马斯克在 CES 现场对着一屋子大公司的首席营销官说了一句话,让整个会场安静了几秒:人类积累的全部知识,去年就已经被 AI 拿去训练用完了。
不是预测,不是警告,是已经发生的事。他把这件事讲得很轻描淡写,好像在说昨天晚饭吃了什么。
这场对话由 Stagwell 董事长马克·佩恩(Mark Penn)主持,台下坐着大约 25 位来自大公司的 CMO。佩恩开场先讲了个自己的故事:他开了四十分钟的自动驾驶,从迈阿密到劳德代尔堡,全程没碰方向盘。他本来是怀疑派,坐完之后信了。
马斯克回了一句:"你只要跟人讲这件事,没体验过的人不会信你。"
整场对话信息量很密。AI 的时间表、自动驾驶的临界点、人形机器人的产能、火星飞船的发射窗口、Neuralink 的患者数量、DOGE 的削减目标、政府对加州山火的荒唐规定、Twitter Files 里那些不合法的事,还有他对 X 平台的真实构想。他几乎没绕弯子,能把话说多直就说多直。
AI 已经无处可学,开始自己教自己
马斯克讲的第一个关键事实是数据枯竭。
你把整个互联网、有史以来所有的书、所有有意思的视频都拿来,拆成一个个 token,这些信息加起来,人类知识的总和,在 AI 训练里已经用尽了。基本上就是去年发生的事。(马斯克)
他还特意说明,这里不算那些重复的猫视频。一千个一模一样的猫视频没有意义,要的是所有有意思的视频。
数据用完了怎么办?转向合成数据。AI 自己写一篇文章,自己提出一个论点,然后自己给自己打分,走一遍自学习的过程。
这条路有个硬伤,马斯克自己也点出来了:你怎么知道 AI 是自己编了一个答案,还是给出了正确答案? 找"真值"(ground truth)变得困难。合成数据能撑多久,能撑到什么质量,是个没有定论的问题。
他说这话的时候,语气没有焦虑,更像是在陈述一个工程约束。AI 的能力边界在往前推,但训练素材这件事本身先撞上了墙。
三到四年内,不碰原子的活 AI 全包了
佩恩问了一个所有人都关心的问题:AI 到底会给普通人的生活带来什么?
马斯克的回答直接:
未来几年内,AI 能完成任何认知类任务。任何不涉及原子的工作,最多三到四年,AI 就能做。(马斯克)
"不涉及原子"这几个字值得停下来想一下。写代码、写报告、做分析、看片子、写法律文书、做客服,这些都只是在处理比特,不碰实物。这部分工作,他给的时间上限是三到四年。
他举了几个已经发生的例子。AI 参加医学考试,能比大约 80% 的医生考得好。AI 看放射影像,比大多数人一辈子都在看片子的人还准。这些话他讲得很平静,好像在报一串已经跑通的数据。
现场有人先做了个调查。问 AI 今天有没有改变你的生活,13% 的人说改变了。问五年后会不会改变,87% 的人说会。从 13% 到 87%,这条曲线本身说明了大家的心态:知道它要来,但还没真正感受到。
马斯克补了一句自嘲式的话:AI 把认知活都干了,那"我们这些人到底要干什么",就成了一个绕不过去的问题。
自动驾驶的临界点就在今年第二季度
特斯拉的自动驾驶,马斯克给了一个具体到季度的判断。
大概今年第二季度,我们就有信心让事故概率低于一个有经验的普通司机。之后会继续往下走,最终做到比人类安全十倍,然后一百倍,到那个程度基本就不会撞车了。(马斯克)
这里有个容易被忽略的重点,他把特斯拉的路线和 Waymo 的路线做了明确对比。Waymo 已经在几个美国城市跑无人车,车里没人,但它被限制在少数城市、少数区域。特斯拉走的是更难、但更通用的路。
他解释了两者的差别:特斯拉只用纯视觉加 AI,不靠昂贵的激光雷达,不靠雷达,甚至不需要提前知道这个地方长什么样。你把车开到一个从没有特斯拉去过的地方,它照样能开。
哪怕是在一个外星行星上,车也照样能开。(马斯克)
这句话听起来像玩笑,其实说的是泛化能力。限定区域的自动驾驶靠的是高精地图和预先建模,通用方案靠的是模型真的"看懂"了路。前者好落地,后者上限高得多。
佩恩追问,政府认证的自动驾驶一年内能不能到位。马斯克没有给死时间,只强调技术上今年就能超过人类平均水平。至于监管,那是另一条线。
Optimus 的产能,按罗马军团来数
整场对话里,人形机器人这块马斯克讲得最兴奋。
我认为人形机器人会是有史以来最大的产品,远远超过其他任何产品。(马斯克)
他的逻辑很朴素:每个人大概率都想要一个自己的 C-3PO 或者 R2-D2。有人想要两个,那就有工业需求去生产、去提供服务。他给了一个比例预测:人形机器人和人类的比例,至少 3 比 1,可能 4 比 1,甚至 5 比 1。
按这个比例算,就是 200 亿到 300 亿台人形机器人。这个数字大到什么程度?马斯克自己都说,到那个阶段,"钱"这个东西还意味着什么,都说不清楚了,经济增长的上限也不存在了。
他顺口抛出一个概念:不是全民基本收入(Universal Basic Income),是全民高收入(Universal High Income)。 前提是 AI 走的是好的那条路。
产能这块给了具体数字。Optimus 目前是他口中的"世界上最先进的人形机器人",一只手有 22 个自由度,看起来、摸起来都像人手。今年的目标是先造几千台,在特斯拉自己的工厂里测试。如果顺利:
- 今年:几千台,先在自家工厂跑
- 明年:5 万到 10 万台
- 后年:再翻十倍,约 50 万台
佩恩开玩笑说,不如用罗马军团来算。一个罗马军团是 5000 人。马斯克笑着接了。这个比喻其实挺贴切,机器人产能爬坡的速度,已经不是按台数在数了。
火星:两年内先送一艘不载人的船
火星的时间表终于有了明确数字。
两年内,我们能送第一艘无人飞船去火星。地球和火星每两年同步一次,现在正好在同步点上。(马斯克)
他没把话说满。第一次飞行的核心目标只有一个:让星舰在火星上安全着陆,不增加火星表面的陨石坑数量。 这句话是他的原话风格,听着像段子,实际是硬指标。着陆成功,下一次才可能送人上去。
再往后,他希望是成千上万艘星舰一起去火星。
我脑子里有个画面,像《太空堡垒卡拉狄加》那样,殖民船一起出发,在太空里是一大片亮点。我觉得那画面会很酷。(马斯克)
但真正的目标不是着陆,是让火星自给自足。他给了定义:如果从地球出发的补给船,因为任何原因停了,火星还能继续活下去、继续发展,才算成功。 地球这边可能是第三次世界大战,可能是自然灾害,也可能是别的什么。只要补给断了,火星不灭,人类文明的平均存活年限就大幅拉长。
他抛了一个判断文明的标准:这个文明是还困在自己母星上,还是已经成了多行星文明。
我们不想当那种只有一个星球的蹩脚文明。就算走不出太阳系,至少得先到另一个星球上。(马斯克)
Neuralink:三个患者之后,是 20 到 30 个
Neuralink 这块,马斯克给了最新的患者数据。
现在有三位植入 Neuralink 的人,都工作正常。升级版的设备电极更多,带宽更高,电池续航也更长。今年希望做到 20 到 30 位患者。(马斯克)
第一代产品解决的是"大脑和身体的连接断了"的问题。四肢瘫痪、截瘫的人,靠着读取大脑运动皮层的信号,只要想着"动一下手",屏幕上的光标就会跟着动。用想法控制电脑和手机。
他用霍金做了类比。如果霍金能像正常人一样快、甚至更快地交流,那对他来说意味着什么,不用多说。
下一个产品叫 Blindsight。即使一个人双眼都没了,视神经断了,甚至从出生就看不见,也能直接和大脑的视觉皮层对接,让他重新"看见"。这项技术已经在猴子身上跑通了,那只猴子到现在应该已经用了两年。
再往后一步更狠。如果在脊髓损伤的位置后面,再装第二个 Neuralink 设备,理论上可以把大脑信号绕过断掉的"线路",传给身体下半部分。让瘫痪的人重新走路。 马斯克说,这在物理上是可行的,他有信心。
长期目标则是另一件事:提升人类和 AI 之间的带宽。
我们说话的时候,带宽是很低的。一个普通人 24 小时持续输出的信息量,不到每秒 1 比特。一天 86400 秒,大部分人一天输出的信息,还不到 86400 比特。(马斯克)
这个对比很扎心。人类接收信息的速度不慢,输出却慢得像蜗牛。Neuralink 的长期野心,是把输出带宽提高一千倍、甚至一百万倍。到那个时候,人的体验会和现在完全不同。他用了"超能力"这个词。
佩恩当场就说:这一切都给我来一份,我是早期采用者。马斯克答得很快:"相信我,你会喜欢这个芯片的,我保证。"
DOGE:目标是两万亿,落在手里可能是一万亿
话题转到政府效率部。佩恩先回忆了一段往事:九十年代他跟着克林顿搞"重塑政府",两年就把预算平衡了。马斯克接话:"那是个好年代。"佩恩自嘲:"不过没持续多久,很快就被推翻了。"
DOGE 能不能省下两万亿?马斯克这次的回答比之前谨慎得多。
我们会冲两万亿去,那是最好的结果。但得留点余量。如果我们冲两万亿,能拿下一万亿,就算打中了。(马斯克)
他补充了另一层逻辑。如果能把预算赤字从两万亿压到一万亿,同时给经济松绑、让增速上来,让商品和服务的产出跟上货币增发的速度,就不会有通胀。他管这个叫"史诗级的结果"。
至于钱从哪儿省,他用了一个画面感很强的比喻:
政府里到处是可以省钱的地方。就像走进一个满是靶子的房间,你闭上眼睛随便一枪,都打不空。(马斯克)
他分析了两层原因。第一层,联邦政府的支票永远不会跳票,它有一台无限印钞机。第二层,花钱的人花的不是自己的钱,让别人对自己的钱上心,本来就很难。
他还讲了一个更荒诞的机制。他在政府里认识一些人,很讲原则,真的想把钱花在刀刃上,但他们做不到,制度不让他们做。到了每个财年末尾,他们会被要求把预算花光,哪怕花在完全没意义的东西上。因为今年不花完,明年的预算就会被砍。
不浪费钱反而会被惩罚。这套激励机制完全是疯的。(马斯克)
佩恩接着补了一刀。他算过一笔账,如果把布鲁克林大桥、林肯隧道的造价按通胀折算,那一轮基础设施法案的钱,够修 4000 条林肯隧道。当然修不出来。政府现在的效率,和当年完全不是一回事。
马斯克同意,但把原因指向了另一个地方:法律的积累。
大量法律和法规累积下来,让几乎所有大型项目在事实上变成"非法"。就算你想做,花在纸面材料上的钱,也比花在工程本身的多。然后就是拖延。(马斯克)
所以 DOGE 还有一块重要工作,很多人没注意到:审查法规,把那些"害处大于好处"的清掉。他承认有些监管是必要的,但相当一部分完全说不通,根本不服务于公共利益。
加州山火、胡瓜鱼和一把不让用的耙子
马斯克举了加州山火这个例子,讲得最具体,也最能说明监管失灵的代价。
要控制山火,办法其实很朴素:留出防火隔离带,把房子周围的灌木清理干净,把水库蓄满。这几件事全是常识。
但加州做不了。
因为一堆环保裁决,他们不允许清理灌木,不允许把灌木从房子边上推开。理由是可能会伤到某种红腿蛙之类的东西。往往是那种你从来没听说过的生物,挡住了这件事。(马斯克)
