Kai-Fu Lee: AI Superpowers – China and Silicon Valley
商业与创业政治与社会AI 与机器学习技术与编程音乐与艺术
📋 章节目录
暂无章节信息
🔑 关键词
datajobschinesecompanieschinadonbettergovernmentvaluecompanyproductshardroutinemoneyautonomousmarketapproachinfrastructureimportantamerican
💬 精彩语录
暂无语录
🎙️ 完整对话（1617 条）
Lex Fridman (00:00.000)
The following is a conversation with Kai Fu Lee.
以下是与李开复的对话。
Lex Fridman (00:02.920)
He's the chairman and CEO of Cinovation Ventures
他是创新工场董事长兼首席执行官
Lex Fridman (00:06.560)
that manages a $2 billion dual currency investment fund
管理着价值 20 亿美元的双货币投资基金
Lex Fridman (00:10.600)
with a focus on developing the next generation
专注于培养下一代
Lex Fridman (00:13.160)
of Chinese high tech companies.
中国高科技企业。
Kai-Fu Lee (00:15.440)
He's the former president of Google China
他是谷歌中国前总裁
Lex Fridman (00:17.880)
and the founder of what is now called
以及现在所谓的创始人
Kai-Fu Lee (00:19.800)
Microsoft Research Asia,
微软亚洲研究院,
Lex Fridman (00:21.680)
an institute that trained many of the artificial
一个培训了许多人工的研究所
Kai-Fu Lee (00:24.920)
intelligence leaders in China,
中国的情报领导人，
Lex Fridman (00:26.520)
including CTOs or AI execs at Baidu,
包括百度的首席技术官或人工智能高管，
Kai-Fu Lee (00:30.280)
Tencent, Alibaba, Lenovo, and Huawei.
腾讯、阿里巴巴、联想、华为。
Lex Fridman (00:34.840)
He was named one of the 100 most influential people
他被评为100位最具影响力人物之一
Kai-Fu Lee (00:38.520)
in the world by Time Magazine.
被时代杂志评为世界第一。
Lex Fridman (00:40.480)
He's the author of seven bestselling books in Chinese
他是七本中文畅销书的作者
Lex Fridman (00:43.840)
and most recently, the New York Times bestseller
最近，《纽约时报》畅销书
Lex Fridman (00:46.680)
called AI Superpowers, China, Silicon Valley,
号称AI超级大国，中国，硅谷，
Lex Fridman (00:50.600)
and the New World Order.
和新世界秩序。
Lex Fridman (00:52.600)
He has unparalleled experience
他拥有无与伦比的经验
Kai-Fu Lee (00:55.760)
in working across major tech companies
在各大科技公司工作
Lex Fridman (00:57.640)
and governments and applications of AI,
Lex Fridman (01:00.040)
and so he has a unique perspective on global innovation
Lex Fridman (01:03.440)
and the future of AI that I think is important
Kai-Fu Lee (01:06.200)
to listen to and think about.
Lex Fridman (01:08.880)
This is the Artificial Intelligence Podcast.
Kai-Fu Lee (01:11.880)
If you enjoy it, subscribe on YouTube and iTunes,
Lex Fridman (01:15.160)
support it on Patreon, or simply connect with me
Kai-Fu Lee (01:17.920)
on Twitter at Lex Friedman.
Lex Fridman (01:20.960)
And now, here's my conversation with Kaifu Li.
Kai-Fu Lee (01:26.040)
I immigrated from Russia to US when I was 13.
Lex Fridman (01:29.360)
You immigrated to US at about the same age.
Kai-Fu Lee (01:32.400)
The Russian people, the American people,
Lex Fridman (01:34.720)
the Chinese people each have a certain soul,
Kai-Fu Lee (01:38.240)
a spirit that permeates throughout the generations.
Lex Fridman (01:42.000)
So maybe it's a little bit of a poetic question,
Lex Fridman (01:45.080)
but could you describe your sense
Lex Fridman (01:48.840)
of what defines the Chinese soul?
Kai-Fu Lee (01:52.000)
I think the Chinese soul of people today, right,
Lex Fridman (01:56.080)
we're talking about people who have had centuries of burden
Kai-Fu Lee (02:01.920)
because of the poverty that the country has gone through,
Lex Fridman (02:05.120)
and suddenly shined with hope of prosperity
Kai-Fu Lee (02:10.480)
in the past 40 years as China opened up
Lex Fridman (02:13.360)
and embraced market economy.
Lex Fridman (02:15.400)
And undoubtedly, there are two sets of pressures
Lex Fridman (02:20.160)
on the people, that of the tradition,
Kai-Fu Lee (02:24.120)
that of facing difficult situations,
Lex Fridman (02:28.000)
and that of hope of wanting to be the first
Kai-Fu Lee (02:31.120)
to become successful and wealthy.
Lex Fridman (02:33.840)
So that's a very strong hunger and a strong desire
Lex Fridman (02:38.320)
and strong work ethic that drives China forward.
Lex Fridman (02:41.120)
And is there roots to not just this generation,
Lex Fridman (02:43.920)
but before that's deeper
Lex Fridman (02:47.280)
than just the new economic developments?
Kai-Fu Lee (02:50.080)
Is there something that's unique to China
Lex Fridman (02:52.520)
that you could speak to that's in the people?
Kai-Fu Lee (02:54.960)
Yeah, well, the Chinese tradition
Lex Fridman (02:58.560)
is about excellence, dedication, and results.
Lex Fridman (03:02.640)
And the Chinese exams and study subjects in schools
Lex Fridman (03:07.200)
have traditionally started
Kai-Fu Lee (03:09.120)
from memorizing 10,000 characters,
Lex Fridman (03:11.080)
not an easy task to start with.
Lex Fridman (03:13.560)
And further by memorizing
Lex Fridman (03:15.120)
his historic philosopher's literature poetry.
Lex Fridman (03:18.960)
So it really is probably
Lex Fridman (03:21.080)
the strongest rote learning mechanism created
Kai-Fu Lee (03:25.040)
to make sure people had good memory
Lex Fridman (03:26.920)
and remember things extremely well.
Kai-Fu Lee (03:30.080)
That's, I think at the same time,
Lex Fridman (03:32.680)
suppresses the breakthrough innovation
Lex Fridman (03:37.360)
and also enhances the speed execution get results.
Lex Fridman (03:42.360)
And that I think characterizes
Kai-Fu Lee (03:44.400)
the historic basis of China.
Lex Fridman (03:47.280)
That's interesting,
Kai-Fu Lee (03:48.120)
because there's echoes of that in Russian education
Lex Fridman (03:50.200)
as well as rote memorization.
Lex Fridman (03:51.960)
So you have to memorize a lot of poetry.
Lex Fridman (03:53.680)
I mean, there's just an emphasis on perfection in all forms
Kai-Fu Lee (03:59.120)
that's not conducive to perhaps what you're speaking to,
Lex Fridman (04:02.120)
which is creativity.
Lex Fridman (04:03.560)
But you think that kind of education holds back
Lex Fridman (04:06.880)
the innovative spirit that you might see
Lex Fridman (04:09.400)
in the United States?
Lex Fridman (04:10.840)
Well, it holds back the breakthrough innovative spirits
Kai-Fu Lee (04:14.720)
that we see in the United States,
Lex Fridman (04:16.360)
but it does not hold back the valuable execution oriented,
Kai-Fu Lee (04:21.760)
result oriented value creating engines,
Lex Fridman (04:25.120)
which we see China being very successful.
Lex Fridman (04:27.840)
So is there a difference between a Chinese AI engineer today
Lex Fridman (04:32.760)
and an American AI engineer,
Kai-Fu Lee (04:34.680)
perhaps rooted in the culture that we just talked about
Lex Fridman (04:36.920)
or the education or the very soul of the people or no?
Lex Fridman (04:41.040)
And what would your advice be to each
Lex Fridman (04:43.600)
if there's a difference?
Kai-Fu Lee (04:45.400)
Well, there's a lot that's similar
Lex Fridman (04:47.000)
because AI is about mastering sciences,
Kai-Fu Lee (04:51.120)
about using known technologies and trying new things,
Lex Fridman (04:54.760)
but it's also about picking from many parts
Kai-Fu Lee (04:58.120)
of possible networks to use
Lex Fridman (05:00.160)
and different types of parameters to tune.
Lex Fridman (05:02.800)
And that part is somewhat rote.
Lex Fridman (05:05.200)
And it is also, as anyone who's built AI products
Kai-Fu Lee (05:08.880)
can tell you a lot about cleansing the data
Lex Fridman (05:12.560)
because AI runs better with more data
Lex Fridman (05:15.080)
and data is generally unstructured,
Lex Fridman (05:18.480)
errorful and unclean.
Lex Fridman (05:22.280)
And the effort to clean the data is immense.
Lex Fridman (05:26.160)
So I think the better part of American engineering,
Kai-Fu Lee (05:31.160)
AI engineering process is to try new things,
Lex Fridman (05:35.520)
to do things people haven't done before
Lex Fridman (05:37.960)
and to use technology to solve most if not all problems.
Lex Fridman (05:43.360)
So to make the algorithm work despite not so great data,
Kai-Fu Lee (05:47.120)
find error tolerant ways to deal with the data.
Lex Fridman (05:50.640)
The Chinese way would be to basically enumerate
Kai-Fu Lee (05:55.920)
to the fullest extent all the possible ways
Lex Fridman (05:58.560)
by a lot of machines, try lots of different ways
Kai-Fu Lee (06:01.000)
to get it to work and spend a lot of resources
Lex Fridman (06:04.560)
and money and time cleaning up data.
Kai-Fu Lee (06:07.680)
That means the AI engineer may be writing
Lex Fridman (06:10.960)
data cleansing algorithms, working with thousands of people
Kai-Fu Lee (06:15.560)
who label or correct or do things with the data.
Lex Fridman (06:19.120)
That is the incredible hard work
Kai-Fu Lee (06:21.680)
that might lead to better results.
Lex Fridman (06:23.960)
So the Chinese engineer would rely on
Lex Fridman (06:26.360)
and ask for more and more and more data
Lex Fridman (06:29.040)
and find ways to cleanse them and make them work
Kai-Fu Lee (06:31.440)
in the system and probably less time thinking
Lex Fridman (06:34.480)
about new algorithms that can overcome data
Kai-Fu Lee (06:38.520)
or other issues.
Lex Fridman (06:39.640)
So where's your intuition?
Kai-Fu Lee (06:40.840)
Where do you think the biggest impact
Lex Fridman (06:42.320)
in the next 10 years lies?
Kai-Fu Lee (06:44.160)
Is it in some breakthrough algorithms
Lex Fridman (06:47.360)
or is it in just this at scale rigor,
Kai-Fu Lee (06:53.160)
a rigorous approach to data, cleaning data,
Lex Fridman (06:55.920)
organizing data onto the same algorithms?
Lex Fridman (06:58.440)
What do you think the big impact in the applied world is?
Lex Fridman (07:02.600)
Well, if you're really in the company
Lex Fridman (07:04.560)
and you have to deliver results,
Lex Fridman (07:06.880)
using known techniques and enhancing data
Kai-Fu Lee (07:09.720)
seems like the more expedient approach
Lex Fridman (07:12.240)
that's very low risk and likely to generate
Kai-Fu Lee (07:15.680)
better and better results.
Lex Fridman (07:17.200)
And that's why the Chinese approach has done quite well.
Kai-Fu Lee (07:20.520)
Now, there are a lot of more challenging startups
Lex Fridman (07:24.280)
and problems such as autonomous vehicles,
Kai-Fu Lee (07:28.480)
medical diagnosis that existing algorithms
Lex Fridman (07:32.080)
probably won't solve.
Lex Fridman (07:34.280)
And that would put the Chinese approach more challenged
Lex Fridman (07:38.720)
and give them more breakthrough innovation approach,
Kai-Fu Lee (07:42.400)
more of an edge on those kinds of problems.
Lex Fridman (07:45.440)
So let me talk to that a little more.
Lex Fridman (07:47.080)
So my intuition personally is that data
Lex Fridman (07:51.000)
can take us extremely far.
Lex Fridman (07:53.680)
So you brought up autonomous vehicles and medical diagnosis.
Lex Fridman (07:56.480)
So your intuition is that huge amounts of data
Kai-Fu Lee (08:00.080)
might not be able to completely help us solve that problem.
Lex Fridman (08:04.000)
Right, so breaking that down further in autonomous vehicle,
Kai-Fu Lee (08:08.080)
I think huge amounts of data probably will solve
Lex Fridman (08:11.280)
trucks driving on highways,
Kai-Fu Lee (08:13.000)
which will deliver a significant value
Lex Fridman (08:15.600)
and China will probably lead in that.
Lex Fridman (08:18.120)
And full L5 autonomous is likely to require
Lex Fridman (08:23.960)
new technologies we don't yet know.
Lex Fridman (08:26.320)
And that might require academia
Lex Fridman (08:28.640)
and great industrial research,
Kai-Fu Lee (08:30.360)
both innovating and working together.
Lex Fridman (08:32.480)
And in that case, US has an advantage.
Lex Fridman (08:35.360)
So the interesting question there is,
Lex Fridman (08:37.200)
I don't know if you're familiar
Kai-Fu Lee (08:38.240)
on the autonomous vehicle space
Lex Fridman (08:39.720)
and the developments with Tesla and Elon Musk.
Kai-Fu Lee (08:42.640)
I am.
Lex Fridman (08:43.480)
Where they are in fact full steam ahead
Kai-Fu Lee (08:49.400)
into this mysterious complex world of full autonomy, L5,
Lex Fridman (08:53.480)
L4, L5, and they're trying to solve that purely with data.
Lex Fridman (08:58.800)
So the same kind of thing that you're saying
Lex Fridman (09:00.800)
is just for highway,
Kai-Fu Lee (09:02.080)
which is what a lot of people share your intuition.
Lex Fridman (09:05.320)
They're trying to solve with data.
Lex Fridman (09:07.200)
So just to linger on that moment further,
Lex Fridman (09:09.320)
do you think possible for them to achieve success
Kai-Fu Lee (09:13.560)
with simply just a huge amount of this training
Lex Fridman (09:17.040)
on edge cases and difficult cases in urban environments,
Lex Fridman (09:20.400)
not just highway and so on?
Lex Fridman (09:22.760)
I think it would be very hard.
Kai-Fu Lee (09:24.520)
One could characterize Tesla's approach
Lex Fridman (09:27.040)
as kind of a Chinese strength approach, right?
Kai-Fu Lee (09:29.840)
Gather all the data you can
Lex Fridman (09:31.560)
and hope that will overcome the problems.
Lex Fridman (09:34.000)
But in autonomous driving,
Lex Fridman (09:36.080)
clearly a lot of the decisions aren't merely solved
Kai-Fu Lee (09:40.400)
by aggregating data and having feedback loop.
