- 8 months ago
Alibaba at a Crossroads
Category
🤖
TechTranscript
00:00Musique
00:44Bonjour everyone, there are people who are shaping our future, there are people who are making history and there are
00:53people who are really building the future of our lives.
00:58Please, welcome Joe Tsai, Chairman of Alibaba.
01:05Thank you.
01:35When it comes to the very difficult decisions.
01:38Joe, it's a real pleasure and an honor to have you.
01:43The pleasure is mine Maurice. Every time I spend time with you I have a lot of fun and also
01:48an absolute honor to be here on this stage.
01:51I understand there's quite a few famous people that will come later this afternoon.
01:57More famous than you?
01:59Absolutely, definitely a lot more famous.
02:02Oh no, you think that Jensen Wang is more famous?
02:05I never heard of him.
02:07Never heard of him.
02:08And President Macron?
02:11He I heard of.
02:13Yeah.
02:14Okay, so we will do a flashback.
02:18Yeah.
02:19Like in the movies.
02:21Yeah.
02:21And we will go back in 1999.
02:251999 in Hangzhou, you met Jack Ma.
02:30You know that he's a very good friend of mine.
02:33Of course.
02:34And you met him in his apartment, which is known today as a little bit the garage of Steve Jobs.
02:46And soon after, you took the leap to join Alibaba.
02:52Can you tell us a little bit how this happened?
02:56It was in 1999, I think it was the month of May.
03:02And I walked into this second story apartment in Hangzhou, China.
03:06It's a city that I have never visited before.
03:09And I just saw there are shoes in front of the apartment.
03:15And then I walk in, there's Jack himself.
03:18But then there were about 10, 12 engineers and customer service people that were working very, very hard.
03:27They were launching a website, Alibaba.com.
03:32And at the time, the site's function was to display the Chinese manufacturers and trading companies online.
03:41So Alibaba's first website was in English language.
03:44And it was a way to market Chinese companies to overseas buyers.
03:50It was a B2B site for wholesale trading.
03:53So very early on, Alibaba's first business was involved in international trade.
03:58But I have to admit, I didn't understand what they were doing quite as much.
04:03The impression that Jack made on me was whenever he spoke, his team listened.
04:10And I thought, this is an incredible leader, someone who has tremendous charisma, as you know.
04:18I also had a little bit of a personal motivation to try to get involved because I was doing investing.
04:25And I had trained as a lawyer, so I went from law to investing.
04:31But as an investor, you don't really get into the operations of a company.
04:36You really don't understand until you're on the side of management.
04:41So I was looking for an opportunity to where it's an intersection of tech, operations, and a leader like Jack.
04:50Jack has been here in Paris, and he has been interviewed by me in this very room.
04:57So you are sitting exactly where I was.
05:01What is interesting, and this is a name that we will share with all the startups,
05:06is that you are investing in business you don't understand.
05:09That's good.
05:10For all the startupers who are in the room who are looking for investors,
05:14I think we should give them your email address.
05:17You don't need to understand the business plan.
05:20You need to understand the people.
05:22Who are you trying to invest in, and who are you following?
05:24If you really are comfortable with the people, and there's that confidence and trust,
05:28I think everything else will take care of itself.
05:31Because a lot of you are founders in the audience.
05:34The business plan will never go according to plan.
05:37That's what you know.
05:38It's never going to be what you told investors last year.
05:41It always will weave into different directions.
05:46And the important thing as a founding team is adaptability, flexibility, and also have the vision to see what's coming.
05:55Because the world changes around you as well.
05:57It's not going to be the same all the time.
05:59What is interesting is that you were 18 people, 18 co-founders, and you were the only known China mainland
06:09citizen.
06:10I was born in Taiwan.
06:12You were born in Taiwan, and you were, I think, the only one who has been educated in the Western
06:19world.
06:20Right, right.
06:22How was the cultural gap, and how have you adjusted to those people that you didn't know?
06:32Yeah.
06:33Well, a little bit about the background is I grew up in Taiwan, but my parents came from the mainland,
06:39from mainland China.
06:40They were from Shanghai, and they moved to Taiwan.
06:44And at the time when I met Jack, I was living in Hong Kong.
06:49So I traveled from Hong Kong to Hangzhou to see him.
06:52So the first thing he told me was, Joe, I am not sure about people like your type.
06:58You're from the finance world.
06:59You're investing.
07:00But the worst thing is, I don't like Taiwanese people.
07:04I don't like Hong Kong people because I can't trust them.
