参议员马克·沃纳：没人准备好应对人工智能可能对我们造成的影响

如果人工智能的进步确实在呈指数级发展，我们做好准备了吗？

If AI progress is actually moving on an exponential, are we ready?

让我们在接下来与美国参议员马克·沃纳讨论一下这个问题。

Let's talk about it with US senator Mark Warner right after this.

财政负责、金融天才、货币魔术师。

Fiscally responsible, financial geniuses, monetary magicians.

这些是人们在把车险转投Progressive并节省数百美元时会说的话。

These are things people say about drivers who switch their car insurance to Progressive and save hundreds.

因为Progressive为一次性付清保费、拥有房产等提供折扣。

Because Progressive offers discounts for paying in full, owning a home, and more.

此外，当您需要帮助时，可以信赖他们出色的客户服务，让您的每一分钱都花得更值。

Plus, you can count on their great customer service to help when you need it, so your dollar goes a long way.

访问progressive.com，查看您是否能节省车险费用。

Visit progressive.com to see if you could save on car insurance.

Progressive意外伤害保险公司及其关联公司，潜在节省金额因情况而异，并非在所有州或情况下都适用。

Progressive Casualty Insurance Company and affiliates, potential savings will vary, not available in all states or situations.

欢迎收听《大科技》播客，本节目致力于对科技世界及其更广泛领域进行冷静而深入的对话。

Welcome to Big Technology Podcast, a show for cool headed and nuanced conversation of the tech world and beyond.

今天我们为您准备了一场精彩的节目。

We have a great show for you today.

美国参议员马克·沃纳今天与我们同在。

US senator Mark Warner is here with us.

我们将讨论政府是否为快速发展的AI做好了准备，政府数据截至目前对AI导致的失业有何反映，以及关于Anthropic公司与五角大楼之间的小型事件的最新进展。

We're gonna talk about whether the government is ready for fast AI progress, what government data says about AI driven job loss so far, and the latest on that small anthropic situation with the Pentagon.

沃纳参议员，很高兴再次见到您。

Senator Warner, great to see you again.

欢迎来到我们的节目。

Welcome to the show.

亚历克斯，非常感谢您邀请我。

Alex, thanks so much for having me.

自从我们上次交谈已经四年了，我联系您是因为我最近感到有些不安。

So it's been four years since we last spoke, and I reached out because I had been getting freaked out.

我实话跟你说吧。

I'll be honest.

我最近和一些在AI实验室内外的人交谈过，他们普遍认为AI技术正在呈指数级发展，并可能带来真正的冲击。

I'd been speaking with some folks in and around these AI labs, and there's a belief among them that AI technology is moving on an exponential and could have real disruptions.

对我和许多一直关注此事的人来说，几个月前这还只是营销话术，但现在至少存在一定的可能性，这种说法是真实的。

And I think for me and many others who've been watching this, that was marketing language a couple months ago, but now there's at least a percentage chance that that's real.

我感到很担忧，因为我不确定政府是否已经为指数级发展做好了准备。

And I'm freaked out because I'm not sure if the government is ready for an exponential.

硅谷或许能应对指数级变化。

Silicon Valley might do exponentials.

华盛顿则常常是线性推进，有时甚至倒退。

Washington does linear or backwards sometimes.

我想听听你的看法，我知道你对这件事非常关注。

I wanted to just get your take on I know you're right into this.

大家都说，去问问沃纳参议员吧。

Everybody says, you know, go to speak to senator Warner.

他是那个知道内情的人。

He's the one that knows what's going on.

但我想要听听你对华盛顿今天整体氛围的看法。

But I wanna get your take on the general vibe in Washington today.

你认为国会议员和参议员中，普通成员是否意识到，可能正在酝酿一场需要采取激烈措施来避免负面后果的事件？

Do you think there's awareness among, you know, rank and file members of Congress, and the Senate that something might be brewing that there will have to be drastic action to head off the negative consequences if it happens?

好吧，亚历克斯，我认为政府还没准备好。

Well, Alex, I don't think government's ready.

我认为社会也没准备好。

I don't think society's ready.

我知道那些同样持AI乐观态度、一直在谈论这件事的人。

And I know the same, you know, AI optimists who are talking about this.

我实际上觉得他们的说法已经改变了，现在变得保守了，因为他们害怕引起恐慌。

I actually think they have changed their pitch and are now holding back because they're freaked out about freaking out people.

我看到过，你知道的，我从价值角度仍然看好AI，但短期来看，未来三到五年，经济动荡将会非常剧烈，我真的觉得我们完全没准备好。

And I've seen, like, you know, the and I am still long AI in terms of value, but, boy, short term, next three to five years, the economic disruption is gonna be I just think we are not ready at all.

我们没有好的数据。

We don't have good data.

我们不知道正在发生什么。

We don't know what's happening.

我常举的一个例子是，看看Anthropic今年的Claude产品，Claude已经某种程度上颠覆了整个软件行业。

An example I like to give is if you just look at Anthropic's Claude products this year, How Claude has already kind of disrupted the whole software business.

现在，市场稍微恢复了一点，但随后又在人力资源领域重演了同样的情况。

Now, the market recovered a little bit, but then it hit the same thing on the HR business.

如果人们没有说行业基础将发生根本性的巨大变化，市场不会这么快做出这样的反应。

The markets don't respond that way that quickly if people aren't saying there's going to be fundamentally dramatic change in kind of industry fundamentals.

而这仅仅是两个领域。

And that's just two areas.

我认为还有更多、更多的变化即将到来。

And I think there's much, much more to come.

所以我听你多次说过，这些首席执行官们可能在淡化其影响。

So I've heard you say this a couple times that these CEOs may be downplaying the impact.

我知道你们私下有交流。

I know they speak with you privately.

他们有没有跟你说过类似这样的话：参议员沃纳，你别跟别人说，但我想告诉你的是，是什么让你得出这个判断的？

Are they telling you things like saying, hey, Senator Warner, don't say this to other people, but here's what we think or what what brings you to that assessment?