Lex Fridman (09:43.480)
There are things that are more akin to human thinking.
Lex Fridman (09:48.000)
And how would those be integrated and built?
Lex Fridman (09:51.640)
There has not yet been a lot of success
Kai-Fu Lee (09:53.960)
integrating human intelligence
Lex Fridman (09:56.400)
or call it expert systems if you will,
Kai-Fu Lee (09:58.760)
even though that's a taboo word with the machine learning.
Lex Fridman (10:02.920)
And the integration of the two types of thinking
Kai-Fu Lee (10:05.560)
hasn't yet been demonstrated.
Lex Fridman (10:07.800)
And the question is how much can you push
Lex Fridman (10:09.920)
a purely machine learning approach?
Lex Fridman (10:12.360)
And of course, Tesla also has an additional constraint
Kai-Fu Lee (10:15.480)
that they don't have all the sensors.
Lex Fridman (10:18.480)
I know that they think it's foolish to use LIDARs,
Lex Fridman (10:21.120)
but that's clearly a one less very valuable
Lex Fridman (10:25.080)
and reliable source of input that they're foregoing,
Kai-Fu Lee (10:28.880)
which may also have consequences.
Lex Fridman (10:32.400)
I think the advantage of course is capturing data
Kai-Fu Lee (10:34.960)
that no one has ever seen before.
Lex Fridman (10:37.080)
And in some cases such as computer vision
Lex Fridman (10:40.680)
and speech recognition,
Lex Fridman (10:42.240)
I have seen Chinese companies accumulate data
Kai-Fu Lee (10:44.880)
that's not seen anywhere in the Western world
Lex Fridman (10:47.360)
and they have delivered superior results.
Lex Fridman (10:50.280)
But then speech recognition and object recognition
Lex Fridman (10:53.760)
are relatively suitable problems for deep learning
Lex Fridman (10:57.120)
and don't have the potentially need
Lex Fridman (11:00.800)
for the human intelligence analytical planning elements.
Lex Fridman (11:04.480)
And the same on the speech recognition side,
Lex Fridman (11:06.440)
your intuition that speech recognition
Lex Fridman (11:08.960)
and the machine learning approaches to speech recognition
Lex Fridman (11:11.480)
won't take us to a conversational system
Kai-Fu Lee (11:14.520)
that can pass the Turing test,
Lex Fridman (11:15.960)
which is sort of maybe akin to what driving is.
Lex Fridman (11:20.080)
So it needs to have something more than just simply
Lex Fridman (11:24.480)
simple language understanding, simple language generation.
Kai-Fu Lee (11:27.520)
Roughly right.
Lex Fridman (11:28.720)
I would say that based on purely machine learning approaches,
Kai-Fu Lee (11:33.120)
it's hard to imagine it could lead
Lex Fridman (11:35.960)
to a full conversational experience
Kai-Fu Lee (11:40.520)
across arbitrary domains, which is akin to L5.
Lex Fridman (11:44.600)
I'm a little hesitant to use the word Turing test
Kai-Fu Lee (11:46.920)
because the original definition was probably too easy.
Lex Fridman (11:50.280)
We probably do that, yeah.
Kai-Fu Lee (11:52.320)
The spirit of the Turing test
Lex Fridman (11:54.400)
is what I was referring to. Of course.
Lex Fridman (11:56.520)
So you've had major leadership research positions
Lex Fridman (11:59.440)
at Apple, Microsoft, Google.
Lex Fridman (12:01.640)
So continuing on the discussion of America, Russia,
Lex Fridman (12:05.120)
Chinese, Seoul and culture and so on.
Lex Fridman (12:09.200)
What is the culture of Silicon Valley
Lex Fridman (12:12.440)
in contrast to China and maybe US broadly?
Lex Fridman (12:16.320)
And what is the unique culture
Lex Fridman (12:18.760)
of each of these three major companies in your view?
Kai-Fu Lee (12:22.080)
I think in aggregate, Silicon Valley companies,
Lex Fridman (12:25.120)
and we could probably include Microsoft in that,
Kai-Fu Lee (12:27.200)
even though they're not in the Valley,
Lex Fridman (12:29.120)
is really dream big and have visionary goals
Lex Fridman (12:33.960)
and believe that technology will conquer all.
Lex Fridman (12:37.960)
And also the self confidence and the self entitlement
Kai-Fu Lee (12:42.280)
that whatever they produce,
Lex Fridman (12:43.560)
the whole world should use and must use.
Lex Fridman (12:47.240)
And those are historically important, I think.
Lex Fridman (12:54.080)
Steve Jobs famous quote that he doesn't do focus groups,
Kai-Fu Lee (12:59.080)
he looks in the mirror and asks the person in the mirror,
Lex Fridman (13:02.360)
what do you want?
Lex Fridman (13:03.520)
And that really is an inspirational comment that says,
Lex Fridman (13:07.880)
the great company shouldn't just ask users what they want,
Lex Fridman (13:11.240)
but develop something that users will know they want
Lex Fridman (13:15.160)
when they see it,
Lex Fridman (13:16.200)
but they could never come up with themselves.
Lex Fridman (13:18.920)
I think that is probably the most exhilarating description
Kai-Fu Lee (13:23.840)
of what the essence of Silicon Valley is,
Lex Fridman (13:26.520)
that this brilliant idea could cause you to build something
Kai-Fu Lee (13:31.800)
that couldn't come out of the focus groups or AB tests.
Lex Fridman (13:35.480)
And iPhone would be an example of that.
Kai-Fu Lee (13:38.000)
No one in the age of Blackberry would write down
Lex Fridman (13:40.520)
they want an iPhone or multi touch.
Kai-Fu Lee (13:42.760)
A browser might be another example.
Lex Fridman (13:44.760)
No one would say they want that in the days of FTP,
Lex Fridman (13:47.480)
but once they see it, they want it.
Lex Fridman (13:49.400)
So I think that is what Silicon Valley is best at.
Lex Fridman (13:53.560)
But it also comes with, it came with a lot of success.
Lex Fridman (13:58.920)
These products became global platforms
Lex Fridman (14:01.960)
and there were basically no competitors anywhere.
Lex Fridman (14:05.120)
And that has also led to a belief
Kai-Fu Lee (14:08.440)
that these are the only things that one should do,
Lex Fridman (14:13.240)
that companies should not tread on other companies territory
Lex Fridman (14:17.960)
so that a Groupon and a Yelp and then OpenTable
Lex Fridman (14:24.080)
and the Grubhub would each feel,
Kai-Fu Lee (14:26.280)
okay, I'm not gonna do the other company's business
Lex Fridman (14:28.560)
because that would not be the pride of innovating
Lex Fridman (14:33.280)
what each of these four companies have innovated.
Lex Fridman (14:36.920)
But I think the Chinese approach
Kai-Fu Lee (14:40.320)
is do whatever it takes to win.
Lex Fridman (14:42.720)
And it's a winner take all market.
Lex Fridman (14:45.000)
And in fact, in the internet space,
Lex Fridman (14:47.200)
the market leader will get predominantly all the value
Kai-Fu Lee (14:50.880)
extracted out of the system.
Lex Fridman (14:53.320)
So, and the system isn't just defined
Kai-Fu Lee (14:57.840)
as one narrow category, but gets broader and broader.
Lex Fridman (15:01.360)
So it's amazing ambition for success and domination
Kai-Fu Lee (15:07.960)
of increasingly larger product categories
Lex Fridman (15:11.760)
leading to clear market winner status
Lex Fridman (15:15.080)
and the opportunity to extract tremendous value.
Lex Fridman (15:19.080)
And that develops a practical, result oriented,
Kai-Fu Lee (15:25.840)
ultra ambitious winner take all gladiatorial mentality.
Lex Fridman (15:31.480)
And if what it takes is to build what the competitors built,
Kai-Fu Lee (15:37.400)
essentially a copycat that can be done
Lex Fridman (15:40.000)
without infringing laws.
Kai-Fu Lee (15:41.920)
If what it takes is to satisfy a foreign country's need
Lex Fridman (15:46.280)
by forking the code base and building something
Kai-Fu Lee (15:48.480)
that looks really ugly and different, they'll do it.
Lex Fridman (15:51.400)
So it's contrasted very sharply
Kai-Fu Lee (15:54.360)
with the Silicon Valley approach.
Lex Fridman (15:56.240)
And I think the flexibility and the speed and execution
Kai-Fu Lee (16:00.040)
has helped the Chinese approach.
Lex Fridman (16:01.960)
And I think the Silicon Valley approach
Kai-Fu Lee (16:05.040)
is potentially challenged if every Chinese entrepreneur
Lex Fridman (16:10.040)
is learning from the whole world, US and China,
Lex Fridman (16:13.240)
and the American entrepreneurs only look internally
Lex Fridman (16:16.320)
and write off China as a copycat.
Lex Fridman (16:19.640)
And the second part of your question
Lex Fridman (16:21.360)
about the three companies.
Kai-Fu Lee (16:23.560)
The unique elements of the three companies perhaps.
Lex Fridman (16:26.080)
Yeah, I think Apple represents
Kai-Fu Lee (16:30.440)
while the user please the user
Lex Fridman (16:33.160)
and the essence of design and brand
Lex Fridman (16:38.160)
and it's the one company and perhaps the only tech company
Lex Fridman (16:43.640)
that draws people with a strong, serious desire
Kai-Fu Lee (16:49.440)
for the product and the willingness to pay a premium
Lex Fridman (16:53.080)
because of the halo effect of the brand
Kai-Fu Lee (16:56.440)
which came from the attention to detail
Lex Fridman (16:59.640)
and great respect for user needs.
Kai-Fu Lee (17:01.760)
Microsoft represents a platform approach
Lex Fridman (17:07.680)
that builds giant products that become very strong modes
Kai-Fu Lee (17:12.760)
that others can't do because it's well architected
Lex Fridman (17:17.280)
at the bottom level and the work is efficiently delegated
Kai-Fu Lee (17:23.640)
to individuals and then the whole product is built
Lex Fridman (17:28.640)
by adding small parts that sum together.
Lex Fridman (17:32.000)
So it's probably the most effective high tech assembly line
Lex Fridman (17:36.520)
that builds a very difficult product
Kai-Fu Lee (17:38.760)
that and the whole process of doing that
Lex Fridman (17:43.040)
is kind of a differentiation and something competitors
Kai-Fu Lee (17:49.040)
can't easily repeat.
Lex Fridman (17:50.800)
Are there elements of the Chinese approach
Kai-Fu Lee (17:53.080)
in the way Microsoft went about assembling
Lex Fridman (17:56.640)
those little pieces and dominating,
Kai-Fu Lee (17:59.000)
essentially dominating the market for a long time
Lex Fridman (18:02.320)
or do you see those as distinct?
Kai-Fu Lee (18:04.040)
I think there are elements that are the same.
Lex Fridman (18:06.680)
I think the three American companies that had
Kai-Fu Lee (18:09.520)
or have Chinese characteristics and obviously
Lex Fridman (18:12.960)
as well as American characteristics are Microsoft,
Kai-Fu Lee (18:16.280)
Facebook and Amazon.
Lex Fridman (18:18.720)
Yes, that's right, Amazon.
Kai-Fu Lee (18:20.160)
Because these are companies that will tenaciously
Lex Fridman (18:23.920)
go after adjacent markets,
Kai-Fu Lee (18:27.640)
build up strong product offering and find ways
Lex Fridman (18:34.360)
to extract greater value from a sphere
Kai-Fu Lee (18:37.960)
that's ever increasing and they understand
Lex Fridman (18:41.480)
the value of the platforms.
Lex Fridman (18:43.520)
So that's the similarity and then with Google,
Lex Fridman (18:47.280)
I think it's a genuinely value oriented company
Kai-Fu Lee (18:54.800)
that does have a heart and soul
Lex Fridman (18:57.000)
and that wants to do great things for the world
Kai-Fu Lee (18:59.800)
by connecting information
Lex Fridman (19:01.920)
and that has also very strong technology genes
Lex Fridman (19:08.840)
and wants to use technology
Lex Fridman (19:13.280)
and has found out of the box ways to use technology
Kai-Fu Lee (19:19.120)
to deliver incredible value to the end user.
Lex Fridman (19:23.720)
If you can look at Google, for example,
Kai-Fu Lee (19:25.240)
you mentioned heart and soul.
Lex Fridman (19:28.040)
There seems to be an element where Google
Kai-Fu Lee (19:31.840)
is after making the world better.
Lex Fridman (19:34.840)
There's a more positive view.
Kai-Fu Lee (19:36.920)
They used to have the slogan, don't be evil.
Lex Fridman (19:39.280)
And Facebook a little bit more has a negative tend to it.
Kai-Fu Lee (19:43.120)
At least in the perception of privacy and so on.
Lex Fridman (19:45.760)
Do you have a sense of how these different companies
Kai-Fu Lee (19:51.080)
can achieve, because you've talked about
Lex Fridman (19:53.040)
how much we can make the world better
Kai-Fu Lee (19:54.520)
in all these kinds of ways with AI.
Lex Fridman (19:56.760)
What is it about a company that can make,
Kai-Fu Lee (19:59.360)
give it a heart and soul, gain the trust of the public
Lex Fridman (1:00:00.880)
would be very dangerous
Kai-Fu Lee (1:00:02.480)
because should two countries rely on AI
Lex Fridman (1:00:05.560)
to make certain decisions
Lex Fridman (1:00:07.400)
and they don't even talk to each other,
Lex Fridman (1:00:10.040)
they do their own scenario planning,
Kai-Fu Lee (1:00:12.120)
then something could easily go wrong.
Lex Fridman (1:00:15.120)
I think engagement, interaction, some protocols
Kai-Fu Lee (1:00:18.960)
to avoid inadvertent disasters is actually needed.
Lex Fridman (1:00:25.120)
So it's natural for each country to want to be the best,
Kai-Fu Lee (1:00:29.120)
whether it's in nuclear technologies or AI or bio.
Lex Fridman (1:00:34.800)
But I think it's important to realize
Kai-Fu Lee (1:00:37.720)
if each country has a black box AI
Lex Fridman (1:00:41.360)
and don't talk to each other,
Kai-Fu Lee (1:00:43.280)
that probably presents greater challenges to humanity
Lex Fridman (1:00:49.280)
than if they interacted.
Kai-Fu Lee (1:00:51.480)
I think there can still be competition,
Lex Fridman (1:00:53.760)
but with some degree of protocol for interaction,
Kai-Fu Lee (1:00:57.080)
just like when there was a nuclear competition,
Lex Fridman (1:01:02.160)
there were some protocol for deterrence
Kai-Fu Lee (1:01:04.880)
among US, Russia, and China.
Lex Fridman (1:01:08.000)
And I think that engagement is needed.