07:07And I certainly do not like the Shanghainese.
07:09Because there's this rivalry between the big city, the metropolitan city, Shanghai, and Hangzhou,
07:15which at the time was a relatively small tourist city because there's a beautiful West Lake.
07:22Today, of course, Hangzhou is the Silicon Valley of China.
07:25It's the software capital of China.
07:27Partly because of Alibaba's development and all that.
07:30But he just said, don't like the Shanghainese, don't like Hong Kong people, and don't like Taiwan.
07:34And they don't like you.
07:37Yeah.
07:37Well, I thought I'm dead because three strikes, I'm already out.
07:41Born in Taiwan, living in Hong Kong, parents from Shanghai.
07:45And I think it speaks to his very sort of salt of the earth, down to earth nature, being someone
07:53that's natively born from Hangzhou.
07:55So fast forward, Alibaba became a very large and impressive company.
08:04And in September 2023, you decided to step in the role of chairman, despite the fact that you announced that
08:14you will try to phase yourself out of the company.
08:18Interesting.
08:19So, somebody who doesn't like Taiwanese, big cities, et cetera, is taking you on board.
08:29And you, you want to phase out, and you step as chairman.
08:33Can you explain a little bit what happened?
08:34Yeah.
08:36But this is maybe around, sort of at the tail end of COVID.
08:42And I had already been with the company for something like 24 years.
08:47And I thought, well, I did buy a basketball team.
08:50You know, maybe we talk about that later.
08:52I own an NBA team.
08:54It's called the Brooklyn Nets.
08:57And, you know, that's my passion.
09:00My passion is in sports.
09:01I wanted to spend more time with the family and with sports.
09:04So, I did sage the company.
09:06I wanted to phase myself out.
09:09I remained on the board of directors of the company.
09:13But then, things changed very quickly.
09:16It's because Alibaba went through a period of huge ordeal.
09:20We got caught in the confluence of competition, very stiff competition, COVID, and regulation,
09:28the regulatory sort of scrutiny within China.
09:32But also, externally, we were facing a lot of geopolitical, you know, the environment today is still very difficult.
09:40Very challenging.
09:41And so, the company, I think I saw a company that kind of lost its direction a little bit.
09:49I think we have expanded too big.
09:52When the board decided that they wanted to put in a new CEO, his name is Eddie Wu, Alibaba's current
09:58CEO.
10:00They said, Eddie is a young guy, you know, relative to us.
10:05He's in his 40s.
10:06And they said, we need to have someone who has been around for a long time, who know the people,
10:13know the culture of the company, to come back and maybe as chairman.
10:18So, at the time, you know, my feeling was, we have a saying at Alibaba, if not now, when?
10:28If not me, who?
10:31And that was the moment.
10:32I thought I could contribute.
10:34I could help the company.
10:35And I knew Eddie because he's one of the founders of the business.
10:39So, I've known him for over 20 years.
10:42And I know I could work with him.
10:44That's when I decided to come back.
10:46Great.
10:47So, today, people are saying that since that moment, Alibaba is going through a huge transformation.
10:57And it looks like a little bit like a rebirth.
11:01So, can you tell us what the company is today and where it is heading?
11:08Yeah.
11:10I think it's a rebirth is probably a little bit too exaggerated in the sense that when you rebirth, you
11:19acquire a new religion.
11:21We did not acquire a new religion.
11:23Our religion remained the same.
11:24Our DNA of innovation, the culture of Alibaba is still very much of an innovative culture.
11:32And we just got a little bit too big.
11:35So, the first thing we did was to shrink the size of our balance sheet.
11:39We sold a lot of investments in, for example, in physical retail.
11:43We owned one of the largest hypermarts in China, which we sold to shrink down the balance sheet and also
11:51think harder about capital management.
11:53More focused.
11:54Much more focused.
11:55So, our CEO is steering the company toward two main businesses.
12:02One is e-commerce.
12:04The other one is cloud computing with a heavy AI element injected into our cloud business.
12:11So, today, Alibaba is about those two things.
12:15And looking out into the future, what we want to do is, in an e-commerce business, we're a consumer
12:23-facing business, right?
12:25So, for us, the user is the most important constituency for us to serve.
12:32When a user comes to use our app, the experience is everything.
12:38So, we say user first.
12:39That's our first operating principle.
12:41The second operating principle is that we want to have AI everywhere.
12:45AI both within the company to help our efficiency productivity-wise, but also AI infused into every app that we
12:54make for consumers.
