普林斯顿-耶鲁“相约周末”品读汇第八场：新一轮技术扩散将如何

普林斯顿-耶鲁“相约周末”品读汇第八场：新一轮技术扩散将如何改变全球权力格局？

发布于 2025-01-09 · 时长 57:17

字幕摘录

时间	英文	中文
0:05	Good day, everyone.	大家好,再见
0:06	And especially to our presenter and commentator,	尤其是对我们的演讲人和评论员来说
0:09	Professor Jeffrey Ding and Yale World Fellow, George Chen.	杰弗里·丁教授和耶鲁世界研究员乔治·陈.
0:14	I'm Carol Lee Rafferty, a Yale College graduate	我是卡罗尔·李·拉费蒂耶鲁大学毕业生
0:17	and also executive director of Yale Center Beijing,	并兼任耶鲁中心北京执行董事,
0:21	Yale University's home here in China.	耶鲁大学在中国的故乡.
0:24	For the past 10 years, since our founding in 2014,	自2014年成立以来的10年中,
0:27	we've been hosting these types of activities	我们一直在主持这些活动
0:30	where we convene thought leaders from all over the world	我们聚集了来自世界各地的思想领袖
0:33	to discuss the most cutting edge issues facing humanity.	讨论人类面临的最尖端问题。

展开字幕全文（1034 条）

序号	英文	中文
1	Good day, everyone.	大家好,再见
2	And especially to our presenter and commentator,	尤其是对我们的演讲人和评论员来说
3	Professor Jeffrey Ding and Yale World Fellow, George Chen.	杰弗里·丁教授和耶鲁世界研究员乔治·陈.
4	I'm Carol Lee Rafferty, a Yale College graduate	我是卡罗尔·李·拉费蒂耶鲁大学毕业生
5	and also executive director of Yale Center Beijing,	并兼任耶鲁中心北京执行董事,
6	Yale University's home here in China.	耶鲁大学在中国的故乡.
7	For the past 10 years, since our founding in 2014,	自2014年成立以来的10年中,
8	we've been hosting these types of activities	我们一直在主持这些活动
9	where we convene thought leaders from all over the world	我们聚集了来自世界各地的思想领袖
10	to discuss the most cutting edge issues facing humanity.	讨论人类面临的最尖端问题。
11	Today, we're so delighted to be co-hosting this book talk	今天,我们很高兴能共同主持这个书评会
12	together with the Princeton University Press Beijing Office.	与普林斯顿大学出版社北京办事处合作。
13	This talk is part of the Princeton Yale Ideas Series	这个演讲是普林斯顿耶鲁思想系列的一部分
14	where we connect the Chinese audience and readers	我们连接中国观众和读者
15	with authors on the most relevant topics.	与作者讨论最相关的专题。
16	So what drives the rise and fall of great powers?	那么什么能推动大国的兴衰?
17	On this very important question, Professor Jeffrey Ding	关于这个非常重要的问题,杰弗里·丁教授
18	will discuss his new book with the Princeton University Press,	将讨论他的新书普林斯顿大学出版社,
19	Technology and the Rise of Great Powers,	技术和大国的崛起,
20	How Diffusion Shapes Economic Competition,	如何传播分裂经济竞争,
21	together with Yale World Fellow, George Chen.	与耶鲁大学世界研究员,乔治·陈。
22	They will discuss the mechanisms of technological diffusion	他们将讨论技术传播的机制。
23	and how they are shaping the pattern of economic competition	以及它们如何塑造经济竞争模式
24	among great powers.	在大国之间
25	Some of you may remember that George recently	你们有些人可能记得乔治最近
26	gave a talk at the Center on AI Governance.	在AI治理中心演讲。
27	So their insights are going to be especially topical	所以他们的洞察力将特别热门
28	to emerging technologies, such as AI,	研究新兴技术,如大赦国际,
29	and their potential impact on the US-China power dynamic.	以及它们对美中电力动力的潜在影响。
30	Let me briefly introduce our presenter and commentator	让我简单介绍一下我们的主持人和评论员
31	today.	今时.
32	Jeffrey Ding is an assistant professor	杰弗里・丁是助理教授
33	of political science at George Washington University	乔治·华盛顿大学政治学系
34	and serves as a reviewer for important academic journals,	并担任重要学术期刊的评论员,
35	such as the American Political Science Review.	例如美国政治科学评论。
36	Previously, he was a postdoctoral fellow	以前,他是博士后研究员
37	at Stanford's Center for International Security	斯坦福国际安全中心
38	and Cooperation.	7月12日至18日,日内瓦
39	George Chen is a practitioner and scholar	陈乔治是行家和学者
40	with over a few decades of expertise	拥有数十年的专门知识
41	across media, technology, and public policy.	包括媒体、技术和公共政策。
42	Currently, he's managing director and co-chair	目前,他是总经理兼联席主席
43	of digital practice for the policy consultancy,	为政策咨询、
44	the Asia Group, which is headquartered in Washington, DC.	亚洲集团,总部设在华盛顿特区。
45	He is also currently a senior fellow	他目前也是高年级同学
46	at the Center of Civil Society and Governance	民间社会和治理中心
47	of the University of Hong Kong, where	香港大学
48	he teaches digital society and governance to master students.	他教数字社会和治理掌握学生。
49	Prior to these roles, he was META, or Facebook's first ever	在这些角色之前,他是META, 或Facebook的第一个
50	and most senior policy leader for Greater China, Mongolia,	以及大中国、蒙古、
51	and Central Asia for almost a decade.	和中亚近十年。
52	And earlier in his career, George held various editorial	在他职业生涯的早些时候,乔治举行了各种社论
53	and management roles at leading media organizations,	以及主要媒体组织的管理作用,
54	such as the South China Morning Post and Reuters.	例如南华早报和路透社.
55	And actually, just when the center was founded in 2014,	事实上,就在2014年中心成立的时候,
56	George was elected as a Yale World Fellow.	乔治被选为耶鲁大学世界研究员.
57	So we're very, very honored and happy to have the two of them	所以,我们非常,非常荣幸和高兴有他们两人
58	at this event today.	今天的这个活动。
59	The event will happen and proceed as follows.	活动将进行如下。
60	Professor Ding will first give a talk on his book,	丁教授首先要谈谈他的书
61	and then George will provide his comments.	然后乔治会提供他的评论。
62	And then after that, the Q&A will begin.	之后,QQA将开始.
63	So before we dive in, I would like	所以在我们潜入之前我想
64	to extend a special thanks also to the Princeton University	还特别感谢普林斯顿大学
65	Press and Yale Center teams for making today's event	新闻和耶鲁中心的团队制作今天的活动
66	and the event series possible.	和可能的活动系列。
67	Professor Ding's book will be available for sale	丁教授的书可以出售
68	after the event.	事发后。
69	So without further ado, let's welcome Professor Jeffrey Ding.	所以,不用多说了,让我们欢迎杰弗里·丁教授。
70	Thank you very much.	谢谢
71	Thank you so much.	非常感谢
72	Thank you so much, Carol.	谢谢你,卡罗尔。
73	And thank you as well, George, for taking the time	也谢谢你乔治慢慢来
74	to join us today.	今天加入我们
75	I'm really looking forward to your comments	我很期待你的评论
76	and the questions from the audience.	还有观众的提问
77	I think when we look at this hot topic of US-China competition	我想,当我们看这个热门话题美中竞争
78	in AI, this image is helpful.	在AI中,这个形象很有帮助.
79	It's helpful for two reasons.	有二义故.
80	The first is oftentimes this competition	第一次经常是这场比赛
81	is seen as this geopolitical chessboard	被视为地缘政治棋盘
82	on which the different countries are trying	不同国家正在尝试的
83	to implement their own strategies.	执行自己的战略。
84	But the second reason it's helpful	但第二个原因是有帮助的
85	is anybody who plays chess or is following	是谁在下棋或正在跟踪
86	Ding Liren in the World Chess Championship	世界国际象棋锦标赛丁立仁
87	knows that this position is not possible.	也知道这个立场是不可能的。
88	The kings can't touch each other on a chessboard.	诸王在棋盘上不能相触.
89	And so I think this image is helpful	所以我觉得这个形象很有帮助
90	because it shows that our understanding of what	因为它表明我们理解什么
91	great power competition in emerging technologies	新兴技术的强大力量竞争
92	looks like is very confused and muddled.	看起来像很困惑和泥巴。
93	Maybe one starting point we can use	也许我们可以用一个起点
94	is Chinese leader Xi Jinping's speech to the BRICS	是中国领导人习近平对金砖国家的讲话
95	summit in 2018.	2018年峰会.
96	This summit of BRICS nations, Brazil, Russia, India, China,	本届金砖国家峰会巴西,俄罗斯,印度,中国
97	South Africa, all emerging economies	南非,所有新兴经济体
98	that are making up an increasingly large portion	占比越来越大
99	of world population and GDP.	人口与国内生产总值。
100	At this BRICS summit, Chinese leader Xi	在金砖国家峰会上,中国领导人习近平
101	talked about the Fourth Industrial Revolution.	谈了第四次工业革命.
