AI与工作的真正未来

你知道什么话题特别无聊吗？

You know what's a really boring conversation?

AI会不会抢走我们所有的工作？

Will AI take all of our jobs?

不幸的是，这正是AI行业长期以来一直想讨论的工作话题。

This, unfortunately, is the conversation about jobs that the AI industry has wanted to have for far too long.

但随着时间推移，我们终于开始更深入地思考，结果发现AI并不会把所有工作都抢走。

But finally, we are starting to get a little more thoughtful consideration as more time passes, and it turns out AI doesn't just take all the jobs.

AI真正做的是彻底改变我们的工作方式，无论是个人层面、团队层面，还是我们个人能追求的目标、团队能达成的目标，以及公司应该如何组织、我们应该优先培养哪些技能。

What AI does do is change the entire landscape of how we work on both individual levels, on team levels, in terms of what we can individually aspire to, in terms of what our teams can aspire to, in terms of how companies should organize themselves, in terms of what skills we should prioritize.

而所有这些才是关于AI和工作真正有趣且有建设性的话题，这正是我们今天要讨论的内容。

And all of those are the really interesting and productive conversations to have about AI and jobs, and that is exactly what we are talking about today.

《AI每日简报》是一个每日更新的播客和视频节目，专注于AI领域最重要的新闻和讨论。

The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

好了，朋友们。

Alright, friends.

在我们深入之前，先快速说几个通知。

Quick announcements before we dive in.

首先，感谢今天的赞助商：Blitzy、Robots and Pencils、Harbor 和 Hyperagent。

First of all, thank you to today's sponsors, Blitzy, Robots and Pencils, Harbor, and Hyperagent.

想要收听无广告版本，请访问 patreon.com/aidailybrief，或者你也可以在 Apple Podcasts 上订阅。

To get an ad free version of the show, go to patreon.com/aidailybrief, or you can subscribe on Apple Podcasts.

如果想了解如何赞助这个节目，请发送邮件到 sponsors@AIDailyBrief.AI。

And to learn more about sponsoring the show, send us a note at sponsors@AIDailyBrief.AI.

另外，请多关注 aidailybrief.ai。

Now keep an eye out in general at aidailybrief.ai.

网站上不仅有每集的完整摘要，还按关键引述、关键数据等分类整理，你还可以找到即将举行的活动信息，比如本周8月26日关于知识工作者代理循环的免费网络研讨会，以及更多关于即将推出的培训项目的信息。

In addition to the website having the full summaries of each episode broken into the key quotes, key numbers, etcetera, You can also find out about upcoming events like our free webinar coming up this week on August 26 about agentic loops for knowledge workers, and you'll also get a lot more information about upcoming training programs.

由 Newfar Gaspar 主持的下一期高管代理领导力项目将在劳动节后启动。

The next iteration of our executive agent leadership program led by Newfar Gaspar is kicking off after Labor Day.

再次提醒，所有这些信息都可以在 aideallybrief.ai 上找到。

Again, you can get all of that information at aideallybrief.ai.

今天的节目当然是一个周末长读兼深度思考的特别节目，而我们的朋友在Every那边正好有非常适合当下的深度思考内容。

Now today's episode is, of course, a weekend long read slash Big Think episode, and our friends over at Every have the perfect Big Think content for right now.

Every最近宣布了他们的第一场会议，叫做Thesis，同时还宣布了一个新项目叫Thesis Statements，这个项目会邀请100位建设者和思想家，就他们预见的AI带来的未来，写一些短篇文章。

Every recently announced their first conference called Thesis, and alongside it, they announced a new project called Thesis Statements that will have 100 builders and thinkers write short essays about the future that they envision coming with AI.

Every的CEO Dan Shipper是这么说的：我们相信，在自动化之后，人类工作仍然有光明的未来，而且我们相信有一小群人知道那是什么样的，因为他们每天都在实践这些答案。

Every CEO Dan Shipper put it this way: We believe there is a bright future for human work after automation, and we believe that there's a small group of humans who know what it looks like because they live the answers every day.

但他们的想法在很大程度上仍然缺席于关于AI的主流讨论。

But their ideas are still largely missing from the mainstream discourse about AI.

这就是为什么我们要创建一个公开记录，记录那些处于前沿的人现在看到的东西，这样我们就能尽可能多地把这些想法传播给更多人。

That's why we're creating a public record of what people at the frontier are seeing now, so we can get these ideas to as many people as possible.

如果你们想自己去看看，可以在every这个网站上找到。

For those of you who want to go check it out for yourself, you can find this at every.

Thesis Statements。

Tothesis statements.

我当然会在节目说明里放一个链接。

I will, of course, include a link in the show notes.

我们要做的是，看看这前25条陈述，读一读它们附带的短文章，然后当然，我会给出我的想法。

And what we're gonna do is look at a set of these first 25 statements, read the short essays that go along with them, and then, of course, I'll give my thoughts.

我们会把它们分成几个类别。

We're gonna organize them into a few categories.

第一个类别，我们称之为“基础”。

And the first category we'll call foundations.

而正是丹本人，我认为他通过他的短文章《自动化之后，人类工作将比以往更多》为这一切的核心论点做了精准的阐述。

And it is Dan himself who I think puts a fine point on the thesis that surrounds all of this with his short essay, After Automation, There Will Be More Human Work Than Ever.

丹写道：CEO、知识工作者和投资者似乎都同意一点，那就是AI对工作、经济、安全和人类意义构成威胁。

Dan writes: One thing CEO, knowledge workers, and investors seem to agree on is that AI is a threat to jobs, the economy, safety, and human meaning.

但如果你和AI行业里的人，或者行业外的早期采用者聊聊，你会听到和我们内部注意到的相同的事情：要做的工作比以往任何时候都多。

But if you talk to anyone in the AI industry or to early adopters outside of it, you'll hear the same thing we've noticed internally here at There's more work to do than ever.

我不认为会有一个临界点，事情突然翻转，工作就消失了。

I don't believe there will be a tipping point where things flip and the jobs are gone.

新的现实恰恰相反。

The new reality is the opposite.

我们越是自动化，需要人类专家做的工作就越多。

The more we automate, the more expert human work there is to do.

原因如下：AI把人类专业知识的残余部分商品化了，只要能被明确到可以训练的程度。

Here's why: AI commoditizes the residue of human expertise whatever can be made explicit enough to train on.

这降低了默认模型输出的价值，同时创造了对差异化内容的需求。

That collapses the value of default model output and creates demand for what's different.

