2019 年 8 月 29 日,上海世界人工智能大会。马云和马斯克在台上对谈了四十多分钟,两个人几乎没有一个观点对得上。
一个说 AI 是人类文明面临的最大威胁,一个说 AI 根本不可能是威胁。一个抬头看火星,一个低头看海里的塑料瓶。一个说人类在每一个维度上都会被超越,一个说人类根本造不出比自己更聪明的东西。
吵完四十分钟,谁也没说服谁。这段对话之所以值得重读,是因为它把两种世界观摆在了同一张桌子上:一种相信技术进步的速度不可控,所以要提前给自己准备后路;一种相信人类的适应能力,所以把焦虑本身当成最大的问题。
文字稿来自现场录音的语音识别,错误不少。比如"Alibaba intelligence"被识别成了"a name AI","Neuralink"被拆成了"neural lace"和"new or linked","book smart"变成了"college smartness"。下面把关键段落按原意还原,顺带说说这两套逻辑各自的软肋。
"我讨厌 AI 这个词"和"AI 研究者最大的错误是以为自己聪明"
开场马云先拆名字。
我讨厌人工智能(Artificial Intelligence)这个词,我叫它阿里巴巴智能(Alibaba Intelligence)。(马云)
这句玩笑藏着他的整套立场。他不把 AI 当成一种异质的东西,而当成人类能力的延伸和放大。既然是延伸,那就没什么好怕的。
马斯克接得很直接:
人们普遍低估了 AI 的能力。他们以为 AI 就是一个聪明的人,实际上它会比最聪明的人类还要聪明得多。我看到 AI 研究者犯的最大错误,就是假设自己是聪明的。跟 AI 比起来,他们不是。所以很多人没法想象一个比自己更聪明的东西。(马斯克)
这段话的重心落在"聪明人往往想不出比自己更聪明的东西"上,而不是"AI 很强"。背后有一层认知上的死结:要准确评估超级智能,你得先拥有接近它的智力。人类判断不了,只能等它出现。
那怎么办?马斯克给的方案很朴素,一句老话:
打不过就加入。Neuralink(脑机接口公司)做的就是这件事,我们能不能跟 AI 一起搭车往前走。(马斯克)
注意这个措辞,"搭车"(go along for the ride)。他没提控制,也没提驯服,说的是把自己变成车上的一员。这个姿态本身就说明他并不认为自己能赢。
99.99% 的预测都是错的,包括刚才那个
马云的反击从方法论开始。
我不是搞技术的,我关心的是生活。未来太难预测了。人类历史上对未来的预测,99.99% 都是错的,包括刚才那个预测。(马云)
马斯克笑着补了一刀:这也说明 80% 的统计数字是编的。这句话是英文世界的一句老笑话,两个人在台上互相拆台,气氛很松。
马云后面那段才是他的核心论点:
人们那么担心人工智能,其实应该对自己更有信心。今天我们没有的解决方案,明天会有。我们连年轻人的问题都还没解决,哪来的解决方案?所以我挺乐观的,我不认为 AI 是威胁。(马云)
他接着把人分成两类。一类是"书本聪明"(book smart)的人,读了很多书,习惯从理论往外推;另一类是"街头聪明"(street smart)的人,像他这样在市场里滚出来。他的说法是:后一类人从来不害怕,只会觉得好玩,然后改自己去拥抱它。
马斯克没有反驳这套人生哲学,他只把自己的位置说清楚:
我天性乐观,这一点要说清楚,我不是在说末日论。我只是说,技术变化的速度极快,已经超过我们理解它的能力。我不确定这是好事还是坏事,我不知道。(马斯克)
两句话放在一起看,分歧很清楚。马云判断的是"人类能不能适应",马斯克判断的是"适应速度跟不跟得上"。前者问结果,后者问斜率。斜率太陡的时候,结果好也没用。
火星是逃生舱,还是给意识做的一份备份
话题转到火星,两个人开始互相打趣。
你为什么对火星这么好奇?
