Name: 人工智能對編程有多大用處？| 馬克-安德森和萊克斯-弗裡德曼 (How useful is AI for programming? | Marc Andreessen and Lex Fridman)
Uploaded: 2025-01-31T21:45:09.000Z
Duration: 12 min 42 s

I think your profile on X states it's time to build, it feels like 2025 is a good year to build.

我覺得你在 X 上的簡介中說是時候建造了，感覺 2025 年是建造的好年份。

程度副詞

So I wanted to ask your advice and maybe for advice for anybody who's trying to build, who's trying to build something useful in the world, maybe launch a startup or maybe just launch apps, services, whatever, ship software products.

所以，我想問問你的建議，也許是給那些想在世界上建立一些有用的東西的人的建議，也許是啟動一家初創公司，也許只是推出應用程序、服務，不管怎樣，推出軟件產品。

情態副詞

So maybe by way of advice, how do you actually get to shipping?

那麼，或許我可以給你提個建議，你究竟該如何出貨呢？

So I mean a big part of the answer I think is we're in the middle of a legit revolution and I know you've been talking about this on your show, but like AI coding, I mean this is the biggest earthquake to hit software in certainly my life, maybe since the invention of software.

所以，我認為答案的很大一部分是，我們正處在一場合法革命的中間，我知道你在節目中一直在談論這個，但就像人工智能編碼一樣，我的意思是，這肯定是我這輩子，也許是自軟件發明以來，軟件遭遇的最大地震。

And I'm sure we're involved in various of these companies, but these tools from a variety of companies are absolutely revolutionary and they're getting better by leaps and bounds right every day.

我相信我們也參與了這些公司中的許多公司，但這些來自不同公司的工具絕對是革命性的，而且每天都在突飛猛進。

And you know all this, but like the thing with coding, there's like open questions of whether AI can get better at, I don't know, understanding philosophy or whatever, creative writing or whatever, but like for sure we can make it much better at coding, right?

你知道這一切，但就像編碼一樣，人工智能是否能更好地理解哲學或其他什麼，創意寫作或其他什麼，這些都是懸而未決的問題，但可以肯定的是，我們能讓它在編碼方面做得更好，對嗎？

Because you can validate the results of coding.

And so there's all these methods of synthetic data and self-training and reinforcement learning that for sure you can do with coding.

是以，所有這些合成數據、自我訓練和強化學習的方法，你都可以通過編碼來實現。

And so everybody I know who works in the field says AI coding is going to get to be phenomenally good and it's already great.

是以，我認識的所有從事該領域工作的人都說，人工智能編碼將變得非常出色，而且它已經很棒了。

And you can, I mean, anybody who wants to see this, just go on YouTube and look at AI coding demos, little kids making apps in 10 minutes working with an AI coding system.

你可以，我是說，任何想看這個的人，只要上 YouTube 看看人工智能編碼演示，小孩子們用人工智能編碼系統在 10 分鐘內就能做出應用程序。

是以，我認為這是一個黃金時代。

I mean, I think this is an area where it's clearly the golden age.

我的意思是，我認為這是一個明顯處於黃金時代的領域。

You know, in a day, as a coder for sure in a day, you can retrain yourself, you know, start using these things, get a huge boost in productivity.

要知道，一天之內，作為一個編碼員，你就可以重新培訓自己，開始使用這些東西，極大地提高工作效率。

As a non-coder, you can learn much more quickly than you could before.

作為一名非編碼員，你可以比以前學得更快。

That's actually a tricky one in terms of learning as a non-coder to build stuff.

作為一個非編碼員，在學習構建東西方面，這實際上是一個棘手的問題。

It's still, I feel like you still need to learn how to code.

我覺得你仍然需要學習如何編寫代碼。

Like you could legitimately be a one person company and get quite far.

就像你可以合法地做一個人的公司，並走得很遠。

So I think for sure for quite a long time, the people who are good at coding are going to be the best at actually having AIs code things because they're going to understand what, I mean, very basic, they're going to understand what's happening, right?

是以，我認為在相當長的一段時間內，擅長編碼的人將是讓人工智能編碼的最佳人選，因為他們會理解，我是說，非常基本的，他們會理解發生了什麼，對嗎？

And they're going to be able to evaluate the work and they're going to be able to, you know, literally like manage AIs better.

他們將能對工作進行評估，並能更好地管理人工智能。

Even if they're not literally handwriting the code, they're just going to have a much better sense of what's going on.

即使他們沒有真正手寫代碼，他們也能更好地瞭解發生了什麼。

So I definitely think like 100%, my nine-year-old is like doing all kinds of coding classes and he'll keep doing that for certainly through 18.

是以，我絕對認為，我九歲的孩子百分之百在上各種編碼課程，而且他肯定會一直上到 18 歲。

But look, having said that, one of the things you can do with an AI is say, teach me how to code, right?

