Hard Fork
The Ezra Klein Show: The A.I. Revolt Is Here
00:0001:19:47
00:00
AWS AI is how the world's leading organizations are transforming their industries,
AWS AI 就是全球领先组织改造他们行业的方式，
00:05
not in theory, but in production, at scale, right now,
不是停留在理论上，而是已经在生产环境中、大规模地、就在现在，
00:09
like turning raced-day data into real-time insights for Formula One fans,
比如把比赛日数据变成 Formula One 车迷的实时洞察，
00:13
reinventing the journey for every United traveler,
为每一位 United 旅客重新构想旅程，
00:17
and personalizing every trip with booking.com.
并通过 booking.com 让每一趟旅行都个性化。
00:20
However big the question for your business or industry, AWS AI is how.
不管你的企业或行业面临多大的问题，AWS AI 就是解决之道。
00:25
So we are taking this week off, along with our producers,
所以这周我们休假，我们的制作人也一起休，
00:51
to get ready for our finale extravaganza,
为我们的收官盛典做准备，
00:54
our last ever episode of Hard Fork, which will be next week.
我们史上最后一集 Hard Fork，会在下周播出。
00:57
And when you see the choreography, you'll understand why we need it a week to prepare.
等你看到编舞，就明白我们为什么需要一周来准备了。
01:01
Exactly. There's pyro, there's choreo, it's going to be a big production.
没错。有烟火，有编舞，会是一场大制作。
01:04
There's Oreos.
还有 Oreos。
01:05
But until then, we have a very special treat to share with you.
但在那之前，我们有一个非常特别的惊喜要分享给你。
01:11
You may remember past Hard Fork guest, Jasmine Sun.
你可能还记得 Hard Fork 以前的嘉宾 Jasmine Sun。
01:14
She recently appeared on the Ezra Klein show to talk about her reporting on data centers.
她最近上了 Ezra Klein show，聊了她关于 data centers 的报道。
01:20
Yeah, she took this really fascinating reporting trip through the Midwest this summer,
对，今年夏天她穿越 Midwest，做了一趟非常引人入胜的报道之旅，
01:25
interviewing people on all sides of the debate throughout Wisconsin and Michigan.
在 Wisconsin 和 Michigan 各地采访这场辩论各方的人。
01:30
She wrote a piece called No Data Centers in My Backyard, Money, Power, and Populism
她写了一篇题为 No Data Centers in My Backyard, Money, Power, and Populism
01:35
in the AI build-out on her sub-stack, really great, really worth a read.
in the AI build-out 的文章，发在她的 sub-stack 上，真的很好，很值得一读。
01:40
And after you read it, you should listen to this conversation with Ezra.
读完它之后，你应该听听这段和 Ezra 的对话。
01:43
Yeah. And then, we'll be back next week, one more time.
对。然后，下周我们会再回来一次。
01:47
And among other things, we'll tell you what we're doing next.
而且，我们还会顺带告诉你我们接下来要做什么。
01:50
We'll be back in a week to say goodbye and bring you our very special finale episode.
一周后我们会回来道别，并给你们带来我们非常特别的最终集。
01:55
But in the meantime, enjoy Jasmine and Ezra talking about data centers.
不过与此同时，请欣赏 Jasmine 和 Ezra 聊 data centers。
02:00
What is big, ugly, and has united Republicans and Democrats at a time when it has felt like nothing else could.
什么东西又大又丑，却能在感觉没什么别的东西能做到的时候，把共和党人和民主党人团结起来？
02:33
AI data centers.
AI data centers。
02:35
Last August, a heat-bap news poll found that about four and ten voters would oppose a data center being built where they live.
去年八月，一项 Heatmap News poll 发现，大约每十个选民中就有四个会反对在自家所在地建 data center。
02:42
By May of this year, opposition grew to seven in ten.
到了今年五月，反对比例上升到了十分之七。
02:46
Florida Governor, Ron DeSantis, or Republican, of course, has proposed a new citizens' bill of rights for AI.
佛罗里达州州长 Ron DeSantis，当然也是共和党人，提出了一项新的、针对 AI 的公民权利法案。
02:52
The incentives of big tech are not the same as what's in the interest of the people and the public.
big tech 的激励逻辑，与人民和公众的利益并不相同。
03:00
Senator Bernie Sanders called for a national data center moratorium.
参议员 Bernie Sanders 呼吁实行全国性的 data center 暂停令。
03:03
This moratorium will give democracy a chance to catch up with the transformative changes that we are witnessing
这项暂停令会给民主一个机会，去赶上我们正在目睹的变革性变化。
03:11
and make sure that the benefits of these technologies work for all of us, not just the wealthiest people on earth.
并确保这些技术带来的好处能惠及我们所有人，而不只是地球上最富有的人。
03:21
There are over a hundred local or statewide moratorium proposals across the country.
全国有超过一百个地方或全州层面的暂停令提案。
03:26
Here in New York, Governor Kathy Hoekle, not usually thought of as a hardcore populist, just imposed a one-year moratorium on data center construction.
在 New York 这里，通常不被视为铁杆民粹主义者的州长 Kathy Hoekle，刚刚对 data center 建设实施了为期一年的暂停令。
03:33
At least, hyper scale AI data centers consume enormous amounts of power.
至少，hyper scale AI data centers 会消耗巨量电力。
03:38
Truly threatening to outpace our grid's capacity and they drive up costs for local rate fairs.
真的有可能超出我们电网的承载能力，还会推高当地电费缴纳者的成本。
03:47
And I refuse to let those costs be passed on to New Yorkers who already paid too much for the utility bills.
而且我拒绝让这些成本转嫁给已经在公用事业账单上付得太多的 New Yorkers。
03:55
So I wanted to get into the fight over data centers.
所以我想深入聊聊这场围绕 data centers 的争斗。
03:58
How much of this is really about water or electricity or aesthetics?
这其中到底有多少是真的跟水、电力或美观有关？
04:02
And how much is about AI itself and the companies that are behind it?
那其中有多少是关于 AI 本身，以及它背后的那些公司？
04:07
My guest today is Jasmine Sun.
我今天的嘉宾是 Jasmine Sun。
04:09
Jasmine has been doing excellent coverage of both the culture inside the AI companies, unusual culture, and the anger that is building against them in the rest of the country.
Jasmine 一直在出色地报道 AI 公司内部的文化——那种很不寻常的文化——以及美国其他地方对这些公司日益增长的愤怒。
04:18
I highly recommend following her sub-stack, but right now she just finished her reporting trip in the Midwest.
我非常推荐大家关注她的 sub-stack，不过她现在刚结束在中西部的报道之旅。
04:24
Talking to the people organizing against these data centers.
去和那些正在组织起来反对这些 data centers 的人交谈。
04:28
And I wanted to hear what she learned.
我想听听她了解到了什么。
04:30
AWS AI is how the world's leading organizations are transforming their industries, not in theory, but in production, at scale, right now.
AWS AI 正是世界领先组织转型各自行业的方式——不是停留在理论，而是此刻就在 production 中、大规模地实现。
04:47
Like turning race day data into real-time insights for Formula One fans, reinventing the journey for every united traveler, and personalizing every trip with booking.com.
比如为 Formula One 车迷把 race day data 变成 real-time insights，为每一位 united traveler 重新构想旅程，并用 booking.com 让每次出行都个性化。
04:57
However, big the question for your business or industry, AWS AI is how?
不过，无论你的业务或行业的问题有多大，AWS AI 就是方法。
05:03
Recently we asked about how you share your New York Times account, and you had a lot to say about New York Times games.
最近我们问过你是怎么共享你的 New York Times 账号的，而你对 New York Times games 有很多话要说。
05:09
I need my own New York Times login, because my sister is so much worse at the crossword than I am.
我需要有自己的 New York Times 登录账号，因为我妹妹做填字游戏比我差多了。
05:15
I discovered that he's already finished connections that day, and I'm like, Jonah, it was my day.
我发现他那天已经把 connections 做完了，我就说，Jonah，那天轮到我啊。
05:21
It doesn't let us play the same game since each other.
它不让我们彼此玩同一个游戏。
05:24
I play the Stoku.
我玩 Stoku。
05:25
I do the crossword.
我玩填字游戏。
05:26
I do the spelling bee.
我玩 spelling bee。
05:27
I do the wordle.
我玩 wordle。
05:29
Please help.
请帮帮忙。
05:30
My kids want to be able to play wordle, but with the wordlebot.
我的孩子们希望能玩 wordle，但要能和 wordlebot 一起玩。
05:34
We would love to be able to have our own puzzles.
我们很希望能有属于自己的谜题。
05:37
I love New York Times games.
我很喜欢 New York Times 的游戏。
05:39
I want to be able to play my own games.
我希望能玩自己的游戏。
05:42
Listeners, we heard you.
听众们，我们听到了你们的声音。
05:44
It's why we created the New York Times Games Family subscription.
这就是我们创建 New York Times Games Family 订阅的原因。
05:47
One subscription, up to four separate logins, and your existing stats and streaks come with you.
一份订阅，最多四个独立登录，你现有的统计数据和连续记录也会一并带过来。
05:52
Find out more at nytimes.com slash family.
了解更多信息，请访问 nytimes.com/family。
05:55
Jasmine Sun, welcome to the show.
Jasmine Sun，欢迎来到节目。
06:00
I'm so excited to be here.
能来这里我太激动了。
06:02
So you just got back from a reporting trip in Wisconsin, Michigan, coming to fight over data centers.
所以你刚从 Wisconsin 和 Michigan 的报道之旅回来，那里正为 data centers 争得不可开交。
06:08
Let's just start with what you see when you're in your data center.
那我们就先从你身处 data center 时看到的东西说起吧。
06:12
What does it look like?
它长什么样？
06:14
I think one of the most important things about rural Wisconsin and rural Michigan is how beautiful it is.
我觉得 Wisconsin 和 Michigan 的乡村地区最打动人的地方之一，就是它们有多美。
06:21
I felt like I was in Eden.
我感觉自己就像在 Eden。
06:23
I felt like I was in paradise.
我感觉自己就像在天堂。
06:24
It's incredibly lush, incredibly green.
那儿特别茂盛，特别绿。
06:26
And as you get closer to a data center, you start to see more power lines.
而当你越靠近一个 data center，你就开始看到更多电线。
06:31
You start to see more towers.
你会开始看到更多高塔。
06:33
And eventually you just see what looks like an extremely large flat warehouse.
到最后，你只会看到看起来像一座巨大、扁平的仓库一样的东西。
06:40
But you just see this sort of like verdant landscape give way to what is a windowless industrial park.
但你只会看到这种郁郁葱葱的景色，被一个没有窗户的工业园区取代。
06:48
Data centers are very ugly.
Data centers 真的很难看。
06:50
I think I didn't appreciate this until I started standing in front of them, getting near them, listening to them.
我觉得直到我开始站在它们面前、靠近它们、听它们的声音，我才真正体会到这一点。
06:55
People will time how long it takes to drive past a data center on the highway,
人们会掐表计时，看开车在高速公路上经过一个 data center 要多久，
07:00
going 70 miles per hour in Port Washington.
在 Port Washington，时速 70 英里。
07:03
I think it's like a minute and 42 seconds.
我觉得大概是 1 分 42 秒。
07:05
The size of these things is a long on the highway.
这些东西的体量，在高速公路上就是长长的一段。
07:08
It is a long, long ride.
这是一段很长、很长的车程。
07:10
These hyper-scale data centers are huge.
这些 hyper-scale data centers 非常巨大。
07:12
They are massive.
它们非常庞大。
07:13
And so I think that the aesthetic questions right around,
所以我觉得，那些审美问题恰恰围绕着，
07:17
is this what I want my state, my community to look like.
这是不是我想要的我的州、我的社区的样子。
07:21
These are really salient to people.
这些问题对人们来说真的非常突出。
07:23
You mentioned hearing them.
你提到过你听到了它们。
07:25
Yes.
是的。
07:26
What do they sound like?
它们听起来是什么样的？
07:27
Oh my gosh.
天哪。
07:28
I mean, they sound like humming, buzzing, worrying.
我是说，它们听起来像嗡嗡声、蜂鸣声、还有让人担忧的声音。
07:32
Every once in a while you'll hear a rattling.
时不时你会听到一阵咔嗒咔嗒的响声。
07:34
The residents who live next door to some of them say they're producing noise.
住在其中一些隔壁的居民说，它们在制造噪音。
07:37
That's not only annoying, but debilitating like this right here.
这不仅烦人，还会像这里这个一样，让人身心俱疲。
07:41
But again, they're windowless.
但话说回来，它们没有窗户。
07:47
They're not that many workers inside.
里面的工作人员并不多。
07:49
So they're very mechanical sounds.
所以那些声音非常机械。
07:51
They are inhuman as a lot of folks would say.
就像很多人会说的那样，它们不像人发出的声音。
07:54
So you spent a lot of time with people organizing against data centers.
所以你花了很多时间和那些组织起来反对 data centers 的人在一起。
07:58
Who were they?
他们是谁？
07:59
You had state home moms.
有州里的全职妈妈。
08:03
You had retired executives.
有退休的高管。
08:06
You had farmers who didn't like the impacts on their land.
有农民，他们不喜欢自己的土地受到影响。
08:10
You had activists, professional activists with environmental groups in the state.
有活动人士，州里环保团体的职业活动人士。
08:16
It was an interesting mix of people.
这是个挺有意思的人群组合。
08:18
But a lot of women, relatively more left-leaning.
但很多是女性，相对更偏左。
08:21
There definitely some right-leaning folks as well.
肯定也有一些偏右的人。
08:23
So one question I've heard people ask is,
所以，我听到有人问过的一个问题是，
08:26
how much is this different than other kinds of industrial installations?
这和其他类型的工业设施到底有多大不同？
08:32
I mean, there are a lot of things that are built all over the country.
我是说，全国各地建了很多东西。
08:36
You wouldn't necessarily want to be right next to.
你不一定想就挨着它。
08:39
Are data centers unusual in this?
data centers 在这件事上算不寻常吗？
08:41
Or are they, you know, from fracking to industrial agriculture?
还是说，它们，你知道，从 fracking 到 industrial agriculture 这一类？
08:46
Does like the latest version of it?
是不是就像它的最新版本？
08:48
It's a good question.
这问题问得好。
08:50
It's when I had and thought a lot about before I went.
这是我去之前就有、也反复想过很多的那件事。
08:53
And I've talked to city officials who were confused by this question of,
而且我跟一些市政官员聊过，他们对这个问题感到困惑：
08:57
we had a chip fab here.
我们这里曾有一个 chip fab。
08:58
We had an auto plant here.
我们这里曾有一个汽车厂。
09:00
We had a fulfillment center here, and nobody cared as much.
我们这里曾有一个 fulfillment center，但没人那么在意。
09:03
Why are data centers so much more unpopular than say solar farms,
为什么 data centers 会比，比如说，太阳能农场更不受欢迎这么多？
09:06
which also faced local opposition in places like Michigan?
而它们在像 Michigan 这样的地方也遭遇过当地反对？
09:10
And so I think while the quality of life concerns around this thing is loud
所以我觉得，虽然围绕这件事的生活质量担忧声音很大
09:15
and annoying and ugly and consumes resources are very similar to other industrial projects.
而且令人厌烦、丑陋、消耗资源，这些都和其他工业项目非常相似。
09:20
There must be some reason that opposition is so much more severe and widespread,
肯定有某种原因，让反对声要严重得多、也广泛得多，
09:26
even beyond the communities where the data centers are literally being built.
甚至超出了那些 data centers 实际正在建设的社区。
09:30
And I think that question has a lot to do with the AI, with the AI industry
而我觉得这个问题跟 AI 本身、跟 AI 行业有很大关系
09:34
and sort of the way that people feel about these companies and these projects.
以及人们对这些公司和这些项目的那种感受。
09:39
When I read your reporting on this, when I've talked to people involved in this,
当我读你关于这件事的报道时，当我和参与其中的人聊过之后，
09:42
it always feels to me there are sort of three layers of concerns that are converging
我总觉得，大概有三层担忧正在汇聚到一起
09:46
into what we call the data center backlash.
汇成我们所说的 data center backlash。
09:48
There's process.
有流程。
09:50
Then there's direct impacts, you know, the environment, water, electricity,
然后还有直接影响，你知道的，环境、水、电，
09:57
and then there's AI itself.
然后还有 AI 本身。
09:59
And maybe let's go through them sort of one by one.
也许咱们可以大概一个一个过一遍。
10:02
One thing that I have been hearing a lot of and I've seen in your reporting as well
有一件事我经常听到，也在你的报道里看到过，
10:07
is the anger over how these processes are going.
就是人们对这些流程推进方式的愤怒。
10:11
And in particular, the use of NDAs, which is not that common, right?
尤其是 NDA 的使用，这其实并不常见，对吧？
10:15
I've covered a lot of what does it take to build a housing development
我报道过很多关于建一个住宅开发项目需要什么的内容
10:18
and you don't tend to hear a lot of the city councilmen got put under an NDA.
而且你通常不太会听说很多市议员都被要求签了 NDA。
10:22
Right.
对。
10:23
So what is happening with these NDAs?
那这些 NDA 是怎么回事？
10:25
Yeah.
嗯。
10:26
I mean, this is also something that really surprised me.
我是说，这也真的让我很惊讶。
10:28
And I think the NDAs that...
而且我觉得那些 NDA……
10:30
I should say non-disclosure agreements.
我应该说是保密协议。
10:32
Right.
对。
10:33
The non-disclosure agreements.
那些保密协议。
10:34
This has really inflamed the amount of local opposition that you see.
这真的加剧了你所看到的当地反对情绪。
10:37
And so basically what would happen oftentimes is there would be some sense
所以基本上，经常会发生的情况是，会有一种感觉
10:41
starting in city council that maybe a big development project was going to show up.
一开始是在市议会里，觉得可能有个大型开发项目要出现。
10:45
But because of the NDAs, the council members would not be able to disclose
但因为这些 NDAs，市议员们没法披露
10:50
that necessarily it was a data center, necessarily who the customers were going to be a company like OpenAI
它到底是不是一个 data center，客户到底会是谁，比如像 OpenAI 这样的公司
10:56
or a company like Anthropic or whoever.
或者像 Anthropic 这样的公司，或者随便谁。
10:59
Or even the size of the project, like how much electricity is this actually going to consume.
或者甚至项目的规模，比如它到底会消耗多少电。
11:03
But whispers would start to get around.
但风声还是会开始传出去。
11:05
I was talking to a VP of a construction union and he was saying, you know,
我之前跟一个建筑工会的 VP 聊过，他说，你知道，
11:09
there's an old Irish saying the only way to keep a secret between three people is to kill two of them,
有一句古老的爱尔兰谚语：三个人之间想保守秘密，唯一的办法就是杀掉其中两个，
11:13
which I thought was hilarious.
我觉得这特别搞笑。
11:15
And so he's saying, when these developers show up, they talk to the general contractor,
所以他说，当这些开发商出现时，他们会跟总承包商谈，
11:18
the general contractor talks to all their subcontractors.
总承包商再跟所有分包商谈。
11:21
The subcontractors talk to all their workers.
分包商再跟所有工人谈。
11:23
Yes, maybe everyone is signing NDAs at every part of that process, but whispers get around.
是的，也许在这个过程的每个环节，每个人都在签 NDA，但风声还是会传出去。
11:28
And as soon as whispers get around, you start to get social media posts,
而一旦有风声传开，你就开始看到社交媒体上的帖子，
11:31
you start to get rumors.
就开始有各种传言。
11:33
And the city council, because they are beholden to these NDAs,
而市议会因为受这些 NDA 约束，
11:36
they lose the ability to get ahead of the social media narrative.
他们就失去了抢先引导社交媒体舆论的能力。
11:39
That was something I repeatedly heard from these local government officials was
这是我从这些地方政府官员那里反复听到的一点：
11:42
we could not get ahead of social media because we had signed an NDA and rumors started getting around.
我们没法抢在社交媒体前面，因为我们签了 NDA，而传言已经开始传开了。
11:47
Why do the companies want these NDAs signed?
为什么这些公司想让人签这些 NDA？
11:49
I mean, I think they didn't think about it.
