李开复 AI
<markdown>
00:00
Kai-Fu, thanks so much for joining us on the show here. Oh, thanks for inviting me. Well, congrats on the new book. Of course, AI 2041. I'm very excited to dig into it. I pre-ordered it. Sorry, I ordered it a couple of days ago and started to read a few of the chapters. I mean, definitely is an interesting perspective. And I love the way you took the fiction route to really get people to really think about and visualize
00:29
the potentials of what we're going to see in the next 20 years. And before we jump into the book, I mean, one of the things that I've heard is that Barack Obama was one of your classmates in Columbia. Is that right? Yes, we were we graduated the same year 1983. Yes. Oh, wow. Did you get to meet him or connect with him while you were there?
00:53
I did not. I was a computer science major and he was political science. So I never met him during my period there. In fact, he's not in, you know, he's actually not in the school book. So I'm not sure what happened, but his picture is not in there. No, he's not. But he's definitely graduated. He didn't show up for his class photos or something. Oh, got it. OK, so he's in the record, but he just isn't in the photo.
01:23
That's right. Yeah. Okay. Okay. This isn't like Trump going back to like him claiming that he's from Kenya or something like that and being born there. It's not a conspiracy theory you're saying. Not at all. He just couldn't find his photo. He's for some reason. Interesting. All right. Well, I'm sure the internet's going to go crazy for that one. Um,
01:43
Cool. Well, I wanted to give you an introduction, of course. We do these beforehand, but just to get people up to speed, some of the key things that you've done before, you know, now at CenoVentures is you've developed a speaker-independent continuous speech recognition system for your PhD. You were the founder and managing director for Microsoft Research Asia. And you've
02:09
And then you became the president for Google China before you got into Cineventures. Tell us a little bit about what you're up to with Cineventures, Ashley, and get your own take on it.
02:21
Sure. We are investors in deep tech. So that includes artificial intelligence, healthcare, life sciences, semiconductor, etc. It leverages my own background in technology. And we are an early stage investor in startup companies. And
02:41
And we will accompany them all the way to IPO. So we're very similar to the likes of Andreessen Horowitz or Greylock or Benchmark in the Silicon Valley. And is the idea because you guys are primarily focused on, let's say, AI, to be focused on companies in China and the US, knowing that they're going to be the duopoly winners in this race that we'll talk about?
03:07
Yeah, we are mostly investing in China. We had earlier efforts to expand globally and including the US, but that became impractical due to recent geopolitical issues. So we're more focused on China now. We do occasionally make an investment abroad. We invested in a French company two months ago, but that's by exception.
03:32
Got it. By impossible, you mean there are regulations set in the US that forbids Chinese investors from investing in the US completely? Or are you just saying it's harder now? It's harder. There's no specific regulation, but there is CFIUS, which means people who take quote unquote Chinese money have to disclose. And that's a hassle to some people.
03:58
And, you know, we're honestly not Chinese money by any means where our investors are almost a half U.S. investors. But nevertheless, because of our primary location, we have been labeled that way. And it's just not convenient for entrepreneurs to take money that requires extra disclosures. You know, entrepreneurs want to build a business, not to spend their life doing disclosures.
04:25
Right, right. Especially with this unpredictability of what will happen in the next five to 10 years. Like we just don't know there might be further turmoil or further restrictions. So it's kind of a marriage, right? When you're taking that money. So I can see why that's the case. Exactly. Is the case the reverse as well, where Chinese companies that may take capital from US investors will also face a lot of scrutiny and challenges? Yeah.
04:55
Not at present. There are some listing requirements, but they're fairly minimal. I think the Chinese government currently has very few regulations on the investor base. In fact, some of the top Chinese IPOs have been largely invested in by U.S. capital and it's still welcome, although obviously there's tension both ways now.
05:24
Got it. Got it. Huh? Yeah, I can, I can see why that would complicate things. Well, I guess in this topic, you know, we wanted to dig into artificial intelligence. It's what the book was about. It's what your previous book was about AI superpowers and would just love to give people some overview background for maybe someone that
05:47
you know, here's AI and headlines, but they're not really sure what AI is exactly. What is the history behind it? And what are the phases that we've gone through to get to where we are now and where we're going? So before we talk about where we're going, which is AI 2041, we'd love to kind of go through what are some of the phases from the beginning until now, meaning what was the first time the term AI or artificial intelligence was
06:16
was coined and what is like the first known AI that people have talked about? Like I've heard that it was the laundromat, like the laundry machine was like the first AI, but we'd love to get your take on this.
06:33
Right. I think it was in the 1950s, at a conference in Dartmouth, the term AI was coined by John McCarthy, who was my PhD advisor's PhD advisor. And he coined the phrase. And at the time, it was very much about, can we get machines to do anything that exhibits human intelligence?