他点了另一个名字:胡瓜鱼(smelt)。为了让这种喜欢半咸水的小鱼活得舒服,加州让大量淡水直接排进海里,而不是留在水库里。他补了一句:没有任何实际证据表明,水库多留一点淡水,胡瓜鱼就会不高兴。
逻辑摆在这儿:留淡水、蓄水库、留隔离带、清灌木,都是低成本高收益的事。真正的小概率物种影响,没人拿数据证明过。结果洛杉矶付了巨大的代价。
他讲这段的时候语气不激烈,只是把一串因果摆出来。但听完之后,你会明白他为什么说政府监管需要系统性清理,而不只是砍预算。
扎克伯格转向、Twitter Files 和 FBI 的后门
佩恩提到扎克伯格最近的"180 度大转弯",以及他承认政府确实在审查内容这件事。马斯克没有绕:
政府百分之百在审查内容,这一点从 Twitter Files 里是确定的事实。有些操作,说实话,已经相当不合法了。(马斯克)
他讲了一个具体机制。FBI 有一个直接连到推特的后门,可以监控任何内容,可以审查任何内容,还有"两星期自动删除"的设置。
所以我们甚至不知道他们到底做了什么。但他们拥有想做什么就做什么的巨大权力。这听起来不像合法的事。(马斯克)
除了政府行为,他还点了一层自我审查。很多人不是被删帖,是自己不敢说。审查的形态不止一种。
他把言论自由和民主绑在一起讲:
没有言论自由,就没有正常运转的民主。你不知道到底发生了什么,就没法做出知情的投票,那你就没有真正的民主。(马斯克)
他回到制宪者的动机。第一修正案被放在第一位,不是随手写的。那批人来自言论被严控的地方,说错话的代价是罚款、坐牢甚至丧命。他们不想要那样的美国。这个位置是有理由的。
他也顺势谈到了公民记者。现在越来越多的事,是普通人先拍下来、先发出来,传统媒体再跟上。他给了一个历史对照:林肯被刺杀的消息,传到亚洲花了大概三周。更早的时候,你的国家参战了,你的村子可能要等一个月才知道。
佩恩补了一个细节:美国当年知道珍珠港被袭,靠的是破译了日本的密码系统,日本那边还不知道自己已经被读穿了。信息传播的速度和渠道,一直是权力的一部分。
X 想成为什么:人类的集体意识
有人问马斯克,你觉得现在的互联网是不是很糟,要怎么修。
他先歪了一下楼,说如果你问的是网络连接,星链可以帮你。他认真讲了星链的意义:很多地方的人有东西想卖,但没有网络就卖不出去;麻省理工把所有课程免费放上网,但你得先有网才能上。网络连接在把一部分人从贫困里拉出来,这件事分量很重。
然后他回到真正的问题:是不是负面内容太多了。
有些时候确实负面内容太多。我提议过调整 X 的算法,让它稍微正面一点,结果一堆人反过来骂我。我心里想,你们到底想要什么。(马斯克)
现场有人问,怎么让"悲观"变得不再时髦。马斯克答得很实在:这个国家以前是很乐观的,现在你问大家走在对的路上还是错的路上,几乎没人会答"对的"。
但他自己给了一个明确的态度:
我对未来四五年其实挺乐观的。我们有可能迎来一个黄金时代。前提是,把压住一切的那座法规大山搬掉。(马斯克)
最后谈到 X 的定位,他把话说得很清楚:
我想把 X 做成一个对世界有益的东西。我把它看成人类的集体意识。既然是集体意识,它就必然包含人性的一切,好的和坏的都会在里面。这很正常。(马斯克)
目标是什么?健康、快乐、清醒的集体意识,而不是相反的那一面。而且要做成最可靠的信息源。想知道世界上发生了什么,大事小事,任何时候、任何地点,X 都能给你最接近真相的答案。
他还说了一句容易漏掉的话:所有关于未来的书里,几乎没人预测到社交媒体会出现。 没有一本书写过社交媒体会怎么发展、会怎么冲击社会、会怎么驱动政治运动。所有人都漏掉了这一块。《星际迷航》也好,《杰森一家》也好,那些对未来的想象里,没有社交媒体的位置。
所以 X 会长成什么样子,其实没有剧本可抄。
最后那句话很马斯克
佩恩问有没有什么想留给在场的这些人。
马斯克说:
我建议大家对未来保持乐观。它变得更大概率的走向是好的,不是坏的。这就是我的预测。(马斯克)
整场听下来,最值得记住的不是哪一个具体时间点。马斯克给的时间表,历史上兑现过的没几个,他自己也知道,所以这次很多地方都加了"如果能""假设顺利"。
更值得记住的是几条判断的走向。认知类工作被 AI 全面接手的窗口是三到四年;自动驾驶的安全拐点就在今年;人形机器人的产能按十倍往上翻;火星的第一艘船两年后出发;Neuralink 从三个患者走向几十个;政府的浪费不是效率问题,是激励机制反着来。
这些判断串起来,指向的是同一个方向:技术这边的曲线在加速,制度那边的曲线在拖后腿。 他自己也是这么看的,不然不会花那么多时间讲胡瓜鱼和防火隔离带。
对看这场对话的人来说,可以落到几件具体的事上:
- 把"AI 会做什么"从抽象讨论,换成"我手上哪些任务只是处理比特、不碰原子",这部分三到四年内会被重构,早点想清楚自己在哪一边。
- 不要用两三年前的体验判断自动驾驶。找一次真正开放道路的长途试驾,比看一百篇评测有用。佩恩就是这么被说服的。
- 关注人形机器人的产能数字,而不是发布会上的演示。几千台、十万台、五十万台,这条曲线比任何 demo 都能说明问题。
- 如果你所在的行业受监管影响大,把"哪些规定已经说不通"列出来。马斯克讲的加州山火那个例子说明,害处大于好处的规则会真实地消耗掉整个行业的效率。
- 对社交媒体的信息质量保持警觉,同时别指望任何一个平台给你完整真相。集体意识包含人性的一切,也包括噪音。
- 对自己保持乐观这件事,可以认真一点。马斯克最后那句不是客套,是他整个判断体系的结果:技术曲线在往上,剩下的变量是制度和人心。
至于他说"你会喜欢这个芯片的,我保证",这句话可以先放一放。但他那句话背后的东西值得记住:人类输出的带宽,一天不到 86400 比特。这个数字很可能是整场对话里最容易被忽略、但最扎人的一个。
0:02chairman of stagwell here with Lindy ocarino and an assembled group of about 25 CMOS of major corporations here at CES during the CES to really see the latest in technology and we're graced with Elon Musk who has consented to give us some of his time to answer some of the these really burning questions about technology how it's going to develop and change our lives lives and how some of his interactions with with government may also change our lives as well yeah and so if I if I might start start off with kind of a a big overall question and I'll do some questions and then we'll get some questions in from the group you know I have to say that
0:49I've been actually a Tesla owner for many years oh great thank you clearly you're clearly someone of Great Taste and well I I I will say that I I I drove from Miami to Fort Lord Dale to meet my brother for dinner about 40 miles and I did not touch I touched the wheel but let me tell you that he
1:13touch the car drove itself is what you're saying and it you did not need to intervene it it did not need me for this entire ride yeah it's pretty magical it's like like when you tell people that they've not experienced it they they don't don't believe you you know yeah yes and I didn't believe me either because I've been a skeptic about how far it would go sure and your latest releases of software are are you know really incredible so than people to to try them as a big question and and in that vein coming in the next decade what do you think is going to be the the greatest advances in technology that will affect people's lives what what should they be expecting to see here
2:01from technology in their lives okay well I don't want to blow your minds but AI is going to be big feel I feel confident in that prediction but but you know the funny thing is if you go back even five years ago certainly 10 years ago if I 10 years ago even 15 years ago I was I was saying AI is gonna be this massive thing that has deep super intelligence smarter than the smartest human people thought I was kidding and they thought that that's ridiculous there's no way a computer's going to be smarter than a human and be able to do all these complicated things and and now you know the it's it's it's the the latest AIS are able to pass complicated
