Transcript
00:00:00 The following is a conversation with Kai Fu Lee.
00:00:02 He’s the chairman and CEO of Cinovation Ventures
00:00:06 that manages a $2 billion dual currency investment fund
00:00:10 with a focus on developing the next generation
00:00:13 of Chinese high tech companies.
00:00:15 He’s the former president of Google China
00:00:17 and the founder of what is now called
00:00:19 Microsoft Research Asia,
00:00:21 an institute that trained many of the artificial
00:00:24 intelligence leaders in China,
00:00:26 including CTOs or AI execs at Baidu,
00:00:30 Tencent, Alibaba, Lenovo, and Huawei.
00:00:34 He was named one of the 100 most influential people
00:00:38 in the world by Time Magazine.
00:00:40 He’s the author of seven bestselling books in Chinese
00:00:43 and most recently, the New York Times bestseller
00:00:46 called AI Superpowers, China, Silicon Valley,
00:00:50 and the New World Order.
00:00:52 He has unparalleled experience
00:00:55 in working across major tech companies
00:00:57 and governments and applications of AI,
00:01:00 and so he has a unique perspective on global innovation
00:01:03 and the future of AI that I think is important
00:01:06 to listen to and think about.
00:01:08 This is the Artificial Intelligence Podcast.
00:01:11 If you enjoy it, subscribe on YouTube and iTunes,
00:01:15 support it on Patreon, or simply connect with me
00:01:17 on Twitter at Lex Friedman.
00:01:20 And now, here’s my conversation with Kaifu Li.
00:01:26 I immigrated from Russia to US when I was 13.
00:01:29 You immigrated to US at about the same age.
00:01:32 The Russian people, the American people,
00:01:34 the Chinese people each have a certain soul,
00:01:38 a spirit that permeates throughout the generations.
00:01:42 So maybe it’s a little bit of a poetic question,
00:01:45 but could you describe your sense
00:01:48 of what defines the Chinese soul?
00:01:52 I think the Chinese soul of people today, right,
00:01:56 we’re talking about people who have had centuries of burden
00:02:01 because of the poverty that the country has gone through,
00:02:05 and suddenly shined with hope of prosperity
00:02:10 in the past 40 years as China opened up
00:02:13 and embraced market economy.
00:02:15 And undoubtedly, there are two sets of pressures
00:02:20 on the people, that of the tradition,
00:02:24 that of facing difficult situations,
00:02:28 and that of hope of wanting to be the first
00:02:31 to become successful and wealthy.
00:02:33 So that’s a very strong hunger and a strong desire
00:02:38 and strong work ethic that drives China forward.
00:02:41 And is there roots to not just this generation,
00:02:43 but before that’s deeper
00:02:47 than just the new economic developments?
00:02:50 Is there something that’s unique to China
00:02:52 that you could speak to that’s in the people?
00:02:54 Yeah, well, the Chinese tradition
00:02:58 is about excellence, dedication, and results.
00:03:02 And the Chinese exams and study subjects in schools
00:03:07 have traditionally started
00:03:09 from memorizing 10,000 characters,
00:03:11 not an easy task to start with.
00:03:13 And further by memorizing
00:03:15 his historic philosopher’s literature poetry.
00:03:18 So it really is probably
00:03:21 the strongest rote learning mechanism created
00:03:25 to make sure people had good memory
00:03:26 and remember things extremely well.
00:03:30 That’s, I think at the same time,
00:03:32 suppresses the breakthrough innovation
00:03:37 and also enhances the speed execution get results.
00:03:42 And that I think characterizes
00:03:44 the historic basis of China.
00:03:47 That’s interesting,
00:03:48 because there’s echoes of that in Russian education
00:03:50 as well as rote memorization.
00:03:51 So you have to memorize a lot of poetry.
00:03:53 I mean, there’s just an emphasis on perfection in all forms
00:03:59 that’s not conducive to perhaps what you’re speaking to,
00:04:02 which is creativity.
00:04:03 But you think that kind of education holds back
00:04:06 the innovative spirit that you might see
00:04:09 in the United States?
00:04:10 Well, it holds back the breakthrough innovative spirits
00:04:14 that we see in the United States,
00:04:16 but it does not hold back the valuable execution oriented,
00:04:21 result oriented value creating engines,
00:04:25 which we see China being very successful.
00:04:27 So is there a difference between a Chinese AI engineer today
00:04:32 and an American AI engineer,
00:04:34 perhaps rooted in the culture that we just talked about
00:04:36 or the education or the very soul of the people or no?
00:04:41 And what would your advice be to each
00:04:43 if there’s a difference?
00:04:45 Well, there’s a lot that’s similar
00:04:47 because AI is about mastering sciences,
00:04:51 about using known technologies and trying new things,
00:04:54 but it’s also about picking from many parts
00:04:58 of possible networks to use
00:05:00 and different types of parameters to tune.
00:05:02 And that part is somewhat rote.
00:05:05 And it is also, as anyone who’s built AI products
00:05:08 can tell you a lot about cleansing the data
00:05:12 because AI runs better with more data
00:05:15 and data is generally unstructured,
00:05:18 errorful and unclean.
00:05:22 And the effort to clean the data is immense.
00:05:26 So I think the better part of American engineering,
00:05:31 AI engineering process is to try new things,
00:05:35 to do things people haven’t done before
00:05:37 and to use technology to solve most if not all problems.
00:05:43 So to make the algorithm work despite not so great data,
00:05:47 find error tolerant ways to deal with the data.
00:05:50 The Chinese way would be to basically enumerate
00:05:55 to the fullest extent all the possible ways
00:05:58 by a lot of machines, try lots of different ways
00:06:01 to get it to work and spend a lot of resources
00:06:04 and money and time cleaning up data.
00:06:07 That means the AI engineer may be writing
00:06:10 data cleansing algorithms, working with thousands of people
00:06:15 who label or correct or do things with the data.
00:06:19 That is the incredible hard work
00:06:21 that might lead to better results.
00:06:23 So the Chinese engineer would rely on
00:06:26 and ask for more and more and more data
00:06:29 and find ways to cleanse them and make them work
00:06:31 in the system and probably less time thinking
00:06:34 about new algorithms that can overcome data
00:06:38 or other issues.
00:06:39 So where’s your intuition?
00:06:40 Where do you think the biggest impact
00:06:42 in the next 10 years lies?
00:06:44 Is it in some breakthrough algorithms
00:06:47 or is it in just this at scale rigor,
00:06:53 a rigorous approach to data, cleaning data,
00:06:55 organizing data onto the same algorithms?
00:06:58 What do you think the big impact in the applied world is?
00:07:02 Well, if you’re really in the company
00:07:04 and you have to deliver results,
00:07:06 using known techniques and enhancing data
00:07:09 seems like the more expedient approach
00:07:12 that’s very low risk and likely to generate
00:07:15 better and better results.
00:07:17 And that’s why the Chinese approach has done quite well.
00:07:20 Now, there are a lot of more challenging startups
00:07:24 and problems such as autonomous vehicles,
00:07:28 medical diagnosis that existing algorithms
00:07:32 probably won’t solve.
00:07:34 And that would put the Chinese approach more challenged
00:07:38 and give them more breakthrough innovation approach,
00:07:42 more of an edge on those kinds of problems.
00:07:45 So let me talk to that a little more.
00:07:47 So my intuition personally is that data
00:07:51 can take us extremely far.
00:07:53 So you brought up autonomous vehicles and medical diagnosis.
00:07:56 So your intuition is that huge amounts of data
00:08:00 might not be able to completely help us solve that problem.
00:08:04 Right, so breaking that down further in autonomous vehicle,
00:08:08 I think huge amounts of data probably will solve
00:08:11 trucks driving on highways,
00:08:13 which will deliver a significant value
00:08:15 and China will probably lead in that.
00:08:18 And full L5 autonomous is likely to require
00:08:23 new technologies we don’t yet know.
00:08:26 And that might require academia
00:08:28 and great industrial research,
00:08:30 both innovating and working together.
00:08:32 And in that case, US has an advantage.
00:08:35 So the interesting question there is,
00:08:37 I don’t know if you’re familiar
00:08:38 on the autonomous vehicle space
00:08:39 and the developments with Tesla and Elon Musk.
00:08:42 I am.
00:08:43 Where they are in fact full steam ahead
00:08:49 into this mysterious complex world of full autonomy, L5,
00:08:53 L4, L5, and they’re trying to solve that purely with data.
00:08:58 So the same kind of thing that you’re saying
00:09:00 is just for highway,
00:09:02 which is what a lot of people share your intuition.
00:09:05 They’re trying to solve with data.
00:09:07 So just to linger on that moment further,
00:09:09 do you think possible for them to achieve success
00:09:13 with simply just a huge amount of this training
00:09:17 on edge cases and difficult cases in urban environments,
00:09:20 not just highway and so on?
00:09:22 I think it would be very hard.
00:09:24 One could characterize Tesla’s approach
00:09:27 as kind of a Chinese strength approach, right?
00:09:29 Gather all the data you can
00:09:31 and hope that will overcome the problems.
00:09:34 But in autonomous driving,
00:09:36 clearly a lot of the decisions aren’t merely solved
00:09:40 by aggregating data and having feedback loop.
00:09:43 There are things that are more akin to human thinking.
00:09:48 And how would those be integrated and built?
00:09:51 There has not yet been a lot of success
00:09:53 integrating human intelligence
00:09:56 or call it expert systems if you will,
00:09:58 even though that’s a taboo word with the machine learning.
00:10:02 And the integration of the two types of thinking
00:10:05 hasn’t yet been demonstrated.