让我这么认为的，是那些在人工智能领域发表这些言论的首席执行官们。

Brings me is, you know, the CEOs who are saying this in the AI space.

我私下从一些知名大公司听到的消息是，他们正在削减实习生或应届毕业生的招聘人数，甚至减半。

And what I'm hearing privately from big brand name firms who are saying they're cutting off or cutting in half the number of interns or first year hires.

我甚至听说一家全国知名的律师事务所决定不再招聘任何初级律师。

I even heard from a nationally known law firm that has decided to hire no first year associates.

他们打算先暂停一下，看看情况如何再决定是否招聘。

They're going to take a pause and see how this works out before they even hire.

这些年轻人辛辛苦苦读完法学院，拿到了大公司的工作offer，结果这份工作却没了，这完全不是他们的错。

All these kids, after they've done everything to get through law school and they got a job offer they thought with a big brand firm, and then it's going away, nothing they did.

是因为人工智能。

Because of AI.

是的，因为人工智能。

Yeah, because AI.

我听说很多中型公司都说，前几天有个人告诉我，他以前有23个人做这个后台工作。

And I hear like so many companies that are midsize who say, I had one guy the other day saying, I had 23 people do this back office function.

现在只剩下三个人了。

Now I got three.

这难道不令人惊讶吗？

Isn't that amazing?

关键是，我们目前甚至还没有收集这方面的数据。

And the thing is, we are not even collecting data on this yet.

这就是为什么我和乔什·霍利共同提出了一项两党支持的法案，要求劳工统计局开始测量这一现象——不仅仅像杰克·多西说他因为人工智能裁掉了40%的员工那样，无论这是否属实我们还无法确认，但也要尝试衡量那些原本会创造的传统岗位。

That's why I've got a bill with Josh Hawley, very bipartisan, that says to BLS, Bureau of Labor Statistics, we need to start measuring this, And not just in terms of firms like Jack Dorsey saying he's cutting 40% of his staff on because of AI, whether that's true or not, we won't know for sure, but kind, but also try to measure jobs that would traditionally have been created.

因为我的观点是，这将特别影响刚毕业的大学生和研究生。

Because my view is that this is gonna particularly hit kids coming out of college, coming out of graduate school.

目前，应届大学毕业生的失业率大约为9%。

We're at about 9% recent college graduate unemployment.

我认为这个数字实际上会达到30%。

I think that number will actually go to 30%.

这种经济动荡不仅会影响那些找不到工作的年轻人，还会影响那些资助他们大学教育的父母，以及我认识的目前在读大学的每个人所感受到的恐惧，我认为人们还没有考虑到这一点。

And the economic disruption that will have not only on those young people that don't get jobs, but their parents who help finance their college education and the level of fear that is amongst everybody I know that's in college at this point, I don't think people are factoring that in.

如果说政府还没做好准备，那都算是轻描淡写了。

And and to say government's not ready would be an understatement.

没错。

Right.

我们会谈谈你正在推动的一些立法。

And we're gonna talk about some of the legislation that you have brewing.

你知道的？

And and and you know?

但这需要的不只是一个或两个参议员。

But it takes more than one or two senators here.

而且，你知道，你已经通过了沃纳参议员的克劳德测试——你是一位知道克劳德是什么的参议员。

And, you know, you've already passed the Claude test, senator Warner, which is you're a senator that knows what Claude is.

100名美国参议员。

100 US senators.

你认为其中有多少人知道Claude是什么？

How many of them do you think know what Claude is?

我希望，比我想象的要多一些。

Well, I hope I hope more than you and I think.

但again，我不确定你现在是否想深入讨论Anthropic的Clyde部分，我们是否现在要走这条路。

But again, I don't know if you want to go now into the whole Clyde's part of Anthropic, whether we want to go down that path now.

但我会说，Anthropic，选你的Anthropic、OpenAI。

But I would argue that Anthropic, pick your Anthropic, OpenAI.

显然，谷歌表现不错。

Obviously, Google is doing well.

我们有六七种大语言模型正在取得重大进展。

We've got a half dozen LLMs that are making major advances.

但目前Anthropic的情况是，他们曾与国防部合作并被广泛使用，而Anthropic的领导层与国防部的Hegseff保持协调，显然，任何公司若要与国防部合作，都必须做出一些妥协。

But what's happening to Anthropic at this point, as they were doing business with the Defense Department and being very well used, And, you know, the Anthropic leadership at Crosswise with Hegseff at DOD, and obviously, any company, if they're gonna do business with with DOD, to make some accommodation.

但认为我们将完全交由皮特·赫格塞斯决定这些人工智能工具可以毫无限制地用于监控，这简直荒谬，

But the idea that we're going to turn over to Pete Hegseff the ability to completely decide that these AI tools can be used totally for surveillance without any guardrails or

没错。

Right.

甚至更糟糕的是，制造无需人类干预的人工智能武器，这可是个天大的问题。

Even potentially worse, creating AI weapons without a human in the loop, that's big freaking deal.

如果我们此刻没有陷入与伊朗的战争，我认为这本应成为重点关注的问题。

And we ought to have if we were not in this war with Iran at this point, I think that would have been a major focus.

而眼下究竟发生了什么？我正试图动员科技界人士：无论你对特朗普和赫格塞斯持何种看法，如果这些决策正在发生，而赫格塞斯竟宣称Anthropic是供应链风险，那意味着不仅Anthropic不能与国防部合作，任何公司——事实上，几乎所有美国主要企业都与国防部有某种程度的合作——也都不能与Anthropic合作了。

And what is even happening with this I'm trying to rally the tech community to say, regardless of what you feel about Trump and HEGSF, if you're having these decisions and then HAGCEF is going to declare, is trying to declare Anthropic a supply chain risk, that would mean that not only Anthropic couldn't do business with DOD, but any company, and virtually every major company in America does some level of business with DOD.

