 AI 基于原始视频与文稿改写,非官方翻译,可能存在偏差或错误;表述以原视频为准。说明

2016 年 1 月,特斯拉推送了一次 OTA 更新。更新完成后,车主站在自家车道上,打开手机下拉菜单点一下按钮,车会自己打开车库门,缓缓开出来停到面前,再转身把身后的车库门关上。

发布会后的问答环节,有记者问 Autopilot 到底安不安全。马斯克的回答留了余地,也给了判断:

在高速公路驾驶这个场景里,我现在的判断是,它可能已经比人类司机开得更好。至少在未来几个月内一定会如此。(马斯克)

他说“可能”,也说“至少在未来几个月内”。两个词叠在一起,能看出他对这件事的态度:数据还没跑够,方向已经很确定了。

整场发布会谈了四个技术话题:召唤功能、蛇形充电器、Autopilot 的安全边界、完全自动驾驶的时间表。每个话题马斯克都给出了具体数字,也都留出了明显的退路。这些数字后来有些兑现了,有些没兑现,把它们放在一起看,会浮现出一条很清晰的判断线索:他知道现在的车还缺什么,也知道补上要花多久。

一台自己从车库开出来的车

v7.1 最直观的新功能叫“召唤”(Summon)。使用场景是这样的:

你站在车道上,打开手机 App 的下拉菜单,点一下召唤按钮,车自己打开车库门,从里面开出来,停到你面前,然后把身后的车库门关上。

反向操作叫“泊车”(Unsummon)。同样一次点击,车自己开回车库,门没开就先打开,开进去之后再关上。

马斯克对这套功能的定性很克制:

这只是召唤方向上的一个婴儿学步。(马斯克)

“婴儿学步”这个说法不算自谦。真正的召唤,按他给的定义,是你在任何地方都能让车来接你,包括你在纽约、车在洛杉矶这种距离。

手机点一下,钥匙按一下

现场有记者追问了一个操作细节:官方新闻稿里写的是用钥匙(fob)召唤,马斯克开场讲的是用手机,到底哪个对?

答案是两个都行。

你可以在钥匙上做一组编码按键,也可以用手机下拉快捷菜单,点一下就触发。(马斯克)

这个细节看起来琐碎,但它决定了这套功能的性质。手机触发的召唤,本质上是把车接入了一个远程指令系统。车不再是等你上车才启动的机器,它可以被一条指令唤醒、移动、归位。这条指令目前只覆盖几十米的范围,但技术路径已经打开了。

蛇形充电器为什么“瘆人”

发布会现场有人问了个有意思的问题:既然召唤都快量产了,那个“邪恶机器蛇充电器”(evil robot snake charger)是不是也该进生产了?

这个梗来自 2015 年特斯拉放出的一个概念视频:充电桩里伸出一根金属蛇形机械臂,主动找到车身上的充电口插进去,全程没有人。

马斯克的回答是:

会做,我们会尽量让它看起来没那么瘆人。有意思的地方恰恰是它的瘆人感。我猜会先在部分超级充电站试水,看看效果。(马斯克)

他补了一句解释,为什么这东西必须存在:车如果要实现完全自主移动,充电这一步就不能有人。那就意味着充电桩一侧也得有自动化系统,能自己完成对接。

这个逻辑当时没多少人认真琢磨,但它把后面所有关于自动驾驶的讨论串起来了。一台自己会开、自己会找路、自己会充电的车,已经超出了驾驶辅助的范畴,它更像一个完整的移动机器人系统。

跨州召唤的前提是有陆桥

马斯克给了一个时间表:

我可能有点乐观,但我认为两年之内,你就能从全国各地召唤你的车。假设你在纽约,车在洛杉矶,它会自己找路到你身边,跟着你手机的位置走。整个旅程中它会自动充电。(马斯克)

他给这个愿景加了一个物理前提:

任何有陆桥连接的地方,任何车物理上能开到的地方。(马斯克)

“陆桥”(land bridge)这个词用得很精确,意思是中间不能有海。从纽约到洛杉矶这条路,车得能一路开过去,中间不能有断点。

他把跨州召唤和完全自动驾驶直接绑在了一起。能跨越全国的自动驾驶,本质上已经是通用意义上的完全自动驾驶了。这也解释了为什么后面记者问“Model S 能不能靠 OTA 升级实现”时,他会把话题引到硬件上。

现在的车缺的是冗余

《Jalopnik》的记者问得很具体:两年后那台能自己横跨美国的车,是现在的 Model S 通过软件升级就能做到,还是得等新硬件?