102	And the whole summit was organized	整个峰会都是组织起来的
103	around the theme of the Fourth Industrial Revolution.	围绕第四次工业革命主题.
104	In the speech, some of the most important parts	在演讲中,一些最重要的部分
105	are he references these past three industrial revolutions	他提到过去三次工业革命
106	and specifically calls out specific technologies	并具体提到具体技术
107	in each of these industrial revolutions.	在每一个工业革命中。
108	What often happens in the Chinese system	中国制度经常发生的事情
109	is leaders give speeches and then analysts and commentators	领导人发表演讲,然后分析家和评论家
110	will try to interpret those speeches	将尝试解释这些演讲
111	a couple months later.	几个月后
112	So here's a publication from the Study Times,	这是研究时报的一篇文章
113	the Xuexi Shi Bao, a relatively authoritative publication	薛世保,相对权威的出版物
114	from the Chinese Communist Party,	从中国共产党,
115	where two months later they analyze this speech	两个月后他们分析这个演讲
116	and they highlight that when countries take advantage	它们强调,当国家利用
117	of industrial revolutions, the main impact	工业革命的主要影响
118	on the rise and fall of great powers	大国的兴衰
119	comes through the country that achieves	穿过国家实现
120	a world-leading productivity advantage.	世界领先的生产力优势。
121	Who gains leadership in advanced productivity	在先进生产力中获得领导的人
122	and becomes the most productive economy?	成为最具生产力的经济?
123	Those of you familiar with international relations	你们熟悉国际关系的人
124	or history know this seminal work	或历史知道这个开创性的作品
125	by historian Paul Kennedy,	历史学家保罗·肯尼迪,
126	where he describes a similar process.	在那里他描述了类似的过程。
127	The rise and fall of great powers	大国的兴衰
128	is a process that involves first,	是一个首先涉及的过程,
129	differentials in technological change,	技术变化的差异,
130	technological change leading to differentials	导致差异的技术变革
131	in economic growth rates,	经济增长率,
132	and then that eventually shapes shifts	然后最终会形成变化
133	in geopolitical and military leadership.	在地缘政治和军事领导中。
134	And so today, and for the purposes of my book,	所以今天为了我的书
135	I focus on the first step of that causal chain.	我着重谈谈这一因果关系的第一步。
136	How do these new technologies actually result	这些新技术的实际结果如何?
137	in a shift in the global economic balances?	全球经济平衡的转变?
138	Which country, which great power is able to take advantage	哪个强国能占上风
139	of these new technologies and sustain economic growth	利用这些新技术并维持经济增长
140	at higher rates than its rivals?	以高于对手的比率?
141	Standard story that we hear a lot today	标准的故事,我们今天听到很多
142	is what I call the leading sector theory.	这就是我所说的主要部门理论。
143	Yeah, leading sector theory could work.	是啊,领先部门理论可以工作。
144	Another way to translate that would be	另一种翻译方式是
145	leading sector theory.	领先的部门理论。
146	So leading sector theory, under leading sector theory,	因此,领先部门理论, 在领先部门理论,
147	one country dominates technological advances	一个国家主导着技术进步
148	in new fast-growing industries.	在新的快速增长的工业。
149	They get to control all the monopoly profits	他们可以控制所有的垄断利润
150	from being the innovation leader,	成为创新的领导者
151	and then they reinvest those,	然后他们再投资,
152	take the gains that they get from exports,	从出口中获得收益,
153	reinvest those into their economy,	重新投资于他们的经济,
154	and achieve world-leading productivity.	并实现世界领先的生产力。
155	Daniel Dresner, a scholar from Tufts University,	丹尼尔·德雷斯纳来自塔夫茨大学的学者
156	summarizes it well.	总结一下
157	A great power becomes a hegemon through a near monopoly	强权通过接近垄断成为霸主
158	on innovation in leading sectors.	关于领导部门创新问题。
159	And this might sound pretty familiar to you.	这听起来很熟悉
160	We all live and work in Beijing or Washington, DC,	我们都住在北京或华盛顿
161	and in policy-making circles,	在决策界,
162	policy-making circles are very much influenced	决策层受到很大影响
163	by this perspective.	从这个角度看。
164	What are leading sectors?	主要部门是什么?
165	Development economist Walt Rostow	发展经济学家沃尔特·罗斯托
166	gives this classic sequence of great leading sectors	给这些优秀的领军部门
167	that includes cotton textile industry	包括纺织业
168	in the first Industrial Revolution,	在第一次工业革命中,
169	the steel and chemical industries	钢铁和化学工业
170	in the second Industrial Revolution,	在第二次工业革命中
171	automobile industry in US-Japan competition	美日竞争中的汽车工业
172	in the third Industrial Revolution.	在第三次工业革命中
173	They oftentimes are very fast-growing single sectors	它们经常是快速增长的单一部门
174	that sometimes become one of the largest industries	有时成为最大的产业之一
175	in an economy.	在经济中。
176	Today, I'm going to argue that	今天,我要说
177	there's a different explanation	有一个不同的解释
178	for how technological revolutions	技术革命
179	lead to the rise and fall of great powers.	导致大国的兴衰.
180	And I want to focus on general purpose technologies,	我想专注于通用技术,
181	通用技術.
182	And my argument here is that general purpose technologies	我的观点是通用技术
183	or GPTs are going to shape economic competition	或GPT将形成经济竞争
184	between great powers in a pathway that's very different	大国之间在一个截然不同的道路上
185	from what I just started this talk with	我刚刚开始的谈话
186	as it relates to leading sectors.	因为它涉及主要部门。
187	And once we have a different sense	一旦我们有了不同的感觉
188	of the technology trajectory,	技术轨迹,
189	that's gonna shape what institutional factors	那会决定什么是制度因素
190	that different countries are going to have to adapt	不同国家将不得不适应
191	to be successful in technological revolutions.	在技术革命中取得成功。
192	My book focuses on the education and training systems	我的书主要讲述教育和培训系统
193	that widen the base of engineering skills	扩大工程技能基础
194	associated with the general purpose technology.	与一般用途技术有关。
195	So that was the two-minute overview of the argument.	因此,这是两分钟的概述。
196	Next, I'm going to fill out the theory in more detail.	接下来,我将更详细地填写这个理论.
197	Then I'm going to present some evidence	那么我要提出一些证据
198	primarily from the second Industrial Revolution.	主要来自第二次工业革命.
199	And then finally, we'll return back to the beginning	最后,我们回到开始
200	and answer this question of	并回答这个问题:
201	what are the US and China actually competing for in AI?	美国和中国在AI里竞争什么?
202	The basis for this argument is	这一论点的依据是:
203	new technologies present demands for countries.	新技术对各国提出了需求。
204	And so whichever country is gonna be most successful	所以无论哪个国家都会最成功
205	in adapting to a new technology	适应新技术
206	has to develop the institutions to meet those demands.	必须发展机构来满足这些要求。
207	And those institutions can range	这些机构可以涵盖范围
208	from a country's political system,	从一个国家的政治制度,
209	the level of centralization or decentralization	集中或权力下放的程度
210	in terms of how it approaches science and technology policy.	在如何对待科学和技术政策方面。
211	It could be the system by which countries	它可以是各国据以
212	organize their businesses.	组织企业。
213	Do they support big business alliances	他们支持大企业联盟吗?
214	like Japan's Koretsu system?	就像日本的科雷苏系统?
215	Or do they support a more competitive industrial landscape?	还是他们支持更具竞争力的工业景观?
216	My book focuses on the institutions of skill formation,	我的书主要讲述了技能的形成
217	talent cultivation.	培养人才。
218	The existing literature, that leading sector theory,	现有的文献, 领先的部门理论,
219	they want to focus on the demands of new technologies	他们想专注于新技术的需求
220	from the perspective of innovation,	从创新的角度出发,
221	創新.
222	Technological change can be mapped out	技术变革可以规划出来
223	into three different phases.	分为三个不同的阶段。
224	Invention is the first time a new discovery	发明是第一次新的发现
225	comes out in the world.	出来的世界。
226	Innovation is the first time that a company	创新是公司第一次
227	pioneers and commercializes that new invention.	开拓者和商业化新的发明。
228	Diffusion is that phase by which a new technology	传播是一个新技术的阶段
229	spreads across a population of users.	分布于用户群中.
230	So leading sector is all about that invention	所以领导阶层都是关于这个发明的
231	and innovation stage, that eureka moment.	和革新阶段, eureka的时刻。
232	And so they focus on a country's ability	所以他们专注于一个国家的能力
233	to adapt its institutional factors	适应其体制因素
234	for monopolizing innovation in leading sectors.	垄断主导部门的创新。
235	So they care a lot about how much is your country	所以他们很关心你的国家多少钱
236	spending on cutting edge R&D?	用于前沿研发?
237	Does your country have the systems to prevent	贵国是否有防止
238	these cutting edge innovations	这些前沿创新
239	from leaking out to other countries?	从泄露到其他国家?