06:58
And for the next 50 years or so, every time something happens,
07:05
that appears to exhibit intelligence, whether it's the telephone switchboard or the smart elevator or laundromat. I don't know. I haven't heard that one before. People very much just say, okay, that's no longer AI. That's engineering. That's a product. So the pursuit of AI has been a very challenging one. When I worked in AI in the 80s and 90s,
07:28
People would look at this as a failure, as a pursuit that is not likely to ever yield results and also continues to write papers but with no workable product.
07:43
So that's been a development for probably the first 50 years or so. There are things that have come out. My PhD thesis was the first speech recognition system that worked for any speaker. And companies that licensed the technology began to put it into, you might recall in the 2000s, late 90s, people who called the
08:07
their stockbroker or airline, were able to reserve an airline reservation or perhaps buy or sell stock by merely using voice on the phone. So that at the time was probably an example of AI.
08:24
But something really big happened about five to 10 years ago, and that's deep learning. And deep learning is arguably invented 30 years ago, but it didn't really work until around 10 years ago.
08:39
And it's a technology that says, well, let's not worry about replicating human thinking exactly, but let's be inspired by the architecture of the human brain and build something that inspired by the human brain with neurons and connections and numbers connecting them. But the way…
09:00
this type of deep learning works is you present many examples of data and then you tell it what the answer should be and it will learn to separate the positive and from the negative exemplars. So if you show it pictures of a cat and a dog and label each one cat or dog, it'll learn to separate cats and dogs.
09:22
And that, in fact, was a Google paper that became quite surprising to people because prior to deep learning, people assumed the human would have to program something like cats have whiskers. They have pointy ears and things like that.
09:37
But it turns out that when you make a deep neural network with a lot of data feeding into a network of many, many layers, up to thousands of layers, it can create its own way of separating dogs from cats. And that can be now extended by Facebook, Twitter.
09:58
by Amazon to separate people who are likely to watch this video from people who are not likely. Then that's why when we see videos on YouTube or TikTok, we like watching more because it already computed and figured out for someone like myself, this is a video that I would likely watch for the whole duration.
10:21
And similarly, Amazon would compute what products I'm likely to buy. Insurance companies and loan companies would figure out to whom the loan should be made. Credit card companies are figuring out which transactions are frauds. And then this list goes on and AI becomes more and more powerful. It has since gained the capability to recognize objects, understand language.
10:44
and autonomous vehicle is beginning to work in constrained environments. So I think we are now at a stage when we have not at all answered the question, how does the human intelligence work? But we have created a set of technologies that are growing very rapidly year after year that can not only do things that we can do that exhibit intelligence, but things that are beyond what we can do, things that we can't imagine us doing.
11:12
So I think it's created a very powerful growth engine. The power of deep learning rests on the fact that if you just throw more data and more compute, it improves itself. So this is very, very powerful. That said, it hasn't really answered the question, how does the human creativity happen?
11:32
Why do we have self-awareness? And why do we have emotions and desires? And can AI have these things? Yet that remains unknown. And perhaps that's what separates us from computed AI.
11:47
Do you feel that similar to the way people didn't really understand what AI was or just refuted the fact and made it into more just an engineering solution that it wasn't really AI? Do you think we'll look in 10, 20 years from now and potentially see that AI also has consciousness and has these emotions and we'll look back and say, oh, like we just didn't have enough data or we just didn't have enough foresight to even think that there would be
12:16
a situation like this? Most people think to really go deeply into true intelligence, human-like capabilities, and with understanding of creativity and self-consciousness will require further breakthroughs. Cannot just be extensions of deep learning and throwing more data at it. And I would fall into that camp. But
12:40
But that said, we are seeing AI mimicking people very quite well. And it's improving rapidly. Deep fakes and fake voices of people generating conversations between questions you can ask Einstein and could be answered in technology.
13:01
Dr. Seuss style with some of the demos are seeing uncanny behavior that is improving. So I wouldn't rule out at some point it really fools people into thinking that it's not only a human, but someone exhibiting emotion creativity. But probably to truly replicate the human process and become a superset that will remain elusive at least in the next 20 years.
13:29
Particularly around consciousness. I can self-claim that I have consciousness. I feel it. But it's hard for me to know from the other person's, for yourself or for my friends, whether they have consciousness. There's really only perspective of myself. Is that ever something we'll ever know from robots or AI in general that they'll have consciousness? How would we ever prove that? Well, the problem is the brain science is still at very formative stages.
13:59
So people who study brain science can't yet answer the question of what is consciousness. And there's still quite a bit of debate. So I think first we have to study our brains and we have to first understand what the question is. Then we can question if AI can exhibit it.
14:17
And I would think, you know, right now, based on current set of technologies, AI can fake a lot of things. Faking visual audio is easier. Faking understanding is harder, but happens sometimes. Faking creativity and consciousness, I think, will be extremely hard sometimes.
14:36
So we'll have many decades to keep working on the one side, but AI scientists will try to fake it. And then the brain scientists will try to figure out what it is that we should fake. Right, right. And it just, everything is moving so fast. And as you mentioned, it seems like it just kind of snuck up on us in the last five to 10 years from the recent paths. And one of the things that you talk about
15:01
And I think you're the perfect person to talk about it because obviously you also went to school in the U.S. and you're very familiar with the culture and everything that's going on around here. And from my understanding is in terms of the AI race, U.S. was in the lead for quite a while until maybe in the recent recession.
