本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
大家好，欢迎收看Sharp Tech的免费试听节目。
Hello, and welcome to a free preview of Sharp Tech.
大家好，欢迎回到Sharp Tech的又一期节目。
Hello, and welcome back to another episode of Sharp Tech.
我是安德鲁·夏普，连线的另一位是本·汤普森。
I'm Andrew Sharp, and on the other line, Ben Thompson.
本，你最近怎么样？
Ben, how you doing?
我刚才在想，可能是在上半场，或者下半场刚开始的时候，突然想到我们Sharp Tech的尼克发来的邮件，然后就直接切入主题了。
Well, I was, I think, at some point, might have been the first half, early in the second half, thinking about our Sharp Tech Nick's emailer, and just, like, coming in hot He
他这是自找的。
did it to himself.
一开始就这样。
Early.
没错。
Yep.
跟他说现在提这个太早了，还在想等事情崩盘的时候得多好笑。
Telling him it was too early, thinking about how funny it was gonna be when it all fell apart.
不过到现在我也不太清楚到底发生了什么。
And then, I'm still not entirely sure what happened.
我的意思是，
I mean,
我到现在还是不敢相信居然出了这种事。
I still can't believe what happened.
我想起一个叫卡梅隆的澳大利亚人，一个月前给我和戈利弗发邮件，问该不该买NBA总决赛的票，从悉尼飞到纽约去看。
I was thinking of a goat emailer, Cameron in Australia, who emailed me and Goliver, like, a month ago asking whether he should buy tickets to the NBA finals and fly to New York City from Sydney, Australia.
我当时就建议他：去吧。
And I recommended at the time, do it.
人生苦短。
Life is short.
对。
Yeah.
这得花一大笔钱。
It's gonna be crazy expensive.
但如果这是你这辈子想做的事，现在就是最好的时机。
But if this is a bucket list item, this is your opportunity.
谁知道尼克斯队什么时候还能再回来呢？
Who knows when the Knicks will be back here?
干脆就去吧。
Just make the move.
结果他真去了，花了不知道多少钱买了第三场和第四场的票。
And then he did it, and he spent God knows how much on tickets to game three and game four.
我当时就坐在那儿。
And, I was sitting there.
他们输掉了第三场，第四场半场就落后30分，我特别后悔，居然劝他去看比赛。
They lost game three and watching game four down 30 at halftime, feeling really guilty that I was encouraging Cameron to go to the game.
结果那场比赛成了他，也是我们所有人记忆中最没意思的一场。
And then it turned into probably the most forgettable game of his life and any of our lives.
所以我们现在在这儿了。
So here we are.
经历了这件事之后，再贵的价钱都不算贵了。
There's no dollar figure that could have made it feel too expensive after this.
天哪。
My god.
我真是太开心了。
I'm so happy
你之前一直担心他，还特意发消息安慰，而且你发得特别及时。
for messaged as such about how worried you were about him, and I think you messaged it exactly at the Nadine.
我记得他们当时落后29分，然后比赛突然翻盘了。
I think they were down 29, and the game turned.
多亏了你的镜头。
Thanks to your camera.
局势逆转了。
Tide was turned.
天哪。
Oh my god.
剩下的就是历史了。
The rest is history.
我本来打算开导你的。
I was gonna counsel you.
听好了。
Look.
你不需要为别人的决定负责。
You need to not take responsibility for other people's decisions.
卡梅隆是独立的个体。
Cameron is his own person.
这一点已经说得很清楚了。
Made that clear.
说明一下，他并没有把责任推给我，但我还是在第四场比赛期间感到内疚。
To be clear, he didn't put it on me, but I still felt guilty during game four.
不。
No.
但也许我错了。
But maybe I'm wrong.
如果你真能扭转局势，那也许你确实应该
If you're if you were able to turn the tide, maybe you do need to be
更关注自己的优势。
caring more about your advantage.
所以，所有在场的尼克斯球迷，接下来几天请记得谢谢我。
So And every Knicks fan in the audience, thank me over the next several days.
不如先让他们赢一场再说。
Well, let's let them win first.
对。
Yeah.
先让他们赢一场再说。
Let's let them win first.
咱们先让他们把系列赛打完吧。
Let's let them finish the series.
因为现在要是他们在总决赛输了，那就更搞笑啦。
It is because now now in finals loss would be even funnier.
所以，你做好心理准备吧。
So, brace yourself.
没错。
Absolutely.
那目前来说，本，咱们先聊一聊 WWDC。
Well, for now, Ben, we will begin with WWDC.
我们待会儿会详细聊聊 Anthropic 和他们的新模型 Fable。
We're gonna talk through Anthropic and its new Fable model later in the show.
不过首先，我每年都很期待 WWDC 的讨论。
But first, I look forward to WWDC convos every year.
作为科技爱好者吗？
As a tech guy?
作为一个科技爱好者，一个Vision Pro的铁粉。
Well, as a tech guy, as a vision pro fanatic.
今年的发布会里提到了几次Vision Pro，真让人高兴。
The vision pro got a couple mentions in this year's keynote, so love to see that.
还没死呢。
Not dead yet.
确实还没死。
It's not dead yet.
你知道的。
You know?
Vision Pro上有了Siri，但总的来说，这更像是苹果的年度状态汇报。
We got Siri on the Vision Pro, but in general, it's sort of an Apple state of the union.
这也是我衡量Sharp Tech发展的一个标志。
It's also how I marked the time on Sharp Tech.
我到现在还经常想起当初发布Vision Pro和Apple Intelligence的那个WWDC。
I still think back to the initial WWDC with Vision Pro, Apple Intelligence.
我还有一个看法，这次WWDC对你来说其实特别重要。
I have an extra I have an extra take, on here that this was actually an incredibly important WWDC for you.
为什么这么说？
Why is that?
