2026 年 6 月,一桩针对 OpenAI 的诉讼被陪审团驳回。驳回的理由跟马斯克说得对不对没关系,是诉讼时效过了。
马斯克在福布斯的这次访谈里把话讲得很直白:陪审团只认定了一件事,时间太久了。至于有没有不当得利,至于那个聊天机器人算不算被偷走,他们一个字都没说。而在他看来,后者才是显而易见的部分。
但他也没把话说满。他承认这案子本身有点含糊,原因跟事实清不清楚无关,问题出在这件事是一点一点发生的。
它不是一次性被偷走的,它是一块一块被偷走的。所以我们得回头看,在那个节点之前,究竟有没有起诉的基础。我觉得其实没有。(马斯克)
一个只判了时效的判决
他接着把"一块一块"的过程拆开讲。OpenAI 最初设立那个营利性实体的时候,加了几道限制:营利部分有利润上限,规模也很小,而且还写了一条,一旦实现通用人工智能(AGI),所有股票都回归慈善机构。
前面两道限制后来松了,最后那道最关键的,也在最近几周被删掉了。
这意味着什么?一个从非营利组织长出来的实体,在实现通用人工智能之后,可以继续作为营利公司存在。它现在的估值是八千亿美元。
马斯克的判断很重:这是危险先例。如果这套路径可以走通,等于告诉所有人,你可以用非营利的身份募钱,做成了再转成营利公司,那美国所有慈善捐赠的根基都会被削弱。
所以他明确表示必须上诉。这跟赌气无关,他看的是判例效应。他自己把后果说得很清楚:
如果不 appeal,这个案子就会成为判例,人们可以拿它当依据,实质上把慈善机构洗劫一空。我认为这是错的。(马斯克)
访谈者顺口接了一句"对啊,反正你也没什么可输的",马斯克回了个"Exactly"。这段对话很能说明他的态度:他不指望这次判决能翻盘,他要的是别让它变成别人以后照着抄的模板。
这里有个容易被忽略的点。很多人看这则新闻,第一反应是"马斯克告输了"。严格讲,他输的是一场程序性的仗,实体问题被绕过去了。而在法律上,程序性驳回往往比实体败诉更难翻,因为它不给你辩论事实的机会。这也是他非上诉不可的原因。
他给自己列的三件事
访谈中途主持人把话题拉回到最初:你当年到底想做什么。
马斯克的回答很朴素。大学时候他想参与那些能真正改变世界的事。他一开始并不打算做 AI,因为他觉得 AI 是把双刃剑。他真正想做的是加速向可持续能源经济的转型,具体就是电动车、太阳能和电池。
第三件事是脑机接口。他用的词很夸张,但举的例子都很具体。
帮助脑部或脊柱受伤的人恢复功能。让多年说不出话的人重新开口,这件事已经做到了。让失去双眼或者视神经受损的人恢复视力,甚至让天生失明的人第一次看见东西,办法是直接跟大脑的视觉中枢对接。还有让人重新走路。
这些差不多是耶稣级别的奇迹了。当技术开始触及耶稣级别的奇迹,那还挺不错的。(马斯克)
"耶稣级别"这个说法听起来像吹牛,但把它拆开看,其实是三件临床上已经有进展的事:语音解码、视觉皮层刺激、运动功能重建。他夸张的只是形容词,事实部分相当扎实。
第四件事是 SpaceX。目标是把生命变成多行星的,把意识扩展到星际。
一百万吨,是多行星文明的及格线
这里他给了一个具体的时间点和一条具体的门槛。
时间点是今年晚些时候,他希望实现第一枚完全可重复使用的轨道火箭。这件事的意义落在运输能力上,省钱只是顺带。没有完全复用,你就没法把足够的吨位送上月球或火星。
门槛是大约一百万吨。
为什么是这个数字?因为要建起足够的工业基础,让那边的文明能够自己生长。他给了一个很狠的检验标准:如果地球的补给船因为任何原因停止出发,月球或火星上的文明还能继续增长,还是就此死掉。