不過，話雖如此，你還是可以對人工智能說，教我如何編碼，對嗎？

And so, and, you know, there's a whole bunch of, you know, I'll name names, like there's a whole bunch of work that they're doing at Khan Academy for free.

所以，你知道，有一大堆人，你知道，我會說出他們的名字，比如他們在可汗學院免費做了一大堆工作。

And then, you know, we have this company, Replit, which was originally specifically built for kids for coding that has AI built in.

然後，你知道，我們有一家名為 Replit 的公司，這家公司最初是專門為孩子們設計的，它內置了人工智能編碼功能。

That's just absolutely extraordinary now.

And then, you know, there's a variety of other systems like this.

然後，你知道，還有各種各樣類似的系統。

And yeah, I mean, the AI is going to be able to teach you to code.

是的，我的意思是，人工智能可以教你編碼。

AI, by the way, is, as you know, spectacularly good at explaining code, right?

順便說一句，眾所周知，人工智能非常擅長解釋代碼，對吧？

And so, you know, the tools have these features now where you can talk to the code base.

是以，你知道，現在的工具都有這些功能，你可以與代碼庫對話。

And so you can like literally like ask the code base questions about itself.

這樣，你就可以像字面意思一樣，向代碼庫提出關於它自己的問題。

And you can also just do the simple form, which is you can copy and paste code into ChatGPT and just ask it to explain it, what's going on, rewrite it, improve it, make recommendations.

你也可以採用簡單的形式，即把代碼複製並粘貼到 ChatGPT 中，然後請它解釋代碼，說明發生了什麼，重寫代碼，改進代碼，提出建議。

And so there's, yeah, there's dozens of ways to do this.

是以，有幾十種方法可以做到這一點。

By the way, you can also, I mean, even more broadly than code, like, OK, you want to make a video game, OK, now you can do AI art generation, sound generation, dialogue generation, voice generation, right?

順便說一下，你還可以，我是說，甚至比代碼更廣泛，比如，好吧，你想做一個視頻遊戲，好吧，現在你可以做人工智能藝術生成、聲音生成、對話生成、語音生成，對嗎？

And so all of a sudden, like you don't need designers, you know, you don't need, you know, voice actors, you know.

所以突然之間，你不需要設計師了，你知道，你不需要配音演員了，你知道。

So, yeah, so there's just like unlimited.

所以，是的，所以就像無限的。

And then, you know, a big part of coding is so-called glue.

然後，你知道，編碼的很大一部分就是所謂的膠水。

You know, it's interfacing into other systems.

你知道，它是與其他系統的接口。

So it's interfacing into, you know, Stripe to take payments or something like that.

是以，它與 Stripe 的接口，你知道，Stripe 接受付款或類似的東西。

And, you know, AI is fantastic at writing glue code.

要知道，人工智能在編寫膠水代碼方面非常出色。

So, you know, really, really good at making sure that you can plug everything together, really good at helping you figure out how to deploy.

所以，你知道，我們真的非常擅長確保你能把所有東西都整合在一起，也非常擅長幫你找出如何部署。

You know, it'll even write a business plan for you.

你知道，它甚至會為你寫一份商業計劃書。

So it's just this, it's like everything happening with AI right now.

就像現在人工智能所發生的一切一樣。

It's just, it's like this latent superpower.

And there's this incredible spectrum of people who have really figured out massive performance increases, productivity increases with it already.

有很多人已經利用它實現了性能的大幅提升和生產力的大幅提高。

There's other people who aren't even aware it's happening.

還有一些人甚至沒有意識到這一點。

And there's some gearing to whether you're a coder or not, but I think there are lots of non-coders that are off the races.

無論你是否是編碼員，都會有一定的匹配度，但我認為有很多非編碼員都不在比賽之列。

And I think there are lots of professional coders who are still like, you know, the blacksmiths were not necessarily in favor of, you know, the car business.

我認為，有很多專業編碼員仍然像鐵匠一樣，不一定支持汽車行業。

So, you know, there's the old William Gibson quote, the future is here.

威廉-吉布森有句名言：未來已來。

And this is maybe the most potent version of that that I've ever seen.

這也許是我所見過的最有力的版本。

Yeah, there's, you know, the old meme with the bell curve, the people on both extremes say AI coding is the future.

是啊，你知道的，老掉牙的鐘形曲線，兩個極端的人都說人工智能編碼是未來。

It's very common for programmers to say, you know, if you're any good of a programmer, you're not going to be using it.

程序員通常會說，如果你是個優秀的程序員，你就不會使用它。

I consider myself a reasonably good programmer and my productivity has been just skyrocketed and the joy of programming skyrocketed.

我認為自己是一個相當不錯的程序員，我的工作效率直線上升，編程的樂趣也直線上升。

Every aspect of programming is more efficient, more productive, more fun, all that kind of  I would also say code is, you know, code has of anything in like industrial society, code has the highest elasticity, which is to say the easier it is to make it the more of it gets made.