因为苹果公司里最负责、也最热衷Vision Pro的人是迈克·洛克威尔，他主动请缨，说让我来改进Siri吧。
Because the person at Apple most responsible and most passionate about the Vision Pro is Mike Rockwell, who basically put up his hand and said, let me fix Siri, please.
让那些不靠谱的人滚开，我们得换人。
Let someone else like, like we need to get rid of the clowns that are in charge.
让我来接手。
Put me in charge.
尽管他原本来自杜比实验室，跟AI开发看起来一点关系都没有。
Despite the fact he's like from like, I think like Dolby Logic origin like no, like, obvious sort of connection to building AI.
如果这次WWDC表现顺利，而且他真能把事情做成，他的地位肯定会大幅提升。
And it's one of those things if WWDC went well and if he's pulls it off, he's going to rise in power.
是的。
Yeah.
他会在这家公司里地位和影响力都会上升。
Well, he's gonna rise in power and prominence at the company.
其实大家对这一点一直有疑问。
There's there's actually been questions about that.
他那个位置上还牵扯不少内部博弈。
Some political intrigue around his position there.
如果他走了，我觉得你的Vision Pro可能会出问题。
And if he were to leave, I think your vision pro might be in trouble.
如果他成功了，对你的Vision Pro反而是好事。
And if he were to succeed, it might be good for your vision pro.
所以这背后赌注不小。
So there's a lot at stake.
有意思的是，古尔曼的报道里提到过这件事，说Rockwell被提拔了，我看到的时候都愣了一下，因为苹果之所以重用Mike Rockwell，是因为Vision Pro虽然商业上不算成功，但技术上是过关的。
It's funny because there was like the line in Gurman's report around that and around the elevation of Rockwell that did prompt a double take in me as I was reading it because Apple looked to Mike Rockwell because Vision Pro was considered a technical success, if not a commercial success.
我没想到公司内部会这么看待这件事。
I'm surprised that's the way it's interpreted internally.
我是觉得这技术确实了不起，但没想到Rockwell会因为苹果这几年的投入而获得这么多认可。
I mean, I think it's a technological marvel, but I'm surprised that Rockwell is getting any credibility out of what Apple's committed to over the last several years.
那就走着瞧吧。
Well, maybe we'll see.
其实几个月前就有报道提到过，你知道的。
I mean, like like the in, you know, there were reports a few months ago.
他对向克雷格·费德里吉汇报这件事并不太满意。
He's not super happy about reporting to Craig Federighi.
明白了。
Okay.
就在前一周，这事又被人提了一次。
That came out again sort of the week before.
我觉得，如果这事能成，他能管更多事情，对大家可能都是好事。
So I think, if this if this works out and he gets put in charge of more stuff, that might be, might be good for everyone.
咱们等着看吧。
We'll see.
瞧好了。
There you go.
而且这儿的宫斗戏还会一直持续下去。
Well, and the palace intrigue will continue indefinitely here.
目前来说。
For now
不过说清楚一点，这可不是我的报道范围。
Not really my beat, to be clear.
但确实如此。
But Yeah.
毕竟我知道这对你有多重要。
Well Because I know how important it is to you.
我一向关注德国的动态。
And I and I look to German.
你知道的，我喜欢看他那种窥探内幕、解读风向的报道，不过我这儿还有六个问题，咱们可以快速过一遍，就在节目开头。
You know, I enjoy reading his sort of, tea leaf reading and insidery reporting, but I have a six pack of questions that we can run through quickly at the top of the show here.
WWDC之后我们收到了大量邮件，但我先说说加布里埃尔的来信，她是在去WWDC的路上发的，上周就发来了。
We got a lot of emails on the heels of WWDC, but I'll begin with Gabrielle who wrote on the way into WWDC, sent this email last week.
本和安德鲁，苹果有没有可能在WWDC之后给外界留下好印象？
Ben and Andrew, is there any way for Apple to come out of WWDC leaving a good impression?
如果他们只保守地兑现2024年就承诺过的功能，大家会觉得这不过是老生常谈。
If they play it safe and only deliver on features that were promised back in 2024, people will think it's just warmed over news.
但要是他们大肆宣传Apple Intelligence的新功能，大家又会担心能不能信得过。
On the other hand, if they show off new capabilities for Apple intelligence, people will be wary of trusting them.
那苹果到底该选哪条路？
Which of these approaches should Apple focus on?
有没有第三种可能，既能让人对他们的发布感到兴奋，又不会让人觉得‘库比蒂诺这地方出问题了’的糟心感？
Is there any third option where they can leave the public excited about what they presented without the bitter aftertaste of the whole something is rotten in the state of Cupertino debacle.
根本不存在第三条路。
So there was no third option.
很多人其实都挺失望的。
A lot of people were pretty underwhelmed.
你觉得怎么样？
What did you think?
你当时在库比蒂诺吧。
You were in Cupertino.
是的。
Yeah.
我认为苹果错失了这个机会。
I think Apple foreclosed the opportunity.
加布里埃尔提到的关键点是，他们让观众无法对这次发布会感到兴奋。
What Gabriel's putting his finger on is they foreclosed the opportunity for people to leave this keynote excited.
原因就跟他讲的一样。
For the exact reasons he said.
他们要是展示一些激动人心的功能，大家也不会信；要是只说些基本功能，又强调这次真的能用，大家也不买账。
They could show exciting things and no one would believe them, or they could show like we're actually like doing basic stuff, but it works this time.
对。
Yeah.
然后大家都觉得没什么大不了的。
And then everyone's underwhelmed.
所以这完全是苹果自己的问题。
So that's Apple's own fault.
2024年的代价就是，到2026年，没人会真正感到兴奋，也不会觉得苹果真的在改变
This is the cost of 2024 is that no one is gonna leave 2026 genuinely excited and feeling like Apple is changing the
世界。
world.
考虑到这一点，
Given
我觉得他们做得还不错。
that constraint, I thought they did a great job.