我刚从火星回来。(马云)
马云说他不是火星的粉丝。爬珠峰要等电梯修好再去,去火星也一样。他还提到马斯克另一家公司 The Boring Company 在地下挖隧道,说每次看到马斯克搞太空的新闻都很尊敬。
马斯克认真解释了动机,这段话说得最实:
我没有把火星当成逃生出口,也不觉得地球要完蛋了。真正的原因是,确实存在一个无法消除的概率,不管我们多努力,地球上都可能发生某些事。地球四十五亿年历史上,生命第一次有可能被延伸到地球之外。我们应该尽快去做。(马斯克)
他给的资源比例也很具体:
占用地球资源不到 1%,用来让生命或者意识成为多行星的。我觉得这个数字应该介于我们花在口红上的钱和花在医疗上的钱之间。至少拿出全球 GDP 的 0.5%,也许 1%,去把生命延伸到多个星球上,这算是资源的合理用法。绝大部分资源还是留在改善地球上。(马斯克)
马云的回应很漂亮:
我佩服你探索火星的勇气。但我不算火星的粉丝,因为去火星太容易了:你从楼顶往外迈一步就到火星了,问题是你永远回不来。(马云)
他接着把话题拉回地面:
我们需要像你这样的英雄,但我们需要更多像我们这样在地球上每天埋头干活的英雄。如果能把资源集中在帮海洋清理垃圾上,那件事比去火星更难。而人工智能可以帮我们做到。(马云)
"捞海里的垃圾比去火星更难",这句听着像抬杠,其实是个真问题。去火星是一次性工程,有明确的目标、预算和终点;海洋塑料治理要持续几十年,靠公共品供给,没有终点,也没有掌声。难点在组织,不在技术。
上海工厂和"中国就是未来"
中间有一段马斯克对中国的表态,值得单独拎出来。
2019 年 1 月特斯拉上海超级工厂动工,同年年底就投产,这个速度在当时让整个汽车行业都吃了一惊。马斯克在现场的原话是,他被特斯拉中国团队的效率震住了,说自己一辈子没见过这么快的事,"中国就是未来"。
他还提到中国民营火箭公司已经实现入轨,说这件事非常难,他对能做到的人怀有极大敬意。
把这段放在整场对话里看有点微妙。马斯克一边在说技术变化的速度超过了人类的承受能力,一边在为另一种速度鼓掌,工厂从荒地到量产只用了十个月。他对"快"本身没有意见,他警惕的是快在某些方向上失控。
工作三天,每天四小时
聊到就业,马云抛出了一整套时间账。
未来二三十年,人类会活得更长。生命科学技术大概能让人活到一百岁、一百二十岁。这可能不是好事,因为你爷爷的爷爷还在辛苦上班。(马云)
他算了三个时代的平均寿命:农业时代三十到三十五岁,工业革命之后七十岁,AI 时代一百到一百二十岁。然后接了一句:
为什么我们要有那么多工作岗位?我觉得人应该一周工作三天,一天四小时。电力的意义就是把人从体力活里解放出来,让人有更多时间。晚上可以去唱歌跳舞。因为人工智能,人会有更多时间去享受作为人的乐趣。(马云)
他还拿自己的家史做对比:他爷爷一辈子去过三个城市,他父亲去过三十个,他自己去过三百个,他的孙辈大概会去三千个,而且总是在特斯拉里、在机器人旁边、一直在移动。
至于那些担心失业的人,他的答案是:人会活得更好,所以不想生孩子,那就用 AI 和机器人来照顾老人。
马斯克没有争论工作时长,他直接跳到结论:
随着时间推移,AI 会让工作变得没什么意义。我建议去学工程、物理这类东西,或者去做那些人们就是想跟人打交道的事,比如艺术。我们还得想办法解决脑机接口这件事,否则我们会被甩在后面。这件事得尽快做,我们时间不多。(马斯克)
"我们时间不多"这句话他在这场对话里说了不止一次。一个在造火箭的人反复说时间不够,另一个在谈生活方式的人说不用着急,整场对话最根本的张力就在这里。
"我从不担心我解决不了的问题"
主持人问马云怎么看待脑机接口这类他觉得棘手的事,他给了一个非常马云的回答:
我从不担心我解决不了的问题。我让别人去解决。如果没人能解决,那就随它去。这就是我的活法。(马云)
这句话听着像耍滑头,其实是另一种资源分配策略:把注意力放在自己有能力改变的边界内。它的前提是世界足够宽容,容得下那些"随它去"的问题。
马斯克的整套逻辑建立在一个相反的假设上。有些问题不能随它去,因为它不会等你准备好才发生。
教育才是马云真正焦虑的地方
聊到教育,马云明显认真起来。他说别人担心就业,他担心教育。中国每年大约有 1800 万新生儿(文字稿里录成了 1800,按上下文应为 1800 万),在他看来这个数字还不够。然后他说了一句值得记住的话:
地球上最好的资源不在煤矿,不在石油,不在电力,在人脑。(马云)
接着他给出了自己最硬的一个判断:
这两年大家一谈 AI 就说人类会被机器控制,我从来不这么想,我认为不可能。因为机器是人造的,人类不可能造出另一个比人类更聪明的动物,尤其是有这么多聪明人在的时候。(马云)
马斯克的回应很平静:
我非常不同意。我们绝对能造出比自己更聪明的东西。以前也没有人类,对吧?所以我说,我看到聪明人犯的最重要的错误,就是假设自己是聪明的,其实不是。(马斯克)
马云要例子,马斯克给了:
计算机在很多维度上已经比人强得多了,我们只是一直在挪动球门柱。跟计算机下围棋,就像跟宙斯打架,赢不了的。人类还占优势的领域,是一块越来越小的角落,而且每年都在缩小。很快我们会在每一个维度上被远远超越,我保证。(马斯克)
马云的收尾是:跟电脑比算数是愚蠢的。电脑聪明,但人类会造出比电脑聪明得多的工具。
这段对话表面在争 AI,实际在争一件事:智能有没有上限。马云把智能理解成一种属性,人有,机器也可以有,造物超不过造物主。马斯克把智能理解成一个连续变量,人类只是这条曲线上的一个点,而且不是最高点。
两种框架都能自洽。区别在于,如果你相信前者,你不需要为 AI 做任何准备;如果你相信后者,你需要做很多准备,而且现在就要开始。
把意识存成游戏存档
主持人问,有 AI 帮忙,人能活多久。
如果我们能把脑机接口做成,年龄就不太重要了。你可以保存自己的状态,也可以恢复自己的状态,就像游戏存档。我真的觉得,如果我们真想解决生物学意义上的衰老,是可以解决的,那需要改 DNA,但那就是一个时钟。(马斯克)
马云的回答很短:他想留在地球上,想在地球上做事,想尽所能让地球变得更好。
他还补了一句对 AI 与环境的态度:人类越了解自己,就会越聪明、越有智慧,AI 能帮上忙。
两种"更好"的定义
结尾马斯克说了他真正的目标:
我希望我们做的任何事,都能让通往这个方向的路更好走一点:让七十四亿人过得更好、更健康。这就是我们的世界。(马斯克)
马云夸了特斯拉,说它让世界更干净、更安静,他很高兴这家工厂建在中国。然后是整场对话里最像总结的一段:
相信人类,相信年轻人。为今天承担责任,但不要把明天的解决方案都用光。人会犯错是好事,从错误里学习是好事,会死也是好事。(马云)
马斯克顿了一下,说:"这大概是真的。"
整场对话最值得回味的就在这儿。马云那句"会死也是好事",是全场唯一一个马斯克没有反驳的观点。一个把赌注押在延长意识存在时间上的人,承认了这句话的分量。
六年过去,再回头看这段对话,几个判断已经被现实部分验证了。大模型的迭代速度确实快到让监管、教育和就业体系都跟不上;AI 在围棋、写作、编程、图像这些具体领域逐个超越人类;"人类能不能造出比自己更聪明的东西"这个问题,已经从哲学问题变成了工程排期问题。
但马云那一半也没有全错。人对工具的适应能力,比技术乐观主义者和悲观主义者估计的都强。AI 真正改变的,往往是"哪些技能还值钱",而"人会不会被取代"反倒没那么要紧。他说的街头聪明,在 AI 时代依然稀缺,用途从"抓机会"变成了"判断什么值得做"。
这场对话真正留下的,是两种应对不确定性的方法:一种提前准备后路,一种当场解决问题。前者需要资源和技术,后者需要信心和运气。
普通人两样都没有,那就只能选一件具体的事,把它做到机器做不到的程度。
- 重读这段对话时,先分清哪一半是预测、哪一半是立场。马斯克关于"技术变化速度超过理解速度"的判断可以验证,马云关于"人类不可能造出更聪明的东西"的断言无法验证。前者值得跟进,后者只能当态度。
- 把"我从不担心我解决不了的问题"这句话用对地方。它适合用来管理情绪,不适合用来做技术选型。判断一件事该不该关心,标准看它会不会在你准备好之前找上门,你现在能不能解决并不重要。
- 马斯克给年轻人的建议可以直接抄:工程、物理,或者那些人们就是想跟真人打交道的事。翻译成人话就是,要么深入理解物理世界,要么提供机器给不了的人际体验,中间地带正在被快速抽空。
- 如果你在做 AI 相关的产品或者团队,把"AI 会让工作变得没有意义"当成一个设计前提。用户真正愿意付钱的部分,正在从"完成任务"转向"过程体验"。
- 给自己留一份"状态存档"。不必等脑机接口,把重要的判断、方法、人际关系定期写下来就够了。技术变化越快,可回溯的个人记录越值钱。
- 看这类对话时,注意谁在说"我们时间不多",谁在说"随它去"。两种人合作时效率最高,因为他们互相补上了对方缺的那一半。
