为什么：代币

讨论让世界持续运转。

Discussion keeps the world turning.

这就是圆桌会。

Is Roundtable.

你正在收听圆桌会。

You're listening to Roundtable.

我是牛红林，今天和史蒂夫与玉顺一起主持。

I'm Niu Honglin, joined by Steve and Yushun.

接下来，我们很高兴为您带来全新的科技深度栏目《为什么》。

Coming up soon, we're happy to present our new tech centered segment, The Why.

在这个环节中，我们会选取一个故事，深入挖掘其背后的意义。

This is where we take one story and go beyond the surface.

在科技报道中，焦点通常放在数据、新品发布和突破性进展上。

In tech reporting, the focus is usually on the numbers, the launches and the breaking throughs.

但在《为什么》中，我们提出的问题旨在揭示更宏大的图景。

But in The Why, we're asking the question that reveals the bigger picture.

为什么这种趋势会存在？

Why does this trend exist at all?

现在，来探讨为什么。

Now, the why.

我们知道是谁、什么、何时和何地。

We know the who, what, when, and where.

但那为什么呢？

But what about the why?

为什么？

Why?

为什么？

Why?

为什么？

Why?

直接从是什么跳到怎么做。

What and then directly go to how.

我们想知道如何使用它。

We wanna know how to use it.

我们想知道它如何帮助我们的日常生活，但我们却忘了问为什么。

We wanna know how it will help us in our daily life, but we forget to ask why.

这一部分我们将花时间深入剖析那些新奇有趣，或早已存在于我们身边却从未停下脚步问过为什么的事物。

This is going to be a segment where we take the time to really break down and chew on things that are either new and intriguing or have been around us for quite a long time, yet we didn't really stop and ask why.

我们并没有真正理解其本质。

We didn't really understand the essence

感谢的意义。

of Thank you.

我正想说，虽然这不是我的环节，但我依然很期待，因为那些我不理解的事物。

I was just about to say, and also things that I it's not gonna be my segment, but still, I'm excited about it because things that I don't understand.

嗯。

Mhmm.

我们对这个话题已经好奇很久了。

We've been curious about this for quite a long time.

嗯，他们没问。

Well, they didn't ask.

每天新闻里都有这么多科技新闻，我们根本没有机会深入挖掘、解释，并用通俗易懂的方式说明它们对人们意味着什么。

There's so much tech stuff that comes out every single day in the news, and we don't have an opportunity to fully dig into it and explain it and break it down in layman's terms what it means for people.

如果世界上有和我处境相似的人，那么希望由像玉顺这样的科技达人来拆解这些内容，能对你有所帮助。

And if there are people out there who are in my shoes, then hopefully, this will be helpful to have a tech guy like Yushun break it down for you.

是的。

Yeah.

所以我真的非常兴奋，这个项目或这个板块终于落地了，因为我们已经准备了很久，而正如你所说，我们终于到了那个时刻——之前一直没时间、也没精力停下来思考‘为什么’？

So I am really, really excited that this project or this segment has really got landed because we've been preparing for a while, and then it is because we kind of, as you said, getting to that kind of moment that we just didn't have the time or, have the energy to actually stop and then think about the Why?

因为我感觉，当你面对一个看似平庸的话题，真正深入探究到底，弄清楚一切背后的真实含义时，那种感觉——

Because I I feel that the feeling of when you're looking at a quite of tacky, topic and then when you really get to the bottom of it and really get to know the the true meaning of everything, that sense of

更深层次的理解。

Deeper understanding.

它

It's

非常令人兴奋且令人满足。

really thrilling and satisfying.

我发现，我们总是不断在各种媒体渠道中寻找和搜索故事，而当我看到科技类文章时，这些文章的写作风格只有科技人士才能理解。

I find the way that, you know, we're constantly looking and searching stories in the media across all different types of media outlets, and I find that when I come across tech articles, the articles themselves are written in a way that tech people can understand.

嗯。

Mhmm.

但非科技人士根本无法理解，这些文章简直就像在说德语一样。

But non tech people would have no idea what you might as well be speaking German in those articles

他们往往会直接从如何使用开始。

a They lot would start from how to use it directly.

这很实用。

That's practical.

这对那些已经具备科技背景的人很有用。

That's useful for those who have already got a background in the tech world.