马斯克把问题拆成了两半:

要实现那个目标,我们需要比现在的车更多的传感器。更多摄像头、更多雷达、冗余的电子设备、冗余的电源总线,让整车具备失效运行(fail operational)能力。(马斯克)

“失效运行”是这场访谈里技术含量最高的一个词。他解释得很直白:

现在如果电子系统出故障,车会失效到驾驶员那里,驾驶员就是 Plan B。但要实现那个目标,车必须失效到另一套车载电子系统,不能依赖人来接管。(马斯克)

这句话信息量很大。2016 年初的特斯拉 Autopilot 是一个单点系统:传感器、计算单元、执行器,任何一环坏了,都由人接手。真正的自动驾驶要求同样的能力在车里至少存在两份,一套坏了另一套顶上,全程不需要人介入。

所以马斯克说的“两年”,落点在硬件冗余的铺开速度上。软件算法跑得快并不构成瓶颈,传感器、雷达、电源总线这些东西要重新设计、重新验证、重新上产线,才是耗时的那部分。

他也承认这一点:

我们现在确实在研发一个新的硬件套件,但它进入量产还需要一段时间。(马斯克)

说得很克制。当天的记者没有追问下去。结合后来特斯拉在 2016 年 10 月发布新一代硬件套件的节奏看,马斯克当时心里已经有时间点了,只是不想提前讲。

限速 5 英里,因为路中间没有隔离带

v7.1 里出现了一个让部分车主不满的限制:在没有中央隔离带的道路上,Autopilot 的最高速度被限制在限速之上 5 英里,用公里制的地方是 10 公里。

有记者当场问:为什么突然加这个限制,是不是出了什么事?

没有特别的事件触发,这个限制其实非常温和。没有中央隔离带的道路,发生更严重碰撞的潜在风险更高,不在这种路上超过限速 5 英里,我认为是合理的。(马斯克)

紧接着他自己补充了一个“我刚才忘说了”的重点:

我们还给 Autopilot 加了弯道速度自适应。车会向前看即将到来的弯道,提前降低速度。以前在 7.0 里哪怕前面是急弯,它也保持固定速度。现在它会自己减速。(马斯克)

这两条放在一起看,逻辑很清楚:v7.1 对 Autopilot 的改动在收缩能力边界。限速是横向的收缩,弯道减速是纵向的收缩。它把车在物理上能做的事情往安全一侧收了一格。

马斯克用了一个具体对比来支撑这个判断:

它不绝对完美,但我认为它现在可能比人做得更好。即便是 beta 版本的召唤和自动泊车,发生轻微刮蹭的概率,我认为也比人低。(马斯克)

每天一百万英里在喂什么

马斯克在开场陈述里反复强调一件事:Autopilot 的定位是增强,接管并不是它的目标。v7.1 引入的“增强版车队学习”(enhanced fleet learning)让每台车的行驶数据都能回传,形成一个持续迭代的数据集。

他给了统计口径:

关键是要收集几千万英里、甚至几亿英里的驾驶数据。目前具备自动驾驶能力的车,每天大约行驶一百万英里,这个数字随着车队规模每周都在上升。(马斯克)

数据拿来干什么:

我们可以对比假阳性和假阴性,在一个统计上显著的数据集上说:这些情况下如果车处于自动驾驶模式,它会避免一起事故;这些情况下如果车处于自动驾驶模式,它会制造一起事故。(马斯克)

假阳性是车在不该介入的时候介入了,假阴性是车在该介入的时候没动。两个方向都得数清楚,才敢谈安全。

他当时的结论是:“目前为止结果很好,但数据积累期还太早。”

他自己做的一个观察很有意思:

我测试这辆车的时候,观察到的结果是,它在保持在车道中心这件事上,明显比路上的其他车做得好。(马斯克)

这句话绕开了所有关于“自动驾驶安全不安全”的抽象讨论,把话题拉到一个可测的维度上。保车道居中这件事,人类司机经常做不到,车做到了。

《The Motley Fool》的记者追问,什么时候能拿出足够证据,证明 Autopilot 在安全上确立了新标准。马斯克的回答是“今年之内大概能提出一些论证”,前提是数据量攒够。他也承认那个量级很大:几千万到几亿英里。

那几起事故的共同点

《旧金山纪事报》的记者问了一个尖锐问题:Autopilot 推出之后,网上到处有人拍视频,把车设成自动驾驶然后自己爬到后座。这些情况下有没有出过事故?

