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。
Meta的首席技术官安德鲁·博斯沃思今天做客我们节目，聊聊公司的人工智能布局，以及为什么决定自主研发新一代AI眼镜。
Meta chief technology officer Andrew Bosworth joins us to talk about the company's AI efforts and why it's building its own new AI glasses.
接下来马上为您揭晓。
That's coming up right after this.
面对持续的变革与机遇，TMT行业的领导者们需要交付实实在在的成果，而不仅仅是想法。
In the face of ongoing disruption and opportunity, TMT leaders need to deliver tangible results, not just ideas.
在速度和表现至关重要的时候，普华永道将市场洞察、行业专长与人工智能、云计算和新兴技术相结合，加速您的转型，从战略到执行，实现可衡量的投资回报。
When pace and performance matter most, PwC combines market insights and deep sector experience with AI, cloud, and emerging tech to accelerate your transformation and drive measurable ROI from strategy to execution.
普华永道能帮您预见未来，领先变革，赢得竞争。
PwC can help you anticipate what's next, outpace disruption, and compete.
如需了解更多，请访问pwc.com。
For more information, visit pwc.com.
欢迎收听《大科技》播客，一档专注于理性、深入探讨科技世界及其影响的节目。
Welcome to Big Technology Podcast, a show for cool headed and nuanced conversation of the tech world and beyond.
今天这期节目内容非常精彩，千万别错过。
We have a great show for you today.
今天我们邀请到了Meta首席技术官安德鲁·博斯沃思，他将和我们聊聊公司的人工智能进展、全新的AI眼镜、企业文化，以及一些深刻的思考。
We're joined today by Meta chief technology officer, Andrew Bosworth, who's gonna talk to us all about the company's AI efforts, its new AI glasses, the company's culture, and some big thoughts at the end.
博兹，很高兴见到你。
Boz, great to see you.
欢迎回到节目。
Welcome back to the show.
谢谢你们邀请我。
Thanks for having me.
我们刚才录之前还在聊，现在真是个特别疯狂的时刻。
We were just talking before we started rolling about what a crazy moment it is Yeah.
在科技圈里，确实如此。
In the tech world.
据我所知，这种级别的进展以前从没出现过。
We haven't seen progress like this as far as I can remember.
核心就在于人工智能模型本身。
The core part of it is the AI model.
AI模型是所有一切的基础。
The AI model underpins everything.
如果没有一个能正常运行的、领先的AI模型，很多事情都很难推进。
Without an a a working AI model or a leading AI model, it's tough to build.
长期以来，大家认为要打造一个出色的AI模型，必须有海量的算力和顶尖的研究人员来优化算法。
The theory for a long time was that to build a great AI model, you needed a ton of compute and great researchers to work on the algorithm.
Meta拥有充足的算力和最优秀的一批研究人员，但迄今为止，领先的AI模型还没真正出现。
Meta has a ton of compute and a team of the best researchers to work on the algorithm, but the leading AI model hasn't materialized yet.
你能谈谈你们在这方面的发现吗？是不是当初那个关于打造优秀AI模型所需条件的核心假设出了问题？
So can you talk a little bit about what you've learned there and whether that core assumption about what it takes to make great AI models is wrong?
我觉得还有一点很重要，那就是高质量的数据。
Well, I the only other ingredient I would add is great data.
你们确实有这方面的优势。
And you have that.
我们也有。
And we do
我们也有。
have that, I think, as well.
所以，确实这里有两个故事。
So, yeah, there's two two two stories here.
第一个是，回顾一下Llama一、二、三版，我们当时确实走在前沿，不断推动进展。
The first one is, you know, I think, you know, go back to Llama one, Llama two, Llama three, we really were, you know, kind of at the forefront and advancing things.
当然，你对此也很清楚。
And and you, of course, know this.
Facebook的基础AI研究团队，历史已经超过十年了。
We the Facebook fundamental AI research group goes back a decade more.
我最早接触到AI动态，就是那时候AI聊天机器人突然出现在我的界面里，我认识了扬，之后开始接触FAIR团队的人。
I've been I mean, that's where I actually first got queued into what was going on with AI is when m the AI messaging bot popped And up in my then I met Jan and started to meet the fair people.
是的。
Yeah.
我当时就觉得，这技术进步得太快了。
And was like, oh, this technology is progressing really fast.
所以Meta很早就切入了这个领域。
So Meta was on it very early.
真正拉开差距的地方——这一点后来也公开了——是我们当时没太重视的，就是在整合Llama三的时候。
And and so the real gap, which I think has been pretty public was what we didn't really raise at the time was when we were pulling Llama four together.
不好意思。
I'm sorry.
我们在整合Llama三的时候，把所有研究资源都集中起来了，把能用的招都用上了，结果无意中把整个研发管线给断了。
When we were pulling Llama three together, we had really pulled in all the research, all the every we pulled out every single stop we had and unwittingly kind of killed the pipeline.
研究人员的工作方式通常是：先搭一个基础框架，然后有人在这个基础上做渐进式改进，也有人在探索全新的方向。
So researchers, you know, the way that it works is you build a base and you've got people pioneering an incremental version of the base, and you've got people out there pathfinding entirely new strategies.
当时我们没太意识到，这也说明我们对这件事的关注还不够——Llama三确实是个很棒的模型，也广受好评，但为了做出它，团队把所有未来可能的突破都押在了这一个版本上。
And kind of unbeknownst to us at the time and kind of speaks to the fact that we weren't focused enough on it, Llama three, which was a great model and was well received, to get to that model, they had kind of pulled forward all the future bets into that digital bet model.
这就导致到了Llama四该做的时候，我们这边已经没有像其他实验室那样在推进的前沿探索了。
Well, that meant when it came time for Llama four, we didn't have any of the pathfinding the other labs still had going.
所以你现在在推理能力上就落后了。
So that makes you now you're behind on reasoning.
现在你在混合专家技术上也落下了。
Now you're behind on a mixture of experts.
现在你在一系列推动技术进步的关键技术上都落后了。
Now you're behind on a bunch of these critical technologies that have been used to continue continue the pace of progress.
这其实是个挺公开的遗憾，我想一年前对我们来说就是这样，也促使马克改变了想法——以前我们觉得AI只是众多方向中的一个。
This is a pretty public, you know, disappointment, I think, a year ago for us and led to Mark shifting from, okay, AI isn't one of our bets, which is how we thought of it up to that point.
以前AI只是我们众多尝试中的一个方向。
AI was just one of the many bets we had.
现在AI成了公司整个战略的基础，所以我们必须彻底改变对它的看法。
AI is a bet that's foundational to the entire company, and so we're to change how we're thinking about this.
这说法听起来有点老生常谈，但我实在想不出更贴切的词了。
And this is such a cliche, but I don't have a better word for it.
幽灵创始人模式。
Ghost founder mode.
他真的切换到了一种只有马克才有的状态，全身心扑在为我们争取算力、招募人才、签下研究人员上，你都能感觉到，这些成果大概是一年前才真正落地的。
Like, he really did flip into a mode that is, like, unique and reserved for Mark that where he just became so focused on getting us all the compute we needed, getting us all the talent that we needed, the researchers that we've signed that you said, you know and they really landed about a year ago.
我觉得亚历山大·王刚满一周年了，和他共事真的特别棒。
I think I think Alexander Wang just hit his one year metaversary, and I have loved working with him.
我已经从他身上学到了很多，现在也开始看到成果了。
I've learned so much from him already, And we are seeing the fruit of that.
如果你看一下Muse Spark的话
So if you look at Muse Spark
这是我们最新推出的模型。
Which is our latest model.
它不是我们最前沿的模型，但却是我们最近发布给《Vogue》用的版本。
Which are not our frontier model, but it's the latest model that we've released to the Vogue.
这个模型反响非常好。
It's a very very well received model.
根据不同的评测标准，它在我们最看重、最能体现产品独特性的那些方面表现得特别出色。
And depending on the benchmark, it does really well on things that we care the most about, that we think are unique to our products.
所以，没错，你在公众对我们的模型认知上，看法完全准确。
And so, yeah, you're absolutely right on where we are in terms of what the public, you know, perception of it is model wise.
我们组建了一支我真正信赖的团队。
We've built the team I really believe in.
计算资源和数据我们都已齐备，所以我很有信心，我们一定能达到目标。
We've got all the compute and the data that we need, so I'm very confident that we're gonna be where we need to be.
我再补充一点，这一点在战略上非常关键：现在市面上的模型其实很多。
I'll add a second piece to this, which I think is strategically very important, though, which is that, you know, models are available.
你可以直接租用一个模型。
Like, you can go rent a model.
你可以用Anthropix。
You can go use Anthropix.
你可以用OpenAI的。
You can use OpenAI's.
你也可以用谷歌的。
You can use Google's.
这些模型都很出色。
They're great models.
你可以直接去获取它们。
You can go get them.
你可以用它们，这已经很不错了。
You can use them, and that's pretty great.
我们真正要在世界上创造的价值，是产品本身。
The real value we're gonna create in the world is is the product.
至于我们对个人超级智能的愿景，我认为，我们是最有资格把它实现的。
And the products that we the vision that we have for personal superintelligence, I think, is a vision that we're uniquely suited to deliver.
这可不只是因为我们有数据。
It's not just that we have data.
这确实很厉害。
That's cool.
我们比几乎任何人都更有可能真正理解你、你想要什么、你是谁，以及什么对你最重要。
We actually have a better chance of understanding you and what you're trying to do and who you are in the world and what matters to you than I think almost anybody else does.
拥有模型只是其中一步，你得战略性地掌握它，避免依赖别人。
So having the model is one piece, and you wanna have that strategically so you don't have a dependency on somebody else.
但你更希望用它来掌握自己的命运。
But you'd mostly wanna be able to control your destiny with that.
模型本身并不是价值所在，我觉得很快消费者就不会在意了。
The model itself isn't the value, and I think we're gonna get to a world very soon where consumers, they don't care.
他们不想关心自己用的是哪个模型。
They don't wanna specify the model they're using.
