Re-upload of Stanford ECON295/CS323 I 2024 I The Age of AI, Eric Schmidt - YouTube
发布时间 2024-08-14 13:38:29 来源
Today's guest really does the introduction. I think I first met Eric about 25 years ago when he came to Stanford Business School as CEO of Novel. He's had done a few things since then at Google starting I think 2001 and Schmidt Futures starting in 2017 and done a whole bunch of other things you can read about. He can only be here until 5.15 so I thought we'd dive right into some questions and I know you guys have sent some as well. I have a bunch written here but what we just talked about upstairs was even more interesting so I'm just going to start with that Eric if that's okay which is where do you see AI going in the short term which I think you defined as the next year or two.
今天的嘉宾真的值得介绍一下。我想我第一次见到埃里克是在大约25年前，当时他作为Novell的CEO来到斯坦福商学院。从那以后，他在谷歌从2001年开始工作，再到2017年成立施密特期货基金，完成了许多其他你可以查阅的成就。他只能待到5:15，所以我想我们直接进入一些问题，我知道你们也提交了一些问题。我这里列了一些问题，不过我们刚刚在楼上谈的话题更有趣，所以就从这开始吧，埃里克，如果你没有异议的话—你认为短期内人工智能的发展趋势如何，我想你将短期定义为未来一两年内。
Things have changed so fast I feel like every six months I need to sort of give a new speech on what's going to happen. Can anybody hear a bunch of computer science in here? Can anybody explain what a million token context window is for the rest of the class? Can you hear? So say your name, tell us what it does. Basically a lot of YouTube problems with like a million tokens or a million words or a bit of just like you just said. So you can ask a million word question. I know this is a very large-generation gem now right now. No no they're going to 10. It's totally ridiculous. Yeah and the topic is that 200,000 going to a million and so forth. You can imagine opening AI just as similar.
事情变化得如此之快，我感觉每隔六个月就需要重新做一次关于未来发展的演讲。有人能解释一下这里所提到的计算机科学概念吗？有人能为大家解释一下什么是百万级别的上下文窗口吗？能听到我说话吗？请说出你的名字，然后告诉我们这是什么东西。基本上，很多YouTube上的问题都有百万级代币或百万级单词的情况，就像你刚才提到的一样。所以你可以问一个含有一百万单词的问题。我知道这是一个当前非常大规模的生成模型。不不，他们要提升到十倍，这实在是太荒唐了。是的，主题是从20万提升到一百万，等等。你可以想象一下，Open AI 也在做类似的事情。
Anybody here give a technical definition of an AI agent? Again, I think you're just not honest. Yes sir. And it's AI agent is right in the side of the house in some kind of way so that might be a name and calling things on the web, finding things on your behalf, be a number of things on a lot of these lines. There are various things. Yeah so an agent is something that does some kind of a task. Another definition would be that it's an LLM, state and memory. Can anybody, again, computer scientists can any of you define text to action? Taking text and turning it into an action. Right here. Go ahead. Yes instead of taking text and turning it into more text. More text. Taking text and having the AI trigger actions based on the problem.
有人可以在这里提供一个AI代理的技术定义吗？我觉得你不诚实。是的，先生。AI代理在某种程度上就像在房子旁边，可能是在网络上找到某些东西，代表你找东西，或者做其他各种事情。一个定义是，它是一个完成某种任务的东西。另一个定义是，它是一个具有状态和记忆的LLM（大语言模型）。有电脑科学家能定义"文本到行动"吗？将文本转换成行动。就在这里。请继续。是的，不只是将文本转换成更多的文本。而是将文本转化为由AI基于问题触发的行动。
So another definition would be language to Python. A programming language I never wanted to see survive and everything in AI is being done in Python. There's a language called Mojo that has just come out which looks like they finally have addressed AI programming but we'll see if that actually survives over the dominance of Python. One more technical question. Why is NVIDIA worth two trillion dollars and the other companies are struggling? Technical answer. I mean I think it just boils down to like most of code needs to run the code optimizations that currently only NVIDIA GPU so quickly how the companies can make whatever they want to but unless they have the 10 years of software there you don't have the machine learning optimization term. I like to think of CUDA as the C programming language for GPUs. Yeah. Right. That's the way I like to think of it. It was founded in 2008. I always thought it was the terrible language and yet it's become dominant.
所以另一个定义将会是将语言转化为Python。Python是我一直不想看到存活的编程语言，但现在所有的人工智能都在用Python。有一种叫Mojo的语言刚刚出现，看起来他们终于在解决AI编程问题了，不过我们还要看看它是否能在Python的霸主地位下存活。再一个技术问题，为什么NVIDIA价值两万亿美元，而其他公司都在挣扎？技术答案。我认为这归结于大部分代码需要运行优化，而目前只有NVIDIA的GPU能快速实现这种优化。公司可以制造他们想要的任何东西，但如果没有10年的软件积累，就没有机器学习的优化。我喜欢把CUDA看作是GPU的C编程语言。对，就是这样。我一直认为它是很糟糕的语言，却还是成了主流。
There's another insight. There's a set of open source libraries which are highly optimized to CUDA and not anything else and everybody who builds all these stacks. This is completely missed in any of the discussions. The come it's technically called VLOM and a whole bunch of libraries like that. Highly optimized CUDA, CUDA. Very hard to replicate that if you're a competitor. So what does all this mean? In the next year you're going to see very large context windows, agents and text action when they are delivered at scale it's going to have an impact on the world at a scale that no one understands yet. Much bigger than the horrific impact we've had on social media. In my view.
这有另一个见解。有一组开源库高度优化了CUDA，但对其他东西没有这样的优化，所有构建这些技术栈的人都使用它们。这完全在任何讨论中都被忽略了。这些库技术上被称为VLOM以及类似的一整套库。高度优化的CUDA，CUDA。如果你是竞争对手，要复制这些非常困难。那么，这一切意味着什么呢？在接下来的一年里，你会看到非常大的上下文窗口、智能代理和文本动作。当它们大规模交付时，将会对世界产生一种还没有人完全理解的影响。比我们在社交媒体上遭受的可怕影响还要大。依我之见。
So here's why. In a context window you can basically use that as short term memory and I was shocked that context windows get this long. The technical reasons have to do with the fact that it's hard to serve, hard to calculate and so forth. The interesting thing about short term memory is when you feed, you ask it a question, read 20 books, you give it the text of the books, is the query and you say tell me what they say it forgets the middle which is exactly how human brains work to. That's where we are. With respect to agents there are people who are now building essentially LLM agents and the way they do it is they read something like chemistry, they discover the principles of chemistry and then they test it and then they add that back into their understanding. That's extremely powerful and then the third thing as I mentioned is text action.
这里是原因。在上下文窗口中，你基本上可以把它当作短期记忆。我很震惊的是上下文窗口竟然能变得如此长。技术上的原因与计算和提供服务的难度有关。短期记忆的有趣之处在于，当你喂给它问题，让它读20本书，把书的内容作为查询文本，然后你问它这些书讲了什么，它会忘记中间的部分，这正是人类大脑的运作方式。我们现在在这方面的进展就是这样。关于智能代理，有一些人正在构建基本上是大型语言模型的代理，他们的做法是读取像化学这样的内容，发现化学原理，然后进行测试，再将测试结果添加到他们的理解中。这非常强大。第三点是我提到的文本操作。
So I'll give you an example. The government is in the process of trying to ban TikTok. We'll see if that actually happens. If TikTok is banned, here's what I propose each and every one of you do. Say to your LLM, the following. Take me a copy of TikTok. Steal all the users, steal all the music, put my preferences in it, produce this program in the next 30 seconds, release it and in one hour if it's not viral, do something different along the same lines. That's the command. Boom, boom, boom, boom, right? You understand how powerful that is. If you can go from arbitrary language to arbitrary digital command, which is essentially what Python in this scenario is, imagine that each and every human on the planet has their own programmer that actually does what they want as opposed to the programmers that work for me who don't do what I ask. The programmers here know what I'm talking about. So imagine a non-arragant programmer that actually does what you want and you don't have to pay all that money to. And there's infinite supply of these programs. And this is all within the next year or two. Very soon. Those three things, and I'm quite convinced it's the union of those three things that will happen in the next wave.
让我给你举个例子。政府正在尝试禁止TikTok。我们拭目以待是否真的会发生。如果TikTok被禁了，我建议每一个人都可以这样做：对你的语言模型（LLM）说以下内容。给我搞一个TikTok的副本。偷取所有用户，偷取所有音乐，把我的偏好放进去，在接下来的30秒内生成这个程序，发布出来，如果一小时内没有火起来，就做一个类似的不同版本。这就是命令。你明白这有多强大吧。如果你可以从任意语言到任意数字命令，这实际上就是这个场景中的Python。想象一下，每个人都有一个程序员，可以按照他们的要求实际执行，而不是像我雇用的那些程序员，他们不按照我的要求做。程序员们都知道我在说什么。所以，想象一下一个不自大且真正按你要求做事的程序员，而且你不需要支付大量费用。这些程序是无限供应的。这一切都会在接下来的一两年内实现，非常快。这三件事情，我相当确信，这将是下一波浪潮的综合体现。
So you asked about what else is going to happen. Every six months I oscillate. So we're on a, is an even odd oscillation. So at the moment, the gap between the frontier models, which they're now only three, I'll review who they are, and everybody else appears to me to be getting larger. Six months ago I was convinced that the gap was getting smaller. So I invested lots of money in the little companies. Now I'm not so sure. And I'm talking to the big companies and the big companies are telling me that they need 10 billion, 20 billion, 50 billion, 100 billion. Stargate is 100 billion, right? They're very, very hard. I talked, Sam Altman is a close friend. He believes that it's going to take about 300 billion, maybe more. I pointed out to him that I'd done the calculation on the amount of energy required. And I, and I then in the spirit of full disclosure, went to the White House on Friday and told them that we need to become best friends with Canada. Because Canada has really nice people, helped in Venn AI, and lots of hydropower. Because we as a country do not have enough power to do this. The alternative is to have the Arabs fund it. And I like the Arabs personally. I spent lots of time there, right? But they're not going to adhere to our national security rules. Whereas Canada and the US are part of a triumpet where we all agree. So these $100 billion, $300 billion of data centers, electricity starts becoming the scarce resource. Well, and by the way, if you follow this line of reasoning, why did I discuss CUDA and NVIDIA? If $300 billion is all going to go to NVIDIA, you know what to do in the stock market. Okay. That's not a stock recommendation. I'm not a licensed.
所以你问了接下来会发生什么事。每六个月我会来回摇摆一下。所以现在我们正处于一个偶数奇数的摇摆周期。目前，在前沿模型之间的差距——现在只有三个前沿模型——和其他所有模型之间的差距似乎在变大。六个月前，我确信这个差距正在缩小，所以我投资了很多钱在小公司上。现在我不那么确定了。我在和大公司谈，他们告诉我他们需要100亿、200亿、500亿甚至1000亿美元。Stargate需要1000亿美元。他们非常非常难搞。我和Sam Altman谈过，他是我的好朋友。他相信这将需要大约3000亿美元，可能还不止。我指出我已经计算了所需的能源量。然后为了完全公开，我在星期五去了白宫，告诉他们我们需要成为加拿大的好朋友。因为加拿大有很好的人民，帮助发明了AI，还有大量的水电资源。而我们国家没有足够的电力来做这件事。另一种选择是让阿拉伯国家资助它。我个人很喜欢阿拉伯人，我在那里待过很长时间。但是他们不会遵守我们的国家安全规则。而加拿大和美国是一个协议的一部分，我们都同意遵守。因此，当涉及到1000亿、3000亿美元的数据中心时，电力开始变成稀缺资源。顺便说一下，如果你按照这个思路推理，为什么我要讨论CUDA和NVIDIA？如果3000亿美元都要给NVIDIA，你知道该怎么在股市上操作了。当然，这不是股票建议，我没有执照。
Well, part of it, so we're going to need a lot more chips. Intel is getting a lot of money from the US government, AMD. And they're trying to build fabs and. Raise your hand if you have an Intel computer in your. Reentel chip in any of your computing devices. Okay. So much for the monopoly. Well, that's the point though. They once did have a monopoly. Absolutely. And NVIDIA has a monopoly now. So are those barriers to entry? Like CUDA, is there something that. So I was talking to Percy, Percy Langhy, the other day. He's switching between TPUs and NVIDIA chips, depending on what he can get access to for training models. That's because he doesn't have a choice. If he had infinite money, today he would pick the B200 architecture out of NVIDIA, because he'll be faster. And I'm not suggesting.
好吧，事情是这样的，所以我们需要更多的芯片。英特尔从美国政府那里获得了很多资金，而AMD也在努力建厂。请举手示意一下你的电脑里是否有英特尔的芯片。好吧，这就不多说垄断的事情了。不过，关键是他们曾经确实拥有过垄断地位。而现在，NVIDIA也有垄断地位。那么，还有进入这个市场的障碍吗？比如CUDA之类的技术？前几天我和Percy Langhy谈过，他在根据能获取到的设备在TPU和NVIDIA芯片之间切换进行模型训练。这是因为他没有别的选择。如果他有无限的资金，他现在会选择NVIDIA的B200架构，因为速度更快。不过我并不是在建议...