而他们做得好的关键，在于这场发布会其实挺让人难受的。
And the most important thing aspect of doing a great job was the extent to which this was a fairly painful keynote.
是的。
Yeah.
一方面，它短得让人震惊。
On one hand, it was shockingly short.
尽管感觉大概有23%的时间都在看加载动画，但它依然短得惊人。
And it was shockingly short despite the fact it felt like approximately 23% of the keynote was watching loading indicators.
就像这个演示一样。
Like this was the demo.
这是所有发布会中，最应该直播的一场。因为正如你所知，他们需要重建信任，而重建信任的唯一方式就是让产品真正稳定运行。
This was the keynote of all keynotes that could have benefited from being live where you have because because what they needed to, they needed to rebuild their credibility and you rebuild your credibility by stuff actually working.
现在他们，我想，大概只会……他们现在应该不会那样做了，
Now they, I guess, will just they're never
但在演示期间我一直在跟几位朋友发信息，说那些AI演示都挺无聊的。
going that do that at this point, but I was texting gentlemen during the presentation that, like, all the AI demos were pretty tedious.
不过，如果你是在现场观众面前做这些演示，然后看着它
However, if you were doing them in front of a live audience and then watching it
运行起来。比如，万一失败了怎么办？
work Like, in real what if it fails?
是的。
Yeah.
那样的话，反而会让人兴奋起来。
Then it becomes kinda thrilling.
他们本可以这么做。
And they could have done
那确实反映了蒂姆·库克时代的一个短板。
that, It's but a this actually is a testament to a weakness of the Tim Cook era.
对。
Mhmm.
敢站出来现场演示，说：我们赌一把，这次真成了。
Like to go out there and do it live and say like, we're go like, we're going to roll the dice.
我们觉得这次真的行了。
We think we actually got it.
敢冒着彻底丢脸的风险去演示，这本身就是在承认：他们早该为现在的状况感到羞愧了。
And we're going to, like, go out there and risk total embarrassment is itself an acknowledgment of the embarrassment they should already feel.
是的。
Yeah.
考虑到2024年发生的事。
Given what happened in 2024.
其次，这就是第三种方式。
And number two, that is the third way.
就是那种你走出来之后，还跟那帮人绑在一起的方式。
That's the way where you come out of it and you're like, hanging on with those guys.
是的。
Yeah.
他们确实走出去了。
They they went out there.
我用这个词是经过斟酌的，但那才是应有的回应。
I like I use that term carefully, but that would have been the response.
就是说，他们走出去了。
Like, they went out there.
他们扔掉了事先录好的主题演讲。
They threw away their prerecorded keynotes.
他们说，我们要加倍投入。
They said, we are doubling down.
我们把所有筹码都押上了，这次要现场演示。
We're putting our chips on the table, and we're gonna do live demos.
管他呢。
Screw it.
结果这一次，真成了。
And you're go it actually works this time.
在这种情况下，结合他们发布的内容，完全就是2024年说的那样。
And in that case, given what they shipped, it's totally what they said in 2024.
几乎没什么新东西。
There was almost nothing new.
你知道吗？
And you know what?
这样也没问题。
That's fine.
对。
Mhmm.
他们2024年展示的东西其实挺吸引人的。
Like, I I I what they showed in 2024 was pretty compelling.
是的。
Yeah.
这就是我们2024年那么期待的原因。
Like, that's why we were excited in 2024.
而且确实很关键的是，
And and it's compelling Well, crucially,
对绝大多数苹果用户来说，这种高端昂贵的AI功能其实并不是他们真正需要的，而这一点很有说服力。
it's compelling for 90% of Apple customers for whom, you know, high end expensive AI is not really what they want.
他们推出的是对高端用户有吸引力的
They're giving It's people compelling for high
高端AI用户其实并不想那样，你知道，你的数据里早就有这些东西，你只是希望把它找出来。
end AI customers who hasn't not wanted to, like, you knew something, it's in your data somewhere and you wanted to surface it.
他们能做到，因为他们掌控着操作系统。
And they can do it because they're the operating system.
正是因为他们对iPhone的掌控地位，才让他们能做别人做不了的AI事情，哪怕技术上并不算特别惊艳。
It's the it's of their position with the iPhone that they are uniquely positioned to do something with AI that no one else can do, even though it's not technically impressive.
但它的差异化非常明显，非常突出。
It's still differentiated, highly differentiated.
这就是为什么我对2024年那么看好。
And that's, that was why I was so enthusiastic about 2024.
因为他们确实有个独特的切入点。
Cause like, they have an angle here.
他们不需要大笔投入资本支出。
They don't need to do a bunch of CapEx spending.
凭借现有的优势，他们就能真正交付一些有意义的东西。
They can actually deliver something meaningful given their preexisting position.
他们正在利用自己作为关键数据拥有者的地位。
They're leveraging their position as the data that matters.
是的。
Yeah.
他们掌控着最重要的数据，同时又拥有足够影响力，能促使开发者使用他们的应用意图系统和索引系统来接入这些数据。尽管开发者可能并不喜欢苹果，也不认同它的各种做法，但这一战略布局非常聪明，我在2024年对此充满期待。
They own the data that matters and they have a position to corral developers to use these app intent system and their index system to give them their data, even though they, you know, even though like the developers don't like Apple and all these bits and pieces, but they were it was a very smart strategic play and I was very enthusiastic about it in 2024.
如果我跳到2026年，回头看所谓的神话或者Fable 5，
And if I step out of the it's 2026 and look what, you know, mythos or Fable 5.
他们真该感到羞愧。
They should be so ashamed.
对。
Yeah.
它依然令人印象深刻，极具吸引力。
It remains impressive and compelling.
而我认为，他们现在的问题在于，没有把这件事做到底。
And, and now I think where they fell short is they didn't go all the way.
我觉得要是来个现场演示，效果肯定完全不同。
I think a live demo would have transformed this.
