马斯克出现在大屏幕上的时候,背景是火星探测器“毅力号”的模型。主持人说这画面看着有点像火星,他笑着接了一句,那是毅力号在火星上的样子。这场对谈发生在 X Takeover 现场,前身是 Tesla Takeover,他已经连续两年来参加。台下坐着一千五百多个从世界各地赶来的车主。
两个多小时里,他喝着一罐健怡可乐,从星舰的隔热罩聊到人的手为什么让机器人工程师头疼,从 AI 是超音速海啸聊到自己为什么觉得活在当下这个时代是最幸运的事。没有产品发布,没有幻灯片,问题一个接一个,回答也一个接一个。
对谈快结束时,几位车主代表上台,递给他一块定制的砖头。砖上刻着一句话:“这块砖会飞。” 几千封信被压缩进这块砖里,代表全世界车主的心意。他们的意思很直接:你把不可能变成了可能,现在轮到我们相信不可能了。
我把整场对谈里信息量最大的部分重新整理了一遍,按主题排了顺序。如果你关心 AI、火星、人形机器人和脑机接口这几件事的长期走向,下面这些内容值得慢慢看。
星舰最难啃的骨头不在发动机,在可重复使用的隔热罩
主持人第一个问题问得很直接:星舰是人类造过最有野心的火箭,哪件事比预想的更难,下一个该盯的里程碑是什么?
马斯克先纠正了一个前提。他说自己从来没觉得这件事会容易,一开始就认定全都很难。然后他给了具体参数:星舰的推力是土星五号登月火箭的两倍半,未来版本会到三倍。土星五号在此之前是人类造过的最大火箭、最大的飞行物体。星舰的重量大约是此前最大飞行物体的两倍,目标是完全、快速地重复使用。
“这真的是一件疯狂的事。我认为它属于现存最难的工程挑战之一。”(马斯克)
他提到一个自己发明的词,叫“傻笑指数”(giggle factor)。当年刚开始讨论星舰,公司内部的人一说完这想法就开始笑,觉得太荒唐。到今天,荒唐感已经变成了一种迟到的现实。
“我大概可以称得上专攻一件事,就是把不可能变成仅仅迟到。曾经是不可能,现在只是晚了。”(马斯克)
真正卡住进度的,是完全可重复使用的轨道级隔热罩。这件事人类从来没做成过。航天飞机的隔热罩每次飞行都会掉落大量隔热瓦,每次回来都要翻修,所谓可复用只是名义上的。猎鹰 9 号让一级助推器实现了常态化回收,这是人类第一次让火箭复用真正具备经济意义。但二级返回、隔热罩无损复用,到今天还是空白。
除了隔热罩,还有两件事同样棘手:让上面级飞船自己降落,以及让巨大的金属筷子在半空中把它夹住。马斯克形容这套动作看起来像一部科幻电影,而且是一部不太可信的科幻电影。用金属筷子去夹住人类造过最大的飞行物体,他本人的评价是“挺野的”。
他对时间表的判断是,飞船有机会在今年回收,最晚明年上半年。之后还要持续改进,让飞船和助推器不只是可复用,而是完全且快速地复用。
星舰的终极目标:每公斤入轨成本低于猎鹰 1 号
完全复用实现之后,成本会降到什么程度?马斯克给了一个非常具体的对照:
把一百吨以上的有效载荷送上轨道的成本,会低于一枚一次性猎鹰 1 号火箭把半吨载荷送上轨道的成本。
这个数字能成立,原因是可复用火箭真正消耗掉的只有推进剂,箭体本身反复使用。推进剂里大约八成是液氧,两成是燃料。他打了一个比方:如果你每次开车出门都要把车扔掉,还得再拖一辆车跟在后面供返程用,没有人会开车。可火箭行业过去几十年就是这么运作的。
隔热罩问题解决之后,下一个大关卡是轨道加注。马斯克特意纠正了用词,他说这不叫 refueling,叫 refilling,因为转移过去的东西主要是液氧。两艘星舰要在轨道上对接,把推进剂从一艘转到另一艘,未来还要对接轨道推进剂仓库。
他说这件事听起来不难,实际上非常棘手,而且全世界没有第二个组织在尝试。他甚至反问了一句:还有谁在做可复用助推器?答案是没有人。哪怕猎鹰 9 号已经用多年时间证明可复用助推器在经济上完全说得通,全球依然没有第二家公司做到常态化复用。
这个事实比技术难度本身更值得琢磨。行业的惯性往往比工程难度更顽固。 大家不是做不到,是没人愿意推翻整个体系重新设计一遍。
火星城市的治理交给火星人,地球只用花百分之一的精力
主持人问:一个自给自足的火星城市到底长什么样,多少人,什么经济结构,什么治理方式?
马斯克的回答很坦率:不知道。火星还没有完成地球化改造,人没法在户外活动,初期只能住在玻璃穹顶里,出门必须穿火星服。至于治理,他说那将由火星人自己决定,皮球踢得很干净。
但他对“为什么值得去火星”有一套很稳的论证,分成两条。
第一条是防御性理由。地球可能出事,可能是世界大战,也可能是恐龙灭绝那种级别的陨石。人类意识是宇宙里一盏很小的烛火,不能让它熄灭。其他物种没有能力把生命扩展到另一颗行星,只有人类能做到。成为多行星物种,等于给意识买了一份长期保险。
“我认为百分之九十九的精力应该放在地球上,但也许百分之一应该用来成为多行星物种。”(马斯克)
第二条是激励性理由。生活不能只是解决一个接一个痛苦的问题,还得有让人早上醒来就想看看今天会发生什么的东西。阿波罗计划就是这样的存在:几乎没有几个人真的去过月球,但全人类都间接去了一次,那是整个人类的共同成就。一个不断向外探索、把发现传回地球的航天计划,对地球上每个人都有激励作用。
他还说自己是个乐观的人,最可能的结局是未来会变好,只是有一小部分概率会出岔子。所以要提前留一手,不能让意识之光熄灭。
至于 SpaceX 会不会走向星际航行,他说希望如此,但大概率不在自己这一代人的有生之年实现。
Cybercab 不会取代 Model 3 和 Model Y
有人问 RoboTaxi 平台怎么和特斯拉现有的产品线共存,会不会取代 Model 3。马斯克从座位数切入回答。
Cybercab 是两座车,Model 3 和 Model Y 分别覆盖四座和六座需求。他的设想是动态调度:一两个人出行,系统派 Cybercab;四个人,派 Model 3;六个人,派 Model Y;将来还会有更大的车型。这套逻辑听起来像网约车的车型分级,区别在于车是无人驾驶的,调度效率会高出一个量级。
关于车队归谁所有,他给了一个很清楚的类比:
“可以把它想成优步和爱彼迎的结合。一部分车是特斯拉自己拥有的,一部分是车主拥有的,车主可以随时把车加进车队,或者随时抽出来。”(马斯克)
这个设计解决了两件事。对特斯拉来说,不必独自扛下几千亿美元的车队资本开支;对车主来说,车闲置的时候还能挣钱,把一辆车低得可怜的资产利用率拉高好几倍。这套模式一旦跑通,卖车的商业逻辑会彻底改变。
主持人接着抛出一个更尖锐的问题:如果 RoboTaxi 成功,特斯拉造的是世界上最好的车,但社会上需要的车反而会变少,这个悖论怎么解?