我是说，我觉得他们根本没考虑过这件事。
11:52
They didn't realize there would be a backlash.
他们没意识到会引来反弹。
11:54
They just thought it would be easier in case they changed their mind.
他们只是觉得，万一以后改主意了，这样会更方便。
11:57
These companies signed lots of NDAs with their own workers with anyone who works with them.
这些公司跟自己的员工，以及所有跟他们合作的人，签了一大堆 NDA。
12:01
I don't think there's a good reason.
我觉得这没什么好理由。
12:03
Microsoft has actually decided to stop using NDAs because of the level of backlash.
Microsoft 其实已经决定停用 NDA，因为反弹太强烈了。
12:07
I've heard from people managing compute at some of the other AI labs
我从其他一些 AI labs 里负责 compute 的人那里听说，
12:11
that they are thinking of making the same decision.
他们也在考虑做出同样的决定。
12:13
One thing that surprised me is all of the pro data center,
让我意外的一点是，所有那些支持 data center 的，
12:17
pro build people who I spoke to, whether workers or AI developers,
我聊过的那些支持建设的人，不管是工人还是 AI developers，
12:21
they all regret the NDAs.
他们全都后悔签了 NDAs。
12:23
They all think that they made the situation much worse.
他们都觉得，自己把情况搞得更糟了。
12:25
What is the impact of a new data center on what are usage,
新建一个 data center 对用水量的影响是什么，
12:30
what are availability in the town?
对镇上的供水可用性又有什么影响？
12:32
They do require some of it, obviously, primarily for cooling the data centers
他们确实需要其中一些，很明显，主要是为了给 data centers 做 cooling，
12:37
because these chips and servers run really hot and they need AC.
因为这些 chips 和 servers 跑起来真的很热，而且它们需要 AC。
12:41
The thing that's gone a bit wrong in the water debate,
在水资源争论里，有点跑偏的地方是，
12:44
I think, is that today's new data centers are almost all closed-loop systems,
我觉得，如今新的 data centers 几乎都是 closed-loop systems，
12:48
closed-loop in the same way that air conditioning is closed-loop,
closed-loop，就像空调是 closed-loop 一样，
12:51
which means they're recycling the water within the system
也就是说他们在系统内部循环利用水
12:54
and they use a fraction of the water that golf courses use.
而且它们用的水只有高尔夫球场用水量的一小部分。
12:57
In fact, in places like Jamesville is constant,
其实，在像 Jamesville 这样的地方，是恒定的，
12:59
we would often see a literal golf course right next to the data center site.
我们经常会看到，data center 的选址旁边真的就有一个高尔夫球场。
13:03
They do some and it has a very sticky icon of these things resource consumption.
他们做了一些，而且它有一个很黏的 icon，就是这些东西的 resource consumption。
13:10
A lot of folks I talked to and was constant in Michigan,
我跟很多人聊过，在 Michigan 一直是这样，
13:13
they would say things like, you know, they're building right by the great lakes.
他们会说，你知道吧，他们就在 Great Lakes 边上建。
13:16
Why would they do that if they weren't trying to drain the lakes?
如果他们不是想把湖水抽干，干嘛要这么做？
13:18
Why would they do that if they didn't need all this fresh water?
如果他们不需要这么多淡水，干嘛要这么做？
13:21
And so your view is at this point, the technology has changed,
所以你的观点是，到了现在，技术已经变了，
13:23
such that water is not as big of a deal.
以至于水已经没那么重要了。
13:26
As maybe it actually was a couple of years ago.
不像它在一两年前可能真的还是个大问题。
13:29
I think the next thing people have heard a lot about is energy usage.
我觉得下一件大家经常听到的事就是 energy usage。
13:33
Yes. So walk me through that.
对。那你给我讲讲这个吧。
13:35
The electricity consumption issue is real.
耗电这个问题是真的。
13:37
So data centers do, in fact, use an incredible amount of electricity.
所以 data centers 实际上真的会用掉惊人数量的电。
13:41
These, these chips and these servers that are processing
这些，这些 chips 和这些 servers 正在处理
13:45
gigantic mathematical calculations to make AI work,
庞大的数学计算，好让 AI 运转起来，
13:49
require tons of energy.
需要巨量的能源。
13:51
All these chips and clusters are talking to each other.
所有这些 chips 和 clusters 都在互相通信。
13:53
You need the interconnections to be really fast.
你需要 interconnections 速度非常快。
13:55
In order to get like a chat you be to answer really quickly with low latency,
为了像 chat 那样，你得回答得又快又有 low latency，
13:59
you need these super fast connections, all that's powered by electricity.
你需要这些超高速的连接，而这一切都得靠电力驱动。
14:04
So we are talking about a really, really significant amount of new electricity
所以我们说的是，新增的电力会非常非常庞大，
14:08
that is going to require new generators, new power plants.
这会需要新的发电机、新的发电厂。
14:11
It's probably going to be natural gas in the near future.
短期内很可能得靠天然气。
14:13
Once the data center is fully operational,
等这个 data center 完全投入运营之后，
14:15
there is not a ton of air and water pollution,
不会有特别多的空气和水污染，
14:18
assuming that everything's working correctly.
前提是一切都正常运转。
14:20
They are relatively clean facilities.
它们算是相对清洁的设施。
14:22
But, you know, like a lot of folks are concerned about the electricity use.
但是，你知道，很多人都在担心用电的问题。
14:25
They're saying, yeah, maybe the data center doesn't use that much water.
他们说，对啊，也许 data center 用的水没那么多。
14:28
Maybe the data center doesn't pollute that much.
也许 data center 造成的污染也没那么大。
14:31
But what about all of these new power plants that they're going to build in order to power it?
但是，他们为了给它供电而要建的所有这些新发电厂呢？
14:36
When the tech companies come into these towns,
当这些科技公司进入这些城镇时，
14:38
and begin talking to the city council members,
开始和市议会议员交谈时，
14:40
when they begin talking to the community,
当他们开始跟社区沟通时，
14:42
what are they sort of promising on the one hand?
他们一方面到底是在承诺些什么？
14:46
This is why this will be good for you.
这就是为什么这对你会有好处。
14:48
And what are they asking for on the other?
那另一方又在要求什么呢？
14:51
Yeah, I mean, so these things do provide an incredible amount of tax revenue.
是啊，我的意思是，这些东西确实能带来相当可观的税收。
14:55
And so, in-mount pleasant,
所以，在 Mount Pleasant，
14:57
that on the old Foxconn site, which ended up being bought out by Microsoft,
就在旧的 Foxconn 厂址上，那块地最终被 Microsoft 买下了，
15:01
they saw, oh, there's all this infrastructure already here.
他们发现，哦，这里已经有这么多基础设施了。
15:03
Why don't we build a hyperscale data center on this unused, industrially-zoned land?
我们为什么不在这块闲置的、规划为工业用途的土地上建一个 hyperscale data center 呢？
15:07
Microsoft is on track to pay $19.6 million in property taxes in 2026.
Microsoft 有望在 2026 年缴纳 1960 万美元的房产税。
15:12
This is expected to continue for many years.
这预计还会持续很多年。
15:15
Again, this is a small village of 28,000 people.
再说一次，这是一个只有 28,000 人的小村庄。
15:19
And so, there are really meaningful property tax benefits.
所以，确实会有很实在的房产税优惠。
15:23
And I do think it, I find it frustrating personally,
而且我确实觉得，我个人对此感到很沮丧，
15:25
when anti-data center organizers say that the job creation is a myth,
当反对 data center 的组织者说创造就业只是个迷思时，
15:28
because I think that 500, maybe six figured jobs for folks who go through apprenticeships
因为我觉得，500 个，也许是六位数薪资的工作，是给那些参加学徒项目的人，
15:34
that maybe don't need college degrees that last two to six years,
而这些学徒项目可能不需要大学学位，会持续两到六年，
15:37
that's enough time to build a family, that's enough time to buy home.
那足够组建家庭，也足够买房了。
15:40
I think it's super meaningful.
我觉得这特别有意义。
15:41
And when I talk to technicians, when I talk to workers,
而且我跟技术人员聊的时候，跟工人聊的时候，
15:44
clearly it was extremely meaningful work.
很显然，那是极其有意义的工作。
15:46
I think one of the things that was the most surprising to me,
我觉得，对我来说最让我意外的事情之一是，
15:49
when I talked to data center activists, was that they responded to so many of the pro arguments
当我跟 data center 活动人士聊天时，他们对那么多支持论点的回应竟然是
15:57
with, I don't believe them.
“我不信他们。”
15:59
So, like, this company says that they are going to treat the water with chemicals in it
所以，就比如说，这家公司说他们会处理那些含有化学物质的水
16:05
so that it doesn't flow into the lakes.
这样它就不会流进湖里。
16:07
People would say, I don't believe them.
人们会说，我不信他们。
16:09
I don't think they can do it.
我觉得他们做不到。
16:10
The companies would say, we are going to cover the cost of our own electricity grid.
这些公司会说，我们会承担自己电网的成本。
16:16
So, you could build out.
这样，你们就可以扩建了。
16:17
We aren't going to ensure that rates do not go up for all Michiganders.
我们不会保证所有密歇根人的电价都不上涨。
16:21
A lot of folks would say, I don't believe them.
很多人会说，我不信他们。
16:24
DTE has raised our electricity rates basically every year since 2022.
DTE 自从 2022 年以来，基本上每年都在涨我们的电价。
16:28
Why would this be the year that they decide not to do it?
凭什么今年他们就会决定不这么做了呢？
16:31
I do not believe them.
我不相信他们。
16:32
They are going to, we are going to create 500 jobs.
他们会，我们会创造 500 个工作岗位。
16:35
And some of these will stick around after the data center is built.
而且其中一些会在 data center 建成后继续存在。
16:38
People just said, I don't believe them.
人们只是说，我不相信他们。
16:40
It was really clear to me that data centers are showing up in an environment
我很清楚地看到，data centers 正出现在一种环境里
16:44
of extremely low trust in both governments and corporations to the extent
这种环境对政府和企业都极度缺乏信任，以至于
16:48
where the pro arguments almost do not land because people just aren't interested
支持方的论点几乎完全落不了地，因为人们根本不关心
16:53
in anything an outside tech company is going to tell them.
任何外部科技公司要告诉他们的东西。
16:55
They say, they have big PR teams.
他们说，他们有庞大的 PR 团队。
16:57
They can say whatever they want.
他们想说什么就说什么。
16:59
One thing that became really obvious to me when I talked about people
有一件事对我来说变得非常明显，当我谈到
17:01
of the AI companies and local officials is that two years ago,
AI 公司的人和当地官员时，那就是两年前，
17:04
no one thought the data center backlash was going to be like this.
没人想到针对 data center 的反对声浪会变成这样。
17:07
And so, these companies were, in fact, looking for things like,
所以，这些公司其实一直在找类似这样的东西：
17:11
does the state have a sales tax exemption as Wisconsin does
这个州有没有像 Wisconsin 那样的销售税豁免
17:15
to ensure that they don't have to pay sales taxes on their very expensive chips
来确保他们不用为那些非常昂贵的 chips 交销售税
17:18
on their very expensive GPUs?
在他们非常昂贵的 GPU 上？
17:20
They were often looking to build in places that might even offer local subsidies
他们常常想在一些地方建设，这些地方甚至可能提供当地补贴
17:25
to the companies for building in that area.
给那些在那个地区建设的公司。
17:28
I think that a lot of the balance of power has shifted as local opposition has ramped up.
我觉得，随着当地反对声不断升级，很多权力平衡已经发生了转变。
17:34
Now, it's the case that I'm hearing from some local officials.
现在的情况是，我从一些当地官员那里听到，
17:37
If we did this again today, we wouldn't have to offer any subsidies
如果今天再这么做一次，我们根本不需要提供任何补贴
17:41
because in fact, now there is so much local opposition
因为事实上，现在当地的反对声太大了
17:45
that the developers are really looking,
以至于开发商真的在找，
17:47
where is there going to be a local community and a government
哪里会有一个本地社区和一个政府
17:50
that is friendly to our project?
对我们的项目友好呢？
17:52
How can we, you know, have a Christmas tree ornaments worth
我们怎么才能，你知道，拿到足够挂满一棵圣诞树的
17:56
of community benefits agreements?
社区福利协议呢？
17:58
And so just the nature of these deals and how much they are skewed
所以，这些交易的性质，以及它们有多偏向
18:01
towards communities versus the AI developers has really changed.
社区而不是 AI 开发者，真的已经变了。
18:05
One of the things that came up a lot in your reporting,
在你的报道里经常出现、我也觉得很有意思的一件事，
18:08
and that I thought was interesting, was the fear that this is a bubble.
就是人们担心这是一个泡沫。
18:11
For sure.
肯定的。
18:12
And what's going to happen is your community will agree to something.
而接下来会发生的是，你们当地社区会先同意某个方案。
18:16
And then in the middle, the bubble is going to pop
然后进行到一半的时候，泡沫就会破掉，
18:18
and you're going to end up with a half-inch data center
最后你得到的会是一个半英寸大的 data center，
18:22
or one that is not being kept up correctly
或者是一个没被妥善维护的 data center，
18:25
or something where the promised benefits don't emerge.
又或者是那种承诺的好处根本没出现的东西。
18:30
I mean, you were in Wisconsin, which had a very, very bad experience with Foxcon.
我是说，你当时在 Wisconsin，那里跟 Foxcon 有过非常非常糟糕的经历。
18:33
Yeah.
是啊。
18:34
Which seems to be structuring the way people are thinking about
这似乎正在塑造人们思考
18:36
at least some of this.
至少其中一部分事情的方式。
18:37
So talk to me a bit about that set of concerns.
所以跟我稍微聊聊那一系列担忧吧。
18:41
Absolutely.
当然。
18:42
Yeah.
嗯。
18:43
I think in terms of what people in these communities
我觉得，就这些社区里
18:46
with these data centers feel when they see the projects come in
拥有这些 data centers 的人而言，当他们看到这些项目
18:49
with their gigantic, like $2 billion of investment,
带着自己的巨额、比如 20 亿美元的投资进来时是什么感受，
18:52
like these gigantic numbers that are being dangled,
就像这些被摆出来的天文数字，
18:54
it feels like a bubble to them.
对他们来说，这感觉就像个泡沫。
18:56
One, they're seeing news articles saying maybe AI is a bubble.
第一，他们会看到新闻文章说 AI 可能是个泡沫。
18:59
We don't know it's a bubble, but it could be.
我们不知道它是不是泡沫，但它有可能是。
19:01
Two, you do have experiences like Foxcon,
第二，确实会有像 Foxcon 这样的经历，
19:04
where you have a big tech company show up in a very small community.
就是一家大科技公司突然出现在一个很小的社区里。
19:08
In this case, Mount Puzzle was in Wisconsin,
在这个例子里，Mount Puzzle 是在 Wisconsin，
19:10
a city of like 28,000 people.
一个大概有 28,000 人的城市。
19:12
It's not very big.
这不算很大。
19:13
Get hundreds of millions of dollars in infrastructure investment
拿到数亿美元的基础设施投资
19:16
and tax subsidy from the town.
以及来自镇上的税收补贴。
19:18
Promise 13,000, high paying manufacturing jobs.
承诺 13,000 个高薪制造业岗位。
19:21
And then pull out because the contract wasn't set up correctly.
然后就因为合同没拟好，他们就撤了。
19:24
They decided they didn't actually want to build a bunch of
他们决定，其实并不想在 Wisconsin 建一堆
19:27
flat screen TVs in Wisconsin.
平板电视。
19:29
And the, you know, the town is left on the hook,
然后，你知道，这个镇就只能自己扛着了，
19:32
having invested all of this money in the grid and in roads.
在电网和道路上投了这么多钱之后，
19:35
They got, I think, a thousand jobs in the end.
我觉得，他们最后拿到了一千个工作岗位。
19:37
Foxconn is still paying back all of this debt that has been accumulated.
Foxconn 还在还所有这些累积下来的债务。
19:41
And so, experiences like that have really, really soured people
所以，像那样的经历真的真的让人们对
19:44
on the question of when and outside big tech company comes in
一家外来的 big tech 公司什么时候进来
19:47
and, you know, promises these gigantic numbers
然后，你知道，承诺这些天文数字
19:50
and all of these jobs and all of this tax revenue.
还有所有这些工作岗位和所有这些税收这件事。
19:52
And it's for this technology that like you can't,
而且这还是为了这种技术，你根本没法，
19:54
a lot of people they don't see, they don't feel,
很多人，他们看不到，也感受不到，
19:56
they don't find personally extremely, extremely useful,
他们自己并不觉得它极其、极其有用，
19:58
not at the levels of these valuations.
配不上这些 valuations 的水平。
20:00
They have a lot of questions about,
他们有很多疑问，
20:02
if the bubble pops are we going to be the ones left
如果泡沫破了，会不会最后是我们
20:04
with a stranded asset in our community?
在社区里留下一个 stranded asset？
20:06
I mean, in Jainsville, Wisconsin,
我是说，在 Jainsville, Wisconsin，
20:08
it was famously the site of this 100 year old GM plant
那里曾以一座有 100 年历史的 GM 工厂而闻名
20:11
that was the centerpiece of the community that employed a ton of people
那曾是整个社区的核心，雇了一大堆人
20:15
when they left during the financial crash in 2008 and the plant closed down.
2008 年金融危机期间，他们撤走了，工厂也关门了。
20:19
Not only did it sort of devastate the community from a work perspective,
这不仅从就业角度算是重创了整个社区，
20:23
but they also left $30 million of contamination
而且还留下了 3000 万美元的污染，
20:27
and hazardous waste in the middle of the city that has never been cleaned up.
和 hazardous waste，就在市中心，而且从来没被清理过。
20:31
There's forever chemicals in there.
那里面还有 forever chemicals。
20:33
This is why developers have not been able to sell this brownfield
这就是为什么开发商一直没法卖掉这块 brownfield，
20:36
is because there's so much waste that GM never cleaned up.
是因为有太多 GM 从来没清理掉的废物。
20:40
And so, I think that people worry about what happens
所以，我觉得人们担心的是会发生什么
20:42
if the AI bubble pops, if maybe it doesn't pop
如果 AI bubble 破了，又或者它没破
20:45
and the data center developers just decide,
而 data center 开发商直接决定，
20:48
never mind, we want to build elsewhere.
算了，我们想去别的地方建。
20:50
Never mind, the data center isn't good enough.
算了，这个 data center 不够好。
20:52
It's better technology and who's going to be left holding the bag?
这是更好的技术，那最后谁来接盘？
20:55
That was the question I heard over and over again.
这是我一遍又一遍听到的问题。
20:57
But this is something that I do think is in people's minds.
但这件事，我确实觉得，是大家心里在想的。
20:59
Yeah.
嗯。
21:00
You bring this in and right now you're at this time of very, very high valuations.
你把这个引进来，而现在你正处在估值非常非常高的时候。
21:03
Right.
对。
21:04
And if AI demand isn't quite what you think,
而且如果 AI 需求并不完全像你想的那样，
21:06
or even just the company that was behind this particular data center
或者哪怕只是这个特定 data center 背后的公司
21:10
is not part of the winter circle,
不属于 winter circle，
21:12
then in a couple of years what you got,
那么过几年，你得到的是什么，
21:14
this is a giant box that's not going to continue being valuable.
这就是一个巨大的盒子，不会继续有价值。
21:21
And so, whatever the promise benefits are from it,
所以，不管它承诺能带来什么好处，
21:23
you know, tax revenue, et cetera.
你知道，税收收入，等等。
21:25
But yeah, maybe they show up for a while.
不过是啊，也许他们会来待一阵子。
21:26
Yeah.
是啊。
21:27
But then what if in five years they're gone,
可然后呢，万一五年后他们走了，
21:29
but you are left with this infrastructure?
却只给你留下这套 infrastructure？
21:31
It's like they can leave Jane's Hill with no real concern.
就像他们可以毫无顾虑地离开 Jane's Hill。
21:34
They're not there.
他们不在那儿。
21:35
They're people don't live there.
他们是不住在那里的人。
21:36
Right.
对。
21:37
But if you're in Jane's Hill, you do live there.
但如果你在 Jane's Hill，那你确实住在那儿。
21:38
Yes.