他们不在乎是4.7还是4.8版本。
They don't wanna they don't care if it's 4.7 or 4.8.
就像你不会在意我用的是Oracle还是SQL数据库一样。
Like, you don't care what Oracle if I'm Oracle or SQL databases.
你只关心它能不能正常工作。
You just want the functionality.
关键是它得靠谱好用。
You are the thing to work well.
这才是我们所有人都会被衡量的标准。
That's the standard to which I think we're all gonna be held.
所以今天讨论的都是模型，这让我觉得，我们对用户这边的关注还不够，没怎么讲清楚普通人能从中得到什么好处。
So today, the discussion is about models, which suggests to me at least that we're a little under indexed on the user side of it and how humans are gonna benefit.
我觉得，这才是我们真正需要讲清楚的故事。
So I think that's the story that we needed to tell.
除了展示我们技术上的成果，还得实实在在地让消费者看到这些价值。
In addition to showing the work that we've done technically, we need to actually demonstrate the value to consumers.
我想先从科学角度聊聊这一点。
So I just wanna talk about this scientifically for a moment.
我一开始提到的是，你靠堆算力硬拼出一个有竞争力的模型。
The thing that I brought up in the beginning was this idea that you kind of brute force your way to a competitive model.
但我听你的意思，现在情况不一样了，因为有了专家混合、推理这些新方法，光靠基础预训练还不够，还得做精细调优，才能做出像今天这些顶尖模型，Meta现在就在干这个。
I think the answer that I'm hearing from you is not anymore because there are new techniques like mixture of experts and reasoning that you actually can you need some level of refinement of that base pre train in order to be able to build the models that were the top tier models that we're seeing today, and that's what Meta is working through right now.
对。
Yeah.
还不只是这样。
It's not just that.
说到底，这关乎整个行业。
It's by way, this is the whole industry.
从Llama 3发布开始，单一大模型的时代就差不多结束了。
The era of the monolithic model kind of died around Llama three launch.
以前人们觉得，只要有一个模型，测测它有多聪明，就能代表它在各种任务上的表现。
Like, the the idea, like, there's one model and just, let's just test how smart this model is, and that was how good it's going to be at lots of things.
现在不一样了，当你用这些工具——比如OpenCode、Cloud Code或者Codex——它们背后会根据任务不同，自动调用多个模型。
We're now in a world where, when you're using these harnesses, whether it's, you know, OpenCode, Cloud Code, Codex, using these harnesses, they're shopping underneath to lots of different models depending on the task.
比如，它可能会调用一个多模态模型。
So they might be going to a multimodal model.
如果你用Gemini，它在做图像生成时，会把任务转给Nano Banana这样的专用模型。
You know, if you're using Gemini, it'll it'll farm tasks out to Nano Banana if it's trying to do image generation.
我们已经彻底告别了‘一个模型统治一切’的时代。
So we've we've really moved past this world where this is just one model that rules everything.
真正该做的，是拥有一个强大但运行成本极高的智能模型，然后通过各种方式把它的能力压缩、分发，只在真正需要时才启用它，毕竟运行这种模型太贵了。
What you really wanna have is a very expensive to run intelligent model that you can distill down in all these interesting ways and places, and use it for its exquisite intelligence only when necessary because it's very expensive to run those models.
否则，在那些不需要顶尖智商的地方，就用更便宜、更快、延迟更低的模型。
Otherwise, have models that are cheaper and faster and have lower latency in all of these other places where it turns out you don't need to have a genius level intellect.
因为如果你想想人类的任务，我真的很相信规模定律。
Because if you think about human tasks, I really believe in scaling laws.
随着算力提升，模型的原始智能也会持续增长，这一点是必然的。
So you're gonna see this continued growth up into the right of as compute scales up, that the the raw intelligence of the model scales up.
但人类的任务并不要求无限的智能。
But, like, human tasks don't have infinite intelligence demands.
很多人类任务其实用普通水平的智能就足够应付了。
Like, there's a lot of human tasks that, like, you can do with conventional levels of intelligence.
所以我认为，未来一定会出现分层，不再只是简单地问：
And so I do think there's gonna be a stratification then where it's not just, okay.
不错。
Cool.
到底哪个模型能一统天下？
What's the one model that rules them all?
这挺不错的。
It's cool.
真正重要的是，如何组合一系列模型，让它们在性能、成本和价值之间取得最佳平衡来解决这些问题。
What is the collection of models that are brought together in such a way that they solve these problems with the right balance of performance and price and value.
是的。
Yeah.
你刚才提到了几点
You said a couple
很有意思的看法。
of interesting things.
首先，真正关键的是产品本身。
First of all, it's the product that matters.
我同意你的观点。
I would agree with you.
而且拥有自己的模型，对实现自主性确实很重要。
And that it's important to have your model, your own model for self reliance.
那我们来聊聊这个。
So let's talk about that.
你肯定注意到了苹果的做法，他们和谷歌达成合作，用了 Gemini 的精简版或者它的某个分支。
I'm sure you saw what Apple did where they made a deal with Google to distill Gemini or do some fork of Gemini.
从早期的反馈来看，Siri 用上这项技术后表现不错。
And it looks like from their early reports, Siri is working pretty well with that technology.
那你有没有考虑过，也和谷歌做个类似的合作，同时自己也在同步研发，既保证短期能快速提升产品，又能实现自主可控？
So have you considered doing a similar deal with Google and then building your own in parallel for that self reliance, but at least being able in the near term to advance your products as fast as you can?
这其实有两个方面。
Well, there's two parts.
我们目前其实已经在用很多不同的模型了。
So we we use lots of different models today.
我觉得，归根结底还是要给用户最好的、最适合他们的模型。
And I think, again, you wanna provide consumers the best model that's gonna work for them.
这里面当然涉及价格和性能的平衡，延迟也是关键因素。
So there's obviously there's a there's a price and a performance that makes that matters here, and there's a latency that matters here.
但拥有自己的模型，不仅能让你掌握主动权，还能在谈合作时争取到更有利的条件，确保用户能得到最优质的回答。
But, like, having your own model gives you the ability to not just control your destiny, you also have much stronger negotiating terms when you're trying to figure out the types of deals that you wanna make to make sure that you're getting the consumers the best available answer.
但他们没花多少钱。
But they didn't spend that much money.
对谷歌来说，也就十亿美元左右。
It was like a billion dollars to to Google.
现在还不好说。
And it's it's too early to tell.
我还不知道实际体验会怎么样。
I don't know what the experience is going to be yet.
我还没用上，得等看看。
I don't have access to it, so we'll find out.
对我们来说，谈的是个人级的超级智能。
I also I for for us at least, we're talking about personal superintelligence.
我们真正看重的是能为产品注入强大的专项能力，不只是通用智能，而是能切实解决用户问题的特定能力。
The ability we wanna be able to have to bring a tremendous specific capability to bear, not just a general intelligence, but a specific capability to bear for the products that we build, that really matters to us a lot.
我们并不认为这只是对现有系统的额外补充。
We're not seeing this as, like, a value add for an existing system.
我们认为，这将是人们与电脑互动的全新方式。
We're seeing this as an entirely new way that people are going to interact with their computers.
这其实和我们在现实实验室长期做的很多工作一脉相承。
It does go back to a lot of the work we've done in Reality Labs for a long time.
我们一直以施乐帕洛阿尔托研究中心、斯坦福研究院或者贝尔实验室这样的先驱为榜样，思考如何把大脑里的信息传给机器——这就是为什么我们投入神经接口，投入这些方向。
You know, we've always tried to model ourselves after pioneers like Xerox PARC or, Stanford Research Institute or Bell Labs where we're we're trying to think about what is the way that we get information from our brains into the machine, and that's hence, our work on neural interfaces, hence, our work on all these things.
反过来，怎么把机器里的信息传回我们的大脑？这也是我们做增强现实和虚拟现实的原因。
And what's the way to get the information from the machine back into our brains, hence, our work on augmented reality and virtual reality?
人工智能可能是我们见过的、把大脑信息传给机器最强大的工具，尤其是当它能感知我们周围大量信息的时候。
AI is potentially the best tool we've ever seen to get information from our brains into the machine, especially if it's able to observe a lot of things around us.
这些独特的能力，我觉得我们正试图实现，目前还没有人能做到。
Those are unique capabilities that I think we're trying to bring to bear that don't have any.
这不仅仅是模型的问题。
It's not just the model.
关键在于模型如何处理这些全新的输入，并从中构建出一个闭环系统。
It's like what's the model's ability to work with all these novel inputs and create a closed loop system out of it.
对。
Mhmm.
我觉得我们正在打造极其强大的模型，我对团队的能力非常有信心。
So I think that we are working on having incredible models, and I'm very confident in the team that we've assembled to do that.
我的意思是，光有模型还不够。
My point is just that it's not enough.
至于苹果公司是否只需租用一个模型就够了，我不确定他们有没有更宏大的愿景，去思考如何让技术真正融入人们的生活。
And whether it's enough for Apple to just go rent that model, I don't know if they have a broader vision for how it integrates with people's lives.
明白。
Okay.
那你是不打算租用模型了？
So you wouldn't rent the model?
不打算。
No.
我们确实会租用模型。
We do rent models.
正如我所说，我们用的模型，你知道的，我们是从哪儿开始的？
Like I said, we we use we we you know, we're From where?
我们内部做开发的时候，其实大量工作都是在我们自己的模型上进行的，根本没有必要去用别人的。
There's no reason for us we we we, you know, we when we're doing development internally, we do have a lot of, development happening on our own models.
当然，也有一些开发工作是基于谷歌、Anthropic或者OpenAI的模型做的。
There's also some areas of development that we do on models that we use from, from Google or from Anthropic or from OpenAI.
对。
Mhmm.
要真正做到不依赖特定模型，而且在经济上划算，关键是你得有自己的竞争性模型。
The ability to be model agnostic and have that be economically sensible actually kind of hinges on you having a competitive model Right.
