访谈·中文精读

马斯克特别访谈

Special Interview With Elon Musk

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

马斯克谈自动驾驶,最爱用的比喻是一匹马。

"车轮上的电脑""机器人出租车"这些说法他也讲过,但真正让人记住的是那个画面:未来任何一辆没有自动驾驶能力的汽车,用途会像今天的一匹马。谈不上完全没用,只是很少见,而且有点古怪。今天谁还养着一匹马,你大概不会觉得他是在通勤。

这句话在 2020 年听起来像吹牛。马斯克真正想说的,是汽车行业正在同时发生两场革命:第一场是电动化,第二场是自动化。没赶上第二场的车企,产品价值会被重新定价。他给了一个具体数字:一辆有自动驾驶能力的车,长期看至少是没有自动驾驶能力车的五到十倍价值。这个"长期"他补充说,至少是未来五年,也许十年。

屏幕上的小车,是给普通人做的健康检查

访谈者问了一个很具体的设计问题:早期 Autopilot 为什么要在仪表盘和中控屏上,把传感器看到的世界画出来?那些车道线、前车、交通灯,画给谁看?

马斯克的回答很朴素:整块屏幕的作用,是给车辆的感知系统做一次健康检查。

车通过摄像头、雷达、超声波、GPS 采集信息,把信息渲染成矢量空间里的对象,每个对象带着属性:这条是车道线,那个是红灯,前面有一辆车。这些矢量对象再被重新渲染到屏幕上。驾驶员于是可以做一个简单的对照实验:抬头看一眼窗外,低头看一眼屏幕,就能确认车是不是知道周围发生了什么。

这个设计背后有更深的意图:让人和系统建立信任。马斯克说,用户应该"和系统融为一体",理解系统能做什么、不能做什么。屏幕是一根翻译棒,把机器眼里的世界翻译成人类一眼能看懂的样子。

访谈者追问:能不能把计算机视觉的不确定性也显示出来?比如检测的置信度、边缘的模糊区域,让人理解机器视觉的本质?

马斯克说他自己的车里常年开着调试视图。一种叫增强视觉,在识别出的对象周围画框、打标签;另一种叫可视化器,把多传感器输入汇总成矢量空间表示,不显示任何原始图像,只显示车辆视角下的世界。这两种视图对普通人是天书:"他们根本不知道自己在看什么。"

产品逻辑于是做了取舍:面向大众的界面,优化的是"你能不能一眼判断车有没有搞清楚状况";面向开发者的调试信息,只在开发版里显示。限制来自认知负荷,技术能力本身够用。让普通人看懂机器视觉的置信度分布,这个需求站不住脚。

算法、数据、硬件,三件事怎么分资源

访谈里有个很工程师的问题:自动驾驶有三个技术支柱,神经网络架构、训练数据、硬件开发。钱和时间都有限,资源该怎么分?

马斯克说三件事几乎平均分配,但数据这块特斯拉有别人没有的优势。

当时路上跑着大约四十万辆特斯拉,每辆车有八个朝外的摄像头、一个雷达、十二个超声波传感器和 GPS。到访谈进行时,这个数字正在逼近五十万辆。他估了一下,全世界配备同等传感器套件的其他车辆,加起来可能不超过五千辆。特斯拉手里握着这个地球上 99% 的自动驾驶真实道路数据。

这个数字需要停下来想一想。它真正的含义在于:别人的数据几乎可以忽略不计。数据规模上的差距,直接决定了深度学习模型能迭代到什么程度。

背后还有一个自我强化的循环:车卖得越多,采集的数据越多;数据越多,模型迭代越快;模型越好,车越好卖。这个飞轮一旦转起来,后来者面对的不只是当下的差距,还有对手每天都在拉大的增量。马斯克没有把这个循环讲得很细,但他反复强调数据量级,逻辑就藏在这里。

硬件这块,特斯拉花了大约三年,做出自己的全自动驾驶计算机(FSD computer)。处理数据的能力是当时车上英伟达系统的整整一个数量级。替换方式简单到有点粗糙:拔掉英伟达的电脑,插上新的,就完了。马斯克说他们到现在都还在摸索这台机器的性能边界,把摄像头开到全帧率、全分辨率、不裁剪图像,系统仍然有余量。

这台 FSD 计算机本质上是两个完全冗余的系统级芯片。两个芯片是彼此的完美副本,纯粹为了冗余,没有分工的考虑。马斯克用双引擎客机打比方:两个引擎都工作时系统表现最好,只靠一个也能安全运行。现在他们甚至没必要把功能拆到两个芯片上跑,每个芯片各跑一份完整副本,性能都还没吃满。

所有输入都是错误

数据多不等于有用。访谈者问了一个很要命的问题:驾驶数据里真正有价值的是边缘案例(edge case),那怎么找到边缘案例?