他们也不能再与Anthropic合作了。

They couldn't do business with Anthropic as well.

这相当于让一个人就能给美国主要科技公司判死刑，人们必须意识到，这些事情正在实时发生。

This would be the ability for a single individual to write a death sentence to major American tech companies, and people need to realize this stuff is happening real time.

好的。

Okay.

所以，我想问一下，我之所以问这个问题，是因为我理解你的观点。

So I I guess, like, the reason why I'm asking the awareness and I and I take your point.

我们将在后半部分再深入讨论一下Anthropic与国防部之间的纠纷，不管你怎么称呼它——国防部还是战争部。

We're gonna talk a little bit more about this in the second half about the Anthropic d o DOD or the war the, dispute they're having with the Department of Defense Department of War, whatever you wanna call it.

我们会再多谈一点这个事。

We'll talk about that a bit more.

我问这个问题的原因是，也许这场纠纷正在让人们对Anthropic产生更多关注。

The reason why I asked is because and and maybe this dispute is giving more awareness to Anthropic.

我只是想确认一下，或者你是在说其实没什么可放心的，当涉及到你同事们的工作优先级时，这件事至少排在前列吗？

I just wanna see if you could reassure me, or maybe you're saying that there is little reassurance, that when it comes to the list of priorities that your your colleagues have, that this at least ranks.

对吧？

Right?

因为我记得五年前、十年前我报道过社交媒体相关的事情。

Because I I remember reporting on the social media stuff, five, ten years ago.

很明显，当时根本就没有什么真正的重视。

It was clear that there was no there was yeah.

你知道吗？

And you know what?

我想我们总算挺过来了。

It it I guess we sort of came out of it okay.

好吧，亚历克斯。

Well, Alex

我们确实挺过来了。

We did.

但这一点

But this

我担心的是，政府对这种可能更快发生的事情同样缺乏意识。

is my worry is the there's the same lack of awareness in the government for something that could happen faster.

说得对。

Well, Amen.

就像社交媒体曾经是个挑战一样。

Like social media was a challenge.

而且你知道吗？我曾推动过关于数据可移植性、互操作性和可委托性的两党法案，而后者现在基本上被称为代理式AI。

And, you know, I had bipartisan bills on data portability, interoperability, delegability, which is now basically called agentic AI.

我们还提出过关于暗黑模式的议案。

We had things about dark patterns.

当时有很多两党合作的行动。

There was lots of bipartisan action.

所有社交媒体公司都说，是的，我们希望有一些有意义的监管，但直到你把条款写进法律文本里为止。

And all the social media companies, they all said, yeah, we want some meaningful regulation until you put words on the page.

而我们一无所获。

And we batted zero.

我们甚至还没有通过那些关于儿童在线安全的法案。

We still haven't even done the freaking kids online safety bills.

所以社交媒体是个挑战。

So social media is a challenge.

我认为它对年轻人有某种心理影响，抱歉，但与人工智能相比，这简直微不足道。

It has, I think, effects you know, psychic effect psychological effects, I'm sorry, on on young people, but it is tiny compared to AI.

当我们想到已经听到的那些关于人工智能可能导致孩子自杀的故事时，再看看六个月前还只是零星报道的、人们与人工智能代理产生浪漫情感的现象。

When we think about, you know, the stories already we're hearing about AI leading kids potentially to suicide, and we're seeing what was kind of a spot story just six months ago of people becoming romantically involved with AI agents.

现在，这已经变成了一种可以统计的数据了。

Now this is actually a statistical thing you can look at.

而这还只是心理和社会影响方面。

And that's just on the kind of psychological societal effects.

但在就业影响方面，我们缺乏可靠的数据。

But on the job effects, we don't have good data.

我觉得，人们正在互相推卸责任。

We have people, I think, shifting blame.

亚马逊宣布裁撤了11000个岗位。

Saw Amazon, they've announced 11,000 job losses.

但他们声称这些裁员与人工智能毫无关系。

They say none of that's due to AI.

但我必须告诉你，如果没有投资者相信这些企业能带来回报，就不会有数以万亿计的私人资本涌入。

But I got to just tell you, you wouldn't have literally not billions, but trillions of private capital coming in if these enterprises don't, the investors don't think they're gonna return.

现在，这其中一些可能是因为我们将迎来医疗领域的重大突破，或者会出现AI创造的新工作，我相信这种情况会发生。

Now, some of this may be because we're gonna have great healthcare breakthroughs or have AI created jobs, which I believe we will.

但从短期来看，AI导致的岗位流失规模将令人震惊，我认为大多数参议员并未意识到这一点。

But in the short term, the amount of AI job dislocation is gonna be jaw dropping, and I don't think the majority of senators understand it.

我认为他们是可以被说服的，但我担心的是，目前特朗普政府的首要议程就是全力推动AI发展，踩下油门，因为我们必须超越中国。

I think they can be convinced, and I just fear that what we've got now is the overriding agenda coming out of the Trump administration is they are kind of AI accelerate accelerators, you know, pedal to the metal because we gotta beat China.

我们确实必须超越中国。

And we do have to beat China.

但如果我们完全不考虑任何监管措施，也不考虑短期经济后果，这真的令人恐惧。

But the idea that we are not gonna think about any guardrails or about the short term economic consequences, I think, is really frightening.

顺便说一句，我依然相信AI的力量——毕竟，我们根本不可能把潘多拉的魔盒再关回去——它可能带来积极影响。

And and as somebody who still believes the power of AI by by the way, there's no way we're putting the genie back in the bottle anyway, is could have positive effects.

我们可能会看到左右两派的民粹主义联手，试图扼杀创新，而且方式笨拙。

We could actually have populism on the left and the right coming together to try to snuff out the innovation and do it ham handedly.

所以啊，我这最后一份工作，真希望能顺利留下来。

So boy, boy, this is, I'm trying to get hired one last time in this job.