对。
21:39
Yeah.
是啊。
21:40
And it's just a real concern.
这真的是一个很现实的担忧。
21:41
You see things like this with Elon's giant colossus data centers
你会看到 Elon 那些巨型 colossus data centers 也有这种情况。
21:44
that he built out in Memphis.
就是他在 Memphis 建起来的那些。
21:46
And what makes people feel better about a new construction project
那是什么让人们对镇上的新建设项目感觉好一点，
21:49
in their town of calling it colossus?
就是把它叫 colossus？
21:51
Oh, yes.
哦，对。
21:52
Yeah.
是啊。
21:53
I mean, these things.
我是说，这些东西。
21:54
I'm airing touch for those people.
我对那些人还真是心太软。
21:56
I mean, usually the thing that happens is they give them very cute sea names,
我是说，通常会发生的是，他们会给它们起一些很可爱的 C names，
22:00
like Project Canoli and the barn.
比如 Project Canoli 和 the barn。
22:02
And they try to make them sound as friendly as possible.
而且他们还尽量让它们听起来尽可能友好。
22:04
But yeah, I mean with colossus,
但是啊，我是说，靠着 colossus，
22:06
XII never really took off.
XII 一直也没真正火起来。
22:08
People were not in fact using Groc as much as Elon thought.
事实上，大家用 Groc 的频率并没有 Elon 想的那么高。
22:10
They were going to be using it.
他们本来是会用的。
22:12
And in that case he was able to get a really good deal
而这样一来，他得以谈成一笔非常划算的交易，
22:15
selling the compute capacity to Anthropic,
把 compute capacity 卖给 Anthropic，
22:17
which was growing like crazy and had not built enough data centers on their side.
因为 Anthropic 当时增长得飞快，自己这边又没建够 data centers。
22:21
But you could totally imagine a world, as you say,
但你可以完全想象这样一个世界，就像你说的，
22:23
where in fact, you know, XII decides we are going to focus on space.
在那里，事实上，你知道，XII 决定我们要专注于太空。
22:28
We don't care about AI anymore.
我们不再关心 AI 了。
22:30
Maybe there's none in Anthropic to pick up the bill because Anthropic has maybe
也许 Anthropic 里没人会来买单，因为 Anthropic 可能
22:34
built enough of their own compute capacity.
已经建好了足够多自己的 compute capacity。
22:36
There is an open question about what happens in that world.
关于那个世界会发生什么，还有一个悬而未决的问题。
22:38
So our abundance last year with the director of public.
所以，去年我们的 abundance 是和 director of public 一起。
22:42
And one thing I've heard been asked by a lot of people
还有一件事，我听到很多人在问
22:46
is like, what does the abundance take on a data center?
就像，abundance 对一个 data center 的看法是什么？
22:49
And the beginning of that book has this line that the question is,
而那本书的开头有这样一句，问题是，
22:55
what do we need more of and how do we get it?
我们需要更多什么，又该怎么得到？
22:58
And I think the question here that has been so hard for the AI companies,
而我认为，这里的问题对 AI companies 来说一直很难，
23:04
for the people trying to build data centers,
对于试图建 data centers 的人来说，
23:07
is actually getting people to believe they need more of them.
其实是让人们相信他们需要更多 data centers。
23:13
When you're talking about building affordable housing,
当你在谈论建可负担住房时，
23:15
when you're talking about building an array of solar panels or wind turbines,
当你在谈论建太阳能电池板阵列或风力发电机时，
23:21
there's a pretty legible argument for why you need that.
有一个相当清楚的论点，说明你为什么需要它。
23:25
People still may not like it.
人们可能还是不喜欢它。
23:27
But we need homes because we need places for people to live.
但我们需要住房，因为我们需要让人们有地方住。
23:31
We need solar panels because we need clean renewable energy.
我们需要太阳能板，因为我们需要清洁的可再生能源。
23:35
How much is this like a normal kind of,
这在多大程度上像一种正常的，
23:38
I don't even exactly want to call it an imbiasm,
我甚至都不太想把它叫作 imbiasm，
23:40
but a normal kind of,
而是一种正常的，
23:43
I don't want the industrial infrastructure built in my backyard
我不想让工业基础设施建在我家后院。
23:47
because what am I going to get out of that?
因为我能从中得到什么好处呢？
23:50
And how much of it is actually something that is more related
而其中又有多少，其实是更关乎
23:53
to people's feelings about AI,
人们对 AI 的感受，
23:56
which is I don't want this built here
也就是：我不想让这个东西建在这里，
23:59
because why would I want to pay the cost for a thing
因为我为什么要为一个东西承担成本，
24:02
that I don't want there to even be more of in the first place?
而这个东西我从一开始就不希望它变得更多呢？
24:07
This I think was one of my big motivating questions going into this trip is,
我觉得，这是我这次出行前最大的动机性问题之一，
24:10
is it, quote, unquote, normal nimbism?
那就是，这算不算是，所谓的，normal nimbism？
24:12
Is it about AI?
这是关于 AI 的吗？
24:13
Is it about something else?
还是关于别的事情？
24:15
And so I spent time both looking at polls
所以我花了一些时间，一边看民调
24:18
and trying to talk to people about,
一边试着跟人们聊，
24:19
would you be excited if this was a chip factory,
如果这里要建一个芯片工厂，你会不会很兴奋？
24:22
which would use a lot more water and pollute a lot more?
那会用掉多得多的水，也会造成多得多污染？
24:25
Would you be excited if it was a solar farm?
如果是个太阳能农场，你会不会很兴奋？
24:27
Also maybe acquiring agricultural land
还有，也许是在收购农业用地
24:29
and turning it into industrial use?
然后把它变成工业用途？
24:31
Would you be excited if it was a million other things?
如果换成其他一百万种东西，你还会兴奋吗？
24:33
And I talked to Nick Bagley,
我还跟 Nick Bagley 聊过，
24:35
who you've had on your show about,
你之前在节目里请他聊过这个的，
24:37
is this just proceduralism?
这难道只是 proceduralism 吗？
24:39
And we talked about the solar farms example,
我们还聊到了太阳能农场的那个例子，
24:42
where solar farms,
在太阳能农场那里，
24:44
the opposition use very similar tactics
反对者用了非常相似的策略
24:46
to the data center opposition.
对于 data center 的反对。
24:47
They were organizing in Facebook groups.
他们在 Facebook 群组里组织起来。
24:48
They were packing town halls.
他们把市政厅会议挤得水泄不通。
24:50
They were talking about the local impacts
他们在谈论对当地的影响
24:52
and the importance of farmland
以及农田的重要性
24:54
and the visions for their communities.
以及他们对自己社区的愿景。
24:56
There were zoning fights, of course.
当然，也有 zoning 之争。
24:58
But like you say, with the solar farms,
但就像你说的，有了 solar farms，
25:00
you do have a very clear pro case.
你确实有一个非常明确的pro case。
25:03
You have a faction.
你有一个faction。
25:04
You have a group of people, a constituency,
你有一群人，一个constituency，
25:06
people who care about the environment,
关心环境的人，
25:08
who want renewable energy,
想要renewable energy的人，
25:09
who understand that that's a thing that,
他们明白那件事，
25:11
yeah, maybe it sucks to have in your backyard,
是的，也许把它建在你家后院很糟，
25:13
but you can take one for the team,
但你可以为了团队牺牲一下，
25:14
because this is important for our planet.
因为这对我们的地球很重要。
25:16
You don't really have a pro-faction with AI,
在 AI 这件事上，你其实并没有一个真正的支持阵营，
25:19
same with like the auto plant, right?
汽车工厂那边也是一样，对吧？
25:21
You have one, you have maybe 7,000 workers
你有一个，你大概有 7,000 名工人
25:23
in the old GM plant in Jamesville,
在 Jamesville 的老 GM 工厂里，
25:25
who all have families who really care about them,
他们都有真正关心他们的家人，
25:27
who see that as a constituency.
他们把这当成一个票仓。
25:29
Everyone drives a car.
每个人都开车。
25:30
They see their car as essential.
他们把车视为必需品。
25:31
I think the fact that it's creating these tangible outputs
我觉得它正在创造这些实实在在的产出，这件事
25:34
really, really matter.
真的、真的重要。
25:36
I don't think that data centers
我不认为 data centers
25:38
have a compelling pro-consituency
有什么强有力的支持群体
25:42
besides the utility companies
除了公用事业公司
25:44
and the AI companies,
和 AI 公司，
25:45
which are already incredibly, incredibly unpopular.
而它们已经极其、极其不受欢迎了。
25:47
And I went in and asked these organizers,
然后我进去问这些组织者，
25:50
do you guys use AI?
你们用 AI 吗？
25:51
I do find it useful.
我确实觉得它挺有用的。
25:52
This is one of the top questions
这是最重要的问题之一，
25:53
that my friends in San Francisco wanted me to ask,
我在 San Francisco 的朋友们想让我问的，
25:55
is are these people like using chat GBT
就是，这些人是不是在用 chat GBT，
25:58
and they don't even realize that the data centers
却没意识到，data centers
26:00
are how they can use it?
才是他们能用上它的原因？
26:01
And what I found there was a lot of these folks did say,
而我在那里发现的是，这些人里有很多确实会说，
26:05
yeah, I've used it to draft an email or make a meme.
对，我用它起草过邮件，或者做过梗图。
26:08
They're not denying that AI might have
他们并不否认 AI 可能
26:10
any possible utility at all.
有任何一点用处。
26:13
But they clearly didn't see it as essential
但他们显然不认为它是必不可少的，
26:15
in the way that cars, energy, and housing are essential.
不像汽车、能源和住房那样必不可少。
26:18
They clearly saw as kind of like a widget, a toy.
他们显然把它看成某种 widget，一个玩具。
26:22
And maybe there are these risks,
而且也许存在这些风险，
26:24
maybe there's the job stuff,
也许是有工作那方面的事儿，
26:25
but fundamentally they were like,
但根本上，他们就觉得，
26:26
this thing is not that useful.
这东西没那么有用。
26:27
I don't really see in my personal life
在我个人生活里，我真的看不出
26:29
how this could justify these gigantic valuations.
这怎么就能让这些巨大的估值站得住脚。
26:32
And so I was talking to, for example,
所以，比如说，我当时在和
26:34
Charles Franklin who runs Marquette Law Polls in Wisconsin
Charles Franklin 聊，他在 Wisconsin 负责 Marquette Law Polls
26:37
and he was explaining that usually
他当时解释说，通常
26:39
you see 50-50 polling on issues
你会看到在一些议题上，民调是 50-50。
26:41
where you have a strong anti-case and a strong pro-case.
也就是反对的一方理由很强，支持的一方理由也很强。
26:43
And the only reason you're kind of seeing
而你之所以多少会看到
26:45
this 70-30 bipartisan opposition to data centers
这种对 data centers 70-30 的两党共同反对
26:47
no matter whether you live near a data center
不管你是住在 data center 附近
26:49
or you don't, which means it's not just nimbism.
还是不住在附近，这意味着这不仅仅是 nimbism。
26:51
You don't want a data center in anyone's backyard.
你不想让 data center 建在任何人的后院。
26:53
It's because he was like,
是因为他当时就像在说，
26:55
there is no strong pro-argument forage.
没有强有力的支持 AGI 的论点。
26:57
There is not even a fight that's really going on.
甚至都没有一场真正在进行中的争论。
27:00
Nobody wants it.
没人想要它。
27:01
It was a phrase I heard over and over.
这是我一遍又一遍听到的一句话。
27:02
You have a very influential definition
你有一个非常有影响力的定义
27:04
of AI populism where you call it a world view
关于 AI populism，你把它称为一种世界观
27:07
in which AI is viewed,
在其中，AI 被视为，
27:08
not only as a normal technology,
不仅仅是一种普通技术，
27:11
but as an elite political project to be resisted.
而是把它当成一个应当抵制的精英政治项目。
27:15
unpack that for me.
给我展开说说。
27:17
The phrase that you hear a lot from AI critics
你经常从 AI 批评者那里听到的一句话
27:20
is why is it being shoved down our throats?
就是：为什么非要把它硬塞给我们？
27:23
Or with, you know, Chatchy BT,
或者拿，你知道，Chatchy BT 来说，
27:26
it's not that people are saying
并不是说人们在说
27:28
there is literally no use for Chatchy BT.
Chatchy BT 真的完全没用。
27:30
It's people are saying,
而是人们在说，
27:31
why are you forcing me at my job to use AI
为什么你在工作中非要逼我用 AI
27:34
to do something worse when I could do it better?
去做一件更差的事，而我明明能做得更好？
27:37
And so, I think that a lot of the public backlash to AI
所以我觉得，很多公众对 AI 的反弹
27:41
that has risen over the past six months
在过去六个月里不断升温
27:43
is not explained by people thinking
并不是因为人们认为
27:46
that the technology has no use at all.
这项技术完全没用。
27:48
It's not explained by them being worried
也不是因为他们担心
27:51
about specific technical properties of LLM's
LLM 的具体技术特性。
27:53
that might lead to rogue AI or misalignment or whatever,
这可能会导致 rogue AI 或 misalignment 之类的，
27:56
which are the sort of safety as arguments.
这些差不多就是那种 safety 论证。
27:58
It's AI as sort of an avatar
这是把 AI 当成某种 avatar，
28:00
for a small group of Silicon Valley billionaires
代表一小群 Silicon Valley 亿万富翁
28:03
ability to impose their vision of the world
拥有把他们那套世界观强加给
28:07
onto everybody else without their consent.
其他所有人、而且不经他们同意的那种能力。
28:10
And I think that's also what I hear echoed
而且我觉得，我在
28:12
in these data center debates.
这些 data center 争论里，也听到了同样的回响。
28:14
It's not just,
这不只是，
28:15
it's going to use this much water, that much water.
而是它要用这么多水、那么多水。
28:17
I frankly think that even if there was no misinformation
坦白说，我觉得就算没有不实信息
28:20
about water use,
关于用水这件事，
28:21
people would be just as angry about the data centers.
人们也会对 data centers 一样愤怒。
28:23
Yeah, I consider the water.
是啊，我考虑水的问题。
28:24
I don't want to say that the water issue is fake.
我不想说水的问题就是假的。
28:27
What I will say is that there was a debate online
我要说的是，网上有过一场争论
28:30
while back about how much water,
前阵子聊到，有多少水，
28:33
like a Chatchy BT query consumed.
比如一次 Chatchy BT query 会消耗掉。
28:36
And somebody was like,
然后有人说，
28:37
if you really care about water,
如果你真的在乎水，
28:39
are you eating beef?
你吃牛肉吗？
28:40
Yes.
吃。
28:41
As somebody who doesn't eat meat,
作为一个不吃肉的人，
28:43
I thought this was quite good.
我觉得这挺好的。
28:44
You could really save a lot of water by going vegetarian.
你吃素的话，真的能省很多水。
28:47
Yes.
是的。
28:48
And relatively much, much more than not using Chatchy BT
而且相对来说，这比不用 Chatchy BT 省的水要多得多。
28:51
and relatively few people in that conversation
而在那场对话里，相对很少有人
28:53
were giving up meat.
放弃吃肉。
28:55
Or giving up YouTube videos.
或者放弃看 YouTube 视频。
28:56
Or giving up YouTube videos.
或者放弃看 YouTube 视频。
28:57
Like use more water than a Chatchy BT query.
比如，比一次 Chatchy BT 查询用的水还多。
28:59
Which is to say that I think sometimes people
也就是说，我觉得有时候人们
29:01
don't like a thing.
就是不喜欢某个东西。
29:02
Yes.
对。
29:03
And they're looking for reasons to justify that dislike.
然后他们就在找理由，来给这种不喜欢找正当理由。
29:06
Yeah.
嗯。
29:07
But what's actually happening at the base
但归根结底，实际发生的是
29:09
is they don't like the thing.
他们就是不喜欢这个东西。
29:10
Right.
对。
29:11
And the data centers,
还有那些 data centers，
29:12
like as you're saying,
就像你说的，
29:13
I think,
我觉得，
29:14
speak to this AI populism question,
更能说明这个 AI populism 的问题，
29:16
like even more precisely.
甚至更准确地说。
29:17
Because the issue with the AI itself
因为 AI 本身的问题
29:19
is that I think people's relationship
在于，我觉得人们跟它的关系
29:22
to it is very complicated.
是非常复杂的。
29:23
I have myself
我自己
29:25
a very complicated relationship to AI.
跟 AI 的关系非常复杂。
29:27
Like I use it a fair amount.
怎么说呢，我其实用得还挺多的。
29:29
I'm not sure I think it's a good thing for society,
我也不确定自己是不是觉得这对社会是件好事，
29:31
the way it is going.
照它现在这个发展势头来看。
29:32
I don't want my kids using it.
我不想让我的孩子用它。
29:34
I know they'll be using it.
我知道他们肯定会用。
29:35
Like maybe it'll make things better,
怎么说呢，也许它会让事情变得更好，
29:37
but I really don't know.
但我真的不知道。
29:38
Like I think that the costs are going to be very,
我是说，我觉得这些成本会非常，
29:41
very high for us,
非常高昂，对我们来说，
29:42
relationally and economically.
在关系上和经济上。
29:44
And so I'm very conflicted.
所以我很纠结。
29:47
But do I want to live next to a data center?
但我真的想住在一个 data center 旁边吗？
29:50
Yeah.
想。
29:51
No.
不想。
29:52
It's totally different.
完全不一样。
29:53
Like that's easier.
就像，那样更容易。
29:54
I mean, one of the most interesting things.
我是说，最有趣的事情之一。
29:55
Somebody is just making you do that.
只是有人让你这么做而已。
29:56
Yes.
是的。
29:57
Of like going to this,
就像是去这个，
29:58
back to back,
一场接一场，
29:59
I went to this Abdul-Bernie AOC rally
我去了这个 Abdul-Bernie AOC 集会
30:01
in Lansing, Michigan.
在Lansing, Michigan。
30:02
And then I went to saw the Selene activists the next day.
然后第二天我去见了Selene的活动人士。
30:04
And I was researching
我当时在研究
30:05
how the Selene Stargate project happened.
Selene Stargate project是怎么发生的。
30:07
And it was really interesting to see these echoes.
而且看到这些回声真的很有意思。
30:10
Of the populist message,
那种民粹主义信息的，
30:11
sort of manifest in the specific project.
某种程度上在这个具体项目里显现出来。
30:13
Like when I'm at this rally,
就像我在这个集会上的时候，
30:15
people are talking about the oligarchy.
人们在谈论寡头集团。
30:16
They're talking about corporate billionaires,
他们在谈论企业界的亿万富翁，
30:18
whether it's big tech,
不管是 big tech，
30:19
or big pharma,
还是 big pharma，
30:20
or DT, the utility companies,
或者是 DT，那些公用事业公司，
30:22
paying off politicians
收买政客
30:24
in order to, you know,
就是为了，你知道，
30:26
screw the people over.
把老百姓坑惨。
30:27
And that's why you need the people to come together
所以这就是为什么你得让民众团结起来
30:31
and to get money out of politics.
还要把金钱从政治里赶出去。
30:33
To prevent DT from donating to these super PACs
防止 DT 给这些 super PACs 捐款
30:37
and paying off Gretchen Whitmer or whatever.
以及收买 Gretchen Whitmer 之类的。
30:39
And then when I learned how the Selene data center
后来当我了解到 Selene data center
30:43
saga played out,
这件事是怎么发展的，
30:44
what happened was the Selene Township City Council,
实际情况是，Selene Township City Council，
30:46
unlike a lot of city councils,
和很多市议会不一样，
30:47
actually voted for one against rezoning their land
实际上投票支持了一个反对重新规划他们土地的人
30:50
for the data center.
用来建 data center。
30:51
So this was the case when local government said
所以情况是这样的，当地政府说
30:53
this is not a vision for our community,
这不是我们社区的愿景，
30:55
it's not worth it to us.
这对我们来说不值得。
30:56
And what happened,
然后发生了什么，
30:57
the data center developers sued Selene Township,
data center 开发商起诉了 Selene Township，
31:00
a town of like,
一个大概……的小镇，
31:01
again, a few thousand people,
又一次，几千人，
31:04
into saying,
被说服说，
31:05
wait, no, this is exclusionary zoning,
等等，不对，这是 exclusionary zoning，
31:07
you can't have no industrial use in your entire township.