对差异化内容的需求，就是对人类专家的需求，即使我们接近通用人工智能也是如此。

Demand for what's different is demand for human experts, even as we approach artificial general intelligence.

Every公司有近30人的团队，我们并没有为了智能体而解雇所有员工。

Every is a team of almost 30 people, and we haven't fired all of our employees in favor of agents.

我们也没有为了用“氛围编码”的应用而放弃软件即服务产品。

We haven't ditched software as a service products in favor of vibe coded apps.

我们仍然雇佣人类来做客户服务，同时有很多智能体助手，我们也仍然雇佣人类作家、编辑和工程师。

We still hire humans to do customer service with a lot of agent assistants and we still hire human writers and editors and engineers.

当然，员工智能体接管了更多稳定、重复、框架清晰的工作，但还有很多工作需要人类参与其中。

Sure, employee agents take over more of the stable, repeatable, well framed layer of work, but there is a lot of work that still requires a human being in the loop.

我们反复发现，对于任何复杂的任务，最好的方式就是让AI和人在同一个工作空间里来回协作。

We've found over and over that for any kind of complex task, the best way to get great work is to have an AI and a human going back and forth in the same workspace.

人仍然至关重要。

Humans are still vital.

在每个例子中，智能体都需要人来配合，工作才能真正完成。

In every example, the agent needs a human in order for the work to, well, work.

智能体离负责确保其正常运作的人越远，它的表现就越差。

The further away an agent gets from a human who is in charge of making sure it works well, the less well it works.

必须有人为它指明方向，判断输出是否合格，找出错误的地方，并把结果转化为现实中的决策或流程。

Someone has to point it at the right thing, decide whether the output is good, catch the places where it is wrong, and turn the result into a real life decision or process.

这是因为当前这一代模型只了解已经完成的工作。

That's because the current generation of models only knows about work that has been done.

而人知道现在需要做什么。

Humans know about what needs to be done, right now.

人能感知到特定的时间、客户、代码库和对话，而训练语料库还做不到这一点。

Humans are alive to a specific time, customer, codebase, and conversation in a way the training corpus isn't yet.

这种“活”的感觉不只是因为数据更新。

The aliveness isn't just having more current data.

我们带着自己的视角来到当下，这个视角是连续的、不断更新的，包含我们当下的欲望、担忧，以及对什么重要的判断，而这些都会影响我们看到的东西。

We come to the moment from somewhere, with a continuous, constantly new perspective of our own running wants, running concerns, and a running read on what matters, which changes what we see.

所以，让专家的工作变得更便宜，并不会简单地取代专家。

Making expert work cheaper does not, therefore, simply replace experts.

反而会创造更多需要专家判断的场景。

It creates more situations where expert judgment is needed.

丹的这个项目开头很棒，但在评论这一部分之前，我想再补充一个。

A great start to this whole project from Dan, but before I comment on this particular section, I actually want to pile on one more.

这个来自前战略顾问兼作家保罗·米勒德，标题是《宣称工作已被解决的人，自己还在工作》。

This one is from former strategy consultant and author Paul Millard, and is called The People Declaring Work Is Solved Will Still Be Working.

保罗写道：除非我们能大幅扩展现代对工作的定义，否则自动化之后并没有所谓的“之后”。

Paul writes: There is no after automation unless we can radically expand our modern conception of work.

目前，我们被困在我所说的“工作形态的现实扭曲场”里。

For now, we are stuck in what I call a job shaped reality distortion field.

我们把所有对工作的理解都压缩成了有职位描述和薪水的工作。

We flatten any conception of work down to activities with a job description and a paycheck.

在这种现实里，我们把人类变成数字，谈论裁员、就业率、薄弱环节、任务包和劳动力这些抽象话题。

In this reality, we turn humans into numbers and talk about abstract topics like layoffs, employment rates, weak links, bundles of tasks, and lumps of labor.

工作远不止这些。

Work is much bigger than this.

这个月我一直很忙：凌晨四点安抚新生儿，送三岁的孩子上学，在家做饭，回邮件，还要挤出一点时间写作。尽管有人说我浪费才华、没有工作，但我的日子充满了工作，我感到很充实。

This month, I have been busy: soothing my newborn at 4AM, transporting my three year old to school, cooking at home, responding to emails, and carving out a little time for writing Despite accusations of wasting my talents and being unemployed, my days are full of work, and I am fulfilled.

与此同时，我们听到硅谷23岁的年轻人高调宣称工作已经解决了。

At the same time, we hear loud pronouncements from 23 year olds in Silicon Valley, declaring that work is solved.

两年，也许三年，就没有什么可做的了。

In two years, maybe three, there will be nothing left to do.

但那时谁来给我女儿换尿布呢？

But then who will change my daughter's diaper?

这些宣言让我觉得好笑的是，它们来自那些沉迷于工作的人，在通宵编码之间大谈哲学。

The thing that makes me laugh about these pronouncements is that they come from people drunk on work, waxing philosophical between all night coding sessions.

但如果你问他们，他们的生活是不是围绕工作来安排的，他们会说不是，因为他们的工作是可选的。

But if you ask them if their lives are designed around work, they'd say no because their job is optional.

他们可以随意跳槽或者休假，同时还能得到人类渴望的东西。

They can job hop or take sabbaticals, all while getting the things humans crave.

比如尊严、挑战、目标感，这些都来自他们生活的环境。

Things like dignity, challenge, purpose from the lives they inhabit.

而他们的生活是围绕工作展开的。

Lives centered around work.

那些没有这种生活的人就感到困惑了。

People without such lives are mystified.

你说工作要结束了是什么意思？

What do you mean work is ending?

对他们来说，工作不是可选的。

To them, a job is not optional.

工作不会因为你离开办公室就结束。

The work doesn't end when you leave the office.

我们还没有足够的语言来描述我们即将前往的地方。

We don't yet have language big enough for where we are going.

我们生活的时代，人们更容易想象世界末日，而不是想象一个空闲的周二下午。

We live in an era in which people have an easier time imagining the end of the world than life with a free Tuesday afternoon.

我们需要摘下工作的眼镜，看到我们的生活已经充满了工作。

We need to take off our job goggles and see that our lives are already full of work.

如果我们花时间去看工作，不仅仅是以工作的形式，我们可能会找到前进的方向。

If we took the time to look at work, not just in job shaped form, we might find our way forward.

所以在这两篇文章中，你有两个非常不同的终点，却朝着同一个中间点努力。

So in these two essays, you have two very different ends working towards the same middle.