马斯克的回答分两层。

第一层,明确否认:

据我所知,没有一起事故是 Autopilot 造成的。有可能有一些小事没传到我这里,或者根本没报给特斯拉,但我确实不知道有。(马斯克)

第二层,解释那些疑似案例:

唯一我们知道的几起,是客户以为自动驾驶模式开着,其实没开。然后他们以为自动驾驶造成了事故,实际上是他们以为它开着,但它没开。(马斯克)

他补了一句更反常识的观察:

有意思的是,就算是那些真的把车设成自动驾驶、然后爬到后座的人,据我所知也没有一次真出事故。(马斯克)

这段回答值得停下来想一想。马斯克给出的框架是:Autopilot 相关的事故,要么源于人的误判,要么源于人的误用。2016 年初这个说法没有第三方数据支持,但它提出了一个归因方法:先看当时是谁在开车。

这个方法和前面那套“假阳性、假阴性”的统计思路是一回事。一起事故到底由算法主动造成,还是人以为算法在开车而实际上没开,两个问题的答案完全不同,混在一起讨论没有意义。

保险责任落在谁头上

最后有记者问了一个非常实际的问题:保险怎么算?车主不在车里,召唤过程中出了事,责任在谁?

马斯克先纠正了提问的前提:

我不太愿意把召唤叫真正的自动驾驶功能,因为你需要手动触发,钥匙必须在有效范围内。现阶段它更像一个遥控功能。(马斯克)

然后给出清楚的定性:

责任在驾驶员。我们非常明确地警告过,这是 beta 功能。停车场的场景千差万别,它不一定都适用。用召唤和自动泊车的人,要留在车内,随时准备按键。(马斯克)

他强调了操作层面的兜底:

任何时候按一下钥匙就能停掉自动泊车和召唤。这是我们强烈建议的做法。(马斯克)

接着补了一句有点凡尔赛的话:

车甚至能检测到一只相当小的狗走过去。它不绝对完美,但我认为它现在可能比人更好。(马斯克)

把这几段放在一起,能看出他当时的基本判断:功能可以新,责任归属不能新。停下来的权力在钥匙上,在驾驶员手里。这是 v7.1 所有新功能共同的兜底设计。

各州各搞一套,车过州界要换驾驶风格

有记者问到加州正在酝酿的自动驾驶立法,以及全国统一标准的必要性。马斯克的态度很明确:

如果所有州都搞自己的一套标准,或者标准差异很大,事情会变得很麻烦。联邦层面的一致性我认为是好的,至少各州要协调,行为要基本一致。否则车一过州界就得换驾驶风格,这会变得非常奇怪。(马斯克)

他对加州当时的进展评价很高:

加州可能是在自动驾驶交通这件事上想得最多的一个州。他们确实想用正确的方式来监管,当然也有明显的理由不希望机器人车乱来。(马斯克)

他用的表述是“我们和加州交通部、DMV 的沟通一直很积极”。这句话承认监管合理,同时暗示特斯拉是主动沟通的一方。

他也没有回避监管的另一面:

出于显而易见的理由,没人希望机器人车乱来。加州想确保的是,在车真正走向自动驾驶之前,安全预防措施已经到位。(马斯克)

24 到 36 个月,附带三个条件

彭博社的记者最后问了一个更根本的问题:你说的两年,是指技术上实现,还是指真正能买到的产品?