2:50tests better than most humans like they can pass the medical tests better than than like 80 80% of doctors or something you know they can AI can diagnose radiography better than U most people whove been doing it their whole life so that that's that's just accelerating obviously I I think probably if you haven't seen jenson's talk it's it's excellent and it really shows how much AI is advancing and it's advancing on the the hardware front on the software front in terms of data this like the the new sort of thing is synthetic data because we've actually run out of all the books and you literally run out of you take the entire internet and all books ever written and all
3:42interesting videos like like you don't need a thousand cat videos that are exactly the same but all the interesting videos and and you sort of distill that down into tokens essentially bits of information and you've we we've now exhausted all of the basically that the cumulative sum of human knowledge has been exhausted in AI training that happened basically last last year and so the only way to then supplement that is with synthetic data where the AI creates it'll sort of write an essay or it will come up with with a thesis and then and then and then it will grade itself and and and and sort of go through this process of self-learning with synthetic
4:24data which is which is always challenging because how do you know it how do you know if it UC hallucinated the answer or it's a real answer so it's challenging to find the ground truth but but it but it is pretty wild that AI at this point has run out of all human knowledge to train on is crazy I I know that you're building for for grock the largest AI Center on the planet and we we already have yeah it's an operation Microsoft is planning 80 billion dollar where where I used to work for a while but that's a lot of money by anyone standards really you know mean you know I did a poll and we asked do AI making a difference in your life today 133% said yes and they said
5:12in five years from now will AI make a difference in your life 87% expect in five years it will make a difference what gigantic difference okay what is it going to do for people is it going to work right will do anything you want and even suggest things you never even thought of so I mean AI really within the next few years will be able to do any cognitive task like it obviously begs the question what what are we all going to do you know but but pretty much any cognitive task that doesn't involve atoms AI will be able to do within I'd say Max three or four years maximum
5:55and and then then the the another element of it is the robotics is that you need so you can't just be thinking AI can't just be thinking in it's in a in a data center it's got to do do things so that's where you need the robots so and you need you know self-driving cars which obviously you've experienced and and and that that rate of improvement is is exponential in in how good the self-driving cars are you know we feel confident in passing the you basically being better than human driving in about three months basically Q2 of this year we feel confident of passing having a probability of of accident that is better than the the average experienced
6:45driver and then and then it'll keep going from there ultimately I think it's going to be 10 times safer than a human driver and then a 100 times safer like it's to the point where really it just won't crash so and that's so so that's happening this year with with Tesla so Tesla and and and this is a software update to you know our cars so as as you've experienced yourself it's the same car it got a software update and suddenly it's way smarter at driving well let me let me try a few timelines then because you know look I'm not the youngest guy around so I want to be as much technology stand is Young get older every year as I get older I used to
7:33build computers and kits right when when you didn't couldn't buy them yet and I don't have to do that anymore so self-driving some timeline self-driving cars certified government certified self-driving you think will be within a year well I mean there there already are autonomous you know in in in some regions like weo has autonomous vehicles with no one in it but they're limited to like like few cities in the US the Tesla solution which is a much more difficult path to go but ultimately much more powerful is is a a general solution to self-driving so the the Tesla software is just purely Ai and vision doesn't rely on any expensive sensors no light offs no Radars or or