而且主要原因可能是，如果我能帮助引导一些人工智能解决方案，我当然不敢声称自己拥有所有答案。

And probably the major reason is if I can help navigate some of these AI solutions, and I don't pretend to have by any means all the answers.

事实上，我和亚历克斯以前也讨论过这个问题。

Matter of fact, think Alex and I have talked about this at one point.

我们认识很久了，大概三年前吧。

We go way back in time, like three years ago.

我认为，至少在那时，人们的想法是，确保就业的保障至少应该是让每个人都掌握基本的编程技能。

The I think, well, at least at that point, thought through, well, guarantee of a job was let's at least make sure everybody has basic coding skills.

这个想法本意是好的，但显然并不正确，因为这些正是最先被取代的工作。

That was well intentioned but was obviously not the right answer since those are the first jobs being eliminated.

那么，当你思考同事们如何看待这个问题时，对他们来说，这是高优先级、中优先级还是低优先级？

And so when you think about the way that your colleagues view this, is it high priority, medium priority, or low priority for them?

这些事情其实很难。

Wasn't this stuff is hard.

我是在开玩笑，但说的也是实情。

I joke, but it's kind of true.

我花更多时间在AI上，并不意味着我对它的理解就更好，这两者之间没有真正的线性关系。

There is no real linear relationship between me spending more time on AI and actually thinking I have a better understanding.

你知道，它发展得太快了。

You know, it it is evolving so quickly.

我认为大多数人都会这样，这是一种人性的反应。

I think most members you know, and this is a human reaction.

如果你看不懂，觉得太复杂，就会想回避它。

If you don't get it and it seems too complicated, you wanna try to punt on that.

这就导致了简单化的解决方案，比如干脆关闭它，或者让所有数据中心停摆一年。

And that allows for simple minded solutions like let's just shut it down, or let's just have a moratorium all data centers for a year.

这并不能回答问题。

That's not going to answer the question.

所以我们必须学会应对它。

So we do have to navigate it.

我希望我能贡献的微小价值是，不要把这个问题变成党派之争，而是努力寻找两党中都认同的人，说：我们必须正视这个问题。

And what small value I hope I can add is not turning this into a partisan issue and trying to find folks on both sides of the aisle that says, Hey, we've got to grapple with this.

中国和世界其他地区正在向前推进。

China and the rest of the world is moving ahead.

我们无法逆转这一趋势，但我们并非无能为力，既可以设立保障措施，也可以应对经济上的动荡。

There is no way we can we can reverse this, but we are not powerless both to put guardrails in effect and also say in terms of the economic dislocation.

我给人工智能界提出的一个挑战是：你们说得对。

You know, in my challenge to the AI community is you guys are right.

如果政府来定义这一切，我们很可能会搞砸。

If government defines us all, we'll probably screw it up.

所以你们帮我们定义这场转型会是什么样子，无论是培训、再培训，还是我们想称作的任何工具，但你们也得帮忙承担费用，因为这背后的成本将是惊人的。

So you guys help us define what this transition looks like, what the training or reskilling, whatever tool we want to call it, but you also got to help pay for it because the cost of this are going to be amazing.

是的。

Yeah.

我会稍微解释一下我提问的思路。

And and I'll I'll explain a little bit about the question my line of questioning here.

我只是想看看，如果我们真的看到这种指数级发展，政府是否能迅速行动。

I just wanted to see if you think the government would be able to move fast if we end up seeing this exponential.

我甚至会采纳你的观点。

I'll even take your words.

你最近在一段精彩的YouTube视频中谈到了人工智能的挑战。

You said recently in a great YouTube video about the AI challenge.

这种变革的剧烈程度，是我一生中见过的最显著的。

This is as dramatic as a change as anything I've seen in my lifetime.

你提到，想想互联网带来的变革，而我们现在看到的这种人工智能变革，可能在接下来两到三年内就结束了。

You said think about the transformation brought by the Internet, this AI AI transformation at the rate we're seeing it could be over in the next two to three years.

而且，你知道，我再次强调，我读过，我知道你有相关立法。

And and, you know, I again, like, I've read I know you have legislation.

目前至少有三项关于人工智能的法案正在推进中，旨在收集数据、理解其影响，并提前应对这些问题。

You have three bills, at least three bills in action right now on AI gathering data, trying to understand, the implications here, trying to head off the issue.

这不一样，你不能以那种方式去阻止它。

And it's different, like, it's not you can attack it in a way that it's not like stop it.

对吧？

Right?

这就像帮助那些可能面临失业风险的人，但听到你描述参议院处理这个问题的方式，我并没有感到多少安心，我觉得这种速度是无法跟上的。

It's like maybe help people who are being who are who are in at risk of job dislocation, but I'm not very reassured hearing the way that you described the way that this issue is being handled in the Senate that that that speed is going to be met.

说实话，我无法指出世界上任何一位政策制定者已经解决了这个问题。

Well, I'm not sure I can point to a policymakers anywhere in the world that's figured this out.

我的意思是，我已经看到了一些良好的两党合作立法。

I mean, I got good bipartisan legislation.

让我们成立一个类似网络太阳圈委员会的机构，真正地取得一些实质进展。

Let's put a commission together similar to the cyber solarium that actually put some points on the board.

比如，成立一个关于未来经济的委员会。

You know, commission of the economy of the future.

我们已经有法案要求劳工统计局开始报告人工智能对就业的冲击。

We've got, you know, bills to get BLS to start reporting on AI job disruption.

我还有一项两党支持的法案，探讨人工智能将如何影响金融市场，以及我们该如何思考这个问题。

I've got a bipartisan bill about how AI is going to affect the financial markets and how we ought to think through this.

我认为这些措施是深思熟虑的，但它们也坦承自己只是些微小的渐进步骤，而实际情况很可能是‘天啊，出大事了’的时刻。

They are, I think, thoughtful, but they are self acknowledging here small incremental steps when it very much could be the holy shit moment.