肯定大受欢迎，毕竟几乎所有反馈都说：能用，而且用起来特别顺。
It would have been a smash hit, especially because almost universally the reports are, it works and it works super well.
我还没安装呢。
I have not installed.
我一般从不装测试版。
I usually never install betas.
不过这次我确实打算安装一下。
I actually do plan on installing this one.
这事儿吧，其实跟我去年还没收拾完摊子是一回事。
This is part and parcel of like, I was still unpacking last year.
我得找找我去年用的那部iPhone，我随身带的备用机，其实是两年前的旧款了。
I have to find like my, my last year's iPhone, my spare iPhone that I carry with me is actually two years ago iPhone.
这周真是忙得不可开交。
It's been a busy week.
所以我很抱歉。
So I apologize.
我这次没做好准备，但所有试过的人都说，这东西真能用。
I was not properly prepared for the show, but by all accounts, everyone who's talked about it says it just works.
真的能用。
And it just works.
苹果现在最该找回的，就是‘开箱即用’的体验。
That's that's what Apple needs to get back to is it just works.
对。
Yeah.
而且看起来好像
And it seems like
确实如此。
it does.
那就走着瞧吧。
Well, we shall see.
博扬在周二回应你的WWDC文章时留言说，苹果并没有真正创新。
Boyan responding to your WWDC article on Tuesday wrote in and said, Apple is not thinking different.
它还停留在过去，陷入了一种‘手里只有锤子，看什么都像钉子’的思维定式。
It is living in the past and becoming a victim of when all you have is a hammer, everything looks like a nail mentality.
这正是微软在移动领域犯过的错误。
That is exactly the mistake that Microsoft made with mobile.
他们硬把所有东西都塞进Windows的框架里。
They tried to shoehorn everything into a Windows viewpoint.
现在苹果正在重走同样的老路。
Apple is replaying the same playbook.
如果人工智能真像它最坚定的支持者所认为的那样彻底改变世界，苹果就麻烦了。
If AI changes the world like its biggest protagonist think it will, Apple will be in deep trouble.
不过幸运的是，AI很可能只是变成另一个非常实用的软件层，这样苹果就能躲过一劫。
Fortunately for them, it is very likely that AI becomes just another very useful software layer, and Apple dodges a bullet.
你更同意博扬的哪个观点？是第一个还是第二个？
What do you agree with one boy on or two boy ons?
一个观点。
One boy on.
因为Boy On的最后一句话，完全推翻了前面那五句话。
Because boy on's last sentence completely refutes like the first, like five sentences.
如果苹果的AI真的只是另一个有用的软件层，那就说明苹果其实是在另辟蹊径，没有砸几千亿进去，没错。
If in fact Apple AI is just another very useful software layer, then that means Apple is in fact thinking different by not spending hundreds of billions of dollars Yep.
而是走自己的路，坚持iPhone的模式。
And going their own way and sticking with the iPhone paradigm.
所以，到底哪一种才是对的，Boy？
So which is it, boy?
我觉得我根本不用回答。
Like, I don't feel like I don't even need to answer.
他自己已经把答案说出来了。
The guy answered himself.
我的意思是，过去几年里，苹果显然和整个科技行业想得不一样。
I mean, I would say they are demonstrably thinking differently than the rest of tech and have been over the last several years.
而且他们明显还是牢牢扎根在 iPhone 的模式里。
And they're also demonstrably anchoring themselves in the iPhone paradigm.
对。
Mhmm.
这并不意味着他们做错了。
That doesn't mean it's wrong.
其实我想表达的是，我这周写的那篇文章自己不太满意，有几个原因。
Like, that was a point I was trying to like, I had I didn't love my article this week for a few reasons.
其中一个就是这点。
One of which was this.
我得反复重写结尾，因为我想说，这些批评——比如波扬的批评——完全有道理。
I had to rewrite, like, the ending, like, multiple times because I was trying to say, look, all these critiques, like, Boyan's critique is totally legitimate.
苹果确实是在做对自己最有利的事。
Apple's totally doing what's best for them.
他们完全是在按自己的模式走，合情合理。
They're totally doing what makes sense for their model.
他们根本不会在自己不太擅长的事情上投入资源。
They're totally not investing in things that they're probably not good at.
所有这些原因都在于
And all these reasons
他们过去几年的思维方式并没有改变。
thinking any differently than they were over the last
过去十到十五年里一直都是这样。
ten or fifteen years.
没错。
That's right.
对。
Yeah.
所有这些情况都是，当你指出公司只是在做对自己有利的事时，后面通常都会跟个‘但是’。
All these things but when you articulate that companies are doing what is good for them and in their interest and like the it's usually followed by a but.
是的。
Mhmm.
现在世界已经不一样了，这种想法完全不对。
The world's different now, and this is totally wrong thinking.
有时候，这反而是正确的想法。
Sometimes it's just right thinking.
你只是运气好，恰好这个模式真的管用。
Like you just you're just fortunate enough like that actually the model works.
事实上，我已经多次提到过这一点。
And the fact of the matter is we I have talked about multiple times.
波扬该给我点认可，或者其中一个波扬，波扬一号。
Boyan should give me a little credit here, or one of the boyans, boyan one.
没错。
Yep.
没错，你可以想象，未来会有那种及时出现的界面，你需要的时候它就在，不需要的时候就消失，后台有智能代理替你做事，更能预判你的需求。
That there's, yes, you can imagine the You've been future warning about the paradigm, just in time UI, there when you need it, gone when you don't, agents working in the background on your behalf, much more anticipatory of your needs.
还有各种无形的传感器和输入设备。
All these ambient sort of sensors and input devices.
所以我才在Solar项目这个背景下提到它，那个演示我真的超喜欢，尽管它很可能永远都不会发布。
This is why I talked about it in the context of like the project Solar a bit, which again, I love that demo, even though it almost probably will never ship.