Lex Fridman (1:01:10.880)
So of course, we're still far from AI
Lex Fridman (1:01:13.520)
presenting that kind of danger.
Lex Fridman (1:01:16.000)
But what I worry the most about
Lex Fridman (1:01:18.400)
is the level of engagement seems to be coming down.
Kai-Fu Lee (1:01:23.000)
The level of distrust seems to be going up,
Lex Fridman (1:01:26.360)
especially from the US towards other large countries
Kai-Fu Lee (1:01:30.240)
such as China and of course, and Russia, yes.
Lex Fridman (1:01:33.400)
Is there a way to make that better?
Lex Fridman (1:01:34.680)
So let's beautifully put level of engagement
Lex Fridman (1:01:37.240)
and even just basic trust and communication
Kai-Fu Lee (1:01:40.680)
as opposed to sort of making artificial enemies
Lex Fridman (1:01:48.880)
out of particular countries.
Lex Fridman (1:01:53.360)
Do you have a sense how we can make it better?
Lex Fridman (1:01:57.160)
Actionable items that as a society we can take on?
Kai-Fu Lee (1:02:01.720)
I'm not an expert at geopolitics,
Lex Fridman (1:02:05.000)
but I would say that we look pretty foolish as humankind
Kai-Fu Lee (1:02:10.800)
when we are faced with the opportunity
Lex Fridman (1:02:13.200)
to create $16 trillion for humanity,
Lex Fridman (1:02:19.520)
and yet we're not solving fundamental problems
Lex Fridman (1:02:26.040)
with parts of the world still in poverty.
Lex Fridman (1:02:29.520)
And for the first time,
Lex Fridman (1:02:31.320)
we have the resources to overcome poverty and hunger.
Kai-Fu Lee (1:02:34.600)
We're not using it on that,
Lex Fridman (1:02:35.960)
but we're fueling competition among superpowers.
Lex Fridman (1:02:38.800)
And that's a very unfortunate thing.
Lex Fridman (1:02:41.880)
If we become utopian for a moment,
Kai-Fu Lee (1:02:44.760)
imagine a benevolent world government
Lex Fridman (1:02:52.240)
that has this $16 trillion and maybe some AI
Kai-Fu Lee (1:02:56.120)
to figure out how to use it to deal with diseases
Lex Fridman (1:02:59.280)
and problems and hate and things like that.
Kai-Fu Lee (1:03:02.640)
World would be a lot better off.
Lex Fridman (1:03:04.840)
So what is wrong with the current world?
Kai-Fu Lee (1:03:07.600)
I think the people with more skill than I
Lex Fridman (1:03:11.040)
should think about this.
Lex Fridman (1:03:13.920)
And then the geopolitics issue with superpower competition
Lex Fridman (1:03:16.920)
is one side of the issue.
Kai-Fu Lee (1:03:19.360)
There's another side which I worry maybe even more,
Lex Fridman (1:03:24.040)
which is as the $16 trillion all gets made by US and China
Lex Fridman (1:03:29.360)
and a few of the other developed countries,
Lex Fridman (1:03:32.040)
the poorer country will get nothing
Kai-Fu Lee (1:03:34.280)
because they don't have technology
Lex Fridman (1:03:36.880)
and the wealth disparity and inequality will increase.
Lex Fridman (1:03:42.440)
So a poorer country with a large population
Lex Fridman (1:03:45.880)
will not only benefit from the AI boom
Kai-Fu Lee (1:03:48.520)
or other technology booms,
Lex Fridman (1:03:50.360)
but they will have their workers
Kai-Fu Lee (1:03:52.360)
who previously had hoped they could do the China model
Lex Fridman (1:03:56.040)
and do outsource manufacturing or the India model
Lex Fridman (1:03:58.840)
so they could do the outsource process or call center.
Lex Fridman (1:04:02.640)
Well, all those jobs are gonna be gone in 10 or 15 years.
Lex Fridman (1:04:05.800)
So the individual citizen may be a net liability,
Lex Fridman (1:04:12.080)
I mean, financially speaking to a poorer country
Lex Fridman (1:04:15.040)
and not an asset to claw itself out of poverty.
Lex Fridman (1:04:19.800)
So in that kind of situation,
Kai-Fu Lee (1:04:22.640)
these large countries with not much tech
Lex Fridman (1:04:26.400)
are going to be facing a downward spiral
Lex Fridman (1:04:30.080)
and it's unclear what could be done.
Lex Fridman (1:04:33.120)
And then when we look back
Lex Fridman (1:04:34.720)
and say there's $16 trillion being created
Lex Fridman (1:04:37.600)
and it's all being kept by US, China
Lex Fridman (1:04:39.720)
and other developed countries, it just doesn't feel right.
Lex Fridman (1:04:43.360)
So I hope people who know about geopolitics
Kai-Fu Lee (1:04:46.920)
can find solutions that's beyond my expertise.
Lex Fridman (1:04:50.880)
So different countries that we've talked about
Kai-Fu Lee (1:04:53.160)
have different value systems.
Lex Fridman (1:04:55.200)
If you look at the United States,
Kai-Fu Lee (1:04:56.920)
to an almost extreme degree,
Lex Fridman (1:04:58.960)
there is an absolute desire for freedom of speech.
Kai-Fu Lee (1:05:03.400)
If you look at a country where I was raised,
Lex Fridman (1:05:05.160)
that desire just amongst the people
Kai-Fu Lee (1:05:06.960)
is not as elevated as it is to basically fundamental level
Lex Fridman (1:05:15.040)
to the essence of what it means to be America, right?
Lex Fridman (1:05:17.560)
And the same is true with China,
Lex Fridman (1:05:19.160)
there's different value systems.
Kai-Fu Lee (1:05:21.480)
There's some censorship of internet content
Lex Fridman (1:05:26.840)
that China and Russia and many other countries undertake.
Lex Fridman (1:05:31.280)
Do you see that having effects on innovation,
Lex Fridman (1:05:36.760)
other aspects of some of the tech stuff,
Kai-Fu Lee (1:05:38.880)
AI development we talked about,
Lex Fridman (1:05:40.920)
and maybe from another angle,
Lex Fridman (1:05:42.520)
do you see that changing in different ways
Lex Fridman (1:05:46.200)
over the next 10 years, 20 years, 50 years
Kai-Fu Lee (1:05:49.560)
as China continues to grow as it does now
Lex Fridman (1:05:53.000)
in its tech innovation?
Kai-Fu Lee (1:05:55.720)
There's a common belief
Lex Fridman (1:05:57.200)
that full freedom of speech and expression
Kai-Fu Lee (1:06:01.040)
is correlated with creativity,
Lex Fridman (1:06:03.040)
which is correlated with entrepreneurial success.
Kai-Fu Lee (1:06:08.520)
I think empirically we have seen that is not true
Lex Fridman (1:06:13.200)
and China has been successful.
Kai-Fu Lee (1:06:15.600)
That's not to say the fundamental values are not right
Lex Fridman (1:06:19.600)
or not the best,
Lex Fridman (1:06:20.840)
but it's just that perfect correlation isn't there.
Lex Fridman (1:06:25.880)
It's hard to read the tea leaves on opening up or not
Kai-Fu Lee (1:06:30.560)
in any country,
Lex Fridman (1:06:31.720)
and I've not been very good at that in my past predictions,
Lex Fridman (1:06:37.000)
but I do believe every country
Lex Fridman (1:06:40.840)
shares a lot of fundamental values for the longterm.
Lex Fridman (1:06:47.440)
So China is drafting its privacy policy
Lex Fridman (1:06:54.480)
for individual citizens,
Lex Fridman (1:06:57.240)
and they don't look that different
Lex Fridman (1:07:00.000)
from the American or European ones.
Lex Fridman (1:07:03.240)
So people do want to protect their privacy
Lex Fridman (1:07:07.600)
and have the opportunity to express
Lex Fridman (1:07:10.680)
and I think the fundamental values are there.
Lex Fridman (1:07:14.200)
The question is in the execution and timing,
Lex Fridman (1:07:17.920)
how soon or when will that start to open up?
Lex Fridman (1:07:21.760)
So as long as each government knows
Kai-Fu Lee (1:07:25.520)
ultimately people want that kind of protection,
Lex Fridman (1:07:29.200)
there should be a plan to move towards that
Kai-Fu Lee (1:07:32.240)
as to when or how and I'm not an expert.
Lex Fridman (1:07:36.280)
On the point of privacy to me, it's really interesting.
Lex Fridman (1:07:39.040)
So AI needs data to create
Lex Fridman (1:07:42.520)
a personalized awesome experience, right?
Kai-Fu Lee (1:07:45.520)
I'm just speaking generally in terms of products.
Lex Fridman (1:07:48.560)
And then we have currently, depending on the age
Lex Fridman (1:07:51.360)
and depending on the demographics of who we're talking about,
Lex Fridman (1:07:54.000)
some people are more or less concerned
Kai-Fu Lee (1:07:55.840)
about the amount of data they hand over.
Lex Fridman (1:07:59.040)
So in your view, how do we get this balance right
Kai-Fu Lee (1:08:04.280)
that we provide an amazing experience
Lex Fridman (1:08:07.120)
to people that use products?
Kai-Fu Lee (1:08:09.920)
You look at Facebook, the more Facebook knows about you,
Lex Fridman (1:08:13.440)
yes, it's scary to say, the better it can probably,
Kai-Fu Lee (1:08:19.360)
better experience it can probably create.
Lex Fridman (1:08:21.200)
So in your view, how do we get that balance right?
Kai-Fu Lee (1:08:25.160)
Yes, I think a lot of people have a misunderstanding
Lex Fridman (1:08:30.400)
that it's okay and possible to just rip all the data out
Kai-Fu Lee (1:08:35.080)
from a provider and give it back to you.
Lex Fridman (1:08:38.240)
So you can deny them access to further data
Lex Fridman (1:08:41.160)
and still enjoy the services we have.
Lex Fridman (1:08:44.080)
If we take back all the data,
Kai-Fu Lee (1:08:46.240)
all the services will give us nonsense.
Lex Fridman (1:08:48.360)
We'll no longer be able to use products that function well
Kai-Fu Lee (1:08:52.360)
in terms of right ranking, right products,
Lex Fridman (1:08:56.160)
right user experience.
Lex Fridman (1:08:57.880)
So yet I do understand we don't want to permit misuse
Lex Fridman (1:09:02.880)
of the data from legal policy standpoint.
Kai-Fu Lee (1:09:07.520)
I think there can be severe punishment
Lex Fridman (1:09:11.160)
for those who have egregious misuse of the data.
Kai-Fu Lee (1:09:16.200)
That's I think a good first step.
Lex Fridman (1:09:19.600)
Actually China in this side on this aspect
Kai-Fu Lee (1:09:22.680)
has very strong laws about people who sell
Lex Fridman (1:09:25.480)
or give data to other companies.
Lex Fridman (1:09:27.960)
And that over the past few years,
Lex Fridman (1:09:30.080)
since that law came into effect,
Kai-Fu Lee (1:09:33.560)
pretty much eradicated the illegal distribution,
Lex Fridman (1:09:38.760)
sharing of data.
Kai-Fu Lee (1:09:40.560)
Additionally, I think giving,
Lex Fridman (1:09:45.440)
I think technology is often a very good way
Kai-Fu Lee (1:09:50.120)
to solve technology misuse.
Lex Fridman (1:09:52.760)
So can we come up with new technologies
Lex Fridman (1:09:56.200)
that will let us have our cake and eat it too?
Lex Fridman (1:09:58.880)
People are looking into homomorphic encryption,
Kai-Fu Lee (1:10:01.960)
which is letting you keep the data,
Lex Fridman (1:10:04.360)
have it encrypted and train on encrypted data.
Kai-Fu Lee (1:10:07.360)
Of course, we haven't solved that one yet,
Lex Fridman (1:10:09.080)
but that kind of direction may be worth pursuing.
Kai-Fu Lee (1:10:13.400)
Also federated learning,
Lex Fridman (1:10:15.160)
which would allow one hospital
Kai-Fu Lee (1:10:17.240)
to train on its hospital's patient data fully
Lex Fridman (1:10:20.000)
because they have a license for that.
Lex Fridman (1:10:22.400)
And then hospitals would then share their models,
Lex Fridman (1:10:24.800)
not data, but models to create a super AI.
Lex Fridman (1:10:28.040)
And that also maybe has some promise.
Lex Fridman (1:10:30.560)
So I would want to encourage us to be open minded
Lex Fridman (1:10:34.160)
and think of this as not just the policy binary, yes, no,
Lex Fridman (1:10:39.640)
but letting the technologists try to find solutions
Kai-Fu Lee (1:10:42.800)
to let us have our cake and eat it too,
Lex Fridman (1:10:44.840)
or have most of our cake and eat most of it too.
Kai-Fu Lee (1:10:48.400)
Finally, I think giving each end user a choice is important
Lex Fridman (1:10:52.920)
and having transparency is important.
Kai-Fu Lee (1:10:55.400)
Also, I think that's universal,
Lex Fridman (1:10:57.400)
but the choice you give to the user
Kai-Fu Lee (1:11:00.560)
should not be at a granular level
Lex Fridman (1:11:02.320)
that the user cannot understand.
Kai-Fu Lee (1:11:04.720)
GDPR today causes all these popups of yes, no,
Lex Fridman (1:11:09.120)
will you give this site this right
Lex Fridman (1:11:10.560)
to use this part of your data?
Lex Fridman (1:11:12.360)
I don't think any user understands
Lex Fridman (1:11:15.000)
what they're saying yes or no to.
Lex Fridman (1:11:17.000)
And I suspect most are just saying yes
Kai-Fu Lee (1:11:18.960)
because they don't understand it.
Lex Fridman (1:11:20.760)
So while GDPR in its current implementation
Kai-Fu Lee (1:11:25.600)
has lived up to its promise of transparency and user choice,
Lex Fridman (1:11:30.360)
it implemented it in such a way
Kai-Fu Lee (1:11:33.000)
that really didn't deliver the spirit of GDPR.
Lex Fridman (1:11:39.080)
It fit the letter, but not the spirit.
Lex Fridman (1:11:41.560)
So again, I think we need to think about
Lex Fridman (1:11:43.760)
is there a way to fit the spirit of GDPR
Lex Fridman (1:11:48.000)
by using some kind of technology?
Lex Fridman (1:11:50.600)
Can we have a slider that's an AI trying to figure out
Lex Fridman (1:11:54.640)
how much you want to slide between
Lex Fridman (1:11:57.520)
perfect protection security of your personal data
Kai-Fu Lee (1:12:01.520)
versus a high degree of convenience
Lex Fridman (1:12:04.080)
with some risks of not having full privacy?
Kai-Fu Lee (1:12:08.120)
Each user should have some preference
Lex Fridman (1:12:10.080)
and that gives you the user choice.