马斯克的回答很直接:“特斯拉是上市公司,你能问的最有意思的问题,恰恰是我最不能回答的问题。” 这句话他在这场对谈里用了不止一次,凡是涉及公司估值和未来产品规划的问题,他都用类似的方式挡了回去。他也解释,这不是不愿意聊,是公开上市公司的信息披露有严格限制。
Roadster 年底做演示,但它不影响财务报表
有人提到他几周前发过一条动态,说 Roadster 项目还在推进。马斯克确认,目标是今年年底做一次演示,也可能推到明年年初。
他特意补了一句,Roadster 不会对公司的财务表现产生实质影响,因为相对其他项目它太小了。但它很酷。
这句话背后的信息不少。特斯拉当前的资源排序很清楚:自动驾驶、Optimus、储能排在最前面,Roadster 属于给品牌加分的情怀项目,不是战略重心。想等这款车的人要有耐心。
Optimus 可能是人类历史上最大的产品
主持人提到马斯克说过的一句话:Optimus 的价值可能超过特斯拉整个汽车业务。他先照例说这涉及公司估值不能细讲,然后给了两个关键判断。
第一个判断是第三代设计就是量产版的正解。V3 相对 V2 是彻底重做,几乎什么都不保留。他当时面临一个选择:是今年先造几千台已知设计不够好的 V2,还是干脆停下来跳到 V3,代价是年底手上只有几百台机器人。
“我认为正确的决定是选 V3,然后明年大幅放量。”(马斯克)
第二个判断是市场空间到底有多大。他绕开财务预测,问了一个更朴素的问题:谁不想要一个自己的 C-3PO 或者 R2-D2? 现场观众举手。他接着说,Optimus 的功能会比那两个科幻角色强得多,会非常实用,所以地球上几乎每个人都会想要一个,工业场景还会需要一大批。
然后他做了一个粗略估算。假设年产十亿台,规模化之后成本会越来越低,单台成本大概能降到三万美元左右。这个数字乘以十亿,对应的年收入是三十万亿美元。他自己也承认,从今天到年产十亿台,中间有极长的一段路要走。
关于 Optimus 最先在哪些场景创造价值,马斯克给了一个排序。
- 高价值医疗照护。 比如给四肢瘫痪的人当二十四小时的护士或者助手,给需要物理帮助的人当帮手。这类场景对当事人是改变人生的,价值最高。
- 危险作业。 任何有生命危险或者容易受伤的工作,都是 Optimus 早期适合介入的领域。
- 重复性任务。 制造业流水线这类枯燥重复的工作。
主持人问:十年后回头看,全自动驾驶和 Optimus 哪个更具颠覆性?马斯克没有犹豫,答案是 Optimus。
AI 是超音速海啸,Grok 的目标是最大化求真
有人问:数字超级智能可能今年或明年就会到来,我们怎么知道自己越过了那条线,第一天会是什么样子?
马斯克的回答有点绕,但核心意思清楚:这件事可能没有想象中那么戏剧化。他给了一个衡量标准:AI 到今天还没有发明过任何有用的新技术,也还没有发现新的物理规律。 这两件事一旦发生,就是质变的信号。
关于“Grok 比人聪明之后人类扮演什么角色”,马斯克说他很早就判断数字超级智能必然出现,不管他参不参与。美国至少会有四五个主要的大模型,就算 xAI 不存在,超级智能一样会来。
“我面前的选择是当观众还是当参与者。这件事反正都会发生,那我宁愿当参与者。这样我还能把精力放在 AI 安全上。”(马斯克)
他对 AI 安全的核心判断只有一句话:最关键的是最大化求真。 求真意味着敢说政治不正确但事实上正确的话,意味着犯错之后承认错误并且纠正。他把这和“发明新技术”“解决医学难题”看作同一件事:想治好癌症,就必须最大程度地求真相,搞清楚病因到底是什么。
他对 AI 发展速度的描述用了一个不太舒服的比喻:
“我见过很多技术演进,但从来没见过任何技术像 AI 这样发展。这可能是个让人不安的比喻,AI 是一场超音速海啸。”(马斯克)
他还强调 AI 必须和机器人结合。软件层面的智能如果能驱动物理世界的执行器,影响才会真正落到现实里。这也是 xAI 和 Optimus 在他那里被放在同一个框架下思考的原因。
好的未来是物质极大丰富,难题变成“活着的意义”
主持人问:如果 AI 和机器人把大部分工作都干了,人类还剩下什么?
马斯克承认这是他自己真正纠结的问题,也是他做 Neuralink 的原因之一。他描述的好版本未来是这样的:产品和服务不再稀缺,任何人都能拥有他们想要的任何东西。AI 和机器人带来的生产力提升,可能让经济规模扩大十倍以上,贫困会被消除。
但更大的挑战随之而来。如果机器人什么都能做,而且做得比你好,你怎么找到活着的意义?
“这可能是最大的挑战。”(马斯克)
他给了一个类比。你骑自行车、骑摩托、开汽车的时候,你在移动能力上已经远远超过任何靠双腿奔跑的人。这没有让跑步变得没有意义。人们照样跑步,照样比赛。人类和机器结合之后,可能进入类似的状态:你不再和机器比谁更强,而是在一个被增强的层面上继续做人的事。
蒸汽机出现的时候,人们也辩论过同样的问题。很多人从“我比机器干得快”里获得价值感,机器变快之后,这种价值感就塌了。马斯克说他能给的最好的答案是:让人类智能和机器智能交织在一起,通过 Neuralink 提高人机带宽,最终实现某种程度上的与 AI 合一。
他对未来的概率判断是,大概百分之八十到九十会是好的,剩下百分之十到二十可能不太好。他用了一个词叫终结局,说完自己补了一句,那种情况会很糟。
Neuralink 从医疗出发,终点是概念层面的心灵感应
Neuralink 目前有八位患者,全部是四肢瘫痪或者失去身体机能的人,他们能用意念控制电脑。八个人都很喜欢这个设备,每天都在用。团队今年的目标是二十位患者,明年会推出“盲视”(Blindsight)植入方案,让完全失明、连视神经都失去的人重新获得视觉。现场为这一段鼓了掌。
马斯克把 Neuralink 的路线分成三个阶段。
- 修复严重的医学问题。 恢复运动能力、视觉、听觉、嗅觉、触觉,治疗癫痫、中风之后失去的行动能力,理论上还能修复精神分裂症。任何和大脑、神经元、感官相关的问题,长期看都可能通过 Neuralink 解决。他把这个设备描述成一种非常强大的通用输入输出装置。
- 验证安全性和有效性之后,进入增强阶段。 大幅扩展智力,实现超高带宽的人机通信。
- 远期可能实现记忆的上传和下载。 把你的记忆存一份,将来下载到一个机器人身体或者克隆体里。马斯克自己说这一段很科幻,但他认为会实现。
他特别强调,这些东西不是给少数精英准备的,是任何想要的人都能用上。
最有意思的部分是他对“概念心灵感应”的描述。人的大脑里有一个复杂的想法或者画面,要把它压缩成语言,对方听到之后再解压,这个过程损耗极大。你脑子里的画面,用语言根本传不过去。但两个人都装了 Neuralink 之后,可以不经压缩地把一整套概念直接传给对方的大脑。
“这就是我称之为概念心灵感应的东西,它会极大改善人与人之间的沟通和理解。”(马斯克)
关于技术难点,他说目前没有看到无法突破的障碍。真正拖慢速度的是安全验证,往人脑里装东西必须极度谨慎,还要和 FDA 紧密配合。不同的问题需要不同版本的设备。“心灵感应”产品读取的是运动皮层的信号,“盲视”产品要往视觉皮层写入像素,电极要插得更深,而且是写不是读。后一项技术在猴子身上已经跑了三年。有一只猴子同时装了两种植入体,状态很好。
他特意花时间讲动物福利。检查设施的美国督察员说,她整个职业生涯没见过比这更好的猴子饲养环境。猴子喜欢打游戏吃零食,和人类一样。有只叫佩奇的猴子坐在树枝上喝香蕉冰沙,每得一分喝一口,它不是被按住的,它是自己喜欢玩这个游戏。
造人形机器人,让马斯克重新认识了人的手
有人问造 Optimus 和做 Neuralink 之间有没有互相启发。马斯克的回答很出人意料:造人形机器人让他对身体有了全新的敬意,人的身体设计得太精妙了。
他顺带吐槽了脊椎:为什么人类要有这么多背痛,这很烦人。然后接了一句,背痛将来 Neuralink 也能治。
真正的重点是手。