能继续长,才算数。这条标准把"建个基地"和"建个文明"彻底分开了。一个靠地球输血的前哨站,哪怕住着一万人,也只是前哨站。
他还特意强调了一句,怕被误读:
成为多行星文明,不意味着我们离开地球去别的地方。那只是把单行星文明搬到了另一个地方而已。(马斯克)
而且那个"另一个地方"要难住得多。他的原话是,跟月球、火星或者几乎任何其他地方比起来,地球都极其容易、极其舒适。这句话放在一个以火星为毕生目标的人嘴里,分量不太一样。
那日常到底在做什么?他说自己并不会每天早上爬起来问"今天我该创新点什么"。
就是造那些把生命延伸到地球之外所必需的技术。(马斯克)
他顺手列了几个正在推进的东西。星链(Starlink)在太空里重建整个互联网,而且更好。特斯拉在做的 Optimus 人形机器人。自动驾驶汽车。大规模太阳能。
主持人接着问了一个很自然的对比问题,想让他在几家公司的成果之间排个高低。他挡回去了,理由是特斯拉是上市公司,另一家也即将上市,这种时候不能评论,属于静默期那类限制。
再追问"最喜欢哪个",他用了"你没法说最喜欢哪个孩子"来回避。这个回答听起来像客套,但放在他这里其实是真的:几家公司共用同一套底层逻辑,能源、制造、运输、算力,彼此咬合。硬拆开排名反而失真。
不过聊到历史上的发明家,他倒是很放松。公司名字来自尼古拉·特斯拉,所以他当然是特斯拉的粉丝。他也喜欢爱迪生,知道有些人不喜欢爱迪生,但他认为爱迪生做过相当了不起的事,通用电气就是爱迪生的公司。科学家那边他点了本杰明·富兰克林、莎士比亚、牛顿、爱因斯坦,"反正就是那些常客"。
真正让他多说几句的是亨利·福特。
福特实际上比人们意识到的更厉害。他基本上发明了复杂物体的规模化制造。在福特之前,人们其实不太知道怎么干这件事。(马斯克)
他讲了一段很多人不太熟的历史:福特最早搞出来的那家公司,后来成了通用汽车的一部分,然后他差不多是被踢了出去,之后才创办福特汽车公司,真正把复杂产品的量产方法做了出来。这事非常难,然后所有人基本都在抄他。
这段话放在整场访谈里很关键。马斯克谈火星、谈太空算力,听起来都像宏大叙事,但他真正佩服的能力是"把复杂的东西大规模造出来"。福特干的是这个,特斯拉的产线干的是这个,SpaceX 把火箭做成流水线产品干的也是这个。他的野心一直挂在制造能力上,概念层面反而说得少。
被问到当代有谁让他佩服,他点了黄仁勋,理由是英伟达在 AI 计算机上的推进。
睡觉时一个突破,午饭时又一个
主持人问他,面对 AI 的进展,你到底是乐观还是悲观。
说实话我同时是两种。到这个阶段,我睡着的时候有一个 AI 突破,醒来又有一个,到午饭的时候还有第三个。已经让人头晕了。(马斯克)
接下来是他给出的时间判断,这部分值得逐条记下来。
第一,会出现远比人类聪明的 AI,这没什么悬念。而且有些地方它已经比人聪明了。
第二,全面超越,连创新能力一起算进去,他估计大概不超过一两年。
第三,五年之后,也就是 2031 年前后,数字智能的总量会超过全人类智能的总和。
第四,那时候至少会有一亿个人形机器人,也可能到十亿。
第五,经济规模大概会在五到七年内翻一倍。他给的理由是你会撞上倍增周期,经济产出增长得太快,前后差几年,变化都是巨大的。
关于"希望它对我们要好"这句,他是在玩笑的口气里说的,但前面那串数字并不玩笑。一个把"一两年内全面超越人类"和"五年内超过全人类智能总和"同时说出口的人,其实是在描述一条非常陡的曲线。