00:10:07 And the question is how much can you push
00:10:09 a purely machine learning approach?
00:10:12 And of course, Tesla also has an additional constraint
00:10:15 that they don’t have all the sensors.
00:10:18 I know that they think it’s foolish to use LIDARs,
00:10:21 but that’s clearly a one less very valuable
00:10:25 and reliable source of input that they’re foregoing,
00:10:28 which may also have consequences.
00:10:32 I think the advantage of course is capturing data
00:10:34 that no one has ever seen before.
00:10:37 And in some cases such as computer vision
00:10:40 and speech recognition,
00:10:42 I have seen Chinese companies accumulate data
00:10:44 that’s not seen anywhere in the Western world
00:10:47 and they have delivered superior results.
00:10:50 But then speech recognition and object recognition
00:10:53 are relatively suitable problems for deep learning
00:10:57 and don’t have the potentially need
00:11:00 for the human intelligence analytical planning elements.
00:11:04 And the same on the speech recognition side,
00:11:06 your intuition that speech recognition
00:11:08 and the machine learning approaches to speech recognition
00:11:11 won’t take us to a conversational system
00:11:14 that can pass the Turing test,
00:11:15 which is sort of maybe akin to what driving is.
00:11:20 So it needs to have something more than just simply
00:11:24 simple language understanding, simple language generation.
00:11:27 Roughly right.
00:11:28 I would say that based on purely machine learning approaches,
00:11:33 it’s hard to imagine it could lead
00:11:35 to a full conversational experience
00:11:40 across arbitrary domains, which is akin to L5.
00:11:44 I’m a little hesitant to use the word Turing test
00:11:46 because the original definition was probably too easy.
00:11:50 We probably do that, yeah.
00:11:52 The spirit of the Turing test
00:11:54 is what I was referring to. Of course.
00:11:56 So you’ve had major leadership research positions
00:11:59 at Apple, Microsoft, Google.
00:12:01 So continuing on the discussion of America, Russia,
00:12:05 Chinese, Seoul and culture and so on.
00:12:09 What is the culture of Silicon Valley
00:12:12 in contrast to China and maybe US broadly?
00:12:16 And what is the unique culture
00:12:18 of each of these three major companies in your view?
00:12:22 I think in aggregate, Silicon Valley companies,
00:12:25 and we could probably include Microsoft in that,
00:12:27 even though they’re not in the Valley,
00:12:29 is really dream big and have visionary goals
00:12:33 and believe that technology will conquer all.
00:12:37 And also the self confidence and the self entitlement
00:12:42 that whatever they produce,
00:12:43 the whole world should use and must use.
00:12:47 And those are historically important, I think.
00:12:54 Steve Jobs famous quote that he doesn’t do focus groups,
00:12:59 he looks in the mirror and asks the person in the mirror,
00:13:02 what do you want?
00:13:03 And that really is an inspirational comment that says,
00:13:07 the great company shouldn’t just ask users what they want,
00:13:11 but develop something that users will know they want
00:13:15 when they see it,
00:13:16 but they could never come up with themselves.
00:13:18 I think that is probably the most exhilarating description
00:13:23 of what the essence of Silicon Valley is,
00:13:26 that this brilliant idea could cause you to build something
00:13:31 that couldn’t come out of the focus groups or AB tests.
00:13:35 And iPhone would be an example of that.
00:13:38 No one in the age of Blackberry would write down
00:13:40 they want an iPhone or multi touch.
00:13:42 A browser might be another example.
00:13:44 No one would say they want that in the days of FTP,
00:13:47 but once they see it, they want it.
00:13:49 So I think that is what Silicon Valley is best at.
00:13:53 But it also comes with, it came with a lot of success.
00:13:58 These products became global platforms
00:14:01 and there were basically no competitors anywhere.
00:14:05 And that has also led to a belief
00:14:08 that these are the only things that one should do,
00:14:13 that companies should not tread on other companies territory
00:14:17 so that a Groupon and a Yelp and then OpenTable
00:14:24 and the Grubhub would each feel,
00:14:26 okay, I’m not gonna do the other company’s business
00:14:28 because that would not be the pride of innovating
00:14:33 what each of these four companies have innovated.
00:14:36 But I think the Chinese approach
00:14:40 is do whatever it takes to win.
00:14:42 And it’s a winner take all market.
00:14:45 And in fact, in the internet space,
00:14:47 the market leader will get predominantly all the value
00:14:50 extracted out of the system.
00:14:53 So, and the system isn’t just defined
00:14:57 as one narrow category, but gets broader and broader.
00:15:01 So it’s amazing ambition for success and domination
00:15:07 of increasingly larger product categories
00:15:11 leading to clear market winner status
00:15:15 and the opportunity to extract tremendous value.
00:15:19 And that develops a practical, result oriented,
00:15:25 ultra ambitious winner take all gladiatorial mentality.
00:15:31 And if what it takes is to build what the competitors built,
00:15:37 essentially a copycat that can be done
00:15:40 without infringing laws.
00:15:41 If what it takes is to satisfy a foreign country’s need
00:15:46 by forking the code base and building something
00:15:48 that looks really ugly and different, they’ll do it.
00:15:51 So it’s contrasted very sharply
00:15:54 with the Silicon Valley approach.
00:15:56 And I think the flexibility and the speed and execution
00:16:00 has helped the Chinese approach.
00:16:01 And I think the Silicon Valley approach
00:16:05 is potentially challenged if every Chinese entrepreneur
00:16:10 is learning from the whole world, US and China,
00:16:13 and the American entrepreneurs only look internally
00:16:16 and write off China as a copycat.
00:16:19 And the second part of your question
00:16:21 about the three companies.
00:16:23 The unique elements of the three companies perhaps.
00:16:26 Yeah, I think Apple represents
00:16:30 while the user please the user
00:16:33 and the essence of design and brand
00:16:38 and it’s the one company and perhaps the only tech company
00:16:43 that draws people with a strong, serious desire
00:16:49 for the product and the willingness to pay a premium
00:16:53 because of the halo effect of the brand
00:16:56 which came from the attention to detail
00:16:59 and great respect for user needs.
00:17:01 Microsoft represents a platform approach
00:17:07 that builds giant products that become very strong modes
00:17:12 that others can’t do because it’s well architected
00:17:17 at the bottom level and the work is efficiently delegated
00:17:23 to individuals and then the whole product is built
00:17:28 by adding small parts that sum together.
00:17:32 So it’s probably the most effective high tech assembly line
00:17:36 that builds a very difficult product
00:17:38 that and the whole process of doing that
00:17:43 is kind of a differentiation and something competitors
00:17:49 can’t easily repeat.
00:17:50 Are there elements of the Chinese approach
00:17:53 in the way Microsoft went about assembling
00:17:56 those little pieces and dominating,
00:17:59 essentially dominating the market for a long time
00:18:02 or do you see those as distinct?
00:18:04 I think there are elements that are the same.
00:18:06 I think the three American companies that had
00:18:09 or have Chinese characteristics and obviously
00:18:12 as well as American characteristics are Microsoft,
00:18:16 Facebook and Amazon.
00:18:18 Yes, that’s right, Amazon.
00:18:20 Because these are companies that will tenaciously
00:18:23 go after adjacent markets,
00:18:27 build up strong product offering and find ways
00:18:34 to extract greater value from a sphere
00:18:37 that’s ever increasing and they understand
00:18:41 the value of the platforms.
00:18:43 So that’s the similarity and then with Google,
00:18:47 I think it’s a genuinely value oriented company
00:18:54 that does have a heart and soul
00:18:57 and that wants to do great things for the world
00:18:59 by connecting information
00:19:01 and that has also very strong technology genes
00:19:08 and wants to use technology
00:19:13 and has found out of the box ways to use technology
00:19:19 to deliver incredible value to the end user.
00:19:23 If you can look at Google, for example,
00:19:25 you mentioned heart and soul.
00:19:28 There seems to be an element where Google
00:19:31 is after making the world better.
00:19:34 There’s a more positive view.
00:19:36 They used to have the slogan, don’t be evil.
00:19:39 And Facebook a little bit more has a negative tend to it.
00:19:43 At least in the perception of privacy and so on.
00:19:45 Do you have a sense of how these different companies
00:19:51 can achieve, because you’ve talked about
00:19:53 how much we can make the world better
00:19:54 in all these kinds of ways with AI.
00:19:56 What is it about a company that can make,
00:19:59 give it a heart and soul, gain the trust of the public
00:20:03 and just actually just not be evil
00:20:06 and do good for the world?
00:20:08 It’s really hard and I think Google
00:20:10 has struggled with that.
00:20:13 First, the don’t do evil mantra is very dangerous
00:20:16 because every employee’s definition of evil is different.
00:20:20 And that has led to some difficult
00:20:22 employee situations for them.
00:20:25 So I don’t necessarily think that’s a good value statement,
00:20:29 but just watching the kinds of things Google
00:20:32 or its parent company Alphabet does
00:20:35 in new areas like healthcare, like eradicating mosquitoes,
00:20:40 things that are really not in the business
00:20:42 of a internet tech company.
00:20:44 I think that shows that there’s a heart and soul
00:20:47 and desire to do good and willingness
00:20:50 to put in the resources to do something
00:20:54 when they see it’s good, they will pursue it.
00:20:58 That doesn’t necessarily mean
00:21:00 it has all the trust of the users.
00:21:02 I realize while most people would view Facebook
00:21:06 as the primary target of their recent unhappiness
00:21:09 about Silicon Valley companies,
00:21:11 many would put Google in that category.
00:21:14 And some have named Google’s business practices
00:21:16 as predatory also.