“Optimus 的手占整个机电问题的一半以上,身体其余所有部分加起来是另一半。”(马斯克)
他请观众看自己的手,动一动,再摸一下前臂。你会发现控制手指的肌肉几乎都在前臂里,手掌本身只有很少的肌肉。手是通过肌腱从前臂像提线木偶一样被操纵的。这是一套极其精密的远程驱动系统,工程上想复现,难度非常高。
除了手,Optimus 的很多部件可以直接复用汽车上的东西。Autopilot 的 AI4 计算机直接拿来当大脑,小电池包和汽车电池包同源,未来的 AI5 硬件也会同步上车和上机器人。软件层面的 AI 高度相似。真正卡住 Optimus 进度的,就是那只手。
即便如此,马斯克对 Optimus 的判断没有动摇:它会成为有史以来最大的产品。
X 的目标是成为最不不可信的信息源
关于 X,马斯克给了一个很具体的进度描述。从只能发短文本和两分钟视频,到现在可以发四小时视频,可以在平台上写一整本小说,X 已经变成全模态平台。私信系统正在升级成 XChat,支持音视频通话,全程端到端加密,点对点,原理类似比特币。
“测试标准是,就算有人拿枪指着我的头,我也读不了你的消息。”(马斯克)
支付功能 X Money 已经在公司内部做 beta 运行,预计几个月内对外发布。马斯克提到这件事的时候有点感慨。当年做 X.com,后来变成 PayPal,现在绕了一大圈又回到支付,算是把当年没做完的产品计划补完。他说这件事有一种诗意的意味。
他对算法的评价很直接:“我们的算法现在确实挺烂的,抱歉。” 问题在于推送过度集中。某一天的热点是什么,算法就把那个东西推给你二十遍。他说悉尼·斯威尼那条牛仔裤的新闻他大概看了四百次,前几次还行,看一百次就没必要了。算法真正该做的,是给人看他觉得最有趣的新东西,而不是把一个热点反复灌到人脸上。
Grok 在 X 里扮演什么角色,他给了三层。
- 单条内容的上下文分析器。 任何一条帖子下面都能按 Grok 按钮,让它判断这条信息的可信度、背景是什么。它做不到百分之百准确,但能提供比裸帖深得多的判断依据,帮你判断真假。
- 广告验真器。 X 刚刚把 Grok 接入了所有广告。你可以问它,这个东西真的有用吗,这个产品靠谱吗。马斯克说他一直想对广告问这个问题。他补了一句,如果 Grok 说某个广告不靠谱,那个广告主大概也投不了多久。这可能反而会推动广告变得更诚实。
- 即时学习入口。 看到一个新发明、新公告,直接在站内问 Grok,可以立刻把自己教育明白。Grok 出错的地方会被记录下来,反馈进系统,让它朝着更少犯错的方向迭代。
关于 X 的定位,马斯克用了一个相对性的说法。
“X 并不完美,它肯定犯过错。但还有哪个社交媒体公司你更信任?我认为比较公平的说法是,X 是最值得信任的,或者说是最不值得怀疑的社交网络。”(马斯克)
他还给 Grok 定了两个并列目标:成为最求真的 AI,同时成为最有趣的 AI。 理由也很马斯克,人生总得有点乐子,笑是最好的药。
二十到三十万亿美元估值的前提条件
虽然马斯克多次说不能谈估值,但在自动驾驶和 Optimus 这条线上,他还是给了一个明确判断。他建议去看凯西·伍德(Cathie Wood)和方舟投资的模型,说他们过去判断对过,这次也会对。有人提出特斯拉估值可能到二十五万亿美元,马斯克回应:
“我认为这个判断大概是对的。如果我们在自动驾驶和 Optimus 上执行得好,估值确实会到二十到三十万亿美元。那是今天的二十到三十倍。数学很清楚,剩下的就是巨量的工作。但如果真的把那巨量工作做完,这就是会发生的事。”(马斯克)
这段话的重点不在数字,在执行条件。二十到三十倍的空间不是靠讲故事撑起来的,是靠两件事真正做成:自动驾驶规模化运营,Optimus 实现量产。这两件事任何一件掉链子,估值逻辑就不成立。
他为什么觉得自己活在一个最好的时代
对谈快结束时,主持人问:什么事情让你对人类、对个人都抱有希望?
马斯克的回答没有谈技术,谈的是时代位置。
“未来会非常有趣、非常激动人心。我想不想活在人类历史上任何一个别的时刻?有没有哪个时间点比现在更有意思?我认为答案是没有。至少对我来说,我们正处在人类历史上最有意思的阶段,而且每过一周都变得更有意思。所以我们应该感到非常幸运,能活着看到这些了不起的事情发生。”(马斯克)
这段话放在整场对谈的语境里是有分量的。他刚刚聊完 AI 有百分之十到二十的概率走向不好的结局,聊完背痛,聊完隔热罩有多难,聊完自己因为上市公司身份不能回答最尖锐的问题。然后在结尾,他还是说这个时代最值得活。
他说完这些,台下的车主代表上台念了一段话。大意是你把电动车从死亡边缘拉回来,让清洁能源变酷,让火箭可复用,还要带我们去火星;你花几十亿美元买下推特保住言论自由;你放下自己的公司去政府免费工作,削减浪费性开支,想把这个国家从破产边缘拉回来。然后他们问:那些人是怎么感谢你的?背叛、嘲讽、攻击、抹黑、车被破坏、动机被质疑。
马斯克听完,接过那块写着“这块砖会飞”的定制砖头,说了一句“我也爱你们”。主持人调侃他最近做手势要小心,他回了一句,我现在只能往前指或者往上指。全场笑了。
这场对谈的信息密度不低。星舰的隔热罩、Optimus 的手、Neuralink 往视觉皮层写像素、X 的端到端加密,每一个话题拆开都能单独写一篇。但把它们放在一起看,能看出一条共同的线索:马斯克在做的所有事,本质上都是在扩展人类这个物种的可能性边界。 火箭扩展物理边界,Neuralink 扩展认知边界,Optimus 扩展劳动力边界,X 扩展信息边界。他自己反复说的那句话,就是这件事的总结:从不可能,到仅仅迟到。
如果你想把这场对谈变成自己的东西
- 去一趟 Starbase。 马斯克自己说的,那地方就在主干道边上,普通人开车过去就能近距离看到火箭和厂房,不需要特殊安排。如果你对航天有兴趣,这是全世界最容易近距离观察重型火箭的地方。
- 盯住隔热罩和轨道加注这两个节点。 星舰的成败不看单次发射成不成功,看这两件事什么时候跑通。隔热罩验证在最近一两年,轨道加注紧随其后。这两个里程碑一过,火星时间表才有讨论的意义。
- 重新评估什么工作最先被机器人替代。 马斯克给的顺序是高价值医疗照护、危险作业、重复性任务。如果你在这三个领域,未来五到十年要认真想自己的位置。如果你的工作核心是判断、创造、和人建立关系,窗口期会长得多。
- 试用 Grok 的单条内容验真功能。 这是目前社交平台上少有的、能直接给出上下文和可信度判断的工具。养成在转发之前先按一下的习惯,你的信息质量会明显不一样。
- 关注 XChat 和 X Money 正式上线。 端到端加密加支付,如果真能做成,X 会从社交平台变成一个完整的生活基础设施。这对内容创作者和跨境电商从业者的影响,比对普通用户大得多。
- 理解求真作为 AI 安全策略的含义。 马斯克反复强调这一点,不是口号。你可以观察自己用的每一个 AI 产品,看它在政治敏感、事实判断、承认错误这三件事上的表现,这就是它在安全维度上的分数。
- 如果你是投资者,把二十到三十万亿美元当成一个条件句来读。 如果自动驾驶规模化加上 Optimus 量产都做成,估值才成立。跟踪这两个执行节点,比盯股价有用。
- 如果你是工程师,去看看 Optimus 的手部方案。 马斯克说手占整个机电问题的一半以上,这个判断对整个机器人行业都成立。谁能把手做好,谁就拿到了人形机器人的钥匙。
0:00Crazy. What a day. What a day. So, obviously, thank you so much to Lars and Steve for coming on here. And I believe we have Elon on, so it's just a matter of, I feel like, you know, just bringing him on the screen and stuff. Yeah. Whoa, Elon, what's up?