万一需要，你随时可以退回去用它，这样别人就没办法随意抬高租用价格。
That you can go back to if you need to, and it creates a real backstop on, like, how much rent somebody can try to charge you on top of that.
但同样重要的是，不管是公司内部的开发者，还是普通用户，我都希望他们不必操心模型会怎么变。
But it's also worth noting whether it's I'm talking about a developer inside of the company or I'm talking about a consumer, I don't want them to worry about the model over time.
现在他们必须这么做。
Today, they have to.
现在这一切都紧密地连在一起了。
Today, it's like it's all very tight tied together.
但长远来看，他们真正该专注的，只是想达成的那个目标。
But over time, they just have a goal they're trying to accomplish, and that's the major focus that they should have.
所以，拥有一个绝对领先、前沿的模型，这种战略定位非常重要。
So there's this, like, strategic construct of having a model and having it be an absolute leading state of the art model, and that's super important.
但拥有它，并不意味着你一下子就赢了。
But it's not like when you have that, suddenly you win.
对。
Yeah.
你得把这模型和产品、分发渠道、用户体验这些环节都串联起来。
There's a bunch of pieces that you have to connect that to in product and in distribution and in the consumer experience.
我认为，正是这四个方面结合起来，才构成了我们的核心优势——而大多数竞争对手，不管是苹果、Anthropic、OpenAI还是谷歌，都只占其中一项。
And I think it is the collection of all four of those things that we see as, our superpower relative to the competitors, most of whom, whether it's Apple or Anthropic OpenAI or or Google, only have one of those things.
是的。
Yeah.
我稍后会更深入地谈产品。
I'm gonna get into product deeper into product in in a moment.
但上次我们聊的时候，你说过你不会和人工智能融合。
But first, last time we spoke, you told me you wouldn't merge with AI.
可你现在说的这种方式，是用技术把想法从脑子里传到电脑，再从电脑传回脑子里。
But the way you're talking about this is you use technology to get your thoughts from your mind to a computer and then from a computer back to your mind.
听起来很像那么回事。
Sounds a lot like that.
你改变主意了吗？
Have you changed your mind?
没有。
No.
我不觉得这是在和人工智能融合。
I I don't see this as merging with AI.
我仍然希望在这些事物之间保持清晰的界限。
I still wanna have a very clear separation between things.
下次我们聊的时候，我还会再问你一次。
I'm gonna ask you again next time we talk.
我知道。
Know.
让它持续下去。
It keep it going.
这其实是一个持续的趋势，而且正在加速——我们和机器之间、机器和我们之间的信息传输速率在不断提升。
I've it's a continuous, it's really a continuation of a trend, an acceleration of a trend where the bit rate between us and machines and machines back to us goes up over time.
其实，我们早就已经在做一些类似的事情了，只是有点搞笑。
And, like, there's funny versions of this that we've already been doing.
自动纠错。
AutoCorrect.
自动纠错就像是一个小小的AI，夹在你和电脑之间，帮你减少错误，实际上就是在提升你和机器之间的信息传输效率。
AutoCorrect is like a little AI that sits between you and the computer that, like, helps improve reduce the loss and effectively improve the bit rate between you and the machine.
还有好多这样的小工具，我们平时一直在用，就是为了加快这个循环。
And there's all these, like, little tools that we use all the time to accelerate the loop.
二维码，我特别喜欢的一个。
QR codes, one of my favorite ones.
二维码。
QR codes.
它就像是一个很酷的方式，让你不用动手输入。
It's like a way of, like, being like, cool.
我想输入一个网址，但绝对不想手动打字，因为出错率太高，搞不好就跳错网站，还得重新找。
I wanna, like, enter a URL, but I definitely don't wanna type a URL because the error rate is gonna be too high and it won't take me to the right website, and I'll have to look.
所以我们才用二维码。
So we use QR codes.
我觉得，如果AI真能用人类的语言、以人类的方式理解东西，那哪怕只是在输入端，也会极大提升我们利用现有算力的能力。
I think if AI if you have an AI that's really able to, in very human terms, in human language terms, understand things, that is a potentially profound improvement of our ability to take advantage of the compute we already have even if it's just on the input side.
再把AI高效地把信息反馈给我们的能力加进来，那我们的信息传输速率就真的大幅提升啦。
Now you combine that with the AI's ability to synthesize information more effectively to get back to us, you really tremendously improved the bit rate.
道格·恩格尔巴特离开NASA后，创办了旧金山研究所，就是这么回事。
This is what Doug Doug Engelbart, when he left, NASA to to, start San Francisco Institute.
他认为，人类面临的问题变得越来越复杂，而人类自身能力的提升却跟不上这个速度，所以他想打造一种人机协同的模式。
His idea was that human problems were getting harder at a steeper rate than human capability was improving, and he wanted to create this human computer symbiosis.
他说，唯一能做到这一点的方式，就是让人和计算机以某种方式紧密协作。
And he said that the only way he could do it is if teams of people could merge with computers in some way Mhmm.
这样才能真正把事情做成。
To make a do.
所以他主导了历史上第一次视频通话、第一次多人协同编辑文档，还发明了鼠标。
And that's why he led, you know, the first ever video call, the first ever joint document editing, the the the mouse.
所有这些创新，都是为了提升信息传输的效率。
Like, all these things came from wanting to increase the bit rate.
我觉得人工智能正是这样一种东西。
I think AI is exactly that kind of thing.
明白了。
Okay.
所以这种方式
And so the way
可能就体现在一个了解你上下文的个人助手身上，对吧？它能帮你把事情办妥。
that it manifests could be in this personal assistant, right, that knows your context Yeah.
主动去为你处理各种事务。
Goes out and gets things done for you.
可能是通过手机或电脑上的聊天界面，也可能是像Meta正在做的那种智能眼镜。
It could happen via a chat interface on a phone or a computer or through glasses like the type that Meta is making.
从产品角度来看，我觉得你之前已经稍微提过一点了，但我还是想再跟你深入聊聊这个话题。
And so from a product standpoint, and I think you've already previewed a little bit of this, but would like to talk to you a little bit about it a little bit more.
所有的产品最终不都会走向融合吗？
Don't all products end up converging?
所有的AI产品，不都会慢慢汇聚到这个个人助手的场景里吗？
Don't all AI products end up converging on this personal assistant use case?
如果你想想OpenAI这家公司，我们刚请过格雷格·布罗克曼来做节目。
If you think about what OpenAI is, we just had Greg Brockman on the show.
OpenAI 想做的，是打造一个超级应用，能帮你做事、理解你，真正地帮上你的忙。
And what OpenAI is trying to do is trying to create this super app that will get things done for you and understand you and really help you out.
你跟它聊天的时候，它就会主动出去为你处理各种事情。
As you talk to it, it will go out and do things in the world for you.
Anthropic 也是类似，Meta 也差不多。
Same thing with Anthropic, similar with Meta.
苹果也有类似的想法，不过我们还得等它真正上线后，看看实际效果怎么样。
And Apple has, again, a similar vision, although we'll wait to see what what it looks like when it's in the wild.
那你怎么看差异化的问题？你同意所有产品最终都会汇聚到这个智能助手的场景吗？
So how do you differentiate, and do you agree that everything sort of converges on this central assistant use case?
是的。
Yeah.
我觉得大家都做得挺有前景的，我们还处在最前沿，所以现在下定论还太早。
Well, I think I think everyone's doing exciting work and we're on the very forefront of it, so it's hard to say.
目前来看，Anthropic 和 OpenAI 正在重点发力企业市场，尤其是 Greg 带领下，他们都在搭建企业级解决方案——毕竟这才是真金白银的地方，我也理解他们需要收入。
I would you know, the work today, the the business that Anthropic is doing and that OpenAI appears to be increasingly pursuing, concentrating things under under Greg is an enterprise business where they're building these harnesses that and that's where the money is, and I understand that they need money.
所以从企业角度切入是个很重要的起点，因为这跟企业需求紧密相关。
So it's it's an important place to start where it's like it's actually very much attached to the enterprise.
从实际来看，所有的收入都集中在这里。
That's where where all the revenue is as a practical matter.
这一点我明白。
And I get that.
大公司确实如此，虽然钱都集中在同一个地方。
That's like that's that's you know, big companies, although there's a lot of money in one place.
这样一来，你只需要搞定少数几个客户，就能拿到大笔资金，毕竟这行是个资金密集型的生意。
So you have a smaller number of sales that you have to make, and you can get, like, larger amounts of capital, and this is a capital intensive game that they're playing.
我觉得他们的核心重点确实放在这些工作场景上。
I think their major focus is definitely on these, like, work use cases.
我觉得这些应用场景非常有价值。
I think those are super valuable.
当然，我们在专业工作中也一直在用这些功能。
Obviously, we take advantage of them as well in terms of our professional work.
这并不是我们的重点。
That's not our major focus.
我们的重点完全放在如何让这些技术真正改善消费者的日常生活上。
Like, our major focus is a 100% on how this is gonna help consumers in their lives.
我觉得真正的问题是，我不认为人工智能之间会变得完全无法区分。
And I think the real question I don't know that the AIs become indistinguishable from one another at all.
我觉得实际上，你自己已经提出了这个问题。
I think there's a real question of, actually, you you framed it yourself.
这些系统有点像私人助手，它们能接触到一些你肯定不希望广泛传播的个人信息。
These these are kinda like a personal assistant, and they have access to information about you that you certainly wouldn't want broadly distributed.
这些信息只对这个私人助手开放。
It's available to that personal assistant.
它是一个值得信赖的助手。
It's a trusted assistant.
如果你曾经雇过私人助手，换新人的时候，都会有个适应期，这正是问题所在。
Well, if you've ever had a personal assistant and hired a new one, there's, a ramp up period, that is that that involves that.
所以，如果你有一个深度融入你生活、表现得很好的个人助手，那就会建立起一种真正的连接，其他竞争者要想取代它，得付出极大的价值才行。
So, like, if you have this personal assistant that's actually quite embedded in your life and is doing well, I think that creates a real connection that you have that requires a lot of value from some other competitor to go replace.