这东西估计要成空气了，但它清晰地展现了一种完全不同于iPhone的计算模式。
It's going be vaporware, but it was a tangible articulation of a completely different model of computing that is not like the iPhone at all.
恰恰是它的反面。
It's the exact opposite.
这就是Boyan one说的未来，而苹果到时候就完蛋了。
It's what Boyan, Boyan one says is coming and Apple is screwed.
但Boyan two又来了，说其实iPhone真的很厉害。
But then Boyan two is here and saying like, actually the iPhone's pretty phenomenal.
那在这样的背景下，消费者到底愿意为什么买单？
And in a context where what what do consumers pay for?
你给我举个例子，消费者真正愿意付费的服务是哪一个？
Give me one service that consumers pay for.
你这话是什么意思？
What do you mean?
比如软件服务吗？
Like a software service?
我不确定。
I don't know.
Netflix。
Netflix.
对。
Yep.
Netflix到底是什么？
What is Netflix?
娱乐服务。
Entertainment.
我本该在文章里提到Netflix，它其实更能说明我的观点。
I should have put Netflix in the article because that helped actually make my point more.
我只是说消费者并不为此付费。
I just said customers don't pay for it.
他们不是为效率付费的。
They don't pay for productivity.
他们喜欢的是娱乐，才愿意为此花钱。
They don't like, they pay for entertainment.
iPhone 是一个极其出色的娱乐设备。
What the iPhone is a phenomenal entertainment device.
它最主要的功能就是娱乐。
That is its number one functionality.
就像社交媒体、竖屏短视频那样。
It's like social media, it's vertical video.
就是和朋友聊天。
It's talking to your friends.
现在它是不是也成了一个提升效率的工具？
It's now, is it also a productivity device?
当然了。
Absolutely.
这是世界上最有用的生产力工具之一，但人们用它来工作。
One of the most useful productivity devices in the world, but it's used by people to do work.
我大部分工作都是在iPhone上完成的。
I do most of my work on the iPhone.
我真正的写作也是在上面做的。
I do my actual writing.
有人问我用什么软件工具，反复问了好几次。
Someone was asking me for like my software stack and like my software stack.
挺讽刺的是，我这台顶级的MacBook Pro，实际上我主要就用个文本编辑器，偶尔开个Excel表格。
It's funny how this super powered MacBook Pro, like when reality, I uses like a text editor and like some Excel spreadsheets maybe.
我大部分阅读、思考，还有构思想法之类的事，都是在手机上做的，我用iPhone就像个Z世代狂热粉丝。
And like, I do most of my reading, most of my like thinking, a lot of my like sketching out ideas or whatever on my phone, I use an iPhone like a gen Z maniac.
它确实是个生产力工具。
And it is definitely a productivity device.
但对大多数人来说，他们平时根本不是在工作。
But for most people, they're not doing work most of the time.
是的。
Yeah.
娱乐的重要性丝毫不亚于生产力，甚至更重要。
They entertainment is just as important, more important in fact than productivity.
从这个角度来看，这些设备有哪一样能比手机更出色呢？
And on that vector, on what perspective is any of these devices better than a phone?
咱们走着瞧吧。
I mean, we'll see.
可能要再过十年，才会出现能真正挑战手机主导地位的设备。
It might be ten years before any sort of device emerges that could actually challenge primacy of the Yeah.
所以我越来越认同博扬的观点了。
So I'm with Boyan I I I'm increasingly with Boyan two here.
我觉得，苹果或许正巧赶上了这么一个好时机。
Like, I'm coming around to the, you know, Apple, maybe Apple's lucky to be in this situation.
运气好总比本事强更幸运。
Better be lucky than good.
不对。
No.
没错。
Exactly.
我看看他们上周发布的东西，苹果其实根本不需要在AI领域打出全垒打，也能在AI领域胜出，并且进一步巩固自己的优势。
And and, like, I look at what they announced this past week, and Apple doesn't need to hit home runs in AI to win in AI and to double down on all those advantages.
而且，他们其实也不必非得在AI上赢。
Well, they also don't need to win in AI.
有时候，只要守住现有的地位就足够了。
They need to sometimes it's sufficient to maintain your position.
是的。
Yeah.
对吧？
Right?
他们根本不可能在AI上全面胜出，真的不会。
Like like, they're not going to win in like, they're not.
他们没必要
Like, they're not
拥有最顶尖的AI或最好的模型。
gonna have the best AI, the best model.
没错。
No.
正是如此。
Exactly.
但换个说法，苹果的做法是
But I guess to change the metaphor, Apple is
不过，他们可能是使用最广泛的AI。
They might have the most used AI, though.
就像昨晚第四节最后阶段的马刺队那样吗？
What about Apple as the Spurs in the second half of game four last night?
他们领先25分，关键是要别连续投丢九个三分球。
They're up 25 points, and the key for them here is to not miss nine threes in a row.
为什么
Why
你明明领先这么多，还连续投九个三分球？
are you taking nine threes in a row like you're behind?
是啊。
Yeah.
像这种时候，上篮和罚球就能搞定，何必呢。
Like, just getting layups in free throws would Yeah.
把事情做完就行。
Get the job
我觉得今年的WWDC就是这样，在我看来完全合理。
And I think that's what WWDC was this year, and it's perfectly sensible as far as I'm concerned.
乔纳森提到，关于苹果和谷歌的AI合作，我想听听更详细的说明。
Jonathan says, goats, I would love to hear some clarification on the Apple and Google AI deal.
我以为苹果是用谷歌的白标版Gemini，或者用谷歌云做设备外的计算。
I thought Apple was using a white label version of Gemini and or using Google Cloud for off device compute.
不过，克雷格·费德里吉的这段话听起来像是苹果根本没用谷歌的任何东西。
This quote from Craig Federighi, though, sounds like Apple is using nothing from Google.
这是费德里吉说的话。
Here's what Federighi said.