15:23
five years ago. What was that transition? And how can you kind of explain what happened there? Like, how do we lose the lead in the US? Well, I don't think US has lost the lead. It's just that China has become very good very quickly. And each country is still quite strong in areas that they're strong. And this was captured in my last book, published in 2018, called AI Superpowers.
15:48
And I think some of the key factors of how China rose up so soon, partly it was the Sputnik moment from the moment that DeepMind, a European technology that beat Chinese top master player at China's own game that claimed to require intelligence.
16:08
And I think that made the entrepreneurs, the VCs, the large companies and the government say, hey, we should rapidly get up to speed on this. And at the time, China already had
16:23
had companies that possessed a lot of data and are in a good position to make use of that data. So these are companies building the super apps, Tencent, Alibaba, ByteDance for some examples. So China has a lot of data and to become very good at AI, one needs to collect data in high quality and large quantities.
16:49
large quantities and structure it and make use of it and connect it to a business metric. And the Chinese super apps are even bigger than the American apps. So Tencent's WeChat is much more powerful than, say, Facebook or Instagram.
17:08
I spent 80% of my time in it. So you can imagine how much data WeChat has for me, as well as for all the other, pretty much everyone I know. And that data becomes really energy and oil to power AI for Chinese companies. That probably is a very important area that the Chinese entrepreneurs were tenacious and developed super apps with a huge amount of data
17:37
The second is that Chinese engineering classes are excellent. Research has been rising up rapidly. And actually, I founded Microsoft Research, as you mentioned, in 1998. And that became really the source of a lot of AI talent who've written a lot of papers and have become quite good and is also catching up with the U.S.,
18:00
and then the entrepreneurial ecosystem with VCs funding tenacious entrepreneurs, as well as favorable Chinese government policies to advance AI and key technologies. So all of these combined together, and also the fact that deep learning really isn't rocket science. For a good computer science student, they can grasp enough deep learning to become a good AI engineer within
18:29
weeks at most months because you don't need everybody to be a deep scientist who writes papers. And China has all these engineers and they see the wonders of AI. And frankly, AI engineer paid a lot more than computer engineering. So people were learning everything on their own and gravitated towards that very, very rapidly. So all of these, I think, played an important role in China becoming very good in
18:59
AI technologies for the uses in internet and financial domains. And more recently, in autonomous vehicles and factory automation, these are some areas where China leads. Also, drones are China's strengths. In the US, I think use of AI in enterprise and cloud technologies is well ahead of China. And I think in terms of
19:25
deep research, US is still ahead, at least in the top 1% of the research. China has caught up in the top maybe 10 to 50, top 10 or 50 or 100% of the research. So I think these two countries lead the world. It's not strictly a…
19:43
a race because companies serve their own sets of customers. Researchers still work together. But if you want to measure what is the market cap of all AI-related companies in two countries or how many people are filing patents or papers, if you want to use those metrics, no doubt China has rapidly caught up.
20:05
Got it. Got it. And particularly just going back to the super apps and the way that it's structured in China, because it's fascinating for me how WeChat is…
20:16
the only app that you can use really to go through your regular day lives. And I would imagine like the average number of apps that a American person has is probably exceeding by double the number of, you know, that someone in, you know, in China may need on a regular basis. And how did it get that way? Is that just like a, do you feel that's just,
20:43
is it almost just a cultural thing to have designed a super app, like one simple app that allows you to do everything? Because from an American perspective, I guess, like there is the product stance of keeping everything minimal and separate and, you know, in its own use case. But what's kind of your take in terms of how the differentiation happened? I think the American culture
21:12
top tech companies follow the Silicon Valley approach, which is it's
21:20
It's a more gentlemanly kind of competition. So you become very deep, very good in what you do. And then there's an ecosystem where everyone works together. So that Grubhub, DoorDash and OpenTable and Yelp and Groupon all have their strengths and they kind of see each other as creating ecosystem of competition and cooperation.
21:46
In China, the entrepreneurial ecosystem has been more of a gladiatorial, tenacious competition, winner take all. And as a result, one single company took all of the spaces in food called Meituan. And it's the leader in the rating, the reservations, the quality.
22:06
the coupons, the delivery, the groceries, and it's going into ride sharing now. So there is a strong desire to expand your empire, leveraging what you have. And of course, the other factor is antitrust laws. I think they were much stronger in the U.S. due to historically seeing the likes of AT&T and Microsoft and Google sort of
22:36
getting into trouble and getting checked by the government, I think has given the US government more of a tougher stance against monopolies. And I think China is now realizing that it too needs to regulate these large companies. But in the early stages, it didn't do so as much as the US government. Yeah. So is it just like this killer culture mentality in China of wanting to
23:05
I've wanted to really dominate and that really drives this monopolistic, you know, super apps that exist today.