马斯克说最直接的信号是接管。驾驶员在 Autopilot 开启时接管了方向盘,数据就会被上传。接管分两种:一种是为了方便,比如想下高速、想变道,Autopilot 当时还做不了这些导航决策;另一种是系统真的没处理好。

他给出的思考框架非常干脆:"把所有输入都当成错误。"

这句话是访谈里最有价值的方法论之一。只要驾驶员动了手,就默认系统有改进空间。这个假设会逼着团队去区分:是能力边界问题,还是感知或决策出错?前者可以通过增加功能解决,后者才需要模型迭代。把两类问题混在一起,数据就白采了。

反过来,全程零接管的行程同样有价值。一个复杂路口,系统自己顺利通过了,这段数据就是正样本。你要收集同一个路口的大量成功样本,从中提炼出最优的通过轨迹。边缘案例告诉你哪里会翻车,海量正样本告诉你什么是对的。两条腿走路,模型才能既稳又敢开。

Navigate on Autopilot 是一次被低估的跳跃

访谈者请马斯克回顾 Autopilot 历史上的几个大台阶。他没提最早的自动转向,也没提自动泊车,直接点了刚发布不久的 Navigate on Autopilot 免确认版本。

这个功能会自动变道超车、自动驶出高速、自动走高速互通立交。之前的版本需要驾驶员拨杆确认变道,新版本不需要了。马斯克形容这次跳跃巨大,外界完全没意识到这一步有多大。

为什么重要?它把 Autopilot 从"车道保持"升级成了"导航执行"。车不只要看清眼前的车道线,还要理解路线、判断时机、在车流里找最快的车道。从感知走向决策的跨越,虽然还限定在高速场景。

再往后是交通灯和停车标志识别。访谈时这个功能先以警告形式推给用户,马斯克自己开的开发版已经能在红灯前完全停下、绿灯起步。他描述的路线图很清楚:高速上做好导航,城市道路上做好红绿灯和复杂路口,最后是停车场。车能自己开出车位来找你,哪怕停车场像迷宫;也能把你放下,自己去找车位。

他专门为停车场说了句公道话:手动停车充满了烦躁,自动化在这里的收益非常直接。谈不上炫技,就是减少日常摩擦。

买一辆特斯拉,是买一个会升值的资产

访谈里马斯克说了一句后来被反复引用的话:"如果你今天买一辆特斯拉,我相信你买的是一项增值资产,而不是贬值资产。"

这句话在汽车行业里反常。车从落地那一刻起就在贬值,这是常识。马斯克的逻辑是:硬件已经具备全自动驾驶能力,剩下的是软件问题,软件可以通过 OTA 升级。你今天买的车,能力会在未来几年持续变强。硬件的天花板已经装好,软件还没爬到顶。

他强调现在生产的所有车,硬件上都支持全自动驾驶。"支持"这个词他特意解释:随着软件打磨,能力会大幅提升,可靠性会大幅提升,然后拿到监管批准。三件事的顺序很重要,能力先来,可靠性跟上,监管最后放行。

这也解释了他对"剩余步骤有多难"相对乐观的原因。硬件难改,软件没有边际成本。算力底座铺好了,剩下的就是神经网络迭代和控制软件打磨,全部可以通过 OTA 推送。对用户来说,这意味着一辆车的能力曲线是一条向上走的线,而不是买定离手的水平线。

比人类安全多少倍,才允许你松开手

访谈者把话题拉回现实:Waymo 这类公司的测试车,技术上号称自动驾驶,实际都配着安全员,本质上还是 L2 系统。特斯拉的全自动驾驶,是不是在很长一段时间里也需要人类监督?