当然，我们并没有把Gemini应用作为自己的应用。
Of course, we don't have the Gemini app as our app.
事实上，所有那些客户端代码都不在我们iOS系统的运行方式里。
In fact, none of that client code is part of how we run on iOS.
对于这些模型，我们既没有使用谷歌提供给客户的任何模型，也没有使用他们部署模型所依赖的基础设施和方式。
For these models, we use none of the models that Google deploys to their customers nor do we use the infrastructure and means by which they deploy models to their customers.
至于知识库，我们当然也不会拿谷歌搜索或者类似的东西作为我们系统的根基。
And then when it comes to the knowledge base, we of course don't use Google search or anything like that as the foundation of our system.
所以，我希望这一点已经说清楚了。
So I hope that's clear.
我们用到的谷歌助手，一点都没有。
The amount of the Google Assistant that we use is none.
乔纳森说，我觉得真相可能在中间某处，但你能帮个普通人理清这种让人摸不着头脑的公司套话吗？
And Jonathan says, I suspect the truth lies somewhere in the middle, but please help a normie make sense of this bizarre corporate speak.
那你能帮帮乔纳森吗？
Can you help Jonathan then?
克雷格·费德里吉说的那句‘我希望这已经说清楚了’其实最讽刺，因为他根本不在乎你听没听懂。
So the funniest line that Craig Federighi said there is so I hope this is clear because he does not hope it's clear.
他这句话里没一个字是假的，可他故意让你产生了完全错误的印象。
Like, he did not lie anywhere in this statement, and he also created an impression that is totally wrong.
明白了。
Okay.
好。
Great.
我不确定。
I don't know.
我要不要一句一句给你拆开讲？
Should I break it down sentence by sentence?
对，没错。
I yeah.
一句一句来吧。
Go sentence by sentence.
来，咱们拆解一下。
Fire Joe Morgan here.
哦，当然。
Oh, yeah.
我不确定要花多久，不过好吧。
I'm not sure how long this is gonna take, but, okay.
当然，我们并没有把 Gemini 当成自己的应用。
Of course, we don't have the Gemini app as our app.
这还用说。
Duh.
Gemini 是一个需要你下载到手机上，或者访问网站才能使用的应用。
Of course, Gemini is an app that you actually download on your phone or go to a website to use.
展开剩余字幕（还有 136 条）
他们没有那个。
They don't have that.
是的。
Oh, yeah.
行吧。
Fine.
明白了。
That's clear.
事实上，那些低调的代码根本不在我们iOS端的运行逻辑里。
In fact, none of that quiet code is part of how we run on iOS.
这又在重复刚才说的内容。
This is saying the same thing again.
也就是说，Gemini应用的代码就是Gemini本身。
Like, the code of the Gemini app is the Gemini.
他们不仅没把Gemini应用拿过去，也没把它的代码拿去加进自己的应用里。
Gemini They did not take the code of the Gemini app in addition to not taking the Gemini app and put it in their app.
所以就是这样，没错。
So that is just that's Yes.
太好了，明白了。
Super Okay.
再说一遍，这个应用的代码和那个后台系统毫无关系。
So again, and the code of the app has nothing to do with the bottle.
那个后台系统就像是核心，而应用只是访问它的入口。
The bottle is like the, the, the app is a means to access the bottle.
好的。
Okay.
其实就是这么回事，无所谓了。
It's, it's just like, whatever.
我的网站只是用来访问我写的内容，对吧？
The, the, my website is a way to access my writing, right?
它就像是一个装东西的容器。
Like it's like a container for, for it.
好的。
Okay.
接下来呢？
What's next?
对于这些模型，我们完全没有使用谷歌提供给客户的任何模型。
For these models, we use none of the models that Google deploys to their customers.
他们用的并不是谷歌给客户那套一模一样的模型。
They don't use the exact same model that Google gives their customers.
这一点没错。
Now that is, that is correct.
我们也不使用他们的基础设施。
Nor do we use the infrastructure.
好的。
Okay.
但它们到底在做什么呢？
So, but what are they actually doing?
苹果有五个模型，其中几个运行在设备上，还有一个在云端。
So Apple has, I think it's five models, several of them on device and then one that's in the cloud.
所有运行在设备上的模型都是苹果自主研发的。
So all the ones that are in device are Apple's own models.
顺便说一下，这些模型确实都是苹果自己开发的。
And by the way, I was, are Apple's own models.
这些模型当然都很小，因为它们运行在设备上。
Those models are of course, because they're on device, very small.
不过，其中有一个还挺特别的。
Now, one of them is pretty interesting.
它的规模其实相当大。
It's actually quite large.
那个200亿参数的模型，每次只加载其中一部分到内存里。
The 20,000,000,000 parameter model that were only part of it is loaded memory at the time.
这叫专家混合架构。
This is called a mixture of experts.
通常来说，专家混合模型会在每个词元时加载不同的专家。
Usually for a of experts, load a different ex or use a different expert every token.
也就是说，你得把整个模型都保留在内存里。
So it's always like, which means you have to have the whole thing in memory.
但这个情况不一样，它是按查询来选择专家。
In this case, you do a expert per query.
所以当你发起一个查询时，系统会说：我们需要模型的这一部分。
So they so when you make a query, it's like we need this part of the model.
它只把这一部分加载到内存，然后在整个词元序列中都保持加载状态。
It loads only that part of the model into memory, and then it's in memory for all the tokens.
不管怎样，这些全部运行在设备上的模型都是苹果自家的。
Anyhow, regardless, they have all those on device models are Apple models.
对。
Mhmm.
那你怎么才能做出这么小的模型呢？
But how do you get small models?
把它精简一下。
Distill it down.
进行精简。
Distilling.
没错。
Exactly.
其实他们不会说‘精简’这个词。
This is there there is a like, they won't say distilling.