12:56As well as, obviously, using our cloud computing service to support the AI applications for enterprises and developers.
13:06Yes.
13:07AI is front and center at Alibaba.
13:11It's something which is fundamental.
13:15And you have even decided to create your own LLM and some foundational models.
13:26So, what drove a commerce-focused business to move that far in AI and not only to use AI to
13:39improve the business?
13:40It looks like it's really two very different businesses.
13:44That's a really great question.
13:46I actually personally never thought about that question until earlier in the green room.
13:52You asked me, you're an e-commerce company.
13:54Why are you doing LLMs, doing foundational models?
14:00It's because the Alibaba core business is all about data.
14:05We capture data.
14:06We cleanse data.
14:08We store data.
14:08We manage data.
14:09We analyze data.
14:10So, very early on, the core technology development is on how to manage the massive amounts of data.
14:18I'm talking about tens of thousands of petabytes of data in our ecosystem.
14:25And it's not just the storage aspect of the data.
14:30It's every day.
14:31There's new users coming in.
14:33There's hundreds of millions of transactions every day coming in.
14:37So, very early on, Alibaba has developed a distributed computing architecture we call Max Compute,
14:44which today is the core basis of our cloud computing business.
14:48What does Max Compute do?
14:49It's a data warehouse and data analytics platform.
14:52And that's what we were good at.
14:57And fast forward to around 2020.
15:01So, this is before chat GPT came out, because GPT came out in November of 22.
15:06So, two years prior to that, our scientists have already started to develop using massive amounts of data,
15:13because they saw the advantage of training a model using the transformer architecture,
15:21but with massive amounts of data that we have access to.
15:25And if you think about a model, it's just an expression.
15:28It's an abstraction of computing power and massive amounts of data.
15:34And that's how we got into the LLM sort of development.
15:38But, of course, thanks to OpenAI, they have woken up the entire world about the power of this transformer architecture
15:49and the future of generative AI.
15:51And every large company jumped into it.
15:56So, from 2022 to around a year ago, there are so many companies in China and also around the world
16:03that say that they developed their own large language model, the foundational model that's going to rule the world.
16:11But over time, there's a little bit of consolidation.
16:15And today, I'm very proud to say that our development team have developed one of the more competitive LLMs in
16:24the Chinese market.
16:26When it comes to open source, we're actually competitive.
16:30We will come to open source a little bit later.
16:32Yeah.
16:32I would like to insist a little bit on these LLMs.
16:37You have built the LLMs in Chinese, in Mandarin.
16:41Is it that different to do it in other Western languages?
16:48Yeah.
16:49So, let me explain it this way.
16:52Mandarin or also…
16:53Yeah.
16:54Yeah.
16:54Let me explain it this way.
16:56There's no really language…
16:57I think it's important that the regions understand.
16:59I think many of you in the audience probably already know the answer.
17:02There's really no language difference in LLM because available data is in all forms of language.
17:08Yes, I know.
17:09And when you think about language, whether it's in Chinese or English or French, it's simply
17:14a representation of knowledge.
17:16The basic laws of physics, math, logic, it can be expressed in different languages.
17:24So, when you train a model with different language, the underlying logic is similar.
17:30Yeah.
17:30So, but the most important thing is in China, English language data is also widely available.
17:37It's all over the internet.
17:39So, there's no difference when you train an LLM in China because Chinese companies have
17:45access to a wealth of data around the world that's basically open source, right?
17:50Yes.
17:50Wikipedia is open.
17:52Same thing in Western companies, whether it's a French company or American company.
17:59And, you know, recently I talked to people in the Middle East.
18:01They want to have developed their own LLMs.
18:03They have data from around the world.
18:06They're also training in different languages.
18:09I am sure some of these American models have excellent question, you know, query and answer
18:18in Chinese.
18:19And there's really not a difference simply because language is simply a representation of
18:25knowledge.
18:25And it's the knowledge that's being trained into these models.
18:30Yes.
18:30It is one, zero, beats, et cetera.
18:35So, at the end of the day…
18:37Tokens.
18:37Exactly.
18:39Whatever that means, right?
18:40Who cares?
18:42But you have been an incredible advocate at very early on, on open source and open
18:49AI.
18:50Yeah.
18:50It's interesting because normally open AI was meant to be open AI.
18:56Yeah.
18:56And to be AI open to the world.
18:59And at the end of the day, they decided to be closed.
19:03And you are supposed to be closed and you decided to be open.
19:09Yeah.