Lex Fridman (1:12:12.000)
But maybe we should turn the problem on its head
Lex Fridman (1:12:14.840)
and ask can there be an AI algorithm that can customize this?
Kai-Fu Lee (1:12:18.840)
Because we can understand the slider,
Lex Fridman (1:12:21.120)
but we sure cannot understand every popup question.
Lex Fridman (1:12:25.080)
And I think getting that right
Lex Fridman (1:12:27.200)
requires getting the balance between
Lex Fridman (1:12:29.720)
what we talked about earlier,
Lex Fridman (1:12:30.760)
which is heart and soul
Kai-Fu Lee (1:12:32.720)
versus profit driven decisions and strategy.
Lex Fridman (1:12:37.120)
I think from my perspective,
Kai-Fu Lee (1:12:40.200)
the best way to make a lot of money in the long term
Lex Fridman (1:12:43.080)
is to keep your heart and soul intact.
Kai-Fu Lee (1:12:46.760)
I think getting that slider right in the short term
Lex Fridman (1:12:50.640)
may feel like you'll be sacrificing profit,
Lex Fridman (1:12:54.080)
but in the long term,
Lex Fridman (1:12:55.760)
you'll be gaining user trust
Lex Fridman (1:12:57.600)
and providing a great experience.
Lex Fridman (1:12:59.520)
Do you share that kind of view in general?
Kai-Fu Lee (1:13:02.000)
Yes, absolutely.
Lex Fridman (1:13:03.240)
I sure would hope there is a way
Kai-Fu Lee (1:13:07.200)
we can do long term projects
Lex Fridman (1:13:09.600)
that really do the right thing.
Kai-Fu Lee (1:13:12.000)
I think a lot of people who embrace GDPR,
Lex Fridman (1:13:15.240)
their heart's in the right place.
Kai-Fu Lee (1:13:16.960)
I think they just need to figure out how to build a solution.
Lex Fridman (1:13:20.680)
I've heard utopians talk about solutions
Kai-Fu Lee (1:13:23.360)
that get me excited,
Lex Fridman (1:13:24.480)
but I'm not sure how in the current funding environment
Kai-Fu Lee (1:13:27.880)
they can get started.
Lex Fridman (1:13:29.320)
People talk about,
Kai-Fu Lee (1:13:30.600)
imagine this crowdsourced data collection
Lex Fridman (1:13:36.440)
that we all trust.
Lex Fridman (1:13:37.880)
And then we have these agents
Lex Fridman (1:13:40.720)
that we ask the trusted agent to...
Kai-Fu Lee (1:13:45.720)
That agent only, that platform,
Lex Fridman (1:13:48.880)
so a trusted joint platform
Kai-Fu Lee (1:13:51.280)
that we all believe is trustworthy,
Lex Fridman (1:13:55.120)
that can give us all the closed loop personal suggestions
Kai-Fu Lee (1:14:03.080)
by the new social network, new search engine,
Lex Fridman (1:14:06.200)
new eCommerce engine that has access
Kai-Fu Lee (1:14:08.600)
to even more of our data,
Lex Fridman (1:14:10.360)
but not directly, but indirectly.
Lex Fridman (1:14:12.400)
So I think that general concept
Lex Fridman (1:14:14.640)
of licensing to some trusted engine
Lex Fridman (1:14:18.520)
and finding a way to trust that engine
Lex Fridman (1:14:20.640)
seems like a great idea.
Lex Fridman (1:14:22.360)
But if you think how long it's gonna take
Lex Fridman (1:14:24.240)
to implement and tweak and develop it right,
Kai-Fu Lee (1:14:27.760)
as well as to collect all the trusts
Lex Fridman (1:14:29.920)
and the data from the people,
Kai-Fu Lee (1:14:31.360)
it's beyond the current cycle of venture capital.
Lex Fridman (1:14:34.960)
So how do you do that is a big question.
Kai-Fu Lee (1:14:38.120)
You've recently had a fight with cancer,
Lex Fridman (1:14:41.840)
stage four lymphoma and in a sort of deep personal level,
Lex Fridman (1:14:48.240)
what did it feel like in the darker moments
Lex Fridman (1:14:51.080)
to face your own mortality?
Kai-Fu Lee (1:14:54.840)
Well, I've been the workaholic my whole life
Lex Fridman (1:14:57.880)
and I've basically worked nine, nine, six,
Kai-Fu Lee (1:15:01.400)
nine a.m. to nine p.m. six days a week, roughly.
Lex Fridman (1:15:04.640)
And I didn't really pay a lot of attention
Kai-Fu Lee (1:15:07.920)
to my family, friends, and people who loved me.
Lex Fridman (1:15:10.680)
And my life revolved around optimizing for work.
Kai-Fu Lee (1:15:14.480)
While my work was not routine,
Lex Fridman (1:15:16.920)
my optimization really what made my life
Kai-Fu Lee (1:15:23.240)
basically very mechanical process.
Lex Fridman (1:15:25.640)
But I got a lot of highs out of it
Kai-Fu Lee (1:15:28.360)
because of accomplishments
Lex Fridman (1:15:30.960)
that I thought were really important and dear
Lex Fridman (1:15:34.400)
and the highest priority to me.
Lex Fridman (1:15:36.720)
But when I faced mortality
Lex Fridman (1:15:38.560)
and the possible death in matter of months,
Lex Fridman (1:15:41.960)
I suddenly realized that this really meant nothing to me,
Kai-Fu Lee (1:15:45.600)
that I didn't feel like working for another minute,
Lex Fridman (1:15:48.560)
that if I had six months left in my life,
Kai-Fu Lee (1:15:51.000)
I would spend it all with my loved ones
Lex Fridman (1:15:54.280)
and thanking them, giving them love back
Lex Fridman (1:15:57.480)
and apologizing to them that I lived my life the wrong way.
Lex Fridman (1:16:01.960)
So that moment of reckoning
Kai-Fu Lee (1:16:05.960)
caused me to really rethink that why we exist in this world
Lex Fridman (1:16:11.520)
is something that we might be too much shaped by the society
Kai-Fu Lee (1:16:17.960)
to think that success and accomplishments is why we live.
Lex Fridman (1:16:22.040)
But while that can get you
Kai-Fu Lee (1:16:26.320)
periodic successes and satisfaction,
Lex Fridman (1:16:29.680)
it's really in the facing death
Kai-Fu Lee (1:16:33.200)
you see what's truly important to you.
Lex Fridman (1:16:35.840)
So as a result of going through the challenges with cancer,
Kai-Fu Lee (1:16:41.960)
I've resolved to live a more balanced lifestyle.
Lex Fridman (1:16:45.720)
I'm now in remission, knock on wood,
Lex Fridman (1:16:48.920)
and I'm spending more time with my family.
Lex Fridman (1:16:52.440)
My wife travels with me.
Kai-Fu Lee (1:16:54.800)
When my kids need me, I spend more time with them.
Lex Fridman (1:16:58.000)
And before I used to prioritize everything around work.
Kai-Fu Lee (1:17:02.680)
When I had a little bit of time,
Lex Fridman (1:17:03.960)
I would dole it out to my family.
Kai-Fu Lee (1:17:06.240)
Now, when my family needs something, really needs something,
Lex Fridman (1:17:09.880)
I drop everything at work and go to them.
Lex Fridman (1:17:12.640)
And then in the time remaining, I allocate to work.
Lex Fridman (1:17:15.880)
But one's family is very understanding.
Kai-Fu Lee (1:17:19.000)
It's not like they will take 50 hours a week from me.
Lex Fridman (1:17:23.160)
So I'm actually able to still work pretty hard,
Kai-Fu Lee (1:17:27.120)
maybe 10 hours less per week.
Lex Fridman (1:17:29.280)
So I realized the most important thing in my life
Kai-Fu Lee (1:17:32.720)
is really love and the people I love.
Lex Fridman (1:17:36.280)
And I give that the highest priority.
Kai-Fu Lee (1:17:38.680)
It isn't the only thing I do,
Lex Fridman (1:17:40.680)
but when that is needed, I put that at the top priority
Lex Fridman (1:17:45.880)
and I feel much better and I feel much more balanced.
Lex Fridman (1:17:50.080)
And I think this also gives a hint
Kai-Fu Lee (1:17:53.080)
as to a life of routine work, a life of pursuit of numbers.
Lex Fridman (1:17:58.080)
While my job was not routine, it was in pursuit of numbers,
Lex Fridman (1:18:02.920)
pursuit of can I make more money?
Lex Fridman (1:18:04.960)
Can I fund more great companies?
Lex Fridman (1:18:07.360)
Can I raise more money?
Lex Fridman (1:18:08.920)
Can I make sure our VC is ranked higher and higher
Lex Fridman (1:18:12.080)
every year?
Lex Fridman (1:18:13.360)
This competitive nature of driving for bigger numbers
Lex Fridman (1:18:18.800)
and better numbers became a endless pursuit
Lex Fridman (1:18:25.200)
that's mechanical.
Lex Fridman (1:18:26.640)
And bigger numbers really didn't make me happier.
Lex Fridman (1:18:31.880)
And faced with death, I realized bigger numbers
Kai-Fu Lee (1:18:34.680)
really meant nothing.
Lex Fridman (1:18:36.360)
And what was important is that people who have given
Kai-Fu Lee (1:18:41.040)
their heart and their love to me
Lex Fridman (1:18:42.840)
deserve for me to do the same.
Lex Fridman (1:18:45.640)
So there's deep, profound truth in that,
Lex Fridman (1:18:48.400)
that everyone should hear and internalize.
Kai-Fu Lee (1:18:52.000)
I mean, that's really powerful for you to say that.
Lex Fridman (1:18:55.420)
I have to ask sort of a difficult question here.
Lex Fridman (1:19:03.060)
So I've competed in sports my whole life,
Lex Fridman (1:19:06.020)
looking historically, I'd like to challenge some aspect
Kai-Fu Lee (1:19:11.780)
of that a little bit on the point of hard work.
Lex Fridman (1:19:15.220)
That it feels that there are certain aspects
Kai-Fu Lee (1:19:18.900)
that is the greatest, the most beautiful aspects
Lex Fridman (1:19:22.020)
of human nature is the ability to become obsessed,
Kai-Fu Lee (1:19:27.740)
of becoming extremely passionate to the point where yes,
Lex Fridman (1:19:31.460)
flaws are revealed and just giving yourself fully to a task.
Kai-Fu Lee (1:19:37.460)
That is, in another sense, you mentioned love
Lex Fridman (1:19:40.800)
being important, but in another sense,
Kai-Fu Lee (1:19:42.380)
this kind of obsession, this pure exhibition of passion
Lex Fridman (1:19:46.660)
and hard work is truly what it means to be human.
Lex Fridman (1:19:50.380)
What lessons should we take that's deeper?
Lex Fridman (1:19:53.140)
Because you've accomplished incredible things.
Kai-Fu Lee (1:19:54.820)
You say it chasing numbers,
Lex Fridman (1:19:57.300)
but really there's some incredible work there.
Lex Fridman (1:20:00.980)
So how do you think about that when you look back
Lex Fridman (1:20:04.980)
in your 20s, your 30s, what would you do differently?
Lex Fridman (1:20:10.020)
Would you really take back some of the incredible hard work?
Lex Fridman (1:20:14.860)
I would, but it's in percentages, right?
Kai-Fu Lee (1:20:19.980)
We're both computer scientists.
Lex Fridman (1:20:22.380)
So I think when one balances one's life,
Kai-Fu Lee (1:20:25.620)
when one is younger, you might give a smaller percentage
Lex Fridman (1:20:30.260)
to family, but you would still give them high priority.
Lex Fridman (1:20:33.720)
And when you get older, you would give a larger percentage
Lex Fridman (1:20:36.340)
to them and still the high priority.
Lex Fridman (1:20:38.500)
And when you're near retirement, you give most of it to them
Lex Fridman (1:20:42.460)
and the highest priority.
Lex Fridman (1:20:43.820)
So I think the key point is not that we would work 20 hours
Lex Fridman (1:20:49.140)
less for the whole life and just spend it aimlessly
Kai-Fu Lee (1:20:52.700)
with the family, but that's when the family has a need,
Lex Fridman (1:20:56.740)
when your wife is having a baby,
Kai-Fu Lee (1:21:00.260)
when your daughter has a birthday or when they're depressed
Lex Fridman (1:21:05.180)
or when they're celebrating something
Kai-Fu Lee (1:21:07.540)
or when they have a get together or when we have family time
Lex Fridman (1:21:11.400)
that it's important for us to put down our phone and PC
Lex Fridman (1:21:14.860)
and be a hundred percent with them.
Lex Fridman (1:21:18.340)
And that priority on the things that really matter
Kai-Fu Lee (1:21:23.740)
isn't going to be so taxing that it would eliminate
Lex Fridman (1:21:29.140)
or even dramatically reduce our accomplishments.
Kai-Fu Lee (1:21:32.020)
It might have some impact, but it might also have
Lex Fridman (1:21:35.180)
other impact because if you have a happier family,
Kai-Fu Lee (1:21:37.940)
maybe you fight less.
Lex Fridman (1:21:39.380)
If you fight less, you don't spend time taking care
Kai-Fu Lee (1:21:43.100)
of all the aftermath of a fight.
Lex Fridman (1:21:45.620)
So it's unclear that it would take more time.
Lex Fridman (1:21:48.220)
And if it did, I'd be willing to take that reduction.
Lex Fridman (1:21:53.260)
And it's not a dramatic number, but it's a number
Kai-Fu Lee (1:21:56.340)
that I think would give me a greater degree of happiness
Lex Fridman (1:22:00.140)
and knowing that I've done the right thing
Lex Fridman (1:22:03.260)
and still have plenty of hours to get the success
Lex Fridman (1:22:08.260)
that I want to get.
Lex Fridman (1:22:09.900)
So given the many successful companies that you've launched
Lex Fridman (1:22:14.460)
and much success throughout your career,
Lex Fridman (1:22:17.340)
what advice would you give to young people today looking,
Lex Fridman (1:22:25.440)
or it doesn't have to be young,
Lex Fridman (1:22:26.780)
but people today looking to launch
Lex Fridman (1:22:28.500)
and to create the next $1 billion tech startup
Lex Fridman (1:22:32.360)
or even AI based startup?
Lex Fridman (1:22:34.280)
I would suggest that people understand
Kai-Fu Lee (1:22:39.800)
technology waves move quickly.
Lex Fridman (1:22:42.720)
What worked two years ago may not work today.
Lex Fridman (1:22:45.920)
And that is very much case in point for AI.