我觉得他们提到过，是谷歌联合开发的，其实就是让大模型回答问题，然后用这些回答来训练你的模型。
I think they said that Google derived or in conjunction with Google, basically, what you do is you ask the big model questions and then you get answers, and that's how you build your model.
这些都是谷歌精简过的模型。
These are Google distilled models.
这种精简做法是公开透明的。
Above board distilling that's happening.
顺便说一句，这么做不仅合法，还有它的道理。
And by the way, if you have the reason it's not just legal.
能直接接触到模型本身。
Having actual access to the model.
所以，生成令牌这个概念背后，其实有多层机制。
So the whole idea of generating tokens, there's actually multiple layers underneath.
如果你只能看到最终输出，虽然也能从中提炼，但效率低得多，而且对母模型的理解和覆盖范围也有限。
If you're only getting access to the final output, it's, you can distill from that, but it's way more inefficient and you're not getting, as broad of a coverage and understanding of this sort of parent model.
如果你能直接接触模型，提炼的效果会好得多。
If you have actual access to the model, your distilling is far more effective.
顺便说一句，当你在线使用大多数模型，包括ChatGPT和Anthropic时，你实际上用的就是一个经过提炼的模型。
And by the way, when you're accessing most of these models online, including ChatGPT Anthropic, you're accessing a distilled model as it is.
当然。
Sure.
没人能接触到真正的原始模型，因为那成本高得离谱，根本没法规模化。
No one's getting access to the real model because it'd be astronomically expensive and impossible to scale.
所以，就连网上那些大型模型，也都已经被不同程度地提炼过了。
So even like the big models that are online are distilled to an extent.
而且他们能做得很好，因为这是他们自己的模型。
And then, and they, they can do a good job of it because it's their model.
所以他们比任何人都更擅长做这种蒸馏。
So they can distill it better than anyone else can.
但苹果设备上的那些模型，并不是谷歌提供的模型。
But all those models on Apple's devices are not the models that Google serves.
它们是苹果提供的模型，因为谷歌和苹果都是从 Gemini 蒸馏出各自的版本来部署的。
They are the models Apple serves because both Google and Apple distill from Gemini for models to serve.
所以他们在谷歌的帮助下做了自己的蒸馏，但这些模型和谷歌自己用的并不一样。
So they're doing their own distilling in with Google's help, but it's it's it's not Google's it's not the same ones that Google has.
对。
Right.
所以云端的那个模型，据我理解，底层应该是某种 Gemini 模型。
It's so the one in the cloud then, the one in the cloud, as I understand it, is some sort of Gemini model at its base.
但现在的模型都是这样，初始模型训练完之后，还要加很多后续的微调和调整，所以虽然本质上不是完全不同的模型，但实际能访问到的版本却有很多种。
But then the way models are today, there's so much post training and things you sort of layer on after the initial model is done that it's effectively, it's not a different model, but it's, there's multiple variations of models that, that can be accessed.
你看看这些公司，动不动就发布新模型、新模型，好像一直在推新东西。
And that, that when you see all these companies, oh, new model, new model, they're releasing models all the time.
确实有些是真正的新模型，比如 Mythos，那是重新训练的，但很多所谓的‘新模型’其实并不是重新训练的。
Now there's genuinely new models like say a Mythos that is like a new, a new, training run, but lots of the new model releases are not new training runs.
它们只是做了后续的微调和强化学习，持续优化而已。你可能一年或两年才做一次预训练，之后就不断在同一个基础上迭代。
They're new post training and new sort of, reinforcement learning that sort of makes they're continually being improved by this post pre like you do one pre training a year, a couple of year, and then you just, iterate on that one that's there.
苹果就是对 Gemini 做了这样的处理。
So Apple has done that to Gemini.
现在云上跑的就是这个版本。
That's what's being served in the cloud.
这跟谷歌原版的不一样，但本来就应该这样。
That is different than Google's and appropriately so.
他们是在根据自己的使用场景调整模型。
They're tuning the model for their use case.
所有这些都跟 Gemini 密不可分。
So all of this is intimately integrated with Gemini.
克雷格·费德里吉特意让你觉得
And Craig Federighi went out of his way to make you think
那根本不是在混淆概念。
that wasn't Confuscate the this.
没错。
Sure.
但他没说谎。
But he didn't lie.
对吧？
Right?
所以，我希望这点说清楚了。
So, so I hope that's clear.
但其实没说清楚。
Well, it's not clear.
你就是故意这么做的。
And you did that on purpose.
那么，我们现在是不是该用他们向客户部署模型的那套基础设施和方式呢？
So now do we use the infrastructure and means by which they deploy their model models to their customers?
不是的，他们有自己的配置，这其实特别有意思，用的是英特尔的头节点和英伟达的GPU。
No, they have their own setup with, and this is actually one of the more interesting parts with, Intel head nodes and NVIDIA GPUs.
明白了。
Okay.
这跟标准的英伟达机架不一样，那种机架是英伟达自己配的头节点，用的是Grace CPU和Blackwell GPU，所以才叫Grace Blackwell。
This is different than a standard like NVIDIA rack where NVIDIA has their own head node with their, with their, Grace, Grace GPU and then like a Blackwell, like that's why it's called Grace Blackwell.
Blackwell其实是GPU，Grace才是CPU。
Blackwell is actually the GPU, Grace is the CPU.
然后还有Vera Rubin。
And then you have Vera Rubin.
Vera是CPU。
Vera is the CPU.
Rubin才是GPU。
Rubin is the GPU.
他们用的是 NVIDIA 的 GPU，前端则是 Intel 的部分。
They're using the NVIDIA GPU part with Intel as the front part.
他们之所以这样搭配，是因为 Intel 和 NVIDIA 都有安全围栏技术，能确保整个计算链路可验证，从而保障私有云计算的安全性。
And the reason they have those is both Intel and NVIDIA have secure enclave technology that lets you like ensure you get the private cloud compute aspects because it's full, the whole chain is fully verifiable.