I mean, it's great to have competition. I've talked to AMD and Lisa Sewick, great length. They have built a thing which will translate from this CUDA architecture that you were describing to their own, which is called RACAM. It doesn't quite work yet. They're working on it. You were at Google for a long time and they invented the transformer architecture. Peter, Peter. Peter, Peter. It's all Peter's fault. Thanks to brilliant people over there, like Peter and Jechtine and everyone. But now it doesn't seem like they've lost the initiative to open AI and even the last leaderboard I saw, Anthropix Cloud was at the top of the list. I asked Sundar this. He didn't really give me a very sharp answer. Maybe you have a sharper or more objective explanation for what's going on there.
我的意思是，有竞争是件好事。我与AMD和Lisa Sewick进行了深入交谈。他们正在开发一种工具，可以将你提到的CUDA架构转换为他们自己的架构，称为RACAM。虽然现在还不能完全正常工作，但他们正在努力改进。你在谷歌工作了很长时间，他们发明了Transformer架构。Peter，Peter。都是Peter的功劳。要感谢像Peter和Jechtine以及那里的所有天才们。但现在看来，他们似乎在与OpenAI的竞争中失去了主动权，甚至我看到的最新排行榜上，Anthropix Cloud位居榜首。我问过Sundar这个问题，他并没有给我一个非常明确的回答。也许你有一个更清晰或更客观的解释，能告诉我到底发生了什么。
I'm no longer a Google employee in this beautiful disclosure. Google decided that work-life balance and going home early and working from home was more important than winning. And the startups, the reason startups work is because the people work like hell and I'm sorry to be so blunt. But the fact of the matter is if you all leave the university and go found a company, you're not going to let people work from home and only come in one day a week if you want to compete against the other startups. In the early days of Google, Microsoft was like that. Exactly. But now it seems to be.
在这个美丽的揭露中，我不再是 Google 员工了。Google 认为在工作与生活之间找到平衡、早早回家以及在家办公，比胜利更重要。而初创公司之所以能够成功，是因为员工拼命工作，对不起，我只能这么直白地说。但事实是，如果你们都离开大学，创办一家公司，你们不可能允许员工在家办公，一周只来公司一天，如果你们想与其他初创公司竞争的话。在 Google 的早期，微软也是这样的。但现在情况似乎变了。
There's a long history of, in my industry, our industry, I guess, of companies winning in a genuinely creative way and really dominating a space and not making the next transition. It's very well documented. And I think that the truth is founders are special. The founders need to be in charge. The founders are difficult to work with. They push people hard. As much as we can dislike Elon's personal behavior, look at what he gets out of people. I had dinner with him and I was in Montana. She was flying that night at 10 p.m. to have a meeting at midnight with x.ai. I think that I was in Taiwan, different country, different culture. And they said that this is TSMC, I'm very impressed with. And they have a rule that the starting PhDs coming out of their good physicists work in the factory on the basement floor.
在我的行业，或者说我们的行业，有着悠久的历史，那就是公司通过真正有创意的方式成功并在一个领域占据主导地位，但却未能完成下一次转型。这点有很多记录。而且，我认为真相是创始人是特别的。创始人需要掌控大局。创始人很难相处，他们对员工要求很高。尽管我们可能不喜欢埃隆的个人行为，但看看他从人们身上挖掘出来的潜力。我曾在蒙大拿州和他共进晚餐，那天晚上十点他还要飞去参加一个午夜的会议，和 x.ai 见面。我认为我当时在台湾，这是一个不同的国家，不同的文化。他们表示这是台积电的规则，我印象非常深刻。他们有一条规定，即刚毕业的物理学博士要在工厂的地下层工作。
Now, can you imagine getting American physicists to do that? The PhDs, highly unlikely. Different work ethic. And the problem here, the reason I'm being so harsh about work is that these are systems which have network effects. So time matters a lot. And in most businesses, time doesn't matter that much. You have lots of time. Coke and Pepsi will still be around and the fight between Coke and Pepsi will continue to go on and it's all glacial. When I dealt with telcos, the typical telco deal would take 18 months to sign. There's no reason to take 18 months to do anything, get it done. We're in a period of maximum growth, maximum gain. And also it takes crazy ideas. Like when Microsoft did the opening ideal, I thought that was the stupidest idea I'd ever heard.
现在，你能想象让美国的物理学家这样做吗？那些博士，几乎不可能。工作态度不同。这里的问题是，我对工作的要求这么苛刻是因为这些系统具有网络效应。所以时间非常重要。而在大多数业务中，时间并没有那么重要，你有很多时间。可口可乐和百事可乐还会继续存在，它们之间的竞争也会继续，但进展非常缓慢。当我与电信公司打交道时，一个典型的电信合同要签18个月。没有理由做任何事情需要18个月，赶紧完成。我们正处在最大增长、最大收益的时期。而且这需要疯狂的想法。就像微软当初进行开放式构想时，我认为那是我听过的最愚蠢的主意。
Outsourcing essentially your AI leadership to open AI and Sam and his team. I mean, that's insane. Nobody would do that at Microsoft or anywhere else. And yet today, they're on their way to being the most valuable company. They certainly head to head and Apple. Apple does not have a good AI solution. And it looks like they made it work. Yes, sir. In terms of national security, or do you look at what you're going to play a role or competition with China as well? So I was the chairman of an AI commission that sort of looked at this very carefully. And you can read it. It's about 752 pages. And I'll just summarize it by saying we're ahead. We need to stay ahead. And we need lots of money to do so. Our customers were the Senate in the house. Out of that came the chips act and a lot of other stuff like that. A rough scenario is that if you assume the frontier models drive forward and a few of the open source models, it's likely that a very small number of companies can play this game. Countries, excuse me. What are those countries? Who are they? Country with a lot of money and a lot of talent, strong educational systems and a willingness to win. The US is one of them. China is another one. How many others are there? Are there any others? I don't know. Maybe. But certainly in your lifetimes, the battle between the US and China for knowledge supremacy is going to be the big fight.
将你的AI领导权外包给OpenAI的Sam及其团队，这简直是疯了。没有人会在微软或其他地方那样做。然而，如今，他们正走在成为最有价值公司的路上。他们显然与苹果不相上下。苹果没有好的AI解决方案。而他们似乎做到了。至于国家安全或与中国的竞争，你是否也会参与其中呢？我曾是一个AI委员会的主席，该委员会非常仔细地审查了这个问题。你可以阅读报告，大约752页。我总结一下，我们现在领先，我们需要保持领先，并且需要大量资金来做到这一点。我们的客户是参议院和众议院。这导致了“芯片法案”等一系列行动。大致情况是，如果假设前沿模型继续推进，加上一些开源模型，可能只有少数几个国家能够参与竞争。哪些国家呢？拥有大量资金、大量人才、强大教育系统和强烈取胜意愿的国家。美国是其中之一。中国是另一个。还有其他国家吗？我不知道，可能有。但在你们的有生之年，美国和中国在知识主导地位上的争夺将是主要的战斗。
So the US government banned essentially the NVIDIA chips, although they weren't allowed to say that was what they were doing, but they actually did that into China. They have about a 10 year chip advantage. We have a roughly 10 year chip advantage in terms of sub-DUV. It is sub-5-10 years. That's roughly 10 years. And so an example would be today we're a couple of years ahead of China. My guess is we'll get a few more years ahead of China and the Chinese are whopping mad about this. It's like hugely upset about it. So that's a big deal. That was the decision made by the Trump administration and further by the Biden administration. Do you find that the administration today and in Congress is listening to your advice? Do you think that it's going to make that scale of investment? I mean, obviously the chips act, but beyond that, building a massive AI system? So as you know, I lead an informal ad hoc non-legal group that's not as different from illegal. Exactly. It's going to be clear. Which includes all the usual suspects. And the usual suspects over the last year came up with the basis of the reasoning that became the Biden administration's AI Act, which is the longest presidential directive in history. You're talking about the special competitive studies project? No. This is the actual act from the executive office. And they're busy implementing the details. So far they've got it right. And so for example, one of the debates that we had for the last year has been how do you detect danger in a system which has learned it, but you don't know what to ask it? Mm-hmm. Okay.
所以，虽然美国政府不允许公开表示在对中国实施禁令，但他们实际上是禁止了NVIDIA的芯片进入中国。他们大概在芯片技术上领先了中国10年左右。我们的芯片技术在先进的光刻技术方面大约比中国领先10年。例如，在目前的情况下，我们比中国领先了几年。我猜我们会再领先中国几年，而中国对此非常愤怒。这是一个大事件，这是特朗普政府作出的决定，且在拜登政府时期进一步强化。你认为当前的政府和国会在倾听你的建议吗？你认为他们会进行那种规模的投资吗？我指的不仅仅是芯片法案，还包括建立一个庞大的AI系统？正如你知道的，我领导一个非正式的临时小组，这与‘非法’不同。确切地说，需要明确的是，这包括所有的惯犯。惯犯们在过去一年中提出了成为拜登政府AI法案基础的理由，这是历史上最长的总统指令。你指的是特别竞争研究项目吗？不，这是来自行政办公室的实际法案，他们正忙着实施细节。目前来看，他们的方向是对的。例如，过去一年我们一直在争论的一个问题是，如何在一个已经学会了的系统中检测出潜在的危险，但你不知道该问它什么。嗯，好吧。
So in other words, it's a sort of a core problem. It's learned something bad, but it can't tell you what it learned. And you don't know what to ask it. And there's so many threats, right? Like it learned how to mix chemistry in some new way, but you don't know how to ask it. And so people are working hard on that, but we ultimately wrote in our memos to them that there was a threshold which we arbitrarily named as 10 to the 26 flops, which technically is a measure of computation, that above that threshold you had to report to the government that you were doing this. And that's part of the rule. The EU to just make sure they were different did it 10 to the 25. Yeah. But it's all kind of close enough. I think all of these distinctions go away because the technology will now, the technical term is called federated training, where basically you can take pieces and union them together. Mm-hmm. So we may not be able to keep people safe from these new things. Well, rumors are that that's how OpenEye has had to train partly because of the power consumption. Yeah. There was no one place where they did.
换句话说，这是一个核心问题。它学会了一些不好的东西，但它无法告诉你它学到了什么，而且你也不知道该问它什么。这就带来了很多威胁，比如它学会了一种新的化学混合方法，但你不知道该如何询问它。所以，人们正在努力解决这个问题，但我们最终在给他们的备忘录中写到，我们设定了一个门槛，任意定义为10的26次方浮点运算（flops），这是一个技术上的计算度量，超过这个门槛，你必须向政府报告你在进行此类活动。而这是规矩的一部分。为了有所区别，欧盟设定的门槛是10的25次方浮点运算。嗯，但这个差别其实很小。我认为所有这些区别最终都会消失，因为现在有一种技术叫做联邦学习，基本上它可以把各个部分联合起来。嗯，所以我们可能没办法保护人们免受这些新技术的威胁。有传言说，这也是OpenAI部分训练方式，因为功耗的原因，没有一个地方能够完成所有训练。
Well, let's talk to about a real war that's going on. I know that something you've been very involved in is the Ukraine war, in particular, on an almost you can talk about white stork and your goal of having a $500,000 drones destroy $5 million tanks. So how's that changing warfare? So I worked for the Secretary of Defense for seven years and tried to change the way we run our military. I'm not a particularly big fan of the military, but it's very expensive. And I wanted to see if I could be hopeful. And I think in my view, I largely failed. They gave me a medal. So they must give medalists a failure or whatever. But my self-criticism was nothing has really changed. And the system in America is not going to lead to real innovation. So watching the Russians use tanks to destroy apartment buildings with little old ladies and kids just drove me crazy. So I decided to work on a company with your friend Sebastian Thrun and a former faculty member here and a whole bunch of Stanford people. And the idea basically is to do two things. Use AI in complicated powerful ways for these essentially robotic war. And the second one is to lower the cost of the robots. Now you sit there and you go, why would a good liberal like me do that? And the answer is that the whole theory of armies is tanks are chilleries in mortar and we can eliminate all of them. And we can make the penalty for invading a country, at least by land, essentially be impossible. It should eliminate the kind of land battles.
好吧，让我们来谈谈一场正在进行的真正战争。我知道你一直非常关注乌克兰战争，特别是涉及到白鹳无人机项目的部分。你希望用价值50万美元的无人机摧毁价值500万美元的坦克。这是如何改变战争方式的呢？
我在国防部长手下工作了七年，试图改变我们运行军队的方式。我并不是特别喜欢军队，但这是一项非常昂贵的开支。我想看看是否能够带来一些希望，但在我看来，我基本上失败了。他们给了我一枚勋章，所以也许他们给失败者发勋章什么的。但就我自我批评而言，什么也没有真正改变，美国的体制不可能引导出真正的创新。
看到俄罗斯用坦克摧毁有老人和孩子的公寓楼，让我非常愤怒。所以我决定和你的朋友塞巴斯蒂安•思鲁恩以及这儿的前任教授和一大群斯坦福的人一起合作，创办一家公司。这个公司的基本理念有两个方面。一是利用人工智能以复杂而强大的方式进行机器人战争，二是降低机器人的成本。
你可能会问，为什么像我这样一个自由主义者会做这样的事？答案是，军队的整个理论是坦克、大炮和迫击炮，而我们可以消除所有这些。我们可以让入侵一个国家的代价，至少是通过陆路入侵的成本变得不可能，这应该可以消除类似的陆战。
Well, this is a really interesting question. Is that does it give more of an advantage to defense versus offense? Can you even make that distinction? Because I've been doing this for the last year, I've learned a lot about war that I really did not want to know. And one of the things to know about war is that the offense always has the advantage because you can always overwhelm the defensive systems. And so you're better off as a strategy of national defense to have a very strong offense that you can use if you need to. And the systems that I and others are building will do that. Because of the way the system works, I am now a licensed arms dealer, a computer scientist, businessman, arms dealer. And I'm sorry to say. Is that a progression? I don't know. I do not recommend this in your career path. I stick with AI. And because of the way the laws work, we're doing this privately. And then this is all legal with the support of the governments.
