为何派拉蒙敌意收购华纳兄弟将成功，ChatGPT负责人回应OpenAI“红色警报” | 2025年12月8日

欢迎各位收看信息台TITV。

Welcome everyone to the information's TITV.

我叫阿卡什·佩斯里特沙。

My name is Akash Pesritsha.

今天是12月8日，星期一。

It is Monday, December 8.

今天我们为大家准备了一场精彩的节目。

We have got a great show planned for you today.

首先，是今早的重大新闻。

First up, the big news of the morning.

派拉蒙针对奈飞的交易，发起了对华纳兄弟探索公司的要约收购。

Paramount has launched a tender offer for Warner Brothers Discovery shares in response to Netflix's deal.

此外，IBM正在收购Confluent。

And separately, IBM is buying Confluent.

我们将为您深入解析这两则新闻，同时也会提前预告本周即将公布的甲骨文和博通的财报。

We're gonna break down both of those stories for you, and we are also going to preview Oracle and Broadcom earnings, are coming up this week.

我们还将邀请我们的OpenAI记者，带来关于该公司在应对红色警报期间实际情况的最新更新。

We're also bringing on our OpenAI reporter for an update on how things are looking on the ground as the company works through its code red.

她与负责带领ChatGPT度过这一阶段的人进行了对话，我们将与大家分享他所说的内容。

She had a conversation with the man leading ChatGPT through it all, and we are going to share with you what he had to say.

接下来，我们将回顾上周举行的一场重要人工智能研究会议。

We're then bringing you a recap of a major AI research conference that happened last week.

当时围绕人工智能在生物学领域的突破，出现了一些重大疑问。

There was some big questions that swirled around AI breakthroughs in biology.

最后，我们将推出我们的每周加密货币板块。

And finally, we are debuting our weekly crypto segment.

本周，我们将深入分析预测市场的动向。

This week, we're gonna be unpacking the momentum in predictions markets.

这将是一期内容丰富的节目，所以我们马上开始。

It's gonna be a busy show, so let's get right on into things.

派拉蒙已发起对华纳兄弟探索公司股份的要约收购，试图进行敌意收购。

Paramount has launched a tender offer for shares of Warner Brothers Discovery in a hostile takeover effort.

该公司在Netflix同意以一笔许多人认为难以通过反垄断监管机构审批的交易收购华纳兄弟探索公司几天后，提出每股30美元的收购要约。

The company is offering $30 a share just days after Netflix agreed to buy WBD in a deal that many have thought would have a lot of trouble getting through antitrust regulators.

为我们详细分析此事的是《信息》杂志的联合执行主编马丁·皮尔斯。

Here to break it all down is Martin Pierce, co executive editor at The Information.

马丁，你周末过得怎么样？

Martin, how was your weekend?

周末挺忙的，一直在关注这件事。

It was a busy one watching this No.

我的周末很愉快，因为Netflix的交易是在周五发生的，所以一切顺利。

My weekend was very pleasant because the Netflix deal was on Friday, so it was fine.

好吧。

Okay.

不过我得告诉你，我们原本以为这会是拉里·埃里森的一周。

I have to tell you though, we thought this was going to be Larry Ellison's week.

我的意思是，甲骨文本该成为焦点，但现在大卫·埃里森却出来表示：不，我要在这里掌控局面，大卫这位

I mean, is Oracle's week in the spotlight, and now you have David Ellison coming in and saying, No, I'm taking control here of David the

这得到了他父亲的支持。

is being backed by his father.

这一点在要约文件中非常明确，文件指出他的父亲在甲骨文公司拥有控股权，价值2500亿美元，因此派拉蒙为收购WBD提供的部分资金实际上就来源于此。

That's very clear from the tender documents because they actually lay out, they point out that his father has a controlling stake in Oracle, which is worth $250,000,000,000 So that's where a portion of the money that Paramount is offering for WBD is actually coming from.

好的。

Okay.

我没想到他们竟然直接说出来了，这真是一个有趣的关联。

So I didn't realize they actually they said that outright, which is an interesting link.

能跟我谈谈你对这次收购行动的反应吗？

Tell me a little bit about what your reactions were to the takeover efforts.

这在预料之中吗？

Was this expected?

是的，周五就有一些传言说这会发生，而且这确实合情合理。

Yeah, there was some speculation on Friday that this would happen, and it honestly makes sense.

派拉蒙表示，他们上周就提出了这个报价，但WBD的董事会几乎没有回应，显然他们根本没有认真考虑过，因为他们选择了奈飞的报价。

Paramount says that they made this offer last week that the board of WBD didn't really respond and they certainly didn't I mean, they don't seem to have really thought it through because they went with the Netflix offer.

这样，股东就有了一个明确的替代选择。

This way shareholders have a clear alternative.

股东现在可以接受这项要约。

The shareholders can now tender into the offer.

这笔交易不会立即完成，因为派拉蒙需要获得政府的批准，但派拉蒙表示他们认为一年内可以完成，而Netflix则认为可能需要长达十八个月。

It won't close immediately as Paramount has to get approval from the government, but Paramount says they think they can close in a year and whereas Netflix thinks they would take up to eighteen months.

当然，Netflix能否获得批准存在巨大的不确定性。

And of course, there's an enormous uncertainty about whether Netflix would get approval.

特朗普已经出来表示，他实际上暗示不认为Netflix应该被允许收购它。

Trump has come out and sort of said that I mean, he's sort of indicated he doesn't think that Netflix should be allowed to buy it.

如果从常规的反垄断分析角度来看，这两家公司都会遇到问题。

If you take this from a regular antitrust analysis, both companies would have issues.