23:14
Yeah, I think if you want to go deeper, it's about the culture that has many families that were poor for 10 or 20 generations. And now the younger generation see that they are the first opportunity in 20 generations to become rich, to become successful.
23:36
there's high expectations from their families. So that's why everything's competitive, getting into a top school, joining a good company, building a great company, winning,
23:47
So that I think is part of the hunger and the desire that comes naturally from having quickly risen from a poor country and poor families into a chance to really change all that. And we've seen that in other cultures. In the U.S., it was perhaps in the early 20th century. Yeah.
24:07
And we've also seen that in Japan, Korea and other countries. So it's a it's a period of time where people see, wow, I can do better than the last 10 generations in my family. I got to grasp this opportunity and give it my all.
24:22
Yeah, and this definitely rings true for immigrants coming into the US or Canada from places like China, Korea, Japan, or anywhere else, India. And even with China, you guys have this thing called 996. Can you explain what 996 means for people that, or 997 rather, I guess, in many startups in China? Yeah.
24:45
Yeah, in the early phases, I think the entrepreneurs encouraged a hardworking environment because people were striving for success and willing to work hard. So 996 means working 9 a.m. to 9 p.m., six days a week. Although that is a little bit frowned upon in China now about not pushing people too hard.
25:10
Yeah, it's frowned upon. Yeah, I think the society is saying, wait a minute, this group of entrepreneurs are pushing people too hard. There have been sudden deaths, heart attacks. And I think there are some efforts, you know,
25:27
in the media and government messaging that pushing people this hard may not be good for their well-being. We'll see how it works out, but certainly it's an element that has driven the environment for the last 30 plus years, which after China opened up.
25:46
What are the things that they're doing to reduce these number of hours? Because 996 is like 9 a.m., 9 p.m., six days a week. But 997, which is seven days a week, was also very common back in the day. Are they adopting like 995 or 965 to make that like a cultural thing in China? Yeah.
26:06
Yeah, roughly, you know, for example, in some large internet companies that used to require working six days every other week is now going to five days every week. So that would be a change in this direction. Yeah. How do you feel about that? Do you feel like the younger generation is a little bit, you know, falling behind in that sense?
26:32
Well, I think millennials globally are a little bit similar. You know, Chinese, China's getting wealthier. So the middle, so the, so I think the middle class is growing and there are more people in the millennials born into a well-to-do families or middle class. And I think they want the same things that as other millennials. So I think, you know, as, so as I mentioned this 996 tenacity hunger for success would
27:01
would eventually fade away as the country becomes wealthier. Right, right. And I guess like nowadays, people can see what other people across the country and what their lifestyle is. And kind of there's this global culture, right? Obviously, everyone has his own independent culture. But with the information age, everyone is so connected that I would imagine with, yeah, people that are just learning to adapt to
27:27
you know, how other millennials are, are going through here in this case. Right. I think, yeah. What defines millennials throughout the world is that they have their own voice and own ways of doing things and they don't conform and they don't listen to a message and say, that's what I want to be too. And that that's individualism at a certain level. And that's,
27:51
And I think 20 years ago, I think people would say, hey, work hard and you'll succeed. So work incredibly hard and people just listen. But nowadays, I don't think people do that, not just in China, but perhaps globally.
28:08
Got it. Got it. And kind of getting back to the, you know, the differences, like one of the things that at least from what I've heard is that like engineering background is something that's emphasized for people that are working in the Chinese government. Is that right? Like there are a lot of people with engineering backgrounds that work in the government in China that support more adoption in terms of, you know, AI and advancements in technology. Yeah.
28:41
I'm not an expert on the demographics of government officials. Obviously, there have been a few top-level leaders that have had engineering degrees, so certainly a degree that's highly respected. If your question is more about why does the Chinese government emphasize tech so much, I think the number one reason is that we're in the fourth industrial revolution.
29:07
And China kind of missed out on the first and the second and a little bit of the third, right? The first was the steam engine. The UK's rise to the top. Second was electricity and other things. US rose to the top. Third, the US led the world again in PC and internet. Uh, China caught a little bit of that and, and, and got a good taste of it's good to have your control your own destiny. So now at the fourth industrial revolution, which is AI automation, um,
29:36
I think China feels like missing out the first three has caused China to pay a dear price of being, becoming, China did quite well having missed the three revolutions, right? Being the outsourcing factory of the world was not a bad outcome. It's the envy of the developing countries, of course, but still it became clear there would be the big winners would be those who are leading in technology and
30:05
And not only inventing them, but putting them to good use, first use. That I think is a very pervasive understanding in the Chinese government that we should not miss out and it's important. And that's why there are these policies basically promoting technologies and for kids to study science, et cetera.