00:21:19 So it’s kind of difficult to have the two parts of a body.
00:21:24 The brain wants to do what it’s supposed to do
00:21:27 for a shareholder, maximize profit.
00:21:29 And then the heart and soul wants to do good things
00:21:32 that may run against what the brain wants to do.
00:21:36 So in this complex balancing
00:21:39 that these companies have to do,
00:21:40 you’ve mentioned that you’re concerned
00:21:42 about a future where too few companies
00:21:45 like Google, Facebook, Amazon are controlling our data
00:21:49 or controlling too much of our digital lives.
00:21:53 Can you elaborate on this concern
00:21:54 and perhaps do you have a better way forward?
00:21:57 I think I’m hardly the most vocal complainer of this.
00:22:02 Sure, of course.
00:22:03 There are a lot louder complainers out there.
00:22:06 I do observe that having a lot of data
00:22:10 does perpetuate their strength
00:22:13 and limits competition in many spaces.
00:22:18 But I also believe AI is much broader
00:22:21 than the internet space.
00:22:23 So the entrepreneurial opportunities still exists
00:22:26 in using AI to empower financial,
00:22:30 retail, manufacturing, education applications.
00:22:34 So I don’t think it’s quite a case
00:22:36 of full monopolistic dominance
00:22:38 that totally stifles innovation.
00:22:43 But I do believe in their areas of strength
00:22:45 it’s hard to dislodge them.
00:22:48 I don’t know if I have a good solution.
00:22:52 Probably the best solution is let
00:22:54 the entrepreneurial VC ecosystem work well
00:22:58 and find all the places that can create the next Google,
00:23:02 the next Facebook.
00:23:03 So there will always be increasing number of challengers.
00:23:07 In some sense that has happened a little bit.
00:23:10 You see Uber, Airbnb having emerged
00:23:14 despite the strength of the big three.
00:23:19 And I think China as an environment
00:23:21 may be more interesting for the emergence
00:23:24 because if you look at companies
00:23:26 between let’s say 50 to $300 billion,
00:23:32 China has emerged more of such companies
00:23:35 than the US in the last three to four years
00:23:39 because of the larger marketplace,
00:23:41 because of the more fearless nature of the entrepreneurs.
00:23:46 And the Chinese giants are just as powerful
00:23:48 as American ones.
00:23:50 Tencent, Alibaba are very strong,
00:23:52 but ByteDance has emerged worth 75 billion
00:23:56 and financial while it’s Alibaba affiliated,
00:23:59 it’s nevertheless independent and worth 150 billion.
00:24:03 And so I do think if we start to extend
00:24:07 to traditional businesses,
00:24:09 we will see very valuable companies.
00:24:12 So it’s probably not the case that in five or 10 years
00:24:17 we’ll still see the whole world
00:24:19 with these five companies having such dominance.
00:24:22 So you’ve mentioned a couple of times
00:24:26 this fascinating world of entrepreneurship in China
00:24:29 of the fearless nature of the entrepreneur.
00:24:31 So can you maybe talk a little bit about
00:24:33 what it takes to be an entrepreneur in China?
00:24:35 What are the strategies that are undertaken?
00:24:38 What are the ways to achieve success?
00:24:41 What is the dynamic of VCF funding
00:24:43 of the way the government helps companies and so on?
00:24:46 What are the interesting aspects here
00:24:47 that are distinct from, that are different
00:24:50 from the Silicon Valley world of entrepreneurship?
00:24:55 Well, many of the listeners probably still
00:24:59 would brand Chinese entrepreneur as copycats.
00:25:03 And no doubt 10 years ago,
00:25:05 that would not be an inaccurate description.
00:25:09 Back 10 years ago,
00:25:10 an entrepreneur probably could not get funding
00:25:13 if he or she could not describe
00:25:16 what product he or she is copying from the US.
00:25:20 The first question is who has proven this business model
00:25:23 which is a nice way of asking who are you copying?
00:25:27 And that reason is understandable
00:25:29 because China had a much lower internet penetration
00:25:34 and didn’t have enough indigenous experience
00:25:40 to build innovative products.
00:25:43 And secondly, internet was emerging.
00:25:47 Link startup was the way to do things,
00:25:49 building a first minimally viable product
00:25:52 and then expanding was the right way to go.
00:25:55 And the American successes have given a shortcut
00:25:59 that if you built your minimally viable product
00:26:02 based on an American product,
00:26:04 it’s guaranteed to be a decent starting point.
00:26:06 Then you tweak it afterwards.
00:26:08 So as long as there are no IP infringement,
00:26:11 which as far as I know there hasn’t been in the mobile
00:26:14 and AI spaces, that’s a much better shortcut.
00:26:19 And I think Silicon Valley would view that
00:26:21 as still not very honorable
00:26:25 because that’s not your own idea to start with,
00:26:29 but you can’t really at the same time
00:26:32 believe every idea must be your own
00:26:35 and believe in the link startup methodology
00:26:38 because link startup is intended to try many, many things
00:26:41 and then converge when that works.
00:26:44 And it’s meant to be iterated and changed.
00:26:46 So finding a decent starting point
00:26:48 without legal violations,
00:26:51 there should be nothing morally dishonorable about that.
00:26:55 Yeah, so just a quick pause on that.
00:26:56 It’s fascinating that that’s,
00:26:59 why is that not honorable, right?
00:27:01 It’s exactly as you formulated.
00:27:04 It seems like a perfect start for business.
00:27:07 Is to take, look at Amazon and say,
00:27:12 okay, we’ll do exactly what Amazon is doing.
00:27:14 Let’s start there in this particular market
00:27:16 and then let’s out innovate them from that starting point.
00:27:20 Come up with new ways.
00:27:22 I mean, is it wrong to be,
00:27:25 except the word copycat just sounds bad,
00:27:27 but is it wrong to be a copycat?
00:27:28 It just seems like a smart strategy,
00:27:31 but yes, it doesn’t have a heroic nature to it
00:27:35 that like Steve Jobs, Elon Musk,
00:27:40 sort of in something completely,
00:27:42 coming up with something completely new.
00:27:43 Yeah, I like the way you describe it.
00:27:45 It’s a nonheroic, acceptable way to start the company
00:27:50 and maybe more expedient.
00:27:52 So that’s, I think, a baggage for Silicon Valley
00:27:58 that if it doesn’t let go,
00:28:00 then it may limit the ultimate ceiling of the company.
00:28:05 Take Snapchat as an example.
00:28:07 I think, you know, Evan’s brilliant.
00:28:09 He built a great product,
00:28:11 but he’s very proud that he wants to build his own features,
00:28:15 not copy others.
00:28:16 While Facebook was more willing to copy his features
00:28:20 and you see what happens in the competition.
00:28:23 So I think putting that handcuff on the company
00:28:27 would limit its ability to reach the maximum potential.
00:28:31 So back to the Chinese environment,
00:28:33 copying was merely a way to learn from the American masters.
00:28:38 Just like we, if we learned to play piano or painting,
00:28:43 you start by copying.
00:28:44 You don’t start by innovating
00:28:45 when you don’t have the basic skill sets.
00:28:48 So very amazingly, the Chinese entrepreneurs
00:28:51 about six years ago started to branch off
00:28:56 with these lean startups built on American ideas
00:28:59 to build better products than American products.
00:29:02 But they did start from the American idea.
00:29:04 And today WeChat is better than WhatsApp,
00:29:08 Weibo is better than Twitter,
00:29:10 Zhihu is better than Quora and so on.
00:29:12 So that I think is Chinese entrepreneurs going to step two.
00:29:18 And then step three is once these entrepreneurs
00:29:21 have done one or two of these companies,
00:29:23 they now look at the Chinese market and the opportunities
00:29:27 and come up with ideas that didn’t exist elsewhere.
00:29:30 So products like Ant Financial,
00:29:34 under which includes Alipay, which is mobile payments,
00:29:38 and also the financial products for loans built on that.
00:29:44 And also in education, VIPKID,
00:29:48 and in social video, social network, TikTok,
00:29:54 and in social eCommerce, Pinduoduo,
00:29:58 and then in ride sharing, Mobike,
00:30:01 these are all Chinese innovated products
00:30:05 that now are being copied elsewhere.
00:30:08 So an additional interesting observation
00:30:13 is some of these products
00:30:14 are built on unique Chinese demographics,
00:30:17 which may not work in the US,
00:30:19 but may work very well in Southeast Asia, Africa,
00:30:23 and other developing worlds
00:30:25 that are a few years behind China.
00:30:27 And a few of these products maybe are universal
00:30:31 and are getting traction even in the United States,
00:30:33 such as TikTok.
00:30:35 So this whole ecosystem is supported by VCs
00:30:42 as a virtuous cycle,
00:30:43 because a large market with innovative entrepreneurs
00:30:47 will draw a lot of money
00:30:49 and then invest in these companies.
00:30:51 As the market gets larger and larger,
00:30:54 the China market is easily three, four times larger than the US,
00:30:58 they will create greater value and greater returns
00:31:01 for the VCs, thereby raising even more money.
00:31:05 So at Sinovation Ventures, our first fund was 15 million,
00:31:09 our last fund was 500 million.
00:31:12 So it reflects the valuation of the companies
00:31:16 and our us going multi stage and things like that.
00:31:19 It also has government support,
00:31:22 but not in the way most Americans would think of it.
00:31:25 The government actually leaves the entrepreneurial space
00:31:28 as a private enterprise, sort of self regulating,
00:31:32 and the government would build infrastructures
00:31:35 that would around it to make it work better.
00:31:38 For example, the Mass Entrepreneur Mass Innovation Plan
00:31:42 builds 8,000 incubators,
00:31:44 so the pipeline is very strong to the VCs.