好吧，这是一个非常有趣的问题。究竟是防御比进攻更有优势，还是能够区分两者？因为我过去一年都在研究这个问题，学到了很多我原本不想了解的战争知识。其中一条就是，进攻总是占有优势，因为你总能压倒防御系统。因此，作为国家防御战略，拥有强大的进攻力量是更好的选择，以便在需要时使用。我和其他人正在构建的系统就是为了实现这一点。由于系统的工作方式，现在我成了一个持证军火商，一个计算机科学家、商人和军火商。对此，我感到抱歉，这是一种进程吗？我不知道。我不建议你们将此作为职业发展路径，坚持做人工智能吧。并且由于法律的规定，我们正在私下进行这些工作，而且这些都是在政府支持下合法进行的。
It goes straight into the Ukraine and then they fight the war. And without going into all the details, things are pretty bad. I think in May or June, if the Russians build up as they are expecting to, Ukraine will lose a whole chunk of its territory and will begin the process of losing the whole country. So the situation is quite dire. And if anyone knows Marjorie Taylor Greene, I would encourage you to delete her from your contact list. Because she's the one a single individual is blocking the provision of some number of billions of dollars to save an important democracy. I want to switch to a little bit of a philosophical question. So there was an article that you and Henry Kissinger and Dan Hutton like I wrote last year about the nature of knowledge and how it's evolving. I had a discussion the other night about this as well.
它直接进入乌克兰，然后他们开战了。具体细节我就不说了，总之情况相当糟糕。我认为，如果俄罗斯在五月或六月按预期集结起来，乌克兰将失去大部分领土，并开始失去整个国家的过程。所以形势非常严峻。如果有人认识玛乔丽·泰勒·格林，我建议你把她从你的联系人列表中删除。因为她一个人正在阻碍数十亿美元的拨款，而这笔钱可以用来拯救一个重要的民主国家。我想转到一个有点哲学的问题上。去年，你、亨利·基辛格和丹·赫顿写了一篇关于知识本质和其演变的文章。我前几天晚上也讨论了这个话题。
So for most of history, humans sort of had a mystical understanding of the universe. And then there's the scientific revolution and the enlightenment. And in your article, you argue that now these models are becoming so complicated and difficult to understand that we don't really know what's going on in them. I'll take a quote from Richard Feynman. He says, what I cannot create, I do not understand the saw this quote the other day. But now people are creating things they do not that they can create, but they don't really understand what's inside of them. Is the nature of knowledge changing in a way? Are we going to have to start just taking the word for these models about them being able to explain it to us? The analogy I would offer is to teenagers. If you have a teenager, you know they're human, but you can't quite figure out what they're thinking. But somehow we've managed in society to adapt to the presence of teenagers, right? And they eventually grow out of it. And I'm just serious. So it's probably the case that we're going to have knowledge systems that we cannot fully characterize. But we understand their boundaries. We understand the limits of what they can do. And that's probably the best outcome we can get. Do you think we'll understand the limits? We'll get pretty good at it. The consensus of my group that meets on every week is that eventually the way you'll do this so-called adversarial AI is that there will actually be companies that you will hire and pay money to to break your AI system. Like Red Team. So it'll be the, instead of human Red Teams, which is what they do today, you'll have a whole company in a whole industry of AI systems whose jobs are to break the existing AI systems and find their vulnerabilities, especially the knowledge that they have that we can't figure out. That makes sense to me. It's also a great project for you here at Stanford because if you have a graduate student who has to figure out how to attack one of these large models and understand what it does, that is a great skill to build the next generation. So it makes sense to me that the two will travel together.
所以在大部分历史中，人类对宇宙的理解是有点神秘主义的。然后出现了科学革命和启蒙运动。在你的文章中，你提出现在这些模型变得如此复杂和难以理解，我们实际上并不真正了解它们的内部运作。引用理查德·费曼的一句话，“我无法创造的东西，我也无法理解。”我前几天看到了这句话。但是现在人们正在创造一些东西，他们知道如何创造，但并不真正理解这些东西的内部。知识的本质是否在以某种方式改变？我们是否必须开始仅仅依赖这些模型所说的内容，而不是让它们向我们解释清楚呢？我提供的类比是青少年。如果你有一个青少年，你知道他们是人类，但你无法完全理解他们在想什么。但无论如何，我们在社会中已经适应了青少年的存在，对吧？而且他们最终会长大。我是认真的。所以，很可能我们将会有一些我们无法完全描述的知识系统。但我们理解它们的边界，知道它们的能力范围。这可能是我们能取得的最好结果。你认为我们能理解这些边界吗？我们会变得相当擅长的。我们每周都会开会的小组的共识是，将来你所谓的对抗性人工智能可能实际上会有公司，你需要支付费用来让这些公司攻破你的人工智能系统。就像“红队”那样。所以未来会有一个整个行业，专门用AI系统来攻破现有的AI系统，找到它们的漏洞，特别是那些我们无法理解的知识。这对我来说很合理。对于你在斯坦福大学的项目也很有意义，因为如果你有研究生需要搞清楚如何攻击这些大型模型并理解它们的运作，这将是一项为下一代培养技能的绝佳机会。所以这对我来说是合情合理的，两者将会一起发展。
All right, let's take some questions from the student. There's one right there in the back. Just say your name. If you mentioned, and this is a legit trick to comment right now, I'm getting AI that actually does what you want. You just mentioned adversarial AI. I'm wondering if you could elaborate on that more. So it seems to be, besides obviously, people increase and you get more performance models but it's getting them to do what you want issue since you're totally going on answering my team. You have to assume that the current hallucination problems become less as the technology gets better and so forth. I'm not suggesting it goes away. And then you also have to assume that there are tests for efficacy. So there has to be a way of knowing that the things exceeded. So in the example that I gave of the TikTok competitor, and by the way, I was not arguing that you should illegally steal everybody's music, what you would do if you're a Silicon Valley entrepreneur, which hopefully all of you will be, is if it took off, then you'd hire a whole bunch of lawyers to go clean the mess up. But if nobody uses your product, it doesn't matter that you stole all the content and do not quote me. Right. You're on camera. But you see my point. In other words, Silicon Valley will run these tests and clean up the mess. That's typically how those things are done. So my own view is that you'll see more and more performative systems with even better tests and eventually adversarial tests, and that will keep it within a box.
好的，让我们跟学生们提一些问题吧。那边最后面有一个问题。请说出你的名字。如果你提到，而且这是个实际建议，我现在正在使用的AI可以实现你想要的功能。你刚才提到了对抗性AI。我想知道你是否可以详细解释一下这方面的内容。除了显然的，人们会不断改进，性能更高的模型以外，似乎关键是要让它们按你的意图行事，因为你完全回答了我的问题。你必须假设随着技术的进步，当前的幻觉问题会减少等等。我并不是说这些问题会完全消失。而且你也得假设有一些有效性测试。因此必须有一种知道这些事情是否超越了预期的方法。举个例子，我提到了一个与TikTok竞争的平台，顺便说一下，我并不是鼓励你们非法盗用所有的音乐。如果你是个硅谷企业家，希望你们都能成为这样的人，如果这个平台成功了，那你就会雇一大批律师来解决这个混乱问题。但如果没有人使用你的产品，那么你盗用了所有内容也没关系，不要引用我说的话。对，你在录像呢。不过你们明白我的意思。换句话说，硅谷会运行这些测试并清理混乱问题。这通常是问题处理的方式。所以我的看法是，你会看到更多更具表现力的系统，有更好的测试，最终出现对抗性测试，这样就能将其控制在一定范围内。
The technical term is called chain of thought reasoning. And people believe that in the next few years you'll be able to generate a thousand steps of chain of thought reasoning. Right. Do this, do this. It's like building recipes. Right. That the recipes, you can run the recipe and you can actually test that it produced the correct outcome. And that's how the system will work. Yes, sir. Yes. The amount of money being thrown around are mind-boggling. And I've chosen, I essentially invest in everything because I can't figure out who's going to win. And the amounts of money that are following me are so large.
这个术语叫做连锁思维推理。人们认为在未来的几年里，你将能够生成一千个连锁思维推理的步骤。对，就像这样做，然后这样做。这就像在构建菜谱一样。对吧，你可以运行这些菜谱，然后你可以实际测试它们是否产生了正确的结果，这就是系统的工作方式。是的，没错。目前投入的资金之庞大令人瞠目结舌。而我选择投资一切，因为我无法预测谁会胜出。跟随我的资金量非常巨大。
I think some of it is because the early money has been made and the big money people who don't know what they're doing have to have an AI component. And everything is now an AI investment, so they can't tell the difference. I define AI as learning systems, systems that actually learn. So I think that's one of them. The second is that there are very sophisticated new algorithms that are sort of post-transformers. My friend, my collaborator for a long time has invented a new non-transformer architecture.
我认为有些原因是早期的钱已经赚到了，那些不懂行的大资金投资者现在必须要有一个AI成分。而现在所有东西都是AI投资，所以他们分不清其中的区别。我认为AI是指那些可以学习的系统，真正能学习的系统。所以我觉得这是其中一个原因。第二个原因是出现了一些非常复杂的新算法，它们可以说是超越了现有的“变压器”（transformer）模型。我的朋友和长期的合作伙伴发明了一种新的非变压器结构。
There's a group that I'm funding in Paris that has claimed to have done the same thing. There's enormous invention there, a lot of things at Stanford. And the final thing is that there is a belief in the market that the invention of intelligence has infinite return. So let's say you put $50 billion of capital into a company. You have to make an awful lot of money from intelligence to pay that back. So it's probably the case that we'll go through some huge investment bubble and then it'll sort itself out.
我在巴黎资助了一个声称完成了同样事情的团队。在巴黎有许多发明，在斯坦福也有很多。而最后一点是，市场上有一种信念，认为智能的发明具有无限的回报。所以，假设你向一家公司投入500亿美元，你必须通过智能带来的收益来收回成本。因此，我们可能会经历一个巨大的投资泡沫，然后它会自我调整。
That's always been true in the past and it's likely to be true here. And what you said earlier was you think that the leaders are pulling away from us. Right now, right now. And this is a really, the question is roughly the following. There's a company called Mistral in France. They've done a really good job. And I'm obviously an investor. They have produced their second version. Their third model is likely to be closed because it's so expensive. They need revenue and they can't give their model away. So this open source versus closed source debate in our industry is huge.
这在过去一直是事实，现在也很可能如此。你之前说过，你觉得领导者正在远离我们。现在，问题大致是这样的：法国有一家叫Mistral的公司，他们做得非常好，而且我显然是投资者。他们已经推出了第二代产品，第三代型号可能会被封闭，因为成本太高。他们需要收入，不能把模型免费提供出去。所以，在我们行业中，开源与封闭源代码的争论非常激烈。
And my entire career was based on people being willing to share software in open source. Everything about me is open source. Much of Google's underpinnings were open source. Everything I've done technically. And yet it may be that the capital costs, which are so immense, fundamentally changes how software is built. You and I were talking my own view of software programmers is that software programmers' productivity will at least double.
我的整个职业生涯都是基于人们愿意分享开源软件。我的一切都与开源有关。谷歌的许多基础也都是开源的，我在技术上所做的一切也是如此。然而，庞大的资本成本可能会从根本上改变软件的构建方式。你我谈到过，我认为软件程序员的生产力至少会翻倍。
There are three or four software companies that are trying to do that. I've invested in all of them in the spirit. And they're all trying to make software programmers more productive. The most interesting one that I just met with is called Augment. And I always think of an individual programmer and they said that's not our target. Our target are these 100 person software programming teams on millions of lines of code where nobody knows what's going on. Well, that's a really good AI. Will they make money? I hope so. There's a lot of questions here. Hi.
有三四家公司正在尝试做这件事。我怀着支持的精神在这些公司都有投资。它们都在努力提高软件程序员的生产力。其中最让我感兴趣的一家公司叫做Augment。我以前总是想着单个程序员，但他们告诉我这不是他们的目标。他们的目标是那些拥有上百名程序员的大型团队，这些团队处理着上百万行代码，没人知道具体情况。嗯，这确实是一个很好的AI。他们能赚钱吗？我希望如此，这里面还有很多问题需要解答。嗨。
So at the very beginning, you mentioned that there's the combination of the context with expansion, the agents and the text action is going to have unimaginable impacts. First of all, why is the combination important? And second of all, I know that you're not like a crystal ball and you can't necessarily tell the future, but why do you think it's beyond anything that we could imagine?
一开始，你提到过结合上下文扩展、代理和文本操作会产生难以想象的影响。首先，为什么这种结合很重要呢？其次，我知道你不像水晶球，无法预见未来，但为什么你认为这会超出我们的想象呢？
I think largely because the context window allows you to solve the problem of recency. The current models take a year to train roughly 18 months, six months of preparation, six months of training, six months of fine tuning. So they're always out of date. Context window, you can feed what happened. You can ask it questions about the Hamas Israel war in a context. That's very powerful. It becomes current like Google. In the case of agents, I'll give you an example. I set up a foundation which is funding a nonprofit. It starts, I don't know if there's chemists in the room.