240	This is based on something called	这是基于一个叫做
241	the leading sector product cycle.	主导部门产品周期。
242	Or the leading sector theory is based on something called	或者主要部门理论是基于所谓的
243	the product cycle theory.	产品循环理论。
244	And essentially this treats the nation like a firm	基本上这把国家当做企业
245	that has a new product innovation.	有了新的产品创新。
246	So your country comes out with a new smartphone innovation.	因此,你的国家有了一个新的智能手机创新。
247	They invested all this R&D.	他们投资了所有这些研发。
248	They came out with this fancy, shiny new product.	他们用这个闪闪发光的新产品出来了
249	And so there's a window of time where you are the only player	所以有一个时间之窗你是唯一的玩家
250	and you get to control monopoly profits.	你就能控制垄断利润
251	But shortly after, other competitors, other countries	但不久,其他竞争者,其他国家
252	gonna develop their own smartphones.	将开发自己的智能手机。
253	And once that competition increases, your profits decline	一旦竞争加剧你的利润就会下降
254	and you lose that window where you had monopoly profits.	你失去了那扇窗户在那里你拥有垄断利润。
255	So these are just a few quotes from international relations	所以这些只是一些国际关系的引言
256	and political science scholars	政治学学者
257	that are using that product cycle as a template.	将产品周期作为模板。
258	I think the last quote is what I want to emphasize	我想最后一句话是我想强调的
259	to you here, which is the leading sector account says	给这里的你, 这是主要部门帐户说
260	the country that is the pioneer in all these new innovations,	国家是所有这些新发明的先驱,
261	they get the benefits of monopoly profits	他们从垄断利润中获益
262	until the diffusion stage, right?	直到传播阶段,对不对?
263	That's where their story ends.	这就是他们的故事结束的地方。
264	They don't care about the diffusion stage	他们不在乎传播阶段
265	because that's when they think the advantages	因为那是他们认为优点的时候
266	to the leader go away.	对领导者走开。
267	For my alternative theory, that's where my story starts,	对于我的另一种理论, 这就是我的故事开始,
268	the diffusion stage.	传播阶段。
269	And so GPT diffusion theory focuses on general purpose	因此,GPT扩散理论侧重于一般目的
270	technologies, not leading sectors.	技术,而不是主导部门。
271	I think what I wanna emphasize here is that we really need	我想在这里强调的是我们真的需要
272	to understand the characteristics of GPTs	了解普通教育的特点
273	and how they are different from leading sectors.	以及它们与主要部门有何不同。
274	Sometimes when I read articles or blogs	有时我读文章或博客
275	on emerging technology policy,	关于新兴技术政策,
276	it feels like everything sounds the same	感觉一切听起来都一样
277	regardless of what technology they're talking about.	不管他们在说什么技术
278	If you saw a paper on quantum computing,	如果你看到一篇关于量子计算的文章
279	you could just find and replace quantum with blockchain	你可以找到量子用块链替换
280	or AI, it's all just boilerplate copied and pasted.	或AI,这一切都只是锅炉板复制和粘贴。
281	We need to take seriously the differences	我们要认真对待分歧
282	between technologies.	技术之间。
283	So let me just give a simple example.	因此,让我举一个简单的例子。
284	Everybody is really interested in China's advances	大家都对中国的进步很感兴趣
285	in electric vehicles.	在电动车辆。
286	EVs are a leading sector, not a GPT.	EVs是一个领先的部门,而不是GPT.
287	Let's compare EVs to something like AI,	让我们比较EVs和AI之类的东西,
288	which I think has that potential to be a GPT.	我认为它有可能是GPT。
289	EVs have one single use case, transportation,	EVs有一个单一的使用案例,运输,
290	getting from point A to B.	从A点到B点
291	AI has countless, I couldn't even list	AI有无数,我甚至不能列举
292	all the different use cases of AI	所有不同的使用案例
293	in terms of how it is affecting the ability to write,	如何影响写作能力,
294	the ability to code, the ability to generate images,	编码能力,生成图像的能力,
295	the ability to perceive landscapes	能够感知风景
296	in terms of applications in autonomous vehicles.	在自动车辆中的应用。
297	One way to really test this is think about	测试的方法之一是思考
298	are advances in electric vehicles	电动车辆的垫款
299	fundamentally transforming what's happening	根本改变正在发生的事情
300	in the AI industry?	在AI行业?
301	No, but think about the other direction.	不,但想想另一个方向。
302	Are advances in AI fundamentally transforming	AI的进步是根本的转变
303	what's happening in the electric vehicle industry?	电动车辆行业发生了什么?
304	Answers are resounding yes.	答案是强烈的是的。
305	One technology is more enabling than the other.	一种技术比另一种技术更有利。
306	One technology is more general purpose than the other.	一种技术比另一种技术更普遍。
307	So if we care about GPTs, the key point here,	所以,如果我们关心GPT,这里的关键点,
308	this is Stanford historian Paul David,	这是斯坦福历史学家保罗·戴维
309	he says that GPTs make their mark on the economy	他说GPT对经济有影响
310	through a regular pattern.	通过常规模式。
311	And this pattern is an extended trajectory	这种模式是延伸的轨迹
312	of incremental technical improvements over time,	随着时间的推移逐步进行的技术改进,
313	which results in a gradual and protracted process	导致一个渐进和旷日持久的进程
314	of diffusion into widespread use across the entire economy.	传播到整个经济的广泛使用。
315	Now, if we care about GPTs,	现在,如果我们关心GPT,
316	then that diffusion stage really matters	那么传播阶段真的很重要
317	because the potential population of users	因为用户的潜在数量
318	or adopters of a GPT is essentially the entire economy.	或收养人基本上是整个经济。
319	And so, of course, maybe if your country	所以,当然,也许如果你的国家
320	is the first to pioneer new technologies,	是率先开拓新技术,
321	that might help you with widespread adoption,	可以帮助你被广泛收养,
322	but it's not the only factor.	但这不是唯一的因素
323	And actually, when you think about competition	其实,当你想到竞争
324	between the most advanced economies,	在最发达经济体之间,
325	for a field as broad as AI or electricity or the computer,	对于像AI或电或计算机一样宽的领域,
326	they're all gonna have frontier firms	他们都会有边境公司
327	and universities who are close	大学和大学
328	to the technological frontier in a GPT, right?	在GPT技术前沿,对不对?
329	Think about the US and China today when it comes to AI.	想想今天的美国和中国,
330	No one country is monopolizing innovation in this space.	在这个空间里,没有一个国家垄断创新。
331	China's top universities, the Tsinghua's of the world,	中国顶尖大学清华世界大学 ,.
332	China's top firms, the Zhipu AIs of the world,	中国顶级企业,世界志普AI,
333	the Baidu's of the world are very close	贝都因人的世界很近
334	to that technological frontier.	在技术前沿。
335	And so, it might be the divergences,	所以,这可能是分歧,
336	the differences in long-term economic growth	长期经济增长的差异
337	might be more about which country can take those innovations	可能更多关于哪个国家可以接受这些创新
338	from their frontier firms	从他们的边境公司
339	and spread them throughout the entire economy.	并把它们传播到整个经济中。
340	Okay, last part of the argument.	好吧,最后一部分的争论。
341	I've shown you how GPTs make their economic impact	我给你们演示过GPT如何产生经济影响
342	in a way that's different from leading sectors.	以一种不同于主要部门的方式
343	Now, that means there needs to be different institutions	现在,这意味着需要不同的机构
344	that meet the demands of GPT diffusion.	满足GPT传播需求.
345	And so, I highlight the institutions	因此,我强调机构
346	that widen the pool of engineering talent linked to GPTs.	这拓宽了与GPT相关的工程人才库.
347	That's gonna address a couple of problems.	这会解决几个问题。
348	That's necessary to address a couple of problems.	这对解决几个问题是必要的.
349	One is pretty obvious, technology always races ahead	很明显,科技总是领先
350	and skills have to keep up.	和技能必须跟上。
351	But I think number two is especially important.	但我认为二号特别重要。
352	For a GPT to spread, you need to have coordination	要传播GPT,需要协调
353	and information flows between the GPT sector	GPT部门之间的信息流动
354	and all the different application sectors	以及所有不同的应用部门
355	that could use the GPT.	这可以用GPT。
356	You don't need that for leading sectors.	你不需要这个领导部门。
357	For electric vehicles,	电动车辆
358	you don't need the electric vehicle industry	你不需要电动车辆行业
359	to coordinate with the construction industry	与建筑业协调
360	or the education industry for electric vehicles	或电动车辆教育行业
361	to make an economic impact.	产生经济影响。
362	But for AI, the construction industry	但对于AI,建筑业
363	needs to know what's happening in the AI industry,	需要知道AI产业发生了什么
364	for AI to spread across sectors.	使大赦国际能够跨越各部门。
365	The education industry needs to know,	教育行业需要知道,
366	okay, what are the latest large language models?	好吧,什么是最新的大语言模型?
367	How do I benchmark whether this LLM is more efficient	我如何衡量LLM是否更有效率
368	or more effective than another LLM?	还是比其他LLM更有效?