我们能有更大的格局吗？

And can we think big?

如果唐纳德·特朗普这个颠覆者，哪怕有一点点同理心或合作精神，一个颠覆者其实可以帮助我们度过这一关。

And, you know, if if Donald Trump the disruptor, if he had an ounce of either empathy or collaborative spirit, somebody that is a disruptor could actually help us through this.

但我真的想更乐观一些，可我却感到恐惧。

But I I I wanna be more optimistic, but I am terrified.

我的意思是，前几天有个人来找我，他说了一件很有意思的事：你把三对父母聚在一起，聊聊他们的孩子。

I mean, I had somebody come in the other day, and it was like, I thought it was a very interesting thing saying you get three couples of parents together who are talking about their kids.

十年前，人们会说：全球化，我不知道我的孩子能不能找到工作。

Ten years ago would have been, this globalization, I don't know if my kid's gonna get a job.

五年前，人们会说：天啊，我真的很担心我的孩子会上瘾于社交媒体。

Five years ago, it'd Oh my gosh, I'm really concerned about whether my kid is getting addicted to social media.

现在，对话已经变成了这样——而且这种讨论已经普遍到我们的政策制定者都意识到了：他们害怕自己的孩子已经做了所有正确的事。

Now the conversation, and this is happening at such a level that our policymakers get it, they are terrified that their kids have done everything right.

他们上着大学，但可能依然没有工作。

They're going through college, and there may not be a job there.

对。

Right.

而且，我能说一下吗？我提到社交媒体作为例子，说明美国立法机构在有效应对技术方面的局限性，但这和社交媒体的情况不一样吗？

And and can I just say so I brought up social media as an example of, our US Legislative body's ability to deal with technology effectively, but it's different than than social media?

我认为我们都同意，在社交媒体问题上，主要分歧在于：你是否要告诉Facebook该如何处理它的新闻推送？

I think we both agree here that with social media, the big disagreement was, are you gonna tell Facebook, like, how to handle its news feed?

该怎么做？

What to do?

这并不一定意味着需要立法或政策来要求AI公司停止开发他们的模型。

This isn't necessarily, legislation that needs to be or policy that needs to be, can we tell the AI companies to stop making their models?

要应对这些负面影响，更关键的是如何促进就业增长或再培训？

To be able to handle the negative effects here, it's more like, how do you stimulate job growth or retraining?

即使这一点也尚未被证实，但让我抱有希望的是，这种情况或许真的可以实现，而且你还能看到这些两党共同提出的法案。

And even that's probably not proven, but, that's what gives me hope is that there's a chance that that that can be and the fact that you have these bipartisan bills

好吧，但让我抱有希望

Well, let's but let's me hope

就是就是

that that

可以成为解决方案。

can be solution.

你知道吗，我跟一些行业里的朋友聊过，他们说，至少我们应该先处理像未经同意传播私密照这样的问题。

You know, I've talked to some of the my friends in the industry who say, like, let's at least deal with things like nonconsensual nudes.

你希望你的女儿或儿子被深度伪造技术滥用吗？

Do you want your young daughter or son to be portrayed with a deep fake out there?

每个人都说好，但接着你就会听到埃隆在Grok上说，不。

And everybody says yes, but then you get, you know, Elon at Grock saying, no.

你知道，我们会成为一个例外。

We you know, we're gonna be an outlier.

在这些问题上，我们总是倾向于采取最低的标准。

And, we default to the lowest common denominator on some of this.

那些关于人们被引导走向自杀的可怕故事。

The idea of these horrific stories of people being guided to suicide.

我们可以说，好吧，我们会试着稍微调整一下模型，但总是落后一步。

We can say, well, we're going to try to correct the model a little bit, but we're always lagging.

我的意思是，我真的对这件事感到有点不安，比如你把浪漫情感寄托在谁身上。

I mean, I do think I'm I'm kinda freaked out about this, you know, the idea of who you turn your romantic interest to.

我想我们都记得几年前那部电影，叫《她》，主角爱上了一个聊天机器人。

I think we all remember that movie a few years back called I think it was called Her, where the the main character fell in love with a a kind of a a chatbot.

这种事情现在正在发生，不是零星出现，而是已经开始在统计上显现出来。

That stuff is happening now, not in tiny numbers, but it's actually starting to appear statistically.

然后我们谈到工作流失的问题。

And then we come to the to the, the job dislocation.

在大多数中等规模的公立大学里，年轻人最热门的专业是商业或工商管理。

In most mid tier public universities, the number one major for most young people is business or business administration.

这些工作就是你毕业后去公司当几年初级分析师之类的职位。

Those are the jobs that you come out, you go work for a firm for a couple years as a young analyst or whatever.

这些工作已经消失了。

Those jobs are gone.

我的意思是，前几天有人提到，杰特，我不确定这是否正确，也许这些公司应该提供激励措施，鼓励更多人选择护理专业，而不是工商管理。

I mean, somebody said, Jet, the other day, I'm not sure this is right, that maybe, you know, some of these companies ought to pay an incentive to get more people into nursing as opposed to business administration.

我们至少应该向人们说明，这些领域中的就业前景将发生巨大变化，而我不确定我们是否已经做好准备。

We ought to at least disclose to people that the job prospects in some of these fields are gonna dramatically change, and I'm just not sure whether we're ready.

我发现的一个令人担忧的事情是，尽管我想更乐观一些，但当你与领先的AI公司和AI思想家交谈时，他们只会给出部分答案，比如：‘哦，我们会建造大量数据中心，这样传统行业就能迎来增长。’

And one of the scary things that I've found, and, again, I wanna be more optimistic, is, like, you talk to the leading AI companies, the leading AI thinkers, and they'll give you a partial answer, well, gosh, we're going to build a lot of data centers so that the traditional trades will have an increase.

而这种增长仅限于建造这些设施的短期需求。

And that will be a short term increase in terms of building those facilities.