但对我们来说，能够理解它是什么、为什么存在，以及它为何对我们的生活有用或有意义，才是激发我刚才在玉顺眼中看到的火花的起点。

But for us, being able to understand what it is, why we have it, and then why it's useful or meaningful to our lives is the starting point of having the sparks that I just saw in Yushun's eyes.

嗯。

Mhmm.

所以，是的。

So, yes.

所以第一集，玉顺，你选了哪个我们应当理解其背后原因的事物？

So for the first episode, Yushun, what did you pick for this something that we should understand the why about?

这个词我想你最近也可能在社交媒体上看到过。

This is the word I think you may have also came across recently on social media.

它叫做代币。

It is called token.

哦，当然。

Oh, sure.

是的。

Yeah.

它经常出现在各种话题标签、热门搜索榜单上。

It is quite frequently appeared on on no matter if it's the hashtag or heated heated searched charts and everything.

首先，让我们从一个足以让你头晕的数字开始，这个数据来自新华社的报道：中国每日AI令牌使用量，现已正式翻译为中文，已达到140万亿，这一数据是国家数据局本周在中国发展论坛上公布的。

So first of all, let's start with a number that should honestly make your head spin, and that's from the news of Xinhua, China's daily AI token usage, now officially translated as in Chinese, has hit 140,000,000,000,000, and that was announced by the National Data Bureau at the China Development Forum just this week.

140万亿。

140,000,000,000,000.

我们这里到底在统计什么？

What exactly are we counting here?

这不是用户数量，不是美元金额，也不是搜索次数，而是我们刚刚提到的令牌。

It's not users, not dollars, not searches, but what we just mentioned, tokens.

这是一种单位。

It's a unit of something.

我一直知道它是一种单位。

I have always known it's a unit of something.

我只是不知道这个单位具体指的是什么。

I just don't know what that something is.

你说这个数字会让人头晕，140万亿，我听到后脑子里想的是：140万亿？

And you said this number is gonna make your head spin 140,000,000,000,000, and my brain said 140,000,000,000,000?

什么？

What?

什么的？

Of Of what?

然后你说的是令牌，我脑子里想的是，那是什么？

And then you said tokens, and my brain said, what's that?

是的。

Yeah.

我知道。

I know.

这就是为什么我们要说所以

And that is why we're having So

我的头还没转晕呢。

my head's not spinning yet.

是的。

Yeah.

所以，你其实答对了。

So basically, you you answered it correctly.

这是AI用来理解和处理信息的最小单位，就像人类用字符和词语进行交流一样。

This it is the smallest unit AI uses to understand and is also process information much like the characters and words humans use to communicate.

所以它被正式翻译为‘词元’，这听起来可能还是有点抽象，但没关系，因为今天我们要帮你弄明白它。

So it's officially translated as and that might still sound a bit abstract as that's totally fine because today we're going to help you make sense of it.

而这个数字之所以令人震惊，是因为早在2024年初，使用的词元数量只有1000亿，而短短两年内就增长了一千倍以上。

And, here's why that number is so shocking because back in early twenty twenty four, it was just 100,000,000,000 tokens that's been used, and then that's more than a 1,000 fold increase in just two years.

所以自然会引发一个问题：是爆炸了还是为什么这么快就爆炸了？

So naturally, this raises a question, what did it explode or why did it explode so fast?

是因为更多人正在与AI聊天吗？因为这本质上就是人们在使用AI。

Is it just because more people are chatting with AI because it is essentially connecting with the use of AI.

对吧？

Right?

还是说，在服务背后有更深层的原因？

Or is there something deeper going on beneath the service?

我猜至少部分原因是人们与AI的聊天更多了。

My guess would be, at least in part, it's because people are chatting more with AI.

我知道我自己就是这样。

I know that I am.

你知道吗，当ChatGPT刚推出的时候，那已经是好几年前了，对吧？

You know, when ChatGPT first came out, it's now years ago, isn't it?

是的。

Yeah.

我当时知道它是什么，甚至那会儿还在我的节目里讨论过它，但我并没有每天都用。

I knew what it was, and even I talked about it on my show at that time, but I didn't really use it every single day.