30:29
Yeah, I wanted to dig into more about AI 2041 now at this point and really just go through some of the stories that you're telling in the book itself. I mean, we can talk about the high-level opportunities starting off. And one of the things that you're talking about is just the amount of wealth that it could potentially create. But I think this is a debate that people are having where
30:58
The opportunities that could come from AI is also going to be some of the jobs that are going to be replaced from that. So I want to dig into that a little bit. In terms of the wealth that it will create, what are some of the opportunities that you're seeing? Or is it going to be one of these things where it's going to be the top few that are able to leverage AI?
31:23
these things like AI or the people that are creating the companies leveraging AI that are going to be claiming the rewards of this? How do you look at that? AI is an optimizing technology that takes a lot of data and is capable of basically doing tasks at a better way than people because they can learn from data and from watching people, from being trained on data, etc.
31:47
So a lot of routine jobs and routine tasks will be replaced by AI, both at a white collar and a blue collar level. And that will lead to some job displacement as well as wealth inequality. So that's one of the topics throughout the book AI 2041, it keeps coming up. And I think people will eventually adapt, but we need to be aware of
32:12
how and why things are happening. So people, so, so AI will be used by top companies to create a lot of wealth and they will displace jobs. And then it'll take away jobs from people at the entry and routine levels who may have a hard time finding another job. So there needs to be mechanisms in society that, that, that,
32:37
bridges the gap of wealth inequality. For example, through taxation and universal basic income can give people a second chance to create a buffer for the people whose jobs may have been taken by AI. Secondly, there needs to be retraining in place to help people to find a new job that they can continue again, because it's not just important from a making a salary standpoint, but also
33:04
people's jobs are an important part of the meaning of their lives. And when the jobs are gone, people may become depressed, fall to substance abuse, et cetera.
33:16
And also education needs to be revamped to help people with help the kids learn skill sets that are not replaceable by AI, whether it's creativity or teamwork or compassion or human to human connections. So those things all need to be built up.
33:36
And I think we can get over these problems of job displacement and also even look forward to AI creating new jobs. Although we don't exactly know which they will be, but we do know they will not be routine jobs. So the blessing for us as a human race is that, uh,
33:55
Once we get over the hump of job displacements, wealth inequality adjustments, policies needed, et cetera, we can look forward to first AI creating a lot of wealth for the world, for society, because no longer do we need routine human labor to create products.
34:14
And secondly, we as a human race are then liberated from having to do routine work. So we can do things that we enjoy or we're good at. So it's a tough intermediary period. Once we get over it, we can look forward to a much better environment and world. Got it. So when you say wealth, you don't necessarily just mean the financial side of things. You're talking about the time, you're talking about
34:45
you know, allowing people to be more creative and having more access, like just overall those benefits that come from that process. Got it. And in terms of the new jobs that could potentially be created, like have you envisioned or thought about what are some of these things that could potentially come up as we get into 2041? Yeah.
35:13
Yes, and I think the book describes some of those possibilities. One, obviously, is in a world full of robots, we're going to need robot repair. We're going to need people who program the AI.
35:27
On the other side, with people having more time on their hands and with goods becoming produced by robots and becoming cheaper, people will want to spend their money more on services. So the services sector should undergo a substantial increase, both in number of employment and number of innovative services that people can have. Imagine if you make a
35:52
a lot of money from AI or technologies. And then you want to spend, you know, more, maybe a more curated vacation with your family or a concierge that gets you the food, the wine, the experience, the massage, whatever it is that you like, or a tour guide who gives you a personalized journey.
36:16
Or maybe for people who are in need, maybe elderly or people in hospitals, there can be volunteers and also paid people who are providing their time, taking care of the elderly, donating their time to
36:33
to foster homes. And there will be parents who want to spend, do homeschooling for their kids, both to help their kids to have the attention they need in an AI economy, but also that may be the best time spent for parents as opposed to getting a service job. So I think these things will likely blossom as well. There will be many other jobs that will be very hard to predict and
37:01
And the reason is that we can't exactly predict every new type of services. Just like when we began the internet revolution, I or any other expert could not possibly have 20 years ago predict the kind of jobs and economy that has come through, let's say, an app.
37:23
called named Uber, right? Uber has created many jobs, disrupted the way transportation works and created lots of opportunities and hurt some, right? Taxi drivers moving to Uber drivers, changing the way of life, et cetera. That could not have been predicted 20 years ago. So for the same reason, I can't really give a complete roadmap
37:47
of what new jobs will be enabled by AI, but there should be many. As history would tell us, every technological revolution has eventually provided more jobs than it decimated. And the same thing, I believe, will be true for AI, but we have to probably wait 15, 20, 25 years, just as we did for the internet, for let's say using Uber again as an example to see jobs blossom in this area.
38:14
Yeah, I do agree with it in terms of these technologies. It's always a double-edged sword, right? So the older people that can't really adapt to, let's say the taxi medallions are going to be displaced and Uber opens up more jobs, but you're also getting into a situation now where Uber drivers are also going to be replaced now in the next probably five years or 10 years, which means there's going to be, it almost seems like as AI gets more
38:44
advanced, you are going to create more jobs. But if it gets so advanced, those jobs that the AI created at that level is also going to be replaced again as the AI gets smarter and smarter. So is it always just going to be this cycle where AI can potentially create new wealth, but it ends up replacing it and that opens up another layer that we just can't foresee right now?