派拉蒙已经拥有一家大型电影制片厂，这将使两家大型制片厂合并。

Paramount already owns a major studio, this would be combining two major studios.

但我们所处的环境并不正常。

But we're not in a normal environment.

特朗普政府显然会代表特朗普的盟友行事，而Ellicence显然是特朗普的盟友。

The Trump administration obviously really will act on behalf of Trump's allies and the Ellicence are clearly allies of Trump.

所以我认为我们可以假设特朗普会批准这笔交易，这给了派拉蒙相对于奈飞的巨大优势。

So I think we can assume that Trump will approve this deal, and that gives Paramount a huge advantage over Netflix.

现在，我只想确认一件事：这个要约收购是针对整个公司，还是仅针对流媒体业务？

Now, one thing I just wanted to make sure I was clear on is that this tender offer, is this for the whole company or is this just the streaming?

不，这是针对整个公司的，这也是它的另一个特点。

No, this is for the entire company, which is the other thing about it.

这对华纳兄弟探索公司的投资者来说要容易得多，因为奈飞的报价仅针对影视工作室和流媒体服务。

It makes it much easier for investors in WBD because the Netflix offer is only for the studio and the streaming services.

因此，该报价假设华纳在交易完成前会进行分拆。

So that offer assumed that Warner would split up before the offer went through.

华纳兄弟探索公司的投资者必须考虑：我将获得的网络资产残值是多少？

Investors in WBD would have to be thinking, well, what is the value of the stub that I'll get in the networks?

而现在，他们可以非常容易地比较这两项报价。

Whereas now they'll know they can actually compare these two offers very easily.

我的意思是，我认为目前派拉蒙是最有可能的赢家。

I mean, I think Paramount has got to be seen as the most likely winner in this right now.

奈飞可能会回来提高报价，但我不清楚他们是否负担得起，而且我认为这会给他们的股价带来巨大压力。

Netflix could come back and raise their offer, but really it's not clear to me that they can afford it and I think it would put a lot of pressure on their own stock price.

所以我怀疑他们不会这么做。

So I doubt they will do that.

他们可能会，但我总觉得他们大概不会。

They might, but I kind of think that they probably won't.

你为什么认为派拉蒙愿意接手整个业务？

And why do you think Paramount is willing to take on the entire business?

我的意思是，你提到过电视有线业务这一块，其实没人真正想要这部分。

I mean, you talked about how the cable side of the business, I mean, it's really the side that nobody really wants.

我的意思是，为什么派拉蒙愿意接手全部业务？为什么这对派拉蒙来说具有战略意义呢？

I mean, why does Paramount why are they willing to take on the whole thing, and why does it make Paramount strategic sense for

他们已经拥有相当规模的有线电视频道业务。

already has a significant cable channel business.

他们拥有CBS，这是广播频道，比有线电视更广泛，但本质上还是类似的东西。

They own CBS, which is broadcast more than cable, but it's still the same kind of thing.

他们拥有CBS、MTV和尼克罗迪恩频道。

They've got CBS, they've got MTV, they've got Nickelodeon.

如果把这些与派拉蒙的有线业务结合起来，我认为可以实现一些规模经济。

If you combine those with Borners, I think there are some economies of scale that you could get out of it.

而且，大多数人认为，尽管这个业务在衰退，但它仍会持续存在，并且确实能赚不少钱。

And I mean, most people think that even though that business is declining, it will continue to exist and it certainly makes a fair amount of money.

这种业务在未来几年内仍会存在。

That could exist for years to come.

所以我认为他们的想法是，不如把这两者合并在一起。

So I think their view is we might as well put these two together.

合并后会更赚钱，而且这

They'll be more profitable combined, and that's

很好。

fine.

最后一个问题是给你提的。

And last question for you.

如果你看看WBD股票最近的交易价格，无论是奈飞的报价还是派拉蒙的报价，WBD股东似乎都是大赢家，因为他们都能拿到好处。

If you look at what WBD shares have been trading at lately, both of these prices, Netflix's offer and the Paramount offer, WBD shareholders seem to be big winners in here regardless because they're getting it.

我的问题是，从估值角度来看，你对奈飞提出的估值怎么看？

My question for you is, I mean, in terms of the valuation, what did you make of the Netflix valuation that they were offering?

你对派拉蒙提出的估值又怎么看？

What do you make of the Paramount valuation that they're offering?

我认为奈飞的报价太高了，因为他们只收购了部分股权。

I mean, I think this value the Netflix offer was I think too high because they're only buying a portion of it.

他们收购的这部分并不是主要的利润来源，而派拉蒙收购的是整个公司，因此能获得更多公司利润的收益。

The portion that they were buying is not the main profit generator, whereas Paramount is buying the entire thing and they will therefore get more access to more of the profits of the company.

所以我认为，从估值角度来看，派拉蒙的报价更合理。

So I think the Paramount offer is a more logical offer valuation wise.

很好。

Great.

马丁，感谢你前来做客。

Well, Martin, I want to thank you for coming on.

我相信这个故事还会继续发展，我们会随着事态进展不断与你交流。

I'm sure that this story will continue to unfold, and so we will talk to you more and more as it happens.

这位是马丁·皮尔斯，我们《信息》杂志的联合执行主编。

That is Martin Pierce, our co executive editor here at The Information.

好的。

Okay.

IBM今早宣布将收购数据流公司Confluent。

IBM announced this morning it is buying Confluent, a data streaming company.

该公司将以110亿美元现金支付收购款项。

The company is paying $11,000,000,000 in cash.