注意这两个判断的量级差别。单个模型比人聪明,是能力问题。智能总量超过全人类,是规模问题。后者一旦成立,经济、科研、军事的底层假设都得重写。他把翻倍周期放在五到七年,逻辑上是一致的:如果智能总量在五年内超过全人类,经济增长不可能还按老节奏走。
太阳占太阳系 99.8% 的质量
访谈里信息密度最高的一段,是他解释为什么 AI 的终局在太空。
先摆现状。SpaceX 现在有一万颗卫星在轨。未来用星舰,每年会发射超过一万颗通信卫星,每一颗的能力都远超现在的卫星。所以从太空提供的通信能力,大概是现在的十到一百倍。
然后他立刻泼了盆冷水:这个数字跟 AI 卫星要送的吨位比,不值一提。
他的思路来自物理学的极限思维法。与其问"我们能做到多少",不如问"理论上最多能做到多少"。如果问题是"人类能调动的最大算力是多少",那答案就在太空,因为那里有远超地球的能源。
接下来这段是全场最好的科普:
地球上的我们,跟太阳比起来,就像一粒灰尘上的微生物。全部人类文明目前使用的能量,不到太阳输出功率的一万亿分之一。真的远不到一万亿分之一。你把人类全部能源用最慷慨的方式去算,也远不到太阳的一万亿分之一。(马斯克)
他还给了一层换算:就算你把文明掌握的能量提高一百万倍,你掌握的仍然不到太阳能量的一百万分之一。
然后是那个更狠的论证。太阳占太阳系 99.8% 的质量。假设你把除太阳之外的所有质量都烧掉,从中提取理论上最大的能量,太阳产生的能量四舍五入仍然是 100%。再假设你从别处传送两个木星进来,也一起烧掉,太阳仍然是 100%。
所以结论很干脆:跟太阳比,其他一切都可以忽略。想在任何有意义的意义上沿卡尔达肖夫指数往上爬,你必须去太空。
这段话的作用是把能源讨论的坐标系换掉。平时争论核电、风电、光伏,都是在人类尺度里比大小。他把尺子换成了太阳,那些争论立刻变成小数点后很多位的事。
从每年一太瓦,到每年一拍瓦
光有坐标系不够,他还给了工程账。
粗略估算:每年一太瓦的太阳能 AI 卫星,从地球发射。怎么算出来的?每年一千万吨的入轨载荷,每吨 100 千瓦的功率,乘起来就是每年一太瓦。
一太瓦是什么概念?美国平均用电功率是半太瓦。也就是说,每年发射上天的 AI 算力,是美国全年用电量的两倍。
他强调这没有违反任何物理定律,全都可实现。这句话的分量要放在他的语境里读:他做的是把"不可能"逐条拆成"要多少吨、多少千瓦、多少次发射"。这些数字一旦落地,剩下的就是工程和成本问题。
想再往上走几个数量级,办法是在月球建基地,造一台质量驱动器(mass driver,一种电磁弹射装置),在月球上生产太阳能板和散热器。芯片可以从地球运,也可以在月球造。这套组合能支撑每年一拍瓦的太阳能 AI。
一太瓦到一太瓦的千倍,就是拍瓦。他没说人类最终需要多少 AI,只说了句"但那是非常多"。
后面那句判断才是重点:那样规模的经济,会远远超过地球经济,地球经济在它面前会变得无足轻重。
这句话在访谈里一带而过,但它是整段的落点。人类历史上所有关于"中心"的假设,都建立在经济活动发生在地球上这个前提。一旦主要的算力和能源生产移到地球之外,那个前提就松了。
所有人都以为隧道是个玩笑
话题转到无聊公司(The Boring Company),马斯克的语气明显轻松了。
他说自己很长一段时间都在讲,人们应该挖隧道。结果所有人都以为他在开玩笑。后来他索性想,行吧,那就成立一家公司来挖。
他给的理由很像一道几何题:建筑是三维的,交通是二维的,所有人还都想在同一时间进出那些三维建筑,指望不堵车本身就不合逻辑。把交通也变成三维的,隧道铺开去,理论上能应对任意程度的车流。