你不能整个镇区里都没有 industrial use。
31:10
And when a town of that size is getting sued
而当一个那么大的镇被告上法庭，
31:13
by a giant AI data center developer,
被一家巨大的 AI data center 开发商起诉时，
31:15
they just settled.
他们就直接和解了。
31:16
They were just like,
他们就像是，
31:17
fine, give us a few million for the fire department
好吧，给我们几百万给消防部门
31:19
and for some schools.
再给一些学校。
31:20
And this fight is not worth it to us.
而且这场斗争对我们来说不值得。
31:21
But that to people felt like a profound,
但这对人们来说感觉像是一种深刻的、
31:23
a profound violation of little DT democracy,
一种对小小 DT 民主的深刻侵犯，
31:26
it felt like the dark money and politics story,
它感觉像是 dark money 和政治的那种故事，
31:28
which is, you know,
也就是，你知道，
31:29
you have some very rich companies show up
有些非常有钱的公司冒了出来
31:32
with the bag of money to your politician,
拿着那袋钱去找你的政客，
31:34
they don't tell anybody else what's happening,
他们不会告诉其他任何人正在发生什么，
31:36
the politicians aren't allowed to tell their citizens
政客们也不被允许告诉他们的公民
31:38
and involve them in the decision making process.
并让他们参与决策过程。
31:40
And they themselves work out a deal,
而他们自己则谈成一笔交易，
31:42
a deal that is fundamentally asymmetric
一笔从根本上就不对等的交易
31:44
because of the amount of money on one side,
因为其中一方有那么多钱，
31:46
that will then transform the image of your community,
而这会进而改变你们社区的形象，
31:48
your lived reality into the world
把你亲身经历的现实带入这个世界
31:50
that these tech companies have decided for you.
这个由这些科技公司替你决定的世界。
31:53
And so I think that the data centers in that sense
所以我觉得，从这个意义上说，data centers
31:55
are a very visceral microcosm of the way
是一种非常切身的缩影，反映出这种方式——
31:58
that a lot of people feel the AI is sharing up in their lives.
很多人觉得 AI 正在他们的生活中出现。
32:01
I would also maybe even take that a little bit further.
我甚至可能还会把这一点再推进一步。
32:04
I think that the way that not all of the AI companies
我觉得，并不是所有 AI 公司都这样，
32:08
and I think anthropic has largely been a good actor here,
而且我觉得，anthropic 在这件事上大体上一直是个正面角色，
32:11
but many of them have acted,
但他们中的许多人已经行动了，
32:13
has opened up such a chasm between what they say
这已经拉开了一道如此巨大的鸿沟，横亘在他们说的
32:18
and then how they act under pressure
和他们在压力下实际怎么做的之间，
32:21
that one should be incredibly, incredibly skeptical of them.
以至于人们应该对他们抱有极其、极其强烈的怀疑。
32:24
And what I mean by this is that, you know,
我这么说的意思是，你知道，
32:27
Sam Altman and all these different people,
Sam Altman 以及所有这些不同的人，
32:29
you know, in front of congressional testimony
你知道，在国会听证
32:31
and interviews will say,
和采访中会说，
32:33
you know, it should not just be us making these decisions.
你知道，这不该只由我们来做这些决定。
32:36
There should be a real deep, small,
这里应该有一个真正深入的、小范围的、
32:39
deep democratic role here in how AI rolls out
深入的民主角色，参与到 AI 如何铺开，
32:43
in, you know, what effects it has on communities
在，你知道，它对社区有什么影响这方面，
32:47
and how it has governed.
以及它是如何被治理的。
32:49
And then when a community or a politician,
然后，当一个社区或一个政客，
32:54
you know, who is representing a community,
你知道，一个代表某个社区的人，
32:57
tries to say, well, we don't want this down here
试着说，好吧，我们不想要这个东西落到这里
33:01
or we want to impose these regulations,
或者我们想把这些监管规定强加下去，
33:04
we have watched repeatedly these companies
我们一次次看到这些公司
33:07
turn tremendous amounts of financial artillery
把巨额资金火力
33:10
against whoever is standing in their way, right?
对准任何挡他们路的人，对吧？
33:13
And to use the expertise and the money
然后用他们的专业能力和金钱
33:17
and the power they are amassing
以及他们不断积聚的权力
33:19
to kind of short-circuit that democratic voice.
来让那种民主声音短路。
33:21
Yeah. I mean, a couple of things.
是啊。我是说，有几件事。
33:23
One is like, I think one big gap I noticed
其中一个就是，我觉得我注意到的一个很大的鸿沟
33:25
between Silicon Valley and the folks
是在 Silicon Valley 和那些人之间——
33:27
in these communities I was talking to is
也就是我当时聊过的这些社区里的人——这个鸿沟就是
33:29
Silicon Valley does tend to think that money
Silicon Valley 确实往往觉得，钱
33:31
solves all problems.
能解决所有问题。
33:32
That if you just make the check bigger,
就是说，只要你把支票金额开得更大，
33:33
everything's going to be okay.
一切就都会好起来。
33:35
And I think people have a sense for
而且我觉得人们能感觉到
33:37
I'm being bribed, this corporation
我被贿赂了，这家公司
33:39
is not offering me a free lunch or whatever.
并不是在请我吃免费午餐之类的。
33:42
There is going to be something that I'm losing here.
我在这里肯定要失去点什么。
33:44
And in fact, sometimes the fact that the data center deals
而事实上，有时候 data center 交易
33:46
were bigger or the amount of political spending was bigger
规模更大，或者政治支出的金额更大
33:48
actually just makes people more suspicious
反而只会让人们更怀疑
33:50
like in the Abdul race.
就像在 Abdul 那场选战里。
33:52
His number one hit on Haley Stevens
他对 Haley Stevens 的头号攻击。
33:55
is how much money she is getting from APAC
就是她从 APAC 拿多少钱
33:57
from DTE from Pharma or whatever.
从 DTE、从 Pharma 或者什么的。
33:59
And so one is just that.
所以其中一个就是这一点。
34:01
I think we're in a political environment
我觉得我们现在处在一个政治环境里
34:02
where making the numbers bigger on the amounts of money bigger
在这种环境里，把金额上的数字弄得更大
34:04
or makes people more suspicious, not less.
或者让人更怀疑，而不是更不怀疑。
34:06
Another one I'll just quickly mention is
另一个我就快速提一下，就是
34:08
I don't even think Anthropics should be left off the hook
我甚至觉得 Anthropics 不该被放过
34:10
for things like the labor market impacts, right?
对于像劳动力市场影响这类事情，对吧？
34:12
They are the ones simultaneously warning
他们正是那些同时在发出警告的人
34:14
that we might see a 50% of white color jobs
说我们可能会看到 50% 的白领工作
34:18
lost by 2030.
到 2030 年消失。
34:19
This is really important to us.
这对我们来说真的非常重要。
34:21
We're breaking out about it.
我们正在为此炸锅。
34:22
Dario has written in his essays.
Dario 在他的文章里写过。
34:24
We might see an underclass of people
我们可能会看到一个底层群体。
34:26
of lower intellectual ability.
智力水平较低的。
34:28
And they are building the agents.
而且他们正在构建 agents。
34:30
They are building the coding agents,
他们正在构建 coding agents，
34:32
the banking agents, the design agents
banking agents、design agents
34:34
that they know are going to displace jobs.
他们知道这些将会取代工作岗位。
34:36
Or they believe at least are going to displace jobs.
或者他们至少相信这些将会取代工作岗位。
34:38
And I think that people feel that hypocrisy as well,
而且我觉得人们也感受到了那种虚伪，
34:40
which is if you are so worried about the inequality,
也就是说，如果你如此担心不平等，
34:43
why are you building the agents to do it?
你们为什么要构建 agents 来做这件事？
34:45
And when I ask executives and researchers
而当我问高管和研究人员
34:48
and whoever had an anthropic this question,
以及不管是谁在 Anthropic，问这个问题时，
34:50
they don't really have a good answer
他们其实并没有一个很好的答案
34:51
because it is true that their business model
因为他们的商业模式确实
34:53
is fundamentally premise on the disruption
从根本上就是建立在那个颠覆之上
34:56
that they say they are causing.
也就是他们说自己正在造成的。
35:21
AWSAI is how the world's leading organizations
AWSAI 就是世界领先组织如何
35:24
are transforming their industries,
正在改变他们所在的行业，
35:26
not in theory, but in production,
不是停留在理论上，而是已经在 production 里落地，
35:28
at scale, right now,
而且是在 at scale 地做，就在当下，
35:30
like turning race day data into real-time insights
比如把比赛日数据变成 real-time insights
35:33
for Formula One fans, reinventing the journey
给 Formula One 车迷，重新定义旅程
35:36
for every united traveler
为每一位 United 旅客
35:38
and personalizing every trip with booking.com.
并借助 booking.com 让每一趟行程都个性化。
35:41
However, big the question for your business or industry,
不过，无论你的企业或行业面临的问题有多大，
35:44
AWSAI is how?
AWSAI 是怎样的？
35:47
I'm Jonathan Knight
我是 Jonathan Knight
35:48
and I'm the general manager of New York Times Games.
也是 New York Times Games 的总经理。
35:51
If you play our games, you probably know
如果你玩我们的游戏，你大概会知道
35:53
there's something a bit different about them.
它们有点不太一样。
35:55
Just like there are writers behind the articles
就像文章背后有作者一样，
35:57
you read in the Times,
你在 Times 上读到的那些，
35:58
there are creators behind our daily puzzles.
我们的每日谜题背后也有创作者。
36:01
Tracy Bennett curates the day's wordless solution
Tracy Bennett 精心编排当天的 wordless solution
36:03
to keep it lively and varied.
让它保持生动又多变。
36:05
When a Lou creates each connections board,
当 Lou 制作每个 Connections 棋盘时，
36:07
including all those categories that try to stump you,
包括所有那些想难住你的类别，
36:10
Samazersky comes through every last letter, word,
Samazersky 都会搞定每一个字母、单词，
36:13
and pan-gram and spelling bee
以及 pan-gram 和 Spelling Bee，
36:15
so that loyal players of all skill levels enjoy it.
让各种水平的忠实玩家都能享受其中。
36:18
Our puzzles are human-made every day
我们的谜题每天都是人工制作的。
36:20
with the standards you'd expect from the New York Times.
带着你对 New York Times 所期待的那种水准。
36:23
And this matters because when you choose
而这一点很重要，因为当你选择
36:25
to spend time with our games,
把时间花在我们的游戏上时，
36:27
it should be time well spent,
那这时间就该花得值，
36:29
solving puzzles that are challenging,
去解开那些充满挑战、
36:31
surprising, and joyful.
又令人惊喜、充满快乐的谜题。
36:33
Handcrafted for you.
为你手工打造。
36:35
We think that's something worth investing in
我们觉得，这值得投入。
36:37
and something worth paying for.
and something worth paying for.
36:39
Subscribe now for a special offer on all of our games.
而且是值得花钱的东西。
36:43
at nytimes.com slash join games.
Subscribe now for a special offer on all of our games.
36:46
You did a big piece for the Times
现在就订阅，享受我们所有游戏的特别优惠。
36:52
on the very widespread belief in Silicon Valley
at nytimes.com slash join games.
36:56
but they will create this underclass.
在 nytimes.com slash join games。
36:58
What does the underclass mean to them?
You did a big piece for the Times
37:01
The idea of a permanent underclass, Cosba AI,
你为时报做了一篇大报道
37:04
is basically a world where
基本上就是这样一个世界：
37:06
any job a person can do,
一个人能做的任何工作，
37:08
either AI or robot can do for them,
AI 或者 robot 都能替他们做，
37:10
which means that workers lose all the
这意味着，劳动者会失去他们所有的
37:13
economic leverage they have
经济筹码，
37:14
and capital owners, people with money,
而资本所有者，也就是有钱人，
37:16
can simply pay machine labor to do all the work
可以直接花钱让机器劳动来做所有工作，
37:19
instead of paying workers.
而不是付钱给工人。
37:20
What that means is anyone who earned their living
也就是说，任何靠自己工作
37:23
by working is no longer able to do that.
谋生的人，都不再能这么做了。
37:25
You end up with a world of runaway inequality
最后你会落到一个失控的不平等世界里，
37:27
where the rich get richer
富人越来越富，
37:29
and the working class gets poor.
而工薪阶层越来越穷。
37:31
Maybe they get some welfare checks,
也许他们还能领到一些福利金，
37:33
but fundamentally it's a loss of economic mobility
但根本上，这是一个社会
37:36
in a society.
失去了 economic mobility。
37:37
When I ask folks in Silicon Valley,
当我问 Silicon Valley 的人，
37:39
do you think by default AI is going to increase
你觉得默认情况下 AI 会加剧
37:41
or decrease inequality?
还是减少不平等？
37:42
I have not yet heard anyone say
我到现在还没听任何人说
37:44
it will decrease inequality or keep it the same.
它会减少不平等，或者让它保持不变。
37:46
They might say the floor will get really high.
他们可能会说，底线会变得非常高。
37:48
They might say AI will bring the cost of consumer goods down
他们可能会说，AI 会把消费品的成本降下来，
37:51
and so people's lives are going to get cheaper
于是人们的生活会变得更便宜。
37:53
and everyone will be super healthy
而且每个人都会超级健康
37:55
so it's okay.
所以没关系。
37:56
But I have not heard a single person in the tech industry
但我还没听到科技行业里有任何一个人
37:58
tell me that they believe that AI
告诉我说他们相信 AI
38:00
is going to decrease inequality.
会减少不平等。
38:02
In fact, many people are very worried
事实上，很多人非常担心
38:04
that instead most workers will lose their leverage
相反，大多数工人会失去自己的议价能力
38:07
and be on a kind of permanent welfare in the far off future.
并且在遥远的未来，靠某种永久福利生活。
38:10
I'm pretty skeptical of this vision,
我对这个愿景还挺怀疑的，
38:13
although I don't rule it out.
虽然我也不排除这种可能。
38:15
It might happen.
它也许会发生。
38:16
Although I just don't think AI is going to be quite as revolutionary
不过我只是觉得 AI 不会那么具有革命性，
38:19
as a lot of these people will think us
像这些人里很多人会以为的那样，
38:21
and will not defuse into the real world as easily.
而且不会那么容易地扩散到现实世界里。
38:23
But the thing you're going to need to adjust
但你需要去适应的，
38:26
to any major technological changes time
是任何重大技术变革的时间。
38:30
and then also it's like the AI industry is an all-out war
然后还有就是，AI 行业就像一场全面战争
38:34
to make sure we have a little time for adjustment as possible.
就是要确保我们用来调整的时间尽可能少。
38:37
And I just find it hard to unnot that.
而我就是很难不注意到这一点。
38:39
How many enterprise sales people are open AI
OpenAI 有多少 enterprise sales 人员
38:42
and then throw up a hiring in order to convince companies
然后还大举招聘，就为了说服公司
38:45
that they can replace their workforce
他们可以取代自己的员工，
38:47
or not maybe not replace but expand their workforce
或者不，也许不是取代，而是扩充他们的员工队伍，
38:50
with agents instead of humans, right?
用 agents 而不是人类，对吧？
38:52
They are having these sales conversations
他们在进行这种销售式对话
38:55
trying to persuade people of these questions.
试图说服人们接受这些问题。
38:57
I don't think that a permanent underclass
我不觉得一个永久性的底层阶级
38:59
is the likeliest outcome economically that we're going to get.
会是我们经济上最可能得到的结果。
39:02
I think that AI is actually just really jagged
我觉得 AI 其实非常参差不齐
39:04
and human jobs are super complex and super hard to automate.
而人类的工作超级复杂，也超级难自动化。
39:07
And most of the folks who are predicting economic apocalypse
而且大多数预测经济末日的人
39:11
haven't actually worked enough real jobs
其实并没有做过足够多的真实工作
39:13
to know how complicated
去了解到底有多复杂
39:15
and how multifaceted most jobs really are.
以及大多数工作实际上有多么多面。
39:18
But I definitely agree on the speed point.
但关于速度这一点，我完全同意。
39:20
I think that like Alexey Moss,
我觉得，就像 Alexey Moss 那样，
39:22
the economist has made this point very well.
这位经济学家把这一点说得非常好。
39:24
One thing I think a lot about are people say,
我经常思考的一件事是，人们会说，
39:26
well, humans can adjust.
嗯，人类可以调整适应。
39:27
Humans can re-skill.
人类可以 re-skill。
39:28
They can retrain.
他们可以重新培训。
39:29
They can just do the new jobs that we're going to develop instead.
他们完全可以转而去做我们接下来要开发的那些新工作。
39:32
But you look at things like software engineering
但你看像 software engineering 这样的领域，
39:34
where people will often say now,
现在人们常常会说，
39:36
senior software engineers are doing great.
senior software engineers 过得很不错。
39:38
They love cloud code.
他们很喜欢 cloud code。
39:39
Junior software engineers have been mostly replaced
junior software engineers 大多已经被取代了，
39:42
and you see hiring in job postings are down in that sector.
而且你会看到，这个领域的招聘和职位发布都在减少。
39:45
Well, do we think that a human software engineer
嗯，我们觉得一个人类 software engineer
39:49
is going to re-skill or upskill themselves faster
会更快地 re-skill 或 upskill 自己吗？
39:52
than the next model is going to get better at software engineering?
比下一个 model 在 software engineering 上变得更好还要快吗？
39:55
That's the question that I really wonder about is,
这就是我真正想知道的问题，
39:57
if AI progress outpaces human's ability
如果 AI 的进步超过了人类的能力
40:00
to re-skill, retrain, upskill, adapt,
去 re-skill、retrain、upskill、adapt 的能力，
40:02
then I'm not really sure what there is going to be left.
那我真的不确定还会剩下什么。
40:06
There will be some jobs left,
会有些工作留下来，
40:08
but it's going to be a really, really painful adjustment.
但这会是一个非常、非常痛苦的适应过程。
40:10
I have had so many people
我遇到过太多人
40:13
at the top of these companies,
在这些公司的最高层，
40:15
right, the very tippy top.
对，就是最最顶层的那些人。
40:17
Tell me they wish all this would slow down.
告诉我他们希望这一切都能慢下来。
40:20
Yeah.
是啊。
40:21
I'm sure you have had them say this to you, right?
我敢肯定，你也听他们这么对你说过，对吧？
40:24
But in this world where in their unguarded moments,
但在这个世界里，在他们卸下防备的时候，
40:28
they will say they wish all this was going slower.
他们会说，他们希望这一切能慢下来。
40:31
Well, one way to slow AI down
嗯，让 AI 慢下来的一种方式
40:34
is to construct the number of data centers you can build.
就是限制你能建多少 data centers。
40:38
Yeah.
是啊。
40:39
Just like you've done, you know,
就像你做的那样，你知道，
40:41
as good reporting as anybody,
报道做得跟任何人都一样好，
40:43
on just how conflicted people,
关于人们到底有多纠结，
40:45
even working for these companies,
甚至是在这些公司工作的人，
40:47
seem to be about what they are building,
似乎是关于他们正在构建的东西，
40:49
and yet they're in this competitive race
可他们却身处这场竞争激烈的竞赛中
40:51
to build it as quickly as possible.
要尽可能快地把它做出来。
40:53
And so it makes them a pretty unconvincing.
所以这让他们显得相当没有说服力。
40:56
Profaction.
Profaction.
40:57
Oh, absolutely.
哦，那当然。
40:58
Yeah.
是啊。
40:59
We're building the thing.
我们正在构建这个东西。
41:00
We're telling you to be afraid of,
我们一边告诉你要害怕它，
41:01
and we need to build it as fast as possible,
又一边需要尽可能快地把它造出来，
41:03
even though we sort of admit
尽管我们多少也承认，
41:05
that we'd better if the whole thing was slowed down.
如果整件事能慢下来，我们其实会更好。
41:07
It's really confusing.
这真的很让人困惑。
41:08
It's a weird argument.
这是个很奇怪的论点。
41:09
Yeah.
是啊。
41:10
It's so confusing.