我认为主要是因为上下文窗口可以解决新近性的问题。目前的模型大约需要一年时间来训练，大约18个月，6个月的准备，6个月的训练，6个月的微调。所以它们总是有点过时。通过上下文窗口，你可以输入最新发生的事情。你可以在上下文中问它有关哈马斯与以色列战争的问题。这非常强大，它变得像谷歌一样最新。在代理方面，我给你举个例子。我设立了一个基金会，资助一个非营利组织。它开始，我不知道这里有没有化学家。
I don't really understand chemistry. There's a tool called chemcro, which was an LOM based system that learned chemistry. What they do is they run it to generate chemistry hypotheses about proteins and they have a lab which runs the tests overnight and then it learns. That's a huge acceleration, accelerant in chemistry, material science and so forth. So that's an agent model. And I think the text to action can be understood by just having a lot of cheap programmers, and I don't think we understand what happens, and this is again your area of expertise, what happens when everyone has their own programmer? And I'm not talking about turning on and off the lights.
我真的不太懂化学。现在有一个叫做chemcro的工具，这是一个基于LOM（学习对象模型）的系统，能够学习化学知识。他们的做法是运行这个工具来生成关于蛋白质的化学假设，然后他们有一个实验室在夜间进行测试，测试结果再反馈给系统学习。这大大加速了化学、材料科学等领域的研究进展。所以这是一个智能代理模型。我认为“从文本到行动”可以通过大量廉价的程序员来实现，但我觉得我们还不真正明白，当每个人都有自己的程序员时会发生什么事情。而且我并不是在说简单的开关灯的操作。
I imagine another example, for some reason you don't like Google. So you say, build me a Google competitor. Yeah, you personally, you don't build me a Google competitor. Search the web, build a UI, make a good copy, add generative AI in an interesting way, do it in 30 seconds and see if it works. So a lot of people believe that the incumbents, including Google, are vulnerable to this kind of an attack. Now we'll see. There are a bunch of questions who are sent over by Slider. I want to give some of them more up-vote. So here's one. We talked a little bit about this last year. How can we stop AI from influencing public opinion, misinformation, especially during the upcoming election? What are the short and long-term solutions from? Most of the misinformation in this upcoming election and globally will be on social media. And the social media companies are not organized well enough to police it. If you look at TikTok, for example, there are lots of accusations that TikTok is favoring one kind of misinformation over another. And there are many people who claim without proof that I'm aware of, that the Chinese are forcing them to do it.
我来举另一个例子，比如你不喜欢谷歌，所以你说，“给我打造一个谷歌的竞争对手。”是的，你个人不可能亲自给我建一个谷歌的竞争对手。你需要搜索网络，构建用户界面，撰写好的文案，以有趣的方式加入生成式AI，然后在30秒内完成，看看是否可行。因此，很多人认为包括谷歌在内的现有公司很容易受到这种挑战。现在我们拭目以待。
这儿有一些由Slider平台提交的问题，其中一些获得了更多的点赞。因此，这里有一个问题。我们去年谈到了一些关于这个问题的内容。我们如何防止AI影响公共舆论并散布误导信息，特别是在即将到来的选举期间？有什么短期和长期的解决方案呢？在即将到来的选举以及全球范围内，大多数误导信息都会出现在社交媒体上。而社交媒体公司并没有足够好的组织能力来监管这些信息。以TikTok为例，有许多指控声称TikTok偏袒某一种误导信息，而没有证据表明这些指控是否真实，还有许多人声称是中国在背后强迫他们这样做。
I think we just have a mess here. And the country is going to have to learn critical thinking. That may be an impossible challenge for the US. But the fact that somebody told you something does not mean that it's true. Well, could it go too far the other way? That there's things that are really are true and nobody believes in what you get. Somebody will call it a pestimalological crisis that now, the law says, no, I never did that. I approve it. Well, let's use Donald Trump. Okay. So I think we have a trust problem in our society. Democracy is can fail. And I think that the greatest threat to democracy is misinformation because we're going to get really good at it. When I managed YouTube, the biggest problems we had in YouTube were that people would upload false videos and people would die as a result. And we had a no-death policy shocking. And we just when we just horrendous to try to address it. This is before generative A on.
我认为我们现在的情况非常混乱。整个国家需要学会批判性思维，但这可能是对美国来说一个不可能完成的挑战。仅仅因为有人告诉你某件事，并不意味着那是真的。那么，会不会走向另一个极端呢？有些事情确实是真的，但没人相信。有人会把它称作一种认识论危机，法律也会说：“不，我从没做过那事，我要证明。”拿唐纳德·特朗普来说，我们的社会确实存在信任问题。民主可能会失败，而我认为民主面临的最大威胁就是错误信息，因为我们在制造错误信息方面会变得非常擅长。当我管理YouTube时，最大的问题是人们上传虚假视频，导致人们因此而丧生。我们有一项“不死亡政策”，非常震惊，而且在应对这类问题时十分艰难。这还发生在生成式人工智能出现之前。
Well, so I don't have a good answer. One technical is not an answer, but one thing that seems like a could mitigate that I understand why it's more widely used is a public key authentication that when Joe Biden speaks, why isn't it digitally signed like SSL is or when, you know, that celebrities or public figures or others, couldn't they have a public key? Yeah, it's a form of public key and then some form of certainty of knowing how the system can. When I send my credit card to Amazon, I know it's Amazon. I wrote a paper and published it with Jonathan Haidt, who's the one working on the anxiety generation stuff. It had exactly zero impact. And he's a very good communicator. I probably am not. So my conclusion was that the system is not organized to do what you said. You had a paper advocating what we just that. Advocating your proposal? Okay. My microphone. What you said.
嗯，我没有一个好的答案。一个技术解决方案并不是答案，但有一点可能会减轻这个问题，我明白为什么它被广泛使用，那就是公共密钥认证。比如，当乔·拜登发表讲话时，为什么不能像SSL那样进行数字签名呢？或者，当名人或公众人物发表言论时，难道他们不能有一个公共密钥吗？是的，这是一种公共密钥形式，然后就可以在某种程度上确定系统的可靠性。当我把信用卡信息发给亚马逊时，我知道那是亚马逊。我曾与乔纳森·海特（Jonathan Haidt）合作撰写并发表了一篇关于焦虑生成问题的论文，但完全没有影响。乔纳森是一个很好的沟通者，但我可能不是。因此，我的结论是，系统并没有组织起来解决你提到的问题。你是不是在倡导我们刚刚讨论的那个解决方案？好的，我的麦克风问题。
Yeah, right. And my conclusion is the CEOs in general are maximizing revenue to maximize revenue. You maximize engagement to maximize engagement. You maximize outrage. The algorithms choose outrage because that generates more revenue. Right. Therefore, there's a bias to favor crazy stuff. And on all sides, I'm not making a partisan statement here. That's a problem. That's got to get addressed in a democracy. And my solution to TikTok, we talked about this earlier privately. Is when I was a boy, there was something called the equal time rule.
好吧，对的。我认为，总体来说，CEO们是在最大化收入，为了最大化收入。他们最大化参与度，就是为了最大化参与度。他们在最大化愤怒。算法选择愤怒，因为那能带来更多收益。没错。因此，存在偏向于疯狂内容的倾向。而且，这不仅仅是某一方的问题，我并不是在发表党派言论。这是一个问题，在民主制度中必须解决。而我对TikTok的解决方案，我们之前私下讨论过的。在我小时候，有一条叫做“平等时间规则”的规定。
Because TikTok is really not social media. It's really television. There's a programmer making you, the numbers, by the way, are 90 minutes a day, 200 TikTok videos per TikTok user in the United States. It's a lot. And the government is not going to do the equal time rule, but it's the right thing to do. Some form of balance that is required. All right. Let's take some more questions. Two more questions. One, economic impact of LMs. Slow or like the market impacts. Slow work. We originally anticipated and checked and couple of service people on them too. Do you think academia deserves or should get AI subsidies? Or do you think they should just partner with the players out there?
因为 TikTok 其实并不是真正的社交媒体。它实际上是电视。从某种意义上讲，有个程序员在为你编排内容。而且，数据显示，美国的 TikTok 用户平均每天使用 TikTok 90 分钟，观看 200 个视频。这已经是很大的一个数字了。虽然政府可能不会执行平等时间法则，但这是应该做的事情，需要某种形式的平衡。好了，让我们再回答几个问题。再回答两个问题。第一个，语言模型（LMs）的经济影响。是缓慢的还是像市场影响一样缓慢的？我们最初预期时做过检查，并且有几位服务人员也参与其中。你认为学术界应该获得 AI 补贴，还是应该与现有的玩家合作？
I pushed really, really hard on getting data centered for universities. If I were a faculty member in the computer science department here, I would be beyond upset that I can't build the algorithms with my graduate students that will do the kind of PhD research. And then I'm forced to work with these. And the companies have not, in my view, been generous enough with respect to that. The faculty members that I talk with, many of whom you know, spend lots of time waiting for their credits from Google Cloud. And they piece that's terrible. This is an explosion. We want America to win. We want American universities. America, you know, there's lots of reasons to think that the right thing to do is to get it to them. So I'm working hard on that. And your first question was labor market impact. I'll defer to the real expert here.
我在推动大学的数据中心化方面付出了非常大的努力。如果我是这里计算机科学系的教师，我会非常不满，因为我无法与我的研究生一起构建做博士研究所需的算法。而且我被迫与这些公司合作，而这些公司在我看来对这方面的支持并不慷慨。我与许多你也认识的教师交流过，他们花了很多时间等待谷歌云的使用权限，他们认为这是个大问题。这是一个巨大的障碍。我们希望美国取得胜利，我们希望美国的大学能受益。美国，有很多理由让我们认为把资源投入到这些大学是正确的选择。所以我在这方面努力工作。至于你的第一个问题——劳动力市场的影响，我会把这个问题交给真正的专家来回答。
As your amateur economist taught by Eric, I fundamentally believe that the sort of college education, high skills task, will be fine because people will work with these systems. I think the systems is no different from any other technology wave. The dangerous jobs and the jobs which require very little human judgment will get replaced. We got about five minutes left. So let's go really quick with some quick, I'll let you pick them, Eric. Yes, ma'am. I'm really curious about the text to action and its impact on, for example, computer science education. I'm wondering what you have thoughts on like how CS education should transform, kind of meet the age.
作为一名由埃里克教的业余经济学家，我基本上认为，高技能的大学教育将会没问题，因为人们会与这些系统协作。我认为这些系统和其他任何技术浪潮没有什么不同。危险的工作和那些需要很少人类判断的工作将会被取代。我们还有大约五分钟时间。所以让我们快速来几个问题，埃里克，你来选。好的，女士。我对文本到行动及其对计算机科学教育的影响非常感兴趣。我很好奇您对计算机科学教育如何转型以适应新时代有何看法。
Well, I'm assuming that computer scientists as a group in undergraduate school will always have a programmer buddy with them. So when you learn your first for loop and so forth and so on, you'll have a tool that will be your natural partner. And then that's how the teaching will go on. That the professor, he or she will talk about the concepts, but you'll engage with it that way. And that's my guess. Yes ma'am, behind you. Yeah, there's something more about the non-transformer architecture that you're excited about. I think one is the talk that I was like, state models, but then now the longer context, the more I'm curious what you're saying in space. I don't understand the math well enough. This is the, I'm really pleased that we have produced jobs for mathematicians. Because the math here is so complicated, but basically they are different ways of doing gradient descent, matrix multiply faster and better. And transformers, as you know, is a sort of systematic way of multiplying at the same time. That's what I think I think about it.
好的，我认为本科阶段的计算机科学学生群体中，总会有一个编程伙伴陪伴着他们。当你学习第一个for循环等内容时，你会有一个自然的伙伴工具。教授会讲解概念，而你会通过这种方式与之互动。我猜测是这样的。女士，你后面的那位。是的，你对非变压器架构还有什么兴奋的地方吗？我觉得一方面是状态模型的演讲，但现在随着上下文的延长，我越来越好奇你在空间中所说的内容。我在数学方面不太理解。这让我很高兴我们为数学家创造了就业机会。因为这里的数学非常复杂，但基本上它们是不同的梯度下降法和矩阵乘法更快更好的方法。而你知道，变压器是一种同时进行乘法的系统化方法。这就是我的理解。
And it's similar to that, but different math. Let's see over here, yes sir. Go ahead. Yeah, you mentioned in your paper on that security, as you have China, the US, and the health of water, and capabilities today. The next 10, the next cost or so, are all other US allies or TIP, nicely, from the US allies. I'm curious what your take is on those 10, there's work in the middle, but aren't formally allies. What is stuff, how likely are they to get involved with securing our superiority done on and what would hold them back from wanting to get involved?
这句话的翻译如下：
“这与那个类似，但涉及不同的数学。让我们看看这边，是的，先生。继续。是的，你在你的论文中提到关于安全问题，有中国、美国、水资源的健康状况和当前的能力。接下来的10个，或接下来的成本中，大多数都是美国的其他盟友或与美国盟友关系良好。我想知道你对这10个国家的看法，它们处于中间位置，但并不是正式的盟友。你觉得它们有多大可能参与保障我们的优势？又是什么因素可能阻止它们参与？”
The most interesting country is India, because the top AI people come from India to the US, and we should let India keep some of its top talent, not all of them, but some of them. And they don't have the kind of training facilities and programs that we are so rich we have here. To me, India is the big swing state in that regard. China's lost, it's not going to come back. They're not going to change the regime as much as people wish them to do. Japan and Korea are clearly in our camp. Taiwan is a fantastic country whose software is terrible, so that's not going to work. Amazing hardware. And in the rest of the world, there are not a lot of other good choices that are big. Europe is screwed up because of Brussels.