显然，我们对电力的需求将大幅增加。

And there's going to be obviously huge needs for more electrons.

因此，我大力主张，如果没有小型模块化核能或其他类型的分布式发电，我们根本无法为这一切提供足够的电力，但即便如此，这类设施的数量仍然相对有限。

So I'm a big advocate that we'll never be able to power this without a small modular nuclear or other kind of decentralized power generation, but that's still gonna be a relatively small number.

然后你会问：我们如何确保无论你从事什么工作，都能利用AI来提升自己的技能？

And then you say, Well, how do we make sure that whatever you're gonna do, you can use AI to become better skilled at it?

而每个人都在使用模糊的术语，因为这些词现在都是他们临时编造的。

And everybody's got soft terms because they're making up right now.

但天哪，我们得马上准备好这些东西。

But gosh, we gotta have that stuff ready yesterday.

等到五月大学临近毕业时，看看这次招聘周期会非常有趣。

And it'll be very interesting to see even this hiring cycle as we get, you know, close to graduation in May in colleges.

这将会

One It's gonna

非常能说明问题。

be very telling.

是的。

Yeah.

这将会非常能说明问题。

It's gonna be very telling.

新毕业生找到工作后，我们会很快了解到情况。

People new grads get jobs, we're gonna learn very quickly.

是的。

Yeah.

我来说几点。

I'll I'll say a couple of things.

首先，作为一个护士的配偶，我同意你的看法。

First of all, as someone who's married to a nurse, I agree with you.

这是一条不错的职业道路，我总是告诉她，至少我们中会有一人长期有工作。

It's a good career path, and I always tell her at least one of us will be employed in the long term.

关于人工智能与浪漫关系，我的意思是，我无法相信成年人不应该能够与AI聊天机器人建立这种关系。

On a on AI romantic relationships, I mean, can't possibly, believe that adults should not be able to enter into these relationships with AI chatbots.

这难道是你认为的更小的问题吗？不。

Is that more of a minor thing that you would No.

就立法而言。

In terms of legislation.

对吧？

Right?

我不知道。

I I don't know.

我的意思是，真的吗？

I mean Really?

关于未成年人，你和我都知道，这事儿很关键。

Obviously on the minors and, you know, you you and and trust me.

我不是那种要当老大哥、禁止成年人行为的人。

I'm not, you know, big brother here in saying we can prohibit behavior of adults.

但作为一个需要繁衍、需要维系人际关系的社会，我认为我们至少应该就此展开讨论。

But, you know, at at at some point, you know, as a functioning society that needs to procreate, that needs to have human relations, I just think we ought to have at least a discussion about this.

我觉得我身边的朋友们要么别轻率地 dismiss 这个问题，说‘我们无能为力’，要么至少该挂个更醒目的警告牌——成年人当然会做他们想做的事，但至少要让他们充分知情，再踏入这些陷阱；同时，我们也要努力维持一种共同的、可共享的真相，以便思考AI如何影响政治辩论。

And I I just think that my friends in the community need to either not blow it off and say, oh, we can't, there's nothing we can do, or, you know, least we put a bigger warning sign, you know, adults are gonna do what they're gonna do, but the full fully informed before you go down some of these rabbit holes, you know, and and the ability to have kind of shared common truth as we think about how AI AI could affect, you know, political debate.

我现在非常担心，我们的第26届选举会受到外国势力的干扰，甚至担心总统本人会试图让联邦机构接管我们的选举。

I am terrified right now of of, you know, disruptions in our twenty sixth election from foreign sources or, frankly, even, you know, the president's willingness to try to say he wants to have the feds take over our elections.

到目前为止，我们还没见过深度伪造技术被大规模使用。

And we have not seen deepfakes used in a massive way so far.

但正如我们所知，这项技术每月都在快速演进。

But as we know, that technology is evolving on a monthly basis.

在选举周期中，只要出现一次重大失误，就足以让那些已经对民主制度失去信心的人彻底丧失信任。

And it only takes one major screw up in a election cycle, for example, for people who already are losing faith to lose faith in our basic democratic processes.

是的。

Right.

你现在正竞选第四个任期，已经是三届参议员了。

So you're running for a fourth term, three term senator at this point.

我跟政治家交谈时最喜欢的一点是，我们可以讨论民调，而没有人比像你这样的人更擅长解读民调了。

And one of the things that I love when I speak with politicians is we can talk about polling, and no one reads polls better than, people like yourself.

我想读给你看几项关于人工智能受欢迎程度或不受欢迎程度的民调，听听你对这些数据可能带来的政治影响有何看法。

So I wanna read to you a couple polls about AI's popularity or lack thereof and sort of get your read on what it could mean politically.

这是来自NBC新闻的民调。

This from NBC new news poll.

你可能已经看过这个了。

You might have seen it.

大多数注册选民，57%的人认为人工智能的风险大于其好处，而且相当一部分选民对人工智能持负面看法，认为无论是民主党还是共和党，都没能妥善应对这一快速发展的技术所带来的政策挑战。

A majority of registered voters, 57% said they believe that the risks of AI outweigh the the benefits, and a plurality of voters view AI negatively and don't believe either Democrats or Republicans are doing a good job handling policy related to the rapidly advancing technology.

我想我们先暂时搁置对民主党与共和党的反应不谈。

I I guess let's leave the the the reaction to, Democrats and Republicans aside for a moment.

那会有什么后果呢？

What are the consequences?

我们在节目中一直试图弄清楚这一点，但没有人比你更合适来谈论这个话题了。

And we've tried to figure this out on the show, but no one better, to speak about it than with you.

如果人工智能行业持续如此不受欢迎，会带来什么后果？

What are the consequences for this AI industry if it continues to pull so low?

他们是否在给自己制造一个需要回应的问题？

Is there are they opening themselves up to put a question?

他们确实是在给自己制造问题。

They're opening themselves up.