00:31:47 For autonomous vehicles,
00:31:48 the Chinese government is building smart highways
00:31:52 with sensors, smart cities
00:31:54 that separate pedestrians from cars
00:31:57 that may allow initially an inferior
00:32:00 autonomous vehicle company to launch a car
00:32:03 without increasing with lower casualty
00:32:07 because the roads or the city is smart.
00:32:10 And the Chinese government at local levels
00:32:13 would have these guiding funds acting as LPs,
00:32:16 passive LPs to funds.
00:32:18 And when the fund makes money,
00:32:21 part of the money made is given back
00:32:23 to the GPs and potentially other LPs
00:32:26 to increase everybody’s return
00:32:30 at the expense of the government’s return.
00:32:33 So that’s an interesting incentive
00:32:35 that entrusts the task of choosing entrepreneurs to VCs
00:32:41 who are better at it than the government
00:32:43 by letting some of the profits move that way.
00:32:46 So this is really fascinating, right?
00:32:48 So I look at the Russian government as a case study
00:32:51 where, let me put it this way,
00:32:54 there’s no such government driven
00:32:57 large scale support of entrepreneurship.
00:33:00 And probably the same is true in the United States,
00:33:04 but the entrepreneurs themselves kind of find a way.
00:33:07 So maybe in a form of advice or explanation,
00:33:11 how did the Chinese government arrive to be this way
00:33:15 so supportive on entrepreneurship
00:33:17 to be in this particular way so forward thinking
00:33:21 at such a large scale?
00:33:23 And also perhaps, how can we copy it in other countries?
00:33:28 How can we encourage other governments,
00:33:29 like even the United States government,
00:33:31 to support infrastructure for autonomous vehicles
00:33:33 in that same kind of way, perhaps?
00:33:36 Yes, so these techniques are
00:33:41 the result of several key things,
00:33:44 some of which may be learnable,
00:33:46 some of which may be very hard.
00:33:48 One is just trial and error
00:33:50 and watching what everyone else is doing.
00:33:52 I think it’s important to be humble
00:33:54 and not feel like you know all the answers.
00:33:56 The guiding funds idea came from Singapore,
00:33:59 which came from Israel.
00:34:01 And China made a few tweaks and turned it into a,
00:34:06 because the Chinese cities and government officials
00:34:09 kind of compete with each other
00:34:11 because they all want to make their city more successful
00:34:14 so they can get the next level in their political career.
00:34:20 And it’s somewhat competitive.
00:34:22 So the central government made it a bit of a competition.
00:34:25 Everybody has a budget.
00:34:26 They can put it on AI or they can put it on bio
00:34:29 or they can put it on energy.
00:34:32 And then whoever gets the results,
00:34:34 the city shines, the people are better off,
00:34:36 the mayor gets a promotion.
00:34:37 So the tools is kind of almost like an entrepreneurial
00:34:41 environment for local governments
00:34:44 to see who can do a better job.
00:34:47 And also many of them try different experiments.
00:34:52 Some have given award to very smart researchers.
00:34:58 Just give them money and hope they’ll start a company.
00:35:00 Some have given money to academic research labs,
00:35:05 maybe government research labs
00:35:08 to see if they can spin off some companies
00:35:10 from the science lab or something like that.
00:35:14 Some have tried to recruit overseas Chinese
00:35:17 to come back and start companies.
00:35:19 And they’ve had mixed results.
00:35:21 The one that worked the best was the guiding funds.
00:35:23 So it’s almost like a lean startup idea
00:35:25 where people try different things
00:35:27 and what works sticks and everybody copies.
00:35:30 So now every city has a guiding fund.
00:35:32 So that’s how that came about.
00:35:34 The autonomous vehicle and the massive spending
00:35:40 in highways and smart cities, that’s a Chinese way.
00:35:46 It’s about building infrastructure to facilitate.
00:35:49 It’s a clear division of the government’s responsibility
00:35:52 from the market.
00:35:55 The market should do everything in a private freeway,
00:36:00 but there are things the market can’t afford to do
00:36:02 like infrastructure.
00:36:04 So the government always appropriates large amounts
00:36:08 of money for infrastructure building.
00:36:11 This happens with not only autonomous vehicle and AI,
00:36:16 but happened with the 3G and 4G.
00:36:20 You’ll find that the Chinese wireless reception
00:36:25 is better than the US because massive spending
00:36:28 that tries to cover the whole country,
00:36:30 whereas in the US it may be a little spotty.
00:36:33 It’s a government driven because I think they view
00:36:36 the coverage of cell access and 3G, 4G access
00:36:44 to be a governmental infrastructure spending
00:36:47 as opposed to capitalistic.
00:36:49 So that’s, of course, the state owned enterprises
00:36:52 are also publicly traded,
00:36:54 but they also carry a government responsibility
00:36:57 to deliver infrastructure to all.
00:37:00 So it’s a different way of thinking
00:37:01 that may be very hard to inject into Western countries
00:37:05 to say starting tomorrow, bandwidth infrastructure
00:37:09 and highways are gonna be governmental spending
00:37:13 with some characteristics.
00:37:16 What’s your sense, and sorry to interrupt,
00:37:18 but because it’s such a fascinating point,
00:37:21 do you think on the autonomous vehicle space
00:37:25 it’s possible to solve the problem of full autonomy
00:37:30 without significant investment in infrastructure?
00:37:34 Well, that’s really hard to speculate.
00:37:36 I think it’s not a yes, no question,
00:37:39 but how long does it take question?
00:37:42 15 years, 30 years, 45 years.
00:37:45 Clearly with infrastructure augmentation,
00:37:49 whether it’s road, the city or whole city planning,
00:37:52 building a new city, I’m sure that will accelerate
00:37:56 the day of the L5.
00:37:58 I’m not knowledgeable enough,
00:38:01 and it’s hard to predict even when we’re knowledgeable
00:38:03 because a lot of it is speculative.
00:38:07 But in the US, I don’t think people would consider
00:38:10 building a new city the size of Chicago
00:38:13 to make it the AI slash autonomous city.
00:38:15 There are smaller ones being built, I’m aware of that.
00:38:18 But is infrastructure spend really impossible
00:38:22 for US or Western countries?
00:38:23 I don’t think so.
00:38:25 The US highway system was built,
00:38:28 was that during President Eisenhower or Kennedy?
00:38:31 Eisenhower, yeah.
00:38:33 So maybe historians can study
00:38:37 how the President Eisenhower get the resources
00:38:40 to build this massive infrastructure
00:38:42 that surely gave US a tremendous amount of prosperity
00:38:47 over the next decade, if not century.
00:38:50 If I may comment on that then,
00:38:53 it takes us to artificial intelligence a little bit
00:38:55 because in order to build infrastructure,
00:38:58 it creates a lot of jobs.
00:39:00 So I’ll be actually interested if you would say
00:39:03 that you talk in your book about all kinds of jobs
00:39:06 that could and could not be automated.
00:39:08 I wonder if building infrastructure is one of the jobs
00:39:13 that would not be easily automated.
00:39:15 Something you could think about
00:39:17 because I think you’ve mentioned somewhere in the talk
00:39:19 or that there might be, as jobs are being automated,
00:39:24 a role for government to create jobs
00:39:26 that can’t be automated.
00:39:28 Yes, I think that’s a possibility.
00:39:31 Back in the last financial crisis,
00:39:34 China put a lot of money
00:39:37 to basically give this economy a boost
00:39:41 and a lot of it went into infrastructure building.
00:39:45 And I think that’s a legitimate way at the government level
00:39:49 to deal with the employment issues
00:39:54 as well as build out the infrastructure
00:39:57 as long as the infrastructures are truly needed
00:39:59 and as long as there is an employment problem,
00:40:02 which no, we don’t know.
00:40:04 So maybe taking a little step back,
00:40:07 if you’ve been a leader and a researcher in AI
00:40:12 for several decades, at least 30 years,
00:40:16 so how has AI changed in the West and the East
00:40:21 as you’ve observed, as you’ve been deep in it
00:40:23 over the past 30 years?
00:40:25 Well, AI began as the pursuit
00:40:27 of understanding human intelligence
00:40:30 and the term itself represents that,
00:40:34 but it kind of drifted into the one sub area
00:40:37 that worked extremely well, which is machine intelligence.
00:40:40 And that’s actually more using pattern recognition techniques
00:40:45 to basically do incredibly well on a limited domain,
00:40:51 large amount of data,
00:40:52 but relatively simple kinds of planning tasks
00:40:57 and not very creative.
00:40:58 So we didn’t end up building human intelligence.
00:41:02 We built a different machine
00:41:04 that was a lot better than us, some problems,
00:41:08 but nowhere close to us on other problems.
00:41:11 So today, I think a lot of people still misunderstand
00:41:16 when we say artificial intelligence
00:41:18 and what various products can do,
00:41:20 people still think it’s about replicating
00:41:22 human intelligence,
00:41:24 but the products out there really are closer
00:41:27 to having invented the internet or the spreadsheet
00:41:31 or the database and getting broader adoption.
00:41:35 And speaking further to the fears,
00:41:37 near term fears that people have about AI,
00:41:40 so you’re commenting on the sort of general intelligence
00:41:45 that people in the popular culture from sci fi movies
00:41:48 have a sense about AI,
00:41:49 but there’s practical fears about AI,
00:41:52 the narrow AI that you’re talking about
00:41:54 of automating particular kinds of jobs
00:41:57 and you talk about them in the book.
00:41:59 So what are the kinds of jobs in your view
00:42:01 that you see in the next five, 10 years
00:42:04 beginning to be automated by AI systems algorithms?