然后，现在ChatGPT、DeepSeek和其他平台已经如此普遍地融入我们的生活，我不但用这些工具工作，而且也越来越频繁地在日常生活中使用它们。

Then, you know, now it's ChatGPT and DeepSeek and other platforms have become so commonplace in our lives now that I don't just and I, of course, I use those tools for work, but I've started to use them more and more and more in my daily life as well.

嗯。

Mhmm.

我最近用得最多的一次是帮我去跟踪健身和饮食计划，真的。

And I so the thing that I used it most recently for was to help me follow a gym and diet pattern, seriously.

我正在把ChatGPT当作我的饮食教练来使用。

I'm using ChatGPT as almost like a a diet trainer for me.

是的。

Yeah.

对。

Yeah.

我会在每顿饭后拍下我吃的食物，它都会记得。

I'm taking pictures of the food that I eat at every single meal, and it remembers.

对吧？

Right?

所以它在一直跟进。

So it's following along.

然后它还会给我做一个大致的热量计算，帮助我保持自律。

And then it's doing, like, a rough calorie count for me, so it's kinda helping to keep me in line.

上周末，我点了一只烤鸭和一罐可乐，然后问它：可乐可以喝吗？

And then last weekend, I I ordered a roast duck and a can of Coke, and I said, is the Coke okay?

他和ChatGPT说，史蒂夫，你知道你希望我怎么告诉你一罐可乐里有多少糖吗？

And he's and the Chatty PT said, Steve, do you know how do you want me to tell you how much sugar is in one can of Coke?

我说：关掉手机，现在好好享受我的食物吧。

And I said, turning phone off so I can enjoy my food now.

但没错，我的意思是，这大概是我表达的一种非常迂回的方式：是的，我认为这是因为人们每天越来越多地与AI聊天。

But, yeah, my point is, I guess that was a very long way of saying, yes, I do think it's because people are chatting with AI more and more on a daily basis.

好的。

Okay.

我的答案是，在我的世界里，我使用AI的工具分为两类。

My answer is that in my world, there are two categories of things that I use token that I use AI with.

一类是免费的，另一类是需要我付费的。

One is the free ones, and the other is the ones that needs me to charge.

嗯。

Mhmm.

当我付费时，我可以购买积分，或者成为某个AI的月度VIP用户，无论是Dova还是DeepSeek，我都能获得一定数量的积分。

And when I charge, I can buy tokens, or I can become a subscriber of a monthly VIP user user of a certain AI, no matter it's Dova or DeepSeek, and I get a certain amount of token.

所以最初，我理解它作为一种单位，是因为我相信我是在购买某种东西，以便提升我的AI水平，让它能更好地帮助我处理更难、更深或更复杂的事情。

So initially, the reason that I understand it, I understood it as a unit of something is because I believe I was buying that something so I can up level, upgrade the level of my AI so that it can help me with harder or deeper or more complicated things or in a better way.

我不确定这是否是一种赌注，对吧。

I don't know if it is a bet yeah.

更好一点说，我觉得这些平台的设计各不相同，但它们通常都有每日使用限制，比如聊天次数或上传文件、照片的次数；当我达到上限后，我就停止使用。

Better, I guess, but these platforms well, they're all designed differently, I guess, but they have a daily limit sometimes where you can chat with or upload files, pardon me, or photos, but then when I reach my daily limit, I just stop.

是的。

Yeah.

然后第二天它会重新充值，我又可以免费使用了。

And then it recharges and I and I use it for free again the next day.

但如果我想付费升级服务，我就是在为这些令牌付费。

But so if I wanna pay for the upgraded service, I'm paying for the tokens

你将要使用的那些。

That you are going to use.

我将要使用的那些。

That I am going to use.

所以这些代币就像一种货币。

So the tokens are like a currency.

这正是我最初认为它是一种东西的单位的原因，我当时是在购买这个单位。

That was why initially I thought it's a unit of something, and I was buying the unit of something.

所以我想知道，以你对科技界的半专家视角，玉顺，你认为整个代币相关世界背后最重要的原因是什么？

So I wonder, in your opinion, from a semi expert in the tech world, Yushun, what do you think is the most important why behind the whole token related world?

所以这个重要的‘为什么’，我认为是：为什么这个微小的、看不见的代币单位，现在被当作定义整个AI经济的数值？

So the important why, I think, is why is this tiny invisible unit token now being treated as the number that defines the entire AI economy?