0:00like a name AI it sort of sounds a bit like that yeah IIIi hate the word AI called artificial intelligence I call it Alibaba intelligence yeah I think generally people underestimate the the capability of AI they sort of think like it's a smart human but it's it's really much it's gonna be much more than that every much smarter than the smartest human the biggest mistake that I see artificial intelligence researchers making is assuming that they're intelligent yeah they're not compared to AI and so a lot of them cannot imagine something smarter than themselves but AI will be vastly smarter vastly in a situation where you if you you know the old saying if you can't if you can't
0:48beat them join them you know that's what neural link is about is like can we go along for the ride with AI I'm not a tech guy I think I'm all about life it's so difficult to predict the future 99.99% of the predictions the human being had in history about the future all wrong including that one oh yeah only you know the the zero point zero zero percent of the prediction are right there right because by accident yeah but it's also true that 80% of statistics are false yeah so my meaning Lauren come on guys jerk when people
1:34worry a lot about artificial intelligence people should have more confidence in themselves a lot of solutions we don't have today that but there will be solutions tomorrow we don't have solutions for the young people we have solutions so I'm quite optimistic and I don't think artificial
1:53intelligence is a threat I don't think that fish intelligent is something terrible but human being a smart enough to learn that people worry a lot about this today are those people I called them college smartness people like us street smart we never scared of we think it's a great fun and we want to change our self to embrace it I'm a naturally optimistic person to be clear I'm not saying hey doom and gloom I'm just saying the rate of change of technology is incredibly fast it is outpacing our ability to understand it I'm not sure is that good or bad I don't know the mind that you want to go on the Mars so we go to the Mars actually I'm not interested in a Mars I just came
2:38back from there so hahaha I'm more interesting on the earth that thinks what's going on happening here so what why you're so curious about the Mars it's not it not from the standpoint of it just being an escape hatch or because I think that Earth is doomed but there's a certain probability that is irreducible that something may happen to earth despite our best intentions despite everything we try to do this is the first time in the four-and-a-half billion year history of Earth that it's been possible to extend life beyond Earth and we should you try to do this as quickly as possible Adam I your courage for explore the Mars I'm not that fan of the Mars because I
3:21think it's easy to go to the Mars when you go on the top of the hills all the out of the building just a one step you go to Mars but you will never be able to come back yeah that's my view that works and I hate to go to the Himalayas to I mean when you climb on them I think someday I would go there when the elevator is ready I would go to have a look I heard you you're gonna dig eternal in deep in the earth which is amazing every time when I read a news about you are interesting the outside space I look at you with a great respect we need a heroes like a you but we need more heroes like us working hard on the earth improving things every day that's
4:07what I want sure I mean to be clear I'm very Pro Earth from a resource standpoint I'm talking about less than 1% of Earth's resources should be dedicated to making life multiplanetary or making consciousness multiplanetary I think it should be like somewhere in between how much we spend on lipstick and how much we spent on healthcare like at least half a percent of Earth's GDP on extending life to be multiplanetary maybe 1% I'd say