这太让人困惑了。
41:11
I remember when I sat down with Abdul,
我记得我和 Abdul 坐下来聊的时候，
41:13
also, yeah, the Michigan Senate candidate.
对，还有，就是那位 Michigan Senate 候选人。
41:15
He cited Darius 50% white collar drawblast up
他引用了 Darius 的 50% white collar drawblast up，
41:18
probably like five times in the 30-minute conversation.
在 30 分钟的对话里大概提了五次。
41:21
It's like he was like,
就好像他在说，
41:22
they're saying that there's going to be
他们说会有
41:23
a recursive self-improvement,
一种 recursive self-improvement，
41:24
and it might kill us all.
而且它可能会把我们都干掉。
41:25
Like, yeah, I get why you would not
就是，是啊，我明白你为什么不想让这东西跑得更快。
41:27
want to make this thing go faster.
这对 data centers 来说是这样，但对任何其他可能让 AI 慢下来的方式也是这样，
41:29
This is true for data centers,
也就是，每个人都只有在能保证其他公司以及 Chinese labs 会跟他们一起慢下来的情况下，才愿意被拖慢。
41:30
but it's true for any other way
但对于任何其他方式来说，这也是真的
41:32
that you might slow AI down,
即你可能用来放慢 AI 的方式，
41:34
which is that everyone only wants to be slowed down
那就是，每个人都只想被放慢
41:37
if they can guarantee that the other companies
前提是他们能保证其他公司
41:39
that the Chinese labs are going to slow down with them.
那些 Chinese labs 会跟他们一起放慢。
41:42
So long as that's not true,
只要那不是真的，
41:43
they are going to keep racing.
他们就会继续竞赛。
41:45
And I think for that reason,
而且我觉得，正因为如此，
41:47
yeah, I mean,
对，我的意思是，
41:48
the thing that I hear when I talk to people
我跟人聊天时听到的说法是，
41:50
at the companies and data center executives
在公司里，还有 data center 高管那里，
41:52
about the build-out is,
关于 build-out，他们说的是，
41:53
how much money do we need to give these cities
我们需要给这些城市多少钱
41:56
to let us build a data center?
好让我们建一个 data center？
41:57
Tell us how to bribe them better.
告诉我们怎么更好地贿赂他们。
41:59
Tell us what we can do.
告诉我们我们能做什么。
42:00
And so when I talk to them about the build-out,
所以当我和他们聊 build-out 的时候，
42:02
I'm not hearing any sort of personal moral reckoning
我没有听到任何形式的个人道德反省
42:05
with slowing AI down.
关于让 AI 慢下来这件事。
42:06
I'm hearing,
我听到的是，
42:07
how do I make the bribes bigger?
我怎么才能把贿赂搞得更大？
42:08
How big do they need to be?
它们得有多大？
42:09
So then, how do you reconcile what many of these executives,
那么，你怎么调和这些高管中的很多人、
42:14
many of these AI company workers are telling you
这些 AI 公司员工中的很多人告诉你的话——
42:17
about their fears of creating an underclass,
关于他们害怕制造出一个底层阶级，
42:19
about their fears of losing control.
关于他们害怕失去控制。
42:20
I mean, we just saw the situation where open AI's model
我是说，我们刚刚看到那种情况：OpenAI 的 model
42:25
was breaking out of a sandbox in order to sort of cheat
竟然从 sandbox 里突破出来，就为了稍微作点弊
42:29
on its evaluation.
——在它的 evaluation 上。
42:30
So the AI's safety people are very worried.
所以 AI 那边的 safety 人员非常担心。
42:32
The safety teams in here,
这里的 safety 团队，
42:34
clearly don't have full control
显然并没有完全掌控，
42:35
or even understanding of what they're building.
甚至都不理解自己在构建什么。
42:37
How do you reconcile?
那你怎么调和？
42:39
If you reconcile,
如果你要调和，
42:40
the way the AI companies talk when they are giving voice
那些 AI 公司说话的方式——当它们表达
42:45
to their fears or the people at them
自己的恐惧，或者为它们内部的人发声时
42:47
talk when they are giving voice to their fears
当他们表达自己的恐惧时，会说出来
42:50
and they're pretty profound hostility
而且他们抱有相当深的敌意
42:54
to anything that would slow down how fast we are building
对任何会放慢我们构建速度的东西
42:58
this thing,
这个东西，
42:59
whose consequences they freely admit they cannot predict.
他们坦率承认，自己无法预测它的后果。
43:03
Yeah, it's fascinating
是啊，这太让人着迷了
43:05
because just on a very personal level
因为单从非常个人的角度来说
43:07
when I talk to people at these companies,
当我跟这些公司里的人聊天时，
43:09
I just think,
我就觉得，
43:10
man, if I thought this thing might have a 10% chance
天哪，如果我觉得这东西可能有 10% 的概率
43:12
of killing a soul or taking everybody's job,
会杀死一个灵魂，或者抢走所有人的工作，
43:15
I wouldn't work on it.
我就不会去做它。
43:16
I would not feel that.
我不会那么觉得。
43:17
I personally could not morally justify taking that chance.
就我个人而言，我在道德上没法为冒这个风险辩护。
43:22
And when I talk to people who are not in the San Francisco
而当我跟那些不在 San Francisco
43:25
and AI world,
和 AI 世界里的人聊天时，
43:26
they feel like I do.
他们和我想的一样。
43:27
They're just like,
他们就会说，
43:28
why would you do it?
你为什么要这么做？
43:29
And I think they're basically three rough buckets
而且我觉得，这些基本上就是三大类
43:31
of rationales that I hear from people
理由，有些是我从人们那里听到的
43:33
or that I hear between the lines from people.
有些则是我从他们的话里听出来的。
43:37
One is this sort of sense of techno-determinism.
其中一种，就是这种 techno-determinism 的感觉。
43:41
Its superintelligence is going to be built inevitably.
superintelligence 注定会被造出来。
43:45
There is no way it's not going to happen.
这件事不可能不发生。
43:47
If it happens,
如果它发生了，
43:48
I want to be part of it.
我想成为其中的一部分。
43:49
I want to make my money from it.
我想靠它赚钱。
43:50
I want to maybe make it happen in the least bad way.
我想，也许让它以最不坏的方式发生。
43:52
I think that's a super common answer.
我觉得这是个超级常见的回答。
43:55
Another is this technology might kill us all,
另一个回答是，这项技术可能会把我们都杀死，
44:02
but it also might be really amazing.
但它也可能真的非常棒。
44:03
It might produce super abundance for everybody.
它可能会为每个人带来 super abundance。
44:06
It might be immortal.
它可能会永生。
44:08
It might double everyone's life spans,
它可能会让每个人的寿命翻倍，
44:10
cure all diseases,
治愈所有疾病，
44:11
bring the cost of every consumer good,
把每一种消费品的成本，
44:13
housing, energy, whatever, to near zero.
住房、能源，随便什么，都降到接近零。
44:15
And that would be utopia.
那就会是乌托邦。
44:17
And so I think all the time to that anecdote
所以我总是会想到那个轶事
44:21
that I think SPF set on a podcast where it's...
我觉得 SPF 在一个播客里说过，当时是……
44:24
Sam Beckman-Fried.
Sam Beckman-Fried。
44:25
Yes, Sam Beckman-Fried set on a podcast where it was like,
对，Sam Beckman-Fried 在一个播客里说过，大意是，
44:27
if you could flip a coin
如果你能抛一枚硬币
44:28
and it was 51% odds,
而且它有 51% 的概率，
44:30
you would double the total amount of human welfare
你会让人类福祉的总量翻倍
44:33
and 49% chance everyone dies.
还有 49% 的概率所有人都会死。
44:35
What do you flip the coin?
你会抛这枚硬币吗？
44:36
He says yes.
他说行。
44:37
And again, I feel compelled to say,
而且，我又忍不住想说，
44:38
like,
就是，
44:39
have you guys here of like,
你们有没有听说过，比如说，
44:40
you know, how do you really know?
你知道吧，你怎么真的知道呢？
44:41
That's what's happening, blah, blah, blah, blah.
事情就是这样，blah, blah, blah, blah。
44:42
Whatever, but that aside,
随便吧，不过撇开这个不谈，
44:43
take the hypothetical,
就拿这个假设来说，
44:44
pure hypothetical.
纯属假设。
44:45
Yeah, yeah.
对，对。
44:48
I think this is a hyperbolic example,
我觉得这是个夸张的例子，
44:50
but I think it's not actually that far off
但我觉得它其实也没那么离谱，
44:52
from what a lot of the people
跟很多
44:54
building super intelligence believe, too,
构建 super intelligence 的人所相信的也差不多，
44:56
that they are basically willing to flip the coin.
他们基本上就是愿意抛硬币。
44:59
Maybe we all die,
也许我们都会死，
45:00
but maybe we're all immortal
但也许我们都是永生的
45:02
and that expected value wise cancels things out.
而从 expected value 的角度看，这会把事情抵消掉。
45:05
And then the final category is just I think folks
然后最后一类，我想就是那些人
45:08
who are so fascinated by the technical endeavor
他们如此着迷于技术上的探索
45:11
of whether we can build this thing
我们能不能造出这个东西
45:12
and how to do it,
以及怎么做出来，
45:13
that they just aren't super worried
以至于他们根本不太担心
45:15
about the consequences
会有什么后果
45:16
or what else might happen.
或者还可能发生别的什么。
45:18
So people have all sorts of self-justifying narratives
所以人们有各种各样的自我辩护式说辞
45:20
as to why it's worth it.
来说明为什么这值得。
45:22
Some I think are better than others.
有些我觉得比另一些更好。
45:24
But it makes sense to me why the public
但我觉得可以理解，为什么公众
45:28
is not particularly sympathetic to any of these.
对其中任何一个都不怎么同情。
45:30
I mean, that middle narrative.
我是说，就是中间那套叙事。
45:32
Well, I heard Sam Beckman
嗯，我听到 Sam Beckman
45:33
for you to say that.
你会这么说。
45:34
I think it was on Tyler Cowan's podcast.
我想那是在 Tyler Cowan 的播客上。
45:36
Oh, that's like,
哦，那就像是，
45:37
that's a psychopath, right?
那是个 psychopath，对吧？
45:39
To actually believe that you would have to be a psychopath.
要真的相信这一点，你得是个 psychopath 才行。
45:42
You would have to have a very, very, very low value
你得对……的评价非常、非常、非常低
45:45
on I think human life.
我想，是对人类生命的。
45:47
Yes.
是的。
45:48
Imagine being the person who flips a coin
想象一下，你是那个抛硬币的人
45:50
and it comes up wrong.
结果却翻到了错的那一面。
45:51
Oh, Jesus, yeah.
哦，Jesus，是啊。
45:52
Like I'm a parent.
比如说，我是个家长。
45:54
The idea that you would do something
一想到你会去做某件事
45:56
that's like 51.49.
差不多是 51.49 那种。
45:58
I'm like,
我就会想，
45:59
your kid is, you know, doubly happy
你的孩子，你知道的，会双倍开心。
46:01
or your kid is gone.
或者你的孩子就没了。
46:02
Yeah.
对。
46:03
You would never.
你绝对不会。
46:04
Of course.
当然了。
46:05
You don't even want to say that out loud.
你甚至都不愿意把这话说出口。
46:06
Of course.
当然了。
46:07
I think that's how almost everybody thinks about it.
我觉得几乎所有人都是这么想的。
46:09
And again, I think 51.49 is obviously
而且，我觉得 51.49 显然
46:11
the most egregious example you could think of.
你能想到的最离谱的例子。
46:13
And so SPF is very unsympathetic.
所以 SPF 非常不近人情。
46:15
But when I think about the super intelligence bet,
但当我想到 super intelligence 这个赌注时，
46:17
a lot of people will characterize it as a 90-10 bet,
很多人会把它形容成 90-10 的赌注，
46:20
as an 80-20 bet.
或者 80-20 的赌注。
46:22
And this question of how much is an acceptable amount
还有这个问题：多少才算可接受，
46:24
of either extinction risk
无论是 extinction risk
46:26
or total disempowerment risk?
还是 total disempowerment risk？
46:28
You know, I think people have very different risk appetites.
你知道，我觉得人们的风险偏好差别很大。
46:32
And Silicon Valley is a play set as always
而 Silicon Valley 一直就是个游乐场，
46:34
prized their high risk appetite.
推崇他们那种高风险偏好。
46:36
I think that makes a lot more sense
我觉得这就合理多了，
46:38
when you're talking about maybe yourself or your company
当你说到的可能是你自己，或者你的公司时，
46:40
full of people who have opted in
里面全是主动选择加入的人，
46:41
to taking a very high risk endeavor.
去从事一项风险极高的事情。
46:43
I think that's extremely different, obviously,
我觉得这显然非常不同，
46:45
when you're talking about the rest of the world.
当你在说世界其他地方的时候。
46:47
And one thing with the data center debates that I'd always hear
而且关于 data center 的争论，我总会听到的一点是
46:49
is like, you know, I get that these people
就是，你知道，我明白这些人
46:51
are making this crazy bet on this technology.
是在这项技术上做这种疯狂的押注。
46:53
They think it's going to change the world.
他们觉得这会改变世界。
46:55
But why do they have to do in our backyard?
但他们为什么非得在我们后院搞这个？
46:57
Why is Mark Zuckerberg not building a data center
为什么 Mark Zuckerberg 不建一个 data center
46:59
in his backyard?
在他自己的后院？
47:00
And so this question of,
所以问题就在于，
47:01
yeah, you guys are going to create
对，你们这些人会带来
47:04
like these very tangible impacts
一些非常切实的影响，
47:06
and to their real harms on specific communities
以及这些影响对特定群体造成的真实伤害，
47:09
that are not the communities benefiting
而那些群体并不是
47:11
from this technology.
从这项技术中受益的群体。
47:12
At least they don't see the benefits yet.
至少他们现在还看不到好处。
47:14
They don't see the cancer cures.
他们看不到治愈癌症的方法。
47:15
They don't see themselves getting these million dollar,
他们不觉得自己能拿到这些百万美元、
47:17
10 million dollar salaries that they are researchers
千万美元的年薪，而这些研究员们
47:19
are getting.
正在拿到。
47:20
It feels like,
感觉像是，
47:21
I think it feels to a lot of these folks
我觉得对很多这样的人来说，
47:23
like they are ponds in some tech billionaires this game.
他们像是某些科技亿万富翁这盘游戏里的棋子。
47:27
And they do not like to feel that way.
而且他们不喜欢有这种感觉。
47:29
Why aren't they building it in their own backyard?
为什么他们不在自己的地盘上搞这个？
47:30
That's why I don't see a bunch of data centers
这也是为什么我在 Northern California 看不到一大堆 data centers。
47:33
in Northern California.
我觉得不用我告诉你，为什么在 Northern California 建设这么难。
47:35
I don't think I need to tell you why it's so hard to build
但我一方面觉得这是真的，在 Northern California 建设确实很难。
47:38
a Northern California.
但另一方面我也觉得，刚说的另一件事好像也有它的道理，
47:39
But I both think that's true
但我既觉得这是真的
47:42
that it's hard to build a Northern California.
要打造一个 Northern California 是很难的。
47:43
But I also think there's like a truth
但我也觉得，这里面好像有某种真相
47:45
to the other thing being said,
再说另一件被提到的事，
47:46
they don't want them there.
他们不想让它们建在那儿。
47:47
I mean, the land is expensive.
我是说，地价很贵。
47:49
It would be very hard,
这会非常难，
47:50
like inexpensive to build a data center.
比如想低成本建一个 data center。
47:51
Like the places where we're talking about.
就像我们说的那些地方。
47:54
But it also gets it a core truth,
但这也触及了一个核心事实，
47:56
which is people don't actually want data centers around them.
那就是，人们其实并不想自己周围有 data center。
47:59
It is a cost.
这是一种成本。
48:00
It is a concentrated cost for a diffuse benefit.
这是一种集中的成本，换来的是分散的收益。
48:03
Like if you believe in the benefit.
就是如果你相信这个收益的话。
48:05
Yeah.
嗯。
48:06
So I think that's part of it.
所以我觉得这也是其中一部分。
48:07
I want to go back to the first bucket
我想回到第一个分类
48:09
you were talking about,
你刚才说的，
48:10
which is like the race dynamics.
也就是 race dynamics 那类东西。
48:11
Yes.
对。
48:12
So at the most generous,
所以，最宽容地说，
48:14
the thing that I've heard repeatedly is sort of what you're describing,
我反复听到的说法，差不多就是你在描述的那种情况，
48:17
which is it would be better if this were going slower.
也就是，如果这件事进展得慢一点会更好。
48:22
But I can't control that
但我控制不了这个，
48:25
because I,
因为我，
48:26
whether I'm at, you know,
不管我是在，你知道，
48:28
anthropic or open AI or Google or meta,
anthropic 还是 open AI 还是 Google 还是 meta，
48:32
you know, if we slow down,
你知道，如果我们放慢速度，
48:34
it just is our less ethical competitors over there
就是那边那些没我们这么讲道德的竞争对手
48:37
who speed up.
会加速。
48:38
And even if you put down legislation
而且就算你出台立法
48:41
slowing down all of America,
拖慢整个美国，
48:43
then it's China.
那赢的就是中国。
48:44
Right.
对。
48:45
You know, the CCP,
你知道，CCP，
48:46
which is going to win the race.
会赢下这场竞赛。
48:49
I guess one question is,
我猜有个问题是，
48:53
do you buy this central metaphor
你买账这个核心隐喻吗，
48:55
of a race that has a like a ticker tape line
就是一场比赛，有一条像 ticker tape 那样的线，
48:59
where at some point somebody passes it
在某个时刻，有人越过了它，
49:01
and then they have the recursive superintelligence
然后他们就拥有了 recursive superintelligence，
49:03
and like the race is over.
然后就好像比赛结束了。
49:05
Or do you see this more as,
还是说你更把它看成，
49:07
like most technologies,
就像大多数技术那样，
49:09
like a kind of like a linear set of gains?
就像一种，像是 linear 的一组 gains？
49:13
I mean, can be fast, can be slow.
我是说，可以快，也可以慢。
49:15
But it doesn't have that.
但它没有那个。
49:17
Somebody is going to win dynamic.
有人会赢下这个 dynamic。
49:19
Yeah.
对。
49:20
I find this really confusing.
我觉得这真的让我很困惑。
49:22
One of the first things that I did
我最早做的事情之一
49:25
when I started reporting more deeply on AI
就是当我开始更深入地报道 AI 的时候
49:27
was try to figure out what AGI meant,
是在试着弄清楚 AGI 到底是什么意思，
49:29
because a lot of the way that this race has been characterized
因为这场竞赛被描述的方式，很大一部分
49:33
is who will build AGI first?
就是谁先造出 AGI？
49:35
Artificial general intelligence.
Artificial general intelligence.
49:37
Artificial general intelligence first.
先造出 Artificial general intelligence。
49:39
The first thing I found,
我发现的第一件事，
49:41
no one agrees on what that means.
就是没人对那到底是什么意思达成一致。
49:43
AGI means everything from AI
AGI 意味着从 AI 到……的一切
49:45
that can build itself to AI
能自我构建成 AI 的
49:47
that can do all human jobs to AI
能完成所有人类工作的那种 AI
49:49
that produces whatever number of economic value.
能产出任意数量经济价值的。
49:51
And so everyone has these different milestones
所以每个人都有这些不同的里程碑
49:54
for what constitutes AGI to them.
对他们来说什么才构成 AGI。
49:56
And also means is that the race has different finish lines
而且这也意味着，这场竞赛有不同的终点线
49:59
and moving finish lines.
而且终点线还在移动。
50:00
And I think that like,
而且我觉得，就...
50:02
you see the way that these models perform differently on benchmarks.
你看，这些模型在 benchmarks 上的表现有多不一样。
50:05
They are extremely jagged.
它们非常参差不齐。
50:07
They can be super good at math.
它们可能数学超强。
50:09
And they can be super bad at poker at the same time.
同时又可能打扑克超烂。
50:12
They can be amazing at cracking cyber security problems,
它们可能特别擅长破解 cyber security 问题，
50:16
but not able to build anything in the physical world.