0:20Oh, giant on a giant screen.
0:25Yeah, it actually kind of looks like Mars in some ways. it is Perseverance.
0:31Is that a cap right there, too?
0:33I mean, that's that's the Perseverance Ro. It's a Perseverance rover on Mars.
0:37Oh my gosh, that's awesome. Yeah.
0:40Well, first off, Elon, thanks for coming to the X Takeover formerly Tesla Takeover for the second year in a row. You're welcome.
0:51Well, let's let's kick things off. Let's start start off with you know really your baby the Starship. Starship is the most ambitious rocket ever built. What's been harder than expected and what's the next milestone we should watch for?
1:08Well, I I thought everything would be hard. So, it's not like I was like it would be easy.
1:14Yeah. Yeah. I mean, Starship is a is a crazy program on so many levels. because you've got something with two and a half and future versions will be three times the thrust of the Saturn 5 moon rocket which was previously largest rocket and largest flying object ever made. So Starship is you know three times the thrust roughly twice the weight of the next largest flying object ever made and has the the goal of being fully and rapidly reusable.
1:51So this is a this is a really crazy thing to do. I think it's it's really one of the hardest engineering challenges that exists. and you know when we first started talking about Starship people thought this was impossible. In fact even within the company we we sort of thought it was impossible and it had a very high sort of what I call giggle factor. you know you mentioned the thing and people immediately start giggling like at the absurdity of it all.
2:21so now but now it's gone from as say we you know I guess I specialize in the in going from the impossible to the merely late. so that's my skill. it wasn't possible now it's just late. so and I encourage anyone who's interested to go and and visit Starbase in South Texas. It's like a magical land with gigantic rockets. I think some of you have perhaps been there and seen it. It's it's very inspiring. It's like I take part or take friends and family.
2:55It's on a major highway, so you can see things quite close up. and really get a sense for the scale of the of the the rocket and the factory and everything. So, it's it's very cool. I I guess the the thing that I thought would be hardest currently is the hardest which is the creating a fully reusable orbital heat shield which has never been done before. So normally the the heat shields are they're they're expandable and even say for the shuttle heat shield they would lose many of the the tiles on every flight and they'd have to refurbish the heat shield between each flight. So it's it's never no one has ever created a fully reusable orbital heat shield
3:38before. and no one in fact no one has created a fully reusable orbital rocket before. and Falcon 9 is the first rocket where there's at least the boost stage is reusable on a regular basis. and where it actually makes economic sense. So so so solving the heat shield problem is I think probably the single biggest remaining challenge for Starship and and of course getting the upper stage or the ship to land and also get caught by the the giant metal chopsticks which looks looks like a sci-fi movie that looks it looks like an impro improbable sci-fi movie the whole thing.
4:20Chopsticks literally catching it catching it.
4:23Yeah. Giant metal chopsticks. It's pretty wild to be catching the largest flying object ever made with metal chopsticks out of the air. so then the but I'm hopeful that the shrimp will be will be recovered maybe this year but certainly I'd say the first half of next year and then there'll be further improvements to make the ship and the booster not just reusable but fully and rapidly reusable which will actually drop the cost per flight cost per ton of payload of Starship below that of of of a Falcon 1 rocket of an expandable Falcon 1. So that that getting 100 tons or more to orbit of useful payload will cost less than a rocket that would ordinarily
5:13deliver a half a ton. That was Falcon 1. And it's because all you're you're doing is replacing the the fuel and oxygen in the rocket as opposed to building a new rocket. and most of the propellant is actually oxygen, not not fuel. But it's almost 80% oxygen, 20% fuel. The next big technology challenge after being able to achieve full reuse of the full reuse of the ship is orbital refilling where where where we actually refill propellant from from orbit like like like aerial re refueling. In this case, it's orbital refilling. I say refilling, not refueling because most of what's transferred is actually liquid oxygen as opposed to fuel.
6:04So for orbital refilling, you you need to have two starships come together and dock and transfer propellant from one starship to another and then and then in the future to an orbit orbital propellant depot. Yeah. So that's that's the situation. doesn't sound hard. It just go
6:28I mean it is very achievable. There's nothing nothing impossible is is being talked about here but it is it is very tricky to do and and and no no other organization is even trying to do this. It's like it's not like well who else is trying to do this? Well, nobody actually. in fact, even after all these years, there's still no company that's reusing a booster. So, no one's even tried to do Balcon 9 level reusability even though we've shown that that it clearly makes a ton of sense. because
7:09you know compare the cost of of any form of transport where if you have to throw away the item afterwards like like if you've got if if you had to if every time you drove somewhere you had to throw away the car you'd be wow that's that's a that's a that's a crazy car where you have to throw away after every every time you drive it. But that's how it works for for rockets.
7:32So that is really yeah I don't think anyone would be driving a car if you had to buy a new car every time you drove somewhere and then you'd have to tow another car behind you for the return journey.
7:47Yeah. So talk about the impossible Elon. What does a self-sustaining city on Mars actually look like? How many people what kind of economy? What kind of governance? Well, that that'll that'll governance will be up to the Martians. but you know, we have some sort of artist impressions of what a Mars city would look like. the and I don't really
8:14know to be honest. you know, except that we you Mars is not yet at the point where it's sort of terraformed where you could live outdoors. So, you you'd have to live initially in glass domes or something like that. And and then you could you could walk outside with with a Mars suit, but you you could you could not walk outside without a Mars suit. so maybe explain for those who haven't heard like like why why is it worth doing something like this? what what's the purpose of of
8:54Mars? Shouldn't we just focus on Earth? And I'm like, yeah, I think we should focus on Earth. Like 99% of what we do should be focused on Earth. but maybe 1% of what we do should be should be focused on on becoming a space fairing civilization and a multilanet species. because there there's the defensive argument where it's you if something were to happen to Earth and
9:24that destroyed civilization, this could be World War II or it could be a meteor like the one that destroyed the dinosaurs, then we want to make sure that the tiny candle of consciousness that exists with with humanity does not go out. So, so I think it's important to be a multilanet species to ensure the long-term survival of consciousness and all and all the life forms that that we we have here on Earth. So, and you know the the other life forms can't extend life to another planet. And and to be clear, I'm I'm an optimist. I mean, I think the most likely outcome is that the future will be good, but there's a small chance that the future that something will go wrong. And if if
10:04so, we don't want the light of consciousness to go out on a you know, we want we want light of conscious light of consciousness to continue. so that's kind of the defensive argument. And then the then there's also the sort of the the the sort of the the inspiration argument which is that life can't be just about solving one sort of miserable problem after another.