马斯克给了一个更精确的窗口:

24 到 36 个月。如果超过这个点,我会非常意外。在这个时间段内,我对存在一个通用的完全自动驾驶解决方案高度自信。(马斯克)

他专门解释了“通用”两个字的含义:

意思是在全国任何地方都能用,而不是限定在某一个城市或某几条路。同时还得设定一个可靠性或安全阈值,在那个 24 到 36 个月的窗口里,车能在几乎所有道路上以显著优于人的安全水平行驶。(马斯克)

这段话比“两年”两个字重要得多。它把目标拆成了三个条件:

  • 地理上通用,不依赖预先绘制的高精地图
  • 时间上随时可用,不限定特定城市或特定路段
  • 安全上有量化标准,显著优于人类司机

三个条件同时满足,才算“解决了”。这个标准在 2016 年定得非常严,他自己也承认“我可能有点乐观”。从后来的时间线看,这个窗口本身没有错,错的是他低估了第三项的达成难度。

四条判断里哪条被推翻了

把整场发布会从头到尾读一遍,马斯克当时对自动驾驶的判断有个清晰结构:能力上他相信两年内能通用,硬件上他知道现在的车不够,责任上他坚持人兜底,监管上他希望统一标准。

四条判断里,三条被后来的现实印证了。硬件冗余成了行业标配,责任归属在 L2 级功能上始终落在驾驶员,统一标准至今仍是各国监管讨论的核心议题。被推翻的那条是时间表。

这个反差点出了自动驾驶行业十年的核心矛盾:算法进步的速度,一直比冗余硬件普及的速度更快,而真正卡住自动驾驶上路的,是后者。到今天,感知算法迭代以月计,一套满足失效运行要求、通过车规验证的冗余硬件,从设计到量产仍然要以年计。

还有一件事值得注意。马斯克当天说的每一句“我可能有点乐观”,事后看都不是客套。他确实知道自己在押一个偏紧的时间表,也知道押错的部分在哪里:数据量、验证周期、硬件产能。他把这些限制条件都当场说出来了,只是听的人当时更关心“两年”那个数字。

对今天读这份文字稿的人,有几条判断可以直接拿来用:

  • 看一家自动驾驶公司,先看它的硬件冗余清单。马斯克当年说得很清楚:要实现失效运行,车上的电子系统、电源总线都至少要有两份。只讲算法不讲冗余的方案,离真正上路还差一层。
  • 功能上线时看清它挂的是不是 beta 标签。召唤和自动泊车在 2016 年就是 beta,马斯克反复强调要留在车里随时按键。一个功能成熟到什么程度,看它在责任归属上的措辞最准。
  • 看事故归因,先看当时是谁在开车。那套“假阳性、假阴性”的分类方法今天依然有效。算法主动造成的事故,和人以为算法在开车而实际没开,是两个完全不同的问题。
  • 听时间表,要听它附带的条件。“24 到 36 个月”这个数字本身没意义,有意义的是它后面那三个条件:地理通用、随时可用、显著优于人类。任何一条不满足,时间表就要往后推。
  • 遇到监管问题,优先寻找统一标准。车过州界就要换驾驶风格,“会变得非常奇怪”。这个判断放在今天的区域准入讨论上依然成立:规范越统一,产品越可能真正规模化。
  • 把“遥控”和“自主”分清楚。需要手动触发、钥匙必须在范围内、随时可以按停的功能,本质上是遥控。判断一个功能属于哪一类,看它在失控时的兜底机制握在谁手里。
YouTube 平台字幕(自动或人工),可能有识别错误,人名与数字请以视频为准。已按语句合并成段落,并去掉 um、uh 等口头填充词。 文稿来源:YouTube(自动/人工字幕)。

0:01in the direction of summon so you can be standing in your driveway and just tap this the the summon button from the pull down menu on your phone and the car will open the garage door come to you and then close the garage door behind it and then you can also unsummon or park the car and again just sort of do the pull down menu tap one quick tap and and the car will go to the garage open the door if it's not open otherwise just go right in and then close the garage door behind it so U it's just a baby step in the direction of summon But ultimately you you'll be able to summon your car from anywhere anywhere where there's a land bridge anywhere that the car can actually

0:59physically get you so I actually think and and you know I might be slightly optimistic on this but I I think that within two years you'll be able to summon your car from across the country so that you know if you're let's say in New York and your car is in Los Angeles it will find its way to you and meet you wherever

1:28your phone is so the phone will just communicate with the car and tell you tell tell the car where to find you and it'll just automatically charge itself along the whole journey so I think that you know like I might be slightly optimistic on that but not not I don't think significantly optimistic that we can do that probably in about two years so this is the first little step in that direction and I think it's it's quite profound to actually experience it even though it is a baby step and then in addition we've made a number of refinements and improvements in the general autopilot and I I I think