但在物理世界里却什么都造不出来。
50:19
And because of that,
也正因为这样，
50:21
I don't think that technology is A as general as people suggested is.
我不觉得这项技术像人们说的那么 general。
50:25
It means that it is much harder to define a finish line to the race.
这意味着，这场竞赛的终点线要难定义得多。
50:30
And my sense is that the race,
而我的感觉是，这场竞赛，
50:32
because you cannot adjudicate it,
因为你没法评判它，
50:34
everyone will always feel that they are falling behind on some dimension.
每个人总会觉得自己在某个维度上落后了。
50:39
I mean, to then make the case for that these AI companies are making
我的意思是，那要为这些 AI 公司正在提出的主张辩护，
50:42
is they do believe in this recursive self improvement.
就是他们确实相信这种 recursive self improvement。
50:45
They think that, you know, open AI Google,
他们认为，你知道，open AI、Google，
50:47
deep-minded and a throwback at all,
deep-minded 和 a throwback at all，
50:49
extremely focused specifically on the question of,
极其聚焦于这个问题，
50:51
can we build AI that built itself?
我们能造出自我构建的 AI 吗？
50:52
Can we build an AI
我们能构建一个 AI
50:53
that can train the next generation model completely from scratch on its own?
它能完全靠自己从零开始 train 下一代的 model 吗？
50:58
And in that sense you get an exponential pace of improvement
而从这个意义上说，你就能获得指数级的改进速度
51:02
for whoever can hit that recursive self improvement curve first.
对于谁先能踏上那条 recursive self improvement curve 的人来说。
51:07
And they think that this might lead to that company pulling ahead.
而且他们认为，这可能会让那家公司一路领先。
51:11
Right now folks think that it's anthropic,
现在大家觉得是 anthropic，
51:13
which has the best coding models,
哪家的 coding models 最好，
51:15
meaning they can code faster,
意味着它们写代码更快，
51:16
meaning that their next models are even better.
也就意味着它们的下一代 models 甚至更好。
51:20
I can see where that argument is,
我能理解这个论点从哪来，
51:24
but I'm not sure when we look at,
但我不太确定，当我们看
51:27
you know,
你知道，
51:28
the latest and the topic models versus the latest open AI models versus the latest,
最新的，以及 topic models，对比最新的 OpenAI models，再对比最新的，
51:32
you know, Chinese open weights models that we see a company pulling ahead
你知道，中国的 open weights models，我们看到有一家公司正在拉开差距
51:37
that decisively in that way,
以那种方式那么果断地，
51:38
especially when every single company in lab is using the same recursive self improvement strategy.
尤其是当 lab 里的每一家公司都在用同样的 recursive self improvement 策略时。
51:43
And so basically,
所以基本上，
51:45
I don't know that the race has a finish line.
我不确定这场竞赛有没有终点线。
51:48
And that's what worries me about it continuing.
而这正是它继续下去让我担心的地方。
51:51
The reason I want to focus in on this race metaphor for a minute is
我想先花点时间聚焦在这个竞赛比喻上的原因是，
51:54
I've come to think it is really one of the central dividing lines
我逐渐觉得它真的是核心分界线之一，
51:58
and how you think about different kinds of AI policy.
以及你如何思考不同类型的 AI policy。
52:02
Whether you think that we are in a race with China
你是否认为我们正在和 China 进行一场竞赛
52:06
to get to the point where one side of the other is going to pull endlessly and decisively ahead
到达某一方会无限地且决定性地领先的地步
52:11
because they hit that recursive self improving level.
因为他们达到了 recursive self improving 的水平。
52:16
Well, then that means what you do in the next one to three years
那么，这就意味着你在接下来的一到三年里所做的事情
52:21
is incredibly incredibly definitely important.
是极其极其绝对重要的。
52:25
But if you don't believe that,
但如果你不相信这一点，
52:27
if you believe something more like,
如果你相信的更像是，
52:29
yes, this is a powerful technology.
是的，这是一项强大的技术。
52:30
It's a powerful technology that might have a lot of downsides,
这是一项强大的技术，可能有很多弊端，
52:32
might come with a lot of social instability.
可能会伴随着很多社会不稳定。
52:35
It's a fact on a society may not be good.
对一个社会来说，它可能并不是好事，这是事实。
52:38
Then
那么
52:40
let's run faster to the bad place
让我们更快地冲向那个糟糕的地方
52:43
is not nearly as compelling an argument
这个论点就远没有那么有说服力了
52:46
and all of a sudden the idea that we should have policy in place
然后突然之间，我们应该让政策到位这个想法
52:50
that slows things down for more voice, for more consideration,
让事情慢下来，为了更多声音、更多考量，
52:55
it's not crazy.
这并不疯狂。
52:56
And I guess one place that goes is that I have begun to notice
而且我猜，这引向的一点是，我开始注意到
52:59
like a really interesting convergence between the SFAI safety people in a way
某种意义上，SFAI 的 safety 那帮人
53:06
and like the AI populace like Bernie Sanders
和像 Bernie Sanders 那样的 AI 大众之间，有一种非常有意思的趋同
53:09
who are in a different way and I'll say,
他们走的是不同的路径，而且我会说，
53:11
who are getting to not that different places
他们到达的地方并没有那么不同
53:15
but through very, very different mechanisms.
但通过非常非常不同的机制。
53:18
They're like the AI safety people who actually believe we are in a race
他们就像那些真的相信我们正处在竞赛中的 AI safety 那帮人
53:22
but they believe that winning that race might bring the end of humanity.
但他们相信，赢得那场竞赛可能会带来人类的终结。
53:26
And so they don't want to move that fast.
所以他们不想推进得那么快。
53:29
If we began to slow down,
如果我们开始慢下来，
53:31
we'd have more credibility for negotiating with China
我们在跟 China 谈判时就会更有可信度
53:33
and trying to come up with international treaties and all the rest of it.
也能试着搞出各种国际条约之类的东西。
53:37
And then you have like the kind of more AI populace side
然后还有那种更偏向 AI 民众阵营的一方
53:40
who just like doesn't want to give all these tech billionaires all this power
他们就是不想把所有这些权力都交给这些科技亿万富翁
53:43
who doesn't believe this technology will be good for people
他们也不相信这项技术会对人们有好处。
53:45
and they're starting to come up with like,
而且他们开始提出一些，就像，
53:47
maybe not the policies that the safety people would
也许不是 safety 那帮人会提出的政策，
53:50
but you know data center moratoriums and things like that.
但你知道，data center 暂停令之类的。
53:53
And so you have this sort of slightly strange
于是你就有了这种有点奇怪的，
53:56
like you would not have considered this coalition.
怎么说呢，你本来根本不会想到的联盟。
53:59
It's super interesting.
这超级有意思。
54:00
I mean you literally have Rhonda Santis doing AI round tables
我是说，你居然真的看到 Rhonda Santis 在搞 AI 圆桌会议，
54:04
with Max Tegmark who's been one of the leading advocates of pausing
跟 Max Tegmark 一起，而他一直是主张暂停的领头倡导者之一。
54:08
and slowing down AI and MIT professor.
还有给 AI 减速，以及 MIT 教授。
54:10
And you have Bernie Sanders doing viral videos with LEAs of you
而且还有 Bernie Sanders 跟 Eliezer Yudkowsky 一起拍爆火视频。
54:15
Kowski, the guy who's telling us that AI is probably going to kill us all
就是那个告诉我们 AI 很可能会把我们全干掉的人。
54:19
if we build it.
如果我们把它造出来的话。
54:20
Which is both an alliance it doesn't and does kind of makes it.
这既算是一种联盟，又不算，但某种程度上又确实算。
54:23
You know what I mean?
你懂我的意思吧？
54:24
Totally.
完全懂。
54:25
I've been spending some time in DC this year to talk to some of these AI populists
今年我在 DC 待了些时间，跟一些这样的 AI populists 聊了聊。
54:28
some from the social conservative right others from say the labor left.
有些来自社会保守派右翼，另一些来自比如说劳工左翼。
54:31
And this band and guy I was talking to told me he's like,
还有这个乐队，还有我聊过的那个哥们儿，他跟我说，他说：
54:34
you know, like I wouldn't send my kids over to a play day at the polycule
你知道，就像我不会把我的孩子送去 polycule 的玩耍日。
54:38
but I can do coalitions.
但我可以搞联盟。
54:39
And so I think it's been one of the most interesting political stories
所以我觉得它一直是最有趣的政治故事之一。
54:43
going on right now as a sort of strange about the bedfellows that have emerged.
现在正在发生，算是某种关于已经出现的那些奇怪盟友的怪事。
54:46
I mean even with the data center stuff I was talking to an activist
我是说，即使是 data center 那些事，我之前跟一个活动人士聊过。
54:49
and they were saying these were liberal women who had gotten into politics
他们说这些是进入政界的自由派女性。
54:52
after the 2016 election of Donald Trump.
在2016年Donald Trump当选之后。
54:54
They said that data centers were the first thing that got them to talk
他们说，data centers 是让他们能够
54:57
productively with their Trump voting neighbors about politics the first thing
和那些投票给 Trump 的邻居就政治进行有成效对话的第一件事，
55:00
in like 10 years almost, which is fascinating to me.
几乎是 10 年来的第一次，这让我觉得特别有意思。
55:03
And in this sense they felt a really strong sense almost of political agency
而在这个意义上，他们感到一种非常强烈的、几乎算是政治能动性的感觉，
55:07
that analysts came out of this fight.
即分析人士从这场争斗中走了出来。
55:10
So yeah, I think one of the big questions that folks in AI safety for example
所以，是的，我觉得，比如 AI safety 领域的人正在思考的一个大问题是：
55:15
are thinking about is do we want to build these alliances
我们想不想建立这些联盟？
55:18
with the rising left and right populist waves in American culture
随着美国文化中左翼和右翼民粹主义浪潮的兴起
55:22
in order to slow AI down?
是为了拖慢 AI 吗？
55:24
Maybe it's okay that we have different reasons and different theories
也许没关系，我们有不同的理由和不同的理论
55:27
for why AI is so dangerous.
来解释为什么 AI 如此危险。
55:29
For one person is big model bad for another person.
对一个人来说，是 big model 不好；对另一个人来说，
55:32
It's big billionaires bad big tech bad.
是那些超级亿万富翁不好，big tech 不好。
55:35
And those folks are sort of linking arms in a lot of ways against the AI accelerationists
而那些人某种程度上在很多方面正联手对抗 AI accelerationists
55:43
and sort of the folks pushing the race faster.
以及那些在某种程度上把这场竞赛推得更快的人。
55:46
So I think this and this we've been sort of living in the data center
所以我觉得，这个，这个，我们一直算是活在 data center
55:51
or moratorium side of the politics.
或者说，政治里暂停令的那一边。
55:54
But what's the other side of it?
但另一边又是什么呢？
55:56
What are the problems with just saying?
直接这么说有什么问题？
55:58
Okay, fine, like let's not build any more data centers.
好吧，行吧，那我们就别再建更多 data centers 了。
56:01
One, like this is not actually the way that you would successfully slow down AI.
第一，这其实并不是你能成功拖慢 AI 的方式。
56:07
If that's what you really wanted.
如果那真是你想要的。
56:09
One, if you if one locality or one state imposes a moratorium,
第一，如果，如果某个地方或某个州实施了暂停令，
56:13
AI companies are very, very happy to go to other states or other countries.
AI 公司非常非常乐意去其他州或其他国家。
56:16
They're already flooding into Texas for example
比如说，它们已经在涌入 Texas 了。
56:19
because it's had such a pro data center environment.
因为那里的 data center 环境一直特别友好。
56:21
People are looking at Louisiana, the Dakotas, Australia.
人们正在关注 Louisiana、the Dakotas、Australia。
56:24
Space of course is a current big interest of Elon Musk's
太空当然是 Elon Musk 目前的一大兴趣。
56:28
because people think that maybe not now, but in five years
因为大家觉得，也许不是现在，但五年后
56:31
we can just put all the data centers in space
我们可以直接把所有 data centers 都放到太空里。
56:33
and solve the political problems that way.
然后用这种方式解决政治问题。
56:35
So one is I'm not really sure that this would stop AI progress that much.
所以，第一点是，我不太确定这能多大程度上阻止 AI 的发展。
56:39
It would just shift the data centers to other places that do welcome them.
这只会把 data centers 转移到其他确实欢迎它们的地方。
56:44
The second thing is I actually do think that
第二点是，我其实确实认为，
56:48
there are ways for these deals to be good.
这些交易是有办法变成好事的。
56:51
Not every community should want a data center.
不是每个社区都应该想要一个 data center。
56:54
I think that many of them may discuss it and say,
我觉得他们中的很多人可能会讨论一下，然后说，
56:57
this isn't what we want.
这不是我们想要的。
56:58
We don't need the tax revenue.
我们不需要这笔税收收入。
56:59
That was bad.
那太糟了。
57:00
But in a lot of the cases with these sites that I visited,
但在我去过的很多这类场地里，
57:04
like you know, in the old GM site in Jamesville,
你知道，就像 Jamesville 那个旧的 GM 场地，
57:07
the British and partners the data center developer was going to clean that brown field up.
British and partners 这个 data center developer 本来打算把那块 brown field 清理干净。
57:11
They were the only ones willing to do so.
他们是唯一愿意这么做的人。
57:13
I talked to a real estate broker who had tried to sell the site five years
我跟一个房地产经纪人聊过，他之前试着卖这个场地，卖了五年，
57:16
and he couldn't do it because not a single other commercial buyer wanted
结果他卖不掉，因为其他商业买家没有一个愿意
57:19
to clean up all of this hazardous waste.
清理所有这些 hazardous waste。
57:21
Only the data centers were willing to do that.
只有那些 data centers 愿意这么做。
57:23
Or with Mount Pleasanton, the Foxconn site, right.
或者像 Mount Pleasanton，也就是 Foxconn 那个场地，对吧。
57:25
They had already cleared all this land.
他们已经把这片地全都清理出来了。
57:27
They had built all this infrastructure.
他们已经把所有这些基础设施都建好了。
57:29
Putting a data center there was a net improvement for the community.
在那里建一个 data center，对社区来说是净改善。
57:34
Most in my opinion personally, from an economic perspective,
大多数情况下，就我个人看来，从经济角度来说，
57:37
from the perspective of there is nothing going on there anyway.
从反正那儿也没什么动静的角度来看。
57:40
So I think that there are ways to do these deals right.
所以我觉得，这些交易是有办法做好的。
57:44
There is enough money in this industry that a lot of communities will decide
这个行业里的钱足够多，所以很多社区都会认定
57:47
it is economically beneficial for them to bring in these jobs and bring in this investment.
引进这些工作岗位、引进这笔投资，对它们来说在经济上是划算的。
57:51
But I think that the way that the data center deals have been done,
但我觉得，data center 交易一直以来的做法，
57:54
nearly guarantee the amount of public backlash that there's been,
几乎就注定了现在会有这么大的公众反弹，
57:57
and the thing that will probably fix it, I suspect, is probably either like a state level streamlining
而我觉得，最终可能解决这个问题的，大概要么是某种州层面的流程精简，
58:03
where someone does the research probably at the state level,
也就是由某个人去做研究，可能是在州层面，
58:06
maybe at the federal level to figure out what is the fair way to do these deals.
也可能是在联邦层面，弄清楚做这些交易的公平方式到底是什么。
58:09
How do we ensure the communities get the most transparency,
我们怎么确保社区获得最大的透明度，
58:12
the benefit out of data center deals when they happen.
data center 交易发生时所产生的好处。
58:14
So it's not case by case, and it's not so asymmetric with the town of 12,000 negotiating
所以这不是逐案处理，也不会像 12,000 人的小镇在跟
58:18
with an open AI or whatever.
一个 OpenAI 或别的什么谈判时那样不对称。
58:20
I would add two other things to that that would be curious to hear you take on.
我想在此基础上再补充两点，也挺想听听你的看法。
58:23
So one you mentioned data centers moving towards other localities.
所以，第一点，你提到 data centers 在向其他地区转移。
58:26
And those localities were not randomly selected.
而那些地区并不是随机选出来的。
58:29
The localities that are going to impose fewer conditions.
那些将要施加更少条件的地方。
58:33
So maybe that is fewer environmental conditions.
所以也许那就是更少的环境条件。
58:36
But in the case of maybe a UAE or some of the Gulf states that are interested here,
但如果是 UAE，或者这里感兴趣的一些海湾国家，
58:40
you're looking at more authoritarian countries.
那你面对的就是更威权的国家。
58:42
So I've heard a lot of people worry about that, or Elon Musk in space.
所以我听到很多人担心这一点，或者担心 Elon Musk 在太空的事。
58:46
So in a sense, if you make it to the data centers,
所以从某种意义上说，如果你能进入那些 data centers，
58:50
can not go into places where there is more democratic control,
却不能进入那些有更多民主控制的地方，
58:54
you might end up with less overall democratic control.
你最终可能会让整体上的民主控制更少。
58:57
The other thing, and I do think this is significant,
另一件事，而且我确实认为这很重要，
59:00
is that there is right now more demand for compute than there is compute.
就是现在对 compute 的需求已经超过了现有的 compute。
59:06
People talk a lot about bubble, but we do not look to have excess AI supply at the moment.
大家都在大聊特聊 bubble，但目前我们看起来并没有过剩的 AI 供给。
59:12
And if demand keeps rising because the coding agents get better
而如果需求持续上升，是因为 coding agents 变得更好，
59:16
and all the rest of the things we know that are happening,
再加上我们知道正在发生的其他所有事情，
59:19
but you are constricting the supply of compute,
但你却在限制 compute 的供给，
59:24
then you end up with more inequality and who can afford it.
那最终就会导致更严重的不平等，以及谁能负担得起的问题。
59:28
So a Goldman Sachs, a JP Morgan, a company with a lot of money to buy compute
所以像 Goldman Sachs、JP Morgan 这种有大笔钱买 compute 的公司，
59:34
is going to have a lot of it, and then ordinary users, small businesses, etc.
就会有很多，而普通用户、小企业等等。
59:39
If you believe AI is important and powerful and I believe it is important and powerful,
如果你相信 AI 重要且强大，而我也相信它重要且强大，
59:44
then you have a problem where you have created much more stratification and who can afford it.
然后就会出现一个问题：你造成了更严重的阶层分化，还有谁能负担得起。
59:49
How do you think about those dimensions of it?
你怎么看待它的这些维度？
59:52
I think with where you build the data centers,
我觉得，就你把 data centers 建在哪儿来说，
59:55
a lot of folks are starting to look at building AI infrastructure as a form of geopolitical leverage, right?
很多人开始把建设 AI infrastructure 看作一种地缘政治杠杆，对吧？
60:00
And so some countries like places like Australia, Canada, countries in Europe are thinking,
所以有些国家，像 Australia、Canada、Europe 的一些国家，都在想，
60:07
actually maybe the way for us to get a slice of Frontier AI,
其实，也许我们分到 Frontier AI 一杯羹的方式，
60:10
for us to negotiate with the countries where the best AI is being developed,
就是让我们能和那些正在开发最好 AI 的国家谈判，
60:14
like the US, in cases like cybersecurity access,
比如 US，在 cybersecurity access 这类事情上，
60:18
is to say, you know, we'll build your data centers here, we'll actually welcome you in,
也就是说，你知道，我们会把你们的 data centers 建在这儿，我们其实会欢迎你们进来，
60:23
and in return maybe you guarantee us access to the Frontier models.
作为回报，也许你们要保证我们能访问 Frontier models。
60:27
So I think that one is that we should look at AI infrastructure as a point of leverage
所以我觉得，其中一点是，我们应该把 AI infrastructure 看作一个杠杆点，
60:34
that both states and countries have.
这是州和国家都拥有的。
60:36
And as you mentioned, if local moratoriums in the US,
而且就像你提到的，如果美国地方上的暂停令，
60:41
if domestic moratoriums or something like that,
如果国内的暂停令或者类似的东西，
60:43
lead to giving that leverage and negotiating power to authoritarian states,
导致把这种杠杆和谈判筹码交给威权国家，
60:48
that's probably something the US should be really worried about.