00:42:09 Yes, this is also maybe a little bit counterintuitive
00:42:13 because it’s the routine jobs
00:42:15 that will be displaced the soonest
00:42:18 and they may not be displaced entirely,
00:42:20 maybe 50%, 80% of a job,
00:42:24 but when the workload drops by that much,
00:42:26 employment will come down.
00:42:28 And also another part of misunderstanding
00:42:31 is most people think of AI replacing routine jobs
00:42:35 than they think of the assembly line, the workers.
00:42:38 Well, that will have some effect,
00:42:41 but it’s actually the routine white collar workers
00:42:44 that’s easiest to replace
00:42:46 because to replace a white collar worker,
00:42:49 you just need software.
00:42:50 To replace a blue collar worker,
00:42:53 you need robotics, mechanical excellence,
00:42:57 and the ability to deal with dexterity
00:43:01 and maybe even unknown environments, very, very difficult.
00:43:05 So if we were to categorize the most dangerous
00:43:10 white collar jobs,
00:43:12 they would be things like back office,
00:43:15 people who copy and paste
00:43:17 and deal with simple computer programs and data
00:43:22 and maybe paper and OCR,
00:43:25 and they don’t make strategic decisions.
00:43:29 They basically facilitate the process.
00:43:32 These softwares and paper systems don’t work.
00:43:34 So you have people dealing with new employee orientation,
00:43:40 searching for past lawsuits and financial documents,
00:43:45 and doing reference check.
00:43:47 So basic searching and management of data.
00:43:50 That’s the most endangered being lost.
00:43:52 In addition to the white collar repetitive work,
00:43:56 a lot of simple interaction work can also be taken care of
00:44:00 such as telesales, telemarketing, customer service,
00:44:04 as well as many physical jobs
00:44:07 that are in the same location
00:44:09 and don’t require a high degree of dexterity.
00:44:12 So fruit picking, dishwashing, assembly line inspection
00:44:17 are jobs in that category.
00:44:20 So altogether, back office is a big part.
00:44:25 And the blue collar may be smaller initially,
00:44:29 but over time, AI will get better.
00:44:32 And when we start to get to over the next 15, 20 years,
00:44:36 the ability to actually have the dexterity
00:44:39 of doing assembly line, that’s a huge chunk of jobs.
00:44:42 And when autonomous vehicles start to work,
00:44:45 initially starting with truck drivers,
00:44:47 but eventually to all drivers,
00:44:49 that’s another huge group of workers.
00:44:52 So I see modest numbers in the next five years,
00:44:55 but increasing rapidly after that.
00:44:58 On the worry of the jobs that are in danger
00:45:01 and the gradual loss of jobs,
00:45:04 I’m not sure if you’re familiar with Andrew Yang.
00:45:06 Yes, I am.
00:45:07 So there’s a candidate for president of the United States
00:45:10 whose platform Andrew Yang is based around,
00:45:14 in part around job loss due to automation.
00:45:17 And also in addition,
00:45:19 the need perhaps of universal basic income
00:45:22 to support jobs that are,
00:45:25 folks who lose their job due to automation and so on.
00:45:28 And in general, support people
00:45:30 under complex, unstable job market.
00:45:34 So what are your thoughts about his concerns,
00:45:36 him as a candidate, his ideas in general?
00:45:40 I think his thinking is generally in the right direction,
00:45:44 but his approach as a presidential candidate
00:45:48 may be a little bit ahead of the time.
00:45:51 And I think the displacements will happen,
00:45:56 but will they happen soon enough
00:45:57 for people to agree to vote for him?
00:46:00 The unemployment numbers are not very high yet.
00:46:03 And I think he and I have the same challenge.
00:46:07 If I want to theoretically convince people this is an issue
00:46:11 and he wants to become the president,
00:46:13 people have to see how can this be the case
00:46:17 when unemployment numbers are low.
00:46:19 So that is the challenge.
00:46:21 And I think I do agree with him on the displacement issue,
00:46:27 on universal basic income.
00:46:30 At a very vanilla level, I don’t agree with it
00:46:33 because I think the main issue is retraining.
00:46:38 So people need to be incented
00:46:41 not by just giving a monthly $2,000 check or $1,000 check
00:46:45 and do whatever they want
00:46:47 because they don’t have the know how
00:46:50 to know what to retrain to go into what type of a job.
00:46:56 And guidance is needed.
00:46:58 And retraining is needed
00:47:00 because historically when technology revolutions,
00:47:03 when routine jobs were displaced, new routine jobs came up.
00:47:06 So there was always room for that.
00:47:09 But with AI and automation,
00:47:11 the whole point is replacing all routine jobs eventually.
00:47:15 So there will be fewer and fewer routine jobs.
00:47:17 And AI will create jobs, but it won’t create routine jobs
00:47:22 because if it creates routine jobs,
00:47:24 why wouldn’t AI just do it?
00:47:26 So therefore the people who are losing the jobs
00:47:30 are losing routine jobs.
00:47:32 The jobs that are becoming available are non routine jobs.
00:47:35 So the social stipend needs to be put in place
00:47:39 is for the routine workers who lost their jobs
00:47:42 to be retrained maybe in six months, maybe in three years,
00:47:46 takes a while to retrain on the non routine job
00:47:48 and then take on a job that will last
00:47:51 for that person’s lifetime.
00:47:53 Now, having said that,
00:47:55 if you look deeply into Andrew’s document,
00:47:57 he does cater for that.
00:47:58 So I’m not disagreeing with what he’s trying to do.
00:48:03 But for simplification, sometimes he just says UBI,
00:48:06 but simple UBI wouldn’t work.
00:48:08 And I think you’ve mentioned elsewhere
00:48:10 that the goal isn’t necessarily to give people enough money
00:48:15 to survive or live, or even to prosper.
00:48:19 The point is to give them a job that gives them meaning.
00:48:22 That meaning is extremely important.
00:48:25 That our employment, at least in the United States
00:48:28 and perhaps it carries across the world,
00:48:31 provides something that’s, forgive me for saying,
00:48:34 greater than money.
00:48:35 It provides meaning.
00:48:36 So now, what kind of jobs do you think can’t be automated?
00:48:43 Can you talk a little bit about creativity
00:48:45 and compassion in your book?
00:48:46 What aspects do you think it’s difficult to automate
00:48:49 for an AI system?
00:48:51 Because an AI system is currently merely optimizing.
00:48:56 It’s not able to reason, plan,
00:48:59 or think creatively or strategically.
00:49:02 It’s not able to deal with complex problems.
00:49:04 It can’t come up with a new problem and solve it.
00:49:08 A human needs to find the problem
00:49:11 and pose it as an optimization problem,
00:49:14 then have the AI work at it.
00:49:16 So an AI would have a very hard time discovering a new drug
00:49:22 or discovering a new style of painting
00:49:26 or dealing with complex tasks such as managing a company
00:49:31 that isn’t just about optimizing the bottom line,
00:49:34 but also about employee satisfaction, corporate brand,
00:49:38 and many, many other things.
00:49:39 So that is one category of things.
00:49:43 And because these things are challenging, creative, complex,
00:49:47 doing them creates a high degree of satisfaction
00:49:51 and therefore appealing to our desire for working,
00:49:54 which isn’t just to make the money, make the ends meet,
00:49:57 but also that we’ve accomplished something
00:49:59 that others maybe can’t do or can’t do as well.
00:50:03 Another type of job that is much numerous
00:50:06 would be compassionate jobs, jobs that require compassion,
00:50:10 empathy, human touch, human trust.
00:50:13 AI can’t do that because AI is cold, calculating,
00:50:17 and even if it can fake that to some extent,
00:50:21 it will make errors and that will make it look very silly.
00:50:25 And also, I think even if AI did okay,
00:50:28 people would want to interact with another person,
00:50:32 whether it’s for some kind of a service or a teacher
00:50:35 or a doctor or a concierge or a masseuse or a bartender.
00:50:40 There are so many jobs where people just don’t want
00:50:44 to interact with a cold robot or software.
00:50:49 I’ve had an entrepreneur who built an elderly care robot
00:50:52 and they found that the elderly really only use it
00:50:55 for customer service.
00:50:56 And not, but not to service the product,
00:50:59 but they click on customer service
00:51:01 and the video of a person comes up
00:51:04 and then the person says,
00:51:05 how come my daughter didn’t call me?
00:51:08 Let me show you a picture of her grandkids.
00:51:10 So people yearn for that people, people interaction.
00:51:14 So even if robots improved, people just don’t want it.
00:51:17 And those jobs are going to be increasing
00:51:20 because AI will create a lot of value,
00:51:22 $16 trillion to the world in the next 10 years.
00:51:26 Next 11 years, according to PWC.
00:51:29 And that will give people money to enjoy services,
00:51:34 whether it’s eating a gourmet meal or tourism and traveling
00:51:39 or having concierge services,
00:51:41 the services revolving around every dollar
00:51:45 of that $16 trillion will be tremendous.
00:51:48 It will create more opportunities
00:51:50 that are to service the people who did well
00:51:52 through AI with things.
00:51:56 But even at the same time,
00:51:58 the entire society is very much short
00:52:01 in need of many service oriented,
00:52:04 compassionate oriented jobs.
00:52:06 The best example is probably in healthcare services.
00:52:10 There’s going to be 2 million new jobs,
00:52:14 not counting replacement,
00:52:15 just brand new incremental jobs
00:52:17 in the next six years in healthcare services.
00:52:20 That includes nurses, orderly in the hospital,
00:52:25 elderly care and also at home care is particularly lacking.