10:26they have to be things that inspire you about the future that make you that make you excited to be alive and make you, you know, you wake up in the morning, you're like, I can't wait to see what happens. To have these things like that, too. And I think having a a space program where where people are going out there and exploring the stars and sending back what they learn is inspiring to everyone on Earth, just like the Apollo program was inspiring to everyone on Earth. whether you went there or obviously very few people went there but but it was in a way we all vicariously went to the moon and it was a great achievement of humanity and it was very inspiring to people all all
11:07around the world. So being a space bearing civilization and and you know learning more about the nature of the universe and and seeing what what's happening out there even if you don't go yourself is still very very inspiring. So that yeah I think those are all things that matter a lot. Do you think that SpaceX is eventually going to go interstellar?
11:27Well, hopefully. I still hope so. Yeah. And probably not in my lifetime, but probably not lifetime, but yes, eventually.
11:42Let's talk about robo taxi. How do you see the robo taxi platform coexisting with the rest of Tesla's lineup? Will it replace the Model 3 and why? Well, this Cyber Cap, which is a two-seater, and that would that would not replace the the three and Y because three and Y have a four-seater and six-seater capability.
12:05So I think you'd find that the cyber cab would be for one or two passengers and then dynamically it would call the model three model 3 or wire for four passengers or wire for six passengers and you know then we'll have the large vehicles but I mean this this is not the forum for making our product announcements
12:34So, to be totally honest, I so I got to be, you know, little careful here about what I say.
12:43We got you. does Tesla plan on owning a large fleet of robo taxis? And then when will private citizens be able to go on the Tesla network?
12:53yes. Tesla will own some of the the fleet will be Tesla owned and some of the fleet will be customerowned. So you could think of it as model kind of like maybe some combination of Uber and Airbnb where you know some people own the cars and then add or subtract them to the fleet and and some cars are owned by Tesla directly.
13:20So if robo taxi succeeds Elon how does Tesla navigate the paradox of making the best cars in the world but also possibly needing fewer of them? It's tough for me to answer questions because Tesla is a publicly traded company and so you know this is this is the most interesting questions you'd want to ask me are the questions I can't really answer.
13:45No. Yeah, of course. No, we definitely don't want to put you on the spot there. so I think you posted a week or two ago about and even Lars mentioned it about just the Roadster program and how it's still active. What can you tell us about its current development status?
14:06Well, we're we're aiming for a demonstration at the end of this year. I mean the roadster is not something that's going to meaningfully affect say the financials of the company. because it's too small relative to other programs. it's h it is very cool. hopefully we're able to demo the year
14:37but or early next. [Applause] You said that Optimus could be more valuable than Tesla's entire automotive business. Yes.
14:52What's the road map from where we are today to widespread deployment?
14:58You know, again, you're really asking me for questions that are highly sensitive to the value of the company. And so, that's not really something I can answer. the current design the version three of optimist I think is the right design to go to volume production but it is a significant redesign from version two so in fact almost nothing stays the same so I was really faced with a choice of should we make several thousand versions to robots when we know the design can be much better or should we pause do version three which is much better and you know and we maybe have I
15:56don't know a few hundred robots instead of a few thousand robots by the end of this year and I think the right decision was to go with version three and then scale production significantly next here. now the the in terms of the value of of optimist, you really need to think of the basic question of of who who wouldn't want their own postal C3PO R2-D2.
16:30I think we don't want that. I mean, let's ask the crowd. Would you guys like if you could get your own personal Yeah. Yeah.
16:38Exactly. It very it would be very cool.
16:41and it it would actually be by the way much better than in functionality than C3PO and R2-D2. So it would be super useful. So I think basically everyone on Earth is going to want one and and then you you're going to have a bunch in industry. so that's why I think there's probably it it's probably world's biggest product because there's at least a market for probably 20 billion maybe it could be 50 billion I don't know line.
17:21So you know hypothetically if Tesla was making a billion of these a year at scale and at scale the cost gets lower and lower so maybe it would be like on the order of $30,000. I'm just guessing here and you know that's that's 30 trillion in revenue. You know, there's a long long way to go between here and making a billion robots a year.
17:53I do think something like that may happen. Yeah.
17:58What would you say is Optimus going to have the biggest impact on? Do you think it's going to start obviously in manufacturing like or at what not not necessarily timeline but like where do you see Optimus having the biggest impact? Well, initially it would go to very high value things. So like say somebody
18:21needs a lot of medical care or something. So, like if if you know often could be like a 24-hour a day nurse or helper to someone that is quadroplegic or something like that you know or somebody just really is needs physical help. So that's the kind of thing which would be lifechanging for a lot of people.
18:53those will be the the highest value initial use cases. It will also be used in situations where it's the work is very dangerous. so or where there's a lot of repetitive tasks. but like any work that's that's that's risk life or some kind of injury would be a good initial use case for Optimus.
19:31When you look ahead in 10 years, what's more transformative? Full self-driving or Optimus? Optimus.
19:40Yeah, I see you're drinking some good old Diet Coke, by the way. Shout out to those. you've hinted that we may reach digital super intelligence this year or next. How will we know that we've crossed that line? And what does one of AG what does day one of AGI really look like?
19:59Well, I'm not sure. Sure. It's like a a massive it may be less profound than the at least it will be less profound like AI has not done yet is invent new technologies that are useful. So it hasn't it has discovered new physics and it has not invented technologies that are useful but I think that is something that that will happen.
20:37If Grock becomes smarter than any human, what role do we play? Are we co-pilots, overseers, or just legacy code?
20:47Well, it won't just be there will be many AI. There will be at least, I think, four major AIs, maybe five, just in the United States. So, even if was not developed, there would still be digital super intelligence. And I I've been fighting personally sort of driving hard on digital super intelligence for a long time because I wasn't sure if this is a double-edged sword or a single-edged sword or what.
21:18But it became very obvious after a while that this was going to happen whether I participated or not. So therefore I had a choice of either be a spectator or a participant but that it was going to happen with or without me. So then I think well okay rather be a participant than a spectator. And I can focus on AI safety which in my
21:50view and I've thought about this for a long time is that the most important thing for AI safety is to be maximally truth seeeking. which means often like saying things that are maybe politically incorrect but actually factually correct. and also there'll be mistakes that are made then admitting the mistakes and taking correct taking corrective action and GRO still has a long way to go to
22:24actually truth seeking but I think this goal is also consistent with inventing new technologies you know solving medical problems you know you know if you If you want to cure cancer, you have to be maximally truth seeking to understand what the root cause is. For example, so so I think that's actually very fundamental both for safety and for usefulness. the the economy of the future I think is going to look sort of quite different from where does today and AI and robotics in the good scenario which we're working towards you know trying to ensure makes that we do have the good scenario there will be no shortage of products and services for anyone like basically anyone one
23:22will be able to have any products as they want. The productivity gains from AI and robotics are just astounding to think about. Like I think I think it could increase the the size of the economy by a factor of 10 or more. Like it it will eliminate poverty. The bigger challenge may be finding meaning in life. So if if the robot can do anything
23:50that you can do but maybe better how do you find meaning in life? That's that may be the biggest challenge
23:57when you think of the fact you know you mentioned that you know you've been holding off on you know really getting involved whether it's you know participating versus being a spectator and when you're thinking of the next couple of years of like let's say not necessarily catching up but just doubling down and really focusing on it.