8:22it doesn't require it doesn't even require knowing the area beforehand like you could put you could have a drive someplace has never been before and no Tesla has ever been before it could even be an alien planet I mean it and the car will still work still drive so and and so that's that's that's this year you know and when can I get a home robot
8:44okay right well so that's that's the other element is our humanoid humanoid robots so I think probably most people if not everyone would like to have their own personal C3PO R2D2 and and and and I I actually think humanoid robust will be the biggest product ever in history by far I agree yeah it's just it's just wild but because you can just say well every human is GNA want one most likely and some will want to and and then there'll be all of industry in terms of making providing products and services so that you have say what's the ratio of humanoid robots to humans my guess is it's least at least 3 to one four to one maybe five one so we're talking about 20
9:3130 billion humanoid robots and you know the it's not even clear what money means at that point or if there's any meaningful cap on the economy I think at that point assuming that things haven't gone arai you know in the in the good AI scenario I think we will have we won't have Universal basic income we'll have Universal High income so do you think five years for my first robot or well for for Tesla we're you
10:07know our Optimus robot really is it's it's unless somebody's got something secret we don't know about the our Optimus robot is the most sophisticated humanid robot in the world it's got a hand that has 22 degrees of freedom it looks and feels like a human hand and you know we're aiming to have several thousand of those bu this year here initially we'll we'll we'll test them out in at Tesla factories but
10:35then assuming things go well we we'll we'll 10x that that output next year so we'll aim to do maybe 50 to 100,000 humanoid robots next year and then 10x at a gain the following year it's like 500,000 robots in three years that's a lot that is a lot yeah guys like maybe we should think of this in terms of Roman Legions how many Legions of robots show what we have like a Roman legion is 5,000 when will we have a colony on Mars well I I think we we'll be able to send the first un crude spacecraft to Ms
11:18in two years so Earth and Ms synchronize every two years and so we're we're at a synchronous point right now so then the next one will be roughly two years from now and there'll be two years from then there'll be another one so for the first trip obviously we want to make sure that we we can land Starship without crashing like we we we need to prove that we can land Starship on Mars without incrementing the crater count and and if if those land safely then maybe on the next trip we would send people and then hopefully that would grow exponentially So eventually there'll be thousands of Starships going to Mars and and I I might have this like
12:01really cool visual like battle Galactica or something you know like Colony ships departing all together with these like bright points of light in space I think it look really cool but so but I think the goal has to be to get to the point where Mars is self-sustaining so the point at which Mars is self- sustaining which is really defined as point at which if the resupply shifts from Earth stop coming for any reason that Mars doesn't die out that Ms can continue to grow so if there's something that happens on earth like that say there's a World War III or some natural disaster or who knows what but for whatever reason this the resupply ships sto coming if if MOS
12:41can still continue to survive then you know the then the probable lifespan of civilization is dramatically greater so you know if you sort of stand back and say how would you evaluate any civilization You' say like well is that Civilization still stuck on its home planet or are they a multiplet civilization and we don't want to be one of those lame One Planet civilizations okay we're have a respectable outcome here even if we don't make it beyond our solar system we're at least got to get to another planet you know well and and finally in my list brain technology communication yeah direct communication am I am I gonna see that also because right now