IBM支付的每股31美元价格，较Confluent上周五的交易价格溢价33%。

The $31 share price IBM is paying is a 33% premium to what Confluent was trading at on Friday.

现在加入我们、为您详细解析此事的是自由资本市场的科技研究主管保罗·米克斯。

Joining me now to break it all down is Paul Meeks, Head of Technology Research at Freedom Capital Markets.

嘿，保罗。

Hey, Paul.

你怎么样？

How are doing?

很棒，老兄。

Great, man.

谢谢你邀请我。

Thanks for having me.

很高兴见到你。

Great to see you.

这周发生了很多事。

So there's a lot going on this week.

我们马上就会谈到甲骨文和博通的财报。

We'll get to earnings for Oracle and Broadcom in just a second.

但你对今天早上IBM收购Confluent的消息怎么看？

But what did you make of the IBM Confluent news this morning?

你知道吗，我觉得这是一笔明智的交易。

You know, I think it's a smart deal.

这是他们持续收购以拓展企业AI布局的又一例证。

It's a continued pattern of some acquisitions to build out their footprint in enterprise AI.

像许多注重估值的分析师一样，我认为他们付的钱太多了，你看，这家公司明年预计每股收益只有48美分，从今年的40美分增长而来。

Like a lot of analysts that are valuation sensitive, I think they paid way too much for it, right, a pretty You big know, here's a company that is only supposed to do 48¢ earnings per share next year, up from 40¢ this year.

所以EPS增长并不多，而你却支付了65倍的市盈率。

So not much EPS growth, and you're paying 65 times earnings.

他们的收入增长不错，但EPS增长却跟不上。

Revenue growth has been good there, but not EPS growth to match.

所以从战略上讲，我喜欢这笔交易。

So strategically, I like it.

这是一套不错的软件产品。

It's a nice product suite.

天啊，他们付的钱真多。

Boy, they're paying a lot.

你认为康夫卢ent合理的收购价格应该是多少？

How much do you think would have been a fair price for Confluent?

实际上，在大幅溢价之前，它周五的交易价格

Actually, where it was trading on Friday before the big premium markup,

我

I

我觉得已经很公平了，尤其是这是一个全现金交易，而康夫卢ent是一家规模较小的公司，将自己交到了IBM及其庞大的企业销售团队手中。

think, is more than fair, you know, particularly because it's an all cash deal and they are a rather small company putting themselves in the hand of IBM and IBM's enormous enterprise sales force.

你自然会期待这笔交易完成后，它的增长会有一个不错的提升。

You would expect it to have a nice growth kicker after the deal.

所以提醒一下我们，康夫卢ent到底做什么业务？然后我们再讨论IBM如何从战略上将其融入其AI产品组合。

So remind us what exactly Confluent does, and then we'll talk about how IBM sees it strategically fitting in its AI product suite.

这将使IBM能够更好地管理数据流和实时数据流。

So it is going to allow IBM to better manage data streaming, real time data streaming.

因此，这是对产品组合的补充。

And so it's a addition to the product suite.

这是箭袋中的另一支箭，正如我所说，这非常合理。

It is another arrow in the quiver, and it makes a lot of sense, as I said.

对。

Right.

你能告诉我一下IBM如今的情况吗？我的意思是，这家公司其实很早就涉足AI了，对吧？

And tell me a little bit about where IBM these days, I mean, you know, it's a company in many ways, I mean, they were very early to AI, right?

我的意思是，早在IBM Watson时代，那就是我们最早见到的AI之一。

I mean, back in the the IBM Watson, that was some of the earliest AI that we saw.

从那以后，它在AI对话中的地位发生了什么变化？

What's happened to that company in terms of being included in the conversation?

我觉得我们现在很少再提到它了。

I feel like we don't talk about it as much.

我们知道它有一个庞大的咨询业务。

We know it has a big consulting practice.

这家公司现在处于什么位置？

Where is the company now?

所以我认为它其实经历了一次不错的复兴，对吧？

So I actually think it's had a nice renaissance, right?

这曾经是我祖父时代的标志性科技公司。

This used to be my grandfather's marquee tech company.

它长期陷入低迷状态。

It went into the doldrums for a long period of time.

每个季度财报公布时，他们的收入都同比下滑，这种情况持续了多年。

Each and every quarter that they reported, their revenues would be down year to year, and that went on for years.

对。

Right.

因此，随着新任CEO上任后迅速扭转公司方向，早在几年前就转向了并非人人都在追逐的公有云，而是混合私有云模式，这实际上非常契合企业客户的需求。

So with the relatively new CEO turning the company on a dime, and several years ago, essentially pivoting to not the public cloud that everybody else was pivoting to, but a hybrid private public crowd, which actually worked very well with enterprise customers.

因此，他们迎来了一波显著的增长，至少是从一家沉寂已久的公司起步的。

And so they have seen a nice burst of growth, at least coming from a very sleepy company.

他们还进行了一些相当精明的补充性收购，比如今天这一桩。

They've made some fairly savvy, like maybe this one today, bolt on acquisitions.

无论何时，只要你是一家大公司的首席技术官，你都会关注他们。

They're always going to be somebody that if you are a chief technology officer of a big company, you're always going to look at them.

几十年来，他们一直是企业领域的标志性公司，现在他们正试图在人工智能领域重现这一地位。

Decades and decades, they've been a marquee in the enterprise, and they're trying to do that here in AI.

当然，他们目前也在大力投入量子计算。

And of course, they're also working very hard on quantum computing right now.

让我们转向本周即将公布的某些新闻。

Let's pivot to talking about some news that we have coming up this week.