369	Where is the AI sector going?	AI部门要去哪里?
370	Because then, in order for me	因为为了我
371	to develop complementary innovations,	发展互补创新,
372	I need to know what's happening in the GPT sector.	我需要知道GPT部门发生了什么
373	And so, engineering knowledge helps systematize	因此,工程知识有助于系统化
374	and standardize the best practices in a GPT	将最佳做法标准化
375	and facilitates those information flows	并促进信息流动
376	GPT 区间
377	and all the different application sectors	以及所有不同的应用部门
378	with their own unique demands.	他们自己独特的要求。
379	That's why, historically,	这就是为什么,历史上,
380	when we see the emergence of a new GPT,	当我们看到一个新的GPT的出现,
381	we often see a completely new engineering discipline follow.	我们经常看到一个全新的工程学科
382	So think electricity and electrical engineering	想想电机工程
383	or the computer and computer science.	或者计算机和计算机科学。
384	It doesn't have the word engineering in it,	里面没有工程这个词
385	but computer science is very much	但计算机科学非常
386	an engineering-oriented discipline.	一个面向工程的学科。
387	Okay, to sum up,	好吧,总结一下,
388	what we're trying to do here is asking	我们想做的是问
389	a very simple question.	一个很简单的问题。
390	How do technological revolutions affect	技术革命如何影响
391	and shape an economic power transition?	并形成经济实力的转变?
392	One great power sustains economic growth	一个大国维持经济增长
393	at higher rates than its rivals over the long run	从长远来看,比对手高
394	because of a productivity advantage.	因为生产力优势。
395	I've given you the conventional wisdom, leading sector,	我给了你传统的智慧,领导阶层,
396	and then I've given you my alternative theory, GPT diffusion.	然后我给你我的另类理论,GPT传播。
397	Now I'm mapping out that the technological trajectories	现在我正在绘制技术轨迹
398	differ on three dimensions.	三个维度不同。
399	When do the technological changes make their greatest impact?	技术变革何时产生最大影响?
400	Does it come very early	来得真早
401	in that brief window of monopoly profits?	在短暂的垄断利润窗口?
402	Or does it come in a very delayed and protracted process	或者是在一个非常拖延和漫长的过程中
403	like it is for GPTs?	就像是GPT吗?
404	Second dimension, what is the phase of technological change	第二方面,技术变革的阶段是什么
405	that is the most important for deciding	这对决定
406	which country gets ahead or not?	哪个国家领先呢?
407	Is it about innovation, monopolizing innovation,	是关于创新,垄断创新,
408	or is it about having that edge in diffusion,	或者说,那是关于传播的优势,
409	spreading that GPT across the most users	将GPT传播到大多数用户
410	within your country?	在你的国家?
411	Third, is technological change and economic growth	第三,技术变革和经济增长
412	concentrated in just a few leading sectors	集中在少数主要部门
413	or is it dispersed and broad-based	或它是分散和广泛的
414	like we would expect with a GPT?	就像我们期待的GPT?
415	Now lastly, that's going to affect	最后,这将会影响
416	the institutional adaptations that countries have to make.	(c) 各国必须作出体制上的调整。
417	If you care more about leading sectors,	如果你更关心领导阶层
418	you might care more about deepening the skill base,	你可能更关心深化技能基础,
419	training the best and the brightest in these new industries.	培训这些新产业中最优秀和最聪明的人。
420	If you care more about GPTs,	如果你更关心GPT,
421	you emphasize what I call GPT skill infrastructure	你强调我所说的GPT技能基础设施
422	and widening the engineering skill base	扩大工程技术基础
423	in these technologies.	这些技术。
424	Okay, enough theory.	好了,够了理论。
425	Let's move into actual evidence.	移入实证.
426	We're gonna rewind to the second industrial revolution.	我们要倒向第二次工业革命.
427	Note how the book covers the same three	注意本书如何涵盖同样的三个
428	industrial revolutions that the speech	工业革命的演讲
429	by Chinese leader Xi Jinping gave at the BRICS Summit.	由中国领导人习近平在金砖国家峰会上给出.
430	And so second industrial revolution, 1870 to 1914.	于是第二次工业革命,1870年至1914年.
431	Here we have all these different innovations.	在这里,我们有所有这些不同的创新。
432	Scholars see it as one of the densest periods	学者认为这是最密集的时期之一
433	of scientific advance.	科学进步。
434	And crucially, we see a shift	最重要的是,我们看到一个转变
435	in the economic balance of power.	在经济实力平衡中。
436	This is the decline of Great Britain	这是英国的衰落
437	and the rise of Germany and the US.	以及德国和美国的崛起
438	What I want to emphasize here is that	我想强调的是
439	when you look at the data on economic efficiency	当你看经济效率数据时
440	and economic productivity,	经济生产力,
441	the US becomes the preeminent economic power before 1914.	美国在1914年之前就成为主要的经济强国。
442	US is the blue line, Britain is the green,	美国是蓝线英国是绿线
443	Germany is the red.	德国是红色。
444	Germany never surpasses the UK	德国从未超越过英国
445	in terms of economic productivity.	在经济生产力方面。
446	So what we're doing in this exercise	所以,我们在做什么在这个练习中
447	is we are trying to ask how did the US	我们想问一下美国怎么样
448	benefit from the technological advances	从技术进步中受益
449	in the second industrial revolution	在第二次工业革命中
450	to become the preeminent economic power	成为强大的经济力量
451	before World War I?	在第一次世界大战之前?
452	Did it follow the GPT diffusion path	它是否遵循了 GPT 传播路径
453	or did it follow the leading sector theory?	还是遵循了主要部门理论?
454	And I'm just gonna go through all of those dimensions	我只是要通过所有这些维度
455	on this table step by step.	在这张桌子上一步一步。
456	So let's start with impact timeframe.	因此,让我们从撞击时间段开始。
457	I think all of the evidence in this era suggests,	我认为所有的证据在这个时代,
458	at least for the US,	至少对美国来说,
459	we don't see the near immediate impact	我们没有看到接近即时撞击
460	of some of these new technologies like chemicals.	包括化学等新技术。
461	Electrification does not have a significant impact	电气化没有重大影响
462	on US economic productivity until after 1914.	直到1914年之后,
463	So we get the advent of the electric dynamo in 1870,	所以我们在1870年有了电动堤坝
464	but it takes all of those decades for electricity	但是这需要几十年的电力
465	to diffuse throughout the US economy.	分散在美国经济中
466	Rather, the key technology in this era,	相反,这个时代的关键技术,
467	in my view, is this general purpose technology	我认为,这是通用技术
468	of interchangeable parts manufacture.	可互换部件的制造。
469	Essentially, we had these earlier innovations	基本上,我们有这些早期的创新
470	in the 1840s, 1850s with machine tools	1840年代,1850年代机器工具
471	like the turret lathe and the universal milling machine.	就像炮塔和通用的磨坊机
472	This is a Jichuang.	这是吉川号
473	And so these machine tools allow you to shape	所以这些机器工具可以让你塑造
474	and cut metal and wood in more precise ways	以更精确的方式切割金属和木材
475	so that you can have standardized parts.	这样就可以有标准化部件。
476	So when a bicycle breaks down,	所以当自行车坏了,
477	you don't have to get a whole new bicycle.	你不必买全新自行车
478	You can just replace that part	你可以换掉那部分
479	because you can cut the material	因为你可以剪掉材料
480	in more precise and standardized ways.	以更精确和标准化的方式。
481	And so these practices,	因此,这些做法,
482	these interchangeable parts manufacturing processes	这些可互换的零件制造工艺
483	are general purpose in that they can be used	属于一般目的,因为可以使用
484	to make not just bicycles, not just sewing machines,	不只是自行车, 不只是缝纫机,
485	but also other machine tools and firearms,	还有其他机器工具和火器,
486	and they affect all these different	并影响所有这些不同的
487	manufacturing industries.	制造业。
488	And so by the time the US is completing its economic rise,	所以到了美国完成经济崛起的时候
489	that's when these machine tools are spreading	这时这些机器工具正在扩散
490	throughout American industry.	整个美国工业。
491	Second question, is the US's advantage	第二个问题是美国的优势
492	based on innovation leadership or diffusion leadership?	* 以创新领导或传播领导为基础?
493	Actually, the US is not the definitive leader	事实上,美国不是最后的领袖
494	in innovation in leading sectors.	在领先部门创新。
495	When it came to electricity, for example,	比如说,在用电的时候,
496	the British Institute of Electrical Engineers	英国电气工程师学会
497	assesses that when it comes to inventive genius,	评估说,当它涉及到发明天才,
498	the UK is not that far behind.	英国没有落后那么远。
499	But when it comes to practical application,	但对于实际应用,
500	that's where the UK has fallen behind the US.	那是英国落后于美国的地方
501	In terms of interchangeable parts manufacturing,	在可互换零件制造方面,
502	a few indicators show that the US is outpacing	一些指标显示,美国的速度超过了美国。
503	Britain and Germany by more than double the rate	英国和德国比这高一倍多
504	by this time period.	截止这一时期。
505	And so Great Britain sends all these study teams	大不列颠派遣了这些研究小组
506	to the US to try to figure out what's happening here.	试着找出这里发生了什么
507	Why are we falling so far behind?	为什么我们落后这么远?
508	They report back that it's about this eagerness of diffusion	他们回报说是这种急切的传播
509	in the United States, spreading machinery	在美国,传播机器
510	across all different industries.	跨越所有不同的行业。
511	Third, the breadth of economic growth	第三,经济增长的广度
512	and technological change.	和技术变革。
513	It's not concentrated in just a few industries.	不只集中在少数行业.
514	The US has broad-based productivity growth in this period.	在此期间,美国生产力有广泛的增长。
515	Lastly, what accounts for the US's advantage?	最后,美国有何优势?
516	What accounts for the US's success	美国成功的原因
517	in the second industrial revolution?	在第二次工业革命中?