因为你们俩都提到了一些

Because both of you mentioned some of

你希望我回答这个问题吗？

the Did you want me to answer that question?

当然不。

Of course not.

我们不能。

We can't.

我在

I am

我们会在这里待一会儿。

We'll be here for a while.

是的。

Yeah.

我会给你一些选项来回答这个问题，看看你能否猜出，或者解释一下，或者分析一下背后的原因。

I'll I'll be giving you some options to answer that questions and see if you can guess or just explain or or just analyze the the the reason.

好的。

Okay.

问题是什么来着？

What what's the question again?

抱歉。

Sorry.

那么，为什么这个微小的、看不见的代币单位现在被当作定义整个AI经济的关键数字呢？

So why is this tiny invisible unit token now being treated as the number that defines the entire AI economy?

好的。

K.

首先，我给你三个选项。

So first of all, I have three options for you.

A，因为词元是计算单词和文本长度最准确的方法。

A, because token is simply the most accurate way to count words and text length.

B，词元代表最小单位，你之前稍微提到了。

B, token represents the smallest unit well, you kind of mentioned it.

A，最小单位是AI用来处理信息的，同时也直接衡量成本、内存和计算量。

A, smallest unit AI uses to process information and also directly measures cost, memory, and computation.

C，因为词元是一种标准化的全球货币，用于在不同平台间交易AI服务。

And c, because token is a standardized global currency created to trade AI services across platforms.

你先说。

You go first.

我有点想把B和C结合起来。

I I kinda wanna say b and c together Mhmm.

因为我的初步理解是C，也就是可以购买令牌。

Because my initial understanding would be C, that is, it can buy token.

但说它是标准化的全球货币并不正确，因为我不认为每个平台的价格都是固定的，所以我把C划掉了。

But the part that a standardized global currency is not true because I don't think the price is fixed for each and every platform, so that's why I'm striking striking off off c C a a little.

划掉。

Little.

我也不选C，因为虽然我之前说过，令牌有点像货币。

I'm gonna say no to c as well because even though I said before, tokens are kind of like a currency.

但我认为令牌对我们来说才像货币。

I think tokens are only like a currency to us to us.

但对于AI来说，我认为令牌根本不代表货币。

But to the AI, I don't think tokens represent currency at all.

对AI而言，它必须代表某种能使其进行计算或执行其功能的东西。

To the AI, it has to represent some type of thing that allows them to compute or do whatever it is that they do.

所以我猜是B。

So I'm gonna guess b.

你知道吗，我总是喜欢史蒂夫在猜这些问题时那种逻辑性的思维方式。

You know, I always like Steve's, like, logical thinking when they were when he was guessing some of these questions.

AI不需要货币。

AI doesn't need currency.

是的。

Yeah.

答案是史蒂夫是对的。

And the answer is Steve was correct.

天哪。

Oh my gosh.

我现在就想退休了。

I'm gonna retire right now.

就这样了。

That's it.

我的职业生涯结束了。

My career is over.

我对这个没问题。

I'm good with it.

实际上，是的，你们两位都提到了关于代币如何运作的一些想法，因为史蒂夫说你上传了图片，然后它会回复一些想法或建议给你，而牛红林则说你在购买代币。

Actually, yeah, both of you touched upon some of the ideas of how token works because Steve said you uploaded pictures and then, they will reply something of the idea or advice for you and Niu Honglin was saying that you were buying token.

对吧？

Right?

这些实际上是我们现在与这些聊天机器人交互时使用的一些高级功能，因为以前我们还记得，只是在和它们对话，它们只是以简单的文本回复我们。

And these are actually some kind of advanced requirements that we are using with these chatbots because previously, we can still remember that we're just talking to them and they're replying to us and these are just simple text responding text kind of behavior.

代币并不是一个真正的单词。

And a token is not really a word.

它也不完全是一个字符，你可以把它看作是AI模型能够理解和处理的最小信息单元。

It's not exactly a character either, and think of it as the smallest chunk of information that an AI model can understand and also process.

你们都说你们几乎每天都在使用AI。

You all said you you are kind of using, AI almost every day.