我们有博通和甲骨文即将公布财报。

We have Broadcom and Oracle reporting earnings.

我们先从甲骨文开始。

Let's start with Oracle.

你对他们的财报数据关注哪些方面？

What are you watching for with that print?

你知道，这个数据非常敏感，因为上个季度，他们宣布了巨额的收入待执行订单，后来我们发现这些几乎全部来自OpenAI，财报发布后股价上涨了三分之一。

You know, this one is very sensitive, because remember last quarter, they announced this huge revenue backlog, which we find out later on is essentially all OpenAI, and the stock after reported went up a third.

然后人们说，天哪，他们得花一大笔钱。

And then people said, Oh my goodness, they're going to have to spend a lot of money.

他们将从产生大量自由现金流转变为消耗现金流。

They're going from generating a lot of free cash flow to burning it.

我们将看到OpenAI是否能支付这笔账单。

And we'll see if OpenAI is going to be able to pay the bill.

于是股价暴跌，甚至跌得比之前还要低。我真正需要关注的是，我并不一定担心他们本季度的营收和每股收益表现。

And so then the stock collapsed and it went down even lower than what So it was the what I need to see here is I'm not necessarily worried about what they report this quarter for revenue and EPS.

我相信它们会不错，但他们必须做几件事，否则就会陷入困境。

I'm sure it'll be fine, but they're gonna have to do a couple of things, or else they're gonna be in some trouble.

他们必须让我们相信，这个OpenAI的未确认收入 backlog 是真实的，并且会逐步确认。

They're gonna have to give us a, comfort that this OpenAI backlog is for real, and it will be debooked.

此外，他们还需要让我们相信，经过所有这些AI基础设施支出后，最终他们仍能恢复现金流并保持不错的利润率。

And also, they're going to have to give us some comfort that at the end of the day, after all this AI infrastructure spending, they are going to go back to generating cash and making a pretty good profit margin.

因为这一直是市场所期待的叙事。

Because that has been the narrative.

从上一次电话会后立即转为正面，到如今变得极其负面，这就是我听到的情况。

It went from immediately positive after their last call to horribly negative, and that's where I heard.

他们必须改变这种叙事。

They're going to have to change the narrative.

没错。

Right.

这简直就像，有时候你可以不听电话会也没关系。

It's almost like, Look, sometimes you can get away with not listening to the call.

你只需要阅读新闻稿、看看数据，就能获取你需要的信息。

You can just read the press release, read the numbers, and get the information you need.

但这一次，我认为恰恰相反。

This one, I actually think it's the opposite.

我认为在某些情况下，你甚至可以跳过这些数据。

I think you can probably skip the numbers in some cases.

你必须重点关注高管们说了什么。

You really got to laser in on what executives say.

这简直就像美联储的声明。

It's almost like the Federal Reserve statements.

我的意思是，人们会非常挑剔地分析：他们为什么这么说？

I mean, people are going to be really nitpicking in terms of, Well, why did they say that?

这是不是意味着经济增速会放缓？

Is that a signal for things going slower?

是的。

Yeah.

你提出的这一点非常好。

It's a very good point you make.

例如，他们报告的收入和每股收益真的重要吗？

For example, does it really matter what they report with revenue and EPS?

我甚至不确定他们对收入和每股收益的指引是否真的重要。

I don't even know if it really matters what they report for their guidance revenue and EPS.

这将是长期的。

It's going to be long term.

这场AI热潮的尽头会有一罐黄金吗？

Is there going to be a pot of gold at the end of this AI rainbow?

而且你极度依赖OpenAI。

And you are super dependent on OpenAI.

这明智吗？

Is that savvy?

我们走着瞧吧。

We'll see.

巴里，快说一下，在我们谈博通之前，你怎么看？我们最近看到很多新闻说甲骨文的信用违约互换价格上升了。

Barry, quickly, before we get to Broadcom, what do you think of We've read a lot in the news about the credit default swaps for Oracle having ticked up in price.

你怎么看待这件事？

What do you make of that?

老实说，这是个负面信号，但我认为即使甲骨文现在负债较重，也不太可能面临无法偿还债务的问题。

Well, it's honestly a negative, but I don't think that you'll ever have a situation where Oracle, even though it's now stretched, will have any problem paying its debts.

我觉得甲骨文会没事的。

I think Oracle is going to be fine.

你现在可能会问，你在花钱吗？

Now you may say, Are you spending your money?

你这样做聪明吗？

Are you being smart?

这确实是个问题，但我认为它们会没事的。

You know, that's in question, but I think they'll be just fine.

对吧？

Right?

这是一家会花钱的公司，但它们一直是巨额自由现金流的创造者，你想想看。

Here's a company that they will spend, but they were a prodigious free cash flow generator, and you think about it.

如果它们任何时候想筹钱，华尔街都会乐意提供资金。

If they ever wanted to raise money, Wall Street would be happy to give it to them.

对。

Right.

快说说，博通怎么样？

Very quickly, what about Broadcom?

你在那方面关注什么？

What are you watching for there?

博通会非常有趣，因为其关键竞争对手马威尔科技正在为超大规模客户设计这些XPUs，因此它们不一定依赖英伟达的GPU。

So Broadcom will be very interesting because Marbell Technology, its key competitor in, essentially designing these XPUs for the hyperscalers so they're not necessarily beholden to NVIDIA's GPUs.

尽管马威尔的股票因华尔街的下调评级而今日大跌，但其业绩报告依然出色。

Marvell had a great report even though the stock's getting hammered today on a Wall Street downgrade.