19:09Interesting.
19:10Yeah.
19:11How did this happen?
19:12And why did you believe that it was a step that you had to take?
19:17And do you still believe that this is the right step?
19:20Well, I would have to credit the researchers and scientists within Alibaba.
19:26You have some.
19:27We have a few of them.
19:28Not that many though.
19:30Not as many as some of the other companies.
19:32But they really espouse the open source.
19:35And it's for a couple of reasons.
19:38Number one, open source, you develop a community of developers.
19:44And they can come and improve your technology, right?
19:48So once you build up an ecosystem, a very rich ecosystem, your technology can get better.
19:53That's number one.
19:54And number two, with open source, it actually kind of democratizes the usage of AI.
20:00People can just download your model and they can use it.
20:03They can develop on top of it.
20:04They could deploy it in their whatever application they have.
20:07They could develop agents on top of it.
20:10And it actually proliferates applications.
20:13And I think AI is useless if you don't have applications.
20:16And that's why we say when we open source, there's going to be more and more applications.
20:23People will use them more.
20:24Now how do we, as a company, right?
20:29Companies have to make profits, general revenue will make profits.
20:33So the way we benefit from open source is that with open source, more proliferation,
20:42more AI applications, it will drive the demand for AI.
20:46It will drive training needs, but of course we see the future, a lot of needs in inference.
20:51And we run a cloud computing business.
20:53What happens when people use AI, they need to have compute, and we provide that service
20:58to them.
21:00One thing that we have seen in China is an incredible level of creativity when it comes to AI.
21:09There is two small operations which are well known, DeepSeek and Manus.
21:19And by the way, we will have Taozan on Friday coming to speak about Manus.
21:27Can you, from your point of view, explain to us what is making the Chinese scene so buoyant,
21:38so creative, and so innovative when it comes to AI?
21:44When you look at the Chinese market, starting with some of the early internet companies,
21:50China has always been very good at adopting, iterating, and adopting new technology, especially
21:58in the consumer internet space.
22:01You go to China, all the consumer internet apps, it's not Google or YouTube or Facebook
22:08or Instagram, it's the Chinese versions of those, and in fact they're actually better.
22:14They provide a better user experience.
22:16So I think China is a very application-rich ecosystem, and companies are very willing to
22:23adapt these applications.
22:25So it's important to have that consumer applications.
22:31So I think having a very vibrant consumer internet space really could drive innovation.
22:40And the other thing is, in China the engineers have this very competitive culture.
22:49So I'll tell you a story.
22:51When DeepSeq came out, this was January, middle of January, it was close to Chinese New Year.
22:59So all of our engineers and researchers had planned to go home.
23:03Chinese New Year is the biggest holiday in China, like Christmas.
23:06Everybody goes home, goes back to their families, travel, all that.
23:13And DeepSeq hit, we read their research papers, we said, holy cow, or we swear a little bit.
23:20How come…
23:21You do that?
23:21Yeah.
23:22The Chinese do.
23:23And Chinese do that too.
23:24So how come we have not, we are fall behind, but we knew we were doing the same things.
23:31You have not said shit.
23:32We do say that as well.
23:34And the F word as well.
23:37But what happened was our engineering leads decided, they said, cancel your Chinese New
23:44Year holiday.
23:45Everybody stay in the company.
23:47Sleep in the office.
23:48We're going to accelerate our development.
23:51So within a few weeks, we launched our version, which is the Quen series of models, which is
24:00not bad.
24:01It's quite competitive.
24:03We also decided to open source of very different versions with smaller parameter models.
24:10So that's the culture, right?
24:12It's a very competitive culture, and I think maybe in other places, if you tell your employees
24:21to cancel their holidays and stay in the office…
24:24Oh, in France, you can do that all the time.
24:27They do that.
24:28They are happy to cancel, including their weekend, to work a little bit hard.
24:34It's easy.
24:36It's easy.
24:36More than 35 hours a week.
24:39Yeah, it's common.
24:41I think if there is a clear vision of what you want to do, then people can get motivated.
24:46I also hear stories of French companies, people working overtime here.
24:51No, no, no.
24:52I'm just kidding.
24:53Yes, this happens, and more than we think, and there are a lot of people who are highly
24:57committed.
24:58You mentioned NBA, and you own, I don't know if he's great, that team.
25:06Are you happy with the Brooklyn?
25:08I would say we're rebuilding right now.
25:12Okay.
25:13Great.
25:13So, and there is Alibaba, and we have seen that you were extremely present at the Paris
25:23Olympics.