你为什么
Why do you
觉得消费级AI的发展这么慢呢？
think consumer AI has been so slow to take off?
我的意思是，之前也有人尝试过。
I mean, have been some attempts.
比如那些角色型AI，还有复刻类的应用。
There have been, like, the character AIs, the replicas.
但你提到OpenAI，你说得对。
But you saw with OpenAI, you're right.
他们确实从商业角度做了调整。
They definitely pivoted from a money standpoint.
他们还是有一些面向消费者的用途，比如营养和健康方面。
They do have some consumer applications that they want, like nutrition, health.
这些消费级应用可能会触及一些更广泛的行业领域。
These are a consumer thing that might tap into some of our broader industries.
但你想想，从娱乐、陪伴，还有帮人处理日常琐事的角度来看，消费级AI本该特别吸引人，这些事你平时不会为了工作去请人帮忙做的。
But this idea, you would imagine that consumer AI would be very appealing to people from an entertainment standpoint, a companionship standpoint, and helping you, I guess, get done things in your life in a way that you wouldn't, you know, call on when you're doing it from a business standpoint.
但发展得一直很慢。
But it's been slow.
对。
Yeah.
我觉得吧，我们当初怎么就以为这一轮能逃过这个规律呢？炒作周期这东西，咱们行业从来都躲不开。
Well, I think, you know, I don't know why we thought this one was gonna be immune, but the hype cycle is is an evergreen concept that our industry continues to fall for.
其实大家也不是真不懂炒作周期。
And it's not that people often misunderstand the hype cycle.
只是对不了解的人来说，大家以为就是有个炒作高峰，然后跌入失望谷底，最后才慢慢找到真正的产品市场契合点。
They think how there's this the hype cycle for those who don't know, you know, there's a there's a peak, of hype, then there's the valley of discontent, then and there's the ultimate eventual product market fit.
炒作周期的重点，不是说这技术是假的。
And the point of the hype cycle isn't that the technology is it's fake.
愿意花这么多功夫去让这东西派上用场的人，其实只占人口中很小一部分。
It's just that people willing to go through a bunch of hoops to make it work are relatively small percentage of the population.
对。
Mhmm.
要把这技术普及到每个人，真正难的不是攻克技术难题，而是要付出大量实际努力。
And the work of bringing it to everybody is actually hard work, and it's hard work that is not just a matter of great, you've done this hard technology problem.
你得把界面设计得真正好用。
It's also you've made the user interface, you know, workable.
还得让它简单到人人都能上手。
You made it easy to use.
得让大家真真切切感受到它的价值。
People understand the value.
因为每个人都在过自己的日子。
Because people are living in their lives.
他们没这工具，日子照样过得挺顺。
They're having great success living their lives without this tool.
你是在要求他们改变习惯。
You're asking to change their habits.
你是在要求他们改变使用电脑的方式，而且是相当大的改变。
You're asking to change how they deal with, computers kind of in a pretty dramatic way.
但他们大多并不喜欢这样。
And they mostly don't like it.
这不会轻易成功。
It's not going to
你不能只靠价值来引导，得先说清楚。
it's not the you have to lead with value.
我们具体能为你做些什么，让你的生活变得更好？
What are we do what are the specific things that we're gonna do for you that are gonna make your life better?
我最喜欢的例子之一就是自主型工作。
Maybe my favorite example of this is the agentic work.
你知道的。
You know?
就像我们行业里的很多人一样，我在去年12月就早早开始用PY和MyClaw，折腾这些智能代理框架。
So like many other people in our industry, I was very early on in December with PY and then MyClaw, you know, using building playing with these agentic frameworks.
我觉得这些工具确实很强大，但用户体验太差了。
And I find them very powerful, but they're not very user friendly.
开发起来特别复杂，维护起来也特别麻烦。
They're very hard to build, to maintain.
而且用久了，效果还会慢慢走样。
They have drift over time.
我想起我曾经给我和我老婆搭过一个，还放到了微信聊天里，她能直接用。
And so when I think about, hey, I built one for my my wife and I, and I put it like on a WhatsApp chat, and she could use it.
但她从来都不用。
She never uses it.
我倒是天天用。
I use it all the time.
她就是不用。
She doesn't use it.
就是很难融入日常的工作流程里。
It's just it's hard to like integrate into a workflow.
她总是直接让我帮她做事情。
She just asks me to do things.
我就是那个执行的人。
I'm the agent.
然后我就自己去处理后续的事。
And then I, like, you know, go from there.
你这就算是分派任务了。
And you delegate.
对。
Yeah.
然后我就去跟那个代理系统对接。
And then I and then I go to the agent.
所以这其实就是个中间传递的过程。
So it's like that's the that's the pass through.
这个路径不错。
It's good path.
其实挺合理的。
It's actually not it's pretty reasonable.
对她来说效果很好。
It's working well for her.
我一点都不怪她。
I don't blame her.
如果我真成功了，反而有点担心——万一我造出一个代理，真把我从这个循环里解放出来了。
If I succeed I'm I'm actually worried if I make an agent that successfully gets me out of that loop.
所以我对这事也没那么迫切。
So I'm I'm not that eager for that.
我的意思是，我们还没让这些东西用起来真正方便。
So my point is like, we have not made these things easy to use yet.
我觉得我们在搜索和研究场景上的表现已经很棒了。
I think we've done a great job of, like, handling search use cases and research use cases.
我觉得人们都懂这些。
I think people understand those.
人们能理解生成式AI在内容创作上的应用。
I think people understand generative AI for content.
比如，我想生成一张搞笑的图片。
Like, I wanna make this funny image.
现在大家已经清楚我们能做什么，也愿意去用这些功能。
I think there's a few use cases that people understand our capabilities now, and they wanna go use those.
但我们还没做好让这些功能融入人们日常生活的工作。
But we have not done the work to make it something that people wanna integrate into their daily life yet.
用起来还不够方便。
It's not easy enough to use.
带来的价值也不够明显。
It doesn't create enough value.
操作太繁琐了，这才是需要解决的问题。
It's too fussy, And so that is the problem to tackle.
这根本是个产品问题。
It's the product problem to tackle.
对。
Mhmm.
要做好这件事，确实需要强大的模型，但光有模型还不够。
You need great models to do it, but great models are not enough.
没错。
Right.
你对AI伴侣怎么看？
Where do you stand on AI companions?
因为说到人们会依赖的助手，有人觉得只要功能做出来，用户自然就会来用。
Because, you know, when it comes to what will be a assistant that people rely on, there is this belief that you build the functionality and then people will come to it.
但另一种观点是，你得打造一个让人觉得像朋友的AI形象。
The other side of it is you build a you build a avatar, an AI avatar that people feel like they're friends with.
这才是真正的差异化所在。
And that is the way that you differentiate.
我的意思是，我们知道个性真的很重要。
I mean We know personality matters a lot.
所以我想说，我们从经验中学到的一点是——我觉得Anthropic在Claude各个版本的迭代中也深有体会——作为人类，我们非常在意自然语言是否让我们感到舒服或反感。
So I will say that one thing we've learned, and I certainly, you know, I think Anthropic has learned over the various generations of Claude, we certainly we care a lot as humans about the way natural language appeals to us or doesn't appeal to us.
因此，个性对这些模型来说确实很关键。
And so personality matters for these models.
不过话说回来，我觉得人群中对此的偏好差异会非常大。
Having said that, I think what you're gonna find is a very big distribution among the population.
有些人真的希望这个AI有实体形象，有个性，甚至有脸。
I think some people absolutely would like this this AI to be embodied and have, you know, a personality and have a a face.
事实上，确实有人在智能体领域里，想创建二十个不同性格的智能体，分别负责生活中的不同方面，比如健身教练、营养师、医生助手等等。
In fact, there's been some people who, when they in the agentic world, they wanna go create 20 different agents that each have a different personality for different parts of their lives, a trainer and a nutritionist and a doctor's assistant and all these different types of things.
但我不是这样的人。
I'm not one of those people.
我其实更喜欢——算了吧。
I actually like, nope.
我就希望我的AI特别可靠、值得信赖，至于它是不是有个具体形象，我根本不介意。
I just want my AI to be, like, extremely reliable and trustworthy, and, like, I'm fine with it being, an amorphous entity.
它不需要长得像人，我照样会信任它。
It doesn't have to have a human structure for me to care about it.
我可不想应付二十个不同的AI。
I certainly don't wanna deal with 20 of them.
我只想用一个，让它把所有事情都帮我搞定。
I just wanna deal with one of them and have it do all the things I need.
所以我觉得现在还太早了。
So I think that what we're it's very early.
现在下定论还为时过早。
It's too early to say for sure.
我觉得大家对这项技术的使用方式和接受程度，会差别很大。
I think you're gonna see a big range of how people want to engage with this technology and how what makes them comfortable with it.
因此，市场自然会提供相应的选择。
As a consequence, would expect the market to deliver that.
你知道吗，未来可能会出现一种情况，这些AI伴侣——这么说可能有点直白——但它们会成为新的社交媒体。
You know, there is a future where these AI companions become, this is a blunt way to put it, but the new social media.
对吧？
Right?
社交媒体是你去看朋友近况、和他们互动的地方。
Social media is a place where you go to see what's going on with your friends and you engage with it.
它可能非常全面，在最好的情况下，花在上面的时间是一个很重要的指标。
It's like, it's very it can be all encompassing and in its best case, And time spent is a pretty important metric.
不过，花完这段时间之后你的感受也同样重要。
Although how you feel after you spend that time is also important.
值得花的时间。
Time well spent.
值得花的时间。
Time well spent.
也许将来，人们会把时间花在和AI互动上，说到底，这不就是你如何与电脑上的东西互动吗？
And maybe that gets replaced by people spending time with I mean, ultimately, it's like how do you engage with something on your computer?
展开剩余字幕（还有 383 条）
也许人们会把时间花在某个非常关心他们的AI身上。
Maybe that gets replaced with people spending time with some AI entity that cares a lot about them.
是的。
Yeah.