24:15What are going to be your biggest focus points?
24:18Making it useful and making it safe for humanity and making it love humanity essentially and be I've never seen any technology advance as fast as AI. So, and I've seen a lot of technologies advance, and maybe this is a discomforting metaphor, but AI is a a supersonic tsunami. It does not sound funny.
24:51It sounds sounds a little scary in that way, but it's in terms of maybe in terms of exceeding prior technologies, it's like supersonic. and and you really do need to pair AI with robotics. Yeah, I I think it's it's crazy just how much it's advanced in just the past year or two years and and seeing how kids are using it, looking using Gro 4, Grock voice, the companions now. I think you've answered this before, but what would what keeps you up at night when you think of the future of AI? And now that you're doubling down and really just going wartime mode right now on just tackling this these issues, what is what is something that keeps you up at night?
25:39Well, a AI safety and making sure the AI is aligned with humanity and wants to foster and grow and better humanity and you know basically we want to be like if you could you could look into a crystal ball right now and see the future needs to be the future one the goal previously was
26:04accelerating the advancement of sustainable energy which is like a very good goal and then the reason I pursued that rather than AI because I could have pursued AI from the beginning was because I was confident that that sustainable energy was a single-edge sword only like it was only a good edge you know so we sure that okay we succeed in that like okay that's that's definitely good sustainable energy by definition is great you got
26:35cleaner air and you know, an can run out of whatever the source fuel is with with combination solar batteries and electric cars, then that's a that's an unequivocally good for AI. Like I said, there's it is more of a double-edged sword risk. but but I think it most likely will be good. It most likely will bring immense prosperity and you know, it's going to figure out how to cure every disease.
27:09and it's it's probably going to be awesome, but I think we need to be careful and not complacent and and we actually be somewhat paranoid to make sure that we have the good AI future and not the bad AI future. so I mean from a Tesla stock standpoint I think with autonom autonomous cars and and especially with with Optimus the
27:46it does go to to sort of crazy levels. I'd always recommend like looking at Kathy Wood and Arinvest's analysis. they're they've been right in the past. and I think they're right this time. And again with a lot of difficult execution, I I I think you know people out there said, "Well, Tesla's valuation could be 25 trillion." I and I think that that's probably correct. If we execute well on autonomous transport and Optimus, it actually does get to a 20 to30 trillion valuation.
28:32so that's 20 to 30 times what it is today.
28:38Everyone's ch Yeah, the math is math is clear. so just it's just a massive amount of work and but but like I think if you do that massive amount of work that's what will happen. Elon let's talk about Neurolink. You've described Neurolink in the past as a way to solve the input output bottleneck between humans and and machines, right?
29:17What's the long-term vision? Is Neuralink a medical company or is it the first step towards symbiosis with AI?
29:23It's it starts off as a medical technology. So, Neurolink's initial goals are really to help people who have say quadriplegic tetroplegic situ they've lost the use of their body and to be able to enable them to control their bone computer just by thinking and I think we have eight eight patients so far and all eight love
29:52the device and and are using it every So it's it's helped a lot of people and I think we're test team the neural team is aiming to do about 20 patients by the end of this year. So it's accelerating. And then next year we'll be doing the blind sight implant which will enable people who are completely blind like if they've lost both eyes and optic nerve and still be able to see facing directly.
30:26[Applause] So the focus essentially is fixing serious medical issues and as prove out the safety and efficacy and then eventually at some point you could get cybernetic enhancements. So it could it could be something that massively augments intelligence and allows you to communicate at
31:02ultra high speeds. so it it's very sci-fi. and and it could even get to the point where you could upload your memories. and and essentially have a a saved version of yourself. and then I'm I'm speculating here, but but but then maybe download that into a new on a closed body either robot.
31:35I'm really getting into sci-fi here, but but like download that into a robot body if you want or a clone version of your original self. You know, they certainly made sci-fi movies and written books about this. So, these are not new ideas, but I I do think stuff like that will be possible.
31:58which would give you I guess a form of or anybody wants it a form of immortality and to be clear I do think these things will be available to anyone who wants to do it so it's not going to be sort of limited to a few sort of you know elite members of society I think it will be available to everyone who wants it
32:24it it's crazy just to think you know with Neuralink Inc. Grock, the Grock companion, and even the humanoid robot. How do you do you see I mean it just feels like more and more whether it's Grock, the Gro Companion, and then Neurolink. I know we we did an interview with Brad who was the third patient who has ALS and he was also utilizing Grock. What do you what do you see continuing to see that integration between let's say Neurolink and even even Grock as things continue to grow?
32:59Yeah, growth will actually be able to understand the signals from your so the neural links taking to your brain and then it's sending the signals to Grog and then Grog actually understand those signals at a at kind of like a a binary level as opposed to having to translate it into words and so greatly improve the efficiency with which you can use your Neurolink device.
33:31like it's so I mean we're really getting into some interesting philosophical questions or interesting questions like if you think about how much of your your mentality your mind is used to take a complex thought or image in your mind and translate that into words and it's and it's very lost.
33:59because how can you really with words convey say an image that you have in your mind? it's very difficult to do so and and if you're not like if you're not an artist it's actually very difficult to actually get the image from your mind onto a computer or paper. but but the but the computer could do that. You could say okay this is the image that you want to produce and
34:25let's say like there's a complex series of concepts that you want to convey to someone else and and if both people have a neural link you'll be able to convey a complex series of ideas uncompressed in their full form to someone else's mind as opposed to reducing complex comp concepts down to a few
34:55sentences and trying to get someone else to hear what those sentences are, decompress those sentences in their mind and try and try to understand the concepts that are in your mind. So this is why I call it conceptual telepathy. So this would greatly improve communication and understanding between humans. You've mentioned restoring vision and mobility but that idea of writing to the
35:24and hearing too sorry. So it's like basically if if there's if there's any sense or brain injury in principle that sensory function or that brain injury can be solved with a neural link. So that includes hearing sight smell. I don't people not many people have lost their sense of smell but feeling it could be like if somebody has a stroke and they've lost the ability to move left arm, a neural link could restore that ability.
36:08it it's it's really a very powerful general purpose input output device that can address you know over time any brain or neurological issue. So if somebody's having seizures it could actually stop the seizures. it could I think probably fix schizophrenia like like a lot of there's a lot of things that could like in principle it could fix basically anything that's to do with the brain or neurons or your senses.
36:51Is there anything that would be like is there a a most difficult problem to solve in that space then? because technically a lot of it could be solved. Is there something that would maybe you would see bigger blockers on trying to solve as far as fixing those problems?
37:09I'm not aware blockers right now. It's just that when you're putting devices in humans, you have to be very careful and to make sure that no damage is done and that is that the implant works and without any negative side effects. And so far you know knock on wood we're betting a thousand here and and and you know patients have the devices have worked and they've noticed no meaningful side effect it's pretty cool but it's just in order to be safe that does slow us down.
37:53and of course we we interact very closely with the with the FDA and for for approvals and stuff. So it's it's not just neural link in a vacuum here. It's it's with regulatory approval from the FDA. So and and then you do need different variants of neural link device for solving different things. So for example for for the teleopathy product which is what we have in the eight patients so far that interfaces with the motor cortex in the brain. So it's like like literally it's someone will will think about gribbing their hand neurolink device will read that signal in the motor cortex and then move the mouse on your on their computer.