518	If my theory is right, we should expect that the US	如果我的理论是正确的, 我们应该期待美国
519	is able to widen the base of engineering talent.	能够拓宽工程人才基础.
520	It has the institutions that are better equipped	它拥有更好的机构
521	to widen the base of mechanical engineering talent	扩大机械工程人才基础;
522	for spreading this interchangeable parts	用于扩展此可互换部件
523	manufacturing method.	制造方法。
524	And that is the case.	这就是事实。
525	Britain did not produce enough mechanical engineers,	英国没有生产足够的机械工程师,
526	and Germany did produce enough,	而德国的产量是足够的,
527	but that mechanical engineering education	但机械工程教育
528	was not connected to industrial and commercial applications.	与工业和商业应用无关。
529	It wasn't standardized and systematized in that way.	它没有标准化和系统化的方式。
530	So here's one piece of evidence on that.	因此这里有一个证据。
531	This is from the US Bureau of Education report	这是美国教育局的报告
532	from around that time period,	从那个时期开始,
533	and they're translating	他们正在翻译
534	a German engineering professor's assessment.	a 德国工程教授的评估。
535	And the top three rows	还有前三行
536	are mechanical engineering schools	机械工程学校
537	represented of the German model.	代表德国模式。
538	The bottom two rows are engineering schools	下两排是工程学校
539	represented of the US model at that time.	当时代表美国模式。
540	The dark shaded lines, the black lines,	黑暗的阴影线,黑线,
541	are the amount of time in the curriculum	课程中的时间
542	dedicated to practical exercises in the lab and shop.	专门从事实验室和商店的实际练习。
543	The gray lines are dedicated	灰色线路是专用的
544	to oral and theoretical instruction.	接受口头和理论教育。
545	So you see how the US has a significant advantage	所以,你可以看到,美国如何有重大优势
546	in terms of the time devoted	时间方面
547	to mechanical engineering education	机械工程教育
548	that is based in practical exercises in the lab and shop.	这是在实验室和商店的实际练习。
549	I'm gonna skip the stuff on chemicals.	我会跳过化学物质。
550	Okay, that's a lot of words on that slide.	幻灯片上有很多字
551	I'm not gonna go through all of them.	我不会去通过所有他们。
552	What I wanna show in these other historical cases	我想在其他历史案件中展示什么
553	that are each their own chapter in the book,	各有自己的一章,
554	I just wanna show you that I'm doing the same exercise	我只是想告诉你,我做同样的练习
555	for all of them,	对于他们所有人,
556	going through these four dimensions	经历这四个维度
557	that we just went through	我们刚刚经历了
558	with the second Industrial Revolution case.	与第二件工业革命案.
559	Third Industrial Revolution, US-Japan competition	第三次工业革命,美日竞争
560	in the 1980s, 1990s.	(原始内容存档于20世纪80年代,90年代).
561	What's interesting here is that everyone expects	这里有趣的是,每个人都期望
562	that Japan will become the number one technological power,	日本将成为第一技术强国,
563	but it doesn't come to be.	但它不会来。
564	The leading sector mechanism can't really explain this	主导部门机制无法真正解释
565	because Japan dominates in all these new industries	因为日本在所有这些新兴产业中占主导地位
566	like consumer electronics, semiconductor components.	比如消费电子,半导体组件.
567	My theory can explain it.	我的理论可以解释
568	Japan does not lead the US when it comes to the adoption	在领养问题上,日本并不领导美国
569	of general purpose technologies, most notably the computer.	通用技术,尤其是计算机。
570	Okay, now let's circle back all the way to the beginning.	好吧,现在让我们回转一路回到开始。
571	What does this mean for US-China competition in AI?	这对AI的中美竞争意味着什么?.
572	I'm going to do the same exercise	我也会做同样的练习
573	where I look at the shape	在那里我看着形状
574	of the technological trajectory across three dimensions	三维技术轨迹
575	and then examine which country will be best prepared	然后再检查哪个国家能做好最好的准备
576	to adapt to the demands of AI.	适应大赦国际的要求。
577	And when most people discuss US-China AI competition,	当大多数人讨论美中AI竞争时
578	they're largely relying on the leading sector model.	他们主要依靠领先的部门模式.
579	My argument here is that on all these dimensions,	我的论点是在所有这些方面,
580	if you take my GPT diffusion model as a basis,	如果你把我的GPT扩散模型作为基础
581	you're going to have very, very different conclusions.	你会有非常,非常不同的结论。
582	So let's start.	所以,让我们开始。
583	Number one, impact timeframe.	第一,影响时限。
584	This is, in the leading sector model,	在主要部门模式中,
585	this is the National Security Commission	这是国家安全委员会
586	on AI's final report, a very influential body	关于大赦国际的最后报告,一个很有影响力的机构
587	set up in the US.	在美国设立。
588	I'm also citing Graham Allison, Harvard professor,	我还引用了哈佛教授格雷厄姆·艾利森的话
589	Eric Schmidt, former Google CEO, Lee Kaifu.	埃里克·施密特(英語:Eric Schmidt),曾任谷歌CEO,李启福.
590	I think most people here know him.	我想这里大多数人都认识他
591	He wrote this book on AI superpowers.	他写了这本书关于AI超能力。
592	They're very much rooted in the leading sector model	他们非常植根于领先的部门模式
593	and their timeframe for when AI could lead China	以及AI何时可以领导中国的时间范围
594	to surpass the US in technological leadership,	超越美国的技术领导,
595	their timeframe is very early.	他们的时间很早
596	It matches this expectation that leading sectors	它符合领先部门的预期
597	will have disproportionate impacts	将产生不成比例的影响
598	early after the sector's development,	在该部门发展初期,
599	all within this decade.	都在这十年里
600	For me, if you take GPT diffusion theory seriously,	对我来说,如果你认真对待GPT传播理论
601	and we know from past GPTs they take multiple decades	我们从过去的GPT中知道他们花了几十年的时间
602	to diffuse and make their mark throughout the entire economy,	在整个经济中传播并留下印记,
603	then let's set 2012 as the date	那我们把2012年定为日期
604	when this new deep learning paradigm was incubated	当这个新的深层学习范式被孵化
605	与 AlexNet 提交
606	(原始内容存档于2017-10-09). to Stanford's ImageNet competition.
607	If it's going to take multiple decades to diffuse,	如果要花几十年时间扩散
608	and remember, electricity took more than four decades	记住,电耗了40多年
609	to diffuse, then we shouldn't see major productivity payoffs	我们不应该看到生产力的回报
610	from AI until after 2030, okay?	从AI到2030年后,好吗?
611	Next, is it going to be about innovation leadership	接下来,这是关于创新领导
612	or diffusion leadership?	还是传播领导力?
613	So a lot of the analysis about whether China can catch up	所以很多关于中国能否赶上的分析
614	or overtake the US in AI focuses	或超过AI焦点中的美国
615	on China's innovation capacity.	中国的创新能力。
616	What is China doing when it comes to investments	中国在投资方面在做什么?
617	and cutting edge AI research and development?	和前沿AI研发?
618	What are the most valuable big firms and startups?	什么是最有价值的大公司和初创企业?
619	For me, I have an alternative perspective	对我来说,我有另一个观点
620	because I care more about is China competitive with the US	因为我更关心的是中国与美国的竞争
621	when it comes to the diffusion of AI	当涉及到AI的传播时
622	and some preliminary evidence comes	还有一些初步证据
623	from China actually ranks pretty middling	从中国来的,其实是名列前茅
624	when it comes to the diffusion of other information	在传播其他信息方面
625	and communications technologies.	和通信技术。
626	I'm citing some indexes collected	我引用一些指数
627	by the International Telecommunications Union.	由国际电信联盟提供。
628	China also lags the US significantly	中国也大大落后于美国
629	in terms of the diffusion of key digital technologies	关键数字技术的传播
630	like cloud computing, smart sensors and industrial software.	比如云计算,智能传感器和工业软件.
631	I'll point you to one of my other articles	我指给你看我的另一篇文章
632	where I argue that China faces a diffusion deficit.	我主张中国面临扩散赤字。
633	On the left, you have all these different indicators	在左边,你有所有这些不同的指标
634	of innovation capacity.	创新能力。
635	How much is your country spending on R&D?	贵国在研发方面的开支是多少?
636	What's the average R&D spending of your top three firms?	你们这三家公司的平均研发开支是多少?
637	Your Huawei's, your Alibaba's.	你的华威,你的阿里巴巴。
638	What's the average global ranking	全球平均排名是多少?
639	of your top three research universities?	你最好的三所研究型大学?