我的意思是，我不太会去评判别人的选择，我并不评判。
I mean, I try not to judge the way people choose I'm not judging.
对。
No.
我同意。
I agree.
没错。
Yeah.
说到技术。
With technology.
我的直觉是，对绝大多数人来说，AI最大的好处其实是腾出更多时间陪伴在乎的人、爱的人。
My my instinct is that for the overwhelming majority of people, the major benefit of AI is gonna be increased time for human contact, for people who they care about, people they love.
我经常在讨论增强现实的时候提到这一点，比如。
And, you know, I talked to this a lot in the context of of augmented reality, for example.
就拿我们现在戴的摄像头眼镜来说吧，陪孩子的时候，我既能拍下画面分享给妻子，让这件事对我们有意义，又能完全沉浸在当下。
You know, even just the the the camera glasses that we have, you know, when I'm with the kids, I'm able to both record something and share it with my wife, which is meaningful to us, and also be fully present.
我手里没有手机，这对我来说很重要。
And I don't have a phone between me and them, and that's an important piece for me.
我之前说过，如果你能更高效地完成工作，就能省下通勤的时间。
I've talked about with if you were able to be more effective with your work, that's more time that you're not spending commuting.
省下的时间就能多陪陪家人，陪陪你爱的人。
That's more time that you're not spending away from your families from the ones that you love.
我个人觉得，对绝大多数人来说，真实的人际连接的价值，只会随着时间越来越重要。
My personal sense is that the overwhelming peep majority of people, the value of authentic human connection, only goes up over time.
它不会越来越低。
It doesn't go down over time.
我觉得，人们早期对人工智能的反应，已经稍微体现出这一点了。
And I think we're seeing that a little bit in how people's reactions to AI early on have been.
我觉得人们担心这是技术在取代人。
I think people are worried that it's a replace of technology.
但我自己并不这么看。
I don't find it that way myself.
我是个很频繁的使用者，实际上，多亏了它，我花在电脑前的时间反而更少了，而不是更多。
I think I'm an avid user of it, and actually, mostly, I'm spending more time not having to be at my computer thanks to it, not the opposite.
我觉得这就是大多数人会如何使用它的趋势，它会如何改变人们与媒体、与亲人的关系——我认为，人们更看重真实的人际连接和真实的时刻。
So I think that's how the I think that's that's my prediction on how the overwhelming majority of people will interact with it and how it will affect their relationship to media and to their loved ones, which I think it's a premium on authentic connection and and authentic human moments.
但我也知道，各种不同的使用方式都会存在。
But I'm sure the entire distribution will exist.
没错。
Yep.
当然，AI眼镜是实现这一愿景的核心。
And, of course, the AI glasses are kinda core to that vision.
是的。
Yeah.
没错。
That's right.
我们马上接着聊这个。
So we'll talk about that right after this.
大家好。
Hi, everyone.
我是亚历克斯·坎特罗维茨。
Alex Kantrowitz here.
我想跟大家介绍一下，我和Gravity合作制作了一部纪录片，探讨AI代理的安全未来。
I wanna tell you about a documentary I've made with Gravity to explore the future of AI agent security.
为了弄清楚我们是否真的准备好迎接自主代理，我采访了麻省理工学院的拉马斯·拉斯卡尔教授、前白宫首席信息官特蕾莎·佩顿、米其林集团首席数据与AI官阿姆比卡·拉贾戈帕尔，以及前阿里巴巴高管莎伦·盖伊。
To find out if we're truly ready for autonomous agents, I sat down with MIT professor Ramesh Raskar, former White House CIO Theresa Payton, Michelin's Group Chief Data and AI Officer Ambika Rajagopal, and Sharon Guy, a former executive at Alibaba.
他们各自都对这个不断变化的领域提供了独特的见解。
They each offer unique insights into this evolving landscape.
最后，我们还对话了Gravity的首席执行官罗里·布伦德尔，一起探讨未来的方向。
We conclude with Rory Blundell, CEO of Gravity, to discuss the path forward.
在Gravity的引领下，一起来探索这段旅程吧。
With Gravity leading the way, join us on this journey.
你可以在节目简介里的链接观看完整纪录片。
You can watch the full documentary at the link in the show notes.
我们又回到《大科技》播客，今天邀请到的是安德鲁·博斯沃思。
And we're back here on big technology podcast with Andrew Bosworth.
博兹，Meta的首席技术官。
Boz, the CTO of Meta.
博兹，很高兴再次见到你。
Boz, great to see you again.
谢谢你抽出时间跟我聊聊。
Thank you for taking the time to speak with me.
如果拉个全景镜头，大家能看到，我们现在就在纽约，你们团队刚刚发布了三款Meta设计的新款眼镜。
If we go to the wide shot, we can see we're here in New York at a moment where you and your team are releasing three new pairs of Meta designed glasses.
我们一直在节目中讨论这个问题：你的手机才是AI设备，还是可穿戴设备才算？
It's it's it's something we've been debating on the show as sort of is your phone the AI device, or is it a wearable?
我们又看到了苹果的时刻，他们似乎准备推出一个真正能用的苹果智能系统，能一定程度上理解你的上下文，甚至帮你完成一些事情。
And we've had this moment again going back to Apple where it looks like they're preparing to release a version of Apple intelligence that actually works, that knows your context to a degree and might be able to get things done for you.
而另一方面，Snapchat推出他们的智能眼镜，结果引发大量负面评价，很多人觉得太差了，根本不用多说。
And then we see, you know, sort of the opposite side is the Snapchat specs release, which got a lot of people saying, maybe we don't mean, those were so bad that people were just you don't have to comment.
我。
I'll
我没法评论。
say I can't comment.
我的意思是。
I mean
我还没见过。
I haven't seen them.
我自己也还没看过。
Haven't seen them myself yet.
我就说吧，我的看法其实反映的是市场的反应。
Let's just say with I'll just my comment reflects what the market did.
埃文·斯皮格尔戴着它们去参加了某个发布会。
Evan Spiegel wore them out to some presentation.
我觉得Snap的股票立马就跌了大概6%。
I think Snapstock went down, like, 6% immediately.
事实就是这样。
It's just what happened.
好吧，这大概是我第一次拍视频戴这个，我也不太确定。
Well, this will be the first video of me wearing I don't know.
这事儿迟早会发生。
It's gonna happen.
你确实有。
You have.
咱们还是让市场来评判吧。
We'll we'll we'll we'll let the market decide.
对。
Yeah.
但我真的很想听听你的看法，毕竟Meta在这件事上投入了不少。
But but I'd love to hear your thoughts on your obviously, Meta has invested a lot in this.
你觉得这是一个很有吸引力的使用场景。
You believe it's a compelling use case.
如果我说，也许我们根本不需要AR或者AI眼镜，
If I were to say, maybe we don't need a r or AI glasses.
手机就够了，你怎么看？
We can just use our phone.
是什么让你觉得另一方的观点更有道理？
What would you say that makes you feel the other side of that better?
是的。
Yeah.
手机确实很棒。
Phones are great.
我的意思是，我很喜欢手机。
I mean, I love phones.
我有两副。
I have two of them.
我觉得它们用起来很舒服。
I think they're happy.
这是很棒的设备。
They're wonderful devices.
从一开始，我们就问自己这个问题：到底需不需要AR眼镜？
I I the the glasses from the very beginning, the question we asked ourselves was this exact question.
我们想，没错。
We said, okay.
手机确实很好用。
Phones are great.
有什么功能，是你希望不用从口袋里掏手机就能直接用上的？
What is something that you wish you could get access to, it's on your phone, without having to your phone out your pocket?
我们想到的答案是摄像头和音频。
And we came up with camera and audio.
这其实很简单。
This is just very simple.
感觉挺酷的。
It's like, cool.
要是我能直接做到这一点就好了。
If I could just do that.
人工智能带来了巨大的推动力，随着时间推移，它所能释放的潜力远超手机通过蓝牙连接所能实现的功能。
The AI has been this tremendous tailwind where actually it unlocks a much larger swath of potential capability over time than what the phone can do, just through, you know, Bluetooth connections.
所以，现在看来前景比两年前或三年前明朗多了。
And so, yeah, it's much more promising now than it looked two years ago or three years ago.
两三年前，这看起来就像是——
Two or three years ago, this looked like, hey.
迟早得给它加个屏幕，变成一个独立的系统，还得配上一堆外设。
At some point, you have to put a display on this, and it has to become a stand alone system, and it has to have all this, you know, kind of accessories attached to it.
现在看来，可穿戴设备市场完全有足够的空间容纳多种不同类型的产品。
Now it actually looks like there's a totally enough room in the market for a big range of of of wearable devices.
眼镜肯定有，但不止是眼镜，可能还有很多其他东西。
Glasses, certainly, probably not just glasses, probably a lot of other things.
人们不想戴眼镜。
People don't wanna wear glasses.
他们想戴别的东西。
They wanna wear different things.
有些设备只是作为手机的输入输出终端。
And some of those devices are just gonna be input and output to your phone.
这挺棒的。
That's cool.
你的手机本来就很好用。
Like, your phone's great.
如果它能让你的输入输出更高效，那简直太好了。
And and if it's just making your life, like, more efficient in terms of how it's doing input output, that's awesome.
有些设备会更完整一些。
Some of them will be more complete.
比如我们刚刚为Meta A Band显示眼镜推出了一套全新的开发平台。
So for our our the Meta A Band display glasses, for example, we just launched a vibe coated platform for it.
任何人都可以自由开发自己想要的应用程序，直接装到眼镜上。
And so anybody who wants to can go literally just build whatever app you want for the glasses.
现在你得先开发好应用，再手动安装到眼镜里。
Now right now, you kind of build the the app and you, like, put them on the glasses.
但未来完全没必要这样——你戴上眼镜，直接告诉它现在想用哪个应用，它就能实时为你生成并运行。
But in the future, there's no reason that couldn't just be you wearing the glasses in real time, telling the glasses what app you want right now and having it on the fly build that app for you.
这挺有意思的。
Interesting.
你明白我的意思吧？
You know what I'm saying?
