1
00:00:02,720 --> 00:00:16,400
Speaker 1: Bloomberg Audio Studios, Podcasts, Radio News.

2
00:00:18,480 --> 00:00:21,079
Speaker 2: Hello and welcome to another episode of The Odd Laws podcast.

3
00:00:21,200 --> 00:00:23,360
Speaker 3: I'm Joe Wisenthal and I'm Tracy Alloway.

4
00:00:23,480 --> 00:00:25,200
Speaker 2: So we were in a Hong Kong recently. That was

5
00:00:25,200 --> 00:00:25,680
Speaker 2: a lot of fun.

6
00:00:25,760 --> 00:00:26,600
Speaker 4: I'd love to be back.

7
00:00:26,880 --> 00:00:29,920
Speaker 3: Yeah, I haven't been back for four years. Not much

8
00:00:29,920 --> 00:00:32,159
Speaker 3: has changed. Actually, I was kind of surprised and you

9
00:00:32,200 --> 00:00:35,280
Speaker 3: expected LUs than I expected. But I am really glad

10
00:00:35,320 --> 00:00:37,800
Speaker 3: we went back because obviously one of the big talking

11
00:00:37,840 --> 00:00:41,919
Speaker 3: points in markets right now is competition between US versus

12
00:00:42,040 --> 00:00:46,040
Speaker 3: Chinese AI. And we finally got a chance to talk

13
00:00:46,120 --> 00:00:49,880
Speaker 3: to a couple of high level executives of Chinese tech

14
00:00:49,920 --> 00:00:53,520
Speaker 3: companies who are actually making all the big capital allocation

15
00:00:53,720 --> 00:00:55,400
Speaker 3: decisions when it comes to the AI race.

16
00:00:55,560 --> 00:00:57,800
Speaker 2: Right, it felt like we're in this moment where there's

17
00:00:57,840 --> 00:00:59,680
Speaker 2: been the I mean, the way I think about it,

18
00:01:00,000 --> 00:01:03,120
Speaker 2: there's been Chinese internet giants, but they concentrated on China, right,

19
00:01:03,280 --> 00:01:05,680
Speaker 2: there's been American internet giants that basically had the rest

20
00:01:05,680 --> 00:01:08,200
Speaker 2: of the world. And whether we're talking about AI or

21
00:01:08,240 --> 00:01:10,920
Speaker 2: self driving cars, we're going to see the first sort

22
00:01:10,920 --> 00:01:15,400
Speaker 2: of like real head to head battle on internet companies specifically,

23
00:01:15,959 --> 00:01:18,160
Speaker 2: and where they're like competing for playing the same game

24
00:01:18,280 --> 00:01:20,160
Speaker 2: on some of the same markets. And of course we

25
00:01:20,240 --> 00:01:23,720
Speaker 2: know American companies can use AI models built by China,

26
00:01:23,880 --> 00:01:26,040
Speaker 2: et cetera, and so there are all kinds of options

27
00:01:26,080 --> 00:01:28,600
Speaker 2: for people. So it's like really interesting to say, like, Okay,

28
00:01:28,840 --> 00:01:30,960
Speaker 2: this clash is actually like it's happening.

29
00:01:31,040 --> 00:01:33,280
Speaker 3: It's a good time to talk to a Chinese tech

30
00:01:33,319 --> 00:01:34,160
Speaker 3: executive for.

31
00:01:34,080 --> 00:01:36,039
Speaker 2: Sure, that's right. So the reason we were back in

32
00:01:36,080 --> 00:01:39,120
Speaker 2: Hong Kong is because we were at the Bloomberg invest Conference.

33
00:01:39,160 --> 00:01:41,760
Speaker 2: Though we also through an Odd Lots trivia night while

34
00:01:41,760 --> 00:01:42,440
Speaker 2: we were in our.

35
00:01:42,560 --> 00:01:45,640
Speaker 3: First non US overseas all Thoughts quiz night.

36
00:01:45,920 --> 00:01:47,760
Speaker 2: Yeah, that was a lot of fun and I'm sure

37
00:01:47,800 --> 00:01:49,880
Speaker 2: we'll come back and do that again. But we were

38
00:01:49,920 --> 00:01:52,320
Speaker 2: at the Bloomberg invest conference and so we had the

39
00:01:52,440 --> 00:01:55,000
Speaker 2: chance to speak with the CFO of by Do, Henry Hoo.