00:52:31 And those jobs are not likely to be filled.
00:52:34 So there’s likely to be a shortage.
00:52:36 And the reason they’re not filled
00:52:38 is simply because they don’t pay very well
00:52:41 and that the social status of these jobs are not very good.
00:52:47 So they pay about half as much
00:52:49 as a heavy equipment operator,
00:52:52 which will be replaced a lot sooner.
00:52:55 And they pay probably comparably
00:52:57 to someone on the assembly line.
00:52:59 And so if we ignoring all the other issues
00:53:03 and just think about satisfaction from one’s job,
00:53:07 someone repetitively doing the same manual action
00:53:10 at an assembly line,
00:53:11 that can’t create a lot of job satisfaction,
00:53:14 but someone taking care of a sick person
00:53:17 and getting a hug and thank you
00:53:19 from that person and the family,
00:53:22 I think is quite satisfying.
00:53:24 So if only we could fix the pay for service jobs,
00:53:28 there are plenty of jobs that require some training
00:53:31 or a lot of training
00:53:33 for the people coming off the routine jobs to take.
00:53:36 We can easily imagine someone
00:53:40 who was maybe a cashier at the grocery store
00:53:43 as stores become automated,
00:53:45 learns to become a nurse or an at home care.
00:53:49 I also do want to point out the blue collar jobs
00:53:52 are going to stay around a bit longer.
00:53:54 Some of them quite a bit longer.
00:53:58 AI cannot be told go clean an arbitrary home.
00:54:02 That’s incredibly hard.
00:54:03 Arguably it’s an L5 level of difficulty, right?
00:54:07 And then AI cannot be a good plumber
00:54:10 because plumber is almost like a mini detective
00:54:12 that has to figure out where the leak came from.
00:54:15 So yet AI probably can be an assembly line
00:54:20 and auto mechanic and so on.
00:54:22 So one has to study which blue collar jobs are going away
00:54:26 and facilitate retraining for the people
00:54:29 to go into the ones that won’t go away
00:54:31 or maybe even will increase.
00:54:32 I mean, it is fascinating that it’s easier
00:54:35 to build a world champion chess player
00:54:39 than it is to build a mediocre plumber.
00:54:42 Yes, right.
00:54:43 Very true.
00:54:44 And to AI and that goes counterintuitive
00:54:46 to a lot of people’s understanding
00:54:48 of what artificial intelligence is.
00:54:50 So it sounds, I mean, you’re painting
00:54:52 a pretty optimistic picture about retraining
00:54:55 about the number of jobs
00:54:57 and actually the meaningful nature of those jobs
00:54:59 once we automate the repetitive tasks.
00:55:02 So overall, are you optimistic about the future
00:55:08 where much of the repetitive tasks are automated?
00:55:11 That there is a lot of room for humans
00:55:13 for the compassionate, for the creative input
00:55:17 that only humans can provide?
00:55:20 I am optimistic if we start to take action.
00:55:23 If we have no action in the next five years,
00:55:27 I think it’s going to be hard to deal
00:55:30 with the devastating losses that will emerge.
00:55:34 So if we start thinking about retraining,
00:55:37 maybe with the low hanging fruits,
00:55:39 explaining to vocational schools
00:55:41 why they should train more plumbers than auto mechanics,
00:55:46 maybe starting with some government subsidy
00:55:49 for corporations to have more training positions.
00:55:53 We start to explain to people why retraining is important.
00:55:58 We start to think about what the future of education,
00:56:00 how that needs to be tweaked for the era of AI.
00:56:04 If we start to make incremental progress
00:56:06 and the greater number of people understand,
00:56:08 then there’s no reason to think we can’t deal with this
00:56:12 because this technological revolution
00:56:14 is arguably similar to what electricity,
00:56:17 industrial revolutions, and internet brought about.
00:56:20 Do you think there’s a role for policy,
00:56:22 for governments to step in,
00:56:24 to help with policy to create a better world?
00:56:28 Absolutely, and the governments don’t have to believe
00:56:32 an employment will go up,
00:56:34 and they don’t have to believe automation will be this fast
00:56:37 to do something.
00:56:39 Revamping vocational school would be one example.
00:56:42 Another is if there’s a big gap
00:56:44 in healthcare service employment,
00:56:47 and we know that a country’s population is growing older,
00:56:51 more longevity, living older,
00:56:54 because people over 80 require five times as much care
00:56:57 as those under 80,
00:56:59 then it is a good time to incent training programs
00:57:03 for elderly care to find ways to improve the pay.
00:57:07 Maybe one way would be to offer as part of Medicare
00:57:11 or the equivalent program for people over 80
00:57:14 to be entitled to a few hours of elderly care at home,
00:57:18 and then that might be reimbursable,
00:57:22 and that will stimulate the service industry
00:57:26 around the policy.
00:57:28 Do you have concerns about large entities,
00:57:33 whether it’s governments or companies,
00:57:35 controlling the future of AI development in general?
00:57:39 So we talked about companies.
00:57:41 Do you have a better sense that governments
00:57:44 can better represent the interests of the people
00:57:49 than companies, or do you believe companies
00:57:52 are better at representing the interests of the people?
00:57:54 Or is there no easy answer?
00:57:56 I don’t think there’s an easy answer
00:57:58 because it’s a double edged sword.
00:58:00 The companies and governments can provide better services
00:58:03 with more access to data and more access to AI,
00:58:06 but that also leads to greater power,
00:58:09 which can lead to uncontrollable problems,
00:58:13 whether it’s monopoly or corruption in the government.
00:58:17 So I think one has to be careful
00:58:21 to look at how much data that companies and governments have
00:58:25 and some kind of checks and balances would be helpful.
00:58:30 So again, I come from Russia.
00:58:34 There’s something called the Cold War.
00:58:36 So let me ask a difficult question here
00:58:39 looking at conflict.
00:58:40 Steven Pinker written a great book
00:58:42 that conflict all over the world is decreasing in general.
00:58:45 But do you have a sense that having written
00:58:49 the book AI Superpowers,
00:58:51 do you see a major international conflict
00:58:54 potentially arising between major nations,
00:58:57 whatever they are, whether it’s Russia, China,
00:59:00 European nations, United States or others
00:59:04 in the next 10, 20, 50 years around AI,
00:59:07 around the digital space, cyberspace?
00:59:10 Do you worry about that?
00:59:12 Is that something we need to think about
00:59:15 and try to alleviate or prevent?
00:59:19 I believe in greater engagement.
00:59:22 A lot of the worries about more powerful AI
00:59:26 are based on a arms race metaphor.
00:59:33 And when you extrapolate into military kinds of scenarios,
00:59:41 AI can automate and autonomous weapons
00:59:46 that needs to be controlled somehow
00:59:48 and autonomous decision making
00:59:51 can lead to not enough time to fix international crises.
00:59:57 So I actually believe a Cold War mentality
01:00:00 would be very dangerous
01:00:02 because should two countries rely on AI
01:00:05 to make certain decisions
01:00:07 and they don’t even talk to each other,
01:00:10 they do their own scenario planning,
01:00:12 then something could easily go wrong.
01:00:15 I think engagement, interaction, some protocols
01:00:18 to avoid inadvertent disasters is actually needed.
01:00:25 So it’s natural for each country to want to be the best,
01:00:29 whether it’s in nuclear technologies or AI or bio.
01:00:34 But I think it’s important to realize
01:00:37 if each country has a black box AI
01:00:41 and don’t talk to each other,
01:00:43 that probably presents greater challenges to humanity
01:00:49 than if they interacted.
01:00:51 I think there can still be competition,
01:00:53 but with some degree of protocol for interaction,
01:00:57 just like when there was a nuclear competition,
01:01:02 there were some protocol for deterrence
01:01:04 among US, Russia, and China.
01:01:08 And I think that engagement is needed.
01:01:10 So of course, we’re still far from AI
01:01:13 presenting that kind of danger.
01:01:16 But what I worry the most about
01:01:18 is the level of engagement seems to be coming down.
01:01:23 The level of distrust seems to be going up,
01:01:26 especially from the US towards other large countries
01:01:30 such as China and of course, and Russia, yes.
01:01:33 Is there a way to make that better?
01:01:34 So let’s beautifully put level of engagement
01:01:37 and even just basic trust and communication
01:01:40 as opposed to sort of making artificial enemies
01:01:48 out of particular countries.
01:01:53 Do you have a sense how we can make it better?
01:01:57 Actionable items that as a society we can take on?
01:02:01 I’m not an expert at geopolitics,
01:02:05 but I would say that we look pretty foolish as humankind
01:02:10 when we are faced with the opportunity
01:02:13 to create $16 trillion for humanity,
01:02:19 and yet we’re not solving fundamental problems
01:02:26 with parts of the world still in poverty.
01:02:29 And for the first time,
01:02:31 we have the resources to overcome poverty and hunger.
01:02:34 We’re not using it on that,
01:02:35 but we’re fueling competition among superpowers.
01:02:38 And that’s a very unfortunate thing.
01:02:41 If we become utopian for a moment,
01:02:44 imagine a benevolent world government
01:02:52 that has this $16 trillion and maybe some AI
01:02:56 to figure out how to use it to deal with diseases
01:02:59 and problems and hate and things like that.
01:03:02 World would be a lot better off.
01:03:04 So what is wrong with the current world?
01:03:07 I think the people with more skill than I
01:03:11 should think about this.
01:03:13 And then the geopolitics issue with superpower competition
01:03:16 is one side of the issue.