那可能是美国真的应该担心的事情。
60:51
On the other hand, there are folks like Anton Leite Carnegie has done work on this,
另一方面，也有像 Anton Leite Carnegie 这样的人，他在这方面做过一些工作，
60:55
where it's, can we give our allies, can we give our democratic allies,
也就是说，我们能不能给我们的盟友，能不能给我们的民主盟友，
60:59
negotiating leverage through them building out compute?
通过让他们建设 compute 来给他们谈判筹码？
61:03
The second thing that you mentioned about pricing is interesting,
你提到的关于定价的第二件事很有意思，
61:06
because I do think one of the big macro trends in AI right now is the closing of the Frontier.
因为我现在确实认为，当前 AI 领域的一个重大宏观趋势，就是 Frontier 正在关闭。
61:12
It's the fact that the very best models, like mythos from Anthropic,
事实就是，最好的那些模型，比如 Anthropic 的 mythos，
61:17
are not being open to everybody.
并没有向所有人开放。
61:20
That is both a safety decision,
这既是一个安全决策，
61:22
as in, we don't want to give really powerful cyber weapons and bio weapons to a bunch of bad actors or just unknown actors.
也就是说，我们不想把非常强大的 cyber weapons 和 bio weapons 给一堆坏人或不明身份的人。
61:28
It is also a pricing question of,
这也是一个定价问题，
61:30
the best models are really, really expensive to run.
最好的 models 运行起来真的非常非常贵。
61:33
They don't have enough compute to run them.
他们没有足够的 compute 来运行它们。
61:35
And so we're going to have to charge a lot of money or only give it to the biggest corporations.
所以我们要么收很多钱，要么只把它给最大的公司。
61:40
And I think that's the reason that startups are worried,
我觉得这就是初创公司担心的原因，
61:44
that countries outside of the US are worried, that normal people are worried.
美国以外的国家担心，普通人也担心。
61:50
Maybe we get superintelligence and it can achieve all these amazing things,
也许我们得到了 superintelligence，它能实现所有这些惊人的事情，
61:54
but I'm not going to get it.
但我不会得到它。
61:56
Maybe my boss is going to get the superintelligence
也许我的老板会得到那个 superintelligence，
61:58
and they're going to automate my job as a worker or as a consumer.
然后他们会把我的工作自动化，不管我是作为工人还是作为消费者。
62:01
I'm not going to be able to do the same thing.
我没法做同样的事情。
62:03
So I think it's also a really good point that if we don't continue the compute buildout,
所以我觉得这也是一个很好的观点：如果我们不继续 compute buildout，
62:07
we do see a world where it is the folks with existing capital and access,
我们确实会看到一个世界，在那里，是那些拥有现有资本和 access 的人，
62:11
probably big corporations in the US and the US government,
很可能是美国的大公司以及美国政府，
62:14
that are going to have access to Frontier AI and all the benefits that it confers.
他们将能够获得 Frontier AI 以及它所带来的一切好处。
62:19
AWS AI is how the world's leading organizations are transforming their industries,
AWS AI 就是全球领先的组织正在变革各自行业的方式，
62:49
not in theory, but in production, at scale, right now,
不是理论，而是在 production 中、规模化地、就在此刻，
62:53
like turning race day data into real-time insights for Formula One fans,
比如把比赛日数据变成 Formula One 车迷的实时洞察，
62:57
reinventing the journey for every united traveler,
为每一位 united traveler 重新定义旅程，
63:01
and personalizing every trip with booking.com.
并通过 booking.com 让每一段旅程都个性化。
63:04
However big the question for your business or industry,
无论你的业务或行业面临多大的问题，
63:07
AWS AI is how?
AWS AI 就是答案？
63:10
The New York Times app unlocked?
The New York Times app 解锁了？
63:13
Everyone knows the times is behind a paywall.
大家都知道 the Times 是在 paywall 后面的。
63:16
Only subscribers have access to all the reporting.
只有订阅者才能看到全部报道。
63:20
But what if you could explore the times for a month, for free,
但如果你能免费探索 the Times 一个月呢，
63:23
without putting in a credit card?
还不用输入信用卡？
63:25
Now you can.
现在你可以了。
63:27
When you download the New York Times app for the first time,
当你第一次下载 New York Times app 时，
63:29
your first month in the app is free, a month to go behind the paywall,
你在 app 里的第一个月是免费的，这一个月可以走到 paywall 后面，
63:34
to see what time subscribers get every single day.
看看 Times 订阅者每天都能看到什么。
63:37
All the investigations, the reviews, the recipes,
所有调查报道、评论和食谱，
63:40
the deeply reported fact-based journalism.
以及那些深度报道、基于事实的新闻。
63:43
If you don't already subscribe to the New York Times,
如果你还没订阅 New York Times，
63:46
download the times app today and get free access for 30 days.
今天就下载 Times app，即可免费访问 30 天。
63:50
You were in China for a trip reporting on AI.
你去中国做了一趟关于 AI 的报道之旅。
63:57
Was there much political AI backlash and ferment there from what you could see?
从你能看到的情况来看，那里对 AI 的政治反弹和发酵多吗？
64:03
I was super interested in this question on this trip
这趟旅行中，我对这个问题特别感兴趣
64:05
because I was finishing my times piece on the permanent underclass while in China
因为我在中国期间，正在完成我给 Times 写的那篇关于永久底层阶级的稿子。
64:09
and so I was basically asking everyone I met there,
所以基本上，我在那儿遇到的每个人我都会问，
64:11
whether it was engineers at the labs or just my family members
不管是实验室里的工程师，还是只是我的家人
64:14
or sort of normal middle class people in Shanghai.
又或者说是上海那种普通的中产。
64:16
Are people worried about AI and jobs?
人们担心 AI 和就业吗？
64:18
Are people worried about AI and social instability?
人们担心 AI 和社会不稳定吗？
64:22
I think the answer is not as much.
我觉得答案是，并没有那么担心。
64:24
I caveat this, of course, with the fact that the information
当然，我这么说有个前提，就是信息
64:27
environment in China is obviously suppressed.
环境在中国显然是受到压制的。
64:29
You can't dissent in public on social media nearly as much as you can in the US.
你没法像在美国那样，在社交媒体上公开发表那么多不同意见。
64:34
You don't have good polling.
你也没有好的民调数据。
64:35
So it's kind of hard to understand the actual level of social discontent
所以有点难了解社会不满情绪的实际程度，
64:38
there is in China.
在中国到底有多高。
64:39
But I would say that for the most part,
但我会说，大体上来说，
64:41
people were not as terrified of AI as they are in the US.
人们并不像在美国那样害怕 AI。
64:45
There's a few explanations for this.
对此有几种解释。
64:47
Some people say that China is more technoloptimistic than the US's.
有些人说，中国比美国更 technoloptimistic。
64:50
I don't love this explanation mostly because the thing that I heard was not exactly optimism.
我不太喜欢这个解释，主要是因为听到的那个说法并不完全是乐观。
64:55
It was not exactly, yeah, we're going to get the cancer cures and the superabundance.
它不完全是那种“是的，我们会得到癌症治愈方法，迎来极度富足”的调子。
64:58
It was something a lot closer to technology as a force that cannot be stopped.
它更接近于把技术看作一种无法被阻止的力量。
65:03
Actually in some ways reminded me more of some of these Silicon Valley beliefs
其实从某些方面说，它更让我想起 Silicon Valley 的一些信念，
65:06
that the future is predetermined that when the state decides that something like AI is a national priority,
就是未来是预先注定的，当国家认定像 AI 这样的东西是国家优先事项时，
65:12
that is going to march forward.
它就会一路向前推进。
65:13
And as an individual, there's not much you can do to resist, especially in a one-party state
而作为个人，你没什么办法去抵抗，尤其是在一党制国家，
65:17
in a authoritarian society.
在威权社会里。
65:19
There is no culture of resistance, really.
真的没有什么抵制文化。
65:21
And so rather than thinking about how do I prevent AI in my workplace or in the world,
所以，与其去想我要怎么在工作场所或全世界阻止 AI，
65:27
that's not really a thing that a lot of people in China think about.
这并不是很多中国人会考虑的事情。
65:30
It's how can I use AI to make sure I don't fall behind?
他们想的是，我要怎么利用 AI 来确保自己不落后？
65:33
In an environment that already has crazy levels of white color competition
在一个白领竞争已经疯狂到不行的环境里，
65:37
and white color unemployment, if you're not upskilling yourself with open claw or whatever,
再加上白领失业，如果你不用 open claw 或者别的什么来提升自己，
65:43
there's a million people online behind you who are going to get on the bus.
网上有一百万人排在你后面，准备上车。
65:46
At the same time, I think that the Chinese state takes a pretty different approach to the US.
与此同时，我觉得中国政府对待这件事的方式和美国相当不同。
65:50
When it comes to AI regulation and also to technology regulation in general.
说到 AI 监管，以及一般意义上的技术监管。
65:54
And so China has past laws sort of banning a lot of kinds of companion chatbots
所以中国已经通过法律，差不多禁止了很多种 companion chatbots
65:58
because they're worried about relationships, they're worried about fertility rates.
因为他们担心人际关系，担心生育率。
66:02
They're worried about addiction.
他们担心成瘾。
66:05
China has past made court rulings that say that AI replaced this worker's job.
中国过去做出过法院判决，说 AI 取代了这个工人的工作。
66:11
AI can do this worker's job is not a good enough reason to lay off a worker.
AI 能做这个工人的工作，并不足以成为裁掉一个工人的理由。
66:15
You have regulations that require all AI generated images to be labeled.
有些规定要求所有 AI 生成的图像都必须标注。
66:20
And you'll see the made with AI sort of language on all of the AI made ads in China.
而且在中国所有 AI 制作的广告上，你都会看到 made with AI 这类字眼。
66:24
And so there's also a sense that some people, some Chinese have,
所以也有一种感觉，就是有些中国人觉得，
66:28
their government is more likely to look out for sort of the social downsides and the labor downsides relative to the US government,
他们的政府跟美国政府比起来，更可能会去关注那种社会层面的负面影响和劳动力方面的负面影响，
66:36
which has thus far been pretty less a fair especially at the national level.
而美国政府到目前为止，尤其是在国家层面，一直没那么公平。
66:40
And that gives some people a bit of solace as well.
这也让一些人多少感到一点安慰。
66:43
There's been some reporting that China and Russia are pushing sort of anti-datacenter memes and social media bots.
有一些报道说，中国和俄罗斯在推那种反 data center 的梗和社交媒体 bot。
66:54
It's hard for me to tell what scale that is, but it has been very much picked up on.
我很难判断那到底有多大规模，但这件事确实已经被很多人捡起来说了。
66:58
But people like Kevin O'Leary, the Shark Tank guy whose big data center project is faced a lot of backlash.
但像 Kevin O'Leary 这样的人，就是 Shark Tank 那个家伙，他的大型 data center 项目遭遇了很多反对。
67:06
Do you buy the sort of growing view among at least some tech elite that the data center backlash is like a Chinese sign-up?
至少在一些 tech elite 里，那种越来越普遍的看法——说对 data center 的反弹像是中国搞的 sign-up——你买账吗？
67:13
I think this is ridiculous to be honest.
说实话，我觉得这挺荒谬的。
67:15
I mean, so I read the Open AI report that was saying this is all a CCP plot.
我是说，我读了 Open AI 那份报告，说这都是 CCP 的阴谋。
67:20
And like it paces in the accounts and the tweets that are doing this side off.
而且就像，它出现在那些账号和推文里，那些账号和推文就是在做这一边的事。
67:25
These tweets have like no likes on them. They have two views per tweet.
这些推文几乎没什么点赞。每条推文只有两个浏览量。
67:28
So I'm not doubting that like I'm sure some clever CCP propaganda person has attempted to inflame the anti-datacenter sentiment.
所以我不是在怀疑这一点，我确信有些聪明的 CCP 宣传人员试图煽动反 datacenter 情绪。
67:37
I have not seen evidence that any of this is working.
我没有看到证据表明这些起了作用。
67:40
I think it feels very organic. I think I also tend to be personally a little suspicious when you just castle your political opponents as being misinformed.
我觉得这感觉非常自然、非常自发。我觉得，当你直接把政治对手说成是被误导时，我个人也倾向于有点怀疑。
67:47
I think there's a way in which people use foreign influence to avoid thinking about the fact that there are people that they live with in society who do not agree with their vision of the world.
我觉得，人们会用境外影响来回避一个事实：社会中跟他们一起生活的人，并不认同他们对世界的愿景。
67:57
And when I talk to these data center activists, for example, they are actually much less tick-tock adult and misinformed than I think people like to caricature.
而且，比如我跟这些 data center 活动人士聊的时候，他们其实远没有人们喜欢讽刺成的那么被 TikTok 搞昏头、那么被误导。
68:04
A lot of them understand sort of the basic facts of what's the difference between an AI data center and the old kind of data center.
他们很多人大概都懂最基本的事实，也就是 AI data center 和过去那种 data center 有什么区别。
68:10
How much was the difference between a closed loop system and an open loop system.
以及 closed loop system 和 open loop system 之间的差别到底有多大。
68:13
Mostly people are not just misinformed. They have just personally decided.
多数情况下，人们不只是被误导了。他们只是自己个人做了决定。
68:17
I'm not that interested in having a data center in my community even if it pays some property taxes.
我没那么想让我的社区里有个 data center，哪怕它能交一些房产税。
68:22
So you talk to people in the AI companies and sort of talk to them about this backlash.
所以你会跟 AI 公司里的人聊，算是跟他们谈谈这种反弹？
68:28
I know they're very worried about this rather talk to them about this.
我知道他们非常担心这件事。更确切地说，去跟他们聊这件事。
68:31
What are they learning from it?
他们从中学到了什么？
68:35
I do think that this year, in 2026, AI executives, AI researchers have started to take the public backlash a lot more seriously than they have in the past.
我确实觉得，今年，也就是 2026 年，AI 高管、AI 研究人员已经开始比过去认真得多地对待公众反弹了。
68:48
I've heard executives ask, can we do better marketing? I don't understand why it is that waymos are so unpopular.
我听到一些高管问：我们能不能把营销做得更好？我真不明白为什么 waymos 这么不受欢迎。
68:56
I've heard executives ask, what do you think are the deals that we should be making?
我听到一些高管问：你觉得我们应该做哪些交易？
69:01
Do you think we should just be mailing checks to every house that lives near a data center project? Will that fix things?
你觉得我们就该给住在 data center 项目附近的每家每户寄支票吗？这样就能解决问题吗？
69:08
Tell us how to make a better deal. Do we need to cut people's electricity prices in half with that work?
告诉我们怎么才能做成一笔更好的交易。我们是不是得靠这个把人们的电费砍一半？
69:15
Unfortunately, the word bribe gets used a lot more than I'm personally comfortable with.
不幸的是，“贿赂”这个词被用得比我个人能接受的多得多。
69:19
I think that when you are framing the thing you're doing, even jokingly, as bribing communities into putting a data center there, I don't think you're starting off on the right foot.
我觉得，当你把你正在做的事——哪怕是开玩笑地——包装成是在贿赂社区，让他们把 data center 放在那里时，我不觉得你一开始就走对了路。
69:27
I think that people can feel when they are being bribed. I've heard these people say these companies are bribing us.
我觉得人们能感觉到自己正在被贿赂。我听到这些人说，这些公司就是在贿赂我们。
69:34
There's a little bit of examination. The thing that I don't think is being examined as much as I want it to be though is, are we building a technology?
已经有一些审视了。但我觉得，有一件事没有像我希望的那样被充分审视，那就是：我们到底是在造一项技术吗？
69:43
Are we building a product that is helping people?
我们是不是在做一个能帮到人的产品？
69:46
The version of this I've heard is we have a marketing problem.
我听到的一种说法是，我们有个营销问题。
69:50
Maybe we should stop saying aloud so often that our technology might take everybody's job and has a 10% chance of upending or destroying humanity altogether.
也许我们不该再那么频繁地公开说，我们的技术可能会抢走所有人的工作，还有 10% 的概率彻底颠覆或毁灭人类。
70:05
What has not been clear to me, even as they begin to move away from that stitching a little bit, is whether or not they no longer believe that.
让我一直没看明白的是，即便他们开始稍微不再那么坚持那套说法，他们到底是不是已经不再相信那件事了。
70:17
I am my personal view is like, I don't think it's going to take everybody's job.
我，我个人的看法是，我不觉得它会抢走所有人的工作。
70:21
But to the extent they do or at least they take that very, very seriously, I keep hearing them say we have a marketing problem.
但就算他们真的这么认为，或者至少他们非常非常认真地对待这件事，我还是总听他们说，我们有个营销问题。
70:28
I keep saying when I talk to them about this, that if you believe the things you have been saying and the fact that things you have told me personally, you don't have a marketing problem.
我跟他们聊这个的时候一直说，如果你相信你一直在说的那些话，还有你私下跟我说的那些事，那你就没有营销问题。
70:36
You have a problematic technology like you have a product problem because people are not going to want that future.
你有一个有问题的 technology，就像你有一个产品问题，因为人们不会想要那样的未来。
70:44
I think I would make some distinctions between different companies, different executives.
我觉得我会区分一下不同的公司、不同的高管。
70:50
One thing that's very interesting to me thinking about the comms in Silicon Valley is that for a very long time, these companies were only marketing to potential recruits and potential investors basically.
对我来说，想到 Silicon Valley 的 comms，有一件很有意思的事：很长时间里，这些公司基本上只在对潜在招聘对象和潜在投资人做 marketing。
71:01
They were trying to win the vibes on AI Twitter in San Francisco and I kind of feel like they never realized everyone else could hear them.
他们当时想在 San Francisco 的 AI Twitter 上赢下 vibes，而我有点觉得，他们从来没意识到其他所有人都能听到他们。
71:09
Now they're trying to take it back, but my sense is that Sam Altman for a long time, part of the reason he was talking about shifting the balance of power from Labour to Capital and Rogay Eye and whatever was also because he was winning points among people who wanted to work out an AI.
现在他们想把它收回来，但我的感觉是，Sam Altman 在很长时间里，之所以会谈论把权力平衡从 Labour 转向 Capital、Rogay Eye 什么的，部分原因也是他想在那些想做出 AI 的人那里赢得加分。
71:22
He had a communicate that he was as agipodism. He was as worried about the same safety things as them.
他得传达出自己也是 agipodism。他也跟他们一样担心那些 safety 的事情。
71:27
Now that his interest is more in political goodwill and IPO-ing and things like that, he sort of changes tune.
现在他的兴趣更多在政治好感和 IPO 之类的事情上，他也就有点改口了。
71:35
I think that, yeah, I'm not sure to what extent every AI industry actor has always believed the things that they've warned about.
我觉得，嗯，我不确定每个 AI 行业里的人是不是都一直相信他们警告过的那些东西。
71:44
I think people can believe things that they don't feel, if that makes sense.
我觉得人是可以相信自己并没有真正感受到的东西的，如果这话说得通的话。
71:49
And I think a lot of people in the AI industry are in a culture and inside arguments where this set of outcomes feels very real or looks very real.
而且我觉得 AI 行业里很多人所处的文化和争论环境，会让这一系列结果感觉非常真实，或者看起来非常真实。
72:04
And so I think they believe it. I think when they make these arguments, I don't think they're just doing it for publicity points.
所以我觉得他们是相信这一点的。我觉得当他们提出这些论点时，我不认为他们只是为了博眼球才这么做。
72:10
In fact, I think it's the opposite. I think when they're now trying to move away from some of these arguments, I think it's actually much more of like a cynical marketing ploy.
事实上，我觉得恰恰相反。我觉得现在当他们试图远离其中一些论点时，那其实更像是一种犬儒式的营销手段。
72:17
But I think they often believe these things without actually in their bones, feeling it, which is why they act relatively heedlessly or at least on the set of things they believe.
但我觉得他们常常相信这些事，却没有真正从骨子里感受到它，所以才会行动得比较轻率，或者至少是依据他们相信的那套东西来行动。
72:31
This speculative, notional set of beliefs about what might happen is way less close to their core than their belief that if they don't build this data center or get this next model out, their competitors or China or somebody, is it going to get in front of them?