640	On all those metrics, China ranks very, very high.	在所有这些衡量标准上,中国排名非常,非常高。
641	Average ranking of 11.	平均排名为11位.
642	When it comes to diffusion capacity,	关于传播能力,
643	how intensive have information communications technologies	信息和通信技术的密集程度
644	permeated across businesses	渗透到企业中
645	and small size and medium size businesses?	小型和中型企业?
646	How well do the different parts of your science	你科学的不同部分有多好
647	and technology ecosystem talk to each other?	和科技生态系统谈话?
648	Is there strong linkages and connective tissue	是否有紧密的联系和连接组织
649	between your universities and your industry?	在你的大学和你的行业之间?
650	China's average ranking is more middling.	中国平均排名比较偏差.
651	Third point, is your technology policy directed	第三点,你的技术政策
652	at just a few leading sectors?	在几个主要部门?
653	Or are you recognizing that technological change	还是你意识到技术变革
654	and economic growth under the GPT model	GPT模式下的经济增长
655	is dispersed throughout all these different industries?	是否分散在所有这些不同的行业?
656	Here's actually a think tank affiliate	这其实是智囊团的附属机构
657	with the State Council, Guo Wuyuan here,	与国务院,郭武远在此,
658	where it's saying, hey, maybe China should reconsider	也许中国应该重新考虑一下
659	picking winners in terms of developing	在发展中挑选赢家
660	a few new technologies and move towards more	新的技术,并走向更多
661	of this horizontal economy-wide effort.	在全经济范围内开展的横向努力。
662	Last point, who's better positioned	最后一点,谁的位置更好
663	to compete in the age of AI?	在AI时代竞争?
664	The question here is, which country has the institutions	这里的问题是,哪个国家有机构
665	that are equipped to widen the base	* 具备扩大基地的装备
666	of engineering skills in AI?	在AI的工程技能?
667	Again, I think there's a preliminary case	再说一遍,我想还有个案子
668	for a US advantage.	为了美国的利益
669	One indicator from the book in the last chapter	上一章书中的一项指标
670	is I try to count up the number of universities	我试图计算大学的数量
671	in both countries that meet a very low quality baseline	达到非常低质量基线的两个国家
672	for training average AI engineers.	用于培训普通AI工程师.
673	So all you have to do is have, your university	你只要有你的大学
674	has to have at least one faculty, just one,	至少有一个系,只有一个
675	that has published in an established AI conference.	在一次既定的大赦国际会议上发表。
676	China, by that metric, has 29 universities.	中国有29所大学。
677	The US has 159.	美国有159个
678	So once you get past the Beidas, the Tsinghua,	所以一旦你穿过贝达人清华人
679	the Nanjing universities of the world,	世界南京大学,
680	by comparison, the US has a broader bench	相比之下,美国有更大的长椅
681	of institutions that can train average AI engineers.	能够培训普通AI工程师的机构。
682	When it comes to industry academic linkages,	关于工业学术联系,
683	remember, diffusion is all about getting different parts	记住传播就是要获得不同的部分
684	of the science and technology ecosystem	科学和技术生态系统
685	to talk to each other and have common standards	彼此交谈并有共同的标准
686	and protocols.	和协议。
687	One indicator of that is which country	其中一个指标是哪个国家
688	has more hybrid AI publications?	有没有更多的混合人工智能出版物?
689	So your AI publications that have at least one co-author	所以,你的AI出版物至少有一个合著者
690	from industry and one co-author from academia.	和一位来自学术界的作者
691	US leads the world in that metric,	美国在这个衡量标准上领先世界,
692	according to Stanford, HAI's AI index.	根据斯坦福,HAI的AI指数。
693	China faces significant challenges in this space.	中国在这方面面临重大挑战。
694	I'll leave you with some of my policy recommendations,	我给你留一些政策建议
695	which are mostly focused at the US government.	他们大多集中在美国政府。
696	This is a foreign affairs op-ed that I published	这是我出版的外事专栏
697	based on the book that says, currently,	根据这本书,目前,
698	the US's approach to competing with China	美国与中国竞争的方法
699	is very much based in the leading sector model.	以主导部门模式为主要基础。
700	We think that we can monopolize innovations in AI	我们认为我们可以垄断AI的创新
701	and prevent these cutting-edge technologies from leaking out.	防止这些尖端技术泄露出去。
702	This is the Fortress America approach.	这是美国要塞
703	I'm saying, instead, if we adopt a diffusion-centric model,	换句话说如果我们采用以扩散为中心的模式
704	it shouldn't be about Fortress America.	不应该是美国要塞的事
705	It should be about running faster than competitors	应该是跑得比竞争对手快
706	and prioritizing how to improve and sustain	并优先考虑如何改进和持续
707	the rate at which AI becomes adopted	采用AI的比率
708	in a wide range of productive processes.	在广泛的生产过程中。
709	And in this op-ed, I lay out a few specific proposals,	我在这篇署名文章中提出了一些具体的建议,
710	such as providing more support for community colleges	例如,为社区学院提供更多的支持。
711	to widen different pathways to training average AI engineers.	扩大培训普通AI工程师的不同途径。