因此，博通必须做到同样的成绩，因为博通的估值相比马威尔有很高的溢价。

So it'll be incumbent upon Broadcom to do the same because Broadcom is priced, is valued at a steep premium to Marvell.

所以它们必须证明自己确实物有所值，否则我认为估值差距会缩小。

And so they're gonna have to show that they're really worth it, or I think the valuation gap will close.

我们都清楚这将是好消息，但它们必须让我们确信，它们正在从这些超大规模客户手中赢得份额，尤其是在帮助他们设计类似英伟达GPU的芯片方面，对抗马威尔。

And so we all know it's going be good news, but they're going to have to make us feel very good that they are gaining share, particularly against Broadcom, with these hyperscalers to help them design their essentially NVIDIA GPU lookalike chips.

对。

Right.

保罗，感谢你前来做客。

Well, Paul, I want to thank you for coming on.

这周很忙，祝你周一愉快。

It is a busy week, and so happy Monday to you.

祝你一切顺利，我们很快再聊。

Good luck with everything, and we'll talk to you again very soon.

谢谢你来参加节目。

Thanks for coming on the show.

兄弟，祝你好运。

Best wishes, brother.

好的。

Okay.

自从萨姆·阿尔特曼上周发出红色警报以来，我们的新闻团队一直在紧急联系公司内部的消息源，以了解实际情况的变化。在所有报道中，一个名字反复出现，那就是负责ChatGPT产品业务的尼克·特利。

Since the Code Red that Sam Altman issued last week, our newsroom has been feverishly talking to sources inside the company to understand how things have been changing on the ground, and through all their reporting, one name has come up time and time again, and that is Nick Turley, who heads up product for ChatGPT.

今天请来斯里·穆皮蒂，为我们详细介绍为什么特利最近在OpenAI的业务中变得如此关键。他上周末刚合作撰写了一篇关于特利的精彩专访。

Here to tell us more about why Turley has become such a central figure at OpenAI's business these days, I want to bring on Sri Mupiti, who co authored a great profile of him over the weekend.

斯里，欢迎来到节目。

Sri, welcome to the show.

周一快乐。

Happy Monday.

周一快乐。

Happy Monday.

很高兴来到这里。

Great to be here.

听起来你上周末花了大量时间与OpenAI的人交谈。

So sounds like you had a busy weekend talking to folks at OpenAI.

给我们讲讲Nick Turley是谁。

Tell us about who Nick Turley is.

当然。

Of course.

Nick Turley是OpenAI的ChatGPT负责人。

So Nick Turley is the head of ChatGPT for OpenAI.

对于一家拥有八亿以上用户的公司来说，这是一个巨大的责任。

And that is a huge responsibility for a company that has 800,000,000 plus users.

尼克·特利的主要职责是支持ChatGPT的增长，以及任何其他支持这一增长的商业模式。

Nick Turley's main role is to support the growth of ChatGPT along with any other types of business models that support that growth as well.

例如，探索广告作为潜在机会。

For example, exploring ads as a as an opportunity.

但同时也包括确保拥有足够的GPU来服务ChatGPT。

But it also includes making sure that you have enough GPUs, for example, to be able to service ChatGPT as well.

他实际上从2020年初或2021年就加入了公司，早在ChatGPT发布之前就已加入。

He actually has been at the company for since the start of '20 or since 2021 before even ChatGPT even launched.

因此，他是从一开始就参与ChatGPT发布的核心人员之一，并见证了该产品在过去几年中的成长。

So he was one of the main folks involved with launching ChatGPT from the beginning and seen has seen that growth of that product over the last couple years.

所以，他并不仅仅参与ChatGPT前端设计和产品方面的工作。

So it's not that he's only involved in sort of the front end design and the product side of Chateapiti.

他还全程参与，确保系统拥有足够的计算资源来运行。

He's also involved end to end in making sure that the thing has enough compute to operate.

也就是说，这是一个非常全面的职责。

I mean, it's very much a comprehensive role.

是的，没错。

Yes, exactly.

我认为他最初是一名个体贡献者产品经理，这意味着你直接参与产品的构建。

I think he began as an IC product manager, which means that you're specifically involved with building the product itself.

但后来

But then

在

over

这些年里，他的职责增加了。

the years, his responsibilities has increased.

因此，他在招聘人员和分配职责方面拥有更多控制权。

So he had more control in terms of hiring people and delegating responsibilities.

今年，他被提升为ChatGPT负责人，

Then this year he was promoted to the head of ChatGPT role,

也就是

which is

这是一个跨职能的角色，旨在确保产品尽可能顺利地推进，同时确保ChatGPT获得成功所需的资源。

more of a cross functional role to be able to ensure that one product is going as well as it could be, but also ensuring that he has the necessary resources for ChatGPT to succeed.

谈谈他在过去一周里角色的变化吧，毕竟我们目前正处于OpenAI的红色警报状态。

Talk to me about how his role has changed now in the last week, given that we are amidst this code red at OpenAI.

为什么他变得如此重要？

Why has he become such a more important figure?

他是不是在统筹全局？

Is he quarterbacking things?

我认为尼克的角色在过去一年左右已经得到了巩固。

I'd say that Nick's role has really been something that has been solidified over the last year or so.

我认为红色警报更凸显了ChatGPT在OpenAI增长中的关键性和重要性。

I think more so Code Red just emphasizes that ChatGPT is such a pivotal, important part of OpenAI's growth.