那就是总会有更多的工作。

The idea that there will always be more work.

对于你刚才听到的保罗来说，这体现在我们生活的实质中充满了各种工作，即使不是那种带着笔记本电脑和领带的工作。

Now, for Paul, who you just heard, this is in the fact that the substance of our life is full of all sorts of work, even if it's not the kind that comes with a laptop and a tie.

丹则更直接一些，他用的是我们大多数人理解的工作，也就是我们为了工作而做的工作。

Dan is a little bit more on the nose, and using work in the way that most of us think about work the work that we do for our jobs.

我自己是坚定地站在“工作只会越来越多”这一边的。

I myself am firmly in the There Will Be More Work camp.

这不仅是因为我觉得丹说得对，我们应该把对工作内容和意义的看法扩展到办公室之外，而且即使就在办公室这个范围内，我也认为人工智能会打开全新的、广阔的天地。

And not only because I think Dan is right that we should expand our perspective on the substance and meaning of what we do beyond the office walls, but because literally even in the context of those office walls, I just think AI unlocks radical new horizons to traverse.

就好像我们一直在一间大得可能超乎想象的大房间里，只在一个小小的亮着灯的角落里活动。

It's as though we've been playing in one tiny lit quarter of what seems like it might be a huge room.

我们一直待在老地方，因为只有那里是亮的。

We stay over where we've always been because that's the only part that's lit up.

但突然有人进来，把灯全打开了，房间的实际大小比我们原来以为的要大得多。

But suddenly, someone comes in and throws the lights, and the size of the room is actually monumental compared to what we had thought.

人类的本质就是会不断去发现、去做、去建设、去创造更多的东西。

The reality of humanity is that we will race to discover and do more and build more and create more.

所以我觉得丹说还需要人类专业能力，这其实还是低估了。

And so I even think that Dan is underselling it when he just points to the continued need for human expertise.

不是说他的观点有错，而是说人类工作会空前增多的更根本原因，是我们的能力范围会得到极大的扩展。

Not that any of his points are wrong, but that the larger reason there will be more human work than ever is because of a radical expansion of what we can all do.

那我们现在来聊聊公司和企业的运作方式吧。

But let's now transition into some specific discourse around companies and how businesses will work.

第一篇是诺亚·布雷尔写的，他是Aliphic的联合创始人，他认为软件公司会比软件工厂表现更好。

The first essay comes from Noah Breyer, co founder of Aliphic, who argues that software companies will outperform software factories.

诺亚说，在智能体工程中，最隐蔽的失败模式不是代码有bug。

Noah says, The most insidious failure mode in Agentic Engineering isn't buggy code.

而是智能体构建了根本不符合目标的功能、产品和系统。

It's agents building fundamentally misaligned features, products, and systems.

解决这个问题比减少缺陷更大也更有趣。

Solving that is a bigger and more interesting problem than reducing defects.

这就是为什么我觉得软件工厂这个比喻对AI工程来说并不合适。

This is why I think the software factory is the wrong metaphor for AI engineering.

挑战不在于如何用六西格玛质量每次冲压出同样的门板，而在于如何让系统与我们的愿景、价值观和架构保持一致地演进。

The challenge is less how to stamp out the same door panel every time with six sigma quality than how to evolve a system in line with our vision, values, and architecture.

从这个角度看，这个过程更像安迪·沃霍尔的工厂，而不是福特的汽车工厂。

In that sense, the process is closer to Andy Warhol's factory than Ford's car factory.

两者都注重产出，但沃霍尔更关心的是确保所有作品都符合一个统一的创意愿景。

Both are focused on throughput, but Warhol was more concerned with ensuring all work aligned with a single creative vision.

对企业来说，最难的还是提出一个愿景，并让整个团队朝着这个愿景努力。正如我在智能体出现之前很久就学到的，实现这一点更像是在创办一家初创公司，而不是在组装汽车。

The hardest problem for a business is still creating a vision and keeping an entire team of humans building toward it As I learned, long before agents existed, achieving this is much more like building a startup than assembling a car.

这个行业有太多人把软件当作一个需要优化和解决的问题。

Too much of the industry treats software as a problem to be optimized and solved.

对于写代码和测试来说，这也许是对的，但更好的比喻其实就在眼前。

That may be true for code writing and testing, but the better metaphor is staring us in the face.

这是软件公司，不是软件工厂。

It's a software company, not a software factory.

工厂只是更大组织的一部分，其中层层相互依赖的系统相互作用，并以不同的速度运转。

A factory is one piece of a larger organization, where layers of interdependent systems interact and move at different speeds.

但公司，一直以来，都是一群智能体聚集在一起，共同构建某种东西。

But a company is, and always has been, a collection of agents brought together to collectively build something.

理解这些智能体的动机，是打造一家成功的软件工厂的关键，而这恰好也是软件公司CEO的工作。

Understanding the motivations of those agents is what it takes to build a successful software factory, and conveniently, exactly the job of a software company CEO.

诺亚的文章在这里开始触及一个我认为企业界的人工智能用户比初创界的人工智能用户更懂的道理：没有人和机构体系围绕的技术，只是一堆流行词。

Noah's essay here starts to point to something that I think that the enterprise world of AI users actually understands better than the startup world of AI users: that technology, without human and institutional systems around it, is just another set of buzzwords.

不过，具体怎么建这些体系、怎么重新构想这些机构，以及为了什么目的，这才是问题所在。

Exactly how and to what ends we'd build those systems and reimagine those institutions, though, is the question.

这就引出了我们的下一篇文章，来自投资人兼作家蒂娜·博林，标题是《基础设施将胜出》。

And that brings us to our next post from investor and writer Tina Boring Infrastructure Will Win.

蒂娜写道：你的客户在凌晨两点决定停止使用你的客户关系管理软件。

Tina writes: Your customer decides to stop using your customer relationship management software at 2AM.

为什么？

Why?

销售代表意识到，你每年花三万美元的CRM合同，实际上只带来一万两千美元的价值。

The sales rep realized that your $30,000 annual contract for your CRM only gives you $12,000 in value.

没有会议，没有谈判——销售代表是个智能体。

There was no meeting, no negotiation the sales rep was an agent.

智能体没有忠诚度。

Agents are not loyal.

它们是理性的行动者。

They are rational actors.

它们能在几毫秒内查看数据，并在对业务有利时做出调整，哪怕是在半夜。

They look at the numbers in milliseconds and make changes whenever it makes sense for the business, even if it is in the middle of the night.