25:24And my understanding is that if the people around the world have been able to watch the
25:34games, it was thanks to Alibaba.
25:38Can you explain?
25:39Right.
25:40So, Alibaba has been a long-term partner for the last three Olympic cycles.
25:46We partner with the IOC, International Olympic Committee, on supporting all of their cloud
25:54technology needs.
25:56One of the areas that we focus on is broadcasting.
25:59So, every Olympic Games, globally, it's a most-watched game, you know, with billions
26:05of people watching, but if you think about the broadcaster's dilemma, they have to bring
26:12their people, the cameras, the trucks, everything, come to the site.
26:17It's from, you know, 180 different countries, and that gets too crowded.
26:22The hotels will not be able to accommodate all the camera people.
26:25And so, what the Olympic IOC have done is they centralized the filming of the games, and they
26:35take all of that footage, put it onto the cloud.
26:40So in the past Paris Olympics, there was about 11,000 hours of games footage.
26:4611,000, that's a lot of data.
26:49I have not seen 11,000 hours, but if you say so, I believe you.
26:55So, we—I'm sure you've seen quite a few hours, but—
27:01A few.
27:03I mean, all of us binge-watched Netflix, so we've seen lots of hours.
27:08So we provide the cloud infrastructure for broadcasters.
27:12They don't need to send their people.
27:13They could just go—let's say you're based in Brazil, or Japan, or Germany.
27:19You just go onto the cloud, edit your—produce your film footage, and then broadcast it back.
27:24So last Olympics was the first time cloud technology had overtaken satellite broadcasting technology.
27:33And that technology was all supported by Alibaba.
27:36And the other thing is, we actually use our AI to make the game presentation more exciting.
27:44Maybe we can roll a tape if I can look at the video.
27:47Sure.
27:49I brought a tape.
27:50I hope—I hope it works.
27:54I hope it works.
27:55It works.
28:06You've got three giants as we see the final site.
28:10Grandidia and Yang have been on that end one.
28:15Huge moment.
28:17That's AI.
28:18That's our AI.
28:23So you are taking the images—you are asking AI to take the images and to turn them into
28:30all directions in order that the people can see the action from very different angles.
28:38That's right.
28:39And the old technology is you have to bring 100 cameras all over the field, and then you
28:44need to stitch it together with the rendering.
28:46It's very compute-intensive, but this is very, very light, very efficient.
28:50and we're very proud to have helped the Paris Olympics to make the viewership really a great
29:00sporting experience.
29:01And the other thing is, congratulations to France for winning the rugby medal.
29:05Yes, thank you for showing that one instead of when we have lost.
29:11Yeah, there was another clip.
29:13No, no, no.
29:14This one was not good.
29:15Yeah.
29:15This one is very expressive.
29:16And by the way, I want to congratulate the PSG fans in Paris for winning the championship.
29:22Yeah.
29:27They have done a great, great job, and it has been a long journey, and they deserve a rose
29:35of applause because they have been absolutely fantastic, and they have shown that a team
29:40is far better than a star.
29:43Yeah.
29:44N'est-ce pas?
29:44That's good.
29:46Thank you.
29:46Joe, I would like to thank you very warmly.
29:49Thank you.
29:50We will have a conversation with somebody you know well.
29:53Yeah.
29:56And I think you will meet him while we leave the room.
30:00I hope he doesn't say bad things about me.
30:03That was the subject.
30:04Yeah.
30:04And we have decided that we will say, okay, now that Joe has spoken, can you tell us a
30:10little bit about all the drawbacks, all the failures, and all the mistakes of Joe?
30:15Yeah.
30:16But I also wanted to say, later on, you will have Jensen coming on stage.
30:21Yes.
30:22I'm one of his biggest fans.
30:23I'm his admirer, and I think he's incredible.
30:28He's great.
30:30So it's great for you to be able to have him on the stage.
30:32He must be incredible because he has now the company with the largest marketer
30:40cap in the world.
30:41I was telling our communications people, they have to make sure, you know, when senior
30:48people get on stage, they say something, it's a mistake, the market cap drops, right?
30:52He has a lot of market cap to protect than the risk I'm taking here.
30:58So now you're the adult in the room of Alibaba.
31:04Well, young at heart, just like you.
31:07Young at heart.
31:08Merci.
31:08Thank you.
31:09Fantastic.
31:10Yes.
31:11Thank you very much.
31:11Joe, thank you.
31:13Thank you so much.