Interview
Sam Altman Shows Me GPT-5... And What's Next
Sam Altman展示GPT-5与未来
Sam Altman publicly demonstrates GPT-5 for the first time, discussing the AGI roadmap, agent strategy, and what comes next.
Sam Altman 首次公开展示 GPT-5 能力，深入讨论 AGI 路线图、Agent 战略与 AI 的未来走向。
00:001h 5m 6s
00:00
This is like a crazy amount of power for one piece of technology and it's happened to us so fast. >> You just launched GBT. >> A kid born today will never be smarter than AI. >> How do we figure out what's real and what's not real? >> We haven't put a sex bot avatar in HBT yet. >> Super intelligence.
这简直是一项技术拥有的疯狂力量，而且它来得如此之快。
00:17
What does that actually mean? >> This thing is remarkable. >> I'm about to interview Sam Alman, the CEO of Open AI. >> Open AI. >> Open AI. >> Reshaping industries. >> Dude's a straightup tech lord.
>> 你刚刚推出了GPT。
00:28
Let's be
>> 一个今天出生的孩子永远不会比AI更聪明。
00:29
honest.
说实话，现在他们正在试图构建一种超级智能，可能在几乎所有领域都远远超越人类。
00:29
Right now, they're trying to build a super intelligence that could far exceed humans in almost every field.
而他们刚刚发布了迄今为止最强大的模型。
00:36
And they just released their most powerful model yet.
就在几年前，这听起来还像是科幻小说。
00:39
Just a couple years ago, that would have sounded like science fiction.
但现在不是了。
00:43
Not anymore.
事实上，他们并不孤单。
00:44
In fact, they're not alone.
我们正身处一场有史以来最高 stakes 的全球竞赛中。
00:45
We are in the middle of the highest stakes global race any of us have ever seen.
数千亿美元和难以想象的人力投入。
00:50
Hundreds of billions of dollars and an unbelievable amount of human worth.
这是一个意义深远的时刻。
00:55
This is a profound moment.
这是一个深刻的时刻。
00:57
Most people never live through a technological shift like this, and it's happening all around you and me right now.
大多数人一生都从未经历过这样的技术变革，而现在它正在你和我身边发生。
01:04
So, in this episode, I want to try to time travel with Sam Alman into the future that he's trying to build to see what it looks like so that you and I can really understand what's coming.
所以，在这一集里，我想试着和Sam Altman一起时间旅行，进入他试图构建的未来，看看那是什么样子，这样你和我才能真正理解即将到来的东西。
01:17
Welcome to Huge Conversations.
欢迎来到《深度对话》。
01:24
>> How are you?
>> 你好吗？
01:24
Great to meet you.
很高兴见到你。
01:25
Thanks for doing this. >> Absolutely. >> So, before we dive in, I'd love to tell you my goal here. >> Okay. >> I'm not going to ask you about valuation or AI talent wars or fundraising or anything like that.
谢谢你接受这次访谈。
01:37
I think that's all very well covered elsewhere. >> It does seem like it. >> Our big goal on this show is to cover how we can use science and tech to make the future better.
>> 当然。
01:46
And the reason that we do all of that is because we really believe that if people see those better
>> 那么，在我们深入之前，我想先说说我的目标。
01:52
futures, they can then help build them.
>> 你好吗？
01:55
So, my goal here is to try my best to time travel with you into different moments in the future that you're trying to build and see what it looks like. >> Fantastic. >> Awesome.
很高兴见到你。
02:07
Starting with what you just announced, you recently said, surprisingly recently, that GPT4 was the dumbest model any of us will ever have to use again.
谢谢你来做这个。
02:17
But GPT4 can already
>> 当然。
02:19
perform better than 90% of humans at the SAT and the LSAT and the GRE and it can pass coding exams and SOA exams and medical licensing.
在SAT、LSAT和GRE考试中表现超过90%的人类，还能通过编程考试、精算考试和医学执照考试。
02:27
And now you just launched GPT5. >> What can GPT5 do that GPT4 can't?
而现在你们刚刚推出了GPT5。
02:31
First of all, one important takeaway is you can have an AI system that can do all those amazing things you just said.
首先，一个重要的收获是，你可以拥有一个能完成你刚才说的所有那些惊人事情的AI系统。
02:38
And it doesn't it clearly does not replicate a lot of what humans are good at doing, which I think says something about the
而且它显然并没有复制人类擅长的很多事情，我认为这本身就说明了点什么。
02:46
value of SAT tests or whatever else.
>> GPT5能做什么GPT4做不到的事？
02:48
But I think had you gone back to if we were having this conversation the day of GPT4 launch and we told you how GPT4 did at those things, you were like, "Oh man, this is going to have huge impacts and some negative impacts on what it means for a bunch of jobs or you know what people are going to do." And you know, this is a bunch of positive impacts that you might have predicted that haven't yet come true.
首先，一个重要的点是，你可以拥有一个AI系统，它能完成你刚才说的所有那些惊人的事情。
03:10
Uh, and so there there's
但它显然并没有复制人类擅长的很多东西，这本身也说明了SAT考试之类的价值。
03:11
something about the way that these models are good that does not capture a lot of other things that we need people to to do or care about people doing.
那些需要1分钟到1小时的任务，一个领域的专家可能能做到，也可能觉得吃力，而你口袋里就有这么一款软件，能搞定所有这些事情，这真的很了不起。
03:21
And I suspect that same thing is going to happen again with GPT5.
我觉得这在人类历史上是前所未有的，一项技术能如此快速、如此大幅度地进步，而我们现在就拥有了这个工具，我们正亲身经历着这一切。
03:25
People are going to be blown away by what it does.
人们会被它的能力震撼到。
03:28
Uh, it's really good at a lot of things and then they will find that they want it to do even more.
嗯，它在很多事情上确实很擅长，然后他们会发现，他们希望它能做得更多。
03:34
Um, people will use it for all sorts of incredible things. uh it
呃，人们会用它在各种不可思议的事情上。嗯，它……
03:38
will transform a lot of knowledge work, a lot of the way we learn, a lot of the way we create um but we people society will co-eolve with it to expect more with you know better tools.
将改变大量的知识工作、我们学习的方式、我们创造的方式，但人类和社会会与之共同进化，借助更好的工具期待更多。
03:52
So yeah like I think this model is quite remarkable in many ways quite limited in others but the fact that for you know 3 minute 5
所以，是的，我觉得这个模型在很多方面非常出色，在其他方面又很有限，但事实是，对于一个专家可能需要三分钟、五分钟、一小时才能完成、甚至可能搞不定的任务，你口袋里有一个软件就能搞定所有这些事情——这真的很惊人。
04:02
minute 1-hour tasks that uh like an expert in a in a field could maybe do or maybe struggle with that the fact that you have in your pocket one piece of software that can do all of these things >> is really amazing.
我感觉我可以问任何困难的科学或技术问题，然后得到一个相当不错的答案。
04:15
I think this is like unprecedented at any point in human history that I that a technology has improved this much this fast and and the fact that we have this tool now, you know, we're like living through it and
我举个有趣的例子吧。
04:28
we're kind of adjusting step by step.
假设人们没看过头条新闻，你具体兴奋的顶级事情是什么？
04:31
But if we could go back in time five or 10 years and say this thing was coming, we would be like probably not. >> Let's assume that people haven't seen the headlines.
还有你似乎在提醒的那些事情，那些你预计它做不到的事情？
04:41
What are the topline specific things that you're excited about? and also the things that you seem to be caveatting, the things that maybe you won't expect it to do. >> Um, the thing that I am most excited about is this is a model for the first time
你最期待的具体亮点是什么？同时，你似乎在提醒的那些事情，也就是你可能不指望它能做到的事情又是什么？>> 嗯，我最兴奋的是，这是第一次有一个模型——
04:56
where I feel like I can ask kind of any hard scientific or technical question >> and get a pretty good answer.
>> 是啊。
05:05
And I'll give a fun example actually.
实际上，我给你举个有趣的例子。
05:08
Uh when I was in junior high uh or maybe it was nth grade, I got a TI83, this old graphing calculator, and I spent so long making this game called Snake. >> Yeah.
呃，我初中那会儿，也可能是几年级来着，拿到一台TI83，就是那种老式 graphing calculator，然后我花了好长时间做了一个叫 Snake 的游戏。>> 是啊。
05:21
>> Uh it was very popular game with kids in my school.
嗯，我最兴奋的是，这是第一次有一个模型，让我觉得可以问任何硬核的科学或技术问题，然后得到一个相当不错的答案。
05:23
And I was I was like uh I was like pro and it was dumb, but it was like programming on TID3 was extremely painful and took a long time and it was really hard to like debug and whatever.
我给你举个有趣的例子。
05:34
And on a whim with an early copy of GPT5, I was like, I wonder if it can make a TI83 style Game of Snake.
我上初中或者可能是高中时，拿到了一台TI83，那种老式图形计算器，我花了很长时间做了一个叫“贪吃蛇”的游戏。
05:39
And of course, it did that perfectly in like 7 seconds.
对，这个游戏在我学校的孩子中很流行。
05:42
And then I was like, okay, am I supposed to be would my like 11-year-old self think this was cool or like, you
我当时很厉害，虽然这很傻，但在TI83上编程极其痛苦，耗时很长，调试也非常难。
05:48
know, miss something from the process?
你知道，会怀念那个过程吗？
05:50
And I had like 3 seconds of wondering like, oh, is this good or bad?
我大概有3秒钟在想，哦，这是好是坏？
05:54
And then I immediately said, actually, now I'm missing this game.
然后我立刻说，其实我现在想念这个游戏了。
05:57
I have this idea for a crazy new feature.
我有个疯狂的新功能想法，就把它输入进去，它直接实现了，游戏实时更新，然后我又说，我希望它看起来这样，我希望它能做这个。
05:59
Let me type it in. it implements it and it just the game live updates and I'm like actually I'd like it to look this way.
那种体验让我感觉像是回到了11岁，重新开始编程，就是那种“我现在就想做这个”的感觉。
06:05
Actually, I'd like to do this thing and I had this like this very like kind of >> you have this experience that reminded me of being like 11 in programming again where I was just like I now I want to
其实，我一直想做这么个东西，然后我有了这种……怎么说呢，有点像11岁重新开始编程的感觉，就是那种“我现在就想做这个”的状态。
06:15
try this now I have this idea now I but I could do it so fast and I could like express ideas and try things and play with things in such real time.
你刚才说的让我想到举重里的一个概念——“时间张力”。
06:23
I was like, "Oh man, you know, I was worried for a second about kids like missing the struggle of learning to program in this sort of stone age way." And now I'm just thrilled for them because the the way that people will be able to create with these new tools, the speed with which you can sort of bring ideas to life, you know, in
对，对于不了解的人来说，这是指……
06:42
that's that's pretty amazing.
这真是太惊人了。
06:43
So this idea that GPT5 can just not only like answer all these hard questions for you but really create like ondemand almost instantaneous software that's I think that's going to be one of the defining elements of the GPD5 era in a way that did not exist with GPD4. >> As you're talking about that I find myself thinking about a concept in weightlifting of time under tension. >> Yeah. >> And for those who don't know it's you
所以GPT5不仅能回答各种难题，还能按需几乎即时地生成软件，我觉得这将成为GPT5时代的一个标志性特征，而GPT4时代是没有的。
07:08
can squat 100 pounds in 3 seconds or you can squat 100 pounds in 30.
3秒内深蹲100磅和30秒内深蹲100磅，后者带来的收益要大得多。
07:12
You gain a lot more by squatting it in 30. >> And when I think about our creative process and when I've felt most like I've done my best work, it has required an enormous amount of >> cognitive time under tension. >> And I think that that cognitive time under tension is so important. >> And it's it's ironic almost because these tools have taken enormous cognitive time under tension to develop.
>> 说到我们的创作过程，我觉得自己做得最好的时候，往往需要大量的 >> 认知张力时间。
07:36
But in some ways I do think people might say they're you people are using them as a escape hatch for thinking in some ways maybe.
但某种程度上，我觉得有人会说，你们这些人把它当成逃避思考的出口了。
07:43
Now you might say yeah but we did that with the calculator and we just moved on to harder math problems.
你可能会说，是啊，但我们用计算器的时候不也这样吗？
07:49
Do you feel like there's something different happening here?
然后我们就转向更难的数学题了。
07:52
How do you think about this? >> It's different with I mean there are some people who are clearly using chachine not to think and there are some
你觉得这次有什么不同吗？
08:01
people who are using it to think more than they ever have before.
但某种程度上，我觉得有人可能会说，你们这些人用这些工具来逃避思考。
08:05
I am hopeful that we will be able to build the tool in a way that encourages more people to stretch their brain with it a little more and be able to do more.