4:36seems like a good a good use of resources towards magnitude more resources spent on earth it seems like a wise investment for the future and obviously I spend a lot of my time on sustainable energy with Tesla with electric cars and solar and batteries and that kind of thing and I'm really excited to be here in Shanghai for the the Shanghai Giga Factory which is I think that Tesla China team has done an amazing job it really might learn like I've just astounded by how good the job is and how much progress has been made and I think it's a good story for the world and to say like look look how much progress you can make in China this is extremely impressive like if I
5:23have my hat is off you know you guys Rock I've never seen anything but so fast in my life before to be totally Frank I've seen some crazy things so you know I think it's like I really think China is the future well that's very impressive and there's also some great progress on entrepreneurial rocket companies in China as well I believe to have made orbit it's very difficult very hard to make orbit achieving orbit I have great respect it was very hard yeah pick up another topic sure yeah job I
5:57think it jobs jobs or life jobs or anything true is the next of 2030 years human being will live much longer the live science technology is gonna make people live probably a hundred a hundred twenty years that may not be a good thing because you get a grandfather's grandfather still working hard but the challenge is my question why should we
6:24have a lot of jobs I think people should work three days a week four hours a day when we have electricity the power of electricity is that we make people more time so you can go to the kala okay in the evening you can go to dancing party in the evening I think because of artificial intelligence people will have more time enjoy being human beings your
6:52life in my life I think I visit the problems 300 cities in my life my graph my father visited 30 cities my grandfather would visit only three cities so my grandchildren private property will visit the 3,000 cities all his life he's always on the Tesla is always on the robots always traveling around so I don't think we need a lot of jobs at that time the jobs we need is make people happier make people experience the life enjoy the human beings so I don't worry about the job jobs a lot first we're gonna have love jobs second we don't need a lot of jobs third there's a very interesting thing in the every culture period average age is like a 3035 years old in the industry
7:48period technology revolution people can live 70 years so in the artificial intelligent period people can live a hundred to the years as I think now the problem comes when people's life is getting better people don't want to have children so we need artificial intelligence the robots to take care of the old guys for sure you will not be happy or you were happy because when
8:18your grand grandfather said all I need to work tomorrow and there's a disaster yeah over time a I will make jobs kind of pointless I would recommend studying engineering physics that kind of thing or working on something where people just want to interact with other people you know art of course we're gonna have to figure out this new or linked situation otherwise
8:44we will be left behind it's very important we do this quickly we don't we don't have much time before what we don't have much time to solve the neuro-link I never worry about the things that I cannot solve I let other people to solve it if nobody can solve it just let it be that's my life oh let's talk about education I'm of quite an interesting about education people worry about jobs but I worry about education China today we have 1800 new