40
00:01:55,200 --> 00:01:57,920
Speaker 2: So check it out. We truly have the perfect guest

41
00:01:58,000 --> 00:02:01,320
Speaker 2: we're going to be speaking with by Do CFO Henry. So, Henry,

42
00:02:01,640 --> 00:02:03,600
Speaker 2: thank you so much for coming on.

43
00:02:03,400 --> 00:02:05,200
Speaker 5: Outlot, thanks for having me.

44
00:02:05,520 --> 00:02:08,079
Speaker 4: And it's a great season, I think Hong Kong and

45
00:02:08,120 --> 00:02:10,440
Speaker 4: it definitely is great to see in both Joan Tracy

46
00:02:10,520 --> 00:02:10,799
Speaker 4: thank you.

47
00:02:11,160 --> 00:02:12,960
Speaker 2: Very nice of you to say. So why do we

48
00:02:13,040 --> 00:02:16,040
Speaker 2: start with this? You know, obviously I feel like half

49
00:02:16,080 --> 00:02:18,840
Speaker 2: the conversations are probably about AI these days. But within AI,

50
00:02:19,680 --> 00:02:22,680
Speaker 2: BYD is a full stack player, right You have cloud,

51
00:02:22,720 --> 00:02:25,000
Speaker 2: you have the application layer, you have your own chips

52
00:02:25,000 --> 00:02:28,680
Speaker 2: and of course your own model. As the CFO, you

53
00:02:28,800 --> 00:02:32,839
Speaker 2: might have to think about prioritization, et cetera. Is there

54
00:02:32,880 --> 00:02:36,840
Speaker 2: one layer of the stack that you feel is a

55
00:02:36,960 --> 00:02:41,000
Speaker 2: must win for by do when you think about resource allocation?

56
00:02:41,160 --> 00:02:43,120
Speaker 2: Is there a layer where it's like, okay, this is

57
00:02:43,160 --> 00:02:45,000
Speaker 2: an area where we have to win.

58
00:02:45,320 --> 00:02:46,080
Speaker 5: Thank you so much.

59
00:02:46,120 --> 00:02:48,920
Speaker 4: And I think you'll probably put the tough question in end,

60
00:02:48,960 --> 00:02:51,160
Speaker 4: but I think it's probably the most difficult question to

61
00:02:51,200 --> 00:02:54,519
Speaker 4: start with. So I think the very unique thing today

62
00:02:54,600 --> 00:02:57,560
Speaker 4: is I think the entire AI has been shifting from

63
00:02:57,680 --> 00:03:02,280
Speaker 4: infrastructure to applications and from model to agents. I think

64
00:03:02,320 --> 00:03:05,120
Speaker 4: that's actually the bad job. I think within that is,

65
00:03:05,120 --> 00:03:07,480
Speaker 4: frankly speaking, right now, is very difficult to say at

66
00:03:07,480 --> 00:03:11,400
Speaker 4: this moment which parties must have because in my view,

67
00:03:12,040 --> 00:03:15,200
Speaker 4: the chap is infrastructures. You need to have a great

68
00:03:15,240 --> 00:03:18,920
Speaker 4: model to bridge the capability. The cloud is a deployment

69
00:03:19,200 --> 00:03:22,600
Speaker 4: of that capability, and obviously the monitization and all the

70
00:03:22,800 --> 00:03:25,440
Speaker 4: ROI questions, especially for the people like me, I say, oh,

71
00:03:25,440 --> 00:03:28,720
Speaker 4: we focus on that is on application layers. So without

72
00:03:28,800 --> 00:03:32,480
Speaker 4: any of that, this ROI cannot work. So to answer

73
00:03:32,480 --> 00:03:34,760
Speaker 4: a question, I think the key thing if I have

74
00:03:34,840 --> 00:03:38,120
Speaker 4: to pick one, is Cloud, because Cloud at this moment

75
00:03:38,400 --> 00:03:41,440
Speaker 4: is a platform. You cannot only hosting Arnie, which is

76
00:03:41,480 --> 00:03:43,520
Speaker 4: our own model, but also I can work in a

77
00:03:43,600 --> 00:03:46,440
Speaker 4: very open to hosting other models. And the my chip

78
00:03:46,600 --> 00:03:50,080
Speaker 4: which connecting to my cloud platform can also help on

79
00:03:50,280 --> 00:03:53,040
Speaker 4: inference because right now the play training is important, but