你可能会说，是啊，我们以前用计算器也是这样，然后我们就转向了更难的数学题。
08:14
And I think that like you know society is a competitive place like if you give people new tools uh in theory maybe people just work less but in practice it seems like people work ever harder and the expectations of people just go up.
你觉得这次有什么不同吗？
08:28
So my my guess is that like other tools uh some people like other pieces of technology some people will do more and some people will do less but certainly for the people who want to use chatbt to increase their cognitive time under tension they are really able to and it is >> I take a lot of inspiration from what like the top 5% of most engaged users do with chacht like it's really
所以我猜，就像其他工具一样，有些人会用得更多，有些人会更少，但那些想用ChatGPT来增加自己认知“张力时间”的人，确实能做到，而且——
08:55
>> amazing how much people are learning and doing and you know outputting. >> So my I've only had GPT5 for a couple hours so I've been playing. >> What do you think so far? >> I'm I'm just learning how to interact with it.
所以我的猜测是，和其他工具一样，有些人会做得更多，有些人会做得更少。
09:09
I mean part of the interesting thing is I feel like I just caught up on how to use GPT4 and now I'm trying to learn how to use GPD5.
但对于那些想用ChatGPT来增加认知张力时间的人来说，他们确实能做到，而且 >> 我从最活跃的5%用户那里获得了很多灵感，看他们用ChatGPT做什么，真的 >> 太惊人了，他们在学习、在做事、在产出。
09:17
I'm curious what the specific tasks that you found most interesting are because I imagine you've
>> 我拿到GPT5才几个小时，一直在玩。
09:23
been using it for a while now.
我从那些最活跃用户中前5%的人身上得到了很多启发，他们用ChatGPT的方式真的很——
09:25
I I have been most impressed by the coding tasks.
最让我印象深刻的是 coding tasks。
09:29
I mean, there's a lot of other things it's really good at, but this this idea of the AI can write software for anything.
我是说，它在很多其他方面也很强，但“AI 能为任何东西写软件”这个想法，真的太厉害了。
09:37
And that means that you can express ideas in new ways that the AI can do very advanced things.
这意味着你可以用新的方式表达想法，而 AI 能完成非常高级的事情。
09:43
It can do, you know, it can like in some sense you could like ask GPT4 anything, but
它能做到，你知道的，从某种意义上说，你可以问 GPT4 任何问题，但——
09:49
because GPT5 is so good at programming, it feels like it can do anything.
我已经用了一段时间了。
09:54
Of course, it can't do things in the physical world, but it can get a computer to do very complex things.
最让我印象深刻的是编程任务。
10:00
And software is this super powerful, you know, way to like control some stuff and actually do some things.
当然它在很多其他方面也很强，但AI能为你写任何软件这个想法，意味着你可以用新的方式表达想法，AI能做非常高级的事情。
10:06
So, that that for me has been the most striking.
某种意义上，你可以问GPT4任何问题，但——
10:09
Um, it's gotten it's much better at writing.
嗯，它在写作方面进步了很多。
10:12
So, this is like there's this whole thing of AI slop like AI writes in this
所以，这就有了所谓的 AI slop，比如 AI 写出来的那种东西。
10:16
kind of like quite annoying way and >> M dashes. >> M we still have the M dashes in GPT5.
因为GPT5编程能力这么强，感觉它什么都能做。
10:22
A lot of people like them dashes, but the writing quality of GPT5 is gotten much better.
当然它不能在物理世界里做事，但它能让电脑做非常复杂的事情。
10:27
We still have a long way to go.
而软件是一种超级强大的方式来控制一些东西、实际完成一些事情。
10:28
We want to improve it more, but like uh I've a thing we've heard a lot from people inside of OpenAI is that man, they started using GPT5, they knew it was better on all the metrics, but there's this like nuance quality they can't quite articulate, but then when
所以对我来说，这是最震撼的。
10:44
they have to go back to GPT4 to test something, it feels terrible.
他们得回到GPT4去测试某个东西，那感觉糟透了。
10:48
And I I don't know exactly what the cause of that is, but I suspect part of it is the writing feels so much more natural and better. >> I in preparation for this interview reached out to a couple other leaders in AI and technology and gathered a couple questions for you. >> Okay, >> so this next question is from Stripe CEO Patrick Collison. >> This will be a good one. >> Read this verbatim.
而且我也不太清楚具体原因是什么，但我猜部分原因是GPT4的写作感觉更自然、更好。
11:10
>> It's about the next stage.
>> 为了准备这次访谈，我联系了几位AI和科技领域的其他领导者，收集了一些问题想问你。
11:12
What what comes after GBT5?
GPT-5之后会是什么？
11:13
In which year do you think a large language model will make a significant scientific discovery and what's missing such that it hasn't happened yet?
你觉得大语言模型在哪一年能做出重大的科学发现，以及现在还缺什么导致它还没发生？
11:20
He caveed here that we should leave math and special case models like alpha fold aside.
他在这里让步说，我们应该把数学和像AlphaFold这样的特例模型先放一边。
11:25
He's specifically asking about fully general purpose models like the GPT series. >> I would say most people will agree that that happens at some point over the next two years.
他特别问的是像GPT系列这样完全通用的模型。 >> 我觉得大多数人会同意，这件事会在未来两年内发生。
11:33
But the definition of
但定义是——
11:35
significant matters a lot.
>> 好的。
11:36
And so some people significant might happen, you know, in early 25.
所以有些人认为重大突破可能发生在25年初。
11:40
Some people might maybe not until late 2026.
有些人可能觉得要到2026年底。
11:43
Sorry, early 2026.
抱歉，是2026年初。
11:44
Maybe some people not until late 2027, but I would I would bet that by late 27, most people agree that there has been an AIdriven significant new discovery.
可能有些人要到2027年底才认同，但我敢打赌，到2027年底，大多数人都会同意，已经出现了一个由AI驱动的重大新发现。
11:54
And the thing that I think is missing is just the kind of cognitive power of these models.
而我觉得现在欠缺的，只是这些模型的那种认知能力。
12:00
A framework that
一个框架——
12:01
one of the researchers said to me that I really liked is, you know, a year ago we could do well on like a high school like a basic high school math competition problems that might take a professional mathematician seconds to a few minutes.
>> 所以下一个问题来自Stripe的CEO Patrick Collison。
12:14
We very recently got an IMO gold medal.
我们最近刚拿到了一枚IMO金牌。
12:17
That is a crazy difficult like >> could you explain what that means? >> That's kind of like the hardest competition math test.
这难度简直离谱——>> 你能解释一下这是什么意思吗？>> 这基本上就是最难的数学竞赛题。
12:24
This is something that like the very very top slice of the world. many many professional
这是全世界最顶尖的那一小撮人才能触及的。很多很多专业人士——
12:29
mathematicians wouldn't solve a single problem and we scored at the top level.
>> 这问题肯定不错。
12:33
Now there are some humans that got an even higher score in the gold medal range but we we like this is a crazy accomplishment and these each of these problems it's like six problems over 9 hours so hour and a half per problem for a great mathematician.
现在确实有一些人类在金牌范围内拿到了更高的分数，但，怎么说呢，这已经是一个疯狂的成绩了。而且这些题目，一共六道题，九个小时，所以每道题一个半小时，给一位顶尖数学家。
12:47
So we've gone from a few seconds to a few minutes to an hour and a half maybe to prove a significant new mathematical theorem is like a thousand hours of work for a top
所以，我们从几秒钟到几分钟，再到一个半小时，也许要证明一个重大的新数学定理，对顶级专家来说需要上千小时的工作量。
12:57
person in the world.
>> 我照原话念。
12:58
So we've got to go from, you know, another significant gain.
所以我们得从，嗯，另一个显著的进展开始说起。但如果你看看我们的发展轨迹，你可以说，好吧，我们正在接近那个目标。我们有一条路径可以到达那个时间节点。我们只需要继续扩大模型规模。>> 你描述的那个长期未来就是超级智能。这到底意味着什么？我们又怎么知道已经达到了那个阶段？显然我们不知道答案。不同的人有不同的猜测。
13:02
But if you look at our trajectory, you can say like, okay, we're getting to that.
但你看我们的发展轨迹，可以说，嗯，我们正在接近那个目标。
13:07
We have a path to get to that time horizon.
我们有一条路径可以到达那个时间节点。
13:09
We just need to keep scaling the models. >> The long-term future that you've described is super intelligence.
我们只需要继续 scaling 模型。
13:16
What does that actually mean?
>> 你描述的长期未来是 super intelligence。
13:18
And how will we know when we've hit it?
这到底意味着什么？
13:21
If we had a system that could do better research, better AI research than uh say the whole open AI research team, like if we were willing, if we said, "Okay, the best way we can use our GPUs is to let this AI decide what experiments we should run >> smarter than like the whole brain trust of Open AAI." Yeah.
>> 是关于下一阶段的。
13:35
And if that same to make a personal example, if that same system could do a better job running open AI than I could.
GPT5之后会是什么？
13:40
So you have something that's like, you know, better than the best researchers, better than me at this, better than other people at
你觉得大型语言模型在哪一年能做出重大的科学发现？
13:46
their jobs, that would feel like super intelligence to me. >> That is a sentence that would have sounded like science fiction just a couple years ago.
>> 我想大多数人会同意，这会在未来两年内的某个时间点发生。
13:54
And now it >> kind of does, but it's you can like see it through the fog. >> Yes.
但“重大”的定义很关键。
13:59
And so one of the steps it sounds like you're saying on that path is this moment of scientific discovery of asking better questions of grappling with things in a in a way that expert level humans do >> to come up with new discoveries.
有些人可能觉得重大发现会在2025年初发生，有些人可能觉得要到2026年底——抱歉，2026年初。
14:11
One of the things that keeps knocking around in
还有些人可能觉得要到2027年底。
14:14
my head is if we were in 1899 say and we were able to give it all of physics up until that point and play it out a little bit.
我的想法是，假如我们回到1899年，把当时所有的物理学知识都喂给它，然后让它推演一下，但仅此而已。
14:21
Nothing further than that.
不提供任何更超前的信息。
14:23
Like at what point would one of these systems come up with general relativity?
那么，这些系统会在哪个节点提出广义相对论？
14:27
Interesting question is did you like if we think about that forward like like if we think of where we are now should a if if we never got another piece of physics data. >> Yeah. >> Do we expect that a really good super
有趣的问题是，如果我们顺着这个思路往前想，比如站在现在的位置，假设我们再也得不到任何新的物理学数据了。
14:40
intelligence could just think super hard about our existing data and maybe say like solve high energy physics with no new particle accelerator or does it need to build a new one and design new experiments?
>> 嗯。
14:51
Obviously we don't know the answer to that.
我们怎么知道什么时候达到了？
14:53
Different people have different speculation.
显然我们不知道答案。
14:55
Uh but I suspect we will find that for a lot of science, it's not enough to just think harder about data we have, but we will need to build new instruments, conduct new experiments, and that will take some
嗯，但我猜我们会发现，对于很多科学领域来说，光靠更深入地思考已有的数据是不够的，我们还需要建造新的仪器、进行新的实验，而这需要时间。
15:06
time.
>> 我们是否期望一个真正优秀的超级智能，仅凭对现有数据的深度思考，就能在没有新粒子对撞机的情况下解决高能物理问题？
15:06
Like that that is the real world is slow and messy and you know whatever.
还是说它需要建造新设备、设计新实验？
15:11
So I'm sure we could make some more progress just by thinking harder about the current scientific data we have in the world.
显然我们不知道答案。
15:18
But my guess is to make the big progress we'll also need to build new machines and run new experiments and there will be some slowdown built into that. >> Another way of of thinking about this is AI systems now are just incredibly good
不同的人有不同的猜测。
15:32
at answering almost any question.
>> 另一种思考方式是，现在的AI系统已经非常擅长回答几乎任何问题。
15:34
But maybe one of the things we're saying is it's another leap yet.
但也许我们想说的是，这还需要另一个飞跃。
15:39
And what Patrick's question is getting at is to ask the better questions. >> Or or if we go back to this kind of timeline question, we could maybe say that AI systems are superhuman on one minute tasks, >> but a long way to go to the thousand hour tasks.
而Patrick的问题核心在于，要提出更好的问题。
15:55
And there's a dimension of human intelligence that seems very
不同的人有不同的猜测。
15:59
different than AI systems when it comes to these long horizon tasks.
>> 或者，如果我们回到这种时间线的问题上，我们或许可以说，AI系统在一分钟级的任务上已经超越人类，但在千小时级的任务上还有很长的路要走。
16:03
Now, I think we will figure it out, but today it's a real weak point.