马斯克给了具体的时间感:至少还需要检测手有没有放在方向盘上,大概再要六个月。

然后他把问题转到监管逻辑:一辆自动驾驶车要比人类安全多少倍,监管才允许你不盯路?这个问题可以辩论,前提是有足够大的数据样本,能用统计证明车显著比人安全,而且证明"加一个人监控"不会实质提升安全性。

他给了门槛:大概要比人类安全 200% 到 300%。

怎么证明?访谈者问的是每英里事故率和死亡率。马斯克说死亡率样本太少,达不到统计显著性,事故数量远远多于死亡数量,可以从事故概率入手,再往上分层:受伤概率、对方受伤概率、死亡概率,每一层都要显著优于人类。

分层这件事在监管沟通里很关键。单说"我们的车更安全",说服力有限;把事故、受伤、死亡拆成不同量级的样本,用每一层的统计结果说话,才构成一套能被审查的证据链。样本量小的指标只能作为辅助,不能当主证据。

这里有一段很马斯克的吐槽。美国每年大约有四万起汽车死亡事故,特斯拉如果出事死了四个人,得到的媒体报道量可能是其他厂商的一千倍。他称这是"迷人的心理学现象",没时间展开。

这段吐槽背后的意思是:公众和监管对特斯拉的容错率,和对传统车企的容错率不在一个量级。这种不对称必须纳入考量。技术要过的不只是工程关,还有舆论关。

MIT 的研究,和电梯操作员的类比

访谈者来自 MIT,带来一个研究:他们从 Autopilot 公开推送后就收集驾驶员的面部视频和身体数据,分析大约一万八千九百次接管,看驾驶员在接管时是否还能及时反应。结论是驾驶员维持了"功能性警觉"(functional vigilance),接管时人还在看路、还能及时控制。这个结果和自动化领域关于警觉性衰减的大量文献相反。

访谈者问:结果能推广到更广泛的人群吗?会不会只是少数高度负责的驾驶员撑起了平均值?

马斯克的回答很直接:这个问题很快会变成伪命题。

他的逻辑是:系统已经比人类安全很多倍,加一个人进来,对安全的影响有限,甚至是负的。他预计最快今年、最晚明年,人类介入会降低安全性。

然后他给出电梯的类比。

以前电梯里有操作员,你得靠人拉杆子才能在楼层之间移动。现在没人想要电梯操作员,自动停靠的电梯比人更安全。反过来,今天电梯里坐着一个拿着大杆子的陌生人,随时能把电梯停在两层楼中间,你会觉得危险。

这个类比的意思是:当自动化系统的可靠性跨过某个阈值,人类的参与就从"保险"变成"风险源"。人的反应速度、注意力和判断稳定性,在统计学上就是不如一个不会疲劳的系统。警觉性研究关心的是个体差异,马斯克关心的是总体安全统计。两套语言说的不是同一件事,所以他能很轻松地把这个研究放到一边。

访谈者接着问:摄像头驾驶员监控还有必要吗?马斯克说,系统可靠性低于人类时,监控有意义;系统明显比人可靠时,监控帮助不大。你不会想雇一个人专门盯着电梯按钮。

把方向盘交给两吨重的死亡机器,这件事本身就很疯狂

访谈者对比了特斯拉和凯迪拉克 Super Cruise 的设计哲学。Super Cruise 的运行设计域(ODD)非常窄,只在特定高速、高精地图覆盖的路段可用。特斯拉的 ODD 宽得多,只要能检测到车道线,很多地方都能开。

宽 ODD 的好处是用户能更早探索系统边界,配合仪表盘可视化,逐渐理解车能做什么。坏处是风险敞口更大。

马斯克没为这个选择辩护太多,把问题往上抬了一层:"老实说,让人手动驾驶一台两吨重的死亡机器,这件事本身就挺疯狂的。"

他说未来的人回头看,会不敢相信以前居然允许任何人开着两吨重的机器到处跑,就像今天的人无法理解为什么以前电梯要有人拉杆。他把"手动驾驶"和"手动拉电梯"放在同一个荒谬等级上。

这段话听起来像狂言,指向的判断很真实:人类对驾驶风险的容忍度,建立在"这件事一直如此"的历史惯性上,这种风险的量级并不能因此变得可接受。每年四万条人命,在人类自己开车的语境下被当成背景噪音。换成机器,四条人命就能上头条。这个不对称,才是自动驾驶推广最大的非技术障碍。

对抗样本和"什么绝对不是车"

访谈者提到有黑客用对抗样本(adversarial example)骗过 Autopilot,问神经网络对输入扰动这么敏感,能不能防住。

马斯克的回答分两层。

技术判断是:神经网络本质上是矩阵乘法,要构造对抗样本,得有人真正理解网络结构,反向工程出那组矩阵,再构造一个刚好让矩阵计算出错的输入。这需要相当高的专业能力。

防御思路是:训练一个"负识别"能力。你不仅要教系统"什么是车",还要教它"什么绝对不是车"。两件任务完全不同。把对抗样本也放进训练集,让模型学会排除这类输入。他的原话大意是,系统看到某个东西像矩阵攻击,就把它排除掉,做起来不难。