从某种意义上说，它们是无情的。

They are, in a way, ruthless.

你的软件可能易于使用、外观漂亮，但AI代理既看不见也不在乎。

Your software can be easy to use and look good, but AI agents neither see nor care.

一些公司会通过构建无头架构成为赢家，这些公司不会有任何人类用户，只有代理。

Some companies will become winners by building headless architecture Those companies won't have any human users, only agents.

其他表现好的企业则处于那些不能快速行动和犯错的领域，比如监管审批、银行和合规系统。

Other businesses that do well are in areas where you can't move fast and make mistakes, like regulatory approval, banking, and compliance systems.

这些公司竞争的不是最复杂，而是在诸如确保电汇安全送达这类事情上最高效。

These companies aren't competing to be the most sophisticated, but to be the most efficient at things like having wire transfers delivered securely.

在未来几年里，拥有这一层的公司将获得优势，能够收集关于谁该做什么以及何时做的情报。

Over the next few years, the companies that own this layer will have an edge in gathering the intelligence about who should do what and when.

随着模型能力趋于一致，开始做出类似的决策，公司之间的竞争将不再主要看谁的模型最好，而更多看谁的系统能把这些决策转化为实际成果。

As the capabilities of the models converge and start to make the same sorts of decisions, companies will compete less on having the best model, and more on the systems that connect those decisions to real world outcomes.

这些关键系统中有一些负责管理任务分配、数据访问、工作流编排和规则执行。

Some of these essential systems manage task routing, data access, workflow orchestration, and rule enforcement.

那些针对特定客户、特定用例提供服务的公司，会比那些泛泛的智能体应用更胜一筹。

Companies that are serving specific customers with specific use cases like these will win over broad Agentic use cases.

这些商业模式就像收费公路。

These business models are like toll roads.

如果你不想付费，就得自己建桥。

If you don't want to pay, you've got to build your own bridge.

而建桥可能要花上好几年，合规成本高达数亿。

And that could take years and cost hundreds of millions for compliance.

所以到头来，那些不那么光鲜的工作可能回报最丰厚。

That's why, in the end, the less glamorous work may be the most rewarding.

诺亚让我们看到公司会如何变化，而蒂娜则从外部视角审视这种变化，其中一个方向就是理性且无情的数字代理会更多地替公司行事，并围绕它们形成一整套全新的支撑体系。

So Noah gives us the idea that companies will change, and Tina takes a look from the outside in as one of the ways that they will change is rational and emotionless digital agents doing more of their bidding, creating a whole new set of support structure around them.

但这对于那些正在思考自己和团队的工作应该如何改变的领导者来说，意味着什么呢？

But what does this mean for the leaders who are trying to figure out how the work that they do and their teams do should change?

为此，我们请来了苏米特·辛格，他曾经是安德森·霍洛维茨基金的合伙人，现在是World Build的创始人和管理合伙人。他写道，那些仅仅用AI来优化现有工作流程的创始人将会失败。

For that, we turn to Sumit Singh, former Andreessen Horowitz partner and now managing partner and founder of World Build, who writes founders who AI f y existing workflows will lose.

过去八年，我作为投资人，一直在看着同样的模式反复出现。

I've spent the last eight years as an investor watching the same pattern repeat.

随着生成式AI的出现，那个时代已经结束了。

That era has ended with the advent of generative AI.

作为投资人，我感到很兴奋。

As an investor, I'm excited.

AI终于打开了真正创新的潜力，这种潜力自移动革命以来一直缺失。

AI has finally opened up the potential for real innovation that's been missing since the mobile revolution.

但我看到一些创始人正在构建专业化的AI产品，就好像他们在打造过去十年里同样的工具一样。

But I see founders building specialist AI products as if they were building the same tools of the last decade.

那些还在沿用旧框架的人，即将犯下大错。

Those who are playing by the old framework are about to make a big mistake.

在AI时代会失败的企业，是那些从现有工作流程出发，然后把它AI化的企业。

The businesses that will fail in the AI era are those that start with an existing workflow and AI ify it.

能活下来的企业，会利用模型独特、细腻的特性，去创造以前技术上不可能实现的新工作流程。

The ones that will survive will leverage models' unique, nuanced properties to invent new workflows that were not technically possible before.

我把这些称为后拟物化应用。

I call these post skeuomorphic apps.

拟物化是个陷阱，就是假设新技术应该模仿以前的样子。

Skeuomorphism is the trap of assuming that a new technology should look like what came before.

早期的移动应用经常掉进这个坑。

Early mobile apps constantly fell into this pattern.

它们复制物理世界，比如垃圾桶图标看起来就像真的垃圾桶。

They replicated the physical world, like the trash can icon that looked like an actual garbage bin.

但它们没有去探索手机能做的独特事情。

But they weren't exploring what our phones could uniquely do.

那些突破性的应用，完全打破了这种陷阱。

The apps that broke through also broke this trap entirely.

优步并没有把出租车调度员的办公桌数字化。

Uber didn't digitize the taxi dispatcher's desk.

它思考的是，当每个人口袋里都有一部知道他们位置的手机时，会有什么新的可能。

It asked what becomes possible when everyone has a phone in their pocket that knows where they are.

手机成了你生活的遥控器，正如投资人马特·科勒所说，点餐用DoorDash，出行用优步，买杂货用Instacart。

The phone became a remote control for your life, as investor Matt Kohler has said, for food, DoorDash, for rides, Uber, for groceries, Instacart.

他们没有去改造现有的工作流程，而是创造了全新的流程。

They didn't adapt existing workflows, they invented new ones.

人工智能正处于完全相同的转折点。

AI is at the exact same inflection point.

最终胜出的创业者们问的是另一个问题：现在有什么新的可能？

The founders who will win are asking a different question: What becomes possible now?

我们能发明什么样的工作，只有人工智能才能实现？

What work can we invent that only AI makes possible?

胜出的应用将发现新的工作流程，而我们甚至还不知道这些流程长什么样。

The winning applications will discover new workflows, and we don't even know what these workflows look like yet.

市面上所有AI编程工具做的第一件事都一样。

Every AI coding tool on the market does the same thing first.

就是开始写代码。

It starts writing code.

但Blitzy反其道而行。

Blitzy does the opposite.

在写一行代码之前，Blitzy会花好几天时间逆向分析你的整个代码库。

Before writing a single line, Blitzy spends days reverse engineering your entire codebase.