最有趣的国家是印度，因为顶尖的人工智能人才从印度来到美国。我们应该让印度保留一部分顶尖人才，而不是全部带走。他们没有像我们这里一样丰富的培训设施和项目。在我看来，印度在这方面是一个关键角色。中国已经丢失，不会再回来。不管人们多么希望，中国都不会改变其政权。日本和韩国显然是在我们这边的。台湾是一个很棒的国家，但他们的软件很差，所以这不行。不过硬件很出色。而在世界其他地方，没有太多规模大又合适的选择。欧洲因为布鲁塞尔搞得一团糟。
It's not a new fact. I spent 10 years fighting them, and I worked really hard to get them to fix the EU Act, and they still have all the restrictions that make it very difficult to do our kind of research in Europe. My French friends have spent all their time battling Brussels, and Macron, who's a personal friend, is fighting hard for this. And so France, I think, has a chance. I don't see Germany coming, and the rest is not beginning. So I'm younger than an engineer by training, and I can follow it from Tyler.
这不是新鲜事。我花了十年时间与他们抗争，努力让他们修改欧盟法案，但他们仍然保留了那些使我们在欧洲进行这种研究非常困难的限制。我的法国朋友一直在与布鲁塞尔作斗争，而马克龙（他是我的朋友）也在努力争取。所以，我认为法国还有机会。我看不到德国的动静，其他国家更是没开始行动。我是工程师出身，现在还年轻，所以能跟上泰勒的步伐。
Given the capabilities that UNVish and these models have, should we still spend time learning to code? Yeah, because ultimately it's the old thing of why do you study English, if you can speak English? You get better at it. Right, you really do need to understand how the assistance work, and I feel very strongly. Yes, sir. Yeah, I'm curious if you've explored the distributed setting, and I'm asking because, sure, like, making a large cluster is difficult, but Mac looks a powerful. There's a lot of small machines across the world. So like, do you think like folding a home or a similar idea works for training? Yeah, we've looked very hard at this.
鉴于UNVish和这些模型具备的能力，我们还需要花时间学习编程吗？ 是的，因为最后这就像是为什么要学习英语，如果你已经能说英语？你会变得更好。对，你确实需要理解这些辅助工具是如何工作的，我非常坚信这一点。 是的，先生。 我很好奇你是否研究过分布式设置，我问这个是因为，当然，建立一个大型集群是很困难的，但Mac 依然很强大。世界各地有很多小型机器。 所以你认为类似 于Folding@home的想法是否适用于训练？ 是的，我们对此进行了深入研究。
The way the algorithms work is you have a very large matrix, and you have essentially a multiplication function. So think of it as going back and forth, and back and forth. And these systems are completely limited by the speed of memory to CPU or GPU. And in fact, the next iteration of NVIDIA chips has combined all those functions into one chip. The chips are now so big that they glue them all together, and in fact, the package is so sensitive, the package is put together in a clean room, as well as the chip itself. So the answer looks like supercomputers and speed of light, especially memory interconnect, really dominated. So I think unlikely for a while.
算法的工作方式是你有一个非常大的矩阵，本质上有一个乘法函数。所以可以把它想象成来回重复操作。这样的系统完全受限于内存到CPU或GPU的传输速度。事实上，下一代 NVIDIA 芯片已经把所有这些功能整合到一个芯片中。这些芯片现在非常大，以至于需要把它们粘合在一起，实际上，这个封装极其敏感，需要在无尘室内组装，就像芯片本身一样。所以答案看起来像是超级计算机和光速，特别是内存互连，真正起到了主导作用。因此，我认为短期内不太可能出现更大的突破。
Is there a way to segment the element? Like, so Jeff Dean, last year when he spoke here, talked about having these different parts of it that you would train separately, and then kind of federate them. Each, in order to do that, you'd have to have 10 million such things, and then the way you would ask the questions would be too slow. He's talking about eight or 10 or 12. Yeah, yeah, yeah, yeah. So not to be a level of. Not at his level. Yeah. See, in the back, yes, way back. I know like after acting for at least New York Times, Sue, opening up for using their works for training, where do you think that's going to go and what that means for data press? I used to do a lot of work on the music licensing stuff, and what I learned was that in the 60s, there was a series of lawsuits that resulted in an agreement where you get a stipulated royalty whenever your song is played. Even they don't even know who you are. It's just paid into a bank. And my guess is it'll be the same thing. There'll be lots of lawsuits and there'll be some kind of stipulated agreement, which will just say you have to pay X percent of whatever revenue you have in order to use. That's cap. BMI. That's cap BMI. Look them up. It's along. It will seem very old to you, but I think that's how it will ultimately.
有办法对这个元素进行分段吗？比如，Jeff Dean去年在这里演讲时谈到了将其不同部分分开训练，然后再将它们联合起来。为了做到这一点，你需要有一千万个这样的东西，然后提出问题的方式会变得太慢。他谈到的是八个、十个或十二个。对，不是要达到某个级别。不是他的级别。对，看后面，是的，很久以前。我知道类似《纽约时报》这样的机构在允许使用他们的作品进行培训后，你认为这将走向何方，这对数据媒体意味着什么？我以前做过很多音乐许可方面的工作，我了解到在60年代，有一系列的诉讼导致了一项协议，无论你的歌曲何时播放，你都会得到约定的版税，甚至他们根本不知道你是谁，钱就直接存入银行。我猜测这次也会是一样的，会有很多诉讼，然后会达成某种协议，规定你必须支付一定比例的收入才能使用这些内容。这就是Cap和BMI，查一下他们的信息，可能会显得很古老，但我认为这就是最终的解决办法。
Yes, sir. Yeah, it seems like there's a few players that are dominating AI, right? And they'll continue to dominate. And they seem to overlap with the large companies that all the antitrust regulation is focused on. How do you see those two trends kind of? Yeah, like do you see regulators breaking out these companies and how will that affect the. Yeah. So in my career, I helped Microsoft get broken up and it wasn't broken up. And I fought for Google to not be broken up and it's not been broken up. So it sure looks to me like the trend is not to be broken up. As long as the companies avoid being John D. Rockefeller, the senior, and I studied this, look it up. It's how antitrust law came. I don't think the governments will act. The reason you're seeing these large companies dominate is who has the capital to build these data centers, right? Right. So my friend Reed and my friend Mustafa. He's coming next week. Reed, two weeks from now. Have Reed talked to you about the decision that they made to take inflection and essentially piece parted into Microsoft? Basically, they decided they couldn't raise the tens of billions of dollars. Is that number of public that you mentioned earlier?
是的，先生。是的，似乎有一些公司正在主导人工智能领域，对吗？而且他们将继续主导下去。而这些公司似乎与那些受到反垄断监管关注的大公司重叠。你怎么看待这两种趋势？你认为监管机构会解散这些公司吗？如果是这样，这将如何影响……对的。我在职业生涯中曾帮助微软避免被拆分，结果并没有被拆分。我还为谷歌不被拆分而努力过，直到现在它也没有被拆分。所以在我看来，趋势似乎是不拆分。只要这些公司不像约翰·D·洛克菲勒那样做事情，我不认为政府会采取行动。我研究过，这是反垄断法的起源。你之所以看到这些大公司主导，是因为谁有资本建设这些数据中心，对吧？对的。我的朋友里德和穆斯塔法。他下周要来，里德再过两周。里德有和你谈过他们决定将Inflection部分划给微软的事吗？基本上是因为他们觉得自己无法筹集到数百亿美元。你之前提到的那个数字是公开的吗？
Yeah. Have Reed give you a good one. Okay, maybe we can say it. I know you're. You gotta go. I don't want to hold you. I want to leave with. Shall we do this? Shall we do this? I have one more question for you. One more. Go ahead. Thank you so much. Thank you so much. Thank you so much. I was wondering where all of this is going to lead countries who are non-participants in development of frontier models and access to compute, for example. The rich get richer and the poor do the best they can. They'll have to. The fact of the matter is this is a rich country's game, right? Huge capital. Lots of technically strong people. Strong government support, right? There are two examples. There are lots of other countries that have all sorts of problems. They don't have those resources. They'll have to find a partner. They'll have to join with somebody else, something like that.
对，Reed 会给你一个好的建议。好吧，也许我们可以说。我知道你要走了，我不想耽误你。我想一起离开。我们要开始吗？我们开始吗？我还有一个问题，一个最后的问题。请问吧。非常感谢，非常感谢，非常感谢。我很好奇这对那些没有参与前沿模型开发和计算资源的国家会有什么影响。富人会变得更富，穷人只能尽力而为。实际上，这是富国的游戏，对吧？庞大的资本，众多技术强的人才，强有力的政府支持，对吧？有两个例子，还有许多其他国家有各种各样的问题，他们没有这些资源。他们必须寻找一个合作伙伴，必须和别人合作，类似这样。
I think the last thing we met you, we were at a hackathon at AGI House. And I know you spent a lot of time helping young people as they create a lot of wealth. And you spoke very passionately about wanting to do that. Do you have any advice for folks here as they're building their. They're writing their business plans for this class or policy proposals or research proposals? At this stage, the careers going forward. Well, I teach a class in the business school on this, so you should come to my class. The I am struck by the speed with which you can build demonstrations of new ideas. So in that, in one of the hackathons I did, the winning team, the command was fly the drone between two towers and it was given a virtual drone space. And it figured out how to fly the drone, what the word between meant, generated the code in Python and flew the drone in the simulator through the tower. I just would have taken a week or two from good professional programmers to do that.
我记得上次见到你是在AGI House的一个黑客松活动上。我知道你花了很多时间帮助年轻人创造财富，并且你对此非常热情。对于在场的人们，他们正在为这个课程编写商业计划、政策建议或研究提案，你有什么建议吗？处于现阶段，如何推进他们的职业生涯？
嗯，我在商学院教一门课，专门讲这些内容，所以你们应该来听我的课。让我印象深刻的是，大家可以很快地将新想法变成展示效果。在我参加的一个黑客松活动中，获胜的团队的任务是让无人机在两个塔楼之间飞行，并在虚拟空间中进行模拟。他们不仅理解了“在……之间”这个词的含义，还用Python生成了代码，并在模拟器中成功让无人机飞越了塔楼。要知道，这可是在一两周内就完成的，而专业程序员可能要花更长时间才能做到。
I'm telling you that the ability to prototype quickly, part of the problem with being an entrepreneur is everything happens faster. Well, now if you can't get your prototype built in a day using these various tools, you need to think about that. Because that's who your competitor is doing. So I guess my biggest advice is when you start thinking about a company, it's find the right of business plan. In fact, you should ask the computer to write your business plan for you. As long as it's. No, we. Yeah, actually talk about that after you leave this. And but I think it's very important to prototype your idea using these tools as quickly as you can because you can be sure there's another person doing exactly that same thing in another company and another university in a place that you've never been.
我告诉你们，快速制作原型的能力很重要，作为企业家问题的一部分是所有事情都发生得很快。如果你无法利用各种工具在一天内完成你的原型制作，你需要认真考虑这个问题。因为你的竞争对手就是这么做的。所以，我最大的建议是，当你开始考虑创业时，一定要找到合适的商业计划。事实上，你可以让计算机为你写商业计划。只要它是合适的。嗯，对，我们其实可以在你离开后再详细讨论这个问题。不过，我认为使用这些工具尽快制作你的想法原型非常重要，因为可以确定的是，在另一家公司、另一所大学里，还有其他人在做同样的事情，而且他们可能在你从未去过的地方。
All right. Well, thanks very much. Thank you all. I'm going to rush off. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. So actually, let me pick up on that very last point because I don't think I talked about in the first class about using LLMs, which is. It's important for the first time. .for each welcome in this class for the assignments, but it has to be. It has to be a full disclosure. So when you use them, whether it's for the weekly assignments or for the final project or whatever, just like you would if you asked your friendly uncle or classmate or anybody else who gave you advice, you should do that. Or if you have notes that you would include in there.
好的，非常感谢大家。我得赶紧走了。谢谢，谢谢，谢谢，谢谢，谢谢，谢谢，谢谢。所以，让我接着讲最后一点，因为在第一节课里我没有提到关于使用大语言模型(LLMs)的事。对每次作业都很重要，但一定要完全公开。无论是每周作业、期末项目还是任何其他任务，使用大语言模型时要像你向朋友、叔叔或同学请教一样，或者像你在使用笔记时一样写明来源。
So what I thought I'd do is. I want to talk a little bit about AIs. Is it GPT and what that means in terms of business and implications? But before we do that, I just want to see if there are any questions you want to pick up on things that Eric brought up that you. I'll try and channel some of his thoughts and we can talk about the things that came up and then we can move on. Yeah, go ahead. One other question I want to ask is, in relation to regulation, if the goal is to maintain supremacy, how do you create the right incentive so that everyone allies and non-allies are motivated, follow it? I mean, among companies that are competing with each other. Countries are in countries. Countries?
我的想法是这样的。我想先聊一点关于人工智能的问题，比如GPT，以及这些在商业上的意义和影响。在我们开始之前，我想看看大家有没有什么问题想针对Eric之前提到的话题提问。我会尽量代替他回答一些问题，然后我们可以讨论这些话题，之后再继续。可以，大家请提问。我还有一个问题想问，关于监管这个问题。如果目标是保持优势，怎样才能创造正确的激励机制，让所有人，无论是盟友还是非盟友，都受到激励并遵守这些规定？我的意思是，比如在彼此竞争的公司之间，或者在国家之间。
You as an MEO and it doesn't just become sort of a ham or obstruct kind of development for the ones that follow the. It's super tricky. There's a book co-opetition that Mary Nellbough wrote about this because there are definitely places where regulation can help companies and help an industry survive. So regulation doesn't necessarily slow. I mean, standards are a good example. And having that clarified can make it easier for companies to compete. So I've talked to a lot of the executives of these companies and there are places where they wish there was some common standards.