他还在访谈里公开鼓励别人也开隧道公司。据他所知,无聊公司是唯一一家认真在做这件事的。他还补了一句实在话:挖隧道很辛苦,机会确实很多,所以希望别人也来做。
这个态度挺有意思。大多数创业者不会主动招呼竞争者对着一块硬骨头一起上。他招呼了,说明他判断这块市场的瓶颈卡在没人愿意干这件事上,竞争压力反倒算不上什么。
医学正在从模拟变成数字
访谈最后聊到还有哪些领域有大机会,他先说了合成医学。
现在的技术条件是可以定制 RNA。他把医学的过去和未来做了个对比:
过去找药的方式,大部分时候像是靠碰运气。就像在森林里捡到一根看起来能当棍子用的树枝,挺原始的。(马斯克)
而一旦能合成 RNA,医学就可以被看成数字化的东西。你只要知道该往那条合成 RNA 链里写什么程序,理论上几乎什么病都能治。
这个类比很准确。传统药物发现是筛选,是在海量分子里试,试中了也未必知道为什么中。可编程的 RNA 是把治疗变成写代码,输入决定输出。他把这条路径叫"医学从模拟走向数字",这个说法比任何术语都好懂。
他还提到电动飞机,说这个方向有机会,已经有一些创业公司在做。主持人重复了一遍"电动飞机",他确认了一遍。
整场访谈到这里收尾。主持人最后说了句"你在文明进步中扮演了有用的角色",马斯克接了个玩笑:谢谢,这是 AI 将来会说的话。
把这场访谈读完之后
这场访谈表面上是在回应一场官司,实际上马斯克借着这个机会把整套逻辑重新讲了一遍。
官司那部分真正的信息在于他要阻止一个判例成立,"他输了"只是个表层说法。OpenAI 从非营利转成八千亿美元营利实体,在他看来是一次缓慢的、分步骤的资产转移,而法律因为它的缓慢而抓不住它。分步骤发生的事,往往比一次性发生的事更难追责,这是制度设计上的一个漏洞。
技术那部分的核心信息是尺度。他把能源讨论从"用哪种发电方式"直接拉到"太阳占太阳系 99.8% 的质量"。他把 AI 讨论从"模型有多强"拉到"五年后智能总量超过全人类、一亿到十亿个人形机器人、经济规模翻倍"。你可以不同意他的时间表,但他给的每个数字都能被追问、被检验、被推翻。这是他的表达方式里最值得学的一点:把判断说成可验证的量化预测,避开价值观表态。
还有一处容易被漏掉:他谈福特的那几句。整场访谈里,火星、太空算力、脑机接口都是未来,只有福特那段是过去。而他夸福特的角度是"发明了复杂物体的规模化制造"。这跟他自己在做的事是一致的。他的宏大叙事下面,压着的是一套制造能力。没有这套能力,前面所有数字都只是 PPT。
- 遇到"某人输了官司"这类新闻,先分清是程序性驳回还是实体判决。程序性驳回不解决事实问题,只解决时间问题,后续上诉和判例效应才是看点。
- 判断一个非营利转营利的案例时,别只看终点,去看每一步的条款变化。利润上限、规模限制、AGI 触发条款,这些是真正决定性质的细节,而它们往往在几年里被逐条删掉。
- 评估任何关于 AI 的预测,先看它给的是能力判断还是规模判断。能力判断可以说"一两年内全面超越",规模判断必须给数字,比如智能总量、机器人数量、经济翻倍周期。后者才可证伪。
- 听到能源路线之争时,试着把尺子换成太阳。人类全部能耗不到太阳输出的一万亿分之一,这个量级一旦建立,很多争论会自动降级为细节问题。
- 读技术愿景时留意作者真正的底层能力。马斯克夸福特夸的是量产方法,这说明他所有项目的共同难点在制造,不在概念。判断一家硬科技公司,也先看它的量产能力。
- 把他的时间表当成校准工具,而不是预言。记下 2031 年这个节点和那几条量化判断,几年后回看,能校准你自己对技术曲线的直觉。
Oh, yeah.