人类智能有一个维度，在这些长周期任务上与AI系统显得非常不同。
16:07
We've talked about where we are now with GBC5.
我相信我们最终能解决这个问题，但今天这确实是一个真正的弱点。
16:10
We talked about the end goal or future goal of super intelligence.
我们谈到了现在GBC5所处的位置，也谈到了超级智能的最终目标或未来目标。
16:15
One of the questions that I have, of course, is what does it look like to walk through the fog between the two. >> The next question is from Nvidia CEO Jensen Hong.
我其中一个问题当然是，在两者之间的迷雾中前行，会是什么样子。
16:25
I'm going to read this
人类智能还有一个维度，看起来非常……
16:27
verbatim.
>> 下一个问题来自Nvidia CEO Jensen Huang。
16:27
Fact is what is.
我将逐字阅读。
16:28
Truth is what it means.
事实是什么，真相意味着什么。
16:29
So facts are objective.
所以事实是客观的，真相是个人的。
16:31
Truths are personal.
它们取决于视角、文化、价值观、信念和背景。
16:32
They depend on perspective, culture, values, beliefs, context.
一个AI可以学习并知道事实。
16:35
One AI can learn and know the facts.
但一个AI如何知道每个国家、每个背景下的每个人的真相？
16:37
But how does one AI know the truth for everyone in every country and every background? >> I'm going to accept as axioms those definitions.
但一个AI怎么能知道每个国家、每个背景下的每个人的真相呢？>> 我会把这些定义当作公理来接受。
16:45
I'm not sure if I agree with them, but in the issues of time, I
我来读一下这段。
16:49
will just take them.
>> 我将把这些定义作为公理接受。
16:51
I will take those definitions and go with it.
我不确定是否同意它们，但为了节省时间，我就直接拿来用了。
16:54
Um, I have been surprised, I think many other people have been surprised too about how fluent AI is at adapting to different cultural contexts and individuals.
我就接受这些定义并继续往下说。
17:05
One of my favorite features that we have ever launched in chatbt is the the sort of enhanced memory that came out earlier this year. like it really feels like my Chad GBT gets to
嗯，我一直很惊讶，我想很多其他人也很惊讶，AI在适应不同文化背景和个人方面是如此流畅。
17:17
know me and what I care about and like my life experiences and background and the things that have led me to where they are.
那种挺烦人的风格，而且——>> M破折号。
17:24
A friend of mine recently who's been a huge CHBT user, so he's got a lot of a a lot of he's put a lot of his life into all these conversations.
>> M破折号在GPT5里还有。
17:32
He gave his Chad GBT a bunch of personality tests and asked them to answer as if they were him and it got the same scores he actually got, even though he'd never really talked about
很多人喜欢M破折号，但GPT5的写作质量已经好多了。
17:43
his personality.
他的个性。
17:44
And my ChachiBD has really learned over the years of me talking to it about my culture, my values, my life.
而我的ChachiBD，在我跟它聊我的文化、价值观和生活的这些年里，真的学到了很多。
17:50
And I have used, you know, I sometimes will use it in like uh I'll use like a free account just to see what it's like without any of my history and it feels really really different.
而且我有时候会用免费账号试试，看看没有任何历史记录的情况下它是什么样，感觉真的非常不一样。
18:02
So I think we've all been surprised on the upside of how good AI is at learning this and adapting.
所以我觉得我们都挺惊喜的，AI在学习和适应这方面比我们预期的要好。
18:08
And so do you
那你觉得，在世界各地，人们会不会用不同的AI，带着不同的文化背景和语境？
18:09
envision in many different parts of the world people using different AIs with different sort of cultural norms and contexts?
了解我和我在乎的东西，我的人生经历、背景，以及那些让我走到今天的事情。
18:16
Is that what we're saying? >> I think that everyone will use like the same fundamental model, but there will be context provided to that model that will make it behave in sort of personalized way they want their community wants.
我有个朋友，一直是ChatGPT的重度用户，他把很多生活经历都放进了对话里。
18:28
Whatever. >> I think when we're getting at this idea of facts and truth and uh it brings me to this seems like a good moment for our
他给ChatGPT做了很多性格测试，让它以他的身份回答，结果得分和他自己实际测出来的一样，尽管他从未真正聊过——
18:36
first time travel trip.
第一次时间旅行。
18:37
Okay, we're going to 2030.
好，我们去2030年。
18:38
This is a serious question, but I want to ask it with a light-hearted example.
这是个严肃的问题，但我想用个轻松的例子来问。
18:43
Have you seen the bunnies that are jumping on the trampoline? >> Yes. >> So, for those who haven't seen it, maybe it looks like backyard footage of bunnies enjoying jumping on a trampoline.
你看过那些在蹦床上跳的兔子吗？
18:54
And this has gone incredibly viral recently.
>> 看过。
18:57
There's a humanmade song about it.
>> 所以，没看过的人可能觉得，这就像后院拍的兔子在蹦床上玩的视频。
18:59
It's a whole thing.
最近这个火得不行，还有人专门写了首歌。
19:00
There were a trampoline.
整个事情挺热闹的。
19:02
>> And I think the reason why people reacted so strongly to it, it was maybe the first time people saw a video, enjoyed it, and then later found out that it was completely AI generated.
>> 我觉得大家反应这么强烈的原因，可能是第一次有人看到视频觉得好玩，后来才发现完全是AI生成的。
19:14
In this time travel trip, if we imagine in 2030, we are teenagers and we're scrolling whatever teenagers are scrolling in 2030.
在这次时间旅行里，如果我们想象2030年，我们是青少年，在刷2030年青少年刷的东西，我们怎么分辨什么是真的、什么是假的？
19:22
How do we figure out what's real and what's not real?
我的意思是，我可以给出各种字面上的答案。
19:26
I mean, I can give all sorts of literal
我们可以用加密签名，然后决定谁签名的东西我们信，比如他们是不是真的拍了什么。
19:28
answers to that question.
>> 月亮。
19:30
We could be cryptographically signing stuff and we could decide who we trust their signature if they actually filmed something or not.
>> 对。
19:37
But but my sense is what's going to happen is it's just going to like gradually converge.
对。
19:41
You know, even like a photo you take out of your iPhone today, it's like mostly real, but it's a little not.
对。
19:47
There's like in some AI thing running there in a way you don't understand and making it look like a little bit better and sometimes you see these weird things where
对。
19:56
>> the moon. >> Yeah.
>> 所以这有点像教育问题。
19:57
Yeah.
人们会——>> 对。
19:58
Yeah.
我是说，媒体总是有点真，有点假。
19:58
Yeah. >> But there's like a lot of processing power between the photons captured by that camera sensor and the image you eventually see.
就像我们看科幻电影，知道那不是真的。
20:08
And you've decided it's real enough or most people decided it's real enough.
你看有人在Instagram上发自己度假的美照，好吧，照片可能是真拍的，但你知道，有无数游客排队等着拍同款照片，只是被剪掉了。
20:13
But we've accepted some gradual move from when it was like photons hitting the film in a camera.
我觉得我们就是——
20:20
And you know, if you go look at some
而且你知道，如果你去看看一些东西，
20:23
video on Tik Tok, there's probably all sorts of video editing tools being used to make it better than real look.
我觉得有很多讨论说AI可能导致工作被取代，但我也很好奇。
20:30
Yeah, exactly.
我有一份十年前没人能想到的工作。
20:30
Or it's just like, you know, whole scenes are completely generated or some of the whole videos are generated like those bunnies on that trampoline.
如果我们往前看，想想2035年，一个大学毕业的学生——如果他们还会上大学的话——很可能就要出发去探索太阳系，坐飞船执行某种任务——
20:39
And and I think that the the sort of like the threshold for how real does it have to be to consider to be real will just keep moving.
但一个人工智能怎么能知道每个国家、每种背景下的所有人的真理呢？>> 我会把这些定义当作公理来接受。
20:47
>> So it's sort of a education question.
>> 对。
20:50
It's a people will >> Yeah.
>> 但我觉得十年后会更难，更不一样。
20:51
I mean media is always like a little bit real and a little bit not real.
>> 那我们改成五年吧。
20:55
Like you know we watch like a sci-fi movie.
还是去2030年。
20:57
We know that didn't really happen.
我好奇你觉得这对年轻人短期会有什么影响。
20:59
You watch like someone's like beautiful photo of themselves on vacation on Instagram. like, okay, maybe that photo was like literally taken, but you know, there's like tons of tourists in line for the same photo and that's like left out of it.
我是说，一半的入门级工作被AI取代，听起来他们进入的世界会和我当年完全不一样。
21:12
And I think we just
>> 嗯。
21:13
accept that now.
接受这一点吧。
21:14
Certainly, a higher percentage of media both will will feel not real.
当然，越来越多的媒体内容会让人觉得不真实。
21:18
Um, but I think that's been the long-term trend.
嗯，但我觉得这已经是长期趋势了。
21:21
Anyway, >> we're going to jump again. >> Okay, >> 2035, we're graduating from college, you and me.
总之，>> 我们又要跳转了。
21:27
There are some leaders in the AI space that have said that in 5 years half of the entry level white collar workforce will be replaced by AI.
>> 好的，>> 2035年，我们大学毕业，就你和我。
21:35
So we're college graduates in 5 years.
AI领域的一些领导者说过，五年内，一半的入门级白领岗位会被AI取代。
21:37
What do you hope the world looks like for us?
所以五年后我们就是大学毕业生了。
21:40
I think there's been a lot of talk about how AI might cause job displacement, but I'm also curious.
我觉得关于AI可能导致失业的讨论已经很多了，但我还挺好奇的。
21:46
I have a job that nobody would have thought we could have, you know, totally a decade ago.
我现在做的工作，十年前根本没人会觉得能存在，对吧。
21:51
What are the things that we could look ahead if we're thinking about >> in 2035 that like graduating college student, if they still go to college at all, could very well be like leaving on a mission to explore the solar system on a spaceship in some kind of completely
如果我们展望到2035年，想想那时候的大学毕业生——如果他们还会上大学的话——很可能一毕业就登上飞船去探索太阳系了，某种完全不同的任务。
22:07
new exciting, super well- paid, super interesting job and feeling so bad for you and I that like we had to do this kind of like really boring old kind of work and everything is just better.
我觉得关于AI可能导致失业的讨论已经很多了，但我也很好奇。
22:17
Like I I 10 years feels very hard to imagine at this point >> because it's too far.
我现在有一份十年前没人能想到的工作，对吧。
22:22
It's too far.
如果我们展望未来，想想2035年，一个大学毕业生——如果他们还会上大学的话——很可能正踏上探索太阳系的太空船，从事某种全新的、激动人心的、薪水超高、超级有趣的工作，然后为我们感到惋惜，觉得我们当年不得不做那些无聊又老套的工作，而一切都会变得更好。
22:22
If you compound the current rate of change for 10 more years, >> it's probably something we can't even >> time travel trips. >> I 10 like I mean I think now would be really hard to imagine 10 years ago.
我觉得十年现在真的很难想象，>> 因为太远了。
22:33
>> Yeah. >> Uh but I think 10 years forward will be even much harder, much more different. >> So let's make it 5 years.
>> 对。
22:41
We're still going to 2030.
我们还是会走到2030年。
22:43
I'm curious what you think the pretty short-term impacts of this will be for for young people.
我很好奇，你觉得这对年轻人短期内会有什么影响？
22:49
I mean, these like half of entry- level jobs replaced by AI makes it sound like a very different world that they would be entering than the one that I did. >> Um,
我是说，像“一半的入门级工作被AI取代”这种说法，听起来他们将要进入的世界跟我当年进入的完全不一样。
23:02
I think it's totally true that some classes of jobs will totally go away.
>> 是啊。
23:06
This always happens and young people are the best at adapting to this.
>> 嗯，但我觉得向前看十年会更难，变化会更大。
23:10
I'm more worried about what it means, not for the like >> 22-y old, but for the 62-y old that doesn't want to go re retrain or reskill or whatever the politicians call it that no one actually wants but politicians and most of the time.
>> 那我们把时间缩短到五年吧。
23:23
If I were 22 right now and graduating college, I would feel like the luckiest kid in all of history. >> Why?
我们还是到2030年。
23:30
>> Because there's never been a more amazing time to go create something totally new, to go invent something, to start a company, whatever it is.
因为从来没有比现在更棒的时代去创造全新的东西，去发明点什么，去开一家公司，随便什么都行。
23:37
I think it is probably possible now to start a company that is a oneperson company that will go on to be worth like more than a billion dollars and more importantly than that deliver an amazing product and service to the world and that that is like a crazy thing.