80
00:03:53,160 --> 00:03:56,000
Speaker 4: eighty percent of the incremental demand today on a token

81
00:03:56,320 --> 00:03:59,200
Speaker 4: are inference related. I think this part of the full

82
00:03:59,240 --> 00:04:01,440
Speaker 4: picture is what I want to emphasize. But you know,

83
00:04:01,480 --> 00:04:03,280
Speaker 4: given a tough question, if I want a big one

84
00:04:03,440 --> 00:04:05,920
Speaker 4: as a student ABCD, I want big number C, which

85
00:04:05,960 --> 00:04:06,480
Speaker 4: is my cloud.

86
00:04:07,560 --> 00:04:09,200
Speaker 3: I'm going to ask you a question which I think

87
00:04:09,320 --> 00:04:11,960
Speaker 3: is going to become standard for financial journalists in the

88
00:04:12,000 --> 00:04:16,200
Speaker 3: same way we ask about headcount and expansion plans. What's

89
00:04:16,200 --> 00:04:20,600
Speaker 3: your token budget? Is it bigger than Joe's? I mean

90
00:04:20,640 --> 00:04:23,880
Speaker 3: the token the token budget for buy does or how

91
00:04:23,880 --> 00:04:26,159
Speaker 3: do you measure? I'll ask it in a slightly different way.

92
00:04:26,200 --> 00:04:29,600
Speaker 3: How do you measure what you're gaining from your token spend?

93
00:04:29,640 --> 00:04:31,040
Speaker 3: How do you measure productivity?

94
00:04:31,160 --> 00:04:31,320
Speaker 5: Yeah?

95
00:04:31,360 --> 00:04:36,440
Speaker 4: Sure, I want to categorize probably two buckets. One, if

96
00:04:36,720 --> 00:04:41,040
Speaker 4: we consume computing power to reach a higher level of

97
00:04:41,120 --> 00:04:45,520
Speaker 4: technology standard you know, the ADI, the how would model

98
00:04:45,760 --> 00:04:48,919
Speaker 4: form and also how harnys can be designed to deliver

99
00:04:49,000 --> 00:04:52,000
Speaker 4: better results, So the are R and D efforts. However,

100
00:04:52,279 --> 00:04:55,280
Speaker 4: as also a tech house, we also deliver those know

101
00:04:55,320 --> 00:04:59,039
Speaker 4: how to our external clients with different verticals. I think

102
00:04:59,520 --> 00:05:03,520
Speaker 4: if I'm measuring our internal consumption of the token, I

103
00:05:03,560 --> 00:05:06,440
Speaker 4: really want to like how better and how efficient our

104
00:05:06,520 --> 00:05:10,000
Speaker 4: technology can be developed. So that's on one side. However,

105
00:05:10,080 --> 00:05:12,520
Speaker 4: on the other end, what I think the ROI is

106
00:05:12,760 --> 00:05:17,240
Speaker 4: more relevant is how many real tasks that the open

107
00:05:17,279 --> 00:05:20,719
Speaker 4: claw and for example our own application called do mate

108
00:05:20,839 --> 00:05:23,599
Speaker 4: also is a real agency in human digits and other

109
00:05:23,640 --> 00:05:26,440
Speaker 4: things can do the task. So I think these two

110
00:05:26,440 --> 00:05:28,960
Speaker 4: different measurements are important in the way that right now

111
00:05:29,320 --> 00:05:33,159
Speaker 4: there are two things are better than last year when

112
00:05:33,200 --> 00:05:36,160
Speaker 4: it's a foundational model getting much better. And number two

113
00:05:36,480 --> 00:05:39,480
Speaker 4: is the framework i e. Open Claw and other things

114
00:05:39,800 --> 00:05:42,600
Speaker 4: can link up the foundation model capability to the rail

115
00:05:42,680 --> 00:05:46,560
Speaker 4: world task from chatting on something to doing something and

116
00:05:46,720 --> 00:05:50,600
Speaker 4: completing on something. I think completion part of the tokens

117
00:05:50,800 --> 00:05:54,080
Speaker 4: is more important today, and I think consumption internally will

118
00:05:54,080 --> 00:05:56,840
Speaker 4: actually encourage the people do that. But think about that,

119
00:05:56,880 --> 00:05:59,520
Speaker 4: even last year or year before, everyone is actually beefing

120
00:05:59,560 --> 00:06:02,200
Speaker 4: up budgets. I think that budget is all there, but

121
00:06:02,320 --> 00:06:03,960
Speaker 4: the completion utask is more important.