You do like ranking things. It’s true.
Yes. Oh, that one. I was like, what are you talking about? Yeah, I guess. Well, they basically just decided that the statute limitations have passed. They did not actually render an opinion on whether, you know, there have been unjust enrichment or the chatbot has been stolen, which I think is obviously the case. But I do think it’s somewhat of an ambiguous situation because, you know, what happened was by degrees, meaning it wasn’t stolen all at once, it was stolen one piece at a time. And so we have to say like, well, was there really even a basis for legal action before, you know, Proposals Go? And I don’t think there is actually.
So, you know, like the first step, I don’t want to belabor this point, but the first step into adding a for-profit thing had a cap for profit and was at a small scale. And also that all stock would revert to the charity upon the invention of artificial general intelligence. That clause has now been removed, so it will continue to be a for-profit after developing artificial general intelligence. That removal was quite recent, just in the last few weeks. So, it is, at this point, it is whatever, an $800 billion for-profit company somehow from a nonprofit.
And I think this is a dangerous precedent to set because if it means that someone can start, take money as a nonprofit, convert that to a for-profit when it’s successful, it undermines all charitable giving in America.
Yeah, exactly. What do you got to lose?
Yeah, I think we necessarily have to appeal because this will become precedent, you know, which people can actually then essentially loot charities and use this case as a basis for doing so, which I think would be wrong.
More or less just happened. I mean, well, let’s just say that when I was in college, I wanted to be part of things that would change the world in a significant way. I didn’t originally think I wanted to be doing things in AI because AI was somewhat of a double-edged sword. But I did want to accelerate the transition to a sustainable energy economy, which is the sort of electric cars, solar and batteries. And, you know, I think some kind of brain-machine interface that can give you cybernetic superpowers, I think is probably good.
It could, you know, help people that have brain or spine injuries, you know, restore— enable people who’ve never spoken for years to speak again, which is— which we’ve done. Give people eyesight who have lost both eyes or the optic nerve or maybe have never even seen at all, blind from birth. By direct interface to the optical centers in the brain, you can actually restore eyesight or give people eyesight that they’ve never had before. And you can enable people to walk again, which I think is profound. I mean, these are kind of Jesus-level things, you know, like I say, like when technologies are hitting like Jesus-level miracles, you know, like, that’s pretty good.
Jesus-level, you’re like, okay, that’s pretty good. So, and then SpaceX is about making life multi-planetary, expanding consciousness to the stars. And hopefully we will make a fundamental breakthrough later this year, which is to have the first fully reusable orbital rocket, which is necessary in order to transport enough tonnage to the Moon or Mars to create a self-growing city, basically to extend civilization to other planets.
I mean, I think probably you need to move a rough order of magnitude of about 1 million tons to the moon or to Mars to create enough of an industrial base to have the civilization there be self-growing, meaning like, the acid test being if the resupply ships from Earth stop coming for any reason, does the civilization on the Moon or Mars continue to grow or does it die out? And I do want to emphasize, like, being a multi-planet civilization doesn’t mean like we leave Earth and go somewhere else, because that would just be a single-planet civilization somewhere else.
Yeah. And a tougher place to live than here, by far. Earth is extremely easy and comfortable compared to the Moon, Mars, or pretty much anywhere else. But, you know, I think at some point we want to be a spacefaring civilization. We want to be out there among the stars and be, yeah.
Well, I guess I don’t really get up and say, “Hey, what shall I innovate today?” You know, type of thing. It’s really just building the technologies necessary to extend life beyond Earth. You know, there’s the Starlink internet, which is rebuilding the entire internet in space, but better, you know? I guess that’s kind of pretty cool. There’s, yeah, the Optimus robot that we’re developing at Tesla, the self-driving cars, solar power at scale, you know?
Well, it’d be difficult for me to sort of comment on that, you know, because, you know, there’s publicly traded companies and one’s publicly traded and one’s, you know, about to be. So it would be like difficult to, you know.
I’m not, I think I’m not allowed to comment on these things. That’s like a quiet period or something to that effect.