所以当你打开像豆包、DeepSeek或JGPT这样的聊天机器人，然后说：‘帮我总结一下这篇文章。’

So when you're opening a chatbot like Dobao or DeepSeek or JGPT and ask, okay, summarize this article for me.

所以当你发送这句话时，你的问题实际上消耗了10个token。

So when you're sending this sentence, your question means token being used, for example, 10 tokens.

然后AI平台会分析这篇文章，而这种行为、这个操作也会消耗token。

And then the AI platform is analyzing this article and then this behavior, this action also consumes token.

啊，原来如此，难怪我们之前讨论时提到，说‘谢谢’其实也会产生额外的token。

Ah, that's why when we were discussing the fact that by saying thank you, we are actually causing some more tokens.

因为当我们对AI说‘谢谢’时，AI需要使用token来理解我们说的话。

It's because when we say this, when we say thank you to AI, they would need to use token to understand what we're saying.

同时也要回复。

And also reply.

是的。

Mhmm.

那么在这种情况下，为什么我不会为所有与AI的互动付费呢？

So in that case, why am I not paying for all of my interactions with AI?

为什么我只用为那些更难、更复杂或升级版的服务付费？

How come I'm only paying for the harder, complicated, or upgraded ones?

是的。

Yes.

这是个非常有趣的问题，因为一些AI聊天机器人甚至图像生成AI，我们都可以免费使用。

That's it is very interesting question because some of the AI's chatbots or even picture generated AI's, we use it for free.

对吧？

Right?

许多AI工具是免费的，因为这些公司采用了多种策略。

And many AI tools are free because these companies use a mix of strategies.

首先，它们提供免费的令牌额度来吸引用户，就像这些免费试用一样。

First of all, they offer free token allowances to attract users like these free trials.

而且，而且，史蒂夫，你说你每天都有一定的额度，用完了就停止了。

And and and Steve, you said you have some kind of allowances daily and then when you use it up, you just stopped it.

对吧？

Right?

不对。

No.

这种行为也很有用，因为它们可以通过观察人们如何使用其产品来训练自己的人工智能。

These kind of behaviors also, it is useful because they can learn from how people use their products to train their own AI.

所以你实际上也在为它们带来利润或好处，即使你没有支付任何费用。

So you're actually also making maybe profits or benefits for them So the even though you didn't pay anything.

所以，是的，我们正在讨论像DeepSeek和ChatGPT这样的东西，但也可以是其他类型的平台。

So, yeah, we're talking about things like DeepSeek and ChatGPT and things like that, but it can be different types of platforms too.

如今我们经常使用社交媒体，看到人工智能生成的视频。

There's a lot we go on social media these days and we see AI generated videos.

对吧？

Right?

还有一些平台可以基于文本提示或你上传的图片为你生成这些视频。

And there are platforms that can generate these videos for you either from a text prompt or from a picture that you upload yourself.

很多这类付费平台，我之前就见过‘积分’这个词。

And a lot of these paid platforms, that's where I saw this word before, tokens.

所以你为AI视频生成器购买积分，本质上你支付的是计算能力？

So you buy tokens for the AI video generators, and so what you're paying for essentially is Computing power?

这不是订阅费用。

It's not the subscription.

不像杂志订阅那样，对吧？

It's not like a magazine subscription, is it?

你付钱是为了使用这个AI工具，让它帮你计算所有内容，而他们用的一个可爱术语就是‘代币’。

You're pay you're paying for the for the tool, for the AI tool to, yeah, compute everything for you, and the the the cute little term that they use for it is token.

是的。

Yeah.

这就是为什么公司不按问题收费。

That's why companies don't charge you per question.

他们是按代币收费。

They charge you per token.

从某种意义上说，代币正成为AI的电费账单。

So in a way, token is becoming the electricity bill of AI.

你用得越多，付得越多。

And the more you use, the more you pay.

这正是我一直思考的某种单位。

That's the something in a unit of something I was thinking about the entire time.

好的。

Okay.

现在，我这边的下一个问题是，对于某些任务，AI执行起来非常顺畅。

Now, the next question from my end would be that for certain mission that I want AI to do, they do it very smoothly.

但对于其他一些任务，不仅无法像我预期的那样快速、顺畅地完成，

But for some other missions, not only can't they do it very quickly and very smoothly in my daily life as I anticipated.