712	Let me leave it there.	让我留在那里。
713	The holidays are coming up,	假期即将到来,
714	so if you want a good gift for a friend or yourself,	所以,如果你想要一个好礼物一个朋友或你自己,
715	please consider getting a copy	请考虑得到一份副本
716	and helping to diffuse the ideas in this book.	帮助传播这本书中的思想。
717	Thanks for your attention	谢谢你的注意
718	and looking forward to George's comments	并期待着乔治的评论
719	and the questions to follow.	和接下来的问题。
720	Can you hear me?	你能听到我吗?
721	All good?	没事吧?
722	Yes, we can hear you, George.	是的,我们可以听到你,乔治。
723	That's quite a lot.	这是相当多。
724	Actually, I'm impressed how you can	其实,我很佩服你能
725	pack the four industry revolutions	收拾四大工业革命
726	into 30 minutes, 40 minutes talk,	进入30分钟,40分钟的谈话,
727	helping us to go through the history,	帮助我们通过历史,
728	the ups and downs and everything.	升降和一切。
729	I'll just quickly give a key takeaways	我马上给钥匙外卖
730	as you were doing the presentation.	你当时在做演示
731	I took some notes.	我做了一些笔记。
732	I think the importance of the book is,	我认为这本书的重要性是,
733	it came at a very interesting time,	在一个非常有趣的时刻,
734	as you introduced it at the very beginning,	当你一开始介绍它时,
735	everybody talk about the US-China competition.	大家都在谈论中美竞争
736	And people know how important technology is,	人们知道技术有多重要,
737	but I think that this is very timely	但我认为这很及时
738	for not just politicians and policymakers	不只是政客和决策者
739	and also for the students and scholars	以及学生和学者
740	to look into how the US-China tensions	调查美中紧张局势
741	started to emerge and get worse	开始浮现变得更糟
742	and how that is related to the competition for technology.	以及这与技术竞争的关系。
743	So I think that in that aspect,	因此,我认为在这方面,
744	the book gives a very fresh perspective	书给出了一个非常新鲜的视角
745	to understand the politics around the two big powers.	了解两个大国的政治
746	We perhaps talk too much about the trade war.	我们也许谈论太多关于贸易战争的问题。
747	I think it's obvious as Biden talk over the White House,	拜登谈论白宫的事时很明显
748	the trade war quickly shifted to more like a technology war	贸易战争迅速转向更像是科技战争
749	and then all kind of like a mixture	然后像混合物一样
750	of trade war plus tech war.	贸易战和科技战
751	Even if you look at the trade war nowadays,	即使你看看现在的贸易战
752	people talking about the export bands, chips and AI,	人们谈论出口带,芯片和AI,
753	this is still very much related to technology.	这仍然与技术密切相关。
754	So that's point one.	这就是第一点
755	Secondly, I think you viewed your research	第二,我觉得你看过你的研究
756	also based on some previous scholars research	也基于以前一些学者的研究
757	Paul Kennedy is a great scholar.	保罗·肯尼迪是一位伟大的学者.
758	I think he was the first to sort of looking	我觉得他是第一个看
759	to how great a country could be	国家是多么伟大
760	and how a country could fall	以及一个国家如何崩溃
761	in terms of its contribution to the world economy.	在对世界经济的贡献方面。
762	Before Paul Kennedy, people talk more about like gun power.	在保罗·肯尼迪之前,人们更多地谈论枪权.
763	As Chairman Mao said, you have guns, you have the regime.	正如毛主席所说,你们有枪,你们有政权.
764	People talk about how many nuclear missiles you have.	人们谈论你有多少核导弹。
765	I think Paul very much looking	我觉得保罗非常看
766	to the economic power perspective.	从经济实力的角度
767	By the way, I think he's still teaching	对了,我想他还在教书
768	at the Yale University,	在耶鲁大学,
769	just three and a half years of advertisement for Yale.	只有三年半的广告为耶鲁。
770	He's getting older,	他变老了
771	but I think he's still doing some lectures.	但我觉得他还在做一些讲座
772	The other finding, and it's just to me,	另一个发现, 它只是对我来说,
773	even I worked in the tech industry for a long time,	甚至我在技术行业工作了很久
774	which was quite inspiring,	这是相当鼓舞人心的,
775	is you talking about the difference between,	你说的是区别,
776	of course, people understand the invention and innovation,	当然,人们理解发明和创新
777	but I think the third dimension is actually more interesting.	但我觉得第三个维度更有趣
778	The diffusion, I think I really like the example	传播,我觉得我真的很喜欢这个例子
779	you used why Chinese EVs are so popular.	你用过为什么中国EV这么流行
780	And then you are in Beijing,	然后你们就在北京
781	our audience in Beijing,	我们在北京的观众,
782	if you go out and then look at public policy	如果你出去看看公共政策
783	in the Beijing Straits, you will be very impressed.	在北京海峡,你会印象深刻。
784	So how quickly a technology can be applied for general use,	因此,技术如何能迅速应用于一般用途,
785	that also affects a lot.	这也影响很大。
786	And then in comparison, I think the US is still hugely	相比之下,我认为美国仍然是巨大的
787	left behind in terms of the EVs for the public,	留下的EV给公众,
788	like for ordinary people to get on it.	喜欢普通人得到它。
789	It's getting there, but the penetration ratio is still	已经到了,但是渗透率还是
790	far lower even than those Nordic countries.	远低于那些北欧国家。
791	So that's very interesting.	这很有趣
792	I think perhaps we should hear more from you,	我想我们该多听听你的话
793	if we look from diffusion perspective,	如果我们从扩散的角度来看,
794	why some countries sort of have the first mover advantage	为何某些国家拥有第一个推动者优势
795	like Japan, right?	像日本,对不对?
796	But how in the end that they didn't really	但最后他们怎么没有
797	became the champion while other countries,	成为冠军而其他国家,
798	and I think now China has sort of the potential,	我想现在中国有潜力了
799	maybe in some ways, China was still behind the US	也许在某些方面,中国仍然落后于美国
800	in terms of AI developments,	关于大赦国际的发展情况,
801	but in terms of how fast the AI adoption could be	但从人工智能的通过速度来看
802	for the society and business,	为社会和企业,
803	China could take a very good lead against the US	中国可以对美国采取很好的领先
804	in the next five to 10 years.	在接下来的五到十年。
805	I think this is also why the US is a bit worried.	我想这也是美国担心的原因。
806	And I also read some literature recently,	我最近还读了一些文献
807	people say the AI adoption in Japan, for example,	比如说日本的人工智能收养
808	is pretty low, like around 10%.	相当低, 大约10%。
809	While in China, the AI adoption in business	在中国,AI在商业上的收养
810	is already around 40%.	已经约40%。
811	It's arguable Japan also has great AI scientists	日本也有伟大的AI科学家
812	and researchers, but how come Japan is a little bit slow	和研究人员,但为什么日本有点慢
813	when it comes to AI adoption for business and society?	当涉及到AI的收养企业和社会?
814	That's also an interesting question,	这也是一个有趣的问题,
815	especially when we think from the diffusion perspective.	特别是当我们从传播的角度思考的时候。
816	I think related to that,	我觉得和那个有关
817	your argument about the leading sector model,	您关于领先部门模式的争论,
818	the so-called the to yang you shi,	所谓的"向阳你什,
819	whether a country should just focus on several sectors,	一国是否应当仅仅侧重于若干部门,
820	like China has also those national champions,	就像中国也有那些全国冠军
821	like BAT and Huawei, that's one aspect.	像BAT和华威, 这是一个方面。
822	The other aspect is still the adoption.	另一个方面仍然是通过。
823	So I think this is also an interesting argument.	所以我认为这也是个有趣的论点。
824	The car industry in the US is actually a quite sad story.	美国的汽车工业实际上是一个相当悲哀的故事.
825	I will argue they used to have very good advantage,	我会说他们以前很有优势
826	the first mover advantage, the leading sector advantage,	领先的行业优势
827	but how come it just completely left behind	但为什么它只是完全留下
828	in the whole competition,	在整个比赛中,
829	when EV emerging as a new technology, right?	当EV作为一种新技术出现,对不对?
830	So that's also an interesting angle.	这也是一个有趣的角度。
831	So I have a couple of questions,	我有几个问题
832	but I don't want to occupy the whole time.	但我不想占用整个时间。
833	I'm pretty sure that the audience will also have	我敢肯定,观众也会有
834	a lot of sorts of questions to share.	有很多问题要问
835	So maybe I start there with one question.	所以也许我先问一个问题
836	I don't know if perhaps less relevant to the book,	我不知道是否与这本书无关
837	but I was also always a bit puzzled.	但我总是有点困惑
838	So you didn't have time to touch on sort of the way	所以,你没有时间接触某种方式
839	how internet and the cyberspace is regulated and governed	因特网和网络空间是如何管理和管理的
840	in different countries.	在不同国家。
841	I think that also matters in terms of innovation.	我认为,就创新而言,这一点也很重要。
842	But one question people always ask is,	但人们总是问一个问题:
843	okay, so if the conventional wisdom will say,	好吧,所以如果传统智慧会说,
844	okay, so innovation will happen	好吧,所以创新会发生
845	when you have a more free environment	当你有一个更自由的环境
846	for research, academic research or scientific research	研究、学术研究或科学研究
847	and this and that, but how can we argue,	和这个,和那个, 但我们如何辩论,
848	like everybody knows like China's internet	就像大家知道中国的互联网
849	is very restrictive, perhaps the most restrictive	限制性很强,也许是最严格的
850	among the major policy in the world,	在世界上的主要政策之一,
851	but how can we argue like even given such high censorship	但是,我们怎么可以争论像即使如此高的审查
852	and tight control of the internet	严格控制互联网
853	and in a way that also affect the innovation,	并以同样影响创新的方式,
854	but we still have quite successful internet companies,	但我们仍然有相当成功的互联网公司,
855	naming the BAT, right?	命名BAT,对不对?
856	Does that mean Silicon Valley,	这是否意味着硅谷,
857	yes, they have the free environment,	是的,他们有自由的环境,
858	but the freedom in governance is not necessary	但管理自由并不必要
859	a key element to push for more innovation.	推动更多创新的关键因素。
860	I'll just leave the question there.	我就把问题留在那里
861	So maybe you can have some thoughts.	所以,也许你可以有一些想法。
862	And then the second question is more about	然后第二个问题更多关于
863	the US-China competition.	中美竞争
864	So crystal ball, but you focus very much	如此的结晶球,但你非常集中
865	on the argument about the diffusion.	关于传播的争论。
866	So I don't know if the US government can get it.	我不知道美国政府能不能拿到
867	If the current attention continues to rise,	如果目前的注意力继续增加,
868	it's going to be lose-lose for both US and China	美国和中国都会输
869	in terms of basically we talk about	基本上,我们谈论
870	how to make the cake bigger, right?	如何使蛋糕更大,对不对?
871	Rather than everyone just try to grab one slice.	而不是每个人都想抓一块
872	I don't have a good answer to that,	我对此没有很好的答案
873	but I'm also curious to hear your thought.	但我也很想知道你的想法。
874	Is there a solution for both sides to see,	有没有一个解决方案让双方看到,
875	this is not a zero-sum game in the,	这不是零和游戏,
876	currently we are in the fourth industrial revolution, right?	我们正处于第四次工业革命中,对吧?
877	Or is this just the fate of the world?	还是这仅仅是世界的命运?
878	Because if you look back into the history,	因为如果你回顾历史
879	every industrial revolution so far	至今为止的每一个工业革命
880	can only have one strong lead like to survive,	只有一个很强的线索能活下来,
881	rather than two countries can share the power	而不是两个国家可以分享权力
882	in some key innovation technology.	在一些关键创新技术中。
883	This might be also a deep concern	这可能也是一个令人深为关切的问题
884	inside the many American politicians' mind.	在很多美国政客的思想中
885	There's no like a number two.	何等为二.
886	The world only recognize number one	全世界只认得第一
887	when it comes to key technologies.	在关键技术方面。
888	So I'll just leave the two questions there.	所以我把两个问题留在这儿
889	I'm pretty sure that the audience	我敢肯定,观众
890	will have their own questions too.	他们也会有自己的问题
891	Thanks a lot.	非常感谢
892	Great book, great presentation.	很棒的书很棒的介绍
893	I think this is very timely a bit	我觉得这很及时
894	because the whole world is so crazy,	因为整个世界都这么疯狂,
895	like people are just confused.	就像人们只是困惑。
896	I think when we're confused,	我想当我们感到困惑时,
897	it's always useful to look into history,	研究历史总是有用的
898	to show some lessons from what happened	给大家一些教训
899	and to see what could happen next.	看看接下来会发生什么
900	Thanks.	谢谢
901	Yeah, thank you so much, George.	是的,非常感谢,乔治。
902	Really appreciate you taking the time to dig into this	非常感谢你花时间去调查
903	and give your own thoughts based on your experience	并根据你的经验表达你的想法
904	and your unique vantage point	和您独特的优势点
905	and also raising some really thought-provoking questions.	也提出了一些令人深思的问题。
906	Let me go through some of your comments	让我听听你的意见
907	and then we'll get into the broader Q&A.	然后我们进入更宽的QQA.