这种关于可能发生什么的、揣测性的、概念性的信念，远没有他们下面这个信念那么贴近他们的核心：如果他们不建这个 data center 或者不把下一个 model 推出来，他们的竞争对手、China 或某个人就会抢到他们前面。
72:45
And they're much more motivated by the push forward.
而且他们更大的动力其实是向前推进。
72:49
I think the technological determinism is just such a big part of it. I think that if I'm trying to think about how my friends in the AI industry would react to this conversation, that's the thing that they would say that we are not focusing on enough.
我觉得 technological determinism 就是其中很大的一部分。我觉得如果我去想我在 AI 行业的朋友们会对这场对话有什么反应，那正是他们会说我们没有足够关注的东西。
72:58
Is they are so sure that there is no way that AGI are super intelligence or whatever it is does not get built and it is only a question of who builds it.
就是他们如此确信，AGI 或 super intelligence 或不管它是什么，根本不可能不被造出来，唯一的问题只是谁来造。
73:06
I think that fundamental sort of underlying belief is what justifies everything else is we have to be the ones to do it.
我觉得，那种根本的、底层的信念，就是用来为其他一切辩护的，那就是必须由我们来做这件事。
73:13
Us pulling back, us stopping is not going to prevent any of the bad stuff.
我们退缩、我们停下，并不能阻止任何糟糕的事情发生。
73:16
I agree with that. And I think that's why the China card in this has been such like a destructive part of the argument.
我同意这一点。而且我觉得，这就是为什么这里面的 China card 一直是这个论证中如此具有破坏性的部分。
73:23
I'm not even sure it's totally untrue like I am completely willing to believe that China and America are in a race for a economically and geopolitically important technology.
我甚至不确定这完全是假的，我完全愿意相信，China 和 America 正在为一项在经济和地缘政治上重要的技术展开竞赛。
73:32
Even if you don't buy like recursive super intelligence.
即使你不买账，比如 recursive super intelligence 这种东西。
73:36
But the way that has been used to not say well, we should enter into international negotiations or something.
但它被用来做的，不是说“好吧，我们应该进入国际谈判之类的”。
73:44
Instead just like we cannot sew down whatsoever no matter what else we worry about or believe I think has been it is acted as a kind of blackmail.
相反，它就像是在说，不管我们还担心什么、相信什么，我们都绝不能放慢；我觉得，这一直以来都像是一种勒索。
73:53
And the thing is that it's not bought by enough people outside of the industry.
而且问题是，行业外还没有足够多的人买账。
74:00
But I think the phase of the politics around now in is out of their control.
但我觉得，现在这波政治已经不受他们控制了。
74:04
And it is just not going to be the case that they have control over the AI narrative like next year that they had.
而且明年他们不可能还拥有他们曾经有过的那种对 AI 叙事的掌控。
74:13
Two years ago. And yeah, I don't really think they know what to do in that space.
两年前。而且，是的，我真不觉得他们知道在那个领域该怎么做。
74:18
And so now it's like either going to have to start benefiting people.
所以现在的情况就是，它要么就得开始让人们受益。
74:23
Right. Some of these like if people begin seeing like drug cures come out right all these things we've actually been promised.
对。其中一些，比如如果人们开始看到药物治愈方法问世，对吧，所有这些我们其实一直被承诺过的东西。
74:29
You could say like we're meeting to see the beginning of like mathematical conjectures like that's been pretty cool.
你可以说，我们聚在一起，就是为了见证 mathematical conjectures 这类东西的开端，那还挺酷的。
74:34
But we're not really seeing the gains. And if you start getting the losses before the gains.
但我们并没有真正看到收益。而如果你在收益到来之前就开始承受损失。
74:40
Right. You start getting the job loss for instance before the promised you know super abundance.
对。比如说，在承诺的、你知道的超级丰裕之前，你就开始遭遇失业了。
74:45
Politically, that's not going to be an equilibrium. You can protect.
从政治上说，那不会是一种均衡状态。你可以保护。
74:48
That's one of the things I'm worried about is I do think that we are pretty likely to see we are seeing a lot of the social instability before we get the cancer cures.
这就是我担心的事情之一：我确实觉得，我们很可能在拿到癌症治愈方法之前，就会看到——其实已经在看到——大量社会不稳定。
74:57
Right. And so or even with the math stuff it's like I think that one thing I notice more and more now is this deep cultural and values gap between Silicon Valley and the rest of America, the rest of the world.
对。所以，或者说即便是在数学那些东西上，也像这样：我觉得我现在越来越注意到的一件事，就是 Silicon Valley 和美国其他地方、世界其他地方之间存在的那种深深的文化和价值观鸿沟。
75:08
I'm not saying the Silicon Valley is wrong that it's cool to like disprove the Jacobian conjecture right like it is cool.
我不是说 Silicon Valley 错了，说推翻 Jacobian conjecture 很酷，对吧，确实很酷。
75:14
But like when you ask a lot of people in the tech industry.
但就像当你问很多科技行业里的人。
75:18
What their utopia looks like they'll say things like we have UBI so no one has to work anymore.
他们的乌托邦是什么样，他们会说，比如我们有 UBI，所以没人需要再工作了。
75:23
We're all immortal and we've discovered all of math and physics.
我们全都永生了，而且我们已经发现了所有的数学和物理。
75:26
And if you do the pulling on UBI and mortality neither are especially popular with the American public.
而如果你对 UBI 和死亡率做民调，这两个在美国公众中都不太受欢迎。
75:32
I think that we pull the mortality is a funny.
我觉得我们民调死亡率这件事挺好笑。
75:35
I mean, it's been pulled. You can look it up.
我是说，它已经被民调过了。你可以去查。
75:38
If you ask people what like they want AI to do for them.
如果你问人们，他们想让 AI 为他们做什么。
75:42
It's not necessarily having you know Karen's towel in your pocket.
不一定是让你，你知道，把 Karen 的毛巾装在自己口袋里。
75:46
It's not disproving math right like they want things to be cheaper.
不是去推翻数学，对吧，他们想要的是东西更便宜。
75:50
They want to be healthier. They want to not do crappy works.
他们想更健康。他们想不用干烂活儿。
75:53
They have more time for the stuff they like.
他们有更多时间做自己喜欢的事。
75:55
But I think it is genuinely true that the stuff that is really cool.
但我觉得，真正酷的那些东西，这确实是真的。
75:59
Also that is oftentimes technically easier to solve like math is not what most people want from this technology.
而且那些东西往往在技术上更容易解决，比如数学，并不是大多数人想从这项技术里得到的东西。
76:08
And I think it's also true that it's just literally technically harder to cure cancer than it is.
而且我觉得，同样没错的是，治愈癌症在技术上就是比这更难。
76:13
It turns out to prove math theorems and the other thing that I hear from the public when I talk about the cancer curators is.
结果就是，证明数学定理，还有当我谈到 cancer curators 时从公众那里听到的另一件事，是：
76:21
Yeah, but are they going to just use it for themselves is Peter teal or whoever just going to buy himself a mortality.
对，但他们会不会就自己用了？Peter teal 或者随便谁，会不会直接给自己买一个 mortality。
76:27
Am I going to be able to afford a mortality?
我能负担得起一个 mortality 吗？
76:29
My view for a very long time has been that a lot of people in these companies overrate how much of the bottleneck in scientific and human progress is raw intelligence.
很久以来，我的观点一直是，这些公司里的很多人高估了 raw intelligence 在科学和人类进步瓶颈中占的比重。
76:40
Yes. Yeah.
对。
76:41
And just I mean, this is a point of abundance.
而且我的意思是，这就是一个富足的节点。
76:43
It's just the point of covering anything anywhere.
它就是一个能覆盖任何地方、任何东西的节点。
76:46
The world is full of friction.
这个世界充满了摩擦。
76:48
Yeah.
对。
76:49
And you know, you want to do drug discovery and I think we should actually do a lot to make drug discovery easier.
而且你知道，你想做 drug discovery，我觉得我们其实应该做很多事，让 drug discovery 变得更容易。
76:54
Make drug testing easier, right?
让 drug testing 变得更容易，对吧？
76:56
Like I have said this many times before I would like to see us prepare like drug development before a world where AI is spitting out way more promising molecular candidates.
就像我之前说过很多次的那样，我希望看到我们把 drug development 提前准备好，去迎接一个 AI 能吐出多得多的、有前景的 molecular candidates 的世界。
77:08
But that's still a world where you need it off monkeys to test things on humans to test things on rats to test things on right and you still need to do all the safety data.
但那仍然是一个你需要在猴子身上测试、在人身上测试、在大鼠身上测试的世界，对吧？而且你仍然得做所有的 safety data。
77:16
And just the amount of the world that is slowed down by we don't have any good ideas.
而且，世界上有多少事情被拖慢，仅仅是因为我们没有什么好点子。
77:21
Like we are out of ideas versus it is hard to organize things amidst humans, you know, amidst, you know, with raw materials like in bureaucracies in organizations you get that like intelligence is important.
就像是我们没想法了，还是在人类之间、在原材料之间把事情组织起来很难，你知道，在官僚体系里、在组织里，你会发现 intelligence 很重要。
77:36
But it is not everything and I think anybody who's like been in organizations like knows it's actually less than you think it is.
但它不是一切，我觉得任何在组织里待过的人都知道，它其实比你想的要少。
77:45
Yeah. Again, I think of a, you know, a lot of these people have been AI researchers for their entire careers, maybe before that they were physics PhDs or they were doing quant trading, which are all these kinds of jobs that are fairly their icy jobs are individual contributors where you're not necessarily working in big teams.
对。再说一次，我想，你知道，这些人很多整个职业生涯都是 AI 研究员，也许在那之前他们是物理学博士，或者做 quant trading，这些都是那种挺 IC 的工作，是 individual contributors，你不一定是在大团队里工作。
78:02
So there's not a lot of politicking and relational work where all of the relevant context lives inside a single code base.
所以没有太多政治操作和人际关系上的活儿，因为所有相关的 context 都在一个 code base 里面。
78:08
And so for AI to sort of understand what's going on and kind of explore all this context that's already been written down.
所以对 AI 来说，它有点能理解发生了什么，也能去探索所有这些已经被写下来的 context。
78:14
I'm not saying there's no tacit knowledge, but a lot more of the context is made explicit.
我不是说没有 tacit knowledge，但多得多的 context 是被明确表达出来的。
78:18
And these are also places where simply applying more thinking and more intelligence just as an individual like as a person a remote worker in a closet or whatever might actually find the more efficient algorithm.
而且这些地方也是，你只要作为一个个体，比如一个在壁橱里远程工作的人之类的，投入更多思考和更多 intelligence，可能真的就能找到更高效的 algorithm。
78:29
Right, like you don't actually need to politic your way to a better algorithm.
对，就像你其实不需要靠搞政治手段来得到一个更好的 algorithm。
78:32
You don't need to do stuff in the physical world to get that.
你不需要在物理世界里做什么就能得到它。
78:35
And so I think a lot of people at these companies don't really realize how hard that is.
所以我觉得这些公司里的很多人并没有真正意识到那有多难。
78:41
I mean, it's funny because people will say things like, yeah, there's like electricity cost and energy cost to AI, but like AI will maybe solve the climate.
我的意思是，这挺好笑的，因为人们会说，对，AI 是有电费成本和能源成本，但 AI 也许能解决气候问题。
78:50
And I ask how and to be clear, I think there are a lot of ways that AI can improve, you know, climate science research help building efficiency.
然后我问怎么解决，而且说清楚一点，我觉得 AI 有很多方式可以改进，你知道，climate science research，帮助提高建筑效率。
78:58
Yeah, absolutely, but like at the same time you ask people and it's just like, oh, I don't know, it's just going to do it, right?
对，绝对是这样，但就像同时你问人们，他们就只是，哦，我不知道，反正它会做到的，对吧？
79:03
Or it's like, oh, how is AI going to improve robotics?
或者就像，哦，AI 要怎么改进 robotics？
79:06
Like, I don't know, AI will figure it out.
就像，我不知道，AI 会自己搞定的。
79:08
You kind of do have this, I find it lazy actually.
你确实有点这个意思，其实我觉得这挺懒的。
79:11
Like I, one of the things that annoys me about this particular approach is it's not that I don't think that AI can contribute to all of these problems.
就是，这种特定做法让我烦的一点在于，我并不是觉得 AI 不能帮助解决所有这些问题。
79:17
I think it definitely can. But what I often hear is a kind of laziness about how it's going to do that.
我觉得它肯定能。但我经常听到的，是在“它要怎么做到”这件事上的一种偷懒。
79:22
And it feels like a Deus Ex Machina of it's super smart. It'll just figure it out.
而且这感觉像是一种 Deus Ex Machina——它超级聪明，它自己就会搞明白。
79:27
I used to say that this was back when, you know, Silicon Valley was a more optimistic place than it has been in recent years.
我以前常说，那还是在，你知道，Silicon Valley 比近几年更乐观的时候。
79:32
But the difference between the culture of DC where I live for a long time and of Silicon Valley.
但 DC 的文化——我在那儿住了很久——和 Silicon Valley 的文化之间的区别。
79:38
Was it, in Silicon Valley, people's worldview is formed by seeing impossible problems prove possible to solve.
是这一点：在 Silicon Valley，人们的世界观是由看到不可能的问题被证明可以解决而形成的。
79:48
And in DC people's worldview is formed by seeing possible problems prove impossible to solve.
而在 DC，人们的世界观是由看到可能的问题被证明无法解决而形成的。
79:55
And I think that is now going to collapse for the AI industry into one worldview because, you know, these are people who, like give them their due, they've invented artificial intelligence.
我觉得现在对 AI 行业来说，这会坍缩成一种世界观，因为，你知道，这些人，平心而论，是他们发明了 artificial intelligence。
80:05
They actually did it. You know, this is amazing. Like I cannot believe how good some of these systems are.
他们真的做到了。你知道，这太惊人了。我简直不敢相信其中一些系统能这么好。
80:11
I'm like shocked to be living through this. They were able to do that. That seemed impossible, prove possible.
我简直震惊自己能活在这个时代。他们居然做到了。那看起来不可能，结果被证明是可能的。
80:16
And now they're going to, now they're finding it's like impossible to build a data center.
而现在他们要——现在他们发现，建一个 data center 简直是不可能的。
80:20
Yeah.
是啊。
80:21
And like that's what doing other kinds of things in the world teaches you.
而这就像，你在现实世界里做其他类型的事情会教给你的东西。
80:26
Yes.
对。
80:27
There are a lot of problems that are not possible to solve, not because you cannot come up with the idea for them.
有很多问题是不可能解决的，不是因为你没法想出解决它们的点子。
80:33
But because you are dealing with the messy realities of societies, of politics, of values, of logistics.
但因为你面对的是社会、政治、价值观、后勤这些乱糟糟的现实。
80:42
And it'll demand a kind of, it'll impose a kind of realism, I think, on the issues that it does not always have.
而且我觉得，它会要求一种——它会强行给这些问题注入一种它们并不总是有的现实感。
80:52
Yeah. I was trying to think about what the difference was between how I would describe Silicon Valley and San Francisco culture a year ago, let's say early 2025 versus now.
对。我一直在想，一年前，比如说 2025 年初，我会怎么描述 Silicon Valley 和 San Francisco 的文化，跟现在相比到底有什么不同。
81:02
And I think the number one thing is that Silicon Valley has really woken up to politics.
而且我觉得，头号变化就是，Silicon Valley 真的开始意识到政治的重要性了。
81:06
You know, in January 2025, Silicon Valley was feeling very triumphant about, about doge, about Elon Musk, about David Sachs and Jerome and the White House.
你知道，在 2025 年 1 月，Silicon Valley 还对 doge、对 Elon Musk、对 David Sachs 和 Jerome 以及 White House 感到非常得意。
81:15
They kind of felt like they were all in control. And actually, if you just build these genius technologies and you get super rich and you have good ideas, you'll just, you know, get the political power to enact your vision.
他们有点觉得自己掌控了一切。而实际上，如果你只要造出这些天才技术、变得超级有钱、又有好想法，你就会，你知道，获得政治权力来实现你的愿景。
81:26
And a year and a half later, a lot of those folks are out of the White House.
而一年半之后，这些人里有很多已经不在 White House 了。
81:30
They failed at reducing the national debt and achieving all these other goals that they thought they could just AI their way into solving.
他们没能减少国家债务，也没能实现其他那些他们以为自己只要靠 AI 就能一路解决的目标。
81:36
Anthropic, for example, has had a lot of problems and its dealings with the Trump administration, fundamentally very political and very relational problems.
Anthropic，比如说，在跟 Trump 政府打交道时遇到了很多问题，本质上是高度政治化、非常讲关系的问题。
81:46
Dario's problem in dealing with the White House was not, I think, that he didn't have good arguments or that he's not very smart or not saying logical things.
我认为 Dario 在跟白宫打交道时的问题，不在于他没有好的论据，也不是因为他不够聪明，或者讲的东西没有逻辑。
81:54
I think that anyone from Anthropic will admit that these are largely relational problems.
我觉得 Anthropic 的任何人都会承认，这些很大程度上是关系问题。
81:58
And so there's a way where I think democracy and politics is a lot more powerful than these very rich and very smart tech people realize.
所以从某种意义上说，我认为民主和政治的力量，比这些非常有钱、非常聪明的科技人士意识到的要大得多。
82:08
And there's some optimism to that, I think, and in looking at it and saying it's actually really hard to buy an election, it's actually really hard to buy out the whole White House at once.
而且我觉得这里面也有一点乐观：仔细想想，其实很难买下一场选举，也很难一次性买通整个白宫。
82:18
But it's an interesting moment, I think, for the tech industry to be realizing how important politics really is and how difficult it is.
但我觉得这是一个很有意思的时刻，让科技行业意识到政治到底有多重要，以及它有多难。
82:26
I think that's a good place to end. Always a final question. What are three books you'd recommend to the audience?
我觉得这是个很好的收尾点。照例问最后一个问题。你会给听众推荐哪三本书？
82:31
Oh, so I think the first one is really relevant to this conversation, which is Benjamin Lobitudes, the maniac, which includes a sort of lightly fictionalized biography of John Von Neumann, the story of AlphaGo.
哦，我觉得第一本跟这次对话特别相关，就是 Benjamin Lobitudes 的《the maniac》，里面包含了某种轻度虚构化的 John Von Neumann 传记，还有 AlphaGo 的故事。
82:42
I think it's very much a sort of halfway novel, halfway non-fiction book about how intelligence is incredibly onspiring and something worth respecting
我觉得它非常像一本一半是小说、一半是非虚构的书，讲的是 intelligence 有多么令人振奋、多么值得尊重
82:54
and at the same time can lead people to some very dark realities.
同时又会把人带向一些非常黑暗的现实。
82:59
My second book is the technology trap from Carl Benedict Ray, which I think is very much about how people's attitudes towards technology and automation depend on to what extent the benefits, the economic growth is shared to what extent they feel like they're getting a piece of the pie.
我的第二本书是 Carl Benedict Ray 的 The Technology Trap，我觉得它很大程度上讲的是，人们对 technology 和 automation 的态度取决于好处、经济增长在多大程度上被共享，以及他们在多大程度上觉得自己分到了一杯羹。
83:17
It goes through a lot of history much more than just the Industrial Revolution.
它讲了很多历史，远不只是工业革命。
83:21
And so that's shaped a lot of my thinking on some of the economic questions and the populist questions.
所以这很大程度上塑造了我对一些经济问题和民粹主义问题的思考。
83:26
And then finally, pre-aperkers that are of gathering.
然后最后是 Priya Parker 的 The Art of Gathering。
83:30
Because I do think that the relational stuff is going to become a lot more important to always was.
因为我确实认为，关系层面的东西会变得重要得多，比以往任何时候都重要。
83:35
And I do think that book has helped me become a better host.
而且我确实认为那本书帮我成为了一个更好的主持人。
83:39
She would be so happy to hear that people should go check out our conversation with the pre-aparker.
她要是听说大家都该去听听我们跟 pre-aparker 的那段对话，肯定会特别开心。
83:43
Jasmine Sun, thank you so much.
Jasmine Sun，太感谢你了。
83:45
Thank you so much for having me. This is fun.
非常感谢你邀请我来。这挺有意思的。