Lex Fridman (1:22:49.700)
I think two years ago, or maybe three years ago,
Kai-Fu Lee (1:22:53.200)
you certainly could say I have a couple
Lex Fridman (1:22:55.320)
of super smart PhDs and we're not sure
Lex Fridman (1:22:58.880)
what we're gonna do, but here's how we're gonna start
Lex Fridman (1:23:01.920)
and get funding for a very high valuation.
Kai-Fu Lee (1:23:05.200)
Those days are over because AI is going
Lex Fridman (1:23:08.480)
from rocket science towards mainstream,
Kai-Fu Lee (1:23:11.520)
not yet commodity, but more mainstream.
Lex Fridman (1:23:14.400)
So first the creation of any company
Kai-Fu Lee (1:23:19.400)
to a venture capitalists has to be creation
Lex Fridman (1:23:22.720)
of business value and monetary value.
Lex Fridman (1:23:26.200)
And when you have a very scarce commodity,
Lex Fridman (1:23:29.800)
VCs may be willing to accept greater uncertainty.
Lex Fridman (1:23:35.040)
But now the number of people who have the equivalent
Lex Fridman (1:23:38.560)
of PhD three years ago, because that can be learned
Kai-Fu Lee (1:23:42.840)
more quickly, platforms are emerging,
Lex Fridman (1:23:46.080)
the cost to become a AI engineer is much lower
Lex Fridman (1:23:49.580)
and there are many more AI engineers.
Lex Fridman (1:23:51.700)
So the market is different.
Lex Fridman (1:23:53.960)
So I would suggest someone who wants to build an AI company
Lex Fridman (1:23:57.440)
be thinking about the normal business questions.
Lex Fridman (1:24:01.480)
What customer cases are you trying to address?
Lex Fridman (1:24:06.120)
What kind of pain are you trying to address?
Lex Fridman (1:24:08.600)
How does that translate to value?
Lex Fridman (1:24:10.680)
How will you extract value and get paid
Kai-Fu Lee (1:24:14.680)
through what channel and how much business value
Lex Fridman (1:24:18.120)
will get created?
Kai-Fu Lee (1:24:19.920)
That today needs to be thought about much earlier upfront
Lex Fridman (1:24:24.520)
than it did three years ago.
Kai-Fu Lee (1:24:26.720)
The scarcity question of AI talent has changed.
Lex Fridman (1:24:30.500)
The number of AI talent has changed.
Lex Fridman (1:24:32.800)
So now you need not just AI, but also understanding
Lex Fridman (1:24:37.640)
of business customer and the marketplace.
Lex Fridman (1:24:41.840)
So I also think you should have a more reasonable
Lex Fridman (1:24:48.000)
valuation expectation and growth expectation.
Kai-Fu Lee (1:24:52.360)
There's gonna be more competition.
Lex Fridman (1:24:54.080)
But the good news though, is that AI technologies
Kai-Fu Lee (1:24:57.840)
are now more available in open source.
Lex Fridman (1:25:00.740)
TensorFlow, PyTorch and such tools are much easier to use.
Lex Fridman (1:25:06.640)
So you should be able to experiment and get results
Lex Fridman (1:25:11.760)
iteratively faster than before.
Lex Fridman (1:25:14.240)
So take more of a business mindset to this,
Lex Fridman (1:25:18.540)
think less of this as a laboratory taken into a company,
Kai-Fu Lee (1:25:23.540)
because we've gone beyond that stage.
Lex Fridman (1:25:26.140)
The only exception is if you truly have a breakthrough
Kai-Fu Lee (1:25:29.740)
in some technology that really no one has,
Lex Fridman (1:25:32.340)
then the old way still works.
Lex Fridman (1:25:34.660)
But I think that's harder and harder now.
Lex Fridman (1:25:37.140)
So I know you believe as many do that we're far
Kai-Fu Lee (1:25:41.140)
from creating an artificial general intelligence system.
Lex Fridman (1:25:45.420)
But say once we do, and you get to ask her one question,
Lex Fridman (1:25:50.420)
what would that question be?
Lex Fridman (1:25:57.640)
What is it that differentiates you and me?
Kai-Fu Lee (1:26:02.180)
Beautifully put, Kaifu, thank you so much
Lex Fridman (1:26:04.260)
for your time today.
Kai-Fu Lee (1:26:05.760)
Thank you.
Lex Fridman (20:03.200)
and just actually just not be evil
Lex Fridman (20:06.120)
and do good for the world?
Lex Fridman (20:08.000)
It's really hard and I think Google
Kai-Fu Lee (20:10.240)
has struggled with that.
Lex Fridman (20:13.120)
First, the don't do evil mantra is very dangerous
Kai-Fu Lee (20:16.720)
because every employee's definition of evil is different.
Lex Fridman (20:20.800)
And that has led to some difficult
Kai-Fu Lee (20:22.600)
employee situations for them.
Lex Fridman (20:25.120)
So I don't necessarily think that's a good value statement,
Lex Fridman (20:29.560)
but just watching the kinds of things Google
Lex Fridman (20:32.320)
or its parent company Alphabet does
Kai-Fu Lee (20:35.320)
in new areas like healthcare, like eradicating mosquitoes,
Lex Fridman (20:40.480)
things that are really not in the business
Kai-Fu Lee (20:42.360)
of a internet tech company.
Lex Fridman (20:44.520)
I think that shows that there's a heart and soul
Lex Fridman (20:47.200)
and desire to do good and willingness
Lex Fridman (20:50.360)
to put in the resources to do something
Kai-Fu Lee (20:54.800)
when they see it's good, they will pursue it.
Lex Fridman (20:58.280)
That doesn't necessarily mean
Kai-Fu Lee (21:00.160)
it has all the trust of the users.
Lex Fridman (21:02.480)
I realize while most people would view Facebook
Kai-Fu Lee (21:06.360)
as the primary target of their recent unhappiness
Lex Fridman (21:09.760)
about Silicon Valley companies,
Kai-Fu Lee (21:11.560)
many would put Google in that category.
Lex Fridman (21:14.040)
And some have named Google's business practices
Kai-Fu Lee (21:16.760)
as predatory also.
Lex Fridman (21:19.720)
So it's kind of difficult to have the two parts of a body.
Kai-Fu Lee (21:24.240)
The brain wants to do what it's supposed to do
Lex Fridman (21:27.440)
for a shareholder, maximize profit.
Lex Fridman (21:29.280)
And then the heart and soul wants to do good things
Lex Fridman (21:32.000)
that may run against what the brain wants to do.
Lex Fridman (21:36.120)
So in this complex balancing
Lex Fridman (21:39.000)
that these companies have to do,
Kai-Fu Lee (21:40.320)
you've mentioned that you're concerned
Lex Fridman (21:42.480)
about a future where too few companies
Kai-Fu Lee (21:45.480)
like Google, Facebook, Amazon are controlling our data
Lex Fridman (21:49.640)
or controlling too much of our digital lives.
Lex Fridman (21:53.360)
Can you elaborate on this concern
Lex Fridman (21:54.840)
and perhaps do you have a better way forward?
Kai-Fu Lee (21:57.560)
I think I'm hardly the most vocal complainer of this.
Lex Fridman (22:02.760)
Sure, of course.
Kai-Fu Lee (22:03.760)
There are a lot louder complainers out there.
Lex Fridman (22:06.240)
I do observe that having a lot of data
Kai-Fu Lee (22:10.760)
does perpetuate their strength
Lex Fridman (22:13.600)
and limits competition in many spaces.
Lex Fridman (22:18.600)
But I also believe AI is much broader
Lex Fridman (22:21.560)
than the internet space.
Lex Fridman (22:23.160)
So the entrepreneurial opportunities still exists
Lex Fridman (22:26.520)
in using AI to empower financial,
Kai-Fu Lee (22:30.600)
retail, manufacturing, education applications.
Lex Fridman (22:34.600)
So I don't think it's quite a case
Kai-Fu Lee (22:36.440)
of full monopolistic dominance
Lex Fridman (22:38.920)
that totally stifles innovation.
Lex Fridman (22:43.160)
But I do believe in their areas of strength
Lex Fridman (22:45.600)
it's hard to dislodge them.
Kai-Fu Lee (22:48.960)
I don't know if I have a good solution.
Lex Fridman (22:52.480)
Probably the best solution is let
Kai-Fu Lee (22:54.960)
the entrepreneurial VC ecosystem work well
Lex Fridman (22:58.320)
and find all the places that can create the next Google,
Kai-Fu Lee (23:02.160)
the next Facebook.
Lex Fridman (23:03.520)
So there will always be increasing number of challengers.
Kai-Fu Lee (23:07.880)
In some sense that has happened a little bit.
Lex Fridman (23:10.680)
You see Uber, Airbnb having emerged
Kai-Fu Lee (23:14.000)
despite the strength of the big three.
Lex Fridman (23:19.440)
And I think China as an environment
Kai-Fu Lee (23:21.760)
may be more interesting for the emergence
Lex Fridman (23:24.520)
because if you look at companies
Kai-Fu Lee (23:26.120)
between let's say 50 to $300 billion,
Lex Fridman (23:32.920)
China has emerged more of such companies
Kai-Fu Lee (23:35.720)
than the US in the last three to four years
Lex Fridman (23:39.320)
because of the larger marketplace,
Kai-Fu Lee (23:41.560)
because of the more fearless nature of the entrepreneurs.
Lex Fridman (23:46.440)
And the Chinese giants are just as powerful
Kai-Fu Lee (23:48.960)
as American ones.
Lex Fridman (23:50.240)
Tencent, Alibaba are very strong,
Lex Fridman (23:52.360)
but ByteDance has emerged worth 75 billion
Lex Fridman (23:56.480)
and financial while it's Alibaba affiliated,
Kai-Fu Lee (23:59.600)
it's nevertheless independent and worth 150 billion.
Lex Fridman (24:03.400)
And so I do think if we start to extend
Kai-Fu Lee (24:07.800)
to traditional businesses,
Lex Fridman (24:09.440)
we will see very valuable companies.
Lex Fridman (24:12.160)
So it's probably not the case that in five or 10 years
Lex Fridman (24:17.600)
we'll still see the whole world
Kai-Fu Lee (24:19.160)
with these five companies having such dominance.
Lex Fridman (24:22.720)
So you've mentioned a couple of times
Kai-Fu Lee (24:26.080)
this fascinating world of entrepreneurship in China
Lex Fridman (24:29.240)
of the fearless nature of the entrepreneur.
Lex Fridman (24:31.080)
So can you maybe talk a little bit about
Lex Fridman (24:33.240)
what it takes to be an entrepreneur in China?
Lex Fridman (24:35.520)
What are the strategies that are undertaken?
Lex Fridman (24:38.240)
What are the ways to achieve success?
Lex Fridman (24:41.120)
What is the dynamic of VCF funding
Lex Fridman (24:43.920)
of the way the government helps companies and so on?
Lex Fridman (24:46.480)
What are the interesting aspects here
Lex Fridman (24:47.880)
that are distinct from, that are different
Lex Fridman (24:50.200)
from the Silicon Valley world of entrepreneurship?
Lex Fridman (24:55.240)
Well, many of the listeners probably still
Kai-Fu Lee (24:59.400)
would brand Chinese entrepreneur as copycats.
Lex Fridman (25:03.000)
And no doubt 10 years ago,
Kai-Fu Lee (25:05.360)
that would not be an inaccurate description.
Lex Fridman (25:09.080)
Back 10 years ago,
Kai-Fu Lee (25:10.760)
an entrepreneur probably could not get funding
Lex Fridman (25:13.640)
if he or she could not describe
Lex Fridman (25:16.000)
what product he or she is copying from the US.
Lex Fridman (25:20.440)
The first question is who has proven this business model
Lex Fridman (25:23.360)
which is a nice way of asking who are you copying?
Lex Fridman (25:27.200)
And that reason is understandable
Kai-Fu Lee (25:29.560)
because China had a much lower internet penetration
Lex Fridman (25:34.880)
and didn't have enough indigenous experience
Kai-Fu Lee (25:40.960)
to build innovative products.
Lex Fridman (25:43.240)
And secondly, internet was emerging.
Kai-Fu Lee (25:47.640)
Link startup was the way to do things,
Lex Fridman (25:49.840)
building a first minimally viable product
Lex Fridman (25:52.960)
and then expanding was the right way to go.
Lex Fridman (25:55.360)
And the American successes have given a shortcut
Kai-Fu Lee (25:59.520)
that if you built your minimally viable product
Lex Fridman (26:02.560)
based on an American product,
Kai-Fu Lee (26:04.240)
it's guaranteed to be a decent starting point.
Lex Fridman (26:06.720)
Then you tweak it afterwards.
Lex Fridman (26:08.400)
So as long as there are no IP infringement,
Lex Fridman (26:11.280)
which as far as I know there hasn't been in the mobile
Lex Fridman (26:14.000)
and AI spaces, that's a much better shortcut.
Lex Fridman (26:19.360)
And I think Silicon Valley would view that
Kai-Fu Lee (26:21.880)
as still not very honorable
Lex Fridman (26:25.120)
because that's not your own idea to start with,
Lex Fridman (26:29.160)
but you can't really at the same time
Lex Fridman (26:32.560)
believe every idea must be your own
Lex Fridman (26:35.120)
and believe in the link startup methodology
Lex Fridman (26:38.080)
because link startup is intended to try many, many things
Lex Fridman (26:41.840)
and then converge when that works.
Lex Fridman (26:44.160)
And it's meant to be iterated and changed.
Lex Fridman (26:46.640)
So finding a decent starting point
Lex Fridman (26:48.640)
without legal violations,
Kai-Fu Lee (26:51.160)
there should be nothing morally dishonorable about that.
Lex Fridman (26:55.240)
Yeah, so just a quick pause on that.
Kai-Fu Lee (26:56.920)
It's fascinating that that's,
Lex Fridman (26:59.760)
why is that not honorable, right?
Kai-Fu Lee (27:01.840)
It's exactly as you formulated.
Lex Fridman (27:04.600)
It seems like a perfect start for business.
Kai-Fu Lee (27:07.600)
Is to take, look at Amazon and say,
Lex Fridman (27:12.080)
okay, we'll do exactly what Amazon is doing.
Kai-Fu Lee (27:14.480)
Let's start there in this particular market
Lex Fridman (27:16.760)
and then let's out innovate them from that starting point.
Kai-Fu Lee (27:20.720)
Come up with new ways.
Lex Fridman (27:22.200)
I mean, is it wrong to be,
Kai-Fu Lee (27:25.200)
except the word copycat just sounds bad,
Lex Fridman (27:27.080)
but is it wrong to be a copycat?
Kai-Fu Lee (27:28.760)
It just seems like a smart strategy,
Lex Fridman (27:31.600)
but yes, it doesn't have a heroic nature to it
Kai-Fu Lee (27:35.760)
that like Steve Jobs, Elon Musk,
Lex Fridman (27:40.720)
sort of in something completely,
Kai-Fu Lee (27:42.200)
coming up with something completely new.