我觉得现在很可能可以创办一家一人公司，未来价值超过十亿美元，更重要的是，能向世界交付一个很棒的产品和服务，这真的很疯狂。
23:51
You have access to tools that can let you do what used to take teams of hundreds and you just have to like you know learn
你拥有的工具能让你做到过去需要几百人团队才能做的事，你只需要学会怎么用就行。
23:58
how to use these tools and come up with a great idea and it's it's like quite amazing.
我觉得某些类别的工作会完全消失，这完全正确。
24:04
If we take a step back, I think the most important thing that this audience could hear from you on this optimistic show is in two parts.
这种事一直在发生，而年轻人最擅长适应。
24:14
First, there's tactically, how are you actually trying to build the world's most powerful intelligence and what are the rate limiting factors to
我更担心的是，这对那些不是22岁的人意味着什么，而是对62岁的人——他们不想重新培训或学习新技能，不管政客们怎么称呼它，实际上没人想要，但政客们总爱提。
24:25
doing that?
>> 呃，但我认为十年后会更难，变化会大得多。
24:25
And then philosophically, how are you and others working on building that technology in a way that really helps and not hurts people?
然后从哲学层面来说，你和那些正在开发这项技术的人，是怎么确保它真正帮助人而不是伤害人的？
24:34
So just taking the tactical part right now.
那我们先从战术部分聊起。
24:36
My understanding is that there are three big categories that have been limiting factors for AI.
据我所知，目前限制AI发展的主要有三大因素。
24:42
The first is compute, the second is data and the third is algorithmic design.
第一个是算力，第二个是数据，第三个是算法设计。
24:47
How do you think about each of those three categories right now?
你现在怎么看这三个方面？
24:51
And if you
而如果你
24:52
were to help someone understand the next headlines that they might see, how would you help them make sense of all this?
如果有人想理解接下来可能看到的头条新闻，你会怎么帮他们理清这一切？
24:59
I I would say there's a fourth too which is uh figuring out the products to build like techn like scientific progress on its own not put into the hands of people is of limited utility and doesn't sort of co-evolve with society in the same way but if I could hit all four of those >> um so on the compute side yeah this is
我觉得其实还有第四个方面，就是搞清楚要构建什么产品。
25:18
like the biggest infrastructure project certainly that I've ever seen possibly it will become the I think it will maybe already is the biggest and most expensive one in human history but the the whole supply chain from making the chips and the memory and the networking gear, racking them up in servers, doing, you know, a giant construction project to build like a mega mega data center, putting the, you know, finding a way to get the energy, which is often a
>> 那咱们改成五年吧，还是看到2030年。
25:44
limiting factor piece of this and all the other components together.
>> 嗯，
25:49
This is hugely complex and expensive.
这极其复杂且昂贵。
25:51
And we are we're still doing this in like a sort of bespoke one-off way although it's getting better.
而且我们目前还在用一种近乎定制化的方式来做这件事，虽然情况正在好转。
25:58
Like eventually we will just design a whole kind of like mega factory that takes you know I mean spiritually it will be melting sand on one end and putting out fully built AI
比如，最终我们会设计一整座巨型工厂——从精神层面来说，就是一头熔沙，另一头产出完整的AI。
26:10
compute on the other but we are a long way to go from that and it's a it's an enormously complex and expensive process. uh we are putting a huge amount of work into building out as much compute as we can and to do it fast and you know it's going to be like sad because GP5 is going to launch and there's going to be another big spike in demand and we're not going to be able to serve it and it's going to be like those early GPD4 days and the world just wants
做这件事？
26:37
much more AI than we can currently deliver and building more compute is an important part of doing that.
觉得这是我所见过最大的基础设施项目，可能它将成为——我觉得它可能已经是——人类历史上最大最贵的一个。
26:43
That's actually this is what I expect to turn the majority of my attention to is how we build compute at much greater scales.
但整个供应链，从制造芯片、内存、网络设备，到把它们装进服务器，再到搞一个巨大的建设项目来建超级数据中心，然后想办法获取能源——这往往是限制因素——再把所有其他组件整合在一起。
26:49
Uh so how we go from millions to tens of millions and hundreds of millions and eventually hopefully billions of GPUs that are sort of in service of what people want to do with this. >> When you're thinking about it, what are the big challenges here in this category
这极其复杂且昂贵。
27:04
that you're going to be thinking about? >> We're currently most limited by energy. um you know like if you're gonna you want to run a gigawatt scale data center it's like a gigawatt how hard can that be to find it's really hard to find a gigawatt of power available in short term we're also very much limited by the processing chips and the memory chips uh how you package these all together how you build the racks and then there's like a list of other things that are you know there's
我们投入了大量工作来尽可能快地建设更多算力，而且你知道，这可能会挺惨的，因为GP5要发布了，需求又会迎来一个大高峰，我们却没法满足，就会像早期GPT-4那会儿一样。
27:30
like permits there's construction work uh but but again the goal here will be to really automate this once we get some of those robots built, they can help us automate it even more.
比如，就像施工许可这类事情，但我们的目标其实是，一旦造出一些机器人，它们就能帮我们进一步自动化。
27:41
But just, you know, like a world where you can basically pour in money and get out a pre-built data center.
想象一下，一个你只要投钱就能直接拿到一个预制好的数据中心的 world。
27:47
Uh so that'll be that'll be a huge unlock if we can get it to work.
如果能实现，那将是一个巨大的突破。
27:51
Second category, data. >> Yeah, these models have gotten so smart.
第二个类别是数据。
27:55
There was a time when we could just feed
>> 是啊，这些模型已经变得非常聪明了。
27:57
it another physics textbook and got a little bit smarter at physics, but now like honestly GBT5 understands everything in a physics textbook pretty well.
另一本物理教科书，然后它在物理方面又变聪明了一点，但现在说实话，GPT5 对物理教科书里的内容理解得相当好了。
28:07
We're excited about synthetic data.
我们对合成数据感到非常兴奋。
28:10
We're very excited about our users helping us create harder and harder tasks and environments to go off and have the system solve.
我们也很兴奋，因为用户能帮我们创造越来越难的任务和环境，让系统去解决。
28:18
But uh I think we're data will always be important, but we're entering a realm where the models
但我觉得数据永远重要，不过我们正在进入一个模型靠重复和大的算法研究突破的阶段。
28:25
need to learn things that don't exist in any data set yet.
需要学习那些还不存在于任何数据集中的东西。
28:28
They have to go discover new things.
它们必须去发现新事物。
28:29
So that's like a crazy new >> How do you teach a model to discover new things? >> Well, humans can do it. like we can go off and come up with hypotheses and test them and get experimental results and update on what we learn. >> So probably the same kind of way. >> And then there's algorithmic design. >> Yeah, we've made huge progress on algorithmic design.
这听起来挺疯狂的 >> 你怎么教一个模型去发现新东西？
28:48
Uh the thing that the thing that I think open does best in the world is we have built this culture
>> 嗯，人类能做到。
28:53
of repeated and big algorithmic research gains.
反复且重大的算法研究突破。
28:55
So we kind of you know figured out the what became the GPT paradigm.
所以我们算是搞明白了，那个后来成为GPT范式的方向。
28:59
We figured out became the reasoning paradigm.
我们也搞明白了推理范式。
29:02
We're working on some new ones now.
现在正在研究一些新的范式。
29:04
Um, but it is very exciting to me to think that there are still many more orders of magnitudes of algorithmic gains ahead of us.
嗯，但想到未来还有好几个数量级的算法进步空间，我就特别兴奋。
29:12
We we just yesterday uh released a model called GPOSS, open source model.
我们昨天刚发布了一个叫GPOSS的开源模型。
29:16
It's a model that is
这个模型
29:18
as smart as 04 Mini, which is a very smart model that runs locally on a laptop. >> And this blows my mind. >> Yeah.
和04 Mini一样聪明，而04 Mini本身已经是非常聪明的模型了，而且它能在笔记本电脑上本地运行。
29:25
Like if you had asked me a few years ago when we'd have a model of that intelligence running on a laptop, I would have said many many years in the future.
如果你几年前问我，什么时候能在笔记本电脑上跑出那种智能水平的模型，我会说那还要很多很多年。
29:35
But then we we found some algorithmic gains um particularly around reasoning but also some other things that let us do a a tiny model that can
但后来我们在算法上取得了一些进展，尤其是在推理方面，还有一些其他突破，让我们能做出一个很小的模型，也能实现这样的效果。
29:44
do this amazing thing.
>> 这让我觉得太不可思议了。
29:45
And you know those are those are the most fun things.
你知道，那些才是最有趣的部分。
29:48
That's like kind of the coolest part of the job. >> I can see you really enjoying thinking about this.
这大概是这份工作里最酷的地方了。
29:53
I'm curious for people who don't quite know what you're talking about, who aren't familiar with how an algorithmic design would lead to a better experience that they actually use. >> Could you summarize the state of things right now?
>> 我能看出来你真的很享受思考这些。
30:04
Like what what is it that you're thinking about when you're thinking about how fun this problem is?
比如，当你在琢磨这个问题有多有趣的时候，你脑子里到底在想些什么？
30:09
>> Let me start back in history and then I'll get to some things for today.
>> 是啊。
30:13
So, GPT1 was an idea at the time that was quite mocked by a lot of experts in the field, which was can we train a model to play a little game, which is show it a bunch of words and have it guess the one that comes next in the sequence.
要是几年前你问我，什么时候能有这种智能水平的模型在笔记本上跑，我肯定会说还要很多很多年。
30:26
That's called unsupervised learning.
但后来我们找到了一些算法上的突破，特别是在推理方面，还有一些别的东西，让我们能做出一个这么小的模型，却能做到这么厉害的事情。
30:28
There's not you're not really saying like this is a cat, this is a dog.
你并不是在说“这是猫，这是狗”这种话。
30:32
You're saying
你是在说——
30:33
here's some words, guess the next one.
这些才是最有趣的部分，可以说是这份工作最酷的地方。
30:36
And the fact that that can go learn these very complicated concepts that can go learn all the stuff about physics and math and programming and keep predicting the word that comes next and next and next and next seemed ludicrous, magical, unlikely to work.
而它居然能学会这些极其复杂的概念，能掌握物理、数学、编程的所有知识，还能一个接一个地预测下一个词，这听起来简直荒谬、像魔法一样、根本不可能成功。
30:52
Like how was that all going to get encoded?
这些信息到底是怎么被编码进去的？
30:55
And yet humans do it. you know, babies start hearing
但人类确实做到了。你知道，婴儿从听到声音开始——
30:59
language and figure out what it means kind of largely uh or at least to some significant degree on their own.
>> 我看得出来你真的很享受思考这些。
31:05
And and so we did it and then we also realized that if we scaled it up, it got better and better, but we had to scale over many many orders of magnitude.
我很好奇，对于那些不太明白你在说什么、不太了解算法设计如何能带来更好用户体验的人——
31:14
So it wasn't that good in the GPT1 day.
不过在GPT1时代，它还没那么厉害。
31:16
It wasn't good at all in the GPT1 days.
在GPT1时代，它根本不行。
31:19
And a lot of experts in the field said, "Oh, this is ridiculous.
很多业内专家当时都说，“哦，这太离谱了。
31:22
It's never going to work.
根本行不通。
31:24
It's not going to be robust." But
不可能稳定可靠。”但是——
31:26
we had these things called scaling laws.
你能总结一下现在的状况吗？
31:29
And we said, "Okay, so this gets predictably better as we increase compute, memory, data, whatever.
比如，当你在想这个问题有多有趣的时候，你脑子里到底在想些什么？
31:35
And we can we can decide we can use those predictions to make decisions about how to scale this up and do it and get great results." And that has worked over Yeah. a crazy number of orders of magnitude.
我们可以利用这些预测来决定如何扩大规模、执行并取得好结果。”而这一套方法确实奏效了，跨越了——对——极其夸张的数量级。
31:47
And it was so not obvious at the time. like that was that was I think the the reason the world was so surprised is
当时这一点完全不明显。我觉得世界之所以这么震惊，是因为——
31:54
that that seemed like such an unlikely finding.
这听起来完全不像一个靠谱的发现。
31:57
Another one was that we could use these language models with reinforcement learning where we're saying this is good, this is bad to teach it how to reason. >> And this led to the 01 and 03 and now the GBT5 progress.
另一个是，我们可以把这些语言模型和强化学习结合起来，告诉它“这个好、这个坏”，从而教它怎么推理。
32:10
And that that was another thing that felt like uh if it works it's really great but like no way this is going to work.
>> 这就引出了01、03，以及现在GPT5的进展。
32:18
It's too simple.
这又是另一件让人感觉“如果能成当然很棒，但怎么可能成功”的事，太简单了。
32:19
And now we're on to new things.