Well, no, I mean, I’m not really going to say that’s like, what’s your favorite kid type of thing, you know? I think you can’t really say that, you know.
Well, I guess, you know, I’m obviously a big fan of Nikola Tesla since named the company after him. But I also like, you know, I don’t know, Edison is great too. Some people don’t like Edison, but I think he did some pretty impressive stuff. And like General Electric is like Edison’s company. I mean, I certainly admire like great scientists, engineers, but like anyone, I mean, you know, Ben Franklin, Shakespeare, Newton, Einstein, you know, the usual suspects.
It was like actually more awesome than people realize because I mean, he basically invented mass manufacturing of complex objects. Like people didn’t really know how to do that before Ford.
And like Ford actually created the automotive industry of Earth, really. And ’cause he pretty much started what became General Motors, and then he kinda got, I don’t know, kicked out of that or something, and then started Ford Motor Company, and really developed mass manufacturing of complex objects. It’s quite a tricky thing. And then everyone just pretty much copied him.
Well, I mean, I guess I mean, I don’t really think about these things that much, but I— well, I mean, I guess, you know, Jensen Huang at NVIDIA is like doing pretty great in terms of inventing, you know, developing AI computers.
Well, I think I’m like simultaneously both, honestly. It’s like, I mean, at this point, like, when I go to sleep, there’s like some AI breakthrough. When I wake up, there’s some AI breakthrough. And by lunchtime, there’s another AI breakthrough. So it’s like a head spinner at this point. It’s pretty obvious that we’re going to have AI that is vastly smarter than humans. And already, in some ways, it already is. I mean, it seems like AI unequivocally smarter than humans in every way, including in innovation, is probably not more than a year or two away, I would say.
This is kind of worrying. I hope it’s nice to us, you know.
In 5 years, we’ll have probably— so 5 years being, say, 2031. I think digital intelligence will exceed the sum of all human intelligence. That there will be in 5 years, probably there might be a, let’s say, at least 100 million humanoid robots, but maybe a billion. Yeah.
And I predict that the economy is probably twice its current size in 5, maybe 6 years. 5 to 7 years. Because you’re going to hit a doubling period, you know, where the economic output is increasing so, so fast that, you know, plus minus a few years, you will see giant changes.
Space is much easier than people may think. I mean, SpaceX at this point has 10,000 satellites in orbit right now. And in the future with Starship, we’ll be launching over 10,000 communication satellites per year, each one of which is much more capable than our current satellites. So you can expect that’s probably 10 to 100 times more communications capability than currently exists from space. But that will pale in comparison to the tonnage of AI satellites. It’s always helpful to use the physics tools of thinking in the limit. If you think of the limit, what is the most amount of AI compute you could possibly bring to bear?
And the way to access power that’s far beyond anything on Earth is in space. So these are some humbling ways to maybe frame the problem and see how tiny we are. Here on Earth, we are like microbes on a dust mote compared to the Sun. All of human civilization currently uses less than a trillionth of the energy, of the power emitted by the Sun. Much less than a trillionth. All human energy counted as generously as you could possibly imagine is much less than a trillionth of the Sun’s energy. So that means if you were to increase civilizational energy harnessed by a million, you would still be much less than a millionth of the Sun’s energy harnessed.
Another way to think about it is, because people sometimes say, “What about this form of power generation or that form of power generation?” I’m like, everything is irrelevant compared to the Sun. The Sun is 99.8% of mass in the solar system. If all mass that was not the Sun was burnt and you extracted the absolute maximum amount of energy from all mass that is not the Sun, the amount of energy produced by the Sun would still round up to 100% of the energy produced in the solar system. And then if you teleported 2 more Jupiters from other— somehow teleported 2 more Jupiters into our solar system and burnt them too, the Sun would still be 100% of power, rounding up.
So really everything is just extremely tiny compared to the Sun. Therefore when you go to space, if you want to climb the Kardashev scale in any meaningful way whatsoever, you have to go to space. Our rough estimate is that you could probably do a terawatt a year of AI-powered, solar-powered AI satellites launched from Earth. That would require something like 10 million tons a year of payload to orbit at 100 kilowatts per ton, and that would be a terawatt per year. The average power usage in the United States is half a terawatt. That would mean you’d be launching twice the amount of AI power than the entire electricity consumption in the United States per year.