作为MEO（中层管理者），你不仅仅只是一个障碍或阻碍后续发展的存在，这是非常微妙的。有一本书叫《合作竞争》，是玛丽·纳尔鲍写的，里面提到一些地方的监管确实能帮助公司和整个行业存续发展。因此，监管不一定会减缓发展。例如，标准就是一个很好的例子，明确的标准可以让公司竞争得更轻松。我和很多公司高管谈过，确实有一些地方他们希望能有统一的标准。
And sometimes there's a bit of a race to the bottom as well on some of the dangerous things. One of the other reasons that the folks at Google say that they didn't move as fast is they felt like these LMs could be misused or dangerous, but their hand was sort of forced. And I was talking to some folks at one of the other big companies and they said, we weren't going to release this feature, but now competitors are doing it. So we're going to have to release it as well. So this is where regulation. There might be some interest in coordinating on regulation, but it's also. Obviously, the more obvious thing is that it is used to hinder competition.
有时在某些危险的事情上也会有一点竞争谁做得更极端的现象。Google 的一些员工说，他们没有那么快推出这些大型语言模型的原因之一是他们觉得这些模型可能被滥用或者存在危险，但他们也被逼得不得不行动。我和另一家大公司的员工聊天时，他们说，我们本来没打算发布这个功能，但现在竞争对手都在做，我们也只能跟着发布。这就涉及到监管的领域了。一方面可能会有人对协调制定监管感兴趣，但显而易见的一个情况是，监管也常被用来抑制竞争。
And a lot of people, for instance, think that the reasons that some of the big companies are very opposed to some of open source and making things more widely open source is they want to slow down competitors. So there's both of those things going on. Yeah, quick question over there. Yeah? I just want to follow up on. I was talking about, should we still learn code? Should we still study English? Are those going to use Google and Eric's? Like, yes, college educated high-skilled jobs or tasks are still going to be safe, but everything else that's in a parking job might not be. That's kind of like in the next one.
很多人认为，一些大公司非常反对开源和让更多东西开源的原因，是想要减慢竞争对手的速度。所以这两方面的因素都在起作用。好，有个快速问题。嗯？我只是想跟进一下。我之前在说，我们还应该学编程吗？我们还应该学习英语吗？这些会用谷歌和Eric的东西吗？是的，受过大学教育的高技能工作或任务仍然是安全的，但其他一些像停车场工作的可能就不那么安全了。下一个问题大概也是这样。
I think we can talk. Meebose talk some more about that in a few minutes, but it is interesting to think about where the AI system sort of just replaced what people are doing versus they compliment them. And in coding right now, it appears that they're not actually that helpful for the really best coders. They're very helpful for moderately good coders. But if you don't know anything at all about coding, they're not helpful either. So it's kind of an inverted U. And you can see why that would be the case that if you don't even understand the code that they generate right now is often buggy or it isn't exactly right.
我认为我们可以谈谈。几分钟后可以再详细讨论这个问题。但有意思的是想一想，AI系统在某些情况下取代了人类的工作，而在另一些情况下则是辅助人类。目前在编程领域，AI对非常优秀的程序员帮助不大，但对中等水平的程序员帮助很大。如果你对编程一无所知，AI也帮不了你。因此，这种情况有点像倒U形曲线。你可以理解为什么会这样：如果你连生成的代码都看不懂，AI生成的代码往往有错误或者不完全正确。
So if you can't even interpret it and understand what's going on, you can't use it very effectively. And for now, the very best coders. It appears that the code that is generated just isn't at that level, so you get that new shape. But that means if you don't know any code, you do need to have some in order to be useful. And I think that's true for a lot of applications right now that you have to have some basic understanding in order to get the most of it. I think it's an interesting open question if that's sort of always going to be the case.
所以，如果你连理解它都做不到，就无法有效地使用它。当前情况是，即使是最优秀的编码人员，生成的代码似乎也达不到那种水平，所以你得到了一个新的形态。但这意味着如果你根本不懂代码，你确实需要懂一些，才能有所作为。我认为，目前很多应用都是这样，你必须有一些基础的理解，才能充分利用它。我觉得这是一个有趣的开放性问题，不知道是否永远都会是这样。
I put up at the last class very briefly this slide that had level zero through five autonomous cars. And one of the things that actually we can talk about now is I'm trying to sort through is what if you took that paradigm and you applied it to all tasks in the economy? Like how many would they go through? So with autonomous cars, we aren't really at level five very much. Although I don't know how many of you guys have written in a Waymo, I mean one of the Waymo cars. So that one seems pretty good, although Sebastian Thrun, who I wrote in it with, says it's just incredibly expensive right now that it's not that they probably lose 50 to 100 dollars. He doesn't know he's not there. He started the program, but he's not there anymore. But just all of the costs of running it, it's not practical. Maybe it'll get down the curve, light hour will get cheaper, etc.
在上节课的末尾，我简要展示了一张幻灯片，讲述了零到五级的自动驾驶汽车。其中一个可以讨论的问题是，如果把这个模式应用到整个经济中的所有任务中，会是什么情况？比如说，有多少任务会经历这些阶段？对自动驾驶汽车来说，我们其实还没有达到第五级。虽然我不知道你们中有多少人坐过Waymo的车，但Waymo的车感觉还是不错的。尽管带我一起乘坐的Sebastian Thrun（他是这个项目的创始人，但已经不在那儿了）说，现在成本非常高，每次可能要损失50到100美元。他也不确定具体数字，因为他已经离开项目了。但总之，由于运行成本高，目前还不实际。也许随着经验积累和技术进步，诸如激光雷达等成本会变得更低。
But we have a lot of sort of autonomous cars at level two, three, even four arguably, where humans are still involved. And you see a lot of other tasks like coding, I just talked about that. On the other hand, chess, that slide before it, I talked about what's sometimes called advanced chess or freaked style chess, when Gary Kasperov after he lost the deep blue in 1998, 97, he started this set of competitions where humans and machines could work together. And for a long time, when I gave my TED talk, it was true in my TED talk in 2012, 2013, it was true at that time that a human working with a machine could beat deep blue or any chess computer. And so the very best chess playing entities were these combinations. That's not true anymore.
但是我们现在有很多属于二级、三级甚至四级的自动驾驶汽车，这些级别的汽车还是需要人类的参与。同时你会看到很多其他任务，比如编程，我刚刚提到过这个话题。另一方面，关于象棋，我之前提到过一种被称为“高级象棋”或“自由式象棋”的玩法，这是加里·卡斯帕罗夫在1998年97年输给深蓝之后发起的一系列比赛，在这些比赛中人类和机器可以协同作战。很长一段时间里，当我在2012年、2013年做TED演讲时，我的演讲内容是基于这样一个事实：人类和机器协同作战可以击败深蓝或任何象棋计算机。所以当时最强的象棋玩家组合就是这种人机协作模式。但现在这种情况已经不再成立了。
Alpha zero and other programs like that, they would get nothing from a human contributing, just be an annoyance to the chess machine. So that went through level zero, machines not being able to do anything, through a period where they work together, to a period where it's fully autonomous in a span of, I don't know, 20 years or so. It would be interesting if anybody wants to work as a research project, or if any of you guys have thoughts right now, what are the criteria for which kinds of tasks in the economy will be in that middle zone? Because that middle zone is kind of a nice one for us humans, where the machines are helping us, but humans are still indispensable to creating value.
Alpha Zero 以及类似的程序，它们无法从人类的贡献中获益，对象棋机器来说，人类的干预只是个烦恼。我们经历了一个阶段，即机器什么也做不了，然后到了一个机器和人类合作的阶段，再到现在机器完全自主，仅用了大约20年左右的时间。如果有人想把这作为一个研究项目来做，或者你们现在有任何想法，哪些经济领域的任务会处于机器和人类合作的中间地带呢？因为这个中间地带对我们人类来说其实蛮好，机器帮助我们，但人类在创造价值的过程中仍然是不可或缺的。
And that would be, that's a zone where you can have higher productivity, more wealth and performance, but also more likely to have shared prosperity because labor is sort of inherently distributed, whereas technology and capital, as Eric was just saying, potentially could be very concentrated. Do you have a thought on that? I was just going to ask a kind of a related question, he was saying also that we have a 10-year like chip manufacturing. I was surprised about that. Yeah, and I think what was interesting, like to me, is like a labor economist, that was really like a green flag I've seen in like literature and news that, okay, if we're on showing all of this chip manufacturing, isn't that going to create some sort of resurgence in blue collar jobs? And I wondered if you had any thoughts about intelligent robotic models or human labor?
这意味着，那是一个可以实现高生产力、更高财富和更好表现的领域，同时也更有可能实现共享繁荣，因为劳动力本身是相对分散的，而技术和资本则可能非常集中。你对此有何见解？我还想问一个相关的问题，他还提到我们有一个类似于10年的芯片制造计划。对此我感到惊讶。而且，从我的角度来看，作为一个劳动经济学家，看到这些文献和新闻让我觉得这是一种积极的信号。因为如果我们重新将芯片制造产业带回国内，这是不是会在蓝领工作中引发某种复苏？我想知道你对于智能机器人模型或人类劳动力有何看法？
I don't think it's going to be much of a, I mean, are you guys a visitor chip fab? Anybody? You guys have a few of you have. How many workers were in that fab? I was like, she has some CV, are you going to go in? So I don't know. Yeah, I mean, well, okay, so the answer is zero. Like the reason they don't let anyone go in, because we humans are too like clumsy and dirty and like, you know, we can't, this just, so it's all robotic. It's sealed inside. So there is like work to, you know, bring stuff to them, etc. And if a robot like falls over or something goes wrong, they have to put on, you've probably seen these like, like space suits, you know, they have to go in and then they kind of maybe adjust something and then they go back out and hope they didn't break anything.
我不认为这会是个大问题。我是说，你们有人参观过芯片制造厂吗？有吗？你们中有几个人去过。那里的工人有多少？我想说，她有一些经历，你会去参观吗？所以我不知道。嗯，我的意思是，答案是零。他们不让任何人进厂，因为我们人类太笨拙、太脏了，所以工厂内都是机器人操作的，整个环境都是封闭的。他们需要处理一些，比如说将物料送进去之类的工作。如果有机器人倒下或者出了什么问题，他们必须穿上那些类似宇航服的防护服，进去调整一下问题，然后出来，祈祷没有弄坏什么东西。
That's, so it's basically lights out. Yeah, I don't think it's, there are some, there is some like more sophisticated labor required that that I don't think it's like a blue collar research. In fact, one of the reasons that Apple reassured MacBook production to Texas is not because labor is so cheap in Texas or anything. It's that they don't actually require a whole lot of labor anymore. So it's a pretty labor, I think US manufacturing is surging in terms of output, but in terms of employment, it's not really growing all that much. Yeah. Let's go over here. Yeah. And what's the point coming for any things or test, test action models in the next? Oh, yeah. Well, he said what Eric, I'm hearing similar thing. Actually, he had a really nice way of putting those three trends. I've heard about them all separately, but I think it was good to bring them all together.
那基本上就是停工了。嗯，我不认为这里需要的是一些复杂的劳动力，而不是蓝领工人的工作。事实上，苹果将MacBook的生产迁回德州的原因之一，并不是因为德州的劳动力很便宜，而是因为他们现在实际上并不需要很多劳动力。所以美国的制造业在产量方面确实在迅速增长，但在就业方面并没有太大的增长。好，我们看看这边。嗯，有什么即将发布的新产品或测试模型吗？哦，是的，他说了什么，埃里克，我听到了类似的事情。实际上，他总结的那三个趋势非常好。虽然我以前分别听到过这些趋势，但我觉得将它们合在一起确实很好。
Earlier today, I was talking to Andrew Ng and he's like, been, been beating this drum about agents in particular as being sort of the wave of 2024 where Andrew had a nice way of describing it that like, as you guys know, like we, if you have an LLM, I don't know, write an essay or something like that, it writes it one word at a time and it just goes through in one pass and writes the essay. And it's pretty good. But imagine if you had to do that, like no backspace, no chance to let you know, make an outline first, you just kind of go through the agents now will say, okay, first make an outline, that's the first step you do when you write an essay, and then fill in each paragraph, then go back and see if the flow is right, now go back and check the voice, is this the right level for our audience? Now, and by iterating like that, you can write a much, much better essay or any kind of a task.
今天早些时候，我和Andrew Ng谈话，他一直在强调"智能代理"将成为2024年的重要趋势。Andrew用一种很好的方式描述了这个概念：大家知道，如果你用大型语言模型（LLM）来写一篇文章，它是一次性逐字写出来的，虽然效果不错，但过程非常机械化，无法进行反馈调整。想象一下，如果你写文章时不能用退格键，也不能先做大纲，只能一次性写完，那会多么困难。现在，智能代理能帮助你解决这个问题。第一步，它会让你先制定大纲，然后逐段填写内容，再检查整体逻辑是否流畅，最后调整语气是否适合目标读者。通过这种反复迭代的方式，你可以写出质量更高的文章或完成其他任务。
This is a real revolution. There's all sorts of things you can just do much better if you do that. Then the thing about the context, when it was also really important. So I'm just going to quote smart people that I know, Eric Corvitz, I was on a panel with him at the GSB, some of you may have been there, it's last week. And he had this nice taxonomy, people were asking him about fine tuning, I think he's asking about fine tuning. And he said, well, it's really, there's really three ways that you can take a model and have it more customized. One is you can fine tune it, which basically like train it some more. Another is with larger and larger context windows. And the third is with rag or techniques like that that are retrieval augmented generation where it goes and accesses external data. But these context windows seem to be like remarkably effective now. I guess, as Eric was saying, we thought it was hard. Maybe Peter can explain.