Lex Fridman (27:43.880)
Yeah, I like the way you describe it.
Kai-Fu Lee (27:45.280)
It's a nonheroic, acceptable way to start the company
Lex Fridman (27:50.440)
and maybe more expedient.
Lex Fridman (27:52.800)
So that's, I think, a baggage for Silicon Valley
Lex Fridman (27:58.920)
that if it doesn't let go,
Kai-Fu Lee (28:00.760)
then it may limit the ultimate ceiling of the company.
Lex Fridman (28:05.120)
Take Snapchat as an example.
Kai-Fu Lee (28:07.200)
I think, you know, Evan's brilliant.
Lex Fridman (28:09.840)
He built a great product,
Lex Fridman (28:11.520)
but he's very proud that he wants to build his own features,
Lex Fridman (28:15.440)
not copy others.
Kai-Fu Lee (28:16.800)
While Facebook was more willing to copy his features
Lex Fridman (28:20.960)
and you see what happens in the competition.
Lex Fridman (28:23.440)
So I think putting that handcuff on the company
Lex Fridman (28:27.440)
would limit its ability to reach the maximum potential.
Lex Fridman (28:31.560)
So back to the Chinese environment,
Lex Fridman (28:33.840)
copying was merely a way to learn from the American masters.
Kai-Fu Lee (28:38.400)
Just like we, if we learned to play piano or painting,
Lex Fridman (28:43.440)
you start by copying.
Kai-Fu Lee (28:44.520)
You don't start by innovating
Lex Fridman (28:45.960)
when you don't have the basic skill sets.
Lex Fridman (28:48.160)
So very amazingly, the Chinese entrepreneurs
Lex Fridman (28:51.000)
about six years ago started to branch off
Kai-Fu Lee (28:56.120)
with these lean startups built on American ideas
Lex Fridman (28:59.480)
to build better products than American products.
Lex Fridman (29:02.120)
But they did start from the American idea.
Lex Fridman (29:04.960)
And today WeChat is better than WhatsApp,
Kai-Fu Lee (29:08.560)
Weibo is better than Twitter,
Lex Fridman (29:10.520)
Zhihu is better than Quora and so on.
Lex Fridman (29:12.920)
So that I think is Chinese entrepreneurs going to step two.
Lex Fridman (29:18.520)
And then step three is once these entrepreneurs
Kai-Fu Lee (29:21.560)
have done one or two of these companies,
Lex Fridman (29:23.720)
they now look at the Chinese market and the opportunities
Lex Fridman (29:27.400)
and come up with ideas that didn't exist elsewhere.
Lex Fridman (29:30.600)
So products like Ant Financial,
Kai-Fu Lee (29:34.200)
under which includes Alipay, which is mobile payments,
Lex Fridman (29:38.280)
and also the financial products for loans built on that.
Lex Fridman (29:44.320)
And also in education, VIPKID,
Lex Fridman (29:48.560)
and in social video, social network, TikTok,
Lex Fridman (29:54.880)
and in social eCommerce, Pinduoduo,
Lex Fridman (29:58.640)
and then in ride sharing, Mobike,
Kai-Fu Lee (30:01.680)
these are all Chinese innovated products
Lex Fridman (30:05.600)
that now are being copied elsewhere.
Lex Fridman (30:08.680)
So an additional interesting observation
Lex Fridman (30:13.000)
is some of these products
Kai-Fu Lee (30:14.240)
are built on unique Chinese demographics,
Lex Fridman (30:17.240)
which may not work in the US,
Lex Fridman (30:19.360)
but may work very well in Southeast Asia, Africa,
Lex Fridman (30:23.120)
and other developing worlds
Kai-Fu Lee (30:25.160)
that are a few years behind China.
Lex Fridman (30:27.800)
And a few of these products maybe are universal
Lex Fridman (30:31.000)
and are getting traction even in the United States,
Lex Fridman (30:33.720)
such as TikTok.
Lex Fridman (30:35.320)
So this whole ecosystem is supported by VCs
Lex Fridman (30:42.040)
as a virtuous cycle,
Kai-Fu Lee (30:43.480)
because a large market with innovative entrepreneurs
Lex Fridman (30:47.640)
will draw a lot of money
Lex Fridman (30:49.360)
and then invest in these companies.
Lex Fridman (30:51.520)
As the market gets larger and larger,
Kai-Fu Lee (30:54.480)
the China market is easily three, four times larger than the US,
Lex Fridman (30:58.440)
they will create greater value and greater returns
Kai-Fu Lee (31:01.120)
for the VCs, thereby raising even more money.
Lex Fridman (31:05.400)
So at Sinovation Ventures, our first fund was 15 million,
Kai-Fu Lee (31:09.960)
our last fund was 500 million.
Lex Fridman (31:12.000)
So it reflects the valuation of the companies
Lex Fridman (31:16.520)
and our us going multi stage and things like that.
Lex Fridman (31:19.840)
It also has government support,
Lex Fridman (31:22.000)
but not in the way most Americans would think of it.
Lex Fridman (31:25.360)
The government actually leaves the entrepreneurial space
Kai-Fu Lee (31:28.720)
as a private enterprise, sort of self regulating,
Lex Fridman (31:32.400)
and the government would build infrastructures
Kai-Fu Lee (31:35.400)
that would around it to make it work better.
Lex Fridman (31:38.520)
For example, the Mass Entrepreneur Mass Innovation Plan
Kai-Fu Lee (31:42.160)
builds 8,000 incubators,
Lex Fridman (31:44.120)
so the pipeline is very strong to the VCs.
Kai-Fu Lee (31:47.560)
For autonomous vehicles,
Lex Fridman (31:48.920)
the Chinese government is building smart highways
Kai-Fu Lee (31:52.600)
with sensors, smart cities
Lex Fridman (31:54.640)
that separate pedestrians from cars
Kai-Fu Lee (31:57.240)
that may allow initially an inferior
Lex Fridman (32:00.440)
autonomous vehicle company to launch a car
Kai-Fu Lee (32:03.640)
without increasing with lower casualty
Lex Fridman (32:07.000)
because the roads or the city is smart.
Lex Fridman (32:10.880)
And the Chinese government at local levels
Lex Fridman (32:13.160)
would have these guiding funds acting as LPs,
Kai-Fu Lee (32:16.720)
passive LPs to funds.
Lex Fridman (32:18.800)
And when the fund makes money,
Kai-Fu Lee (32:21.400)
part of the money made is given back
Lex Fridman (32:23.880)
to the GPs and potentially other LPs
Kai-Fu Lee (32:26.720)
to increase everybody's return
Lex Fridman (32:30.480)
at the expense of the government's return.
Lex Fridman (32:33.080)
So that's an interesting incentive
Lex Fridman (32:35.800)
that entrusts the task of choosing entrepreneurs to VCs
Kai-Fu Lee (32:41.080)
who are better at it than the government
Lex Fridman (32:43.240)
by letting some of the profits move that way.
Lex Fridman (32:46.080)
So this is really fascinating, right?
Lex Fridman (32:48.720)
So I look at the Russian government as a case study
Kai-Fu Lee (32:51.800)
where, let me put it this way,
Lex Fridman (32:54.240)
there's no such government driven
Kai-Fu Lee (32:57.840)
large scale support of entrepreneurship.
Lex Fridman (33:00.840)
And probably the same is true in the United States,
Lex Fridman (33:04.000)
but the entrepreneurs themselves kind of find a way.
Lex Fridman (33:07.640)
So maybe in a form of advice or explanation,
Lex Fridman (33:11.640)
how did the Chinese government arrive to be this way
Lex Fridman (33:15.560)
so supportive on entrepreneurship
Kai-Fu Lee (33:17.720)
to be in this particular way so forward thinking
Lex Fridman (33:21.520)
at such a large scale?
Lex Fridman (33:23.120)
And also perhaps, how can we copy it in other countries?
Lex Fridman (33:28.280)
How can we encourage other governments,
Kai-Fu Lee (33:29.800)
like even the United States government,
Lex Fridman (33:31.600)
to support infrastructure for autonomous vehicles
Lex Fridman (33:33.800)
in that same kind of way, perhaps?
Lex Fridman (33:36.040)
Yes, so these techniques are
Kai-Fu Lee (33:41.040)
the result of several key things,
Lex Fridman (33:44.440)
some of which may be learnable,
Kai-Fu Lee (33:46.040)
some of which may be very hard.
Lex Fridman (33:48.040)
One is just trial and error
Lex Fridman (33:50.520)
and watching what everyone else is doing.
Lex Fridman (33:52.920)
I think it's important to be humble
Lex Fridman (33:54.680)
and not feel like you know all the answers.
Lex Fridman (33:56.920)
The guiding funds idea came from Singapore,
Kai-Fu Lee (33:59.480)
which came from Israel.
Lex Fridman (34:01.440)
And China made a few tweaks and turned it into a,
Kai-Fu Lee (34:06.120)
because the Chinese cities and government officials
Lex Fridman (34:09.320)
kind of compete with each other
Kai-Fu Lee (34:11.360)
because they all want to make their city more successful
Lex Fridman (34:14.680)
so they can get the next level in their political career.
Lex Fridman (34:20.280)
And it's somewhat competitive.
Lex Fridman (34:22.280)
So the central government made it a bit of a competition.
Kai-Fu Lee (34:25.160)
Everybody has a budget.
Lex Fridman (34:26.800)
They can put it on AI or they can put it on bio
Kai-Fu Lee (34:29.840)
or they can put it on energy.
Lex Fridman (34:32.160)
And then whoever gets the results,
Kai-Fu Lee (34:34.120)
the city shines, the people are better off,
Lex Fridman (34:36.160)
the mayor gets a promotion.
Lex Fridman (34:37.960)
So the tools is kind of almost like an entrepreneurial
Lex Fridman (34:41.680)
environment for local governments
Kai-Fu Lee (34:44.840)
to see who can do a better job.
Lex Fridman (34:47.480)
And also many of them try different experiments.
Kai-Fu Lee (34:52.480)
Some have given award to very smart researchers.
Lex Fridman (34:58.480)
Just give them money and hope they'll start a company.
Kai-Fu Lee (35:00.840)
Some have given money to academic research labs,
Lex Fridman (35:05.840)
maybe government research labs
Kai-Fu Lee (35:08.080)
to see if they can spin off some companies
Lex Fridman (35:10.760)
from the science lab or something like that.
Kai-Fu Lee (35:14.080)
Some have tried to recruit overseas Chinese
Lex Fridman (35:17.120)
to come back and start companies.
Lex Fridman (35:19.000)
And they've had mixed results.
Lex Fridman (35:21.000)
The one that worked the best was the guiding funds.
Lex Fridman (35:23.400)
So it's almost like a lean startup idea
Lex Fridman (35:25.880)
where people try different things
Lex Fridman (35:27.560)
and what works sticks and everybody copies.
Lex Fridman (35:30.600)
So now every city has a guiding fund.
Lex Fridman (35:32.880)
So that's how that came about.
Lex Fridman (35:34.600)
The autonomous vehicle and the massive spending
Kai-Fu Lee (35:40.400)
in highways and smart cities, that's a Chinese way.
Lex Fridman (35:46.080)
It's about building infrastructure to facilitate.
Kai-Fu Lee (35:49.520)
It's a clear division of the government's responsibility
Lex Fridman (35:52.880)
from the market.
Kai-Fu Lee (35:55.400)
The market should do everything in a private freeway,
Lex Fridman (36:00.560)
but there are things the market can't afford to do
Kai-Fu Lee (36:02.920)
like infrastructure.
Lex Fridman (36:04.480)
So the government always appropriates large amounts
Kai-Fu Lee (36:08.680)
of money for infrastructure building.
Lex Fridman (36:11.960)
This happens with not only autonomous vehicle and AI,
Lex Fridman (36:16.840)
but happened with the 3G and 4G.
Lex Fridman (36:20.800)
You'll find that the Chinese wireless reception
Kai-Fu Lee (36:25.320)
is better than the US because massive spending
Lex Fridman (36:28.560)
that tries to cover the whole country,
Kai-Fu Lee (36:30.600)
whereas in the US it may be a little spotty.
Lex Fridman (36:33.040)
It's a government driven because I think they view
Kai-Fu Lee (36:36.880)
the coverage of cell access and 3G, 4G access
Lex Fridman (36:44.120)
to be a governmental infrastructure spending
Kai-Fu Lee (36:47.080)
as opposed to capitalistic.
Lex Fridman (36:49.880)
So that's, of course, the state owned enterprises
Kai-Fu Lee (36:52.240)
are also publicly traded,
Lex Fridman (36:54.280)
but they also carry a government responsibility
Kai-Fu Lee (36:57.760)
to deliver infrastructure to all.
Lex Fridman (37:00.240)
So it's a different way of thinking
Kai-Fu Lee (37:01.920)
that may be very hard to inject into Western countries
Lex Fridman (37:05.440)
to say starting tomorrow, bandwidth infrastructure
Lex Fridman (37:09.320)
and highways are gonna be governmental spending
Lex Fridman (37:13.880)
with some characteristics.
Kai-Fu Lee (37:16.120)
What's your sense, and sorry to interrupt,
Lex Fridman (37:18.280)
but because it's such a fascinating point,
Lex Fridman (37:21.680)
do you think on the autonomous vehicle space
Lex Fridman (37:25.600)
it's possible to solve the problem of full autonomy
Lex Fridman (37:30.160)
without significant investment in infrastructure?
Lex Fridman (37:34.120)
Well, that's really hard to speculate.
Kai-Fu Lee (37:36.440)
I think it's not a yes, no question,
Lex Fridman (37:39.000)
but how long does it take question?
Kai-Fu Lee (37:42.000)
15 years, 30 years, 45 years.
Lex Fridman (37:45.200)
Clearly with infrastructure augmentation,
Kai-Fu Lee (37:49.040)
whether it's road, the city or whole city planning,
Lex Fridman (37:52.360)
building a new city, I'm sure that will accelerate
Kai-Fu Lee (37:56.480)
the day of the L5.
Lex Fridman (37:58.320)
I'm not knowledgeable enough,
Lex Fridman (38:01.280)
and it's hard to predict even when we're knowledgeable
Lex Fridman (38:03.920)
because a lot of it is speculative.
Lex Fridman (38:07.160)
But in the US, I don't think people would consider
Lex Fridman (38:10.440)
building a new city the size of Chicago
Kai-Fu Lee (38:13.280)
to make it the AI slash autonomous city.
Lex Fridman (38:15.960)
There are smaller ones being built, I'm aware of that.
Lex Fridman (38:18.880)
But is infrastructure spend really impossible
Lex Fridman (38:22.040)
for US or Western countries?
Kai-Fu Lee (38:23.720)
I don't think so.