成千上万个代理读取数百万行代码，梳理每一个依赖关系、每一个未记录的约束条件，以及过去十年里做出的每一个架构决策。

Thousands of agents ingest millions of lines, mapping every dependency, every undocumented constraint, every architectural decision made over the last decade.

最终形成一个动态知识图谱，就像一位在系统里摸爬滚打三十年的首席工程师那样，对你的软件了如指掌。

The result is a dynamic knowledge graph that understands your software the way a principal engineer would after thirty years in the building.

其他工具靠grep搜索和markdown文件来猜测上下文。

Other tools guess at context with grep searches and markdown files.

Blitzy从不猜测。

Blitzy never guesses.

它先建立真正的理解，然后自主交付超过80%的完整软件史诗。

It builds true understanding first, then delivers over 80% of entire software epics autonomously.

经过验证、端到端测试、生产级质量的拉取请求。

Validated, end to end tested, production grade pull requests.

这就是为什么财富500强的工程团队把最重要的代码库托付给Blitzy。

That's why Fortune 500 engineering teams trust Blitzy with the code bases that matter most.

你自己去blitzy.com看看吧。

See for yourself at blitzy.com.

就是blitzy.com。

That's blitzy.com.

我经常在企业AI公司里看到一种现象：他们在每个云、每个模型、每个框架上都做对冲，或者花钱请GSI做一个永远结束不了的试点项目。

One thing I keep seeing in enterprise AI companies hedging across every cloud, every model, every framework or paying a GSI for a pilot that never ends.

真正在交付的团队，他们选了一条路，然后快速前进。

The team's actually shipping, they've picked a lane and they move fast.

这也是我喜欢今天赞助商Robots and Pencils的原因之一。

That's one of the reasons I like today's sponsor Robots and Pencils.

他们全力投入了AWS。

They've gone all in on AWS.

他们是AWS的高级合作伙伴，并且在四十五天内交付了生产级AI同事。

They're an advanced tier and AWS pattern partner and they ship production AI coworkers in forty five days.

这让他们在AI同事方面做出了一些我见过的最有趣的工作。

That's led to them doing some of the more interesting work I've seen on AI coworkers.

我说的不是聊天机器人，而是真正嵌入企业架构、能实际干活的智能体系统。

And by that I'm not talking about chatbots, I'm talking about actual Agentic systems that sit inside a business architecture and do real work.

这种专注很重要，无论你是想真正落地项目的企业领导者，还是想把客户从感兴趣变成实际部署的AWS代表。

That kind of focus matters if you're an enterprise leader trying to get something real into production, or an AWS rep trying to move a customer from interested to deployed.

在robotsandpencils.com上申请AI简报。

Request an AI briefing at robotsandpencils.com.

和robotsandpencils聊一次，你就明白了。

One conversation with robotsandpencils and you'll know.

每期节目，我们都会讨论OpenAI、Anthropic、SpaceX AI、Google和Meta之间的竞争。

Every episode, we cover the competition between OpenAI, Anthropic, SpaceX AI, Google, and Meta.

很可能你已经对谁领先有了自己的看法。

Chances are, you've already formed an opinion about who's leading.

但每个AI实验室都在采取不同的方法，构建不同的技术，建立不同的合作伙伴关系，并发展独特的生态系统。

But every AI Lab is taking a different approach, building different technologies, forging different partnerships, and developing a unique ecosystem.

Harbor Capital Advisors的AI实验室生态系统ETF系列为投资者提供了一种方式，让他们能够接触到他们认为最有可能成功的AI生态系统。

Harbor Capital Advisors' AI Lab Ecosystem ETF Suite gives investors a way to gain exposure to the AI ecosystem they believe is best positioned for success.

无论你在哪里投资，都可以搜索Harbor AI实验室生态系统ETF，或者关注HarborCapital的X账号了解更多信息。

Search Harbor AI Lab Ecosystem ETFs wherever you invest, or follow HarborCapital on X to learn more.

请访问harborcapital.com获取包含投资目标、风险、费用、支出及其他重要信息的招股说明书。

Visit harborcapital.com for a prospectus containing investment objectives, risks, fees, expenses, and other important information.

投资前请仔细阅读并考虑这些信息。

Read and consider it carefully before investing.

风险包括本金损失和与人工智能相关的风险。

Risks include principal loss and artificial intelligence related risks.

Harbor ETFs由Foreside Fund Services LLC分销。

Harbor ETFs are distributed by Foreside Fund Services LLC.

Harbor与AI Daily Brief没有关联，这些基金也不隶属于任何AI实验室，也没有得到它们的赞助或认可。

Harbor is not affiliated with AI Daily Brief, and the funds are not affiliated with, sponsored by, or endorsed by any AI lab.

这是一则付费广告，并非个性化投资建议。

This is a paid advertisement and not personalized investment advice.

投资有风险，包括可能损失本金。

Investing involves risk, including possible loss of principal.

本期AI Daily Brief由Hyperagent赞助播出，Hyperagent让你可以运行多组智能体，你的团队可以一起管理它们。

This episode of the AI Daily Brief is brought to you by Hyperagent, where you run fleets of agents your team can manage together.

新用户可获得1000美元的推理额度。

New users get $1,000 in inference.

别再依赖本地智能体和等待笔记本电脑提示的聊天工作流了。

Forget local agents and chat workflows waiting on your laptop to be prompted.

Hyperagent在云端部署始终在线的智能体，在你团队已经使用的工具中完成实际工作。

Hyperagent deploys always on agents in the cloud, doing real work across the tools your team already uses.

营销部门的智能体能把竞争对手的动向转化为落地页。

Marketing's agent turns competitor moves into landing pages.

销售部门的智能体负责丰富潜在客户信息、起草邮件，并更新客户关系管理系统。

Sales's agent enriches leads, drafts emails, and updates the CRM.

运营智能体负责跟进文书工作，并跟踪预算。

Ops agent chases the paperwork and tracks the budget.

每个智能体都能访问共享上下文，并遵循你关于权限范围和审批的规定。

Every agent has access to shared context and follows your rules about scope and approvals.

是时候引入一些像团队成员一样的智能体了。

It's time you add agents that feel like teammates.

去Hyperagent雇佣你的智能体吧，它由Airtable团队打造。

Hire yours at Hyperagent built by the team at Airtable.

在hyperagent.com/aidailybrief领取你的1000美元推理额度。

Claim your $1,000 in inference at hyperagent.com/aidailybrief.