2:11at this point with 7.1 and with the enhanced Fleet learning that comes with 7.1 we know we'll have to see what the statistics say but I I my I I believe that it is probably better than human at this point in highway driving or certainly will be as the fleet learning gets more and more sophisticated that that's actually been my observation when I've been testing the car is that it's certainly better than human at staying in the center of the the lane compared to other cars on the road so I think that that's sort of quite quite exciting and and if you know if it isn't quite better than human yet it certainly will be in the coming months I think those are the two main

3:08things and and of course the car can automatically parallel park with a tap of a button it can also now reverse Park in so if you want to back into a space it'll it'll do that too just automatically yeah I think those are the main things we can just jump into questions then all right ladies and gentlemen it's time for live Q&A just click a little raised hand icon at the bottom of the Q&A pound first question is from Dean deran from Associated Press thanks for taking the call hi thanks for taking the call I'm wondering why the the speed restrictions is there other particular things that happened that made you decide to do that well I don't think

4:00the speed restriction we introduced with 7.1 is particularly uous it's really just capping the maximum speed on roads without a center divider at 5 miles above the speed limit or 10 km above the speed limit in Europe so or place with the things are in kilometers so it's a it just means like on

4:22roads without a s divider where there's a potential for a more serious Collision it kind of makes sense to not go more than maybe five miles above the speed limit that's that's really that's the only restriction really I think it's quite reasonable oh and I forgot to mention one important point which is we've added curve speed adaptation to the autopilot capability so the car will look ahead to the oncoming curves and and and reduce its speed whereas previously in 7.0 it would maintain a fixed speed even if you went around a sharp corner and now it will automatically reduce the speed I think those are both quite important safety enhancements and we have a question from

5:09Chris Perkins at Roden track hey there with the the the new summoning feature it is it seems more likely now that we'll actually see a a production version of the the evil robot snake charger is that something that you're that could reach production in the near future yeah we'll yes we will do some version of that we'll try to make it look less creepy maybe I mean it is fascinatingly it's sort of fascinating in its creepiness I think but

5:47yeah I think we' probably roll it out first on some of the superchargers and and see how that goes but I think it's it's quite yeah quite interesting but yeah essentially car will need to sort of automatically charge and so that that means the you need some sort of automated system on the charger side to connect and we have one from Mike Balaban from jalop hey Elon thanks so much for having us I just have a quick question about you mentioned that in two years the car would actually be able to drive itself from New York to Los Angeles and I'm just wondering would that be limited to Something Like the Model X and model 3 or that be would the model S be able

6:34to do that from an overthe A firmware update or something like that well I mean we do need more senses than the car currently has to achieve that goal so this is partly a hardware partly a software thing because in order to do that really by itself you need like you need a lot of redundancy and you know more cameras more radar and and like said kind of redundant Electronics redundant Power buses and that kind of thing essentially to be so that the the cars fail operational it can any system can fail in the car doesn't need to to go to back up to a driver I so right right now the the the driver is sort of Plan B if if if if something goes wrong with the electronics it it it

7:26fails to driver and and for for that thing it would need to F fail to another set of electronics on the car and we have a question from Nikki Gordon Bloomfield transport evolved good morning Elon I was wondering if you would care to comment on the proposed legislation and what legislative changes that California is considering and how you view an importance of a nationwide standard for autonomous Drive in the future yeah that that's a good point I mean certainly it would be would certainly make things difficult if if all states adopted separate standards or if there was a wide range of Standards so some consistency at the federal level I think would be would be

8:16good or at least for the states to coordinate their activity and and be fairly consistent in what they do otherwise it's going to get pretty weird if if the if the cause Behavior has to change from one when AC crosses the state border but I mean so far our conversations with the the California Department of Transport and DMV have been quite positive and they they do recognize that this is what's coming and they just want to I think approach it in the right way so so I mean I think I think the yeah California's probably put the from what I can see the most amount of thought into the into autonomous transport and but they they do want to sort of regular ated in you know you

9:00don't I mean for obvious reasons you don't want kind of robot cause misbehaving and doing crazy things so I think that that California just wants to make sure that there have been adequate safety precautions before cars really go autonomous and next we have a question from Mark chidiac from Bloomberg News hi yes that my question is regarding this software update and it sounds like you feel like cars may be autonomous within two years is that fair or is it simply I was just trying to get a sense of when we might see fully aous car I think so within two years I think it's I think it's sort of cool 24 to 36 months I mean it's really I'd be very