122
00:06:03,960 --> 00:06:06,040
Speaker 2: But let me just press you on this question a

123
00:06:06,040 --> 00:06:08,840
Speaker 2: little bit further. So let's say Tracy and I are

124
00:06:09,240 --> 00:06:12,840
Speaker 2: Let's say we worked for Baydo in the same department,

125
00:06:13,120 --> 00:06:16,479
Speaker 2: I don't know, some department of yours. Would we have

126
00:06:17,040 --> 00:06:20,520
Speaker 2: identical token budgets or would you have one way of saying,

127
00:06:20,680 --> 00:06:23,400
Speaker 2: you know what, Joe, Tracy is actually finding ways to

128
00:06:23,400 --> 00:06:26,160
Speaker 2: get more value out of AI than you are, So

129
00:06:26,200 --> 00:06:28,599
Speaker 2: I'm going to increase her budget. Like do you make

130
00:06:28,680 --> 00:06:32,440
Speaker 2: decisions like that? And do you have measurement techniques to

131
00:06:32,520 --> 00:06:35,520
Speaker 2: see like this person really should get ten times the

132
00:06:35,520 --> 00:06:38,320
Speaker 2: token budget of another person because they have figured out

133
00:06:38,320 --> 00:06:40,599
Speaker 2: how to get a lot of juice from the squeeze,

134
00:06:40,640 --> 00:06:41,159
Speaker 2: so to speak.

135
00:06:41,240 --> 00:06:43,560
Speaker 4: You know, Joe, given the question asking I seeing pro

136
00:06:43,680 --> 00:06:47,279
Speaker 4: next time, I give you multiple con Yeah, I think

137
00:06:47,600 --> 00:06:50,520
Speaker 4: right now, the technology evolved very fast. I think that's

138
00:06:50,560 --> 00:06:53,000
Speaker 4: the beauty part of AI. So we don't want to

139
00:06:53,040 --> 00:06:56,080
Speaker 4: constrain by us out by before thinking through something, we

140
00:06:56,240 --> 00:06:58,520
Speaker 4: just install certain policy by saying, you know, these are

141
00:06:58,560 --> 00:07:03,360
Speaker 4: employees with defined the token number by the titles or similarities.

142
00:07:03,760 --> 00:07:06,120
Speaker 4: I don't think that's the way it works. So I

143
00:07:06,160 --> 00:07:09,080
Speaker 4: think we want to more open and more nimbo in

144
00:07:09,120 --> 00:07:13,000
Speaker 4: a way that given enough token to the individuals to

145
00:07:13,120 --> 00:07:17,200
Speaker 4: empower their internal RD efforts. But on the other end,

146
00:07:17,360 --> 00:07:19,760
Speaker 4: we do have a lot of efforts to make sure

147
00:07:19,760 --> 00:07:22,800
Speaker 4: the token costs become dramatic coming down. I think the

148
00:07:22,840 --> 00:07:26,040
Speaker 4: cost is coming out very fast before you even think

149
00:07:26,040 --> 00:07:29,040
Speaker 4: about getting a policy. Maybe the unit cost is coming

150
00:07:29,080 --> 00:07:31,440
Speaker 4: down half in like a few weeks. So we need

151
00:07:31,480 --> 00:07:34,320
Speaker 4: to think about the speed and the cost and output

152
00:07:34,320 --> 00:07:37,360
Speaker 4: efficiency these three parameters as a package, not only on

153
00:07:37,400 --> 00:07:39,280
Speaker 4: the number of tokens. That's when I say, on the

154
00:07:39,320 --> 00:07:42,040
Speaker 4: other end, you feel very interesting facts. So right now

155
00:07:42,080 --> 00:07:44,520
Speaker 4: you know we are recruiting a lot of younger talents

156
00:07:44,560 --> 00:07:47,160
Speaker 4: to then even for Baydu, which is you know, twenty

157
00:07:47,240 --> 00:07:50,600
Speaker 4: plus years listed a public company. So my seeing is

158
00:07:51,240 --> 00:07:54,840
Speaker 4: the kids actually getting smarter than people expected, so they

159
00:07:54,880 --> 00:07:57,440
Speaker 4: were not wasted the tokens you'll give to them, so