908	So first, before I address the two questions,	所以首先在我回答这两个问题之前
909	I think it's important to go in depth on this EV example.	我认为必须深入探讨EV的例子。
910	EVs are not a general purpose technology.	EV不是通用技术.
911	So yes, China might be ahead in terms of EV adoption,	是的,中国在EV收养方面可能领先
912	but I don't see that as being connected	但我不认为这是连接
913	to widespread productivity growth	实现生产力普遍增长
914	as much as the potential of AI.	就像AI的潜力一样
915	So I do think China has shown	所以,我认为中国已经显示
916	that it can be a strong diffusion leader	它可以是一个强大的传播领导者
917	in two types of settings.	在两种类型的设置中。
918	One is when it's about scaling up infrastructure,	一是扩大基础设施规模
919	which can be done with a top-down	用自上而下的方法完成
920	coordinated approach from the government.	政府的协调方针。
921	So example there is high-speed rail.	因此,有高速铁路。
922	Two is, and maybe EVs fall under this category,	二是,也许EV属于这个类别,
923	when adoption by consumers is relatively cheap	当消费者的收养相对便宜时
924	and doesn't have too many barriers to entry.	并且没有太多的进入障碍。
925	And so think mobile payment systems	所以想想移动支付系统
926	or food delivery systems.	或食物输送系统。
927	Diffusion of GPTs is much more difficult	GPT的传播要困难得多
928	because you need businesses to adopt it	因为你需要企业收养
929	and embed it into productive processes	并把它嵌入生产过程
930	to improve productivity.	提高生产率。
931	It can't be coordinated from the top down.	从上而下无法协调.
932	See example A of the Soviet Union.	参见苏联的例子A.
933	And it's much more difficult to do	更难做到
934	than just consumer-facing adoption.	不只是消费者的收养。
935	With EVs, the consumer doesn't need,	用EVs,消费者不需要,
936	the user doesn't need to have an engineer	用户不需要工程师
937	to help them implement the technology.	帮助他们实施技术。
938	For AI, if you're a small-sized,	对于AI,如果你是一个小尺寸,
939	medium-sized firm in Hunan province,	湖南省中型企业,
940	you're gonna need at least one AI engineer	你至少需要一位人工智能工程师
941	in your firm to be able to implement AI.	在你的公司里能够执行AI。
942	And you have to make the profit case.	且须作利物案.
943	There has to be sort of a fast-acting	一定有快动作
944	natural diffusion mechanism.	自然扩散机制。
945	So that's one thing I just,	所以,这是一件事,我只是,
946	I think we need to get to that level of detail	我想我们得说清楚
947	rather than just China scales up everything	不只是中国把一切都放大
948	means it's gonna diffuse faster.	意味着它会扩散更快。
949	Okay, on your two really important questions.	好吧,关于你的两个非常重要的问题。
950	First is what is the role of governance and regulation?	第一,治理和监管的作用是什么?
951	And how do we think about concerns	我们如何看待关注
952	that China's need to control and regulate information flows	中国需要控制和管制信息流动
953	and censor information means	和审查信息意味着
954	that it won't be able to innovate in this space?	无法在这个空间创新?
955	I think you drew a really good example	我想你画了一个很好的例子
956	that we've seen in the internet space	我们在互联网上看到的
957	that China can still be very, very innovative	中国仍然可以非常,非常创新
958	despite strict censorship controls.	尽管严格的审查控制。
959	And we're starting to see that with AI as well.	我们开始看到AI。
960	I was just talking to someone who has talked with Alibaba	我刚刚和阿里巴巴谈过
961	and how they're controlling their large language models.	他们如何控制他们的大语言模型。
962	And essentially, it's just daily calls	基本上,这只是每天的电话
963	with the censorship authorities	与审查当局的合作
964	using the same practices that they do for other censorship	使用与其他审查相同的做法
965	such as on like Taobao forums.	像陶宝论坛一样
966	And so China seems to have developed a way	所以中国似乎发展了一种方式
967	to still innovate in the context of these censorship demands.	仍在这些审查要求的背景下进行创新。
968	Let me give one example that touches on all the points	让我举一个涉及所有问题的例子。
969	you raised on that first thread	你在第一线上长大了
970	because I think it unpacks the complicated picture.	因为我觉得它能解开复杂的画面
971	So let's go back to April, 2023,	那么我们回2023年4月吧
972	China with the Cyberspace Administration, CAC,	中国与网络空间管理局,CAC,
973	releases the draft new generative AI regulations.	发布新基因AI条例草案.
974	And this draft imposes some really, really strict regulations	这份草案规定了一些非常严格的规定
975	and requirements on generative AI service providers	对基因AI服务供应商的要求
976	that have public opinion properties	拥有公共舆论财产的人
977	for social mobilization capacity.	社会动员能力。
978	One of the requirements at the time in that draft in April	4月草案中当时的一项要求
979	was generative AI service providers need to verify	人工智能服务提供商需要验证
980	the objectivity and accuracy of their training data.	培训数据的客观性和准确性。
981	Now, these LLMs are essentially training	现在,这些LLMS基本上是训练
982	on all of the available Chinese language texts	关于所有中文文本
983	on the internet or essentially like the equivalent	在互联网上或基本上类似
984	of Wikipedia.	维基百科中的相关条目。
985	It would be like trying to verify the accuracy	就像试图验证准确性
986	and objectivity of the Wikipedia corpus	以及维基百科的客观性
987	or the Baidu Baika corpus.	或贝都贝卡人
988	Over the course of that summer,	在那个夏天,
989	then we have the revised version that comes out in August	然后我们有了8月推出的修订版
990	and relaxes a lot of those overly stringent restrictions.	并且放松了这些过于严格的限制。
991	And so you start to see the push and pull	所以你开始看到推力和拉力
992	between how much can we control these models	我们能控制多少这些模型
993	to prevent sensitive political outputs	防止敏感政治产出
994	and how much do we need to balance that	以及我们需要多少平衡
995	with ensuring we can still compete in this space.	保证我们仍能在这个空间里竞争
996	As for your second question,	至于你的第二个问题
997	in terms of where do I see this space going	从我所看到的空间的走向来看
998	and what is the potential for getting away	和什么是潜逃的潜力
999	from zero-sum competition?	从零和竞争?
1000	So I think what I would say to that is	所以我想我要说的是
1001	if both countries adopt a GPT-diffusion-focused approach,	如果两国都采用以全球预防扩散为重点的方法,
1002	then it's not necessarily about controlling the other side	那不一定要控制对方
1003	from being able to run the race at all.	从能够运行比赛的完全。
1004	It's about let's see who runs faster	看看谁跑得更快
1005	based off of our institutions.	从我们的机构。
1006	I'm not gonna try to control you	我不会试图控制你
1007	and prevent you from running the race at all	并阻止你运行比赛完全
1008	because doing that would actually hurt my ability to run.	因为这样做会伤害我的跑步能力
1009	So let me give a specific example for that.	因此,让我举一个具体的例子。
1010	In the US right now, there's a debate over	在美国,现在有一个争论
1011	whether the US should try to expand	美国是否应该扩大
1012	the export control regime to AI open source models,	* 出口管制制度对AI的开放源码模型,
1013	open source AI models.	开源AI模型.
1014	So think the meta-Lama series	所以,想想元拉玛系列
1015	or the best Chinese open source large language model	或最优秀的中文开源大语言模型
1016	is Alibaba's Q1 or DeepSeq's version.	是阿里巴巴的Q1或DeepSeq的版本.
1017	And so I think if the US were to do that,	所以我想如果美国这样做,
1018	that's an example where you become more closed	举例来说,你变得更加封闭
1019	and when you hurt your own open source ecosystem,	当你伤害了自己的开源生态系统
1020	that's gonna hinder the US's diffusion capacity in AI.	这会妨碍美国在AI的传播能力.
1021	So efforts to become more closed might actually trade off	所以,努力关闭可能实际上交换
1022	with your own country's ability to run faster.	拥有自己国家更快运行的能力.
1023	And so there, it's not like this magical world	所以,那里,它不是像这个神奇的世界
1024	of win-win where everything's in harmony,	在一切和谐的双赢中
1025	but maybe there is a world where the US and China	但也许有一个世界美国和中国
1026	are just trying to compete on even terms	只是想在公平的条件下竞争
1027	to see who can run faster on GPT diffusion.	看看谁能在GPT扩散上跑得更快。
1028	Let me leave it there.	让我留在那里。
1029	And George, feel free to chime in	George 随便打个电话吧
1030	when we get to the Q&A session as well,	当我们到达QQA会议,
1031	but let's open it up to the folks in the room	但是让我们打开它在房间里的人
1032	and really looking forward to seeing	并真正期待着看到
1033	what you all have to say.	你们都得说些什么
1034	I have to say.	我必须说。

该视频共有字幕 1034 条。解锁更多字幕为会员功能，请移动到价格

下载安装