13:23this is looking pretty good for me yeah so we've got to neuralink we've got now three patients with three three humans with neural links implanted and they all working well and we've upgraded devices that that where the devices will have more more electrodes basically higher higher bandwidth longer battery life and everything and so we we expect to you
13:49know hopefully do I know 20 or 30 patients next year or this year I should say with the upgraded neuralink devices and and this our first product is we're we're trying to enable people who have lost their brain Body Connection so they're are tetraplegic or paraplegic or it basically like can imagine like say Stephen Hawking if Stephen Hawking could communicate as fast or even faster than a normal human that would be transformational yeah so that that's sort of our first product is is
14:27being able to read the the the the motor cortex of the brain and say that if you think about moving your hand it will move the cursor on the screen and it enables enables people to control their computer or their phone just by thinking and then our next part will be blindsight that even if somebody has lost both eyes or has lost the optic nerve or even if they've never they've been blind from both we can interface directly with the visual cortex in the brain and enable them to see and we already have that working in monkey actually have a monkey who's now had that for I think two years and so that enabling sort of basically enabling people to control devices and
15:14ultimately we think if you have a second neuralink device that is past the point where the the spinal damage occurred we can actually transmit the signals from the brain past the where essentially the wires are broken and abled someone to walk a game so that that would really be profound obviously but I'm I'm confident that that is physically possible and but and then the longterm goal for neur link is to be able to improve the bandwidth so so right
15:47right now when when we're speaking our bandwidth in bits per second is quite low and the sustained bandwidth of a human is less than one bit per second over a 24-hour period so there 86,4 seconds in a day and the average human put outputs much less than 86,400 bits in a day if if if somebody's a writer they might they'll exceed that but most people do not output more than the number of seconds in the day and but with with a neural link you you could increase that output capability by a thousand or maybe a million so it would be a profoundly different experience like you'd be superum essentially put put me down for all of this so far an early adapter yeah and and trust
16:35me you really you really like the chip I can guarantee it and let me kind of bring us down to earth for for a question on on Doge yeah I I worked very closely actually with President Clinton in the 90s where we did have Reinventing government we did balance the budget in two years actually that's awesome oh those were the days well it didn't last very long because got blown up very quickly have you identified Some Cuts that you're really that you're really looking at that you think will be successful do you think the two trillion is is a realistic number now that you're looking more closely at it yeah
17:19well I think I think we we will we'll try for two trillion I I think that's like the best case outcome but I do think that you kind of have to have some overage I think if we try for two trillion we we've got a good shot at getting one and and if if we can get drop the budget deficit from two trillion to one trillion and and kind of free up the economy to you know have additional growth such that the output of goods and services keeps Pace with the increase in the money supply then there will be no inflation right so that's that I think would be an epic outcome and in ter in
18:04terms of saving money in the government well as you you as you know it's it's a it's a very target-rich environment for saving money like if you if you look in any direction it's like people like where will you find places to save money I'm like it's like being in a room full of targets like you could close your eyes and and you can't miss so there's just a lot of waste in government because especially the federal government you you just got a situation where the the checks never bounce like they've got like the infinite money computer and and and then the the people that spend the money are not the people it's it's not their money you know it's it's it's very