This is not breaking any physics here. This is all achievable. Now, if you want to take that up several more orders of magnitude, then the thing to do is to build a base on the moon and construct a mass driver. Build the solar panels and radiators on the moon. Chips you could probably bring from Earth, but build those on the moon too. And that would enable you to do about a petawatt per year of solar-powered AI. Now I’m not sure how much AI we need at the end of the day, but that’s a lot.
It would be vastly greater than the Earth’s economy. Earth’s economy would be really inconsequential by comparison with that.
Well, I did say for the longest time that people ought to build tunnels, and then everyone thought I was joking. And then I was like, “Fine, well, let’s just create a company, The Boring Company, to build tunnels.” Because this is an obvious way to reduce traffic to zero by going 3D with your transport system. Just as we go 3D with the buildings, it’s obviously illogical to expect that you won’t have traffic if you have buildings that are 3D and a transport system that is 2D, and everyone tries to go in and out of the 3D buildings at the same time.
But if you have a transport network that is 3-dimensional with tunnels going all over the place, then you can address any arbitrary level of traffic. Now, I encourage other people to start tunneling companies because The Boring Company is the only one, as far as I can tell, that’s taking this seriously.
Yeah, honestly, there’s a lot of opportunity in tunnels. There really is. And it’s quite hard work building these things. So I hope other people do it too. And I guess there’s probably a lot of opportunity in synthetic medicine. These days, you can basically create custom RNA. I think we’re going from a situation where medicine has been analog, where we kind of just find drugs kind of by accident most of the time, like finding sticks in the forest that seem like a useful club. It’s quite primitive how we found medication in the past. But once you can— and since we can now construct synthetic RNA, you can effectively think of the future of medicine as being digital.
If you just know what to program into that synthetic RNA strand, you can basically, I think, cure almost anything.
Electric aircraft are probably an opportunity there. I think there’s some startups in that field.
Electric aircraft. Electric aircraft.
“You played a useful role in the advancement of civilization.”
Thank you for human. That’s what the AI will say. Fair.
我最初并不觉得自己想从事AI,因为AI某种程度上是一把双刃剑。
I didn’t originally think I wanted to be doing things in AI because AI was somewhat of a double-edged sword.
我想强调,成为多行星文明并不意味着我们离开地球去别处,因为那只会是在别处的一个单行星文明。
And I do want to emphasize, like, being a multi-planet civilization doesn’t mean like we leave Earth and go somewhere else, because that would just be a single-planet civilization somewhere else.
与月球、火星或几乎其他任何地方相比,地球都极其容易且舒适。
Earth is extremely easy and comfortable compared to the Moon, Mars, or pretty much anywhere else.
很明显,我们将拥有远比人类聪明的人工智能。而且在某些方面,它已经如此了。
It’s pretty obvious that we’re going to have AI that is vastly smarter than humans. And already, in some ways, it already is.
我认为数字智能将超过所有人类智能的总和。
I think digital intelligence will exceed the sum of all human intelligence.
我预测,经济规模在5年、也许6年后可能是当前的两倍。5到7年。
I predict that the economy is probably twice its current size in 5, maybe 6 years. 5 to 7 years.
目前,全部人类文明使用的能量还不到太阳释放能量的万亿分之一。远不到万亿分之一。
All of human civilization currently uses less than a trillionth of the energy, of the power emitted by the Sun. Much less than a trillionth.
但如果你有一个三维的交通网络,隧道遍布各处,那么你就能应对任意水平的交通流量。
But if you have a transport network that is 3-dimensional with tunnels going all over the place, then you can address any arbitrary level of traffic.
如果你知道该往那段合成RNA里编写什么,你基本上就能,我认为,治愈几乎任何疾病。
If you just know what to program into that synthetic RNA strand, you can basically, I think, cure almost anything.
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