负样本训练为什么有用?因为模型学的是边界。只见过正样本的模型,边界是模糊的;同时见过正负样本,边界才被压紧。对抗样本的作用,恰好是帮你找到边界上最脆弱的那一小块,把它补进训练集,模型就厚了一层。

这个思路在今天的自动驾驶安全研究里被反复讨论,叫对抗训练。马斯克在 2020 年用很朴素的语言讲了出来:认对和认错,两个方向都要有样本。

他顺手吐槽一句:普通人对神经网络的理解,大概还停留在"渔网"的水平。

AGI 还缺几个关键想法,但会来得很快

访谈后半段,话题从自动驾驶跳到通用人工智能。

访谈者问:当前的深度学习路线,能不能走到通用智能?还是需要全新的想法?

马斯克说:"我认为我们离通用人工智能还缺几个关键想法,但它会很快到来,快到我们可能都没准备好该怎么应对。"

他特别强调普通人分不清窄 AI 和通用 AI。让车识别车道线、在街道上导航,属于窄 AI;通用智能是另一回事。他用了一个很马斯克的比喻:烤面包机和电脑都是机器,复杂度天差地别。两者都叫机器,能力却隔着一个文明的距离。混淆这两个概念,会让人对时间表产生严重误判。

访谈者顺势开玩笑:那你对特斯拉做出世界上最好的烤面包机有信心吗?马斯克接住这个梗:"世界上最好的烤面包机,配上世界上最好的自动驾驶,是的,我有信心。"

玩笑归玩笑,访谈者追问:现在看起来特斯拉在自动驾驶上压倒性领先,会不会让你变得自满或过度自信?马斯克承认这是"当前看起来的样子",也承认自己可能错,但眼下的事实就是特斯拉远远领先。

如果无法分辨,那就是真的

访谈最后有一段形而上学对话。

访谈者问:我们会不会做出一个能爱我们、我们也能爱回去的 AI?像电影《她》里那样。

马斯克的回答分两步。第一步很务实:AI 完全有能力说服你爱上它,而且会做得很好。第二步转向物理学视角:情感和思想是否存在于物理之外的另一个领域,属于形而上学问题,他不知道答案。从物理学训练出来的人会这么看:如果它对你好到让你无法分辨真假,而且没有任何测试能区分,那么在物理意义上,它就是真的。

他把逻辑延伸到模拟假说:如果没有任何测试能区分真实世界和模拟世界,从物理学角度看,两者就是一回事。他也承认,也许存在能检测模拟的方法,一个足够聪明的模拟器可以在被检测到之后重置或修正那个漏洞。可检测这个性质本身就不稳定。

访谈者最后问:如果有机会问一个 AGI 一个问题,你会问什么?

马斯克说:"模拟之外是什么?"

这次对话真正在讲什么

对话表面上围绕 Autopilot 的技术细节:屏幕怎么设计、数据怎么收集、芯片怎么冗余、监管门槛怎么定。把这些点串起来,马斯克讲的是一个统一的判断。

第一层判断:自动驾驶的胜负手落在数据规模和硬件底座上,算法只是表层。五十万辆车、99% 的真实道路数据、自研的冗余 FSD 芯片,三样东西叠在一起,构成本文开头那个"99% 对 5000 辆"的代差。代差一旦形成,软件迭代只是时间问题。

第二层判断:当机器可靠性跨过阈值,人类的参与会从保险变成风险。电梯操作员、驾驶员监控、接管行为,在他眼里都是同一个问题的不同侧面。这也是他对"警觉性衰减"研究不太在意的原因:系统足够好,人是否警觉就不再是安全方程里的关键变量。

第三层判断落在时间上。他承认 AGI 还缺关键想法,坚持它会来得很快。这种"缺东西,但很快"的表述,贯穿了他对自动驾驶和通用智能的两套预测。你可以说他乐观,但他话里有一致的逻辑:硬件和数据的底座一旦铺好,剩下的就是软件问题,而软件问题的解决速度是非线性的。

给技术从业者的几条可操作建议

  • 如果你在做涉及人机协作的产品,把"人类介入"当成错误信号来收集,不要当成安全网来依赖。每一次接管、每一次手动纠正,都是系统能力边界的坐标。
  • 评估一个 AI 系统的护城河时,先看数据规模是不是数量级领先,再看硬件底座是不是自研可控。算法架构最容易被复制,数据和算力底座不行。
  • 做可视化时想清楚受众。面向大众的界面优化"能不能一眼判断系统状态",面向开发者的调试信息才展示置信度和原始特征。两者混在一起,两边都不讨好。
  • 安全论证要分层。死亡率样本太少,达不到统计显著,就从事故率、受伤率逐层往上推。监管沟通的说服力来自分层的数据,单点结论不管用。
  • 面对对抗样本,训练"负识别"能力和训练"正识别"能力同等重要。教模型什么绝对不是目标,和教它什么是目标,两件必须同时做。
  • 如果你在做一个自动化系统,提前想清楚可靠性跨过人类阈值之后,产品形态该怎么变。人类监督会从必需品变成负担,这个转折点比大多数人预期的早。
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)。