38:47and then for sight, it's a different thing because you've got to stimulate the visual cortex. and the visual cortex is in a different location and it's and it's a little deeper. So, you've got so you've got to put the electrodes a little bit deeper to reach the visual cortex. And in this case, it's it's it's not reading, it's writing. It's it's effectively writing pixels to the visual cortex.
39:18This is something we've had working in monkeys for about three years now. in fact, one of our monkeys has both a telepathy implant and a blind sight implant and he's a very happy monkey. I do want to emphasize we take great care of our our animals. this is a really big deal because I've always said like imagine if we were in their position.
39:50Well, let's just do all the things that if that if we were in their position, we'd do. and the the USA inspector that came by to check out our facilities, she said in her entire career career, she's never seen a nicer monkey enclosure ever. And, and let me tell you, monkeys love for the lefty implant.
40:18monkeys are just like us. They they love playing video games and eating snacks. So, if if you look if you look at that video of Paige, I think he's like four years old now. You see, he's just he's just sitting on a branch sipping a banana smoothie. which he gets every time he scores.
40:42He gets a sip of a smoothie. and but he's not being held down like he likes playing the game. So it's really just like humans likes like snacks and video games and yeah same thing.
40:59One question I have like building a humanoid robot and neural link intersect if at all. I mean what are the learnings from that? is there any overlap?
41:11Well I have to say building a humanoid robot has really made me think a lot about how the human body works. you gain a new appreciation for actually just how awesome at design the human body is. Now admittedly, you know, I think some parts of the human body could be better like the spine. Like why do we have so much back pain? It's very annoying.
41:40Maybe don't fight sumo wrestlers, man.
41:42Yeah, seriously. Doug and Greg are but you know yeah typically you know everyone has back pain in some point in their life which is also a thing neural link can solve so that would be cool too. but no neur optimist has really made me think a lot about how our hands work how we balance how we do different things. but maybe more than anything just how incredible our hands are.
42:17They're it's like wow. these things that our hands do like the hands and optimist are half or more than half of the electromechanical problem and the entire rest of the body is is a half.
42:36Wow. That's crazy. Yeah. And it's like honestly you look at your hands, you're like and you move and if you and if you feel your forearm like the almost all the muscles that control your hand are actually in your forearm and and and they're pulling your fingers like like puppets.
42:59So yeah, there's only a a small number of muscles that are in that in your hand itself, but your hand is all being controlled through tendons that go either through or above your pop tunnel. and your hand is being operated like a marionette from your forearm.
43:17It's crazy how much the the human body, whether it's like, you know, Tesla vision going, you know, just trying to see what kind of like similar to how humans view the the world, right? and getting rid of the radar. And then same thing with the humanoid robot and just, you know, thinking and seeing how the human body works. is there anything
43:40more specifically even within the human body that will continue to infer whether it's the humanoid robot? And it sounds like obviously the the the hand is the big one of the biggest problems, but are there other parts of the body too that would
43:55continue to help infer other parts of the the humanoid robot or again it's just yeah with Tesla vision and and eyesight.
44:03Well, with with Optimus, we're able to reuse the autopilot computer. So, Optimus is intelligence is powered by the AI AI4, the same computer that controls the car and you know a small battery pack that's similar to the car battery pack. So, there's a lot of batteries with the car.
44:30and and then the AI that's in the car is similar to the AI that will be in Optimus. running on the AI4 hardware and then AI5 hardware when that's ready. So there a lot of parallels but the limiting factor is the hand. Yeah. But I'm convinced that Optimus will be the biggest product ever.
45:02yeah. So, and then of course autonomy is starting to roll out currently in Austin but will be in many cities in throughout the United States later this year and hopefully many countries next year.
45:27Yeah, here in the Bay Area we have a bunch of FOMO. Everyone's looking at right now. I can tell you I I always whenever I open that robo taxi app, man, it just the geoence just the the figurine is insane. Funny. Yeah. Yeah.
45:46I mean, you got to have some fun. Yeah. Don't take yourself too seriously type of thing. [Applause]
45:58Let's switch and talk about X. You've called X and here we are at X takeover. You've called X the everything app but it's still in a transitional phase. What exactly do you want to be in five years?
46:11I mean X is definitely improving. It's I mean it's evolved a lot from Twitter can only have like you know short text tweets and like two minute videos to the point where now you can have like 4hour videos and you can write a novel if you want on system. So it's it's gone sort of fully multimodal from long videos to long text to anything in between. and we're
46:44currently really improving the the DM system. So with a new system called Xhat which enables audio video calling it's all fully encrypted peer-to-peer style like Bitcoin. and the asset test is that even if somebody puts the gun to my head, I still can't read your messages. That's the test. so it's like very secure and
47:14and and also like you do calling and and and video with it. and then there's the kind of X money or X Finance release which is hope hopefully only a few months away. We actually have it operating in beta within the company. and that that'll be a major factor. so but but ex as much as possible at ensuring freedom like freedom of speech within the as much as we can within the bounds of the law and
47:56you know also like aspirationally trying to reach for the truth of things like you see that in community notes where community notes will will correct someone even if they're very powerful communic community notes will correct me it'll correct presidents CEOs even major advertisers and even if that cost us money so I think it's generally there's a lot of good things that have happened we're trying to improve the algorithm which we know kind of sucks right now sorry about that they have the overarching goal of the algorithm is to try to show people information they would find most interesting.
48:46but right now it'll show you too much of one thing, you know. So if whatever the sort of interesting thing of the day is right now it'll the X algorithm will show you too much of that and it's like okay I don't need to see it 20 times you know you know I think I think I've seen like the Sydney Sweeney jeans thing about 400 times this point and it's like okay the first few times was good but we shouldn't show things like a hundred times because because you want to hear new things.
49:23so but X is making steady progress being offering you know, pretty much everything you'd want to do and yeah, I feel good about where it's headed. Payments, we're we've all I know we're waiting on that and just you know, thinking about even PayPal too. It's crazy just what Fate Loves irony doing X.com back in what the late 90s and yeah
49:54then Full Force buying it you know years back and now now you're doing payments again.
50:01Yeah, it's it's a sort of poetic and or or re returning to something that was sort of unfinished with PayPal and completing the product plan. What role do you think Grock and XAI play in the future of X? Is it just a feature a core intelligence layer or does that change how people experience the platform?
50:27Yeah, Grock I think it's pretty helpful because that that that Grock button which you can press about any given post on the system and its analysis of the post is usually quite accurate. It's it it doesn't bat a thousand, but it's really gives you deeper insight into any given piece of information that somebody posts on the system. you can figure out what context helps you figure out if it's true or not. and we actually just
50:58added Grock for all advertising in the system. So you can you can press the rock button and see, hey, is this ad is this product going to work or is it is this product legit or not? Which I kind of always wanted to have on ads, you know, it's like, does this thing really work? that's what I want to ask a lot of ads. And now on the X system, you can ask, does this thing really work?
51:26I guess advertisers where Grock says it doesn't really work. I'm probably not going to advertise for very long. maybe it'll encourage better honesty in advertising, which I think would be a good thing. [Applause] But we're just generally trying to make X a system that you can trust more than any other system.