39:08
Yes, I think so. That's why in one of the stories in the book, we created a new type of company called job reallocation companies. And what they will do is on an individual basis, if your job were displaced by AI, it would find something that's individually more suitable for you to gain the retraining and take that on.
39:30
And there are characters in the story who have been job reallocated two, three, four times in two decades. And we should expect that to happen. But if society and the government and companies will provide that cushion and retrain people, we will stop looking at a job as something you do for life, but as a new experience, new skills that you learn and reallocate
39:58
As long as the danger of unemployment and the loss of subsistence is removed by social welfare, then I think people might look forward to the retraining as opposed to being painted into a corner. Yeah. And how does that look for the way society is going to be made up? Are we just going to have a massive middle class, particularly with things like UBI and taxation systems where
40:27
people that aren't able to find jobs or make a lot of income are able to survive with these basic necessities of being able to pay for their stuff. Is that just going to lead to like a one big middle class and then maybe like a percentage of people that are just insanely wealthy, meaning like there's just not going to be a lower class anymore?
40:50
Well, this is something we can all speculate, but I believe one of the likely outcomes would be, as you described, that is larger numbers of people who are no longer fearful of losing their jobs and they get the retraining they need. And then they rotate jobs as society technologies change.
41:13
And also, in addition, I think some number of people may choose to go into a type of work that isn't just generating economic value, but generates societal value, health value.
41:29
going into healthcare services, working in elderly home, homeschooling your children would all fall into that category. So that would somehow fit. Of course, the society needs to accept these jobs and these people who are, I think, adding value to the society and not be looked down upon, nor to be economically discriminated. That is, they need to earn a good
41:56
living doing these things so that requires some adjustment in the society and then i think the ultra rich there the question needs to be asked does there really need to be some form of equalizing taxation that's been proposed in the u.s by elizabeth warren and others uh the rich tax basically uh yeah what are your thoughts on that
42:22
Obviously, you're a very rich man, of course. Well, obviously, I think the rich people will fight back against that. And I think if there were a better way, maybe we can find it. But if not, that might be a last resort that we need to accept. Because what's the point of having one person possess $1 trillion? Yeah.
42:49
They can never use it up. It's not a economic or from a whole society standpoint, it doesn't help the society. So putting some taxation for the ultra-rich corporations and people seems like an approach that we, the human race, understand how to put forward, regulate, regulate.
43:15
If there were clever ways, great. I mean, people talk about UBI, but UBI needs to be funded somehow. Unless you have this ultra-rich taxation, I don't see how UBI can be funded. So it seems to me it's probably unavoidable because the likelihood for the human race to invent another mechanism to rebalance society
43:37
The wealth inequality seems so unlikely. People have talked about death and taxation are the two unavoidable things in life. And taxation has a lot of, we have a lot of practice throughout the years doing taxation. So time to just use that from our arsenal as a necessary step.
43:59
Yeah. There must be something that is in between where if someone that is at that top 0.1% can deploy the capital in an efficient way that gives back some sort of value to society, which is hard to quantify or hard to classify, I guess. I guess one example is like Jeff Bezos, obviously he gets a lot of knack for being the wealthiest person in the world, but he's
44:23
There is something to be said, I feel, that the person that ended up, unless it was nepotism, but someone that actually ended up making their own wealth and was self-made is somewhat deservant of being able to deploy that capital in the most effective way for society. Jeff Bezos can do what he does now, which is with Blue Origin.
44:50
because he has the capabilities and the reason why he got to the billions of dollars that he has is because he just has those capabilities, the work ethic, the knowledge, the experience to do that. And I think there is some sense in terms of someone that has that amount of capital, as long as they're deploying it into something that can push society forward.
45:13
deserves to keep that capital and continue to deploy it because that person can make better decisions in some sense than the average person, certainly better than me. So yeah, I don't know. It seems like it's a hard thing to quantify, but
45:29
I feel like there's an in-between there rather than just taxating for the sake of taxating, you know? Yeah, like the Gates Foundation is another example. That's right. Yeah, with the dedication and the research they do, deploying the money in the most efficient way. It's the money that you made, so you care more about it than the government. And there's something just to be said about that. But of course, if you start creating all kinds of different ways you can quote-unquote contribute to society without being taxed,
45:58
then the tax, the loopholes will emerge as well. So you have to balance these. Yeah. Yeah. And just in terms of the job displacement, I mean, I think what's inevitable is that AI will accelerate continuously and kind of as brash as it sounds, like the people, like when I have kids, those kids will just grow up in a generation that, you know,
46:21
is built around learning these skill sets that they need to thrive in this new world. And there's just going to be this temporary step that we have to go through where the existing generations that grew up without AI are going to have a little bit of trouble, but like the world's going to move on, right? Eventually. And our kids and our grandchildren, they're just going to be able to thrive in this world, which is,
46:48
Yeah. I mean, for this generation, it's scary because you're saying in 15 years, 40 to 50% of jobs will be replaced by AI. And you kind of have these steps and processes. Can you actually break down the four quadrant that you have in terms of the repetitive jobs and the ones that are going to be non-creative and which types of jobs can't be replaced? Yeah.