它们甚至会说：你的电脑没有能力完成这项任务，无法提供我想要的。

They would even say something like, your computer does not have the ability to conquer that mission, to deliver I that

我不明白你的意思。

don't know what you mean.

你能举个例子吗？

Can you think of an example?

例如，如果我让AI生成一段短视频，而我面前已经有一个模型，比如一只猫随着某段音乐跳舞，这没问题。

For example, if I am asking AI to generate a short video that is already that I have a model already in front of my face, for example, a cat dancing to a certain piece of music, that is okay.

它们可以在几分钟内，最多几小时内给我最终的视频。

They can give give me the video, the final result, in several minutes or several hours at most.

但如果我让AI生成一段相同长度的视频，同时给出非常具体的指令，比如我希望这只猫是这种类型的猫。

But if I tell the AI to generate a the same length of a video, but I give it very specific instructions, like I want the cat to be this type of cat.

我希望它以这个角度移动。

I want it to move to this angle.

我是在从零开始创作，而不是使用它提供给我的模型。

I am creating something from scratch, not from the model that it provides me.

那么事情就变得复杂了。

Then things got complicated.

是的。

Yeah.

而且你想一想。

And because think about it.

它们正在思考或分析你的需求和请求，实际上，这正是令牌或AI的工作方式，因为它们会实时生成这类令牌。

They they are thinking or analyzing your requirements, your requests, and then actually, that's how token or AI works because they are generating that kind of token in real time.

是的。

Yeah.

但我认为还有另一个问题。

But I think there's another issue.

如果我错了，请纠正我，玉顺。

Correct me if I'm wrong, Yushun.

我认为还有一个问题，那就是它不仅需要具备计算你所要求内容的算力，还需要能够理解你所提出的要求。

I think there's another issue there is that it's not just what it needs to be able to compute the power to compute what you've requested, but it needs to be able to understand what you've requested as well.

并将我的话转化为它能够计算或权衡的内容，对吧。

And transform my words into something that it can compute or to weigh Right.

一起，而且

Together And

我经常使用这些视频创作工具，但这里有个问题，或者说我遇到的情况是。

I've used these video creation tools a lot, but here here's the problem or here's what I experienced.

所以，我想我们现在有点转向另一个话题了。

So and I guess we're kind of on a different topic now.

对不起，玉顺。

Sorry, Yushun.

但在我脑子里，我非常清楚地知道我想要的视频是什么样子。

But so I have in my brain very clearly what I want the video to be.

对吧？

Right?

我能在脑海中清晰地看到它。

I can see it in my mind's eye.

我知道，而且我认为我在提示框里给出了完美的说明，我期待提示框里的内容能完全实现。

I know, and I think I give a perfect explanation in the prompt box, and I expect exactly what I put in the prompt box.

但现在提示框输出的结果，完全不是我想要的视频样子。

Now what comes out of the prompt box is not at all what I wanted the video to look like.

这是AI工具的问题，还是我的问题？

Is that the AI tool's fault, or is it my fault?

依我看来，这某种程度上是我的错，因为我并不知道该如何用准确的语言与AI沟通，而这也是如今的一个问题。

Well, in my thinking, it's kind of my fault because I don't know the specific language to use to talk to the AI, and this is an issue these days as well.

要让AI做你想要的事情，就必须知道如何与它沟通。

How to get the AI to do what you want to do, have to know how to talk to it.

你不是在和另一个人类交谈。

You're not talking to another human being.

你是在和一台只理解非常、非常、非常特定表达方式的计算机对话。

You're talking to a computer that understands a very, very, very specific way of ex explanation.

但我想回到你的主题，玉顺，回到那140万亿个令牌的问题上。

But I guess to tie it back to your topic, Yushun, and going back to the 140 it was 140,000,000,000,000 that Tokens.

是的。

Yeah.

当我们无法与它有效沟通、第一次请求时得不到想要的结果时，就会浪费越来越多的令牌，这大概就是你遇到的问题。

That we're wasting more and more tokens when we don't speak to it, when we don't get what we want from it the first time we request something, which is the problem, I guess, you had.