我认为第一波冲击会来自俄勒冈州的数据中心。

I think the first line will be Oregon's data centers.

你知道的？

You know?

而且它们规模很大。

And they are big.

它们消耗大量电力，这几乎成了人们对人工智能整体担忧的代名词。

They have a lot of power, And that becomes almost a proxy for the overall concerns about AI at large.

它们必须确保人们的电费不会上涨，供水不会紧张，并且要更好地进行监管。

And they're going to have to go ahead and make sure that people's electric bills don't go up, that the water supplies don't go up, that they are better that they are better screened.

我了解到弗吉尼亚州的一个县，把人工智能带来的收入全部投入到了经济适用房建设中，让民众能切实受益。

I got a county in Virginia that took their AI revenues and put it all into affordable housing so people see a tangible benefit.

在弗吉尼亚，我们正处于前沿。

And in Virginia, we're on the front line.

我们在数据中心方面堪称天堂，

We're data center heaven in terms of

最大的州

biggest state

我们目前在州一级正就从该行业每年抽取5亿到10亿美元的问题展开激烈辩论。

by We're data centers in the having a major debate right now at the state level about trying to extract somewhere between 500,000,000 and $1,000,000,000 a year from the industry.

我希望这个行业能积极应对这些问题，说：是的，我们会自愿提供帮助，并将这些资源投入到这场经济转型中。

I would hope the industry would lean into some of these things and say, we're gonna yes, we will voluntarily help and we will dedicate that to this economic transition.

事情发生得太快了。

It's happening so quickly.

我不确定我们能否就此达成一致，而且我也不再是州长或政府官员了。

I'm not sure we're gonna get that together, and I'm no longer the governor or the state official.

但整个科技行业基本上已经表明了态度，而我作为一个支持科技的人这么说。

But the tech industry writ large has basically said, and I say this as a pro tech guy.

我的商业背景是科技，我非常相信科技，但到目前为止，科技行业普遍认为：政策制定者根本不理解我们，我们可以无视他们。

My business background was tech, I'm a big believer, but the tech industry so far has generally said, rightly so, policymakers don't get us, we can blow them off.

社交媒体平台之所以能成功，正是因为它们完全没有任何监管基础。

That was clearly the success of the social media platforms to never have any regulatory basis at all.

而当你面临过度监管时，比如来自欧盟的监管，它们就会拿欧盟来举例，说：听着，我们不想变得像欧洲人那样。

And then when you do have overregulation, say from the EU, and they'll point to the EU and say, Listen, we don't wanna be like Europeans.

他们根本没有任何创新。

They have no innovation at all.

所以把这件事做对很难。

So getting it right is hard.

但在这方面，如果他们忽视并认为我们可以无视任何监管或框架，我们没有任何义务，我认为这可能会反噬他们。

But on this one, if they kind of ignore and say, We can blow off any regulatory or any framework, we have no obligation, I think it could bite them.

现在，人工智能不会消失。

Now, it's not gonna disappear, AI.

这些模型已经存在了。

These models are out there.

而且你知道，中国显然正在以惊人的速度投入其中。

And the fact that, you know, you know, in China clearly is investing at an amazing rate.

但即使美国停止发展，这些模型也可以转移到其他拥有计算能力的实体手中。

But even if America closed down, the models can transfer to another entity that has a compute power.

所以从某种意义上说，这不会消失。

So this is not going away in a certain way.

别让我听起来像个怀旧的老派人物，但如果说曾经有一个时刻，全世界本应共同思考这个问题，而不是陷入国家间的竞争，那现在就是这个时候。我坚信，无论我们是否正迈向通用人工智能或完整的过程，我们都已越来越接近这些模型内部所发生的奇迹。

Not to sound old school wistful, but if there was ever a time where the world as a whole ought to be thinking through this rather than nation state competition, It is on this issue, and I absolutely do believe that we are now, whether it's full AGI or full process, we're getting close to where the magic that happens inside these models.

我认为，至少我听很多人——主要是男性——说过，我们其实并不理解这一切究竟是怎么发生的。

I don't think, at least I've heard from many of, and they're mostly guys have said, we don't really understand how this is what all is happening.

这远远超出了预测下一个单词的层面，那种AI入门模型早在两年前就广为人知了。

This is way beyond just predictive of the next word, which was kind of the AI one zero one model that people got educated on long, long time ago, like two years ago.

好的。

Okay.

在结束这段之前，我想跟你确认一下，因为你提到了数据中心。

Let me run this by you before we end this segment because you mentioned the data centers.

我感到非常震惊。

I was stunned.

最近有一系列关于人工智能的负面民调，而人们对数据中心的看法让我感到震惊。

So there was a series of negative polls about AI that came out recently, and I was stunned at the way people feel about data centers.

这是来自皮尤研究中心的数据。

So, this is from Pew.

我相信你一定看过这项民调。

I'm sure you've seen this poll.

远比美国人为数据对环境的影响持负面看法，比例为39%对4%；对家庭能源成本的影响为38%对6%；对附近居民生活质量的影响为30%对6%。

Far more Americans say data centers are mostly bad than good for the environment at 39% to 4%, for home energy costs at 38% to 6%, and the quality of life for those that live nearby, 30% to 6%.

我的天啊。

I mean, goodness.

是的。

Yeah.

这些数据中心的民意调查结果真是糟糕透顶。

That is a terrible, terrible polling numbers for these data centers.

这意味着它们将因为反对声音太大而根本无法建成吗？

Does that mean that they're just they're gonna be places where they are just not gonna be built because the opposition is so high?

我看到Axios有一篇报道说，今年计划建设的数据中心有一半被推迟了。

Saw there was an Axios report that said something like half of the data centers that are expected to be built this year are delayed.

其中一部分是由于零部件短缺，但我认为社区的反对会是主要因素。

Now some of that is part shortages, but I think community opposition is gonna be Yeah.

非常重要的一个部分

Big big part of

是的。

it.