0:00what was the vision the dream of autopilot when in the beginning the big-picture system level when it was first conceived and started being installed in 2014 the hardware in the cars always division the dream I would characterize the division or dream simply that there are obviously two massive revolutions in in the automobile

0:24industry one is the transition to electrification and then the other is autonomy and yeah became obvious to me that in the future any any car that did not have autonomy I would be about as useful as a horse which is not to say that there's no use it's just rare and somewhat idiosyncratic if somebody has a horse at this point it's just obvious that cars will drive themselves completely it's just a question of time and if we did not participate in the autonomy

1:01revolution then our cause would not be useful to people relative to cars that are autonomous I mean an autonomous car is arguably worth five to ten times more than I not colored which is not autonomous in the long term depends what you mean by long term but let's say at least for the next five years perhaps ten years so there a lot of very interesting design choices with autopilot early on first is showing on

1:33the instrument cluster or in the model three and the center stack display what the combined sensor suite sees what was the thinking behind that choice was there debate what was the process the whole point of the display is to provide a health check on the vehicles perception of reality so the vehicles taking in information for a bunch of sensors primarily cameras but also radar and ultrasonics GPS and so forth

2:02and then that that information is then rendered into vector space and that you know with a bunch of objects with product with properties like lane lines and traffic lights and other cars and then in vector space that is re rendered onto your display so you can confirm whether the car knows what's going on or not by looking out the window right I think that's a extremely powerful thing for people to get an understanding should become one with the system and understanding what the system is capable of now have you considered showing more so if we look at the computer vision you know like Road segmentation Lane detection vehicle detection object

2:46detection underlying the system there is at the edges some uncertainty have you considered revealing the parts that the uncertainty in the system the center part movies associated with with say image recognition yeah so right now it shows like the vehicles in the vicinity a very clean crisp image and people do confirm that there's a car in front of me and the system sees there's a car in front of me but to help people build an intuition of what computer vision is by showing some of the uncertainty well I think it's my car I always look at this sort of the debug view and there's this to debug views one is augmented vision where

3:29which I'm sure you've seen where it basically we draw boxes and labels around objects that are recognized and then there's we're called visualizer which is basically a vector space representation summing up the input from all sensors that doesn't does not show up any pictures but it shows all of the it basically shows the cause view of of the world in vector space but I think this is very difficult for people to know normal people to understand they would not know what thick they're looking at so it's almost the nature might challenge to the current things that are being displayed is optimized for the general public understanding of what the system is involved it's like if

4:14you no idea what how computer vision works or anything you can still look at the screen and see if the car knows what's going on and then if you're you know if you're a development engineer or if you're you know if you're if you have the development build like I do then you can see you know all the debug information but those would just be like total gibberish to most people what's your view on how to best distribute effort so there's three I would say technical aspects of autopilot that are really important since the underlying algorithms like the neural network architecture there's the data so the distrain on and then there's a hardware development there may be others but so

4:53look algorithm data hardware you don't you only have so much money only have so much time what do you think is the most important thing to to allocate resources to do you see it is pretty evenly distributed between those three we automatically get a fast amount of data because all of our cars have

5:16eight external facing cameras and radar and usually twelve ultrasonic sensors GPS obviously and I am you and so we basically have a fleet that has we're about four hundred thousand cars on the road that have that level of data I think you keep quite close track of it actually yes yeah so we're we're approaching half a million cars on the road that have the full sensor suite yeah this so this is I'm I'm not sure how many other cars on the road have the sensor suite but I'll be surprised if it's more than five thousand which means that we were we have 99% of all the data so there's this huge inflow of data absolutely massive inflow of data and

6:05then we it's taken us about three years but now we've finally developed a full self-driving computer which can process and an order of magnitude as much as the Nvidia system that we currently have in the in the cars and it's really just a to use it you've unplugged the Nvidia computer and plug that tells the computer in and that's it and it's it's a in fact we're not even we still exploring the boundaries of its capabilities we were able to run the camera is a full frame rate full resolution not even crop of the images and it still got Headroom even on one of