18:41very hard for for people to care about spending someone else's money and and then even if you you know I certainly I know people in the government who do care about on just as a matter of principle spending money effectively and they try to do so and they can't the system prevents them from doing so and and they even get told to do crazy things as as again Probably sounds familiar where you get towards the end of the budget cycle and and they're they're told to spend up to their budget and and even on on nonsense stuff because if they don't spend their budget the budget gets reduced so it's actually sort of sort of a perverse incentive to waste
19:20money and and and then they they kind of get punished for not wasting money so it's just totally bananas well well I agree you will find you know I I I did a mathematical analysis in terms of how government used to do things so if you take the Brooklyn Bridge or the Lincoln Tunnel and you inflation the in infrastructure bill should actually get you 4,000 adjusted bre Lincoln tunnels which of course of course it's not because government is not as efficient as it used to be no exactly essentially that we've got had an accumulation of laws and regulations that make basically any large project essentially illegal and and and even if you've
20:07tried to do it you've got satisfi you've got to spend way more money on the paperwork than on the thing itself so and and then it gets delayed and so there's there's there's an element of like Doge which is very important which is looking at regulations and and getting rid of ones that where the the the the H is worse
20:28than the good like you say like well any given regulation is like okay there's some amount of good some amount of harm but you know what's that ratio is it like you know and there's a lot of regulations where frankly just they're completely nonsensical and we want to get rid of nonsensical regulations that do not Ser the public good well Linda Yaro in her keyot here mentioned the Doge thing and she got enormous Applause so I think the country is really waiting to to see this effort they're they're behind it they're optimistic let me try to get in one more topic here before I get one or two other questions out there which is obviously Mark Zuckerberg made
21:07an amazing 180 degree turn yeah you know that cool what's your reaction to what he did and and his acknowledgement frankly that the government was in fact censoring things or he was censoring things or the government or some combination thereof yeah I mean there there there's no question I mean 1,00% government was was censoring things we know that for a fact from from the Twitter files I mean some of the stuff was like like pretty illegal frankly I mean the FBI had this portal into Twitter where they could spy on anything and and censor anything and had a two- we Auto delete so we don't even know what they did except that they had immense power to do whatever they wanted which
21:53doesn't sound legal and you know that sounds pretty pretty crazy and you know there was also a lot of self-censoring and I don't know there was just a lot of censoring going on and I think you I feel very strongly that the you have to have freedom of speech to have a functioning democracy you know if you don't have freedom of speech freedom of expression then how do you know what's really going on and and if you can't make an informed vote then you don't have you don't have a real democracy so it's you know just it's incredibly important to I think listen to the wisdom of the founders of the country and say why why did they make that the First Amendment
22:38you know and and it's they did it for a reason because they came from places where that that where there was massive censorship and the penalties for speaking your mind would be you know findes imprisonment or death and they're like we really don't like that we want that to not be the case in America so then they you know freedom of speech so you just posted you are the media are citizen citizen journalism I mean really becoming more and more important in in this AG of media and you see it with the with the wildfires here yeah exactly it really takes the citizens to to report and tell people yeah actually you can think of it like before there was before the