47:16
Sure. In both of my books, I used two dimensions to create four quadrants. And basically the dimensions are on the horizontal axis is the level of creativity. So the more creative, the more human-oriented,
47:31
involvement is required and the less creative would be more routine work where ai can take over more quickly then on the y-axis would be the trust compassion empathy human to human connection element higher up would be jobs that require a lot of that human connection lower would be the ones that you can do isolated and and clearly in these quadrant on the lower left
47:58
would be the low compassion, low creativity that's likely to be all replaced by AI. And that's a good percentage of the tasks that we do on a day-to-day basis.
48:08
But on the high compassion, lower creativity level, that's where the service jobs will emerge. The healthcare services, tens of millions of jobs like that. Also some jobs like teachers and doctors will become more human connection and maybe less using AI as a way of doing things.
48:32
teaching or diagnosis on that front. So jobs will be shifting from the routine work up to a lot of it will be up to the service work because it's gonna be hard to teach a routine worker to become creative or to become a CEO or a scientist, but probably becoming someone in the service industry seems more feasible.
48:54
And then on the lower right are the high creativity, but jobs that don't require a lot of compassion and human connection, jobs like scientists, researchers, et cetera. So that will continue to be human occupied, but perhaps using more AI tools in a symbiotic way. And then finally, the high compassion, high creativity.
49:17
And the high creativity jobs, maybe a CEO's job or someone running a country requires both of that to do a good job. Then those will also be using AI tools, but that's where the human unique qualities of creativity and compassion will come together and shine.
49:37
Got it. Got it. So for someone that is about to enter high school or university and their parents that are listening to this to figure out what they should, what are some of the skill sets that they should be teaching their children? Is the focus going to be more on soft skills and that things like, I don't know, creative arts or philosophy or those things are going to be
50:00
Or psychology, those things are going to be the ones that might be more relevant for adapting into this new world than, you know, kind of the hard skills that maybe were taught in university traditional system.
50:15
Yeah, directionally, what you said is basically right. But I think even more importantly, there will need to be new courses and a new learning process to gain the skill sets that really need to be gained. For example, how do you make someone more creative? How do you hone critical thinking skills and increase curiosity?
50:38
And how do you increase the level of why and why not in the classroom and less about what's the answer? So that shift needs a whole new design of education. Similarly, the soft skills, compassion, empathy, ability to communicate, persuade, and work in a team, and win people's respect and trust,
51:02
That has always been, I think more than 50% of success in any corporate world anyway, but the classrooms haven't been good at teaching those. So how do we pivot the education process? And if it's not done in a public school, maybe parents will want to provide that. So I think that's the most important thing for parents is to make sure that the kids can learn new skills, become creative,
51:30
and have a strong ability to communicate and empathize with other people. Also, I would say it's incredibly important for the kids to follow their heart, to follow their passion, because you're not just competing with other people, but AI that is becoming increasingly good. So if you don't do what you love, it's hard to become the best in what you do.
51:56
So those are the suggestions we give. And in the book, AI 2041, we talked about how AI tools can be used to enhance education. One example is in the story Twin Sparrows, when AI companions are working with each kid, trying to make learning fun and trying to watch the kid from
52:20
all day long and then gain enough understanding so that it knows how to motivate the child, what they like, and making learning engaging and fun in an individualistic kind of way. So that kind of paints a possible technology future for education. But of course, parents need to also do what only a parent and a human can do to help the child
52:49
grow in this particularly challenging age, but also full of technology infusion and opportunities. Do you see a world where robots and humans can fall in love with each other? It's quite likely that there will be humans who fall in love with robots that are maybe software-based.
53:12
Maybe chatbot at first, then conversational, because AI is just getting better and better at it. It fakes emotion, connection, and understanding. And then eventually, probably beyond 20 years, there can be robots that have those capabilities too.
53:28
But I would remind people that if you're thinking you might be falling in love with a robot, remember that it doesn't love you back. It is merely parroting phrases back that it thinks will gain your attention and love.
53:45
and it doesn't have true feelings. So I hope this doesn't become any kind of mainstream phenomenon. There will always be people who are lonely, who feel they found someone who understands, and robots will be better and better at it. But I think that is against our human capacity
54:05
culture and intuition and preference that we should love other human beings and not robots. And this will remain in a very small fraction of the society. Yeah, it's worth debating, I think, because I mean, loneliness is such a pandemic now that
54:26
it might be worth just letting people fall in love with robots and despite the robot perhaps not liking it, but it still might serve their needs in some sense. And, you know, what if there was data to point that people do feel less lonely with robots? I mean, that could be a business, separate business on its own, where you have your own companion in some sense, or even just a friend or, you
54:53
you know, he's a caregiver in this sense. Yeah, I wonder if that's something that could potentially come into. Is that a story in the book at all? Like someone falling in love with the robot?