160
00:07:57,480 --> 00:08:00,760
Speaker 4: they have a sensible judgment about what are the task

161
00:08:00,800 --> 00:08:03,800
Speaker 4: they need to prioritize because they're working out agents and

162
00:08:03,840 --> 00:08:07,320
Speaker 4: the models. The model actually helping them also prioritize autopaths

163
00:08:07,360 --> 00:08:09,960
Speaker 4: they have. So I think the power of the technology

164
00:08:10,000 --> 00:08:12,160
Speaker 4: today is it is not only at the tools. So

165
00:08:12,200 --> 00:08:15,120
Speaker 4: that's actually the key concept on to mention it is

166
00:08:15,160 --> 00:08:18,120
Speaker 4: not only a tool, it's a mindset, and the mindset

167
00:08:18,200 --> 00:08:21,480
Speaker 4: become automatic and more intelligent. That if people can work

168
00:08:21,520 --> 00:08:24,040
Speaker 4: on that well and have a new relationship with agents

169
00:08:24,120 --> 00:08:27,280
Speaker 4: and with a model, some of the old questions we

170
00:08:27,400 --> 00:08:29,680
Speaker 4: kind of struggle ourselves will be kind of diminished and

171
00:08:29,760 --> 00:08:30,400
Speaker 4: less important.

172
00:08:30,600 --> 00:08:33,880
Speaker 3: Okay, so no token maxing at BYDEO. But since you

173
00:08:33,960 --> 00:08:36,160
Speaker 3: brought up talent, one thing I'm very curious about is

174
00:08:36,200 --> 00:08:39,439
Speaker 3: we know the competition for the top engineers is so

175
00:08:39,520 --> 00:08:41,920
Speaker 3: intense right now, and in the US we see these

176
00:08:41,960 --> 00:08:45,760
Speaker 3: headlines where engineers are treated like sports stars. You know,

177
00:08:45,800 --> 00:08:49,520
Speaker 3: they're being treated for millions of dollars or whatever. What

178
00:08:49,679 --> 00:08:53,640
Speaker 3: is Bydeo's pitch to top talent, Like, if you're trying

179
00:08:53,640 --> 00:08:56,440
Speaker 3: to attract someone to the company, what is it you say,

180
00:08:56,480 --> 00:08:58,079
Speaker 3: to them that makes them want to work for buy

181
00:08:58,160 --> 00:09:00,640
Speaker 3: do versus another tech.

182
00:09:00,480 --> 00:09:03,679
Speaker 4: Firm's a great question, So let's bring a different perspective.

183
00:09:03,920 --> 00:09:07,400
Speaker 4: I think, you know, we are a technology company. Previously,

184
00:09:07,679 --> 00:09:10,440
Speaker 4: I think the priorities we empower our clients to be

185
00:09:10,520 --> 00:09:13,960
Speaker 4: more intelligent. We give them more technology tools to help

186
00:09:14,000 --> 00:09:17,080
Speaker 4: them to remove a move from the you know, the

187
00:09:17,120 --> 00:09:20,719
Speaker 4: traditional it to the cloud environment, you know, such as that.

188
00:09:21,320 --> 00:09:24,440
Speaker 4: But right now I think AI, especially for the big

189
00:09:24,480 --> 00:09:28,280
Speaker 4: corporation like us, also change us as well. We need

190
00:09:28,320 --> 00:09:32,320
Speaker 4: to think about the cultural change organization change not only

191
00:09:32,400 --> 00:09:35,839
Speaker 4: as an organization and a company, but also how AI

192
00:09:35,920 --> 00:09:39,440
Speaker 4: empower ourselves. So it's actually equally important to do something

193
00:09:39,480 --> 00:09:42,719
Speaker 4: for client versus think about new tools affecting ourselves.

194
00:09:42,920 --> 00:09:43,800
Speaker 5: So Trace is right.

195
00:09:44,040 --> 00:09:45,720
Speaker 4: I think there are a few things we actually make

196
00:09:45,800 --> 00:09:49,400
Speaker 4: a lot of different thinkings and some of the new initiatives.

197
00:09:49,440 --> 00:09:52,520
Speaker 4: First of all, we're probably among a few companies in

198
00:09:52,600 --> 00:09:56,040
Speaker 4: China still very open and even increasing the campus recruiting

199
00:09:56,200 --> 00:10:00,000
Speaker 4: and the focus on the younger talents and number two recents.