23:24internet you kind of had to have the what I call Legacy media of which which is you have to have some aggregation points where you know you'd have reporters going going and find things out then they would come they would go to to to their office they would write up U articles they would then print those articles on paper that paper would then be distributed and it was kind of the only it was the only way to know what was going on but it was very slow especially in the old days you know I think when like when when Linn Lincoln was the assassinated I think it took like 3 weeks for the that news to reach Asia or something like that you know and in fact in the old days like
24:07you wouldn't even know that your country had gone to war because it would take like a month for the fact hey we're at War to reach your village yes well we actually learned about Carl Harbor because we decoded the Japanese they didn't know that we decoded their their their cyber system right at the time and that's how we actually learned about it yeah and and you know we have just if I can squeeze in I just got three minutes here it's fine you got a little longer if you want okay thank you I I
24:40so just I got three questions in from the audience that Dan Gardner and Toby Daniels have on discour said want to know do you think the internet sucks and what do you think we need to do to fix it you mean the content on the Internet or or your internet connection because starlink can help you on the internet connection starlink is great for internet connectivity so you know especially for like faces that that have bad connectivity in fact I think star link is is is really having a significant effect in terms of lifting people out of poverty in fact in in many parts of the world where people have a product that they want to sell but if they don't have an internet
25:19connection they they can't do it and or or if they want to learn things like basically you can learn anything on the internet for free like MIT has all its lect that available for free on the internet but you need an internet internet connection and so once you have an internet connection you you've got access to education you've got access to a global market so I think it's it's very significant how connectivity makes difference in people's lives know with with respect to maybe the question was geared at like is there too much negativity on the internet the I think at times there is there is too much negativity and you know actually on on the X for I
25:58proposed like well we're going to tweak the algorithm to be a bit more positive and then people got upset about me for that I'm like I'm like okay like what do you guys want you know well actually the second question from Zach Moffett CE of Target victory was how do you make pessimism uncool again so I think it ties to to maybe exactly what you're saying which is that you know it be it's become people people are afraid you know this used to be a can dotion yeah always very positive and now now the country when I ask if we on the right track or the wrong track they never say that we're on the right track anymore okay there is well I hope hopefully you know hopefully with
26:40this you know I'm I'm actually pretty optimistic about the next four or five years I think we're actually GNA I think we have the potential for a golden age so we need to it's very important to get rid of of of the the mountain of of of regulations that are holding things back and I don't mean there there's some good regulations but there's just so much that we just can't get anything done I mean you take sort of the California wildfires for example you we really need to have fire brakes and we need to clear the brush back away from houses and and we need to make sure the reservoirs are full these are all kind of obvious things but but due to a
27:21bunch of environmental rulings you can't actually do that in California so they're not allowed to do the five rakes and they're not allowed to push the the brush back away from houses because it might it might hurt some like red-legged frog or something like that you know there's like some sort of creature that usually a creature you've never heard of that is preventing this from occurring you know this this there's like this fish called the smelt for example and so there's so we have far more fresh water run off into the ocean than than we should really and on the theory that it helps this one little fish that likes slightly Briny freshwater salt water mix and if if we