我觉得我们正朝着一个非常棒的方向发展，确实如此。
And so I think we are headed towards a very cool zone Mhmm.
未来的应用生态不会像现在这样封闭，但内容中心还是会存在。
Where it's a little less like app garden specific, you're still gonna have these content homes.
内容始终是永恒且重要的，就像在电视上、在社交媒体上、在任何地方一样。
Content continues to be an evergreen and important thing as it has been on TV, as it has been on social media, as it has been everywhere.
总会有一些地方，你想要触达的媒体内容就存在于那里，它们看起来有点像应用、频道，或者不管怎么说，就是这么个东西。
So there's still gonna be places where media that you wanna reach lives, and those are those look kinda like apps or or for, or channels or whatever lack of a better term.
但还有大量小众的需求。
But there's a long tail of things.
我的烤面包机干嘛非得有个应用？
Like, why does my toaster need an app?
我认真问你一个问题。
Let me ask you this in seriousness.
我的烤面包机确实有个应用。
Like, my toaster has an app.
我觉得它根本不需要这个。
I don't think it needs one.
我可不想要这种东西。
I don't want that.
没错。
Right.
我只想告诉我那个AI助手，给我弄点我想吃的吐司。
I just wanna tell my AI agent, get me the toast that I want.
就是我每天吃的那种吐司。
It's the same toast I have every day.
你直接帮我拿过来就行。
Just get it for me.
我不想去管每样东西都怎么操作。
I don't wanna have to go do what everything is.
你的吐司app能干嘛？
What does your toaster app?
它能让你做什么？
Is it does it let you
我真的一点都不想安装它。
I honestly I I refuse to install it.
我也不愿意安装，我尊重你的决定。
I refuse to install I respect that choice.
我就是不干。
I refuse.
我绝对不这么做。
I absolutely won't do it.
所以你得为某些事站出来。
And so You have to stand up for something.
是的。
Yeah.
你听我说。
Listen.
有个底线。
There's a line.
有个谁都知道的底线。
There's a line that nobody you know?
是的。
Yep.
我觉得吧，说实话，你确实可以做到。
I think I think you can actually it's a I have to admit.
有时候，能有个专门的App来控制所有功能，其实挺棒的，我也认可这一点。
Sometimes, like, the it's so cool that you can have a specific app to control every aspect of the thing, and I I respect that.
而且我本来就是个科技爱好者。
And I'm a I'm a tech guy.
对吧？
Right?
所以我喜欢这种折腾的乐趣。
So I like the fidgety nature of it.
但说实话，现在这情况已经有点过头了。
But it's like, literally, at this point, it's kind of gotten out of hand.
我其实就想跟一个智能系统说：喂。
What I really just wanted to is to tell an intelligent system, hey.
帮我把想要的东西找出来，然后让它自动帮我搞定。
Get me the thing that I want, and it can do that for me.
我们现在已经能看到这种模式的雏形了。
And we see an early form of this.
比如我们和Spotify的合作，你只要对眼镜说一声，想听歌就行。
You know, our partnership with with Spotify, you ask the glasses to play music.
只要你绑定了Spotify账号，它就会自动把你想听的音乐找出来。
It go if you have a Spotify account linked, it goes and gets the music you want.
这确实挺棒的。
And it's like, yeah, this is great.
这正是我想要的。
This is what I wanted.
我不希望为了做这么一件事还得折腾好几步。
I didn't wanna have to go through a bunch of steps to do this.
至少对我来说，我的想法是，手机当然很好，而且还会继续好下去。
So for me, at least, the way I'm thinking about this is not that, phones are great and they're gonna continue to be great.
我不觉得未来会是那种靠App主导的样子。
I don't think the app y thing is the way the future's gonna look.
我觉得未来会是各种有价值的服务。
I think the future's gonna be valuable services Right.
这些服务会主动提供给你，你想什么时候用、怎么用，都能自由选择，然后为这些服务付费，费用可以提前谈好，也可以随时按需支付。
That are provided to you, and you getting access to those services the way that you want when you need it, and paying money to the people who provide those valuable services, all negotiated either in advance or on demand.
我真的很相信
I really believe in
这一点。
this.
我今天用了Meta AI应用，看到你们已经把佳明的设备和眼镜连起来了。
I saw you had the I was on the Meta AI app today, and I saw there's a Garmin connector to the glasses.
对我来说，跑步训练的时候，要是能直接说：
And for me, you know, as I'm training, I'd love to be able to say, well, I'm building up to this, like, half marathon.
Meta AI，帮我找找附近这个时间段内有没有五公里的跑步活动，帮我报名。
Meta AI, find me a five k in my neighbor in my area, in in this window, and sign me up.
没错。
Totally.
而且是在跑步的时候就能完成。
And to do that as I'm on a run.
对。
Yeah.
这样我就不用花一个小时自己慢慢琢磨了。
So I don't need to spend an hour figuring it out on my own.
我完全同意。
Agree completely.
再往上说，你的Meta AI最好能早就知道你在训练，而且清楚你设定的目标。
And taking it at a higher level, your Meta AI ideally would already know that you're training, and you have a goal that you're trying to reach.
它还能连上所有相关的信息，比如你的饮食，这才是我们想实现的方向。
And it's tied into all the pieces that matter, your nutrition and your know, it's like that's like that's the direction we want to get this thing.
从现在到那个目标之间还有很多步骤，但方向就是这样。
There's a lot of steps between now and then, but that is where we're going.
上次我们聊过那个Orion眼镜。
The Orion glasses, we talked about those last time.
对。
Yeah.
现在进展到什么地步了？
Where do those stand?
就是那种完整的AR眼镜。
Those are the full AR Full AR glasses.
使用体验。
Experience.
是的。
Yeah.
Orion对我们来说是个特别重要的里程碑，毕竟我们早就有了AR的愿景。
So Orion was such a a important moment for us, you know, having had this AR vision for such a long time.
终于，他们给了我们一台设备，可以开始拿来做软件实验了。
Finally, they gave us the device that we could use to start to play with the software on.
虽然我们没能把价格降到觉得适合面向消费者推出的水平，但我们在设计和开发时，初衷就是把它当作一款消费级产品来做的。
And even though we couldn't get the price to be one that we felt comfortable launching as a consumer product, we did intent when we designed it and developed it, it was a consumer design and intention.
所以这个产品本身，戴起来舒服吗？用起来方便吗？
And so the product itself is, like, is it quite wearable, quite workable?
我家里就有一副。
Like, I have a pair at home.
我们用它来测试软件。
We use it to test the software.
我们一直在迭代软件，而且在软件上取得了大得多的进展。
So we've continued to iterate in the software, and we've made so much more progress in the software.
这不只是因为AI变强了——虽然这确实带来了巨大改变，更重要的是，我们有了Orion这个开发平台，这才是关键。
Not just because AI has gotten better, but that makes a huge difference to what that software is, but also because you have Orion to develop on, which makes a big difference.
所以，我们依然非常专注于整个产品线。
So, yeah, we continue to be very focused on the entire spectrum.
我们之前也提过，除了显示眼镜和摄像眼镜，其实还有一系列价格更低的产品在规划中。
You know, we've hinted here that, you know, in addition to display glasses and camera glasses, you know, there's a whole range of glasses that may be below that in the price range.
但我依然相信，全功能增强现实将是这个领域未来的发展方向。
Well, there also may be I really still believe in full AR as a a future for the space.
我觉得我们会继续沿用至今的策略，原因和我们当初没有推出Orion是一样的。
I think we're gonna continue to take the same approach we have so far and the same reason we didn't launch Orion.
光有这些功能还不够。
It's not just enough that it does all this functionality.
它还得外观漂亮、佩戴舒适，让人愿意天天戴着，价格也得让普通人觉得值这个钱。
It has to look great, has to be comfortable enough that you wanna wear it, It has to be at a price point that that a reasonable person would say, yeah, this is a this is a good value.
那么，离那一天还有多远？
So So how far
这个我说不准。
away is that?
我不能
I'm not
给出具体的时间。
gonna say exact number.
我得说，我喜欢
I will say I like
我们取得的进展。
the progress we're making.
是以年还是以月来衡量？
Measured in years or months?
我不会回答这个问题。
I'm not gonna answer that.
好吧。
Alright.
这说得通。
That's fair.
我很欣赏你们的努力。
I appreciate the hustle.
总得问一问。
Have to ask.
我知道你确实这么想。
I know you do.
我有点犹豫，是因为像我们这样在类似公司待过的人，都懂这个。
It's and some some of my reticence is, you know, we we people who have been in companies like ours know this.
我们一直在评估各种项目，不断问自己：这个是不是对的？
We're constantly looking at vehicles and, like, asking ourselves, is this the one?
它准备好了吗？
Is it ready yet?
这个是不是对的？
You know, is this the one?
而且，我们真的快要进入状态了。
And, man, we're we're getting in
进入状态了。
the zone.
这真的很让人兴奋。
It's pretty exciting.
好的。
Okay.
不错。
Cool.
我们来聊聊元文化吧。
Let's talk about metaculture for a moment.
对。
Yeah.
你负责应用人工智能部门，对吧？
You've you're running this applied AI division That's right.
这个部门最近引起了一些报道。
Which has been the subject of some reporting.
我负责代理式转型加速器。
I run the agentic transformation accelerator.
没错。
Right.
那个部门里有一个团队，就是AI团队。
One of the groups in that is the is the AI team.
对。
Yeah.
好的。
Okay.
我念一段《连线》杂志的引述。
I'm just gonna read the quote from Wired.
有位员工对《连线》说，那地方简直就是劳改营。
One employee told Wired, it's literally the gulag.
你突然就没了人生目标。
You have zero purpose in life all of a sudden.
你几乎不跟任何人交流。
You barely interact with anyone.
你每周就只有一堆任务，据说那些员工被安排做一些AI谜题，用来训练人工智能。
You just have these tasks every week, apparently talking about how, you know, the employees there have been put on some, like, AI puzzles that they they have to try to accomplish that helps train the AI.
那边到底怎么回事？
What's going on there?