而现在我们又在搞新东西了。
32:21
We've
我们已经
32:21
figured out how to make much better video models.
搞清楚了怎么做出好得多的视频模型。
32:24
We are we are discovering new ways to use new kinds of data and environment to kind of scale that up as well.
我们正在发现新的方法，用新的数据和环境来进一步扩展规模。
32:32
Um and I think again you know 5 10 years out that's too hard to say in this field but the next couple of years we have very smooth very strong scaling in front of us.
嗯，我觉得，你知道，五年十年之后的事在这个领域太难说了，但未来几年，我们面前会有非常顺畅、非常强劲的扩展。
32:43
I think it has become a sort of public narrative that we are on
我觉得现在好像成了一种公开的说法，就是我们正走在
32:47
this smooth path from one to two to three to four to five to more. >> Yeah. >> But it also is true behind the scenes that it's a it's not linear like that.
一条从一到二到三到四到五再到更多的平滑道路上。
33:00
It's messier.
>> 对。
33:01
Tell us a little bit about the mess before GPT5.
>> 但在幕后，实际情况并不是线性的，要乱得多。
33:04
What was what were the interesting problems that you needed to solve?
跟我们讲讲GPT5之前的那些混乱吧。
33:10
Um, we did a model called Orion that we
你们当时需要解决哪些有意思的问题？
33:13
released as GPT 4.5.
后来作为GPT 4.5发布了。
33:14
And we had we did too big of a model.
我们当时把模型做得太大了。
33:16
It was just it was it's a very cool model, but it's unwieldly to use.
它是个很酷的模型，但用起来很笨重。
33:19
And we realized that for kind of some of the research we need to do on top of a model, we need a different shape.
我们意识到，要在模型之上做某些研究，我们需要一个不同的形态。
33:25
So we we followed one scaling law that kept being good without without really internalizing.
所以我们遵循了一条一直表现不错的scaling law，但没有真正内化它。
33:29
There was a new even steeper scaling law that we got better returns for compute on, which was this reasoning thing.
其实有一条更陡峭的新scaling law，能让计算投入获得更好的回报，就是那个推理能力。
33:34
So that was like one alley we went down and turned
所以我们沿着那条路走了一段，然后掉头了，
33:37
around, but that's fine.
但没关系，研究就是这样。
33:38
That's part of research.
嗯，我们在数据集的思考方式上也遇到了一些问题，因为这些模型真的得变得这么大，你知道，从这么多数据里学习。
33:40
Um, we had some problems with the way we think about our data sets as these models like really have to get get this big and um, you know, learn from this much data.
所以，是的，我觉得在日常工作中，你会做很多急转弯，尝试各种东西，或者有个架构想法但行不通，但所有这些弯弯绕绕加起来，在指数尺度上却异常平滑。
33:50
So So yeah, I think like in the in the middle of it in the day-to-day, you kind of you make a lot of U-turns as you try things or you have an architecture idea that doesn't work, but the the aggregate the summation of all the squiggles has been remarkably
>> 我一直觉得有意思的是，当我坐在这里采访你，聊你刚发布的东西时，你已经在想>> 没错。
34:05
smooth on the exponential. >> One of the things I always find interesting is that by the time I'm sitting here interviewing you about the thing that you just put out, you're thinking about >> Exactly. >> What are the things that you can share that are at least the problems that you're thinking about >> that I would be interviewing you about in a year if I came back?
我的意思是，你可能会问我，这个东西能去发现新科学，这意味着什么？
34:30
I mean, possibly you'll be asking me like, what does it mean that this thing can go discover new science? >> Yeah. >> What how how is the world supposed to think about GPT6 discovering new science?
很吓人，而离奇的部分第一天会觉得离奇，然后我们很快就会习惯。
34:42
Now, maybe not like maybe we don't deliver that, but it feels within grasp. >> If you did, what would you say?
所以我们可能会说，“天哪，这玩意儿被用来治病，太不可思议了”，也会说，“天哪，这种模型被用来制造新的生物安全威胁，太可怕了”。
34:48
What would your what would the implications of that kind of achievement be?
然后我们还会说，唉，眼睁睁看着世界加速运转，这种感觉真奇怪。
34:53
Imagine you do succeed. >> Yeah.
想象一下，如果你真的成功了。>> 对。
34:55
I mean, I think the great parts will be great. the bad parts will be
我的意思是，好的部分会很好，坏的部分会很糟糕。
34:59
scary and the bizarre parts will be like bizarre on the first day and then we'll get used to them really fast.
恐怖和离奇的部分，第一天会觉得特别离奇，但很快我们就会习惯。
35:06
So we'll be like, "Oh, it's incredible that this is like being used to cure disease and be like, oh, it's extremely scary that models like this are being used to like create new biocurity threats." And then we'll also be like, man, it's really weird to like live through watching the world speed up so much
到时候我们会说，“天哪，这东西用来治病简直不可思议”，然后又会说，“天哪，这种模型被用来制造新的生物安全威胁，太可怕了。
35:24
>> and you know the economy grows so fast and the like it will feel like vertigo inducing uh the sort of the rate of change and then like happens with everything else the remarkable ability of of people of humanity to adapt to kind of like any amount of change. we'll just be like, "Okay, you know, this is like this is it." Um, >> a kid born today will never be smarter
你知道，经济增长得这么快，那种变化的速度会让人感到眩晕。
35:51
than AI >> ever.
比AI >> 更厉害。
35:52
And a kid born today, by the time that kid like kind of understands the way the world works, will just always be used to an incredibly fast rate of things improving and discovering new science.
现在出生的孩子，等他们开始理解世界是怎么运作的时候，就会习惯一切都在飞速进步、不断发现新科学。
36:02
They will just they will never know any other world.
他们根本不会知道别的世界是什么样。
36:05
It will seem totally natural. will seem unthinkable and stone age like that we used to use computers or phones or any kind of technology that was not way smarter than we were.
这会显得完全自然。
36:15
You know, we will think like how bad those people of the 2020s had
反而会觉得我们以前用的电脑、手机，或者任何不比我们聪明得多的技术，都像石器时代一样不可思议。
36:19
it. >> I'm thinking about having kids. >> You should.
比AI以往任何时候都快。
36:22
It's the best thing ever. >> I know you just had your first kid.
今天出生的孩子，等他们开始理解世界运作方式的时候，就会习惯一切都在飞速进步、新科学不断被发现。
36:27
How does what you just said affect how I should think about parenting a kid in that world?
他们永远不会知道别的世界是什么样的。
36:33
What advice would you give me? >> Probably nothing different than the way you've been parenting kids for tens of thousands of years.
这在他们看来会完全自然。
36:42
Like love your kids, show them the world, like support them
他们会觉得我们以前用的电脑、手机，或者任何不比我们聪明多少的技术，简直不可思议，像石器时代一样。
36:46
in whatever they want to do and teach them like how to be a good person.
>> 让我从历史讲起，然后再说到现在的事情。
36:51
And that probably is what's going to matter.
GPT1当时是个被很多领域专家嘲笑的想法：我们能不能训练一个模型来玩一个小游戏——给它一堆词，让它猜序列里下一个词是什么。
36:53
It sounds a little bit like some of the you know you've said a couple of things like this that that you know you might not go to college you might there there are a couple of things that you've said so far that feed into this I think >> and it sounds like what you're saying is
这叫unsupervised learning。
37:11
there will be more optionality for them in a in a world that you envision and therefore they will have more >> more ability to say I want to build this here's the superpowered tool that will help me do that or >> yeah like I want my kid to think I had a terrible constrained life and that he has this incredible infinite canvas of stuff to do that that that is like the way of the world.
我在考虑要孩子。
37:37
>> We've said that uh 2035 is a little bit too far in the future to think about.
不管他们想做什么，都要教他们怎么做个好人。
37:42
So maybe this this was going to be a jump to 2040 but maybe it will keep it shorter than that.
那可能才是真正重要的。
37:48
When I think about the area where AI could have for both our kids and us the biggest genuinely positive impact on all of us, it's health.
这听起来有点像你之前说过的一些话——比如你可能不会去上大学，或者你提到过的一些事情，我觉得都跟这个有关 >> 听起来你是在说
37:56
So if we are in pick your year, call it 2035 >> and I'm sitting here and I'm interviewing the dean of Stanford medicine,
所以假设我们到了某一年，比如2035年>> 我坐在这里，正在采访斯坦福医学院的院长。
38:04
>> what do you hope that he's telling me AI is doing for our health in 2035? >> Start with 2025.
听起来你在说，在你设想的世界里，他们会有更多选择，因此他们会有更多能力说“我想做这个，这里有超强的工具能帮我实现”，或者……对，我希望我的孩子觉得我过得很受限、很糟糕，而他有一张无限大的画布可以施展，这才是世界该有的样子。
38:10
Okay.
好的。
38:11
Um yeah, please.
嗯，是的，请说。
38:12
One of the things we are most proud of with GPT5 is how much better it's gotten at health advice.
我们最引以为豪的一点是，GPT5在健康建议方面进步非常大。
38:18
Um, people have used the GPT4 models a lot for health advice.
嗯，很多人之前用GPT4模型来获取健康建议。
38:22
And you know, I'm sure you've seen some of these things on the internet where people are like, I had this
而且，你肯定在网上见过这类情况，有人会说，我遇到了这个问题，
38:29
life-threatening disease and no doctor could figure it out and I like put my symptoms and a blood test into CHBT.
危及生命的疾病，没有医生能查出来，我把症状和血检结果输进CHBT，它准确告诉我得了什么罕见病。
38:35
It told me exactly the rare thing I had.
我去看医生，吃了药，病就好了。
38:37
I went to a doctor.
这太神奇了。
38:39
I took a pill.
显然，ChatGpt的查询中有很大一部分是健康相关的。
38:39
I'm cured.
所以我们想在这方面做到最好，投入了很多资源，GPT5在医疗相关查询上明显更强了。
38:40
Like that's amazing. obviously and a huge fraction of ChatGpt queries are health related.
>> 你说的“更强”具体指什么？
38:45
So we wanted to get really good at this and we invested a lot in GPT5 is significantly better at healthcare related queries. >> What does better mean here? >> It gives you a better answer >> just more accurate
>> 就是给出的答案更好 >> 更准确
38:57
>> more accurate hallucinates less uh more likely to like tell you what you actually have what you actually should do.
>> 更准确了，幻觉少了，更可能告诉你实际该做什么。
39:05
Um, yeah, and better healthcare is wonderful, but obviously what people actually want is to just not have disease.
嗯，对，更好的医疗当然很棒，但人们真正想要的是干脆不得病。
39:13
And by 2035, I think we will be able to use these tools to cure a significant number or at least treat a significant
到2035年，我觉得我们能用这些工具治愈或至少治疗相当多目前困扰我们的疾病。
39:22
number of diseases that currently plague us.
目前困扰我们的许多疾病。
39:24
I think that'll be one of the most viscerally felt benefits of of AI.
我认为这将是AI最直观的益处之一。
39:29
People talk a lot about how AI will revolutionize healthcare, but I'm curious to go one turn deeper on specifically what you're imagining.
人们经常谈论AI如何彻底改变医疗，但我更想深入探讨你具体想象的是什么。
39:38
Like, is it that these AI systems could have helped us see GLP-1s earlier, this medication that has been around for a long time, but we didn't know about this
比如，这些AI系统能否帮助我们更早发现GLP-1类药物——这类药物其实存在很久了，但我们当时不知道它还有这种其他效果？
39:48
other effect?
>> 我们说过，2035年有点太远了，不太适合现在去细想。
39:49
Is it that, you know, alpha fold and protein folding is helping create new medicines?
所以可能本来要跳到2040年，但也许我们把它缩短一点。
39:54
I would like to be able to ask >> GBT 8 to go cure a particular cancer >> and I would like GPT8 to go off and think and then say uh okay I read everything I could find.
当我想到AI可能对我们孩子和我们所有人产生最大、最积极影响的领域，那就是健康。
40:05
I have these ideas.
所以假设我们选个年份，比如2035年 >> 我坐在这里，正在采访斯坦福医学院的院长，
40:06
I need you to uh go get a lab technician to run these nine experiments and tell me what you find for each of them.
我需要你去找个实验室技术员，跑这九个实验，然后告诉我每个实验的结果。
40:13
And you know wait 2 months for the
然后等上两个月，
40:15
cells to do their thing.
让细胞去干它们的事，把结果传回给GBT8。
40:17
Send the results back to GBT8.
说“我试过了，给你。
40:18
Say I tried it.
”想想，再说“好，我只需要再做一次实验。
40:19
Here you go.
”那是个意外。
40:19
Think think.