浪费只是一种视角，我认为另一种视角是，AI正逐渐或悄然成为基础设施，因为越来越多的AI、工具和平台正在成为众多行业的基础设施。

Waste is one, I think, perspective of of of it, and I think another perspective is that it is becoming or silently becoming the infrastructure because more and more AI or more and more tools and platforms are becoming the infrastructure of so many industries.

因此，我们才使用如此多的令牌，而它正成为一种事实——即它是一套24小时不间断为我们运行的基础设施。

So and that is why we're using so many tokens, and then it is becoming a fact that it is the infrastructure that is working continuously twenty four seven for us.

对。

Right.

因为这不仅仅是你我这样的用户，林玉顺，我们在制作什么来着？

Because it's not just people like you and I, Yushun Lin, making our what was it?

跳舞的猫咪视频。

Dancing cat videos.

跳舞的猫咪视频。

Dancing cat video.

嗯哼。

Mhmm.

我需要知道

I need to know

稍后。

later.

而且

And

但等一下。

But hold on.

不好意思打断一下。

So sorry to interrupt.

这不仅仅是你和我用这些AI工具来做我们想做的事。

It's not just you and I talking to these AI tools to do what we want.

现在医院已经在使用它了。

This is being used at hospitals now.

它也被用于无人机配送。

This is being used for drone delivery.

它被用于自动驾驶汽车。

This is being used for autonomous vehicles.

它被用于餐厅，比如咖啡机器人制造商。

This is being used in restaurants, for coffee robot makers.

现在它无处不在，时刻都在使用，每当AI在运行、在工作、在运行、在工作时，它都在消耗越来越多、越来越多、越来越多的令牌。

This is being used everywhere all the time now, and every time AI is working and doing and working and doing, it's using more and more and more and more and more tokens.

我这么说对吗？

Am I correct?

是的。

Yeah.

那这140万亿个令牌在现实世界中到底意味着什么？它们之于AI，就像千瓦时之于电力。

And what, like, what does that 140,000,000,000,000 tokens actually mean in the real world is that, they are to AI what kilowatt hours are to electricity.

好的。

Okay.

令牌之于AI，就像千瓦时之于电力。

Tokens are to AI what kilowatt hours are to electricity.

好的。

Okay.

抱歉。

Sorry.

你继续说。

Go ahead.

嗯嗯。

Mhmm.

当我们说140万亿个标记时，我们真正想表达的是，人工智能每天都在以巨大且工业化的规模被使用。

When we say 140,000,000,000,000 tokens, what we are really saying is AI is being used at a massive and also industrial scale every single day.

我还有另一个问题，因为当我们谈论千瓦时和电力时，我们从政府和媒体的信息中了解到。

Then I have another question because when we talk about kilowatt hours and electricity, and we've learned through messages from governments and media.

离开房间时，要把灯关掉。

When you leave a room, turn off your lights.

那样你就是在浪费能源。

Then you're wasting energy.

嗯嗯。

Mhmm.

对吧？

Right?

我们应该尽自己所能保护环境，以节约尽可能多的能源。

We should do what we can for the environment to helps conserve as much energy as we can.

这不也是同样的问题吗？

Is that not the same question here?

我们难道不应该在这里也问同样的问题吗？

Should we not ask that same question here?

我们是不是应该停止每天这么大量地使用令牌？

Stop using as many tokens as we're using all day every day?

或者我们是不是应该问，为什么我们现在使用的令牌数量如此之多，未来会不会继续越来越多？

Or maybe are we asking why are we using so many more tokens now, and are we going to use more and more and more tokens in the future?

简短回答，首先，大概是的，但并不是一个简单的问题，因为这两种力量正在同时发生。

And short answer, first of all, probably yes, but not a very simple way because these two forces are happening at the same time.

一方面，对AI的需求正在激增，因为我们有了更多应用、更多使用，所以令牌使用量也增加了。

One, the demand of using AI is exploding because we're having more applications, more usage, so more tokens.

另一方面，效率也在提高。

And also efficiency is also improving.

更好的模型，每个任务所需的令牌更少，所以未来有点矛盾。

Better models, fewer tokens needed per task, so the future is a bit of a paradox.

我们可能每个任务使用的令牌更少，但总体上使用的更多。

We may use fewer tokens per task, but we may use more overall.

是的。

Yeah.