现在，Sumeet，考虑到他在a16z的经历，他自然是从初创公司及其所构建内容的角度来看待这个问题。

Now Sumeet, understandably given that he's coming from his experience at a16z, is looking at this from the standpoint of startups and what they build.

但我认为，对于正在摸索AI在哪里能发挥最大价值的传统企业来说，这一点同样成立，甚至更加成立。

But I think that this is every bit as true, if not more true, for legacy enterprises who are in the midst of figuring out where AI can provide the most value.

我经常把人工智能分成两类：效率型AI和机会型AI。

I have often broken this into two different categories of AI, efficiency AI and opportunity AI.

需要说明的是，效率型AI本身并没有什么问题。

There is, to be clear, nothing wrong with Efficiency AI.

把需要做的事情做得更快、更便宜、更好，这是好事，也是一个非常合理的起点。

Doing the things that you need to do faster, cheaper, better is a good thing, and it is a perfectly reasonable place to start.

然而，如果认为这种极其强大、能解锁新能力的技术，只会被用来以完全相同的方式做事，只不过把人类换成智能体，那就从根本上误解了它的潜力。

However, to assume that this incredibly powerful, capability unlocking technology will be constrained to doing things the exact same way that they've always been done, just by an agent instead of a human, is to fundamentally misimagine what is possible.

机会型AI的核心在于，去问那些以前因为有了AI而能去探索的新方向。

Opportunity AI is all about asking those questions about where you can go that you couldn't go before because you have AI in your corner.

要找出机会型AI对你特定业务的最大机会在哪里，将会困难得多，也需要更多的迭代和实验。

It is going to be much more difficult and take much more iteration and experimentation to figure out where the greatest opportunities of opportunity AI actually lie for your particular business.

这就是为什么我一直强调，要在企业内部为实验创造空间，不要过早地限制或过度强调AI投入的ROI，因为那样做会让人们偏向那些只是老一套、只是更便宜或更快的应用。

This is why I'm constantly beating the drum of needing to create space for experimentation and not prematurely cut off or overly ROI ify your AI efforts inside the enterprise, as to do so will be to bias people towards those applications which are just the same old things, but a little cheaper or faster.

这个问题的另一个维度我谈得不多，但这也是为什么我总体上对那种只观察人类行为、然后训练智能体去重复同样事情的初创公司持怀疑态度。

There is another dimension of this that I haven't spoken about as much, but is why I'm a little skeptical in aggregate of startups whose only job is to watch what people do to train agents to do the same thing.

我一直觉得，说智能体会像我们一样做同样的工作，这种想法本身就不太合理。

The idea that agents are going to do the same work that we do in the same way has never struck me as making a lot of sense.

智能体几乎必然会用它们自己特有的方式去做事情。

Agents are almost inevitably going to do things in different agent native ways.

这样可能会更高效，也更符合它们的优势，但同时也需要更高层次的组织整合，确保这些新的智能体流程能和人类同步。

Presumably that will be more efficient and native to their strengths, but which will also require another level of organizational integration to make sure that those new agentic processes can be synced with humans as well.

不过，我觉得苏米特那个更宏观的观点很重要，就是说我们不能认为新世界只是旧世界的高效版，这个观点非常非常重要。

Still, I think Sumeet's broader pointthat we can't think that the new world is just gonna look like the old but more efficient is a very, very important one.

作家汤姆·克里奇洛在他的论文陈述里开始探讨这一点，他说，拥有最好时钟的公司会打败拥有最好模型的公司。

Writer Tom Crichlow starts to explore this a little bit in his contribution to the thesis statements, the company with the best clock will beat the company with the best model.

汤姆说：在19世纪之前，没有一个全球统一的时间标准。

Tom says: Before the 1800s, there was no universally agreed upon time.

但随着火车速度越来越快，每个车站统一时间就变得更重要了。

But as trains got faster, it became more important for each station to agree on what time it was.

否则，伦敦和布里斯托的“正午”含义不同，你坐车就总会不是早到就是晚点。

Noon couldn't mean two different things in London and Bristol otherwise, you'd always be late or early for your ride.

因此，铁路系统统一了时钟，结果就是今天仍然存在的标准时间和时区。

The railway system therefore synced the clocks, and the result was the standard time and time zones that still exist today.

现在，我们需要再次调整时钟。

Now, we need to change the clocks again.

人工智能带来了更快的思考速度，但我们还没有协调好工作流程来跟上这个速度。

AI has introduced faster thinking, but we haven't yet coordinated our workflows to keep pace.

智能体以秒为单位运作。

Agents operate in seconds.

团队每周开会。

Teams meet weekly.

财务部门按季度规划。

Finance plans quarterly.

领导层每年重新审视战略。

Leadership revisits strategy annually.

组织的每个部分都处于不同的当下，每个当下都基于不同的现实版本。

Each part of the organization inhabits a different present, each made using a different version of reality.

我们需要一个新的标准时间，一个适用于AI时代的协调机制。

We need a new standard time, a new coordination mechanism for the age of AI.

就叫它标准状态吧。

Call it standard status.

这将是一个持续更新的记录，包含目标、决策、权限和约束，人类和智能体都能共享。

This will be a continuously updated record of goals, decisions, permissions, and constraints, shared by humans and agents alike.

这听起来像是个工程问题。

This sounds like an engineering problem.

吸收上下文，生成状态，并不断更新，让员工保持同步。

Ingest the context, generate a status, and continually update it so that workers can stay in sync.

但保持同步不等于保持对齐。

But staying in sync is not the same thing as staying aligned.

下一个挑战是应对我们工作方式中固有的模糊性。

The next challenge is to grapple with the ambiguity inherent in the way we work.

团队会议、财务计划和年度规划一直带有仪式感，它们被设计成让个人和团队保持对齐、被倾听和有动力的形式。

Team meetings, financial plans, and annual planning were always part ritual, designed as formats to keep individuals and teams aligned, heard, and motivated.

我们能不能放下这些形式，改用标准状态呢？

Can we just let go of this theater in favor of standard status?

还是说，我们仍然需要保留人类那种定期沟通的流程？

Or will we need to continue the human process of just checking in?

在协调一致方面，拥有最佳时钟的公司可能很快就会胜过拥有最佳模型的公司。

When it comes to alignment, the company with the best clock may soon beat the company with the best model.

但前提是，它要明白知道时间并不等于知道当下需要什么。

But only if it remembers that knowing the time is not the same as knowing what the moment requires.