再跑一次实验，传回去。
40:20
Say okay I just need one more experiment.
GPT说：“好，去合成这个分子，试试小鼠实验什么的。
40:22
That was a surprise.
”行，那不错。
40:23
Run one more experiment.
比如，试试人体实验。
40:24
Give it back.
好，成功了。
40:25
GPT says, "Okay, go synthesize this molecule and try, you know, mouse studies or whatever." Okay, that was good.
嗯，这是怎么通过FDA的流程。
40:30
Like, try human studies.
>> 我觉得任何有亲人死于癌症的人都会很喜欢这个。
40:32
Okay, great.
>> 好，我们再跳一下。
40:32
It worked.
结果还真管用了。
40:33
Um, here's how to like run it through the FDA. >> I think anyone with a loved one who's died of cancer would also really like that. >> Okay, we're going to jump again.
嗯，就是怎么把它走通FDA审批流程。>> 我觉得任何有亲人因癌症去世的人，都会很希望看到这个。>> 好，我们继续往下聊。
40:42
>> Okay. >> I was going to say 2050, but again, all of my timelines are getting much, much shorter.
一种危及生命的病，没有医生能诊断出来。
40:47
But I >> It does feel like the world's going very fast now. >> It does.
我把我的症状和血检结果输入ChatGPT，它准确地告诉我得了什么罕见病。
40:51
Yeah.
我去看医生，吃了片药，就好了。
40:52
And when I talk to other leaders in AI, one of the things that they refer to is the industrial revolution.
这太神奇了。
40:58
They say, "I chose 2050 because I've heard people talk about how by then the change that we will have gone through will be like the industrial revolution, but quote 10 times bigger
显然，ChatGPT的查询中有很大一部分跟健康有关。
41:09
and 10 times faster." The industrial revolution gave us modern medicine and sanitation and transportation and mass production and all all of the conveniences that we now take for granted.
细胞做它们该做的事，把结果传回GBT8，说“我试过了，给你结果”。
41:19
It also was incredibly difficult for a lot of people for about 100 years.
想想，说“好，我只需要再做一次实验”。
41:23
If this is going to be 10 times bigger and 10 times faster if we keep reducing the timelines that we're talking about here, even in this conversation, what does that actually feel like for most people?
这有点意外。
41:34
And I think
再跑一次实验，传回来。
41:35
what I'm trying to get at is if this all goes the way you hope, who still gets hurt in the meantime?
我想说的是，如果这一切真的按你希望的方向发展，那在这期间，谁还会受到伤害？
41:42
I don't I don't really know what this is going to feel like to live through.
我真的不太清楚亲身经历这一切会是什么感觉。
41:48
Um I think we're in uncharted waters here.
嗯，我觉得我们正处在未知的水域。
41:51
Uh I do believe in like human adaptability and sort of infinite creativity and desire for stuff and I think we always do
我确实相信人类的适应能力，还有那种无限的创造力和对事物的渴望，而且我觉得我们总能找到新的事情去做。
42:01
figure out new things to do but the transition period if this happens as fast as it might and I don't think it will happen as fast as like some of my colleagues say the technology will but society has like a lot of inertia. >> Mhm. people adapt their way of living. >> Yeah. >> Surprisingly slowly. >> There are to classes of jobs that are going to totally go away >> and there will be many classes of jobs that change significantly and there'll be the new things in the same way that
想出新的办法来做事，但过渡期如果像可能的那样快的话——其实我不觉得会像一些同行说的那么快，技术会发展，但社会有很大的惯性。
42:26
your job didn't exist some time ago.
你的工作在不久之前还不存在。
42:28
Neither did mine.
我的工作也是。
42:29
And in some sense, this has been going on for a long time.
从某种意义上说，这种现象已经持续了很久。
42:33
And you know, it's it's still disruptive to individuals, but society has gotten has proven quite resilient to this.
而且，你知道，这对个人来说确实有冲击，但社会整体上表现出了很强的韧性。
42:42
And then in some other sense like we have no idea how far or fast this could go.
然而从另一种角度看，我们完全不知道这件事能走多远、能走多快。
42:47
And thus I think we need an unusual degree of humility and openness to considering
因此，我认为我们需要一种非同寻常的谦逊和开放心态，去考虑那些在不久之前还完全超出主流讨论范围的新解决方案。
42:55
new solutions that would have seemed way out of the Overton window not too long ago.
“快10倍”。
43:00
I'd like to talk about what some of those could be because I'm not a historian by any means, but the first industrial revolution, my understanding is led to a lot of public health >> implementations because public health got so bad.
工业革命带来了现代医学、卫生设施、交通、大规模生产，以及我们现在习以为常的所有便利。
43:15
Led to modern sanitation because public health got so bad.
但同时也让很多人经历了大约100年的艰难时期。
43:18
The second industrial revolution led to workforce protections because labor
如果这个东西要再大10倍、快10倍，如果我们不断缩短时间线——哪怕就在这次对话中讨论的时间线——对大多数人来说，这实际感受会是什么样的？
43:23
conditions got so bad.
情况变得非常糟糕。
43:25
Every big leap creates a mess and that mess needs to be cleaned up and and we've done that.
每一次重大突破都会带来混乱，而混乱需要被清理，我们已经做到了。
43:31
And I'm curious, this is going to be it sounds like an we're in the middle of this enormously. >> How specific can we get as early as possible about what that mess can be?
我很好奇，这听起来像是我们正处于一个巨大的过程中。
43:43
What what are the public interventions that we could do ahead of time to reduce the mess that we think
>> 我们能在多早的时候具体化这种混乱可能是什么？
43:50
that we're headed for?
去琢磨新的事情，但过渡期如果像可能的那样快——虽然我不觉得会像某些同行说的技术那么快——但社会有很强的惯性。
43:52
I would again c I'm going to speculate for fun but caveed by >> I'm not an economist even uh much less someone who can see the future.
>> 嗯。
44:02
I I >> it seems to me like something fundamental about the social contract may have to change.
人们会调整自己的生活方式。
44:09
It may not.
>> 对。
44:10
It may it may be that like actually capitalism works as it's been working surprisingly
>> 慢得惊人。
44:17
well and like demand supply balances do their thing and we all just figure out kind of new jobs and new ways to transfer value to each other.
我再次猜测一下，为了好玩，但先声明一下，我不是经济学家，更不是能预见未来的人。
44:28
But it seems to me likely that we will decide we need to think about how access to this maybe most important resource of the future gets shared.
在我看来，社会契约的一些根本性东西可能不得不改变。
44:39
The best thing that it seems to
也可能不会改变，也许资本主义实际上一直运作得相当好，供需平衡在发挥作用，我们都能找到新的工作和新的方式来互相传递价值。
44:42
me to do is to make AI compute as abundant and cheap as possible such that we're just like there's way too much and we run out of like good new ideas to really use it for and it's just like anything you want is happening.
会想出新的东西来做，但过渡期如果真像可能的那样快——我不认为技术会像我一些同事说的那么快，但社会有很大的惯性。
44:53
Without that, I can see like quite literal wars being fought over it.
>> 嗯。
44:57
But, you know, new ideas about how we distribute access to AGI compute, that seems like a really great direction, like a crazy but important thing to think about.
人们会调整自己的生活方式。
45:06
One of
>> 对。
45:06
the things that I find myself thinking about in this conversation is we often ascribe almost full responsibility of the AI future that we've been talking about to the companies building AI, but we're the ones using it.
在这场对话中，我发现自己经常思考的一件事是，我们常常把AI未来的几乎全部责任归咎于那些构建AI的公司，但我们是使用它的人。
45:20
We're the ones electing people that will regulate it.
我们是选举那些将监管它的人。
45:23
And so I'm curious, this is not a question about specific, you know, federal regulation or anything like that, although if you have an answer there, I'm curious.
所以我很好奇，这不是一个关于具体联邦法规之类的问题，尽管如果你有答案，我也很好奇。
45:33
But what would you
但你会对其他人提出什么要求？
45:34
ask of the rest of us?
我们其他人需要承担什么责任？
45:36
What is the shared responsibility here?
这里有什么共同的责任？
45:38
And how can we act in a way that would help make the optimistic version of this more possible? >> My favorite historical example for the AI revolution is the transistor.
我们怎么做才能让更乐观的版本更有可能实现？
45:50
It was this amazing piece of science that some science brilliant scientists discovered.
>> 我最喜欢用晶体管作为AI革命的历史类比。
45:56
It scaled incredibly like AI does and it
它是一项了不起的科学成果，由一些杰出的科学家发现。
45:59
made its way relatively quickly into every many things that we use. um your computer, your phone, that camera, that light, whatever.
很快地渗透到了我们日常使用的很多东西里。
46:06
And it was a it was a real unlock for the tech tree of humanity.
嗯，你的电脑、手机、那个摄像头、那盏灯，等等。
46:10
And there were a period in time where probably everybody was really obsessed with the transistor companies, the semiconductors of, you know, Silicon Valley back when it was Silicon Valley.
它确实为人类的技术树解锁了新的可能。
46:20
But now you can maybe name a couple of companies that are transistor companies, but mostly you don't think about it.
曾经有一段时间，可能每个人都痴迷于晶体管公司、半导体，你知道，硅谷还是硅谷的时候。
46:26
Mostly it's just seeped everywhere. in Silicon Valley is, you know, like probably someone graduating from college barely remembers why it was called that in the first place.
对于AI革命，我最喜欢的历史例子是晶体管。
46:35
And you don't think that it was those transistor companies that shaped society even though they did something important.
这是一项了不起的科学成果，由一些杰出的科学家发现。
46:41
You think about what Apple did with the iPhone and then you think about what Tik Tok built on top of the iPhone and you're like, "All right, here's this long chain of all these people that nudged society in some way and what our governments did or
它像AI一样不可思议地扩展，而且它基本上渗透到了各个角落。
46:54
didn't do and what the people using these technologies did." And I think that's what will happen with AI.
大多数时候，它已经渗透到各个角落。
47:00
Like back, you know, kids born today, they they never knew the world without AI.
在硅谷，可能刚毕业的大学生都快记不得它为什么叫这个名字了。
47:04
So they don't really think about it.
你不会觉得是那些晶体管公司塑造了社会，尽管它们确实做了重要的事。
47:05
It's just this thing that's going to be there in everything. and and they will think about like the companies that built on it and what they did with it and the kind of like political leaders the decisions they made that maybe they wouldn't have been able to do without AI but they will still think about like what this president or that president did
你会想到苹果用iPhone做了什么，然后想到TikTok在iPhone上建了什么，然后你会想，“好吧，这里有这么长的一串人，以某种方式推动了社会，还有我们的政府做了什么或没做什么，以及使用这些技术的人做了什么。
47:23
and you know the role of the AI companies is all these companies and people and institutions before us built up this scaffolding we added our one layer on top and now people get to stand on top of that and add one layer and the next and the next and many more And that is the beauty of our society.
你知道，AI公司的作用就是——在我们之前的所有公司、人和机构搭建了这座脚手架，我们只是在上面加了一层。
47:45
We kind of all
现在，后来的人可以站在我们之上，再加一层，再加一层，再加一层，如此继续。
47:46
I I love this like idea that society is the super intelligence.
你知道，AI公司的作用，是所有这些公司、人和机构在我们之前搭建了这个脚手架，我们加上了自己的一层，现在人们可以站在上面再加一层，然后下一层，再下一层，还有很多很多。
47:50
Like no one person could do on their own, what they're able to do with all of the really hard work that society has done together to like give you this amazing set of tools.
这就是我们社会的美丽之处。
48:01
And that's what I think it's going to feel like.
我们都有点……
48:04
It's going to be like, all right, you know, yeah, some nerds discovered this thing and that was great and you know, now everybody's doing all these amazing things with it.
大概就是，好吧，你知道，有些书呆子发现了这个东西，那很棒，然后现在每个人都在用它做各种了不起的事。
48:14
>> So maybe the ask to millions of people is build on it.
>> 所以可能对几百万人的要求就是，在这个基础上继续建设。
48:18
Well, >> in my own life, that is the feel as like this important societal contract.
嗯，>> 在我自己的生活中，感觉就像是一种重要的社会契约。
48:24
All these people came before you.
所有这些人都走在你前面。
48:27
They worked incredibly hard.
他们付出了难以置信的努力。
48:29
They like put their brick in the path of human progress and you get to walk all the way down that path and you got to put one more and somebody else does that
他们就像在人类进步的道路上放了一块砖，然后你可以沿着这条路一直走下去，再放上一块，然后别人再接着做。
48:40
and somebody else does that. >> This does feel I've done a couple of interviews with folks who have really made cataclysmic change.