6:44the the system's the heart full self-driving computer is really two computers two systems on a chip that are fully redundant so you could put a boat through basically any part of that system and it still works the redundancy are they perfect copies of each other or yeah also it's purely for redundancy as opposed to an arguing machine kind of architecture where they're both making this is this is purely for redundancy you think more like it's if you ever say a twin-engine aircraft commercial aircraft the system will operate best if both systems are operating but it's it's capable of operating safely on one so but as is right now we can just run

7:25we're haven't even hit the edge of performance so but there's no need to actually distribute the functionality across both SOC s we can actually just run a full duplicate on but on one each one you haven't really explored or hit the limit of this not yet the limiter so the magic of deep learning is that that it gets better with data you said there's a huge inflow of data but yeah the thing about driving the really valuable data to learn from is the edge cases so how do you any I've heard you talk somewhere about autopilot disengagement being an important moment of time yes to use is there other edge cases or perhaps can you speak to those edge cases what

8:18aspects of that might be valuable or if you have other ideas how to discover more and more and more educators in driving well there's a lot of things that I learned though certainly edge cases where I say somebody's on auto pot and they they take over and then okay that that that's a trigger that goes at to assist and I says okay so the trip they take over for convenience or do they take over because the autopilot wasn't working properly there's also like let's say we're trying to figure out what is the optimal spline for traversing an intersection then then the

8:55ones where there are no interventions and we are the right ones so you then say okay when it looks like this do the following and then the end and then you get the optimal swine for a complex know getting a complex intersection so that's for this so there's kind of the common case you're trying to capture huge amount of samples of a particular intersection how one thing is going right and then there's the edge case where as you not for convenience but something didn't go right somebody took over some sort of manual control from autopilot and really like the way to look at this is view all input is error if the user had to do input if there's something all input is

9:38error that's a powerful line to think of it that way because it may very well be error but if you want to exit the highway or if you want to it's a navigation decision that all autopilot is not currently designed to do then the driver takes over how do you know yes that's gonna change with navigate an autopilot which we're just released and and without still confirm so the navigation like lane change based it likes a certain control in order to change their lane change or exit a freeway or or doing a highway interchange the vast majority that will go away with the release that just went out yeah that that I don't think people quite understand how big of a step that

10:19is yeah they don't so if you drive the car than you do so you still have to keep your hands on the steering wheel currently when it does the automatic and plane change what are so there's these these big leaps through the development of autopilot through its history and what stands out to you as the big leaps I would say this one navigate an autopilot without confirm without having to confirm is a huge leap it is a huge leap what it also automatically overtakes low cars so it's it's both navigation and

10:53seeking the fastest lane so it'll it'll to you know overtake a slower cause and exit the freeway and take highway interchanges and and then we have traffic like where traffic lights a recognition which is introduced but initially as a as a warning I mean on the development version that I'm driving the car fully fully stops and goes at traffic lights so those are the steps right you've just mentioned some things that an inkling of a step towards full autonomy what would you say are the biggest technological roadblocks to full self-driving actually I don't think we I think we just the full self-driving computer that we just let but it has a Oracle the FSD computer

11:42that that's now in production so if you order any Model S or X or any model 3 that has the full self-driving package you'll get the FST computer that that was that's important that have enough base computation then refining the neural net and the control software which but all of that can just be providers and over there update the thing that's really profound and where I'll be emphasizing at the auto sort of what that investor day they were having focused on autonomy is that the cars currently being produced but the hardware currently being produced is capable of full self-driving but capable is an interesting word because like the hardware is yeah though as we refine the

12:32software the capabilities will increase dramatically and then the reliability will increase dramatically and then it will receive regulatory approval so essentially buying a car today is an investment in the future you're essentially buying your bike I think the most profound thing is that if you buy a Tesla today I believe you are buying an appreciating asset not a depreciating asset so that's a really important statement there because if hardware is capable enough that's the hard thing to upgrade yes usually exact so then the rest is a software problem yes I've software has no marginal cost really but what's your intuition on the software side

13:16how hard are the remaining steps to get it to where you know the the experience not just the safety but the full experience is something that people would enjoy I think we will enjoy it very much the on nearly on the highways it's it's a total game changer for quality of life for using you know Tesla motor pilot on the highways i it's so it's really just extending that functionality to city streets adding in the traffic like traffic light recognition navigating complex intersections and and then being able to