Lex Fridman (38:25.760)
The US highway system was built,
Lex Fridman (38:28.920)
was that during President Eisenhower or Kennedy?
Lex Fridman (38:31.960)
Eisenhower, yeah.
Lex Fridman (38:33.160)
So maybe historians can study
Lex Fridman (38:37.240)
how the President Eisenhower get the resources
Kai-Fu Lee (38:40.280)
to build this massive infrastructure
Lex Fridman (38:42.800)
that surely gave US a tremendous amount of prosperity
Kai-Fu Lee (38:47.560)
over the next decade, if not century.
Lex Fridman (38:50.800)
If I may comment on that then,
Kai-Fu Lee (38:53.080)
it takes us to artificial intelligence a little bit
Lex Fridman (38:55.440)
because in order to build infrastructure,
Kai-Fu Lee (38:58.080)
it creates a lot of jobs.
Lex Fridman (39:00.520)
So I'll be actually interested if you would say
Kai-Fu Lee (39:03.400)
that you talk in your book about all kinds of jobs
Lex Fridman (39:06.560)
that could and could not be automated.
Kai-Fu Lee (39:08.920)
I wonder if building infrastructure is one of the jobs
Lex Fridman (39:13.080)
that would not be easily automated.
Kai-Fu Lee (39:15.760)
Something you could think about
Lex Fridman (39:17.040)
because I think you've mentioned somewhere in the talk
Kai-Fu Lee (39:19.920)
or that there might be, as jobs are being automated,
Lex Fridman (39:24.280)
a role for government to create jobs
Kai-Fu Lee (39:26.200)
that can't be automated.
Lex Fridman (39:28.160)
Yes, I think that's a possibility.
Kai-Fu Lee (39:31.040)
Back in the last financial crisis,
Lex Fridman (39:34.280)
China put a lot of money
Kai-Fu Lee (39:37.840)
to basically give this economy a boost
Lex Fridman (39:41.200)
and a lot of it went into infrastructure building.
Lex Fridman (39:45.480)
And I think that's a legitimate way at the government level
Lex Fridman (39:49.920)
to deal with the employment issues
Kai-Fu Lee (39:54.360)
as well as build out the infrastructure
Lex Fridman (39:57.040)
as long as the infrastructures are truly needed
Lex Fridman (39:59.840)
and as long as there is an employment problem,
Lex Fridman (40:02.160)
which no, we don't know.
Lex Fridman (40:04.920)
So maybe taking a little step back,
Lex Fridman (40:07.840)
if you've been a leader and a researcher in AI
Kai-Fu Lee (40:12.840)
for several decades, at least 30 years,
Lex Fridman (40:16.200)
so how has AI changed in the West and the East
Kai-Fu Lee (40:21.000)
as you've observed, as you've been deep in it
Lex Fridman (40:23.080)
over the past 30 years?
Kai-Fu Lee (40:25.080)
Well, AI began as the pursuit
Lex Fridman (40:27.840)
of understanding human intelligence
Lex Fridman (40:30.280)
and the term itself represents that,
Lex Fridman (40:34.160)
but it kind of drifted into the one sub area
Kai-Fu Lee (40:37.520)
that worked extremely well, which is machine intelligence.
Lex Fridman (40:40.840)
And that's actually more using pattern recognition techniques
Kai-Fu Lee (40:45.080)
to basically do incredibly well on a limited domain,
Lex Fridman (40:51.280)
large amount of data,
Lex Fridman (40:52.760)
but relatively simple kinds of planning tasks
Lex Fridman (40:57.120)
and not very creative.
Lex Fridman (40:58.680)
So we didn't end up building human intelligence.
Lex Fridman (41:02.440)
We built a different machine
Kai-Fu Lee (41:04.560)
that was a lot better than us, some problems,
Lex Fridman (41:08.080)
but nowhere close to us on other problems.
Lex Fridman (41:11.720)
So today, I think a lot of people still misunderstand
Lex Fridman (41:16.320)
when we say artificial intelligence
Lex Fridman (41:18.040)
and what various products can do,
Lex Fridman (41:20.720)
people still think it's about replicating
Kai-Fu Lee (41:22.960)
human intelligence,
Lex Fridman (41:24.200)
but the products out there really are closer
Kai-Fu Lee (41:27.600)
to having invented the internet or the spreadsheet
Lex Fridman (41:31.680)
or the database and getting broader adoption.
Lex Fridman (41:35.360)
And speaking further to the fears,
Lex Fridman (41:37.640)
near term fears that people have about AI,
Lex Fridman (41:40.360)
so you're commenting on the sort of general intelligence
Lex Fridman (41:45.080)
that people in the popular culture from sci fi movies
Kai-Fu Lee (41:48.080)
have a sense about AI,
Lex Fridman (41:49.600)
but there's practical fears about AI,
Kai-Fu Lee (41:52.000)
the narrow AI that you're talking about
Lex Fridman (41:54.840)
of automating particular kinds of jobs
Lex Fridman (41:57.280)
and you talk about them in the book.
Lex Fridman (41:59.440)
So what are the kinds of jobs in your view
Kai-Fu Lee (42:01.560)
that you see in the next five, 10 years
Lex Fridman (42:04.480)
beginning to be automated by AI systems algorithms?
Kai-Fu Lee (42:09.280)
Yes, this is also maybe a little bit counterintuitive
Lex Fridman (42:13.040)
because it's the routine jobs
Kai-Fu Lee (42:15.280)
that will be displaced the soonest
Lex Fridman (42:18.360)
and they may not be displaced entirely,
Kai-Fu Lee (42:20.880)
maybe 50%, 80% of a job,
Lex Fridman (42:24.080)
but when the workload drops by that much,
Kai-Fu Lee (42:26.320)
employment will come down.
Lex Fridman (42:28.760)
And also another part of misunderstanding
Kai-Fu Lee (42:31.520)
is most people think of AI replacing routine jobs
Lex Fridman (42:35.720)
than they think of the assembly line, the workers.
Kai-Fu Lee (42:38.760)
Well, that will have some effect,
Lex Fridman (42:41.000)
but it's actually the routine white collar workers
Kai-Fu Lee (42:44.240)
that's easiest to replace
Lex Fridman (42:46.200)
because to replace a white collar worker,
Kai-Fu Lee (42:49.280)
you just need software.
Lex Fridman (42:50.720)
To replace a blue collar worker,
Kai-Fu Lee (42:53.120)
you need robotics, mechanical excellence,
Lex Fridman (42:57.200)
and the ability to deal with dexterity
Lex Fridman (43:01.880)
and maybe even unknown environments, very, very difficult.
Lex Fridman (43:05.640)
So if we were to categorize the most dangerous
Kai-Fu Lee (43:10.600)
white collar jobs,
Lex Fridman (43:12.400)
they would be things like back office,
Kai-Fu Lee (43:15.640)
people who copy and paste
Lex Fridman (43:17.920)
and deal with simple computer programs and data
Lex Fridman (43:22.200)
and maybe paper and OCR,
Lex Fridman (43:25.600)
and they don't make strategic decisions.
Kai-Fu Lee (43:29.040)
They basically facilitate the process.
Lex Fridman (43:32.040)
These softwares and paper systems don't work.
Lex Fridman (43:34.600)
So you have people dealing with new employee orientation,
Lex Fridman (43:40.520)
searching for past lawsuits and financial documents,
Lex Fridman (43:45.200)
and doing reference check.
Lex Fridman (43:47.920)
So basic searching and management of data.
Kai-Fu Lee (43:50.960)
That's the most endangered being lost.
Lex Fridman (43:52.800)
In addition to the white collar repetitive work,
Kai-Fu Lee (43:56.440)
a lot of simple interaction work can also be taken care of
Lex Fridman (44:00.240)
such as telesales, telemarketing, customer service,
Kai-Fu Lee (44:04.840)
as well as many physical jobs
Lex Fridman (44:07.280)
that are in the same location
Lex Fridman (44:09.520)
and don't require a high degree of dexterity.
Lex Fridman (44:12.200)
So fruit picking, dishwashing, assembly line inspection
Kai-Fu Lee (44:17.840)
are jobs in that category.
Lex Fridman (44:20.320)
So altogether, back office is a big part.
Lex Fridman (44:25.400)
And the blue collar may be smaller initially,
Lex Fridman (44:29.840)
but over time, AI will get better.
Lex Fridman (44:32.520)
And when we start to get to over the next 15, 20 years,
Lex Fridman (44:36.880)
the ability to actually have the dexterity
Kai-Fu Lee (44:39.120)
of doing assembly line, that's a huge chunk of jobs.
Lex Fridman (44:42.600)
And when autonomous vehicles start to work,
Kai-Fu Lee (44:45.600)
initially starting with truck drivers,
Lex Fridman (44:47.440)
but eventually to all drivers,
Kai-Fu Lee (44:49.360)
that's another huge group of workers.
Lex Fridman (44:52.000)
So I see modest numbers in the next five years,
Lex Fridman (44:55.600)
but increasing rapidly after that.
Lex Fridman (44:58.080)
On the worry of the jobs that are in danger
Lex Fridman (45:01.240)
and the gradual loss of jobs,
Lex Fridman (45:04.080)
I'm not sure if you're familiar with Andrew Yang.
Kai-Fu Lee (45:06.680)
Yes, I am.
Lex Fridman (45:07.840)
So there's a candidate for president of the United States
Kai-Fu Lee (45:10.600)
whose platform Andrew Yang is based around,
Lex Fridman (45:14.400)
in part around job loss due to automation.
Lex Fridman (45:17.720)
And also in addition,
Lex Fridman (45:19.600)
the need perhaps of universal basic income
Kai-Fu Lee (45:22.600)
to support jobs that are,
Lex Fridman (45:25.280)
folks who lose their job due to automation and so on.
Lex Fridman (45:28.600)
And in general, support people
Lex Fridman (45:30.400)
under complex, unstable job market.
Lex Fridman (45:34.360)
So what are your thoughts about his concerns,
Lex Fridman (45:36.760)
him as a candidate, his ideas in general?
Kai-Fu Lee (45:40.080)
I think his thinking is generally in the right direction,
Lex Fridman (45:44.680)
but his approach as a presidential candidate
Kai-Fu Lee (45:48.480)
may be a little bit ahead of the time.
Lex Fridman (45:51.040)
And I think the displacements will happen,
Lex Fridman (45:56.120)
but will they happen soon enough
Lex Fridman (45:57.800)
for people to agree to vote for him?
Kai-Fu Lee (46:00.480)
The unemployment numbers are not very high yet.
Lex Fridman (46:03.800)
And I think he and I have the same challenge.
Kai-Fu Lee (46:07.600)
If I want to theoretically convince people this is an issue
Lex Fridman (46:11.560)
and he wants to become the president,
Kai-Fu Lee (46:13.920)
people have to see how can this be the case
Lex Fridman (46:17.520)
when unemployment numbers are low.
Lex Fridman (46:19.720)
So that is the challenge.
Lex Fridman (46:21.400)
And I think I do agree with him on the displacement issue,
Kai-Fu Lee (46:27.080)
on universal basic income.
Lex Fridman (46:30.280)
At a very vanilla level, I don't agree with it
Kai-Fu Lee (46:33.920)
because I think the main issue is retraining.
Lex Fridman (46:38.360)
So people need to be incented
Kai-Fu Lee (46:41.520)
not by just giving a monthly $2,000 check or $1,000 check
Lex Fridman (46:45.480)
and do whatever they want
Kai-Fu Lee (46:47.160)
because they don't have the know how
Lex Fridman (46:50.960)
to know what to retrain to go into what type of a job.
Lex Fridman (46:56.840)
And guidance is needed.
Lex Fridman (46:58.640)
And retraining is needed
Kai-Fu Lee (47:00.440)
because historically when technology revolutions,
Lex Fridman (47:03.160)
when routine jobs were displaced, new routine jobs came up.
Lex Fridman (47:06.880)
So there was always room for that.
Lex Fridman (47:09.400)
But with AI and automation,
Kai-Fu Lee (47:11.840)
the whole point is replacing all routine jobs eventually.
Lex Fridman (47:15.320)
So there will be fewer and fewer routine jobs.
Lex Fridman (47:17.840)
And AI will create jobs, but it won't create routine jobs
Lex Fridman (47:22.640)
because if it creates routine jobs,
Lex Fridman (47:24.840)
why wouldn't AI just do it?
Lex Fridman (47:26.880)
So therefore the people who are losing the jobs
Kai-Fu Lee (47:30.360)
are losing routine jobs.
Lex Fridman (47:32.280)
The jobs that are becoming available are non routine jobs.
Lex Fridman (47:35.680)
So the social stipend needs to be put in place
Lex Fridman (47:39.280)
is for the routine workers who lost their jobs
Kai-Fu Lee (47:42.040)
to be retrained maybe in six months, maybe in three years,
Lex Fridman (47:46.120)
takes a while to retrain on the non routine job
Lex Fridman (47:48.560)
and then take on a job that will last
Lex Fridman (47:51.360)
for that person's lifetime.
Kai-Fu Lee (47:53.360)
Now, having said that,
Lex Fridman (47:55.280)
if you look deeply into Andrew's document,
Kai-Fu Lee (47:57.120)
he does cater for that.
Lex Fridman (47:58.200)
So I'm not disagreeing with what he's trying to do.
Lex Fridman (48:03.280)
But for simplification, sometimes he just says UBI,
Lex Fridman (48:06.320)
but simple UBI wouldn't work.
Lex Fridman (48:08.720)
And I think you've mentioned elsewhere
Lex Fridman (48:10.560)
that the goal isn't necessarily to give people enough money
Kai-Fu Lee (48:15.720)
to survive or live, or even to prosper.
Lex Fridman (48:19.080)
The point is to give them a job that gives them meaning.
Kai-Fu Lee (48:22.760)
That meaning is extremely important.
Lex Fridman (48:25.560)
That our employment, at least in the United States
Lex Fridman (48:28.560)
and perhaps it carries across the world,
Lex Fridman (48:31.120)
provides something that's, forgive me for saying,
Kai-Fu Lee (48:34.520)
greater than money.
Lex Fridman (48:35.600)
It provides meaning.
Lex Fridman (48:36.880)
So now, what kind of jobs do you think can't be automated?
Lex Fridman (48:43.280)
Can you talk a little bit about creativity
Lex Fridman (48:45.080)
and compassion in your book?
Lex Fridman (48:46.640)
What aspects do you think it's difficult to automate
Lex Fridman (48:49.640)
for an AI system?
Lex Fridman (48:51.920)
Because an AI system is currently merely optimizing.
Kai-Fu Lee (48:56.440)
It's not able to reason, plan,
Lex Fridman (48:59.240)
or think creatively or strategically.
Kai-Fu Lee (49:02.040)
It's not able to deal with complex problems.
Lex Fridman (49:04.400)
It can't come up with a new problem and solve it.