这真的是一场革命。通过这种方式，你可以做很多事情，而且会做得更好。至于上下文的问题，其实也非常重要。所以，我要引用一个我认识的聪明人——Eric Corvitz。我上星期和他一起参加了一个在GSB（商学院）的论坛，有些人可能也在场。他提出了一个很好的分类法。当时人们在问关于微调的问题，我想他就是在谈论微调。他说，其实有三种方法可以让模型更个性化。第一种是对它进行微调，基本上就是对它进行更多的训练。第二种是使用越来越大的上下文窗口。第三种是使用像RAG（检索增强生成）这样的技术，它会访问外部数据。但这些上下文窗口现在似乎特别有效。正如Eric所说，我们原本以为这是很难的。或许Peter可以解释一下。
But for some reason, we're able to make much, much bigger ones. And now as you can load like a whole book or whole set of books, you can load all sorts of informations in there. And that can give you all of the context around this. So that's a pretty big revolution. It opens up a bunch of capabilities that we just we just didn't have before, including having things much more current, as Eric was saying. Did you want to follow that? There's certainly a lot more capital going here, but that kind of begs the questions and comments. Why is all this capital going there as opposed to somewhere else? And I think, you know, if you look at the arc of history, sometimes it looks kind of smooth, but if you look more closely, there's a lot of jumps. There are certain big inventions and smaller inventions.
但是，出于某种原因，我们现在能够制造更大得多的东西。而现在你可以加载整本书或一整套书进去，你可以在里面加载各种各样的信息。这可以为你提供所有相关的背景信息。所以，这是一个相当大的革命。它开启了许多我们以前没有的能力，包括让信息变得更加及时，正如埃里克所说。你想继续说吗？这里确实有更多的资金涌入，但这也引发了我们对一些问题和评论的思考。为什么所有的资金都流向这里，而不是其他地方？我认为，如果你看看历史的轨迹，有时候它看起来比较平缓，但如果你仔细观察，会发现有很多跳跃。有些是重大的发明，还有一些是较小的发明。
And Andrew Carparthi was saying that he was playing around with physics and to really make progress in physics, you know, to be like a top physicist, you have to be incredibly smart, study a whole lot. And maybe if you're lucky, you could make some small incremental contribution, and some people do. But he says that right now in AI machine learning, we seem to be in an era where there's just a lot of low hanging fruit that there've been some breakthroughs. And instead of exhausting...
Andrew Carparthi 说他在研究物理，但要在物理学领域真正取得进展——成为顶尖的物理学家，你不仅需要非常聪明，还需要花大量时间学习。而且即便如此，如果你足够幸运，可能也只是做出一些小的增量贡献，虽然确实有些人做到了。但他表示，现在在人工智能和机器学习领域，我们似乎处于一个有很多“低垂的果实”（即容易实现的成果）的时代，已经有了一些突破。与其耗尽资源去追寻很难的问题...
the space, like picking all the food off of a tree, it's more like combinatorics in the second machine that talked about building blocks. When you put two building blocks together or Lego blocks, you can make more and more. Right now, we seem to be in an era where there's just a lot of opportunity, and people are recognizing that. And one discovery begets another discovery begets another opportunity. And because of that, it attracts the investment. And more people are involved. And in economics, sometimes when you more resources go in, you get diminishing returns, like in, I don't know, in agriculture or in mining.
这段话的意思是，当前我们处在一个充满机遇的时代，就像从树上摘果子一样，这些机遇更像是组合学中的构建模块。当你把两个构建模块或者乐高积木拼在一起，就能创造出越来越多的东西。现在，人们逐渐认识到这些机遇，而且一次发现会带来另一次发现，进而带来新的机遇。这种连锁效应吸引了更多投资，也让更多人参与进来。然而，在经济学中，有时候资源投入更多时回报会递减，比如在农业或采矿业中。
Other places, there's increasing returns. And more engineers coming to Silicon Valley makes the existing engineers more valuable, not less valuable. So we seem to be in an era where that's happening. And then the flywheel of the additional investment, the additional dollars for training, all of that makes them more and more powerful. I don't know how long this will continue. But I don't, I don't, you know, it just seems that there are some technologies that they hit this really fertile period. And there's, you know, positive feedback and some help that we seem to be in one of those right now.
在其他地方，收益越来越高。而更多工程师涌入硅谷，使得现有工程师变得更加有价值，而不是价值降低。所以，我们似乎正处在这样一个时代。然后，额外的投资、更多的培训资金等一切都在让他们变得越来越强大。我不知道这种情况会持续多久。但是，我觉得，有些技术进入了一个非常有利的时期，存在正反馈的作用。我们似乎正处在这样一个阶段。
So people who are trained in getting in the field are making contributions that are often quite, quite significant in a, you know, faster time than they might have in some other fields. Cursing all of you guys. I think they're doing the right thing right now. Yeah. Let's take a couple more questions and then, okay, how about over here? So not all, not everyone can sit in the room and have all these discussions and debates around AI. And so I'd like to get your thoughts on AI literacy for non technical stakeholders, whether they're policymakers, a happy make it in somewhat of judgment or the general public like music tech.
所以，那些在实地训练过的人往往会在比其他领域更短的时间内做出相当重要的贡献。诅咒你们所有人。我认为他们现在做的是正确的事情。好的，让我们再回答几个问题，然后，我们看看这边的情况。并不是每个人都可以在房间里参加关于人工智能的讨论和辩论。所以我想听听你对非技术相关方的人工智能知识的看法，不论是政策制定者，会作出某些判断的人，还是普通大众，比如音乐科技领域的人。
How do you think about explaining technical basics versus discussing abstract implications of domestic upper right now? Well, that's a hard one. I have to say, there's been a sea change recently in terms of how much people in Congress and elsewhere are paying more attention to this topic. It used to be not something that they were interested in. Now everyone's trying to understand it a little bit better.
你怎么看待解释技术基础和讨论国内高层抽象意义之间的区别？这个问题有点难回答。我得说，最近在国会及其他地方，人们对这一话题的关注度发生了巨大的变化。过去他们对此不感兴趣，现在每个人都在努力更好地理解它。
And I think that there are a lot of margins where people can make contributions. They can make contributions in the technical side. But if anything, I mean, my bet is that that the business and economic side is where the bigger bottleneck is right now, that, you know, even if, you know, if you made enormous contributions to technology side, you still, there's still a gap converting that into something that will change policy. So understand if you're into political science or a politician understanding what are the implications for democracy and for misinformation and power and concentration, those are things that are not well understood at all. I don't know that a computer scientist is necessarily the right person to try to understand that, but understanding enough about the technologies, you know, what might be possible.
我认为在人们可以做出贡献的领域有很多。技术方面是他们可以做出贡献的一个领域。但如果说有更关键的瓶颈，我认为是商业和经济方面。即使你在技术方面做出了巨大的贡献，仍然存在将其转化为能够改变政策的东西的差距。所以，如果你是研究政治学的，或者是一名政治家，了解这些技术对民主、虚假信息、权力和集中化的影响是非常重要的。这些影响目前还没有被很好地理解。我不确定计算机科学家是否是最适合来理解这些影响的人，但至少要了解技术方面的可能性。
And then thinking through what are the dynamics like Henry Kissinger was doing with Eric Schmidt in his book. If you're an economist thinking through the labor market implications, implications for concentration, implications for inequality, jobs, implications for productivity and what drives productivity, those are things that are very ripe right now. And you could go through lots of different fields where there's, you know, understanding well enough what the technology might be capable of, but then thinking through the implications, that's, I think, were some of the biggest payoffs. Sorry. I mean, let me give you a little bit more of a concrete example. And this is something I was going to talk about last week.
然后，像亨利·基辛格和埃里克·施密特在他们的书中所做的那样，深入思考这些动态。如果你是一位经济学家，你会考虑劳动力市场的影响、集中度的影响、不平等的影响、就业的影响、生产力以及驱动生产力的因素。这些方面现在都非常值得关注。你可以在许多不同的领域进行探讨，先充分理解技术的潜力，然后思考其影响，我认为这是最大的收获之一。不好意思，让我给你一个更具体的例子，这是我上周打算谈论的内容。
Electricity was also a general purpose technology. And general purpose technologies have this characteristic that they're probably in and of themselves. But one of the real powers of general purpose technologies, GPTs, as I was saying, is that they give complimentary, they ignite complimentary innovations. So, you know, electricity, light bulbs and computers and electric motors and electric motors give you compressors and refrigerators and air conditioning. You can just kind of have a whole set cascade of additional innovations from this one innovation. And most of the value comes from these complimentary innovations. One thing people don't appreciate enough is that some of the most important complimentary innovations are organizational and human capital complementarities.
电力也是一种通用技术。通用技术有一个特点，就是它们本身可能并不特别引人注意。但通用技术的真正力量之一，如我所说，是它们能够引发配套的创新。比如电力，它带来了电灯泡、计算机、电动机，而电动机又带来了压缩机、冰箱和空调。这样，从一个创新可以引发出一系列额外的创新。大部分的价值来自这些配套的创新。大家并没有充分认识到的是，一些最重要的配套创新是组织和人力资本的互补创新。
So, with electricity, when they first introduced electricity into factories, Paul David's here at Stanford studied what happened to those factories. And surprisingly, not much. The factories, when they started electrifying, they were not significantly more productive than the previous factories that were powered by steam engines. He's like, well, that's kind of weird, because this seems like a pretty important technology. Is it just a fad? Obviously not. The factories before electricity were powered by steam engines. They typically had a big steam engine in the middle and then crankshafts and pulleys that powered all the equipment. And it was all distributed. But it was, you tried to have it as close to the steam engine as possible, because if you make the crankshaft too long, it would break the torsion. When they introduced electricity, he found that in factory after factory, they would pull out the steam engine and they would get the biggest electric motor they could find and put it where the steam engine used to be. And fire it up. But, you know, it didn't really change production a whole lot. You can see that that's not a big deal.
那么，在工厂首次引入电力时，斯坦福大学的保罗·大卫研究了这些工厂的变化。让人惊讶的是，变化并不大。当工厂开始电气化时，它们的生产效率并没有显著高于之前使用蒸汽机的工厂。他觉得很奇怪，因为电力看起来似乎是一项很重要的技术。难道只是昙花一现吗？显然不是。在电力出现之前，工厂通常依靠蒸汽机运作，这些蒸汽机通常放在工厂中央，通过曲轴和皮带轮来驱动所有的设备。为了提高效率，设备尽量靠近蒸汽机，因为如果曲轴太长，会因为扭力而断裂。
当引入电力后，大卫发现各个工厂会将蒸汽机拆掉，换上他们能找到的最大的电动机，放在原来蒸汽机的位置，然后启动它。但这并没有对生产效率产生多大的影响。所以你可以看到，这不是一个大问题。
So then they started building entirely new factories from scratch in a new location. What did those look like? Just like the old ones. They would take the same model. Some engineer would make a blueprint, you know, maybe take a big X where the steam engine says, no, no, put electric motor here and they'd go and build a fresh factory. Again, not a big improvement in productivity. It took about 30 years before you started seeing a fundamentally different kind of factory, where instead of having the central power source, you know, a big one in the middle, you had distributed power because electric motors, as you guys know, you know, you can make them big, you can make a medium, you can make them really, really small, you can have them all connected in different ways. So they started having each piece of equipment have a separate piece of a separate motor instead of one big one. They called it unit drive instead of a group drive. I went and read the books in Baker Library at Harvard Business School from 1914 and it was like this whole debate about unit drive versus group drive.
于是他们开始在一个新地点从头建造全新的工厂。这些新工厂长什么样呢？跟旧的几乎一样。他们会用相同的模型。有工程师会做一份蓝图，也许在蒸汽机的位置打个大叉，说不，不，要在这里换成电动机，然后他们就去建一个新的工厂。不过，这对生产力提升并没有太大帮助。大约过了30年后，才出现了一种根本不同类型的工厂，不再使用一个中央电源，比如一个中央的蒸汽机，而是使用分布式电源。因为电动机的优点大家都知道，你可以把它们做得很大，也可以做中等大小，还可以做得非常非常小，并且可以以不同方式连接。所以他们开始让每台设备都有一个单独的电动机，而不是用一个大的电动机。这种方法被称为单驱动（unit drive），而不是群驱动（group drive）。我去哈佛商学院的贝克图书馆看了1914年的书，里面正在讨论单驱动和群驱动之间的争论。
Well, when they started doing that, then they had a new layout of factories where it was typically on a single story where the machinery was not based on how much power it needed, but based on something else, the flow of materials. And you started having these assembly line systems. That led to a huge improvement in productivity, like a doubling of productivity or tripling in some cases. So the lesson is not that electricity was a fad or a dud and was over hyped. Electricity was a fundamentally valuable technology, but it wasn't until they had that process innovation, that organizational innovation of rethinking how to do production that you got the big payoff. There's a lot of stories like that. I only told you one of them. We don't want that much time. So I tell you the other ones, but in some of my books and articles, if you look at the steam engine or others, you had similar generational legs decades before people realized that this technology could allow you to do something completely different than you used to do.