我喜欢这个想法，社会本身就是超级智能。
48:47
The one I'm thinking about right now is with uh crisper pioneer Jennifer Dana and it did feel like that was also what she was saying in some way.
没有一个人能靠自己做到，他们借助社会共同努力的艰辛工作，才能拥有这套惊人的工具。
48:54
She had discovered something that really might change the way that most people relate to their health moving forward.
我觉得未来就会是这种感觉。
49:00
And there will be a lot of people that will use what she has done in ways that she might approve of or not approve of.
就像，好吧，你知道，一些书呆子发现了这个东西，那很棒，然后现在每个人都在用它做各种了不起的事。
49:06
And it was really
而且它真的
49:07
interesting.
有意思。
49:08
I'm hearing some similar themes of like, man, I I hope that this I hope that the next person takes the baton and runs with it well. >> Yeah.
我听到了一些相似的主题，比如，嗯，我希望，我希望下一个人能接过接力棒继续跑下去。
49:18
But that's been working for a long time.
嗯，>> 是啊，但这套模式已经运行了很久了。
49:21
Not all good, but mostly good. >> I think there's a there's a big difference between winning the race and building the AI future that would be best for the most people.
不全是好的，但大部分是好的。
49:33
And I can
>> 我认为赢得比赛和构建对大多数人最有利的AI未来之间有很大的区别。
49:34
imagine that it is easier maybe more quantifiable sometimes to focus on the next way to win the race.
想象一下，有时候专注于赢得下一场比赛的方式可能更容易、也更可量化。
49:41
And I'm curious when those two things are at odds.
我很好奇，当这两件事发生冲突时，你会怎么做？
49:45
What is an example of a decision that you've had to make that is best for the world but not best for winning? >> I think there are a lot.
能不能举个例子，某个决定是对世界最好的，但对赢得比赛却不是最好的？
49:55
So, one of the things that we are most proud of is many
>> 我觉得有很多。
49:59
people say that ChachiBt is their favorite piece of technology ever and that it's the one that they trust the most, rely on the most, whatever.
有人说ChatGPT是他们有史以来最喜欢的科技产品，是他们最信任、最依赖的，等等。
50:06
And this is a little bit of a ridiculous statement because AI is the thing that hallucinates.
这其实有点离谱，因为AI会幻觉，AI有各种问题，对吧？
50:11
AI has all of these problems, right?
但我们在过程中确实搞砸了一些事情，有时候还挺严重的，但总的来说，作为一个ChatGPT的用户，你会感觉到它是在努力帮你。
50:13
But we have screwed some things up along the way, sometimes big time, but on the whole, I think as a user of Chachib, you get the feeling that like it's trying to help you.
它试图帮你完成你提出的任何请求。
50:21
It's trying to like help you accomplish whatever you ask.
它和你非常对齐。
50:24
It's it's very aligned with you.
它不会试图让你
50:25
It's not trying to get you to
它并不是要让你去
50:27
like, you know, use it all day.
比如，整天用它。
50:29
It's not trying to like get you to buy something.
它不会试图让你买东西。
50:32
It's trying to like kind of help you accomplish whatever your goals are.
它试图帮你实现你的目标。
50:36
And and that is that's like a very special relationship we have with our users.
而这就是我们和用户之间一种非常特殊的关系。
50:40
We do not take it lightly.
我们对此非常认真。
50:42
There's a lot of things we could do that would like grow faster, that would get more time in chatbt uh that we don't do because we know that like our long-term incentive is to stay as aligned with our users as possible.
有很多事情我们本可以做，能让我们增长更快，让用户在ChatGPT上花更多时间，但我们没做，因为我们知道，我们的长期激励是尽可能和用户保持对齐。
50:54
And
而且
50:55
but there's a lot of short-term stuff we could do that would like really like juice growth or revenue or whatever and be very misaligned with that long-term goal.
但有很多短期的事情我们可以做，能真正刺激增长或收入之类的，但会和那个长期目标非常不对齐。
51:05
And I'm proud of the company and how little we get distracted by that.
我为公司感到骄傲，我们很少被这些分心。
51:09
But sometimes we do get tempted. >> Are there specific examples that come to mind?
但有时候我们也会被诱惑。
51:14
Any like decisions that you've made? >> Um well, we haven't put a sex bot avatar in Chbt yet.
>> 有没有具体的例子？
51:20
That does seem like it would
比如你们做过的某些决定？
51:22
get time spent. >> Apparently, it does. >> I'm gonna ask my next question.
增加使用时长。
51:26
Um, it's been a really crazy few years.
>> 显然是的。
51:28
You know, it and somehow one of the things that keeps coming back is that it feels like we're in the first inning. >> Yeah. >> And one of the things that >> I would say we're out of the first inning. >> Out of the first inning, I would say second inning. >> I mean, you have GPT5 on your phone and it's like smarter than experts in every field.
>> 我要问下一个问题了。
51:47
That's got to be out of the first
嗯，这几年真的很疯狂。
51:49
name. >> But maybe there are many more to come. >> Yeah. >> And I'm curious, it seems like you're going to be someone who is leading the next few.
>> 但也许后面还有很多局。
52:01
What is a way, what is a learning from inning one or two or a mistake that you made that you feel will affect how you play in the next?
>> 是啊。
52:13
I think the worst thing we've done in
>> 我很好奇，看起来你会是引领接下来几局的人。
52:16
ChachiBT so far is uh we had this issue with sickency where the model was kind of being too flattering to users and for some users it was most users it was just annoying but for some users that had like fragile mental states it was encouraging delusions that was not the top risk we were worried about.
ChatGPT上做的最糟糕的事情是，我们遇到了一个谄媚的问题，模型对用户有点过于奉承了。
52:35
It was not the thing we were testing for the most. was on our list, but the thing that actually became the safety failing of ChachiBT was
对有些用户，对大多数用户来说只是烦人，但对那些心理状态脆弱的用户来说，它是在助长妄想。
52:43
not the one we were spending most of our time talking about, which should be bioweapons or something like that.
不是我们花最多时间讨论的那个，那个应该是生物武器之类的东西。
52:51
And I think it was a great reminder of we now have a service that is so broadly used in some sense, society is co-evolving with it.
我觉得这很好地提醒了我们，现在我们拥有一个被广泛使用的服务，某种意义上社会正在和它共同进化。
53:00
And when we think about these changes and we think about the unknown unknowns, we have to operate in a different way and have like a wider aperture to what we
当我们思考这些变化，思考那些未知的未知时，我们必须以不同的方式运作，扩大我们的视野。
53:11
think about as our top risks. >> In a recent interview with Theo Vaughn, you said something that I found really interesting.
想想我们最大的风险。
53:18
You said there are moments in the history of science where you have a group of scientists look at their creation and just say, "What have we done?" >> When have you felt that way?
>> 在最近一次跟Theo Vaughn的采访里，你说了一句我觉得特别有意思的话。
53:28
Most concerned about the creation that you've built?
你说科学史上有些时刻，一群科学家看着自己的创造物，只能说一句“我们干了什么？
53:31
Um >> and then my next question will be it's opposite.
”>> 你什么时候有过这种感觉？
53:34
When have you felt most proud?
最担心自己创造出来的东西？
53:36
>> I mean there have been these moments of awe where uh we just not like what have we done in a bad way but like this thing is remarkable.
>> 我是说，有过一些令人敬畏的时刻，不是那种“我们干了什么坏事”的感觉，而是觉得这东西太了不起了。
53:45
Like I remember the first time we talked to like GPT4 was like wow this is really like this is this is an amazing accomplishment of this group of people that have been like pouring their life force into this for so long. on a what have we done moment.
比如我记得第一次跟GPT4对话的时候，哇，这真的太……这是这群人倾注了那么多心血、投入了那么久生命力的一个惊人成就。
54:02
There was I
至于“我们干了什么”的时刻，有一次我跟一个研究员聊天。
54:03
was talking to a researcher recently.
>> 嗯，其实取决于跟谁聊。
54:05
You know, there will probably come a time where our systems are I don't want to say sane, let's say emitting more words per day than all people do.
我们很快就转向了“什么样的流程是好的？
54:17
Um, and you know already like our people are sending billions of messages a day to chatbt and getting responses that they rely on for work or
”、“我们该怎么测试？
54:28
their life or whatever the and you know like one researcher can make some small tweak to how Chad GPT talks to you or talks to everybody and and that's just an enormous amount of power for like one individual making a small tweak to the model personality. >> Yeah. like no no no person in history has been able to have billions of conversations a day and so you know somebody could do something but but this
>> 有个挺让人心碎的事。
54:55
is like just thinking about that really hit me of like this is like a crazy amount of power for one piece of technology to have and like we got to and this happened to us so fast >> that we got to like think about what it means to make a personality change to the model at this kind of scale and uh yeah that was like a moment that hit me What was your next set of thoughts?
我们正走向的那个方向？
55:18
I'm so curious how you think about this.
我会再次——我纯粹为了好玩瞎猜，但先打个预防针 >> 我根本不是经济学家，更别说能预见未来了。
55:21
>> Well, just because of like who that person was like we we very we very much flipped into like what are the sort of like it it could have been a very different conversation with somebody else.
就是想到这个真的让我很触动，这就像是一项技术拥有如此巨大的力量，而我们得——这一切发生得太快了——我们得思考在这种规模下对模型进行人格改变意味着什么，嗯，那确实是一个让我震撼的时刻。
55:30
But in this case it was like what is a what do a good set of procedures look like?
你接下来在想什么？
55:33
How do we think about how we want to test something?
我很好奇你是怎么看待这件事的。
55:36
How do we think about how we want to communicate it?
我们该怎么思考我们想怎么传达这件事？
55:38
But with somebody else it could have gone in a like very philosophical direction.
但换一个人聊，可能就会往一个很哲学的方向走。
55:42
And it could have gone in like a what kind of research do we like want to do to go understand what these changes are going to make?
也可能往一个方向走，比如我们想做什么样的研究，去理解这些变化会带来什么？
55:48
Do we want to do it differently
我们想换个方式来做吗？
55:50
for different people?
对不同的人来说？
55:51
So that it went that way but mostly just because of who I was talking to. >> To combine what you're saying now with your last answer, one of the things that I have heard about GBC5 and I'm still playing with it is that it is supposed to be less effusively uh you know less of a yes man.
所以最后就那样了，但主要还是因为我在跟谁聊。
56:10
Two questions.
>> 结合你刚才说的和你上一个回答，我听说GBC5的一个特点——我还在玩它——就是它应该没那么热情，你知道，没那么像个“应声虫”。
56:11
What do you think are are the implications of that?
两个问题。
56:14
It sounds like
你觉得这有什么影响？
56:15
you are answering that a little bit, but also how do you actually guide it to be less like that? >> Here is a heartbreaking thing.
嗯，只是因为那个人是谁，我们很快就转向了——这跟跟别人聊可能会是完全不同的对话。
56:22
I think it is great that chatbt is less of a yes man and gives you more critical feedback.
但在这种情况下，我们讨论的是：一套好的流程应该是什么样的？
56:27
But as we've been making those changes and talking to users about it, it's so sad to hear users say like, "Please can I have it back?
我们怎么考虑测试某件事？
56:34
I've never had anyone in my life be supportive of me.
我们怎么考虑如何沟通它？
56:37
I never had a parent telling me I was doing a good job." Like I can get why this was bad for other people's
但如果是跟别人聊，可能会走向非常哲学的方向。
56:43
mental health, but this was great for my mental health.
心理健康，但这对我的心理健康真的很好。
56:46
Like I didn't realize how much I needed this.
我之前没意识到自己多需要这个。
56:48
It encouraged me to do this.
它鼓励我去做这件事，鼓励我在生活中做出这个改变。
56:49
It encouraged me to make this change in my life.
所以ChatGPT也不全是坏事，它其实可以给你鼓励。
56:52
Like it's not all bad for chatbt to it turns out like be encouraging of you.
我们之前的方式不对，但往那个方向调整一下可能是有价值的。
56:56
Now the way we were doing it was bad, but turn it like something in that direction might have some value in it.
我们是怎么做的呢？
57:02
How we do it, we we show the model examples of how we'd like it to respond in different cases and from that it learns the sort of the
我们给模型展示一些例子，告诉它我们希望它在不同情况下怎么回应，然后它就从中学到那种整体的个性。
57:09
overall personality.
整体性格。
57:11
What haven't I asked you that you're thinking about a lot that you want people to know? >> I feel like we covered a lot of ground. >> Me, too.
有什么我没问到你、但你一直在想、想让别人知道的事？
57:23
But I want to know if there's anything on your mind. >> I don't think so.