在这个方面，今年秋天我会重点探索的一个方向，就是从单人AI转向多人AI，从个体智能体转向团队智能体。

Now on this front, one of the things that I am going to be exploring a lot this fall is a shift from single player AI to multiplayer AI, from individual agents to team agents.

我觉得很多关于协调的方面，以及围绕新型协调设计的新系统，都会成为这些对话的重要内容。

And I think a lot of these aspects of coordination and new systems designed around a new type of coordination are going to be the substance of some of those conversations.

到目前为止，我们只谈了组织层面，那个人呢？

So far, we've been talking just about the organization, but what about the individual?

尽管这些人大多都确信AI之后还会有工作存在，但其中很多人认为这些工作会有所不同。

Even though these folks are pretty much all in the group being quite sure that there will be jobs that remain after AI, many of them think that those jobs will look somewhat different.

Oboe的首席执行官Nir Zickerman写道，AI无法胜任的工作将会规模化。

Oboe CEO, Nir Zickerman, writes that jobs AI can't do will scale.

技术革命总是会引发同样的循环。

Technological revolutions always trigger the same cycles.

工作从容易自动化的岗位转向不容易自动化的岗位，并释放出前所未有的生产力。

Jobs shift from ones that are easy to automate to ones that aren't, and unprecedented productivity is unlocked.

AI也会带来同样的转变。

We'll see the same shift with AI.

与具体且可验证的任务相关的工作会明显减少，但AI做不好、却能在AI帮助下以前所未有的规模完成的事情，也会显著增加。

There will be an obvious reduction in jobs related to concrete and verifiable tasks, but there will also be a significant rise in things that AI cannot do well but which can now be done at unprecedented scale with AI.

大型语言模型在接管具体且易于验证的任务方面表现出色，这些工作将会消失。

LLMs have proven exceptionally good at taking over concrete tasks that are easily verifiable These jobs will fade away.

但由于结构性的缺陷，大型语言模型在处理需要模糊性、开放性和创造性的任务时能力相当有限。

But due to structural flaws, LLMs are quite limited in their ability to perform tasks that require ambiguity, open endedness, and creativity.

这些工作将会蓬勃发展。

These are the jobs that will thrive.

创意产业会起飞，就像数码相机、数字音频工作站、互联网和社交媒体让我们都变成了创作者一样。

Creative enterprises will take off for the same reason that digital cameras, digital audio workstations, the internet, and social media turned us all into creators.

需要组织管理和沟通的岗位会比以往任何时候都更重要，需要人际互动的岗位也是如此。

Roles requiring organizational management and communication will become more important than ever, as will those that require human interaction.

系统层面的岗位也会蓬勃发展，因为监督大量智能体的需求在增加。

Systems level roles will also flourish as the need to oversee droves of agents rises.

就拿电影行业来说吧。

Take the film industry.

最近，视频生成模型的兴起引发了对电影制作持久性的猜测。

The rise of video generation models has recently stirred up speculation about the durability of movie making.

但是，由人类编剧创作故事、由人类演员表演，这一点不会改变。

But human actors acting out stories created by human screenwriters aren't going anywhere.

这些是只有人类才能做好的创意工作。

Those are creative enterprises that only humans can do well.

然而，电影制作中的其他所有环节，从融资到选角再到后勤管理，都会被简化到前所未有的程度。

Yet all the other stuff that goes into a movie, from financing to casting calls to managing logistics, will be streamlined to levels never before seen.

这将让艺术家们敢于承担更多创作风险，去做那些以前在技术上和经济上都不可行的事情。

This will allow artists to take more creative risk and to do things that were previously neither technically nor economically feasible.

这需要几十年的时间，但最终就业市场上那些琐碎忙碌的工作会大幅减少，取而代之的是更多发挥人类独特能力的工作。

It will take several decades, but job markets will eventually have far fewer roles made up of busywork and more roles made up of capabilities that humans uniquely possess.

我喜欢这个系列文章的一点是，我对AI行业的未来预测一直有个不满，就是他们历来不愿意深入具体细节。

Now one of the things that I like about this whole thesis series is that one of my beefs with the future projections from the AI industry has been a historic unwillingness to get into the specifics.

这正是我在今年早些时候那期节目《AI将创造的新工作》中试图解决的问题。

That's what I tried to address with an episode from earlier this year called The New Jobs AI Will Create.

无论你是否同意这些分析，我很高兴看到很多作者都在深入探讨这个具体问题。

Whether you agree or not with the analysis, I'm glad to see a lot of these authors digging into that exact question.

另一位这样做的作者是乔·哈德森，他是《成就的艺术》的创始人。

Another who does that is Joe Hudson, founder of The Art of Accomplishment.

他写道，智慧工作将取代知识工作。

He writes that wisdom work will replace knowledge work.

乔说，在AI出现之前，知识让你脱颖而出。

Before AI, says Joe, knowledge set you apart.

知道得越多，赚得就越多。

Knowing more meant earning more.

但随着模型一夜之间吞掉整个领域，像情绪清晰度、辨别力和人际连接这样的智慧技能，才是让你不可或缺的东西。

But as models swallow entire fields overnight, wisdom skills like emotional clarity, discernment, and connection is what keeps you indispensable.

AI模型不会睡觉，也不会倦怠。

AI models don't sleep or burn out.

很快，一个训练有素的模型就能在任何时间同时超越物理学、法律和工程学领域的专家。

One highly trained model will soon be able to outperform an expert in physics, law, and engineering simultaneously at any hour.

想象一个世界，你所有的知识都变得无关紧要，就像今天会生火一样。

Imagine a world where all your knowledge is irrelevant, akin to the ability to build a fire today.

偶尔有用，但在有电灯、中央供暖和炉灶的世界里，大部分时候都不需要。

Occasionally useful, but mostly unnecessary in a world with light bulbs, central heating, and stove tops.

AI还会让那些才华横溢的人更难逃避文化破坏性行为的后果。

AI will also make it harder for brilliant people to get away with culturally destructive behavior.

几十年来，非凡的知识或技能为难以相处的同事筑起了一道护城河。

For decades, extraordinary knowledge or skill created a protective moat around difficult colleagues.

人们嘀咕着“他们就是这样”，然后继续维持和平。

People muttered That's just how they are and kept the peace.

但当模型能在几秒钟内起草简报、诊断异常或优化市场策略时，为什么还要继续为才华横溢的混蛋付出情感代价呢？

But when a model can draft the brief, diagnose the anomaly, or optimize the market strategy in seconds why keep paying the emotional tax of a brilliant jerk?