好的，当他们开始这样做时，新工厂的布局就有了新的变化，通常是单层的，机械设备的布局不再是根据所需的电力，而是根据另一种东西——材料的流动。于是，装配线系统开始出现。这带来了生产力的巨大提升，在某些情况下，生产力翻倍甚至三倍。所以，结论不是电力是一个昙花一现的潮流或夸大其词的东西。电力是一项具有根本价值的技术，但直到他们有了这个流程创新，即重新思考生产方式的组织创新，才得到了巨大的回报。有很多类似的故事。我只给你讲了其中一个。我们没有那么多时间去讲全部。所以我告诉你其他的故事在我的一些书和文章中，如果你看看蒸汽机或其他技术，它们在几十年前也有类似的阶段，人们才意识到这些技术可以让你做一些与以前完全不同的事情。
I think AI is a bit like that in some ways that there's going to be a lot of organizational innovation. It's going to be new business models, new ways of organizing an economy that we hadn't thought of before. Right now, people are mostly just retrofitting. I could go through a whole other set of skill changes that are complementary. I don't know what they all are. You have to be creative to think about them, but that's what the gap is. In the case of early computers, it's literally like 10 times more investment in organizational capital and human capital if you look at the size of the investments to the hardware and software.
我认为，某些方面来说，人工智能有点类似于这种情况，会有大量的组织创新。会出现新的商业模式，新的经济组织方式，是我们以前没有想到过的。目前，人们大多只是在进行改造升级。我可以列出一整套互补的技能变化，但我不确定它们全部是什么。你需要有创意来思考这些内容，这正是目前的差距所在。以早期计算机为例，如果你看一下硬件和软件的投资规模，实际上在组织资本和人力资本上的投资多了十倍。
So that's very big. That said, I'm open to adjusting my thoughts on this a bit because Chachi, PT, and some of the other tools, they have been adopted very quickly, and they have much more quickly been able to change things. In part, because you don't need to learn Python to the same degree, you can do a lot of things just in English, and you can get a lot of value just by putting them on top of the existing organization. Some of it's happening faster, and in some of the papers that you may have read for the readings here, we had 15, 20, 30 percent productivity gains pretty quickly. My suspicion is that even bigger once people figure out these complementary innovations, and that's a long way of answering your question about it. It's not just that the technical skills is figuring out all the other stuff, all the ways of rethinking things.
所以，这非常重要。话虽如此，我愿意稍微调整一下我对此的看法，因为Chachi、PT和其他一些工具被采用得非常快，它们改变事情的速度也更快。部分原因是你不需要同样程度地学习Python，你可以用英语完成很多事情，只需要将它们集成到现有的组织中就能获得很多价值。有些事情变化得更快，你可能在一些阅读材料中看到，我们很快就能实现15%、20%、30%的生产率提升。我猜测，一旦人们搞清楚这些互补创新，会带来更大的提升。这也是我对你这个问题的长篇回答。不仅仅是技术技能，还需要搞清楚其他所有的东西，所有重新思考的方式。
So those of you who are at the business school or in economics, there's a lot of opportunity there to rethink your areas now that you've been given this amazing set of technologies. Yeah, question. It seems like you're expressing more content than there was regard to the speed of transformation. Is that my correction? Well, yeah, I would make a distinction between two things. I'll defer to him and others on the technologies that we're going to hear from several other folks, and there are people who are equally optimistic as him, or even more optimistic on the technology side. There's also people who are less optimistic.
所以，对于那些在商学院或经济学领域的人来说，现在是重新思考你们领域的好机会，因为你们已经拥有了一套很棒的技术。好，有问题吗？你似乎在表达对转型速度比实际情况更满意。这个问题理解对吗？嗯，是的，我会在两件事之间做个区分。我会把技术方面的讨论留给他和其他几位专家，这里有些人对技术持非常乐观的态度，甚至比他还乐观。当然，也有些人没有那么乐观。
But technology alone is not enough to create productivity, so you can have an amazing technology. And then for various reasons, maybe people just don't figure out an effective way to use it. Another is it may be regulatory things. I mean, some of my computer science colleagues introduced, developed better radiology systems for reading medical images. They weren't adopted because of cultural, you know, people just didn't want them. They didn't want, and there are safety reasons. When I did an analysis of which tasks AI could help the most, and which professions were most affected, I was surprised that airline pilots was kind of near the top.
但是，仅靠技术不足以创造生产力。即使你拥有惊人的技术，由于各种原因，人们可能无法找到有效的使用方法。例如，可能会有监管方面的问题。我有一些计算机科学的同事，他们引入并开发了更好的医学影像读片系统，但由于文化等原因，这些系统并没有被采用。人们就是不想使用它们，而且也有安全方面的顾虑。当我分析哪些任务最能受益于人工智能、哪些职业受影响最大时，我很惊讶地发现航空飞行员居然排在前列。
But I think that a lot of people would not feel comfortable not having the pilot go down with you. They sort of, you want to have the human in there. So there are a lot of different things that might slow it down significantly. And I think that's something we need to be conscious of. And if we could address those bottlenecks, that would probably do more for productivity than just working on the technology alone. Yeah, question. So Eric, an interesting comment on data centers in universities. I think this is a larger point of like, I know it's the whites and keep writing track. All right. What is the role of the university ecosystem? Obviously, there is this larger, I'm sure all of the CS professors here. So I'll take, I mean, I think it'd be great if they were more funding into the federal government has something called the national AI resource that is helping a little bit, but it's in like the millions of dollars, tens of millions of dollars, not billions of dollars, let alone hundreds of billions of dollars. Although Eric did mention to me before class that they're working on something that could be much, much bigger. He's pushing for something much, much bigger. I don't know if it'll happen.
但是我认为，很多人会觉得没有飞行员一起降落会让他们感到不安。他们会希望有一个人类在里面。所以，有很多不同的因素可能会显著减缓这一进程。我认为这是我们需要意识到的问题。如果我们能够解决这些瓶颈，可能比仅仅在技术上做文章能更大地提升生产力。好的，问题来了。埃里克，你提到大学里的数据中心，这是一个有意思的话题。我认为这是一个更大的问题。我知道有一些对白保持关注。大学生态系统的角色是什么？显然在场的所有计算机科学教授...所以我想，如果联邦政府能提供更多的资金，那就太好了。他们有一个叫做“国家人工智能资源”的项目，略有帮助，但资金规模只有几百万，几千万美元，而不是几亿甚至几千亿美元。虽然埃里克课前跟我提到，他们正在努力争取一个更大规模的项目，不知道是否能实现。
That's for training these really large models. I had a really interesting conversation with Jeff Hinton once. Jeff Hinton, as you know, is sort of like one of the godfathers of deep learning. And I asked him like, what kind of like hardware he found most useful for doing his work. And he was sitting at his laptop and kind of just tapped his Mac book. And it just reminded me there's a whole other set of research that maybe universities have a competitive advantage in, which is not training $100 billion models, but it's elevating new algorithms like whatever comes after Transformers. And there's a lot of other ways that people can make contributions. So maybe there's a little bit of a division later. I'm all for, I'm support my colleagues asking for more budgets for GPUs. But that's not always where academics can make the biggest contributions.
这是为了训练这些非常大的模型。我曾经和Jeff Hinton进行过一次非常有趣的对话。你知道，Jeff Hinton被称为深度学习的教父之一。我问他，什么样的硬件对他的工作最有帮助。当时他坐在笔记本电脑前，轻轻敲了敲他的MacBook。这让我想起，大学在某些研究领域可能具有竞争优势，不是训练价值百亿的模型，而是提升新的算法，比如未来可能取代Transformers的算法。而且人们还有很多其他可以做出贡献的方式。所以可能将来会有一些分工。我完全支持我的同事们争取更多GPU预算，但这不一定是学术界能够做出最大贡献的地方。
Some of it comes from ideas and new ways of different perspective about thinking about things, new approaches. And that's likely where we have an advantage. I had dinner with a Sendom Alana Thon last week. He just moved from Chicago to MIT. And he was a researcher. We were talking about what is the comparative advantage of universities. And he made the case, you know, patience is one of them. That there are people universities who are working on very long term projects. You know, there's people working on fusion. They've been working on fusion for a long time, not because they're going to get, you know, a lot of money this year or 10 years from now probably from building a fusion plant or even 20 years. I don't know how long it is for fusion. But you know, it's just something that people are willing to work on even if the timelines are a little further.
其中部分优势源自于对事物的新理念和新视角，新的方法。这可能就是我们的竞争优势所在。上周我与Sendom Alana Thon共进晚餐。他刚从芝加哥搬到麻省理工学院，他是一名研究人员。我们讨论了大学的比较优势是什么。他提出耐心是其中之一。在大学里，有些人从事的是长期项目。他们在研究可控核聚变，已经做了很久，不是因为他们今年或未来十年、甚至二十年能从建设核聚变电厂中获得大笔资金。我不知道核聚变需要多长时间，但有人愿意在这个领域长期投入，即使时间表很远。
It's harder for companies to afford to have those kinds of timelines. So there's a comparative advantage or divisional labor in terms of what universities might be able to do. We have just a couple minutes left. This is kind of fun. So we'll just do one or two more questions. And I want to talk a little bit about the projects. Yeah. Yeah. I was wondering about the emerging abilities of a other group. Yeah. It seemed that Eric was leaning more towards the architectural differences and designing better models versus the last class we talked about in a more small instance. I wonder how you sort of write it. We said all three. So you guys remember the scaling laws? It had like three parts to, I think I put the scale in law that like Dario and team. So it was more compute, more data and algorithmic improvements, including more parameters.
公司负担不起那种时间表是越来越困难的。因此，在大学能做什么方面存在一个比较优势或分工。我们只剩下几分钟时间了。这有点有趣，所以我们再进行一两个问题。我想稍微谈谈一些项目。是的，是的。我想了解一下另一组的新兴能力。看起来埃里克更倾向于讨论架构的差异和设计更好的模型，而我们在上一节课更多地讨论了小例子。我想知道你是怎么概括的。我们提到了所有三种方法。你们还记得那个扩展定律吗？我好像提到了Dario和团队提出的扩展定律。它包含了三部分：更多的计算，更大的数据量和算法改进，包括更多的参数。
And all three of them, I think I heard Eric say all three of them were important, but not to be dismissed. This last one like new architectures. All three of them I think are being important. So I think there was a question in there though also. Was it like how much closer are we to like, like, HCI types system with these looking models? Eric doesn't think we're like that close to AGI type systems. Although I don't think it's like a sharp definition. You know, in fact, that was one of the I was going to ask him that question, but we're in a time would have been good to hear him, him describe it. But when what I was talking to him, it's just not that sharply a defined thing. You know, in some ways AGI is already here.
他们三个都很重要，我记得埃里克也说过，三个都很重要，但不能被忽视。最后一个像是新架构。我觉得三个都很重要。所以我认为这里应该也有一个问题。是关于像HCI类型系统与这些模型的接近程度吗？埃里克认为我们离AGI类型系统还不是很近。不过我觉得这个定义并不那么明确。事实上，我本来还准备问他这个问题，但是时间不够了，没办法听他详细解释。不过，当我与他交谈时，他也说这并不是一个明确的定义。从某种意义上说，AGI已经在这里了。
Peter Norvig would article called AGI is already here. I don't know if it's in the reading packet. I think if it's not, I should I'll put it in there. It's a it's a it's a fun little article with Blaze E. ARCA. Gary ARCA. And a lot of the things that, you know, 20 years ago, people would have said, this is what AGI is. That's kind of what LLMs are doing. Not as well maybe, but it's sort of solving problems in a more general way.
彼得·诺维格（Peter Norvig）会写一篇名叫《AGI已经来了》的文章。我不确定它是否在阅读材料中。如果没有，我会把它放进去。这是一篇有趣的小文章，提到了Blaze E. ARCA和Gary ARCA。许多年前人们认为的AGI（人工通用智能）的特征，现在大型语言模型（LLMs）已经做到了。虽然可能不会那么完美，但这些模型正在以更普遍的方式解决问题。
On the other hand, there's obviously many things they do much worse than humans currently. Ironically, physical tasks are one of the ones that humans have a comparative invention right now. And there's something you guys may know of more of X paradox. Hans Moravek pointed out that often the kinds of things that a three year old or a four year old can do, like, you know, buttoning a shirt or walking upstairs or very hard to get a machine to be able to do.
另一方面，显然有很多事情目前它们做得比人类差得多。具有讽刺意味的是，体力任务是人类目前拥有比较优势的领域之一。你们可能听说过莫拉维克悖论。汉斯·莫拉维克指出，通常三四岁的孩子能够做到的事情，比如扣钮扣或上楼梯，却很难让机器做到。
Whereas a lot of things that a lot of PhDs have trouble doing, like solving convex optimization problems are things that machines are often quite good at. So it's not quite a things that are easy for humans and hard for computers and other things that are hard for humans and easy for computers. They're not like the same scale.
很多博士生在解决凸优化问题时会遇到困难，而这恰恰是机器常常很擅长的事情。因此，并不是所有事情都可以简单地分类为对人类来说容易但对计算机来说困难，或者对人类来说困难但对计算机来说容易。其实它们之间并没有一个统一的衡量标准。
And next week, we have Mira Moravek, Chief Technology Officer of OpenAI, briefly the CEO of OpenAI. And so come with your questions for her. We'll see you.
下周我们将迎来OpenAI的首席技术官米拉·莫拉韦克，她曾短暂担任过OpenAI的CEO。请带上你的问题来见她。到时见。