9:52surprised if it's beyond that point and now that's that that's when it's technically possible for for a generalized I think I'd be highly confident that there's a generalized solution for fully self-driving vehicles in that time frame so meaning meaning it would work anywhere in the country as opposed to limited to a particular City or set Road and and and one also needs to

10:20set a a reliability threshold or a safety threshold for self-driving autonomous cars I I think like in that 20 2 4 36 month time frame the the the it will be able to drive on call it virtually all roads at a safety level significantly better than human and another question from Dean Durban from the associate press I'm sorry I didn't actually have another question I guess the hand was still raised or something but I did want to ask you can summon from your phone or from the fob the press release said the fob but I also heard you at the beginning talking about the phone so which it's both or it's both yeah so you can you can basically do a coded key press yeah you

11:14can do a coded key press on your key fob or you can you can you'll be able to just pull down the the the quick action menu on your phone and and tap someo can we have a written question from Daniel Sparks or mle fool he because his question is how long do you think it will take for the case to be very clear that the more autonomous likee features of autopilot such as Auto steer to make a compelling case that Tesla is setting new standards in safety with this technology compared to human

11:51drivers I I think we'll probably be able to make some of those arguments this year I mean the key is to collect tens of millions or perhaps hundreds of millions of miles of driving Pro probably in the hundreds of millions of miles and and right now for cars that have the autonomous capability we we see about a million miles a day of driving and that that number is increasing obviously with each passing week as more as the fleet increases so what we're able to do is compare the false positives and false negatives over a a statistically significant data set and say okay

12:33these are the cases where if if the car had been an autonomous mode it would have prevented an accident or or conceivably this is a case where if it was in autonomous mode it would have it it did it it caused an accident I should say thus far the results have been very good although it's still early in the in the in the data accumulation

12:58period and actually the the the only cases that we're really aware of where that where there was some accidents is where in a couple of cases customers thought the autonomous driving mode was on but actually it wasn't so then they they thought that that the autonomous driving caused the accident but actually it was just that they thought it was on but it wasn't that that you yeah that was a couple cases but but no no no serious injuries or

13:37anything we have a question from David Baker San Francisco Chronicle hi I actually wanted to follow up on that last point there I mean you know when you first put out the autopilot features there were people posting their videos online of doing truly crazy borderline stupid things with the cars have you seen any cases where the autopilot did actually cause an accident or is it only things where it's there's an element of human error

14:12there I'm actually not aware of any cases where the autopilot course an accident it's possible that there may have been some small accidents that were that that didn't reach me or were never reported to Tesla or something like that but no I'm not aware of actually of any in fact even the cases where people did pretty crazy things like you set the car in autonomous mode and then got into the back seat and that kind of thing oddly enough n none of those actually resulted in an accident that I'm aware of okay and we have a question from Mike Ramsey from The Wall Street Journal his question is when will you build in new autopilot hard Hardware into the Model S

15:02well I mean we don't want to comment on on future product transitions but it's it's it's not certainly not not immediately we are working on a new hardware Suite but it's it it'll be some time before that enters

15:27production all right sounds like maybe okay hello yes thank you elen what have you figured out about insurance costs or Insurance situations if the insured driver is not in the vehicle if something happens to it do you mean during the parking well I mean in this case it it's I would hesitate to call the parking a truly autonomous feature since you have to actuate it and the key has to be within a certain range it's more like a remote control feature at this point so I mean I guess the liability would be with the driver and we do take you know we are very clearly warning that this is beta feature you know we don't there are many different parking garage

16:17scenarios and it's not it may not work in all of them so it's important for people using the summon and and self Park capability to stay inside of the vehicle and be prepared to tap the key if they see it's going wrong so you can stop the auto park and summon function at any time just by tapping the key and that's really what we strongly recommend at this point so so yeah I mean the driver would be responsible for anything that that that happens although I I I suspect it's going to be somebody quite rare when that there's any issues they like very rare I mean the car can even detect if

17:02like a fairly small dog walks past so it's it's not absolutely perfect but I I I think it's you know probably probably better than a person right now even in you say like what's the probability of having a little offender scrape even with the beta version of summon an auto park I think it's probably less than a person and that's all the questions we have at this time all right thanks

17:49everyone ladies and gentlemen that concludes today's event you may now disconnect your lines

资源与出处

相关视频

同频道、同主题或相邻年份