但你不必是个才华横溢的吹牛大王，你的技能才面临被AI颠覆的风险。

But you don't have to be a talented blowhard for your skills to be at risk of AI disruption.

杠杆已经从你能做什么转移到了你做事时的表现方式。

The leverage has shifted from what you can do to how you show up while doing it.

当知识不再稀缺时，什么仍然有价值？

When knowledge is no longer scarce, what remains valuable.

智慧。

Wisdom.

智慧是如何生活。

Wisdom is how to live.

它是错误经过时间和反思消化后的残留物。

It is the residue of mistakes metabolized by time and reflection.

这不能急于求成，也不能直接复制粘贴。

It can't be rushed, and it can't be copy pasted.

这是一种切身体验，来自内心的指引。

It is an embodied, as in felt in the body, experience guidance from the inside.

无论AI变得多聪明，它都不能替你过你的人生。

No matter how intelligent AI becomes, it can't live your life for you.

它无法在高风险谈判中感受到你身体发出的信号，无法察觉董事会上隐藏的恐惧，也无法听出客户礼貌言辞背后未说出口的拒绝。

It can't feel your body's signal in a high stakes negotiation, sense the hidden fear in a boardroom, or hear the unspoken no behind a client's polite words.

这就是为什么未来的经济会重视智慧型工作者。

That's why tomorrow's economy will prize wisdom workers.

你可以从AI那里得到答案，但如何运用这些答案需要智慧。

You can get answers from AI, but how you use those answers takes wisdom.

Built First的创始人Bethany Crystal认为，这不仅仅是智慧的问题。

Built First founder Bethany Crystal thinks it goes beyond just wisdom.

她说，独特之处将成为人类最大的优势。

Weirdness, she says, will be the best human advantage.

贝瑟尼写道，回到高中时期，我只以我的外号“拉杆箱女孩”为人所知。

Bethany writes, back in high school, I was known only by my appointed nickname, wheelie bag girl.

作为唯一一个背着带轮子书包的孩子，我被无情地嘲笑，经常独自一人，要么重新按字母顺序整理我的名片收藏，要么给我的篮球大小的球加橡皮筋。

As the only kid with a rolling backpack, I was made fun of mercilessly and spent a lot of time alone, re alphabetizing my business card collection or adding rubber bands to my basketball sized ball.

我并不介意。

I didn't mind.

作为早期互联网的孩子，古怪和特别才是常态。

As a child of the early Internet, being strange and peculiar was the standard.

后来网络长大了，自我优化了，也要求我们所有人都这样做。

Then the web grew up, optimized itself, and asked us all to do the same.

我听从了。

I listened.

我变得正常了。

I got normal.

然后一切又变了，这次是AI站在了最前沿。

And then it all changed again, this time with AI at the forefront.

如今，我一个人经营一家公司，这在AI出现之前需要五个人才能完成。

Today, I run a company solo that would have required five humans in the pre AI age.

但最让我惊讶的不是AI让我更高效了，而是它让我又变得古怪了。

But the biggest surprise isn't that AI has made me more productive, it's that it's made me weird again.

AI重新带来了那种好玩、实验性的工作方式。

AI has reintroduced a playful, experimental way of working.

朝九晚五的工作正在被拆分成一系列以项目为基础的兴趣，重点在于手艺和热情。

The nine to five is becoming decoupled into a series of project based interests with a focus on craft and passion.

我在为朋友们做宾果卡应用，为我的客户定制超个性化的播放列表。

I'm building bingo card apps for my friends and hyper personalized playlists for my clients.

我重新启动了我的博客，把它变成了一个“选择你自己的冒险”的体验。

I relaunched my blog as a choose your own adventure experience.

这没给我带来任何收入，但这是我多年来梦想做的事情。

It's not making me any money, but it's something I dreamed about doing for years.

回到学校那会儿，那些古怪的习惯和极端的爱好会让你在操场上被欺负。

Back in school, those quirky habits and extreme niches would get you pushed around the playground.

但在后AGI时代，这些才是让你保持人性的东西。

But in a post AGI world, they're the only things that keep you human.

如果你把这种回归怪癖看作AI带来的意外好处，那么这种更广泛的、可能意想不到的好处，也贯穿在其他几个论点中。

And if you view the return to weirdness as a, call it, unexpected benefit of AI, this idea of broader, perhaps unexpected benefits is one that runs throughout a number of the other thesis statements as well.

比如，Abstract Group的联合创始人艾米丽·弗农就认为，AI会激发人们对平庸创意的抵制。

Abstract Group co founder Emily Vernon, for example, argues that AI will spur a resistance to mediocre ideas.

艾米丽说，品牌这个概念其实没那么复杂。

Emily says, brand as a concept isn't that complex.

它就是一个好主意，只要持续执行，就会随着时间不断增值。

It's a good idea that compounds over time when executed consistently.

AI确实在帮助解决持续执行的部分，但好主意本身仍然很难得。

AI is certainly helping with the consistency part, but a good idea is still hard.

现在甚至更难了，因为AI让平庸的创意变得难以抗拒。

It's now even harder because AI makes mediocre ideas irresistible.

既然用几个快速提示就能免费生成一个过得去甚至还挺有品味的品牌，为什么还要花六位数、用六个月去打造自己的品牌呢？

Why spend 6 figures in six months building your brand when you can generate something passable, even tasteful, with a few quick prompts for free?

我们真正该关注的不是品味。

It's not taste that we should be tracing.

越界电影导演约翰·沃特斯呼吁大家挣脱“好品味”的束缚。

Transgressive filmmaker John Waters' call to arms is to break free of the tyranny of good taste.

如今，品味不错但毫无记忆点的品牌生成起来又便宜又容易，这种束缚比以往任何时候都更让人窒息。

Now that tasteful but forgettable brands are cheap and easy to generate, that tyranny is more oppressive than ever.

相反，我们必须变得不可预测。

Instead, we have to be unpredictable.

创意专业人士一直都知道，伟大的想法是不可预测的，神经科学也证实了这一点。

Creative professionals have always known that great ideas are unpredictable, and neuroscience confirms it.

大脑天生就会注意并记住那些出人意料的事物。

The brain is tuned to both notice and remember surprises.

它会把可预测的东西标记为容易遗忘的。

It registers predictable as forgettable.

既然AI的整个前提就是可预测，这正是我们该介入的地方。

As AI's whole premise is being predictable, this is where we fit in the loop.