我不太清楚。
I'm not sure
这个人根本没查过古拉格到底是什么样子，和硅谷六位数的软件工作有多像或者有多不像。
this person's ever googled what a gulag was like and how similar or not it is to a six figure software job in Silicon Valley.
看起来确实没有，但居然有人会这么说，
Doesn't seem like it, but the fact that they would say that
抛开夸张的成分不谈。
Setting set aside setting aside the hyperbole.
明白了。
Okay.
是的。
Yeah.
所以我们内部已经花了很多时间讨论这个问题。
So we've been spending a lot of time on this internally.
这是一个非常重要的议题，你们一直长期关注我们，所以你们知道，这家公司一旦进入紧急状态，就会全力以赴。
It's a hugely important topic for You've covering us a long time, so you know Like, this is a company that goes into lockdowns.
当我们面前有个紧迫的机会时，我们就会这么做。
Like, when we have an urgent opportunity ahead of us, we, like, do this.
我们对移动业务就这么做过。
We did it with mobile.
对视频功能也是这么做的。
We did it with video.
对动态消息功能同样如此。
We did it with stories.
我们确实这么做过，虽然这些举措并没有每次都彻底改变公司方向，但有时候我们会意识到，如果现在集中精力把某件事做到极致，市场里就藏着巨大的机会。
We've done it and it's not that they every one of these things pivots the entire company, but there are moments where like, wait, if we put exquisite effort on something right now, we think there's a tremendous opportunity for us in the market.
这次我们就看到了这样的机会。
And in this case, we saw that.
当我们推出 Muse Spark 的时候，我们真的觉得……但我要小心表达。
We we really feel like, you know, the when we came out with Muse Spark and I wanna be careful.
Muse Spark 是个很棒的模型，我们对它特别期待。
Like, Muse Spark is is a great model, and we're really excited about it.
原本我们没把编程能力当作这个模型的重点，但它在编程上的表现居然比我们预期的还要好。
It's coding had not been a focus for us on the model, but it actually was a better out of the box at coding than we had expected it to be.
早期通过实验我们发现，只要给它一些不算多、但经过精心设计的示例，再做一次微调，就能大幅提升它的竞争力。
And we found early on through experiments that, like, actually giving it just a relatively modest number of trained, kind of expertly guided examples and we could post train the model, we could dramatically improve its competitiveness.
当你真正算清楚这笔账，就会意识到：这对我们来说是个绝佳的机会——不仅能让我们在运营上拥有更多自主权，还能打造一个对用户极具价值的编程工具。过去一年，能写代码的AI已经成了这类系统里最强大的功能之一。
And so when you start to, like, run the numbers and the math on this, you're like, oh, this is an incredible opportunity for us to build a coding model that not only, allows us to have independence to how we operate the company, but also something that we think is gonna be valuable, both inside of if you, you know, give users AI that's able to code, that's obviously one of the very, very powerful tools that's kind of become very common in these AI systems over the last year.
同时，未来我们也能让这个模型本身更广泛地开放给更多人使用。
And then also, for us to be able to make the model itself more widely available over time.
所以我们看到了这个巨大的机会，立刻做出了调整，动员了公司里成千上万的人，全身心投入到这个AI项目中，专门去做这些高质量的示范训练。
So we basically saw this huge opportunity, such a big opportunity that we pivoted kind of on a dime and brought a lot of people across the company, thousands of people, out into this AI organization to to do these expert traces.
我们真的非常需要他们的专业能力。
We absolutely need their expertise.
如果做得不够好，这事根本行不通。
It doesn't work if you do a bad job.
结果发现，如果用糟糕的代码来训练模型，反而会损害它的性能。
It turns out if you use a bad piece of coding to train the model, you do some damage to it.
对。
Yeah.
你不想让错误的行为被强化。
Have You don't wanna reinforce failure.
这些例子必须做得很好。
They have to be well done.
必须由专家精心指导。
They have to be expertly guided.
我们当时做得非常快。
Now we did it very quickly.
因此，整个过程缺乏系统性。
And as a consequence, it did not have a lot of structure.
也没有良好的沟通机制。
It did not have great communication around it.
我之前说过这话。
I've been on record.
其实那不是真的。
Actually, that's not true.
我并没有公开说过。
I wasn't on record.
是被泄露出去的。
I was leaked.
内容被泄露了。
It was leaked.
说这事儿太恶劣了。
Calling it atrocious.
你之前说的是‘可能’，不是二十年来最糟的，但确实很严重。
You said maybe, not the worst it's ever been in twenty years here, but it's up there.
确实非常严重。
It's definitely up there.
那根本不是我的原话，我也不清楚你从哪儿听来我说过这话。
That actually was not a quote for me, and I don't know where that You didn't say that.
我从来没说过这话。
I didn't say that.
好吧。
Okay.
不过我确实说过类似的话。
But I've said things like it.
都是紧急的事情。
Urgent stuff.
我对这个没什么意见。
I'm fine with it.
所以，好吧。
And but So the Okay.
它作为一个大公司的程度。
The degree to which it's a big company.
我们当时看到了这个紧急的机会，做出了我认为在战略上完全正确的调整，但却没花时间去一一跟每个人谈一谈，跟他们说清楚。
The degree to which we saw this urgent opportunity and made the change that I think strategically was absolutely the right change, but did not do the work to kinda go to each person and be like, let let me talk to you about Mhmm.
这到底是什么，为什么我们需要它，为什么它这么重要。
What this is and why we need it and why it's important.
没错。
Yep.
我们知道，他们原本在做自己特别感兴趣的工作，现在为了做这个项目，只好暂时搁置了。
Knowing that they had they had other work that they were excited about that they were putting on pause to come do this work.
你懂的。
You know?
但这是我们公司一贯的做法。
But this is that is something our company does
对。
Mhmm.
当我们觉得在某个瞬间看到了难得一遇的绝佳机会时，就会这样行动。
When we feel like we see these, like, unbelievable opportunities that exist in moments of time.
所以，我们确实正在应对这场席卷整个行业的变革，每个公司也都身处其中。
And so, yeah, we are, like, navigating this this change that's happening in the industry is happening inside every company as well.
没错。
Right.
这真是前所未见，你刚才说得很对。
And, it's like nothing we've ever seen you said you let out with this.
在我们的职业生涯里，从没见过这样的情况，我觉得这让很多人开始犹豫了。
It's like nothing we've ever seen before in our careers, and I think that is giving people pause.
因此，对我和其他领导者来说，必须更好地沟通：到底发生了什么？为什么会这样？对你有什么影响？长远来看会怎么发展？
And so it raises the bar on me and other leaders do a much better job than we have done communicating what's going on, why is it happening, how does it affect you, how do we see it playing out long term.
要让他们明白，他们所扮演的角色至关重要，不可或缺。
Make sure they understand that the role they're playing is one that we consider very critical, very important.
如果不是这样，我们也不会做出这个调整，这很明显。
Otherwise, we wouldn't have made that change, obviously.
我们能简单聊聊追踪机制吗？
Can we talk about the tracking briefly?
是的
Yeah.
说实话，如果我是员工，我可能也不会喜欢这种做法。
I actually you know, I under if I was an employee, I don't think I'd be a fan of it.
但最近我们在节目里其实已经讨论过，为什么你们可能会这么做。
But I actually sort of made the case for why you might be doing it on our show recently.
现在我们坐在一起了，正好聊聊这个事。据报告，Meta已经开始追踪员工的键盘输入和打字方式，并用这些数据来训练模型。
And now that we're sitting next to each other, let's talk about it because so basically the the reports have been that that Meta has started to track some keystrokes and the way that employees type and and basically use that as a way to train models.
我的看法是，随着模型训练转向强化学习——就像Scale AI的亚历山大·韦恩说的，他们现在大部分训练都已经是强化学习，而不是预训练了。
And my perspective on this was as model training moves into reinforcement learning where I think Scale AI where Alexander Wayne came from said most of their training is reinforcement learning now as opposed to pre training.
没错。
That's right.
我们之前也提到过。
We talked about previously.
当技术进入强化学习阶段后，让模型学会如何在所谓的‘环境’中完成任务就变得特别重要，这些环境其实就是各种模拟现实场景的练习场，模型在里面尝试完成具体目标。
As the technology moves into reinforcement learning, it's very valuable for these models to learn how to accomplish tasks in what's typically called as gyms or like different areas that different like simulations of real world activity that they go in and try to accomplish.
所以我的理解是，这个项目其实就是把刚才说的那种方式大规模放大了，让模型观察员工如何完成任务，然后自己学会去完成这些任务？
And so am I right in thinking that this program is basically a just massively scaled up version of that where the models watch employees work through their tasks and then learn how to accomplish tasks on their own?
是的。
Yeah.
不过这事得从两个方面来看。
Well, there's two parts to this.
你说得一点都没错。
The first one is, you're absolutely right.
如今的人工智能里，强化学习的作用比两三年前大家预想的要大得多。
Reinforcement learning is playing a much bigger role in today's kind of AI than people had maybe predicted two or three years ago that it would.
但这还不是全部。
It's not just that though.
人类的知识和行为的长尾部分，其实非常长。
There's also the long tail is long.
也就是说，人类做事的方式和积累的经验，远远超出我们能想到的范围。
Like, the long tail of human knowledge and and behavior is very long.
实际上，互联网上所有的文本加起来，也只覆盖了我们所知内容的一小部分。
And most of it, as much as for all the text, for the entire corpus of text on the Internet, most of the stuff that we know is still not on the Internet.
很多东西其实就在我们脑子里。
It's, like, in our heads.
那是经验。
It's experience.
是日积月累形成的。
It's built up over time.
是一些我们早已习以为常的行为。
It's behaviors that are second nature to us.
所以，我觉得这个系统在某种程度上特别巧妙。
And so this system was, in some ways, I thought quite genius.
员工根本不需要改变任何工作方式，照常做事，就能自然产生这些独一无二的数据。
You've got employees who need to change nothing about how they go about their day, can go about it as they always have, and in doing so, produce this corpus of unique data.