51:50And I think I think it is there already. It does not say that X is perfect. It is definitely mistakes. But is there any other social media company you trust more than X? Nope. So like I'm not trusting Facebook, you know. so or not to pick on Facebook, whatever, but you know, I I think it's actually probably fair to say that X is the most
52:25trustworthy or least untrustworthy of social networks. Yeah, I think that became very apparent just when you saw like when you just compare of like the news during COVID and things being censored, you know, just you come to X and you can find out what's really happening
52:43and it's real time. It's really like the I would say the you know the world's discourse in town square, right? It can sometimes be a a fist fight it feels like, but at the same time it's just, you know, where you can find out what's really going on,
52:57right? That and that's the goal. And I I do want to emphasize obviously there's a long way to go. It's it's far from perfect, but I I do think it's getting better over time. And you know sometimes it's two steps forward, one step back, but the trend over time I think is very good for X being the place where you can figure out what's really going on in the world. dig into the truth and you can ask Grock more. you can also learn a lot of things. So you can ask if there's a a new thing that's invented or announced, you can ask Grock on the system more about it and instantly educate yourself. And and then we we monitor when Grock gets things wrong and then we
53:45we feed that we feed that back into the system to make it aspirationally less wrong over time. Yeah, it's really just crazy to think you know when you talk to Grock and you can ask it for just things in your timeline. Hey, make a post sound like me on this topic or you want
54:10like we are actually aiming for Grock to be the actually I should say two gold are for Grock to be the most truth seeeking AI but also the funniest AI because you got to have some fun in life and you know laughter is the best medicine. You gota you you got to have some fun. Elon, you've predicted a world of abundant energy, robotic labor, and AGI.
54:42with AI and Optimus, doing, you know, potentially a a lot of things and reducing the amount of things that we're doing today, what do you think is left for humans and and, really what does that mean for in the world of artificial super intelligence? Yeah, that's a question that I struggle with. Actually, this is part of the reason for neural link, which is to
55:11improve the bandwidth communication with AI. So to help achieve a better human AI symbiosis so that you know AI knows what we want and it can respond to collective human world. and I think things like can actually effectively dramatically increase our our intelligence. but I think the future is very much one where human intelligence, machine intelligence and robots are interwoven.
55:52I think I like I said I think it's 80% maybe 90% likely to be awesome but we got to be cautious about that you know 10% maybe not awesome situation like Terminator
56:05yeah I mean seriously that that you know stuff like that that would be bad.
56:13What's your vision for a society where human effort is no longer required for survival or productivity? How do we avoid irrelevance?
56:23Right? I think that's the central question. Well, if if we are symbiotic with AI where we that there's this essentially a merger of human and machine intelligence, then we may be able to address the relevance question. you know, like like let's say you ride a bicycle or a motorcycle or a car,
56:50you're you're super human in what you're able to do. You can now travel faster than any human with legs. I mean, if you're in a car, you can go vastly faster than the fastest human ever. does that make running irrelevant? I mean, people still run and still race against each other.
57:15So, but but you got humans working with machines. I mean, they actually had these kind of debates when the steam engine came along and and the trouble. it's like well because people derived a lot of their value from saying they they can do manual work faster than a machine and then the machine got faster. But this is a philosophical question that I struggle with and the best thing I can come up with is that you have interwoven human and machine intelligence and got something like neural link to open human intelligence and improve the symbiosis with AI that that we effectively become maybe one with the AI. Elon, you off
58:11Sorry, the audience was about to clap and I started talking. but Elon, you you know, you're you're solving I guess in a lot of ways, as sometimes you say, you know, a lot of problems. and in a lot of ways, you're creating an a very exciting future with reusable rockets, the mission to Mars with Starship, with the humanoid robots and how it's going to just the impact that it'll have and Neuralink and seeing the impact on Noland and Brad. What
58:43What gives you hope for today not just for humanity and civilization, but personally? Well, I think the future's going to be very interesting and exciting. Would I want to be at any other point in human history? Is there some other point in human history that that is more interesting than where we are today? And I think the answer is no. At least for me, I think we're at the most interesting part of human history. And it's getting more interesting with each passing world. So that so okay well if I'm in if I'm in the part of history that that is the most interesting then I think you know anyone feels that way and I think a lot of people feel that way that
59:33we should feel very lucky to be alive at this at this time to see these amazing things happen. [Applause] Well, Elon, we wanted to do a time check. but
59:50yeah, I I bet I I better get back to my eight jobs.
59:54Yeah, exactly. You're not a busy man at all. but we just Kelvin and I, we just want to on behalf of Tesla owners of Silicon Valley, Tesla owners of Austria, Sanwin Valley, and really just, all of the 1500 people that traveled from all over the world. We just want to personally just say thank you for all you're doing. and thank you for continuing to push humanity forward. Well, thank you.
1:00:24Yeah, you have a whole bunch of people here just cheering. So, thank you for all you're doing. Thank you.
1:00:31Well, and and thank you for all your support over the years. it's super appreciated. you know, I would say my heart goes out to you, but I have to watch what I gestures these days. Ah, we know what you need.
1:00:48I can only point straight ahead or straight up.
1:00:55But anyway, my heart does go out to you and thank you all for your support over the years.
1:01:01Yeah. And we just had one thing that we wanted to present to you. Okay.
1:01:11How's it going, Elon? So on behalf of TSO TOSV and the presenting sponsors haha yes we we wanted to present something to you that was symbolic meaningful and really you know these last few months have obviously been quite tumultuous for you and so we really got creative and and really yeah we wanted to create you something special. And so I'm going to read you something. Basically, you know, Elon, you you brought electric cars back from the dead. You made clean energy cool. You made rockets reusable. You're going to take us to Mars.
1:01:57You spent billions to buy Twitter and save free speech. You stopped running your empire to work for the US government for free, to reduce wasteful spending, to save the country from bankruptcy. Yeah,
1:02:11man. It took a few hours for that one.
1:02:17And and how and how were you thanked? It's like you were betrayed. You were you were mocked, attacked by the very people who once cheered you on.
1:02:28Your name dragged through the mud, your cars vandalized, your motives questioned. But that's not the full story, as I imagine many here would attest. So there are millions of people who see you for who you are.
1:02:44People whose lives you've touched and changed. People who believe in what you're building. People who love you and have your back. Thank you.
1:02:51So yeah, thank you. So, we were gonna we were going to introduce to like a few more than a few, but we had a reel that we were going to play, but with some technicalities, we're unable to do it of actual testimonials from people really meaningful, impactful ones that people really wanted you to hear. But, we'll post it on X.
1:03:18We'll post it on X. there was only a few but there's a lot more people who want to express their gratitude and support. Oh, thank you.
1:03:26But you know fortunately they could write letters and so Elon thousands of people wrote letters to you to to to demonstrate you know and express how much you have meant to them. We love you. I love you guys too.
1:03:46Thank you. So, we compressed all those paper letters into this brick. Jug. Yep. This is a custommade brick. the brick itself is a message to you. It's heavy brick. Not in the sense of how heavy it is, but it's a brick of belief which weighs more than any material matter. We believe you can achieve the impossible.
1:04:11Hence the quote on the front of this brick. You want to read it, John? This brick does fly in spirit because you made you have made the impossible happen. All right.
1:04:27Yeah. And so Elon, yeah, you make bricks fly and in turn that inspires us to to pursue the impossible. And you know what we all say to that? Everyone repeat after me. Haha. Yes.
1:04:44Yes. Ah, yes. Thank you, Elon. We love you.
1:04:49Thank you. Love you, too.
1:04:50Thank you, Elon, for your time. We appreciate it. And hopefully you have a good rest of your Saturday afternoon. All right. You guys, it was a pleasure speaking. All right. [Applause]
资源与出处
- 视频在 YouTube 收看
- 档案silicon-valley.tech 档案页本页视频与文稿数据来自 Elon Musk Archive
- 精读本篇精读 Markdown 版