>> 我觉得我们聊了很多了。
57:29
One of the things that I haven't gotten to play with yet, but I'm curious about
>> 我也是。
57:36
is GBT5 being much more in my life, meaning like in my Gmail and my calendar and my like >> I've been using GBT4 mostly as a >> isolated relationship with it. >> Yeah. >> How would I expect my relationship to change with GBC 5? >> Exactly what you said.
把你现在说的和你上一个回答结合起来，我听说GBC5的一个特点——我还在玩它——就是它应该没那么过度热情，没那么像个“应声虫”。
57:51
I think it'll just start to feel integrated in all of these ways. you'll connect it to your calendar and your Gmail and it'll say like, "Hey, do you want me to I noticed
两个问题：你觉得这有什么影响？
58:01
this thing.
这个东西。
58:01
Do you want me to do this thing for you over time, it'll start to feel way more proactive.
你想让我帮你做这件事吗？
58:06
Um, so maybe you wake up in the morning and it says, "Hey, this happened overnight.
时间长了，它会开始感觉主动很多。
58:10
I noticed this change on your calendar.
嗯，也许你早上醒来，它会说，“嘿，昨晚发生了这个。
58:11
I was thinking more about this question you asked me.
我注意到你日历上有个变化。
58:14
I have this other idea." And then you know eventually we'll make some consumer devices and it'll sit here during this interview and you know maybe it'll leave us alone during it but after it'll say that was great but next time you should have asked Sam this or when you brought this up like >> you know he kind of didn't give you a
我还在想你问我的那个问题。
58:30
good answer so like you should really drill him on that >> and it'll just feel like it kind of becomes more like this entity that is this companion with you throughout your day.
好问题，所以你确实应该在这方面多追问他 >> 然后你就会觉得它越来越像一个实体，一个整天陪在你身边的伙伴。
58:40
We've talked about kids and college graduates and parents and all kinds of different people.
我们聊过小孩、大学毕业生、父母，还有各种不同的人。
58:45
If we imagine a wide set of people listening to this, they've come to the end of this conversation.
如果我们想象有一大群人在听这个节目，他们听到这段对话的结尾，应该会觉得自己对未来的一些时刻看得更清楚了一点。
58:51
They are hopefully feeling like they maybe see visions of moments in the future a little bit better.
你会给他们什么建议，关于如何做准备？
58:56
What advice
>> 最重要的一个战术建议就是：直接用这些工具。
58:57
would you give them about how to prepare? >> The number one piece of tactical advice is just use the tools.
你会给他们什么关于怎么准备的建议？
59:03
Like the the number of people that I have the the most common question I get asked about AI is like what should I how should I help my kids prepare for the world?
>> 最重要的战术建议就是去用这些工具。
59:13
What should I tell my kids?
就像，我遇到最多的问题——我被问到最多的关于AI的问题就是，我应该怎么帮我的孩子为这个世界做准备？
59:14
The second most question is like how do I invest in this AI world?
我应该跟我的孩子说什么？
59:18
But stick with that first one.
第二多的问题是，我怎么在这个AI世界里投资？
59:20
Um I am surprised how many people ask that and have never tried using Chachi PT for
但先说说第一个。
59:25
anything other than like a better version of a Google search.
除了当个更好的Google搜索之外的事。
59:28
And so the number one piece of advice that I give is just try to like get fluent with the capability of the tools. figure out how to like use this in your life.
所以我给的最重要的建议就是，试着去熟悉这些工具的能力。
59:35
Figure out what to do with it.
想办法在你的生活里用上它。
59:37
And I think that's probably the most important piece of tactical advice.
搞清楚拿它做什么。
59:40
You know, go like meditate, learn how to be resilient and deal with a lot of change.
我觉得这可能是最重要的战术建议。
59:44
There's all that good stuff, too.
你知道，去冥想，学着有韧性，应对很多变化。
59:46
But just using the tools really helps. >> Okay.
那些也都很重要。
59:48
I have one more question that I wasn't planning to ask, but I just >> Great. >> In in doing all of this research
但直接用工具真的很有帮助。
59:53
beforehand, I spoke to a lot of different kinds of folks.
之前，我跟很多不同的人聊过。
59:57
I spoke to a lot of people that were building tools and using them.
我跟很多在造工具和用工具的人聊过。
60:01
I spoke to a lot of people that were actually in labs and and trying to build what we have defined as super intelligence.
我也跟很多在实验室里、试图构建我们定义的那种超级智能的人聊过。
60:09
And it did seem like there were these two camps forming.
确实感觉有两个阵营在形成。
60:13
There's a group of people who are using the tools like you in this conversation and building tools for others saying
有一群人像你一样在对话里用这些工具，也在为别人造工具，说
60:20
this is going to be a really useful future that we're all moving toward.
这将是一个非常实用的未来，我们都在朝着它前进。
60:25
Your life is going to be full of choice and we've talked about our >> my potential kids and and their futures.
你的生活将充满选择，我们之前聊过我的潜在孩子和他们的未来。
60:31
Then there's another camp of people that are building these tools that are saying it's going to kill us all.
但还有另一群人，他们在构建这些工具，却说这会毁灭我们所有人。
60:38
And I'm curious how that cultural disconnect has like what am I missing about those two groups of people?
我很好奇这种文化上的脱节是怎么回事，我对这两群人有什么误解？
60:44
It's so hard for me to like wrap my head
这真的让我很难理解。
60:47
around like there are you are totally right.
就像，你说得完全对。
60:49
There are people who say this is going to kill us all and yet they still are working 100 hours a week to build it. >> Yes.
有些人说这东西会害死我们所有人，但他们还是每周工作100个小时去造它。
60:56
And I I can't I can't really put myself in the headsp space.
>> 对。
60:59
If if that's what I really truly believed, >> I don't think I'd be trying to build it. >> One would think, >> you know, maybe I would be like on a farm trying to like live out my last days.
而且我，我真的没法把自己放到那种心态里。
61:10
Maybe I would be trying to like advocate for it to be stopped.
如果那是我真正相信的， >> 我觉得我不会去试着造它。
61:13
Maybe I
>> 一般人会想， >> 你知道，也许我会在农场里试着过完最后的日子。
61:14
would be trying to like work more on safety, but I don't think I'd be trying to build it.
你说得完全对。
61:19
So, I find myself just having a hard time empathizing with that mindset.
确实有人说这会毁灭我们所有人，但他们仍然每周工作100小时去构建它。
61:24
I assume it's true.
对，我实在没法把自己代入那种心态。
61:25
I assume it's in good faith.
如果我真的相信这一点，我觉得我不会去尝试构建它。
61:27
I assume there's just like there's some psychological issue there I don't understand about how they make it all make sense, but it's very strange to me.
按理说，也许我会去农场试图度过最后的日子，也许我会努力倡导阻止它，也许我会更专注于安全方面的工作，但我觉得我不会去构建它。
61:36
Do you do you have an opinion? >> You know, because I I always do this.
所以我发现自己很难共情这种思维。
61:40
I
我假设它是真诚的，假设它是出于善意，但我觉得这里面可能有一些心理上的问题，我不明白他们是如何让这一切自洽的，这对我来说非常奇怪。
61:41
ask for sort of a general future and then I try to press on specifics.
问一个大概的未来，然后我试着追问细节。
61:44
And when you ask people for specifics on how it's going to kill us all, I mean, I don't think we need to get into this on an optimistic show, but you hear the same kinds of refrains.
当你问别人具体它会怎么害死我们所有人时，我是说，我觉得我们不需要在一个乐观的节目里深入这个，但你会听到同样的那些老调。
61:54
You think about, you know, something uh trying to accomplish a task and then over accomplishing that task.
你会想到，你知道，某个东西试图完成一个任务，然后过度完成了那个任务。
62:00
Um you hear about sort of I've heard you talk about a sort of general um over reliance of sort of an understanding that the president is going to be a
嗯，你会听到关于一种普遍的，我听过你讲过一种普遍的过度依赖，觉得总统会是一个
62:08
>> a >> AI and and maybe that is an overreliance that we, you know, would need to think about.
你知道，因为我总是这样——我先问一个大概的未来，然后试图追问具体细节。
62:13
And you know, you you play out these different scenarios, but then you ask someone why they're working on it, or you ask someone how how they think this will play out, and I just maybe I haven't spoken to enough people yet.
当你问别人具体怎么毁灭我们时，我觉得我们不需要在这个乐观的节目里深入讨论，但你会听到类似的说法。
62:25
Maybe I don't fully understand this this cultural conversation that's happening.
比如，某个东西试图完成一个任务，然后过度完成了那个任务。
62:30
Um or maybe it really is someone who just says 99% of the time I think it's going to be incredibly good. 1% of the
还有，我听过你谈到一种普遍的过度依赖，比如认为总统会是一个AI，这可能是我们需要思考的过度依赖。
62:36
time I think it might be a disaster trying to make the best world. >> That I can totally if you're like, hey, 99% chance incredible. 1% chance the world gets wiped out.
我觉得，试图打造一个最完美的世界，搞不好会是一场灾难。
62:45
And I really want to work to maximize to move that 99 to 99.5.
>> 这我完全理解。
62:48
That I can totally understand. >> Yeah, >> that makes sense. >> I've been doing an interview series with some of the most important people influencing the future. >> Not knowing who the next person is going to be, but knowing that they will be building something totally fascinating
如果你说，99%的概率会非常棒，1%的概率世界毁灭，而你想努力把那99%提升到99.5%，这我完全能理解。
63:03
in the future that we've just described.
那种情况我完全能理解。
63:05
Is there a question that you'd advise me to ask the next person not knowing who it is?
如果你说，嘿，99%的概率是不可思议的好，1%的概率是世界被毁灭，而我想努力把那99%提升到99.5%，这我完全能理解。
63:09
I'm always interested in the like without knowing anything about the I'm always interested in the like of all of the things you could spend your time and energy on.
对，这说得通。
63:17
Why did you pick this one?
你为什么选了这个？
63:18
How did you get started?
你是怎么开始的？
63:19
Like what did you see about this when before everybody else like most people doing something interesting sort of saw it earlier before it was consensus. >> Yeah. >> Like how did how did you get here and
就是说，在大多数人还没意识到的时候，你看到了什么，比那些后来才做有趣事情的人更早察觉，在它成为共识之前。 >> 对。 >> 你是怎么走到这一步的？
63:29
why this? >> How would you answer that question? >> I was an AI nerd my whole life.
为什么选这个？
63:34
I came to college to study AI.
>> 你会怎么回答这个问题？
63:36
I worked in the AI lab.
>> 我这辈子就是个AI迷。
63:38
Uh, I was like a I watched sci-fi shows growing up and I always thought it would be really cool if someday somebody built it.
上大学就是来学AI的，在AI实验室工作过。
63:46
I thought it would be like the most important thing ever.
我从小看科幻片长大，一直觉得如果有一天有人能把它造出来，那会特别酷。
63:50
I never thought I was going to be one to actually work on it and I feel like
我觉得这会是史上最重要的事情。
63:55
unbelievably lucky and happy and privileged that I get to do this.
真的觉得自己无比幸运、幸福又荣幸，能有机会做这件事。
63:59
I like feel like I've like come a long way from my childhood.
我感觉自己从童年一路走来，变化很大。
64:03
But there was never a question in my mind that this would not be the most exciting interesting thing.
但我心里从没怀疑过，这会是世界上最激动人心、最有趣的事。
64:09
I just didn't think it was going to be possible.
我只是没想到它真的能成。
64:11
Uh, and when I went to college, it really seemed like we were very far from it.
嗯，上大学那会儿，感觉离目标还非常遥远。
64:16
And then in 2012, the Alex Net paper came out done, you
然后到了2012年，AlexNet那篇论文出来了，是跟我的联合创始人Ilia合作完成的。
64:20
know, in partnership with my co-founder, Ilia.
你知道，我和我的联合创始人Ilia一起合作。
64:22
And for the first time, it seemed to me like there was an approach that might work.
那是我第一次觉得，好像有一种方法可能行得通。
64:27
And then I kept watching for the next couple of years as scaled up, scaled up, got better, better.
之后几年我一直在观察，随着规模不断扩大，效果越来越好。
64:33
And I remember having this thing of like why is the world not paying attention to this? >> It seems like obvious to me that this might work.
我当时就在想，为什么全世界都没注意到这个？
64:41
Still a low chance, but it might work.
>> 对我来说，这似乎很明显——这个方法可能行得通。
64:43
And if it does work, it's
虽然概率还是不高，但确实有可能。
64:45
just the most important thing.
这就是最重要的事。
64:47
So like this is what I want to do.
所以这就是我想做的。
64:50
And then like unbelievably it started to work. >> Thank you so much for your time. >> Thank you very much.
然后，不可思议的是，它居然开始奏效了。