9:18babies born every year which which is not enough we need we need to have like much more than that but I think the best resources of the human beings on the or the best resources on the earth and not the coast not the oil not the electricity it's it's the human brains in my life and especially last two years when people talk about an AI say human human being will be controlled by mushrooms I never think about it I think it's it's it's it's impossible right it's impossible because because human beings they are different machines are invented by human beings and according
10:03to the size right humans can never create another animal that is smarting humans especially where you have so many smart people it's impossible to make another smart people I very much disagree with that ok yeah that's good yeah we can definitely make things smarter than ourselves I mean the they didn't used to be humans right so the then like I said the most important was the mistake I see smart people making is assuming that they're smart they're not yep so give me example what animals or things that a human being made that is smart a human beings well computers actually are already much smaller than people on so many dimensions we just keep moving the goal posts trying to
10:49play a computer go is like trying to fight Zeus it's a not going to work where there's just a smaller and smaller corner of what of intellectual pursuits that humans are better than computers and that every year it gets smaller and smaller and and soon we will be far far surpassed in every single way guaranteed it's stupid to compete with a computer
11:16on plague old computers us a clever but there will be more tools that human beings will create it much clever than computers that's my view life or human being machines so you want to talk about your cars ultimately autonomy what do you want to talk feel this feels like one of those steps in a videogame we would like pick a path a choice I choose life okay life so how much longer do you
11:48think people can live for with the helper AI can a I help with environment sustainability okay if if we can if we endure the neuro lace then I think or the neural link essentially the age will not matter that much you can simply save your state and restore your state like just like a saved game essentially something very close to that I do think we can we can
12:23solve a biological aging if we really wanted to you'd have to make DNA changes but it's obviously just on a on a clock well I think AI can definitely help the environment suspend sustainability and when human beings know themselves better human being will be smarter and will be wiser I want to stay in this earth I want to work on this earth I want to do anything I can to help this earth to better because even in a go to this space is great but if we can spend at resources just to focus on helping pick up the garbage from the oceans that thing is more difficult than to go to the Mars but artificial intelligence can help us achieving that and solve the
13:10problems my hope is that anything we can do to improve this road to helping seven point four billion people live better and live healthier and this is all about our world and I think we will be working very happily because I love your product Tesla you know making war cleaner and no
13:33noise and those such great technology I'm happy you have a de Factory in China and I think we need to do more things to improve this earth improve this world and make sure that people are happier and people care about the family people care about the house that's all we should do and trust us trust a human
13:55beings and trust young people let's take responsibility for today but let's not take away all the solutions for tomorrow it's great human being make mistakes it's great human being learn from mistakes it's great to die ha ha that's probably true thank you yeah you can live long I said like yeah thank
14:21you fight for the lights of consciousness thank you thank you so much for this
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