01:03:19 There’s another side which I worry maybe even more,
01:03:24 which is as the $16 trillion all gets made by US and China
01:03:29 and a few of the other developed countries,
01:03:32 the poorer country will get nothing
01:03:34 because they don’t have technology
01:03:36 and the wealth disparity and inequality will increase.
01:03:42 So a poorer country with a large population
01:03:45 will not only benefit from the AI boom
01:03:48 or other technology booms,
01:03:50 but they will have their workers
01:03:52 who previously had hoped they could do the China model
01:03:56 and do outsource manufacturing or the India model
01:03:58 so they could do the outsource process or call center.
01:04:02 Well, all those jobs are gonna be gone in 10 or 15 years.
01:04:05 So the individual citizen may be a net liability,
01:04:12 I mean, financially speaking to a poorer country
01:04:15 and not an asset to claw itself out of poverty.
01:04:19 So in that kind of situation,
01:04:22 these large countries with not much tech
01:04:26 are going to be facing a downward spiral
01:04:30 and it’s unclear what could be done.
01:04:33 And then when we look back
01:04:34 and say there’s $16 trillion being created
01:04:37 and it’s all being kept by US, China
01:04:39 and other developed countries, it just doesn’t feel right.
01:04:43 So I hope people who know about geopolitics
01:04:46 can find solutions that’s beyond my expertise.
01:04:50 So different countries that we’ve talked about
01:04:53 have different value systems.
01:04:55 If you look at the United States,
01:04:56 to an almost extreme degree,
01:04:58 there is an absolute desire for freedom of speech.
01:05:03 If you look at a country where I was raised,
01:05:05 that desire just amongst the people
01:05:06 is not as elevated as it is to basically fundamental level
01:05:15 to the essence of what it means to be America, right?
01:05:17 And the same is true with China,
01:05:19 there’s different value systems.
01:05:21 There’s some censorship of internet content
01:05:26 that China and Russia and many other countries undertake.
01:05:31 Do you see that having effects on innovation,
01:05:36 other aspects of some of the tech stuff,
01:05:38 AI development we talked about,
01:05:40 and maybe from another angle,
01:05:42 do you see that changing in different ways
01:05:46 over the next 10 years, 20 years, 50 years
01:05:49 as China continues to grow as it does now
01:05:53 in its tech innovation?
01:05:55 There’s a common belief
01:05:57 that full freedom of speech and expression
01:06:01 is correlated with creativity,
01:06:03 which is correlated with entrepreneurial success.
01:06:08 I think empirically we have seen that is not true
01:06:13 and China has been successful.
01:06:15 That’s not to say the fundamental values are not right
01:06:19 or not the best,
01:06:20 but it’s just that perfect correlation isn’t there.
01:06:25 It’s hard to read the tea leaves on opening up or not
01:06:30 in any country,
01:06:31 and I’ve not been very good at that in my past predictions,
01:06:37 but I do believe every country
01:06:40 shares a lot of fundamental values for the longterm.
01:06:47 So China is drafting its privacy policy
01:06:54 for individual citizens,
01:06:57 and they don’t look that different
01:07:00 from the American or European ones.
01:07:03 So people do want to protect their privacy
01:07:07 and have the opportunity to express
01:07:10 and I think the fundamental values are there.
01:07:14 The question is in the execution and timing,
01:07:17 how soon or when will that start to open up?
01:07:21 So as long as each government knows
01:07:25 ultimately people want that kind of protection,
01:07:29 there should be a plan to move towards that
01:07:32 as to when or how and I’m not an expert.
01:07:36 On the point of privacy to me, it’s really interesting.
01:07:39 So AI needs data to create
01:07:42 a personalized awesome experience, right?
01:07:45 I’m just speaking generally in terms of products.
01:07:48 And then we have currently, depending on the age
01:07:51 and depending on the demographics of who we’re talking about,
01:07:54 some people are more or less concerned
01:07:55 about the amount of data they hand over.
01:07:59 So in your view, how do we get this balance right
01:08:04 that we provide an amazing experience
01:08:07 to people that use products?
01:08:09 You look at Facebook, the more Facebook knows about you,
01:08:13 yes, it’s scary to say, the better it can probably,
01:08:19 better experience it can probably create.
01:08:21 So in your view, how do we get that balance right?
01:08:25 Yes, I think a lot of people have a misunderstanding
01:08:30 that it’s okay and possible to just rip all the data out
01:08:35 from a provider and give it back to you.
01:08:38 So you can deny them access to further data
01:08:41 and still enjoy the services we have.
01:08:44 If we take back all the data,
01:08:46 all the services will give us nonsense.
01:08:48 We’ll no longer be able to use products that function well
01:08:52 in terms of right ranking, right products,
01:08:56 right user experience.
01:08:57 So yet I do understand we don’t want to permit misuse
01:09:02 of the data from legal policy standpoint.
01:09:07 I think there can be severe punishment
01:09:11 for those who have egregious misuse of the data.
01:09:16 That’s I think a good first step.
01:09:19 Actually China in this side on this aspect
01:09:22 has very strong laws about people who sell
01:09:25 or give data to other companies.
01:09:27 And that over the past few years,
01:09:30 since that law came into effect,
01:09:33 pretty much eradicated the illegal distribution,
01:09:38 sharing of data.
01:09:40 Additionally, I think giving,
01:09:45 I think technology is often a very good way
01:09:50 to solve technology misuse.
01:09:52 So can we come up with new technologies
01:09:56 that will let us have our cake and eat it too?
01:09:58 People are looking into homomorphic encryption,
01:10:01 which is letting you keep the data,
01:10:04 have it encrypted and train on encrypted data.
01:10:07 Of course, we haven’t solved that one yet,
01:10:09 but that kind of direction may be worth pursuing.
01:10:13 Also federated learning,
01:10:15 which would allow one hospital
01:10:17 to train on its hospital’s patient data fully
01:10:20 because they have a license for that.
01:10:22 And then hospitals would then share their models,
01:10:24 not data, but models to create a super AI.
01:10:28 And that also maybe has some promise.
01:10:30 So I would want to encourage us to be open minded
01:10:34 and think of this as not just the policy binary, yes, no,
01:10:39 but letting the technologists try to find solutions
01:10:42 to let us have our cake and eat it too,
01:10:44 or have most of our cake and eat most of it too.
01:10:48 Finally, I think giving each end user a choice is important
01:10:52 and having transparency is important.
01:10:55 Also, I think that’s universal,
01:10:57 but the choice you give to the user
01:11:00 should not be at a granular level
01:11:02 that the user cannot understand.
01:11:04 GDPR today causes all these popups of yes, no,
01:11:09 will you give this site this right
01:11:10 to use this part of your data?
01:11:12 I don’t think any user understands
01:11:15 what they’re saying yes or no to.
01:11:17 And I suspect most are just saying yes
01:11:18 because they don’t understand it.
01:11:20 So while GDPR in its current implementation
01:11:25 has lived up to its promise of transparency and user choice,
01:11:30 it implemented it in such a way
01:11:33 that really didn’t deliver the spirit of GDPR.
01:11:39 It fit the letter, but not the spirit.
01:11:41 So again, I think we need to think about
01:11:43 is there a way to fit the spirit of GDPR
01:11:48 by using some kind of technology?
01:11:50 Can we have a slider that’s an AI trying to figure out
01:11:54 how much you want to slide between
01:11:57 perfect protection security of your personal data
01:12:01 versus a high degree of convenience
01:12:04 with some risks of not having full privacy?
01:12:08 Each user should have some preference
01:12:10 and that gives you the user choice.
01:12:12 But maybe we should turn the problem on its head
01:12:14 and ask can there be an AI algorithm that can customize this?
01:12:18 Because we can understand the slider,
01:12:21 but we sure cannot understand every popup question.
01:12:25 And I think getting that right
01:12:27 requires getting the balance between
01:12:29 what we talked about earlier,
01:12:30 which is heart and soul
01:12:32 versus profit driven decisions and strategy.
01:12:37 I think from my perspective,
01:12:40 the best way to make a lot of money in the long term
01:12:43 is to keep your heart and soul intact.
01:12:46 I think getting that slider right in the short term
01:12:50 may feel like you’ll be sacrificing profit,
01:12:54 but in the long term,
01:12:55 you’ll be gaining user trust
01:12:57 and providing a great experience.
01:12:59 Do you share that kind of view in general?
01:13:02 Yes, absolutely.
01:13:03 I sure would hope there is a way
01:13:07 we can do long term projects
01:13:09 that really do the right thing.
01:13:12 I think a lot of people who embrace GDPR,
01:13:15 their heart’s in the right place.
01:13:16 I think they just need to figure out how to build a solution.
01:13:20 I’ve heard utopians talk about solutions
01:13:23 that get me excited,
01:13:24 but I’m not sure how in the current funding environment
01:13:27 they can get started.
01:13:29 People talk about,
01:13:30 imagine this crowdsourced data collection
01:13:36 that we all trust.
01:13:37 And then we have these agents
01:13:40 that we ask the trusted agent to…
01:13:45 That agent only, that platform,
01:13:48 so a trusted joint platform
01:13:51 that we all believe is trustworthy,
01:13:55 that can give us all the closed loop personal suggestions
01:14:03 by the new social network, new search engine,
01:14:06 new eCommerce engine that has access
01:14:08 to even more of our data,
01:14:10 but not directly, but indirectly.
01:14:12 So I think that general concept
01:14:14 of licensing to some trusted engine
01:14:18 and finding a way to trust that engine
01:14:20 seems like a great idea.
01:14:22 But if you think how long it’s gonna take
01:14:24 to implement and tweak and develop it right,
01:14:27 as well as to collect all the trusts
01:14:29 and the data from the people,
01:14:31 it’s beyond the current cycle of venture capital.