Kai-Fu Lee (49:08.800)
A human needs to find the problem
Lex Fridman (49:11.480)
and pose it as an optimization problem,
Kai-Fu Lee (49:14.760)
then have the AI work at it.
Lex Fridman (49:16.640)
So an AI would have a very hard time discovering a new drug
Kai-Fu Lee (49:22.360)
or discovering a new style of painting
Lex Fridman (49:26.440)
or dealing with complex tasks such as managing a company
Kai-Fu Lee (49:31.440)
that isn't just about optimizing the bottom line,
Lex Fridman (49:34.440)
but also about employee satisfaction, corporate brand,
Lex Fridman (49:38.000)
and many, many other things.
Lex Fridman (49:39.840)
So that is one category of things.
Lex Fridman (49:43.400)
And because these things are challenging, creative, complex,
Lex Fridman (49:47.360)
doing them creates a high degree of satisfaction
Lex Fridman (49:51.160)
and therefore appealing to our desire for working,
Lex Fridman (49:54.480)
which isn't just to make the money, make the ends meet,
Lex Fridman (49:57.360)
but also that we've accomplished something
Lex Fridman (49:59.560)
that others maybe can't do or can't do as well.
Kai-Fu Lee (50:03.320)
Another type of job that is much numerous
Lex Fridman (50:06.320)
would be compassionate jobs, jobs that require compassion,
Kai-Fu Lee (50:10.240)
empathy, human touch, human trust.
Lex Fridman (50:13.600)
AI can't do that because AI is cold, calculating,
Lex Fridman (50:17.760)
and even if it can fake that to some extent,
Lex Fridman (50:21.920)
it will make errors and that will make it look very silly.
Lex Fridman (50:25.120)
And also, I think even if AI did okay,
Lex Fridman (50:28.320)
people would want to interact with another person,
Kai-Fu Lee (50:32.000)
whether it's for some kind of a service or a teacher
Lex Fridman (50:35.920)
or a doctor or a concierge or a masseuse or a bartender.
Kai-Fu Lee (50:40.920)
There are so many jobs where people just don't want
Lex Fridman (50:44.440)
to interact with a cold robot or software.
Kai-Fu Lee (50:49.400)
I've had an entrepreneur who built an elderly care robot
Lex Fridman (50:52.480)
and they found that the elderly really only use it
Kai-Fu Lee (50:55.240)
for customer service.
Lex Fridman (50:56.880)
And not, but not to service the product,
Lex Fridman (50:59.480)
but they click on customer service
Lex Fridman (51:01.840)
and the video of a person comes up
Lex Fridman (51:04.120)
and then the person says,
Lex Fridman (51:05.960)
how come my daughter didn't call me?
Kai-Fu Lee (51:08.080)
Let me show you a picture of her grandkids.
Lex Fridman (51:10.240)
So people yearn for that people, people interaction.
Lex Fridman (51:14.000)
So even if robots improved, people just don't want it.
Lex Fridman (51:17.840)
And those jobs are going to be increasing
Kai-Fu Lee (51:20.120)
because AI will create a lot of value,
Lex Fridman (51:22.840)
$16 trillion to the world in the next 10 years.
Kai-Fu Lee (51:26.640)
Next 11 years, according to PWC.
Lex Fridman (51:29.520)
And that will give people money to enjoy services,
Kai-Fu Lee (51:34.640)
whether it's eating a gourmet meal or tourism and traveling
Lex Fridman (51:39.680)
or having concierge services,
Kai-Fu Lee (51:41.480)
the services revolving around every dollar
Lex Fridman (51:45.600)
of that $16 trillion will be tremendous.
Kai-Fu Lee (51:48.080)
It will create more opportunities
Lex Fridman (51:50.080)
that are to service the people who did well
Kai-Fu Lee (51:52.800)
through AI with things.
Lex Fridman (51:56.240)
But even at the same time,
Kai-Fu Lee (51:58.280)
the entire society is very much short
Lex Fridman (52:01.720)
in need of many service oriented,
Kai-Fu Lee (52:04.240)
compassionate oriented jobs.
Lex Fridman (52:06.200)
The best example is probably in healthcare services.
Kai-Fu Lee (52:10.360)
There's going to be 2 million new jobs,
Lex Fridman (52:14.320)
not counting replacement,
Kai-Fu Lee (52:15.400)
just brand new incremental jobs
Lex Fridman (52:17.520)
in the next six years in healthcare services.
Kai-Fu Lee (52:20.560)
That includes nurses, orderly in the hospital,
Lex Fridman (52:25.000)
elderly care and also at home care is particularly lacking.
Lex Fridman (52:31.480)
And those jobs are not likely to be filled.
Lex Fridman (52:34.840)
So there's likely to be a shortage.
Lex Fridman (52:36.840)
And the reason they're not filled
Lex Fridman (52:38.600)
is simply because they don't pay very well
Lex Fridman (52:41.600)
and that the social status of these jobs are not very good.
Lex Fridman (52:47.520)
So they pay about half as much
Kai-Fu Lee (52:49.720)
as a heavy equipment operator,
Lex Fridman (52:52.160)
which will be replaced a lot sooner.
Lex Fridman (52:55.600)
And they pay probably comparably
Lex Fridman (52:57.560)
to someone on the assembly line.
Lex Fridman (52:59.880)
And so if we ignoring all the other issues
Lex Fridman (53:03.480)
and just think about satisfaction from one's job,
Kai-Fu Lee (53:07.160)
someone repetitively doing the same manual action
Lex Fridman (53:10.520)
at an assembly line,
Kai-Fu Lee (53:11.640)
that can't create a lot of job satisfaction,
Lex Fridman (53:14.640)
but someone taking care of a sick person
Lex Fridman (53:17.680)
and getting a hug and thank you
Lex Fridman (53:19.600)
from that person and the family,
Kai-Fu Lee (53:22.000)
I think is quite satisfying.
Lex Fridman (53:24.680)
So if only we could fix the pay for service jobs,
Kai-Fu Lee (53:28.720)
there are plenty of jobs that require some training
Lex Fridman (53:31.880)
or a lot of training
Kai-Fu Lee (53:33.400)
for the people coming off the routine jobs to take.
Lex Fridman (53:36.960)
We can easily imagine someone
Kai-Fu Lee (53:40.320)
who was maybe a cashier at the grocery store
Lex Fridman (53:43.480)
as stores become automated,
Kai-Fu Lee (53:45.640)
learns to become a nurse or an at home care.
Lex Fridman (53:49.160)
I also do want to point out the blue collar jobs
Kai-Fu Lee (53:52.760)
are going to stay around a bit longer.
Lex Fridman (53:54.720)
Some of them quite a bit longer.
Kai-Fu Lee (53:58.120)
AI cannot be told go clean an arbitrary home.
Lex Fridman (54:02.200)
That's incredibly hard.
Lex Fridman (54:03.920)
Arguably it's an L5 level of difficulty, right?
Lex Fridman (54:07.560)
And then AI cannot be a good plumber
Kai-Fu Lee (54:10.080)
because plumber is almost like a mini detective
Lex Fridman (54:12.840)
that has to figure out where the leak came from.
Lex Fridman (54:15.640)
So yet AI probably can be an assembly line
Lex Fridman (54:20.200)
and auto mechanic and so on.
Lex Fridman (54:22.800)
So one has to study which blue collar jobs are going away
Lex Fridman (54:26.760)
and facilitate retraining for the people
Kai-Fu Lee (54:29.240)
to go into the ones that won't go away
Lex Fridman (54:31.120)
or maybe even will increase.
Kai-Fu Lee (54:32.960)
I mean, it is fascinating that it's easier
Lex Fridman (54:35.000)
to build a world champion chess player
Kai-Fu Lee (54:39.480)
than it is to build a mediocre plumber.
Lex Fridman (54:42.040)
Yes, right.
Kai-Fu Lee (54:43.160)
Very true.
Lex Fridman (54:44.000)
And to AI and that goes counterintuitive
Kai-Fu Lee (54:46.160)
to a lot of people's understanding
Lex Fridman (54:48.000)
of what artificial intelligence is.
Lex Fridman (54:50.120)
So it sounds, I mean, you're painting
Lex Fridman (54:52.480)
a pretty optimistic picture about retraining
Kai-Fu Lee (54:55.400)
about the number of jobs
Lex Fridman (54:57.000)
and actually the meaningful nature of those jobs
Kai-Fu Lee (54:59.560)
once we automate the repetitive tasks.
Lex Fridman (55:02.080)
So overall, are you optimistic about the future
Lex Fridman (55:08.160)
where much of the repetitive tasks are automated?
Lex Fridman (55:11.640)
That there is a lot of room for humans
Kai-Fu Lee (55:13.840)
for the compassionate, for the creative input
Lex Fridman (55:17.360)
that only humans can provide?
Kai-Fu Lee (55:20.080)
I am optimistic if we start to take action.
Lex Fridman (55:23.400)
If we have no action in the next five years,
Kai-Fu Lee (55:27.640)
I think it's going to be hard to deal
Lex Fridman (55:30.760)
with the devastating losses that will emerge.
Lex Fridman (55:34.200)
So if we start thinking about retraining,
Lex Fridman (55:37.120)
maybe with the low hanging fruits,
Kai-Fu Lee (55:39.360)
explaining to vocational schools
Lex Fridman (55:41.800)
why they should train more plumbers than auto mechanics,
Kai-Fu Lee (55:46.640)
maybe starting with some government subsidy
Lex Fridman (55:49.680)
for corporations to have more training positions.
Kai-Fu Lee (55:53.600)
We start to explain to people why retraining is important.
Lex Fridman (55:58.160)
We start to think about what the future of education,
Lex Fridman (56:00.760)
how that needs to be tweaked for the era of AI.
Lex Fridman (56:04.560)
If we start to make incremental progress
Lex Fridman (56:06.760)
and the greater number of people understand,
Lex Fridman (56:08.960)
then there's no reason to think we can't deal with this
Kai-Fu Lee (56:12.360)
because this technological revolution
Lex Fridman (56:14.360)
is arguably similar to what electricity,
Kai-Fu Lee (56:17.280)
industrial revolutions, and internet brought about.
Lex Fridman (56:20.440)
Do you think there's a role for policy,
Kai-Fu Lee (56:22.680)
for governments to step in,
Lex Fridman (56:24.640)
to help with policy to create a better world?
Kai-Fu Lee (56:28.000)
Absolutely, and the governments don't have to believe
Lex Fridman (56:32.280)
an employment will go up,
Lex Fridman (56:34.040)
and they don't have to believe automation will be this fast
Lex Fridman (56:37.560)
to do something.
Kai-Fu Lee (56:39.440)
Revamping vocational school would be one example.
Lex Fridman (56:42.520)
Another is if there's a big gap
Kai-Fu Lee (56:44.760)
in healthcare service employment,
Lex Fridman (56:47.480)
and we know that a country's population is growing older,
Kai-Fu Lee (56:51.840)
more longevity, living older,
Lex Fridman (56:54.280)
because people over 80 require five times as much care
Kai-Fu Lee (56:57.560)
as those under 80,
Lex Fridman (56:59.680)
then it is a good time to incent training programs
Kai-Fu Lee (57:03.480)
for elderly care to find ways to improve the pay.
Lex Fridman (57:07.600)
Maybe one way would be to offer as part of Medicare
Kai-Fu Lee (57:11.720)
or the equivalent program for people over 80
Lex Fridman (57:14.600)
to be entitled to a few hours of elderly care at home,
Lex Fridman (57:18.760)
and then that might be reimbursable,
Lex Fridman (57:22.160)
and that will stimulate the service industry
Kai-Fu Lee (57:26.720)
around the policy.
Lex Fridman (57:28.920)
Do you have concerns about large entities,
Kai-Fu Lee (57:33.400)
whether it's governments or companies,
Lex Fridman (57:35.480)
controlling the future of AI development in general?
Lex Fridman (57:39.240)
So we talked about companies.
Lex Fridman (57:41.000)
Do you have a better sense that governments
Kai-Fu Lee (57:44.360)
can better represent the interests of the people
Lex Fridman (57:49.480)
than companies, or do you believe companies
Lex Fridman (57:52.320)
are better at representing the interests of the people?
Lex Fridman (57:54.960)
Or is there no easy answer?
Kai-Fu Lee (57:56.840)
I don't think there's an easy answer
Lex Fridman (57:58.120)
because it's a double edged sword.
Kai-Fu Lee (58:00.200)
The companies and governments can provide better services
Lex Fridman (58:03.760)
with more access to data and more access to AI,
Lex Fridman (58:06.760)
but that also leads to greater power,
Lex Fridman (58:09.440)
which can lead to uncontrollable problems,
Kai-Fu Lee (58:13.600)
whether it's monopoly or corruption in the government.
Lex Fridman (58:17.760)
So I think one has to be careful
Kai-Fu Lee (58:21.400)
to look at how much data that companies and governments have
Lex Fridman (58:25.040)
and some kind of checks and balances would be helpful.
Lex Fridman (58:30.400)
So again, I come from Russia.
Lex Fridman (58:34.040)
There's something called the Cold War.
Lex Fridman (58:36.840)
So let me ask a difficult question here
Lex Fridman (58:39.240)
looking at conflict.
Kai-Fu Lee (58:40.720)
Steven Pinker written a great book
Lex Fridman (58:42.160)
that conflict all over the world is decreasing in general.
Lex Fridman (58:45.440)
But do you have a sense that having written
Lex Fridman (58:49.680)
the book AI Superpowers,
Lex Fridman (58:51.800)
do you see a major international conflict
Lex Fridman (58:54.440)
potentially arising between major nations,
Kai-Fu Lee (58:57.840)
whatever they are, whether it's Russia, China,
Lex Fridman (59:00.400)
European nations, United States or others
Kai-Fu Lee (59:04.120)
in the next 10, 20, 50 years around AI,
Lex Fridman (59:07.720)
around the digital space, cyberspace?
Lex Fridman (59:10.200)
Do you worry about that?
Lex Fridman (59:12.120)
Is that something we need to think about
Lex Fridman (59:15.520)
and try to alleviate or prevent?
Lex Fridman (59:19.600)
I believe in greater engagement.
Kai-Fu Lee (59:22.680)
A lot of the worries about more powerful AI
Lex Fridman (59:26.720)
are based on a arms race metaphor.
Lex Fridman (59:33.280)
And when you extrapolate into military kinds of scenarios,
Lex Fridman (59:41.560)
AI can automate and autonomous weapons
Kai-Fu Lee (59:46.560)
that needs to be controlled somehow
Lex Fridman (59:48.800)
and autonomous decision making
Kai-Fu Lee (59:51.640)
can lead to not enough time to fix international crises.
Lex Fridman (59:57.560)
So I actually believe a Cold War mentality
🔗 相关节目