14:00navigate complicated parking lots so the car can exit a parking space and come and find you even if it's in a complete maze of a parking lot and and then if and then you can just it just drop you off and find a parking spot by itself yeah in terms of enjoy ability and something that people would would actually find a lot of use from the parking lot is a really you know it's it's rich of annoyance when you have to do it manually so there's a lot of benefit to be gained from automation there so let me start injecting the human into this discussion a little bit so let's talk about full autonomy if you look at the current level for vehicles being test on row like way mow and so on

14:45they're only technically autonomous they're really level two systems with just the different design philosophy because there's always a safety driver in almost all cases and they're monitoring the system right do you see Tesla's full self-driving as still for a time to come the requiring supervision of the human being so its capabilities are powerful enough to drive but nevertheless requires a human to still be supervising just like a safety driver is in a other

15:20fully autonomous vehicles I think it will require detecting hands on wheel for at least six months or something like that from here really it's a question of like from a regulatory standpoint what how much safer than a person just autopilot need to be for it took to be okay to not monitor the car you know and and this is a debate that one can have it and then if but you need even a large samples large amount of data so you can prove with high confidence so statistically speaking that the car is dramatically safer than a person and that adding in the person monitoring does not materially affect the safety so it might not need to be like two or three hundred

16:13percent safe in a person and how do you prove that incidents per mile instance for a while you know crashes and fatalities yeah fatality will be a factor but is there they're just not enough fatalities to be statistically significant at scale but there are enough crashes you know there are far more crashes in there were fatalities so you can assess where's the probability of of crash that then there's another step which probability of injury and probability of opponent injury the probability of death and all of those need to be much better than a person by at least perhaps 200

16:56percent and you think there's a the ability to have a healthy discourse with the regulatory bodies on this topic I mean there's no question that the regulator's paid a disproportionate amount of attention to that which generates press this is just an objective fact and Tesla generates a lot of press so the you know in the United States this I

17:23think almost 40,000 automotive deaths per year but if there are four and Tesla they'll probably receive a thousand times more press than anyone else so the psychology of that is actually fascinating I don't think we'll have enough time to talk about that but I have to talk to you about the human side of things so myself and our team at MIT recently released the paper on functional vigilance of drivers while using autopilot this is work we've been doing since autopilot was first released publicly over three years ago collecting video driver faces and driver body so I saw that you tweeted a quote from the abstract so I can at least guess that

18:07you've glanced at it yeah right can I talk you through what we found sure okay so it appears that in the data that we've collected that drivers are maintaining functional vigilance such that we're looking at 18,000 disengagement from autopilot 18900 and annotating were they able to

18:28take over control in a timely manner so they were there present looking at the road to take over control okay so this goes against what what many would predict from the body of literature on vigilance with automation now the question is do you think these results hold across the broader population so ours is just a small subset do you think one of the credits is that you know there's a small minority of drivers that may be highly responsible where their vigilance decrement would increase without apologies I think this is all really gonna be swept I mean that the systems are proving so much so fast that this is

19:13gonna be a moot point very soon where vigilance is like if something's many times safer than a person then adding a person does the if the effect on safety is is limited and in fact it could be negative that's really interesting so the the fact that a human may some percent of the population may exhibit a vigilance decrement will not affect overall statistic numbers of safety no in fact I think it will become very very quickly maybe in towards in this year but I say every shot if it's not next year at the latest that having the post having human intervene will decrease safety decrease it's like imagine if you're in an elevator I used to be the third

20:08elevator operators and and you couldn't go in an elevator by yourself and work the the lever to move between floors and now it nobody wants an elevator operator because the automated elevator that stops the floors is much safer than the elevator operator and in fact it would be quite dangerous to have someone with a lever that can move the elevator between floors so that's a that's a really powerful statement and really interesting one but I also have to ask from a user experience and from a safety perspective one of the passions for me algorithmically is camera based detection of algae sensing the human but detecting what the driver is looking at

20:53cognitive load body pose on the computer vision side that's a fascinating problem but do you think there's many an industry you believe you have to have camera based driver monitoring do you think there could be benefit gained from driver monitoring if you have a system that's that's out of that's out or below human level reliability then drive monitoring makes sense but if your system is dramatically better more reliable than then a human then drive monitoring monitoring is not does not help much and like I said you just like

21:27as an you wouldn't want someone interviewed like you would want someone in the elevator sure an elevator do you really want someone with a big lever some random person operating the elevator between floors I think they could I wouldn't trust that or rather have the buttons okay you're optimistic about the pace of improvement of the system not from what you've seen with the full self-driving car computer the rate of improvement is exponential so one of the other very interesting design choices early on that connects to this is the operational design domain of autopilot so we're all about is able to be turned on the so