01:14:34 So how do you do that is a big question.
01:14:38 You’ve recently had a fight with cancer,
01:14:41 stage four lymphoma and in a sort of deep personal level,
01:14:48 what did it feel like in the darker moments
01:14:51 to face your own mortality?
01:14:54 Well, I’ve been the workaholic my whole life
01:14:57 and I’ve basically worked nine, nine, six,
01:15:01 nine a.m. to nine p.m. six days a week, roughly.
01:15:04 And I didn’t really pay a lot of attention
01:15:07 to my family, friends, and people who loved me.
01:15:10 And my life revolved around optimizing for work.
01:15:14 While my work was not routine,
01:15:16 my optimization really what made my life
01:15:23 basically very mechanical process.
01:15:25 But I got a lot of highs out of it
01:15:28 because of accomplishments
01:15:30 that I thought were really important and dear
01:15:34 and the highest priority to me.
01:15:36 But when I faced mortality
01:15:38 and the possible death in matter of months,
01:15:41 I suddenly realized that this really meant nothing to me,
01:15:45 that I didn’t feel like working for another minute,
01:15:48 that if I had six months left in my life,
01:15:51 I would spend it all with my loved ones
01:15:54 and thanking them, giving them love back
01:15:57 and apologizing to them that I lived my life the wrong way.
01:16:01 So that moment of reckoning
01:16:05 caused me to really rethink that why we exist in this world
01:16:11 is something that we might be too much shaped by the society
01:16:17 to think that success and accomplishments is why we live.
01:16:22 But while that can get you
01:16:26 periodic successes and satisfaction,
01:16:29 it’s really in the facing death
01:16:33 you see what’s truly important to you.
01:16:35 So as a result of going through the challenges with cancer,
01:16:41 I’ve resolved to live a more balanced lifestyle.
01:16:45 I’m now in remission, knock on wood,
01:16:48 and I’m spending more time with my family.
01:16:52 My wife travels with me.
01:16:54 When my kids need me, I spend more time with them.
01:16:58 And before I used to prioritize everything around work.
01:17:02 When I had a little bit of time,
01:17:03 I would dole it out to my family.
01:17:06 Now, when my family needs something, really needs something,
01:17:09 I drop everything at work and go to them.
01:17:12 And then in the time remaining, I allocate to work.
01:17:15 But one’s family is very understanding.
01:17:19 It’s not like they will take 50 hours a week from me.
01:17:23 So I’m actually able to still work pretty hard,
01:17:27 maybe 10 hours less per week.
01:17:29 So I realized the most important thing in my life
01:17:32 is really love and the people I love.
01:17:36 And I give that the highest priority.
01:17:38 It isn’t the only thing I do,
01:17:40 but when that is needed, I put that at the top priority
01:17:45 and I feel much better and I feel much more balanced.
01:17:50 And I think this also gives a hint
01:17:53 as to a life of routine work, a life of pursuit of numbers.
01:17:58 While my job was not routine, it was in pursuit of numbers,
01:18:02 pursuit of can I make more money?
01:18:04 Can I fund more great companies?
01:18:07 Can I raise more money?
01:18:08 Can I make sure our VC is ranked higher and higher
01:18:12 every year?
01:18:13 This competitive nature of driving for bigger numbers
01:18:18 and better numbers became a endless pursuit
01:18:25 that’s mechanical.
01:18:26 And bigger numbers really didn’t make me happier.
01:18:31 And faced with death, I realized bigger numbers
01:18:34 really meant nothing.
01:18:36 And what was important is that people who have given
01:18:41 their heart and their love to me
01:18:42 deserve for me to do the same.
01:18:45 So there’s deep, profound truth in that,
01:18:48 that everyone should hear and internalize.
01:18:52 I mean, that’s really powerful for you to say that.
01:18:55 I have to ask sort of a difficult question here.
01:19:03 So I’ve competed in sports my whole life,
01:19:06 looking historically, I’d like to challenge some aspect
01:19:11 of that a little bit on the point of hard work.
01:19:15 That it feels that there are certain aspects
01:19:18 that is the greatest, the most beautiful aspects
01:19:22 of human nature is the ability to become obsessed,
01:19:27 of becoming extremely passionate to the point where yes,
01:19:31 flaws are revealed and just giving yourself fully to a task.
01:19:37 That is, in another sense, you mentioned love
01:19:40 being important, but in another sense,
01:19:42 this kind of obsession, this pure exhibition of passion
01:19:46 and hard work is truly what it means to be human.
01:19:50 What lessons should we take that’s deeper?
01:19:53 Because you’ve accomplished incredible things.
01:19:54 You say it chasing numbers,
01:19:57 but really there’s some incredible work there.
01:20:00 So how do you think about that when you look back
01:20:04 in your 20s, your 30s, what would you do differently?
01:20:10 Would you really take back some of the incredible hard work?
01:20:14 I would, but it’s in percentages, right?
01:20:19 We’re both computer scientists.
01:20:22 So I think when one balances one’s life,
01:20:25 when one is younger, you might give a smaller percentage
01:20:30 to family, but you would still give them high priority.
01:20:33 And when you get older, you would give a larger percentage
01:20:36 to them and still the high priority.
01:20:38 And when you’re near retirement, you give most of it to them
01:20:42 and the highest priority.
01:20:43 So I think the key point is not that we would work 20 hours
01:20:49 less for the whole life and just spend it aimlessly
01:20:52 with the family, but that’s when the family has a need,
01:20:56 when your wife is having a baby,
01:21:00 when your daughter has a birthday or when they’re depressed
01:21:05 or when they’re celebrating something
01:21:07 or when they have a get together or when we have family time
01:21:11 that it’s important for us to put down our phone and PC
01:21:14 and be a hundred percent with them.
01:21:18 And that priority on the things that really matter
01:21:23 isn’t going to be so taxing that it would eliminate
01:21:29 or even dramatically reduce our accomplishments.
01:21:32 It might have some impact, but it might also have
01:21:35 other impact because if you have a happier family,
01:21:37 maybe you fight less.
01:21:39 If you fight less, you don’t spend time taking care
01:21:43 of all the aftermath of a fight.
01:21:45 So it’s unclear that it would take more time.
01:21:48 And if it did, I’d be willing to take that reduction.
01:21:53 And it’s not a dramatic number, but it’s a number
01:21:56 that I think would give me a greater degree of happiness
01:22:00 and knowing that I’ve done the right thing
01:22:03 and still have plenty of hours to get the success
01:22:08 that I want to get.
01:22:09 So given the many successful companies that you’ve launched
01:22:14 and much success throughout your career,
01:22:17 what advice would you give to young people today looking,
01:22:25 or it doesn’t have to be young,
01:22:26 but people today looking to launch
01:22:28 and to create the next $1 billion tech startup
01:22:32 or even AI based startup?
01:22:34 I would suggest that people understand
01:22:39 technology waves move quickly.
01:22:42 What worked two years ago may not work today.
01:22:45 And that is very much case in point for AI.
01:22:49 I think two years ago, or maybe three years ago,
01:22:53 you certainly could say I have a couple
01:22:55 of super smart PhDs and we’re not sure
01:22:58 what we’re gonna do, but here’s how we’re gonna start
01:23:01 and get funding for a very high valuation.
01:23:05 Those days are over because AI is going
01:23:08 from rocket science towards mainstream,
01:23:11 not yet commodity, but more mainstream.
01:23:14 So first the creation of any company
01:23:19 to a venture capitalists has to be creation
01:23:22 of business value and monetary value.
01:23:26 And when you have a very scarce commodity,
01:23:29 VCs may be willing to accept greater uncertainty.
01:23:35 But now the number of people who have the equivalent
01:23:38 of PhD three years ago, because that can be learned
01:23:42 more quickly, platforms are emerging,
01:23:46 the cost to become a AI engineer is much lower
01:23:49 and there are many more AI engineers.
01:23:51 So the market is different.
01:23:53 So I would suggest someone who wants to build an AI company
01:23:57 be thinking about the normal business questions.
01:24:01 What customer cases are you trying to address?
01:24:06 What kind of pain are you trying to address?
01:24:08 How does that translate to value?
01:24:10 How will you extract value and get paid
01:24:14 through what channel and how much business value
01:24:18 will get created?
01:24:19 That today needs to be thought about much earlier upfront
01:24:24 than it did three years ago.
01:24:26 The scarcity question of AI talent has changed.
01:24:30 The number of AI talent has changed.
01:24:32 So now you need not just AI, but also understanding
01:24:37 of business customer and the marketplace.
01:24:41 So I also think you should have a more reasonable
01:24:48 valuation expectation and growth expectation.
01:24:52 There’s gonna be more competition.
01:24:54 But the good news though, is that AI technologies
01:24:57 are now more available in open source.
01:25:00 TensorFlow, PyTorch and such tools are much easier to use.
01:25:06 So you should be able to experiment and get results
01:25:11 iteratively faster than before.
01:25:14 So take more of a business mindset to this,
01:25:18 think less of this as a laboratory taken into a company,
01:25:23 because we’ve gone beyond that stage.
01:25:26 The only exception is if you truly have a breakthrough
01:25:29 in some technology that really no one has,
01:25:32 then the old way still works.
01:25:34 But I think that’s harder and harder now.
01:25:37 So I know you believe as many do that we’re far
01:25:41 from creating an artificial general intelligence system.
01:25:45 But say once we do, and you get to ask her one question,
01:25:50 what would that question be?
01:25:57 What is it that differentiates you and me?
01:26:02 Beautifully put, Kaifu, thank you so much
01:26:04 for your time today.
01:26:05 Thank you.