編程的方方面面都更有效率、更有生產力、更有趣，所有這些。我還想說的是，在工業社會中，代碼是任何東西中彈性最大的，也就是說，代碼越容易製作，就會有越多的代碼被製作出來。

Like I think effectively there's unlimited demand for code.

就像我想的那樣，實際上對代碼的需求是無限的。

Like in other words, like there's always some other idea for a thing that you can do, a feature that you can add or a thing that you can optimize.

換句話說，總會有一些其他的想法，比如你可以做的事情，你可以添加的功能，或者你可以優化的東西。

And so, like overwhelmingly, you know, the amount of code that exists in the world is a fraction of even the ideas we have today.

是以，就像絕大多數人一樣，你知道，世界上存在的代碼量甚至只是我們今天所擁有的想法的一小部分。

And then we come up with new ideas all the time.

然後，我們不斷提出新的想法。

And so, I think that like, you know, I was in the late 80s, early 90s when sort of automated coding systems started to come out, the expert systems, big deal in those days.

是以，我認為，在 80 年代末 90 年代初，自動編碼系統開始出現，專家系統也開始出現，這在當時是件大事。

And there were all these, there was a famous book called The Decline and Fall of the American Programmer, you know, that predicted that these new coding systems were going to mean we wouldn't have programmers in the future.

有一本很有名的書叫《美國程序員的衰落與沒落》，書中預言這些新的編碼系統將意味著我們未來將不再有程序員。

And of course, the number of programming jobs exploded by like a factor of 100.

當然，編程工作的數量也以 100 倍的速度激增。

Like my guess will be we'll have more, my guess is we'll have more coding jobs probably by like an order of magnitude 10 years from now.

我的猜測是，10 年後，我們會有更多的編碼工作，可能會是一個數量級。

But there will be, we will be creating so much more software that the whole industry will just explode in size.

但是，我們將會創造出更多的軟件，整個行業的規模將會爆炸式增長。

Are you seeing the size of companies decrease in terms of startups with the landscapes of little tech?

您是否看到，隨著小技術的發展，初創公司的規模有所縮小？

All we're seeing right now is the AI hiring boom of all time.

我們現在看到的是有史以來人工智能的招聘熱潮。

Everybody's trying to hire as many engineers as they can to build AI systems.

每個人都在想方設法僱傭儘可能多的工程師來構建人工智能系統。

I mean, there's a handful of companies, you know, there's a little bit, there's customer service.

我的意思是，有少數幾家公司，你知道，有一點，有客戶服務。

You know, we have some companies and others, I think it's Klarna that's publicizing a lot of this in Europe, where, you know, there are jobs that can be optimized and jobs that can be automated.

你知道，我們有一些公司和其他公司，我想是 Klarna 公司在歐洲宣傳了很多這方面的資訊，你知道，有一些工作可以優化，有一些工作可以自動化。

But like for engineering jobs, like it's just an explosion of hiring.

但就像工程類工作一樣，招聘人數激增。

At least so far, there's no trace of any sort of diminishing effect.

至少到目前為止，還看不出任何遞減效應的跡象。

Now, having said that, I am looking forward to the day, I am waiting for the first company to walk in saying yes, like the more radical form of it.

話雖如此，但我還是期待著這一天的到來，我在等待著第一家公司走進來說 "是的"，就像更激進的形式一樣。

So basically, the companies that we see are basically one of two kinds.

是以，我們看到的公司基本上有兩種。

We see the companies that are basically, sometimes used weak form, strong form.

我們看到的公司基本上都是這樣，有時採用弱形式，有時採用強形式。

So the weak form companies, I sometimes use the term, it's called the sixth bullet point.

是以，弱形式公司，我有時會用這個詞，叫做第六個要點。

AI is the sixth bullet point on whatever they're doing.

無論他們在做什麼，人工智能都是第六個要點。

So they've got the, you know, whatever, dot, dot, dot, dot, and then AI is the sixth thing.

是以，他們已經有了，你知道，不管是什麼，點、點、點、點，然後人工智能是第六件事。

And the reason AI is the sixth thing is because they had already previously written the slide before the AI revolution started.

而人工智能之所以是第六件事，是因為在人工智能革命開始之前，他們就已經寫好了幻燈片。

So they just added the sixth bullet point on the slide, which is how you're getting all these products that have like the AI button up in the corner, right?

是以，他們在幻燈片上添加了第六個要點，這就是你如何獲得所有這些在角落裡有人工智能按鈕的產品，對嗎？

And all of a sudden, Gmail is offering to summarize your email, which I'm like, I don't need that.

突然間，Gmail 提供了郵件摘要功能，我想，我不需要那個。

Like, I need you to answer my email, not summarize it.

比如，我需要你回覆我的郵件，而不是總結它。

好吧，我們看到了這些，這很好。