然后还有一种非常非常有趣的心态转变，就是为什么有些人说，别给我钱。

And then there is a really, really fascinating shift in mindset that is why are some people saying just, you know, don't pay me money.

用令牌支付我。

Pay me in tokens.

因为我也从一些从业者那里听说过这种说法。

Because that is I I also heard that from some of the practitioners.

我从未见过这种情况。

I've never seen that.

对。

Yes.

这些‘用令牌支付我’的情况。

These Pay me in tokens.

嗯嗯。

Mhmm.

这又回到了货币理念的问题上。

That come back to kind of the currency idea again.

这听起来一开始很奇怪，但实际上很有道理，因为在人工智能经济中，代币就像获取智能、构建和创造以及自动化的能力，而这同样代表了生产力。

This sounds weird at first, but it actually makes sense because in the AI economy, tokens are like access to intelligence, ability to build and create and to automate, and that also equals to that productive capacity.

因此，对于开发者、初创公司或企业来说，拥有代币可能比拥有现金更有价值。

So for developers and startups or companies, having tokens can be more valuable than having cash.

所以这就像在说，我不需要钱。

So it's kind of like saying, I don't need money.

我只要给我无限的电力。

I just give me unlimited electricity.

或者更简单地说，直接给我无限的计算能力。

Or in in a easier way, just give me unlimited computing power.

这真是太有趣了。

Well, this is fascinating.

我可以继续问很多为什么或怎么做，比如，既然我们现在已经有了一些技术，可以通过更优、更环保的编码来减少完成相同任务所需的代币数量。

I can continue asking a lot of whys or hows in the sense that, well, since we do have the some of the technology to allow us to use less tokens for the same mission because we're doing or using better coding, greener coding.

嗯。

Mhmm.

那么，未来我们是否可能看到代币价格不仅上涨，反而下降呢？

And is it possible that in the future, we get to see not the rising of prices in the token world, but also see a lowering of it?

既然我们在讨论价格，那么对于我们普通人来说，是否有可能面临这样的风险：我们只能接触到不够智能的AI，因为最智能的AI需要消耗更多代币，意味着它会更昂贵？

And since we're talking about prices, is it possible, or is it a threat, a risk for common people like us to think about the fact that we can get access to the not, well, smartest AI because the smartest AI would cost more token, which means the smartest AI might be a bit more pricey.

我们是否正走向这样一个世界：未来我们需要政府出台更完善、更细致的AI发展指南，以确保AI使用世界的质量，而这又对我们的能源供应商意味着什么？

Are we looking at a world where in the future we need the government to give well, we definitely would need the government to give better and more detailed guidelines in the development of AI to make sure the quality in the AI using world, and also what does this mean for our energy suppliers?

这对建设一个更绿色的未来意味着什么？

What does this mean for building a greener future?

这些都可以继续追问，但关于‘为什么’这一部分的精彩之处在于，它让我们能够观察趋势、了解最新的技术突破，并提出‘为什么’的问题。

These are all the questions we can continue to ask, but what's fascinating about the Why segment is that it's the segment where you get to observe the trends, the latest technological breakthroughs, and ask the Why's.

实际上，我今天最大的收获——虽然不介意对宇顺说——是你拍了太多跳舞的猫视频，多到你的电脑都受不了了。

Actually, my favorite takeaway today, no offense to Yushun, is that you make so many dancing cat videos, so many in fact, that your computer says, I can't handle this anymore.

至少这是一种人们能够清楚理解代币用途的方式。

Well, at least that is a very obvious way people can understand the use of token.

是的。

Yeah.

解释得很清楚。

Well explained.

谢谢。

Appreciate it.

谢谢。

Thank you.

谢谢。

Thank you.

这为我们今天的圆桌讨论画上了句号。

And that brings us to the end of today's roundtable.

我们很期待听到您对今天话题的看法。

We'd love to hear your thoughts on today's topics.

请在您收听播客的平台留下评论，或发送邮件至 roundtablepodcast@qq.com。

Leave us a comment wherever you listen to your podcast or send us an email to roundtablepodcast@qq.com.

我是牛红林，和Steve、Yushun一起。

I'm Niu Honglin with Steve and Yushun.

谢谢

Thank

感谢收听。

you for listening.

再见。

Bye bye.