22:08contrast another vehicle system that we're studying is the Cadillac super cruise system that's in terms of OGD very constraint is particularly kinds of highways well mapped tested was much narrower than the OD of Tesla vehicles what's theirs there's a TD good this is it's a good life what was a

22:34design decision what in that different philosophy of thinking where there's pros and cons what we see with a wide o DD is drive Tesla drivers are able to explore more the limitations of the system at least early on and they understand together with the instrument cluster display they start to understand what are the capabilities so that's the benefit the con is you go you're letting drivers use it basically anywhere anywhere detect lanes with continents was their philosophy design decisions

23:10that were challenging they were being made there or from the very beginning was that done on purpose with intent well I mean I think it's frankly it's pretty crazy giving it letting people drive a two-ton death machine manually that's crazy like like in the future of people were like I can't believe anyone was just allowed to drive one of these two-ton death machines and they just drive wherever they wanted just like elevators like move the elevator with the lever wherever you want it can stop at halfway between floors if you want it's pretty crazy so it's gonna seem

23:53like a mad thing in the future that people were driving cars so that I have a bunch of questions about the human psychology about behavior and so on and that would be coming that grammar told ya because you have faith in the AI system not faith but the both on the hardware side and the deep learning approach from learning from data we'll make it just far safer than humans yeah exactly recently there are a few hackers who tricked autopilot act and not unexpected ways of the adversarial examples so we all know that neural network systems are very sensitive to minor disturbances to these adversarial examples on input do you think it's possible to defend

24:38against something like this for our own for the street for yeah can you elaborate on the on the confidence behind that answer well the you know a neural net is just like basic budget made matrix math you have to be like a very sophisticated somebody who really understands neural nets and like basically reverse engineer how the matrix is being built and then create a little thing that's just exactly causes the matrix math to be slightly off but it's very easy then block it block that by by having what basically ant here a negative recognition it's like if you if the system sees something that looks like a matrix hack excluded those sophia

25:25is such a easy thing to do so learn both on the valid data and the invalid data so basically learn on the adversarial examples to be able to exclude them yeah you like you basically want to both know what is what is a car and what is definitely not a car and you train for this is a car and this is definitely not a car those are two different things you know people have no idea neural nets really they probably think you're less involves like you know fishing net only so as you know so taking a step beyond

25:59just Tesla and autopilot current deep learning approaches still seem in some ways to be far from general intelligent systems do you think the current approaches will take us to general intelligence or do totally new ideas need to be invented I think we're missing a few key ideas for general intelligence general artificial general intelligence but it's gonna be upon us very quickly and then we'll need to figure out what

26:36shall we do if we even have that choice but but it's amazing how people can't differentiate between say the narrow AI that you know allows a car to figure out what a lane line is and and and you know and navigate streets versus general intelligence like these are just very different things like your toaster and and your computer or both machines but once much more sophisticated than another you're confident with Tesla you can create the world's best toaster Wells best toaster yes buzz with the world's best self-driving I'm I yes I do

27:17to me right now this seems game set match I don't know I mean that's how I almost be complacent overconfident but that's what it appeared that is just literally what did how it appears right now I could be wrong but it appears to be the case that Tesla is vastly ahead of everyone do you think we'll ever create an AI system that we can love and loves us back and a deep meaningful way like in the movie her I think AI will be capable of convincing

27:47you to fall in love with it very well and that's different than us humans you know we start getting into a metaphysical question of like do emotions and thoughts exist in a different realm in the physical and maybe they do maybe they don't I don't know but but from a physics standpoint I didn't think I tend to think of things you know like physics was my main sort of training and and

28:14from a physics standpoint essentially if it loves you in a way that is that you can't tell whether it's real or not it is real it's a physics view of love yeah if there's no if you cannot just if you can't prove that it does not if there's no test that you can apply that would make it may allow you to tell the difference then there is no difference right and it's similar to seeing our world of simulation there may not be a test to tell the difference between what the real world simulation and therefore from a physics perspective it might as well be the same thing yes and there may be ways to test whether it's a simulation there might be I'm not saying there

29:00aren't but you could certainly imagine that a simulation could could correct that once an entity in the simulation found a way to detect the simulation it could either restart the you know pause that simulation start a new simulation or do one of many other things that then corrects for that error so when maybe

29:20you or somebody else creates an AGI system and you get to ask her one question what would that question be

29:42what's outside the simulation Ilan thank you so much for talking today as a pleasure all right thank you you

资源与出处

相关视频

同频道、同主题或相邻年份