55:07
No, no, not something I believe that should happen. So we didn't include in this story. But there were companions. There were companions. There were AI keeping people company, AI companion teacher for kids. So there are some of that.
55:24
And there are also research today that shows AI can be good in things like even a suicide hotline or people who have psychological problems but could not reach a human psychiatrist for help in time. That robot provides a reasonable backup, but only when the human is not available. And it seems to show that it's more effective than not having it.
55:53
Got it. Got it. I just want to end with some questions here, Kaifu. We talked a bit about work-life balance and, you know, just kind of segmenting this with the leverage that people will have with automation and AI, you know, in that world of leverage and automation is important.
56:13
Is working smarter and making better decisions going to be a more important factor for people to have more success in their careers and just their overall success versus just simply working hard? Yeah.
56:32
I think it's a combination of working hard and working smart. I certainly believe in working smart, doing things in a non-road learning kind of way and developing new skills and spending one's time in a balanced way. This is all good, but sometimes, you know, don't work hard, work smart, become something people say as an excuse not to work hard. And, and I think, um,
57:00
in an exciting environment, one should find something people, one is excited about that's worth working hard and working smart, not to hurt your health, but to do things that you love and that you're self-motivated to be thinking about it all the time and to, so that you can become more creative and become more knowledgeable.
57:29
So I think finding something you're passionate about is something I believe in. I also believe if you found something you're passionate about, you're likely to want to work hard, not as a result of people telling you or being forced to compete and survive, but out of your own volition to want to do that. That I think is a place that if you found yourself to be in that kind of work hard and smart, but not…
57:56
but not hurt your health kind of a combination. That would be, I think the best combination if, if one could find that thing that you're passionate about. Yeah. So let's say someone has found something they're passionate about. If you were to give advice to your 30 year old self, looking back where you're, you've clearly found your passion and the things that you want to do, what's like a piece of advice that you would share?
58:23
I would tell my 30 year old version of myself to take more risks because risks should be taken when one's young. And once you're older, it's, um, you don't have as much ahead of you. The cost you pay could be higher, um, and you become more set in your ways. And, and I think society in many ways teach more of us to not take risks. Um,
58:48
Maybe because the Maslow hierarchy does have survival at a very basic level. But if you don't take your risks when you're young, you may not have a chance to do that. And that will be regret when you're older.
59:04
And do you think like the definition of young and older is changing? I know you had a pretty hard health scare with fourth stage lymphoma. I don't know how much, how deeper you got into longevity. Like we've had a Sergey young on the show. I'm not sure if you're familiar with him. He's got the longevity fund and talks a little bit about longevity as well. And this idea that we can live to 200 years old. Like, is that, is that an industry apart from AI that's,
59:33
been interesting for you, particularly with the events that have happened recently? Sure. So my last question was referring to 30-year-old asking a 60-year-old, which I think is what we just did. But I also believe in longevity. The two are not mutually exclusive. I think new advances in life science and also using AI can be well combined with longevity.
01:00:01
We have invested in a company called Deep Longevity that uses a tool that will take all of your body metrics that includes primarily your blood right now, but also potentially your full body MRI, your multi-omics inputs,
01:00:16
and your wearables input into an AI system that compares you with other people of your own age, younger age, healthier, not so healthy, and give you some feedback on how you can improve your lifestyle, which includes certainly new types of medicines and nutrients, but also your stress management, exercise, sleep,
01:00:42
and the nutrition as an overall combination to get your metrics to be healthier. I don't think we, the human race, have ever had a complete data-centric way to manage ourselves.
01:00:57
If we believe data-centric ways are better ways to manage companies, better ways to make internet companies, financial companies more successful, then we should use it on us. So I am engaged in using this technology right now. I do deep measures every quarter,
01:01:18
And then some of them every year, and then I can see progress. So as someone who's, um, data driven and who's, uh, very competitive, uh, it becomes kind of a game like for me, leaderboard or something. Uh,
01:01:33
just competing with myself. Yes. It's private personal data. So I can't see anybody else's data, but it would show, you know, your blood looks, you know, six years younger than it did one year ago, which was the number I got. And I feel like, you know, I've won the game or something. Congrats. Yeah. That's amazing. Well, a big step in terms of,
01:01:53
being healthy and longevity is prevention. So something like what you're mentioning is definitely a step forward for society. So, well, Kai-Fu, I really appreciate your time and all the insights that you shared. I would highly recommend people to check out AI 2041. Where else can people learn more about you, what you're up to? And are you working on the next book now?
01:02:20
Oh no, I'm still getting over the promotion for this one, but there might be another one who knows AI 2061 or something like that one day. In 20 years from now, right? Right, yeah. Love it, love it. Well, really again, appreciate your time. Highly recommend people to check it out and thanks so much for tuning in guys.
</markdown>
D:2025.06.04
<markdown> </markdown>