例如，拥有八亿以上的用户，并且能够持续增长，到2030年实现每周活跃用户达到26亿，这是一个非常艰巨的目标，但我们需要一个可以依靠的人。

For example, having 800,000,000 plus and being able to continue that growth to be able to hit 2,600,000,000 weekly active users by 2030 is a really intense goal, but to have someone to be able to look to.

例如，尼克似乎每周都会分享一些他认为最重要的Slack消息，提醒大家各自的优先事项。

For example, Nick apparently has weekly top of mind Slack messages that he shares with folks on their priorities.

他为人们撰写备忘录，帮助理解战略，并深度参与日常与公司各部门的协作。

He writes these memos for people to understand the strategy and gets really involved with working with folks day to day across different types of the org.

我认为尼克现在被推到前台，主要是因为他作为ChatGPT负责人的角色，确保ChatGPT能够以前所未有的速度持续增长。

I think that why Nick is sort of being put to the front now is just because of his role as the head of ChatGPT to make sure that ChatGPT is able to continue to grow more than ever.

在我们之前的多次对话中，有一个观点反复出现：OpenAI始终处于紧急状态，始终在推进事务。

Across the different conversations that we've had, one of the things that came up is that OpenAI is always in an urgent mode and always towards things.

因此，有些人认为，上周的情况与以往几周并无不同。

And so some folks, for example, have described it as like this last week being no different from weeks prior.

没错。

Right.

他们其实并不完全认同‘代码红色’这个说法，但我认为萨姆的备忘录确实强调了这确实是一场真正的红色警报。

They didn't really actually agree with the code part, but I think that Sam's memo really emphasized that there is truly a code red.

所以我认为这里存在不同的视角。

So I think there's a different perspectives there.

你现在和他谈过了。

Now you talked to him.

你几天前跟他进行了一次采访。

You had an interview with him a couple days ago.

从那次对话中你学到了什么？

What did you learn from that conversation?

是的，正是如此。

Yes, exactly.

与尼克交谈有助于了解他的优先事项，这非常好。

So it was great to talk to Nick in terms of understanding his priorities.

要让ChatGPT持续增长，关键之一是尼克的愿景，即让它成为人们的超级助手，帮助他们完成原本无法做到的事情，他称之为几乎赋予他们超能力。

So one of the key things about making ChatGPT to continue to grow is Nick's vision of being able to have it be a super assistant for people, where he described it as almost giving them superpowers to do things that they weren't able to do.

例如，如果你不会编程，它可以帮你写代码；如果你无法看医生，它可以提供医疗建议。

For example, coding if you aren't able to code or getting medical advice if you aren't able to visit a doctor.

我认为这种超级助手的愿景与他其他优先事项高度契合，其中之一是让产品高度个性化。

And I think that super assistant vision really fits into sort of his other priorities of making one, the product very personalized to you.

这意味着，随着你使用产品的次数增多，它会越来越了解你，从而提供像谢丽所必需的那些功能。

So that means that as you use the product more, it understands you more and can provide, for example, the things that Sheree specifically needs.

第二点是让ChatGPT具备上下文能力。

The second is to have the context of ChatGPT.

这意味着将它与不同的应用程序以及对你重要的事物连接起来。

So that means connecting it to different apps and things that are important to you.

例如，我的Slack消息或我的照片，如果我需要查找我狗狗的不同照片的话。

So for example, my Slack messages or my photos, if I needed to pull up different types of photos of my dog, for example.

对。

Right.

第三点是代表用户采取行动。

And the third is to actually take action on behalf of the user.

这是在搜索功能基础上的延伸。

And so this is building upon the search functionality.

尼克实际上是我们在开发SearchGPT（即集成到ChatGPT中的搜索功能）时发现的关键人物之一，以确保这些功能

And Nick was actually one of the key folks that we found out of leading SearchGPT, which is the search features that is incorporated into ChatGPT to make sure that they're

并非全部都是最新的。

not all up to date.

那我问你一个问题。

So let me ask you this.

你和他有过这样的对话，而且我相信你和我们的同事在报道过程中，曾与20位过去或现在与尼克共事过的人交谈过。

You had this conversation with him, and you certainly spoke with, I think you and our colleagues spoke with 20 people who either worked with Nick in the past or are currently working with him through the course of your reporting.

从这些对话和你的所有报道中，你是否觉得尼克已经掌控了局面？

From that conversation, from all your reporting, did you walk away thinking that Nick's got it under control.

我的意思是，他们最终会解决这个问题的。

I mean, they're going to figure this out.

OpenAI 目前领先，Code Red 是关于这个主导权的逐步聚焦，但这是为了看看他们是否能扩大优势。

OpenAI, they have this lead that Code Red is about the lead sort of narrowing, but this is to sort of see if they can widen it.

你离开时是否感到有信心？

Did you leave away with a sense of confidence?

还是你离开时觉得，你知道吗？

Or did you leave the conversation thinking, You know what?

我觉得现在球还在空中飞着呢。

I think the balls are in the air right now.

我的意思是，我们仍然需要观察事情如何发展。

Mean, we still need to see how this plays out.

我认为人工智能发展得太快了，你真的无法预测谁会保持领先。

I think that AI is moving so quickly that you really don't know who's going to stay in the lead.

我觉得看到OpenAI、谷歌和其他公司能够竞争并迅速前进，将会非常令人兴奋。

I think it's just going to be really exciting to see OpenAI and Google and others be able to compete and move move ahead quickly.

我认为，人们在与我们的对话中给尼克的反馈之一是，有时他过于扎根于研究人员正在处理的具体研究问题。

I think one of the areas of feedback that, like, folks had given Nick in terms of our conversations was that at times that he as it has technically rooted in some of the research specific research problems that researchers were working on.