那过去的是什么内容？
Like what got past there?
数据有没有被清空？
Did it get flushed?
它到底传到哪儿去了？
Where it went?
所以从你的设备，到 Intel 的头节点，再到 NVIDIA 的 GPU，整个链条都通过加密握手实现了完全保护，这样他们才能继续兑现私有云计算的承诺——尽管这和两年前他们推销的私有云完全不一样，那时候他们说的是‘我们的服务器、我们的芯片，跑在云上’。
And so it's a whole cryptographic sort of handshake between your device, the Intel head node, the NVIDIA GPU that it's fully secured, that whole chain, so they could continue to make their promise around private cloud compute, even though it's totally different than the private cloud compute they tried to sell two years ago, which is our servers on our chips, like, you know, running in the cloud.
所以，
So,
但这里的基础设施不是谷歌的，而是他们自己的。
so But it they've is couple infrastructure here, not Google's infrastructure.
不，这是谷歌的基础设施。
No, it's Google infrastructure.
它运行在谷歌的数据中心里。
It is running in Google data centers.
所以它是在谷歌的数据中心，用英特尔和英伟达的芯片运行的。
So it's running in Google data centers on Intel and Nvidia chips.
这跟以前对私有云计算的定义完全不一样了。
And so it is, this is a total switch from what private cloud compute was.
但因为这些英特尔和英伟达芯片里嵌入了这项技术，能从密码学层面保证完全私密，所以它依然是可信的私有云计算。
It's still credibly private cloud compute because of this technology that's embedded in these Intel and Nvidia chips that lets you cryptographically guarantee that it's totally private.
对。
Yeah.
但基础设施彻底变了。
But it's a total change in infrastructure.
而这并不是谷歌自己用的方案。
And that's not what Google uses.
谷歌用的是TPU集群。
Google uses fleets of TPUs.
它们没有那种加密保障，没法确保完全私密。
They don't have the cryptographic guarantees as far as it being private and all those sorts of things.
所以我们的知识库，根本不用谷歌搜索。
So knowledge base, we don't use Google search.
是的，他们好像有自己的搜索索引，这还挺有意思的。
Yes, they had, apparently they had their own search index, which is sort of interesting.
我希望这一点已经说清楚了。
And then we get to, so I hope that's clear.
至于谷歌助手，我们完全不用。
And then the amount of the Google Assistant we use is none.
谷歌助手只是一个应用程序。
Google Assistant is an app.
对，这个我们明白。
Like, yeah, we get it.
我们不用Google助手，它就跟以前的那个应用一样。
We don't you don't use the Google Assistant is like the old app.
所以我们既不用新应用，也不用旧应用。
So it's like, we don't use the new app and we don't use the old app.
谢谢，克雷格。
Thanks, Craig.
我们已经弄清楚了。
We we figured that out.
这完全是胡说。
So nonsense.
胡说八道。
Nonsense.
第三句话的前半部分也很有误导性，因为他们并不使用向客户部署模型所依赖的基础设施。
First clause of the third sentence, is also pretty misleading, but they don't use the infrastructure and means by which they deploy models to their customers.
这似乎是目前最准确的说法。
That seems to be the most accurate claim there.
而且这里所有内容都是对的。
And then Everything in here is correct.
至少这个说法是最不容易误导人的。
It's Well, I the least misleading claim.
所以不管怎样，乔纳森，希望这能帮你理清思路。
So in any event, Jonathan, I hope that clears things up for you.
保罗提到，如果Siri AI只是白标Gemini的一个消费者端接口，虽然底层模型本身不是苹果的，但考虑到本今天提到的接口整合价值，这种情况下，苹果凭借其渠道优势、操作系统与设备的深度整合以及庞大的用户基础，是否足以让它在一段时间内获得一定的喘息空间？
Paul says if Siri AI is some sort of consumer harness for white labeled Gemini, although the underlying model itself isn't Apple, given Ben's note today about the integration value of harnesses, is that circumstance enough to give Apple AI breathing room for a while to come given their distribution advantages, OS device integration mode, installed base, etcetera?
假设Siri AI完全是苹果自主研发的，如果苹果能为Gemini打造一个不错的接口层，这种能力本身难道不足够让它在主流模型之间保持一定的竞争张力吗？
And assuming Siri AI was developed all in house, if Apple was able to create a good harness light for Gemini, isn't that skill valuable enough to let them keep some competitive tension among the leading models?
最后，如果苹果能让一个基于Gemini优化的小型模型在设备端运行，这会不会进一步帮助它维持各大模型厂商之间的竞争，从而全面有利于苹果？
And finally, if Apple was able to make a smaller Gemini derived model run on device, does that further help maintain competitive tension among model makers all to Apple's advantage?
你对此有什么看法吗？
Any thoughts there?
是的。
Yeah.
保罗，不好意思啊。
So I'm sorry, Paul.
我得稍微挑你一下，但我觉得这个问题其实代表了很多对当前AI发展水平一无所知的人。
I'm gonna have to pick on you a little bit, but I think this question is, I think, stands in for a lot of people who don't have an awareness of how advanced AI is right now.
好的。
Alright.
免费试听部分到此结束。
And that is the end of the free preview.
如果你想继续收听本恩和我的内容，节目说明里有订阅链接，你也可以直接访问 sharptech.fm。
If you'd like to hear more from Ben and I, there are links to subscribe in the show notes, or you can also go to sharptech.fm.
无论选择哪种方式，你都能获得一个专属播单，每周我们所有节目都会包含在内，还有来自 Stratechery 和 Stratechery Plus 套餐的更多优质内容。
Either option will get you access to a personalized feed that has all the shows we do every week, plus lots more great content from Stratechery and the Stratechery Plus bundle.
去试试吧。
Check it out.
如果你有任何反馈，欢迎发邮件到 email at sharp tech dot FM。
And if you've got feedback, please email us at email at sharp tech dot FM.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。