而且，有趣的是，一些科技公司会正确地指出，看看那些已经大量建设数据中心的州的电价。

Well and and with the interesting thing too, and the rightfully, some of the tech companies will say, well, you actually look at the electric rates for states that have done a lot of this.

它们并没有出现显著上涨。

They've not seen a dramatic rise.

但我认为，它们不能仅仅说：我们会确保你的公用事业和电费不会上涨。

But I think they have to do more than just say, hey, we're going to cover your no increase in utility and electric rates.

我认为它们必须将这一点写入法律。

I think they've to put that in statute.

我认为它们应该更多地转向在人工智能设施附近自发电，这样电力就不需要接入整个电网。

I think they've got to move more towards self generation that is adjacent to the AI facility so it doesn't go into the full grid.

我认为我们必须对此进行记录。

And I think we have to document that.

我认为它们在用水方面还需要做更多工作。

I think they have to do more on the water usage.

我认为他们在视觉筛查方面需要做得好得多。

I think they need to do a much better job on just visual screening.

这些是又大又丑的建筑。

These are big, ugly buildings.

是的，确实如此。

Yes, they are.

但有一点是，它们确实在进步，但当你进入一个社区去推广时，如果人们脑海中唯一联想到的是像弗吉尼亚州北部那种老旧的上一代数据中心，那就很难说服他们。

But the thing a little bit is they are making progress, but to go into a community and sell that when the only image you have is of, say, data centers in Northern Virginia that are still old school last generation, that's a hard sell.

当然，总会有一些地区需要这笔额外收入来维持运转。

Now, there always will be a jurisdiction that needs that additional revenue to get by.

它们确实能产生收入，而且不会带来很多孩子，因为周边就业岗位不多，但我们需要重新思考这个问题。

They do generate revenue and they don't bring a lot of kids because they don't have a lot of jobs adjacent, but there needs to be a rethinking on this.

我认为弗吉尼亚州目前正在发生的这场州级博弈，我已经告诉过这个行业：你们得密切关注，因为我们是数据中心的重镇。

And I do think that the state battle that's going on in Virginia right now, I've said to the industry, you guys got to watch this because we are the mother load of data centers.

如果弗吉尼亚州在经济协议上做出任何调整，全美其他各州都会效仿。

If there is some adjustment of the kind of economic deal that's going to happen in Virginia, that is going to be copied by every other state around the country.

我对人工智能行业的建议是，别只是拼命地对抗。

And my pitch to the AI industry is, you know, don't just fight it like mad.

要主动一点，说好的。

Be proactive and say, yes.

我们会投入更多资金，不仅确保你们的电价不会上涨并得到适当保护，还会实际拿出资金来帮助度过这场经济转型。

We're gonna chip in more, and we're gonna chip in more, not only make sure your electric rates don't go up and they are appropriately shielded, but we're gonna actually put money on the table to help through this economic transition.

我经常看到人们点头赞同，但缺乏具体政策建议。

I get a lot of head nods, but the lack of specific policy ideas.

亚历克斯，我已经和我能接触到的每个人谈过了，大多数政策专家和其他人要么在观察这个问题，要么想做一些事情，比如收集数据，但关于实际的再培训计划是什么样子，我们目前还没有太多好的例子。

Alex, I've talked to everybody I can, and most policy experts and others are observing the problem or want to do things like, I'm trying to collect data, but what the actual reskilling retraining program looks like, we don't have a lot of good examples so far.

是的，我感受到科技公司的一些挫败感。

Yeah, I'm sensing some frustration with tech companies.

对。

Yeah.

但我能理解。

But I I kinda get it.

你知道，如果你想想那些大公司，它们多年来一直受到抨击，因为大多数大公司都是超大规模云服务商。

You know, if you think about the big the big guys, you know, they've been pounded on for years since since most of the big guys act the hyperscalers.

它们中的大多数最初都是从社交媒体起步的，比如亚马逊、微软这些公司。

Most of them actually started, as either social media or you got Amazon and Microsoft and some of those.

但它们一直靠着空谈应付过去，却没有制定任何实际的规则或监管措施。

But they've kind of gotten through with kind of good lip service, but no rules or regulations in place.

而这次，我觉得情况不同了，回到你提到的数据中心数据上。

And it's kind of like this time, I think the seriousness and back to your numbers on data centers.

这种担忧是真实且明显的，我不希望这种创新就此停滞，但我确实认为，以更积极主动的方式坐下来解决这个问题至关重要。

The fear is real and palatable, and I don't want this innovation to stop, but I do think sitting down and figuring this out in a more forward leaning way is really essential.

这正是我在这里拼命想要推动的事情。

That's what I'm desperately trying to do here.

至少这份工作不允许它变成民主党和共和党之间的对立。

And this job at least is not allow this to be kind of divided d's versus r's.

好的。

Alright.

我们先休息一下，然后回来再聊聊Anthropic、五角大楼以及人工智能在战争中的现状。

Let's take a quick break and then come back and talk a little bit more about Anthropic and the Pentagon and the state of AI in warfare.

广告后马上回来。

Back right after this.

在这档节目中，我采访过很多优秀的科技创始人，其中一个出人意料的普遍挑战反复出现：找到合适的域名。

I've interviewed a lot of great tech founders on this show, and one surprisingly universal challenge comes up again and again, finding the right domain name.

当我推出《大科技》时，自己也遇到过这个问题。

It's something I ran into myself when launching Big Technology.

你想要的名字往往已经被注册了，很容易就想随便选一个凑合了事。

The names you want are often taken, and it's tempting just to settle and move on.

但我最尊重的那些创始人，从不在基础问题上妥协，而你的品牌名称正是其中之一。

But the founders I respect most don't settle on fundamentals, and your name is one of them.

它应该能立刻传达出你真正打造的是什么。

It should immediately signal what you actually built.

这正是我欣赏.dot.tech域名的原因。

That's what I appreciate about dot tech domains.