28:02if we if we keep more fresh water then the Smelt fish will will not be happy U but but there's no actual evidence that that the Smelt fish is going to be unhappy if we we keep a bit more fresh water in fact so we should we should keep more fresh water keep the reservoirs full and and just have some sensible fire brakes and move the brush away from houses that's just an example of like would have saved a lot of trouble you a lot of tragedy in La if we've done that so you know and yeah and I think I
28:40think Ai and Robotics is going to lead to a higher standard of living for for people beyond what they can imagine so I think we have ai doctors and medicine that are pretty incredible so that's the F the final question she says from from Katherine heritage heritage if I got that right if if all the robots and everything freeze up time for humans what is it they will do with that time or what can we ask them to do with that time and I think that's a yeah question rounds out the circle technology yeah I guess it would be a bit like being retired I mean there's now this will take a few years but at some point if if this as AI gets as Ai and Robotics get better
29:34eventually AI will be able to do everything that humans can do so any task you do will be optional like it'll be like a hobby or you know so now it does be a question like will this will our life lives have meaning if if the computers and the robots can do everything better than we can and that is that is a real question I I I do wonder about that myself and that's maybe that's why we need the neural link so we can enhance human capability so we can keep up with the yeah well I I just want to thank
30:12you for well this obviously incredibly optimistic view about technology and where it's going I I feel reassured that the the kind of leaps that that are going to be made particularly what you're working on are just incredible the next few years not not decades away but very CL really talked in terms of years consistently and very thankful for that and and I think that that's you know absolutely tremendous message here at the you know where we are in Las Vegas kind of studying what what technology is going to be available to people and and really again thank you for the tremendous role in opening up free speech and for the role that that X and you know Linda is PL in terms
30:58of that I know that we're we're working to get full recognition by everyone of the of a the platform as it should be because free speech is is I think the proper the proper way to go yeah I think it unbalance is good and you know one of the things I try to create conceptually with the xplatform is like it it's like a global Consciousness it's like Collective conscious consciousness of of humanity now if you have a collective consciousness of humanity well you're going to get every aspect of humanity good and bad that's just naturally what happens so but but I I you know hope I do want it to be a good and productive thing and you know have this the aspiration is to
31:41to maximize the Jets you know it's either in Star Trek or the Jetson more or less right and but the one thing that was never predicted in any book that I could find is social media so it's the one thing that they never really there's no book built around how social media would develop how it would really impact Society how would move political movements and so forth and it's interesting that everybody missed that in their in their and so I guess the the last closest closest question maybe is is what do you want X to be and where is that platform going to go well I mean I do want XP Force for good that
32:48exactly so I do want to x a Force for good and and I do view it as as sort of like the the group mind of humanity and we want to have a sort of a healthy happy and sane group mind versus the opposite and I mean or yeah so and I I wanted to be just like the best source of Truth like if you're trying to understand what's going on in the world that it has the most most accurate the most upto-date information about anything large or small so so it it gives you the
33:33best understanding of of what's going on in the world you know anywhere anytime yeah terrific thank you any any closing thought you want to leave us with otherwise I want to thank you for being so generous with your time you're welcome I I think I would encourage people to be optimistic about the future I think it is much more likely to be good than bad so that's my prediction
34:01[Applause] all right thanks guys
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