因此，随着研究不断深入和快速进步，我很想知道他如何适应并将其融入其中。

And so I think as research continues to increase and rapidly improve, I'm I'm curious to see sort of how he's able to adapt to that and be able to incorporate it.

到目前为止，他在将ChatGPT转化为基于GPT-3研究的产品方面做得非常出色。

So far, he's been able to do a great job of it so far in terms of actually turning ChatGPT into a product that folks are using based on GPT-three research.

但我认为，公司仍有继续迭代的空间。

But I think that there's still sort of opportunity for the company to continue iterating.

因此，我很期待看到接下来的发展。

So I am curious to see.

我认为还有很多问题是我很想得到解答的。

I think there's a lot of questions still that I'd love to have answered.

例如，你们将如何实现超级助手的愿景，同时兼顾OpenAI的其他优先事项？

For example, how will you be able to pull off the super assistant vision and with all the sort of other priorities that OpenAI has?

我知道萨姆已经试图让大家专注于ChatGPT，但我知道，比如……

I know Sam has sort of tried to focus everyone to focus on ChatGPT, but I know for example Right.

还有一些努力，比如在广告方面进行投入。

Efforts, for example, to focus on advertising in some ways as well.

所以这肯定将是一个令人兴奋的时期。

So it'll be an exciting time for sure.

这是一项艰巨的任务。

It's a big job.

所以我们非常感谢你向我们推荐尼克·特利，并邀请他来节目中谈论他。

So we we appreciate you bringing Nick Turley to our attention and coming on the show and talking about him.

非常感谢你。

Thank you so much.

我们将在本周晚些时候再与您联系，我相信届时会有更多新闻被揭露。

We will get back to you later this week as I'm sure there'll be more news to uncover.

这是来自《信息》杂志的斯里·穆皮尼，他负责报道OpenAI和Anthropic的所有动态。

That is Sri Mupini, who covers all things OpenAI and Anthropic here at The Information.

好的。

Okay.

当OpenAI的研究人员在应对红色警报时，许多其他研究人员正聚集在人工智能研究领域最重要的会议之一——神经信息处理系统大会。

While researchers at OpenAI were responding to a code red, many other researchers were gathered at one of the most important conferences for AI research, the Neural Information Processing Systems Conference.

这并不是一个关于神经科学的会议，而是一个关于人工智能的会议。

It's not a conference about neuroscience, it's a conference about AI.

我的同事斯蒂芬妮·帕拉佐洛一整周都在现场。

My colleague Stephanie Palazzolo was there on the ground all week.

她今天早上写了一篇文章，讲述了她所了解到的内容，我想请她来告诉我们，现场最热议的话题是什么。

She wrote a story this morning about what she learned, and I want to bring her on to tell us more about what the talk of the town was on the ground.

斯蒂芬妮，早上好。

Stephanie, good morning.

很高兴你在这里。

It's great to have you here.

嗨，早上好。

Hi, good morning.

所以，这个会议名字真差劲，说实话，但据说是研究者必去的地方，它到底是什么？

So what is this conference that has a terrible name, if I'm being honest, but apparently is the place to go for research?

是的。

Yeah.

它更常被称为NeurIPS，比全称容易说多了。

So it's much more commonly known as NeurIPS, which is a lot easier to say than the full name.

但本质上，NeurIPS是每年举办的规模最大AI研究会议之一。

But essentially, NeurIPS is one of the kind of largest conferences held each year for AI researchers.

过去，它主要聚焦于学术研究者。

And so in the past, it's been really more focused on academic researchers.

但近年来，它变得越来越受欢迎，很多人都去参加。

But in more recent years, it's become, you know, much more popular for people to go to.

今年，参会人数实际上超过了25,000人，不仅包括研究人员，还有初创公司创始人、希望投资他们的投资者，当然还有像我这样的记者。

And so this year they actually had more than 25,000 people there, including not only researchers, but now startup founders, investors that want to back them, and of course, journalists like myself.

你用亚马逊的AI研究员大卫·劳恩的一句话作为你文章的开篇。

So, you opened your story with this line from David Lawn, who does AI research at Amazon.

你开头引用的那句话是：‘我敢保证，我们今天训练模型的方式不会持续下去。’

The quote that you started with is, he said, I will guarantee that the way we're training models today will not last.

为什么这句话引起了你的注意？

Why did that line stand out to you?

他这句话到底是什么意思？

What did he even mean by that?

是的。

Yeah.

大卫之前是Adept公司的首席执行官，再之前，他还长期担任OpenAI的工程负责人。

So David, who, you know, previously was the CEO of a company called Adept, and then also, prior to that, he had also led engineering at OpenAI for for quite a while.

每当他发表这类评论时，他都迅速澄清，他并不是说预训练、规模扩展或AI已经终结了。

He was very quick to say, you know, whenever he says this sort of comment, he doesn't mean that pre training is dead or scaling is dead or AI is over.

但他基本上是在说，我们显然缺少一些关键要素，而这些要素是实现通用人工智能所必需的。

But he's basically just saying that we're clearly missing some kind of key ingredients that we need to have in order to reach artificial general intelligence.

他相信，为了实现通用人工智能，模型的发展在未来必须发生巨大变化。

He does believe that model development is going to have to change a lot over time to be able to reach AGI.

因此，例如，他和会议上许多其他研究人员都谈到了一种叫做持续学习的重要概念，我在今天早上的文章中也写到了这一点。

So for instance, one thing that he and a lot of other researchers at the conference talked about, which I wrote about in this morning's piece, is the importance of something called continual learning.