比如，人类到底是怎么使用电脑的。
In this case, designing how do how do humans use computers?
人工智能在使用电脑这件事上，其实还是特别不擅长。
AIs are actually still really weirdly bad at just using computers.
这问题看起来简单，实际上却出乎意料地难，目前还没人真正解决得好。
Like, it's a it's a it's a it's it's a surprisingly hard problem that is not well solved.
现在大家在电脑使用上投入最多精力的，就是
And that's where all the energy is going with computer use and
身份识别。
identity.
所有电脑操作都围绕这个。
That's all computers.
你当然可以先在前端大幅提升智能，再慢慢提炼简化。
And you can you can ramp up, you know, the intelligence in the front end for sure and then try to distill down from that.
但我们确实相信，有了这些数据，能让人生活更轻松。
But we do think having this data has the potential of making people's lives easier.
这根本不是关于内容本身，那个曾经难以传达的东西——我们之前没说清楚——重点根本不在你正在做的具体事情上。
It's not even about the content, the the the thing that was a challenge to communicate, and, again, we did a poor job, was not even about the content of the thing that you're doing.
关键是电脑怎么理解这个数字界面里发生的一切，而我们现在用的很多工具，都是通过这个界面来操作的。
It's about how is the computer able to understand what's happening inside this digital interface, which is the way we access a lot of our tools, like, in the in the world today.
第二点是，我觉得这个数据集挺有意思，但到底行不行，现在还说不准。
The second thing is, and and so I it's so I think this dataset is interesting, but it's we won't know.
这更像是一个长期积累的数据集。
It's like a it's a long running dataset.
对于那些需要长期训练的专家级任务，还是像我们应用AI团队现在做的这种方式更有效。
So the second part of this is you're still for a long tail expert training, you're better off doing work like we are doing with our applied AI team, the AI team.
就是一小部分经过详细记录的任务，用来对模型做后续训练就足够了。
Like, that is a relatively small number of really well documented, you know, tasks that can post train a model.
但这个情况不一样。
This is a different thing.
这个是真正长期持续的。
This is, like, very long running.
等我们积累了一年的数据，就有可能用它来优化模型了。
Once we have, like, a year of data, you have something that's potentially interesting to to bring to bear on the model.
我想补充一下，自上线以来，我们对这个项目做了不少调整。
I do wanna add, we've also made a bunch of changes to the program since the launch.
对。
Right.
我们增加了三十分钟的休息时间。
We've added We take a thirty minute break.
暂停一下。
Pausing.
人们选择退出的原因有很多。
People can opt out for a bunch of reasons.
是的。
Mhmm.
所以我们针对那些有顾虑的人，对项目做了一系列调整，比如
So, like, we've we've made a bunch of changes to the program for people who had concerns about
你最近把很多以前的博客文章发到了 Substack 上。
So you are posting a lot of your old blog posts to Substack.
我一直在邮箱里收到这些文章，也都读了。
And I've been getting them in my email and reading them.
最近我读到一篇特别有意思的文章，讲的是你做生物研究时，医生说疼痛就是康复的一部分。
And there was very interesting one that I read recently talking about how you were doing some biology research and the doctor said the pain is rehab.
要康复，就得经历这种疼痛。
You need that pain in order to be able to heal.
你提到过，要想真正进步，就必须学会接纳疼痛。
You write as at some point, have to be able you have to embrace the pain to make real progress.
在两个其他方面都一样的故事里，人更容易记住那个引发更强情绪的那个。
Given two otherwise equal stories, humans remember the story that evoked stronger emotion.
情绪决定了我们的大脑如何筛选记忆。
Emotion is how our brain triages memories.
有时候，只有疼了，大脑才会把它列为优先处理的事。
Sometime it has sometimes it has to hurt for your brain to prioritize it.
特别感谢
Shout out
就是BS80，我在哈佛上的神经生物学课。
to BS 80, a class my my neurobio class at Harvard.
人工智能就是我们的进化生物学。
AI is Our evolutionary bio.
人工智能正在帮我们减轻很多痛苦。
AI is taking away a lot of the pain.
对吧？
Right?
现在人类用人工智能做的，很大程度上就是把工作中那些痛苦的部分交给了它。
Like big part of what humanity is doing with AI right now is a lot of the painful parts of our work.
我们把这些都交给AI了。
We're giving it to AI.
如果这个目标实现了，那我们到哪儿去寻找痛苦呢？
If that goal is accomplished, where do we find the pain?
所以，我特别喜欢这一点。
So the this so I love this.
顺便说一句，我让我的智能助手把我的博客文章迁移到Substack，这样以后就能同时在两个平台发布。
And a very small aside, my so I I one of things I did is I assigned my agent the task of bringing my sub my blog posts over to Substack so that at some point I could do both.
直到最近我才意识到，它根本没保留任何项目符号。
I didn't realize until very recently that it wasn't any bulleted list.
它直接把项目符号全删了。
It would just strip out.
我的智能助手根本不懂项目符号列表。
My my agent did not understand bulleted list.
智能助手还有很长的路要走。
So we have a long ways to go on agents.
我的意思是，还行吧。
Is my okay.
第一阶段。
Phase one.
所谓的痛苦，其实是康复过程中的必经阶段。
The the the pain is the rehab came yeah.
我们当时在研究一个课题，关于戒断毒品时大脑的神经生物学变化。
There was a a question we were studying, the neurobiology that would occur during withdrawal from, drug use.
有个学生突然问：嘿。
And a student asked, hey.
我们有这么多症状，
We have all these symptoms.
那为什么不直接给病人开止痛药呢？
Why don't we just give people the pain a pain medicine?
教授听了就说：你没明白。
And the professor was like, you don't understand.
这种痛苦本身就是药。
The pain is the medicine.
比如，当你产生想吸毒的冲动，去寻找毒品，结果却经历巨大的痛苦，这正是大脑重新学习、摆脱毒瘾的方式。
Like, experiencing desire to pursue drugs, drug seeking behavior, and then having it be immensely painful is the way you reprogram your brain to, like, overcome the drug seeking behavior.
如果你把痛苦消除了，那这个人就永远不可能真正戒掉。
And if you get rid of the pain, then the person is, like, never gonna do it.
所以这种痛苦其实是一种有建设性的痛苦。
There is so this is a productive form of pain.
顺便说一句，我认为AI领域现在发生的这些动荡——不只是在Meta，而是在每一家公司——就是我所说的这种痛苦。
By the way, I would argue AI, all these paroxysms happening not just at Meta, but at every company is the pain I'm talking about.
这种痛苦无路可逃，只能硬着头皮穿过它，你得自己摸索出一条路，弄清楚什么有效、什么无效，过程特别艰难。
That is the pain that there is no way out but through, and you have to figure out the path through it to figure out what works and what doesn't work, and it's just gritty.
我们社会里还有很多其他类型的痛苦，跟真正创造价值毫无关系。
We do have lots of other types of pain in our society that have nothing to do with real value being created.
这个问题经常被提到，教育就是一个很好的例子。
This this comes up a lot and education is a good example.
我记得上学的时候有人跟我说，我相信你们也听过：别用计算器。
I remember being told, I'm sure you were when I was in school, hey.
这次考试不准用计算器。
You can't use a calculator in this test.
你走出校门以后，日常生活中也不会随身带着计算器。
You will not have a calculator with you in your as you go about your day in the real world.
胡说八道。
Bullshit.
我身上至少带着三个计算器，更别说还能直接问我的眼镜数学题。
I have at least three calculators on my person at all times, not to mention I can just ask my glasses, math problems.
我简直被计算器包围了。
I got I'm filthy with calculators.
结果发现，不用计算器做数学题、做考试，是一种特定类型的痛苦，但没什么实际价值。
It turns out doing a math problem doing a math test without a calculator is a certain kind of pain, not a particularly useful kind.
真正有价值的，反而是用计算器来做那些需要深度思考的难题。
Doing a harder math test that requires critical thinking with a calculator is probably the more valuable way to do that thing.
对吧？
Right?
所以我确实觉得，我们经历的痛苦，应该和我们想创造的价值对得上号。
So I do think it's important to align the pain that we're experiencing with the value we're trying to create in the world.
我觉得，学会把AI融入学习，其实就能避开这种无谓的痛苦。
I think, like, learning to integrate AI, you could avoid that pain.
你直接跳过它。
You skip it.
你根本不做。
You don't do it.
你我都清楚，这会让你面临真正的风险。
You and I both know that it puts you at real risk.
你会落后于那些会用AI、又想做同样工作的人。
You're gonna fall behind people who, you know, are able to do AI and wanna do the same job as you.
你会落后于那些已经把AI融入运营或产品的公司。
You're gonna fall behind other companies that have integrated AI either economically or in the products that you offer.
你知道吗？谢丽尔·桑德伯格有句名言，说公司通常不是因为设定高目标却没达成而失败。
You know, there's this Cheryl always had this Cheryl Sandberg has this great quote, which is that companies don't usually fail by setting tough goals and missing them.
而是因为设了太低的目标，一路轻松达成，最后垮掉。
They fail by setting easy goals and hitting them all the way down.
所以我觉得，今天你完全可以轻松避开这种痛苦，只要说一声：没错。
And so, like, I think you could easily avoid the pain today by just being like, yeah.
我们就是不打算做，根本不会去做。
We're just not gonna we're just not gonna do it.
我们就任由它发生，等以后再说吧。
We're just gonna let it happen, and then we'll figure it out later on.
我觉得有建设性的痛苦，也有没必要的痛苦，关键是要能分辨清楚哪一种是哪一种。
So I think there is productive pain and unproductive pain and maybe a little bit of judgment to know which one's which.
博兹，每次和你聊天都真的很愉快。
Boz, it's really always a pleasure to speak with you.
非常感谢。
Thanks so much
谢谢你邀请我。
for coming on this Thanks for having me.
各位，非常感谢收听和观看，我们下一期《大科技播客》再见。
Everybody, thanks so much for listening and watching, and we'll see you next time on Big Technology Podcast.
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。