简单来说，这就是指人工智能在训练完成后仍能继续学习的能力。

Basically, that is AI that's able to keep on learning even after it's been trained.

大卫还深入探讨了模型训练步骤的顺序问题，并指出当前的顺序未必合理。

David also talked a lot about the order of steps in training a model and how it doesn't necessarily make a ton of sense.

目前，在称为预训练的阶段，需要耗费大量的计算资源和时间，模型会浏览来自互联网和书籍的海量数据。

Right now there's a ton of compute and time that's spent in a phase called pre training, where a model just looks at a ton of data from like the Internet and books.

但大卫的核心观点是：既然我们未来要让模型完成特定任务，为什么还要花大量时间向它展示这些随机、宽泛但最终可能并不相关的互联网数据呢？我们应当更多地展示与这些任务相关的数据。

But David basically made a point which is, you know, why are we spending so much time showing models this, kind of random, very general, but like ultimately maybe not super relevant data from the internet when we should be showing it more data that's related to the sorts of tasks that we're going to be asking the model to do in the future.

那么，所有这些如何与你提到的关于物理科学和生物学等领域所引发的问题联系起来呢？

Now, how does all this connect to some of the questions that you wrote about being raised around the physical sciences and biology, for example?

人工智能的一大承诺是，聊天机器人只是其中之一，但它实际上可能带来我们至今尚未找到的某些疾病的治疗方法。

This has been a big promise of AI that, hey, the chatbots are one thing, but this could actually lead to cures for some of the diseases that we haven't been able to find them for yet.

看完你的报道后，我感觉人们至少对近期的前景持怀疑态度，也怀疑以目前人们研究AI的方式，这一承诺能否真正实现。

I walked away from your story thinking that maybe people were a little bit skeptical of at least the near term promise and also that promise actually being achieved given the way people are working on it.

跟我们聊聊这方面的情况吧。

Talk to us a little bit about that.

是的。

Yeah.

我想强调的是，显然AI公司如今已经通过现有的AI产品赚取了大量利润，即使我们还没达到AI发现新药或进入超级智能阶段的地步。

So, I mean, I think it's important to mention that clearly AI companies are making a lot of money today from existing AI products, even if we're not to the point where AI is finding new cures for diseases and we're clearly not at this kind of super intelligence phase.

所以，无论未来如何发展，即使AI的进步今天就停止了，依然有大量利润可图。

So there's a lot of money to be made, you know, no matter what happens, even if the AI progress were to stop today.

但总的来说，不仅在这次大会上，从过去几周的新闻报道以及我们接触的信源来看，人们对AI模型是否能超越其训练数据实现泛化，正日益产生怀疑。

But yeah, I think in general, and not only at this conference, just from news stories we've seen in the past couple of weeks and sources that we've talked to, there has been kind of this growing skepticism around whether AI models will be able to kind of generalize beyond the data that they were trained on.

因此，它们能否在训练数据中从未出现过的情况下提出新发现，并超越训练它们的研究人员的知识边界？

And so, you know, will they be able to come up with new discoveries that they've never seen before in training data and go beyond kind of the knowledge of the researchers that are training them.

还有像OpenAI和Anthropic这样的公司正在聘请人类专家，让他们编写生物学和医学等领域中困难问题和答案的示例。

And also the human experts that companies like OpenAI and Anthropic are hiring to, you know, write examples of tough answers and questions and fields like biology and medicine.

所以我认为，许多研究人员都认为，如果人工智能能够实现某种持续学习，它就能更像人类那样学习——我们不需要看到成千上万个解决数学问题的例子，就能学会如何解数学题。

So I think a lot of the researchers, a lot of the researchers think that if AI is able to do some sort of continual learning, then it would be able to learn more similar to the way that humans do, where we don't need to be shown like 10,000 examples of how to solve a math problem to be able to learn how to solve a math problem.

我们基本上是通过经验和试错来学习的。

We kind of just learn from experience and from trial and error basically.

对。

Right.

我知道这是一场非常热门的会议，所以非常感谢你来跟我们分享这一切。

Well, I know that it was a very buzzy conference, and so I want to thank you for coming on to tell us all about it.

我们很快再跟你聊。

We'll talk to you again very soon.

这位是我们在《The Information》的AI记者斯蒂芬妮·帕拉佐洛。

That is Stephanie Palazzolo, our AI reporter here at The Information.

好的，我们将在节目中推出一个关于加密货币的全新每周栏目。

Okay, we are launching a new weekly segment here on the show about crypto.

每周，我们都会邀请我们的加密货币记者杨越驰，来讨论加密世界中最热门的话题以及她对当日最重要新闻的分析。

Every week, we'll bring on our crypto reporter, Yuechi Yang, to talk about what the hottest topics are in the crypto world and what her analysis is of the biggest stories of the day.

在过去一周里，Kalshi 因其 10 亿美元的融资和 110 亿美元的估值成为关注的焦点。

Over the past week, Kalshi has been one of the central points of attention with its $1,000,000,000 funding round at an $11,000,000,000 valuation.

这一估值比它在十月时的估值高出两倍以上。

That is more than twice the valuation that it fetched in October.

我想请杨越驰来帮我们深入解析这一切。

And I wanna bring on Yuecchi to help us break this all down.

杨越驰，很高兴见到你。

Yuecchi, it's great to see you.

欢迎再次回到节目。

Welcome to welcome back to the show.

嘿，阿卡什。

Hey, Akash.

所以，在我们的首个每周加密货币环节中，我们将讨论预测市场。

So for our first segment, the weekly crypto segment, we are going to talk about prediction markets.