200
00:10:00,200 --> 00:10:03,280
Speaker 4: We also task the senior people not only look at

201
00:10:03,600 --> 00:10:06,840
Speaker 4: you know, the current reporting structure, but also in the

202
00:10:06,920 --> 00:10:10,480
Speaker 4: real mentor relationship with the younger, growing piece of the

203
00:10:10,559 --> 00:10:13,959
Speaker 4: human capital in the company. But more importantly, I think

204
00:10:14,200 --> 00:10:17,280
Speaker 4: it is really about giving people more autonomy to work

205
00:10:17,320 --> 00:10:20,600
Speaker 4: in a company, so more trust and more autonomy and

206
00:10:20,760 --> 00:10:23,720
Speaker 4: give them more real work. And you know, there's a

207
00:10:23,720 --> 00:10:26,520
Speaker 4: one concept called one person company, right, so we're very

208
00:10:26,520 --> 00:10:29,080
Speaker 4: happy to working with one person company because they actually

209
00:10:29,160 --> 00:10:31,360
Speaker 4: use our AI to very nicely and willing to pay

210
00:10:31,400 --> 00:10:35,280
Speaker 4: a lot of revenue to our products given the quality. However,

211
00:10:35,800 --> 00:10:38,720
Speaker 4: within the company, we'll also encourage people to be the

212
00:10:38,720 --> 00:10:42,680
Speaker 4: one person team so they can actually use the agents

213
00:10:42,760 --> 00:10:44,960
Speaker 4: to work on a lot of internal tasks. So internally

214
00:10:44,960 --> 00:10:46,480
Speaker 4: you have listen a little to call the doo Doo

215
00:10:46,760 --> 00:10:49,040
Speaker 4: in Chinese is a very kind of nicky name, which

216
00:10:49,080 --> 00:10:52,360
Speaker 4: is actually our internal kind of open clock, similar tools

217
00:10:52,400 --> 00:10:55,080
Speaker 4: and which actually enhancing people's efficiency. And the true point

218
00:10:55,080 --> 00:10:58,280
Speaker 4: I think give me more people, more autonomy, more trust

219
00:10:58,880 --> 00:11:01,559
Speaker 4: and the more room to grow and attracting new talents

220
00:11:01,559 --> 00:11:04,360
Speaker 4: I think equally all kind of very important to change yourselves.

221
00:11:04,360 --> 00:11:07,800
Speaker 4: But also you know, reporting lines and organized instructure need

222
00:11:07,840 --> 00:11:09,640
Speaker 4: to come with it to make sure that people can

223
00:11:09,720 --> 00:11:12,480
Speaker 4: deliver the results. And the last note, I want to

224
00:11:12,480 --> 00:11:15,680
Speaker 4: mention the key things that people see the application is important.

225
00:11:15,720 --> 00:11:17,439
Speaker 4: They can work on a full stack in by DO,

226
00:11:17,520 --> 00:11:20,120
Speaker 4: which is very unique value in the China tax space.

227
00:11:20,240 --> 00:11:23,440
Speaker 2: You know, it just occurred to me. American companies are

228
00:11:23,480 --> 00:11:27,000
Speaker 2: kind of becoming more Chinese in the sense that they're

229
00:11:27,200 --> 00:11:29,959
Speaker 2: doing more vertical integration. Like that's sort of a long

230
00:11:30,200 --> 00:11:33,960
Speaker 2: history here of sort of the whole thing, and now

231
00:11:33,960 --> 00:11:37,640
Speaker 2: we see one of this phenomenon is that every American

232
00:11:37,679 --> 00:11:40,520
Speaker 2: company like they want to even start designing and selling

233
00:11:40,520 --> 00:11:43,120
Speaker 2: their own chips, their own silicon, which is something that

234
00:11:43,200 --> 00:11:45,160
Speaker 2: you're doing your own business and you've had it for

235
00:11:45,200 --> 00:11:47,800
Speaker 2: a while, and I'm trying to wrap my head, like

236
00:11:48,160 --> 00:11:51,320
Speaker 2: what is the rationale? How much is it about just

237
00:11:51,760 --> 00:11:54,440
Speaker 2: wanting to be able to control your own fate more

238
00:11:54,840 --> 00:11:57,720
Speaker 2: and so wanting to control more of the supply chain

239
00:11:58,240 --> 00:12:03,800
Speaker 2: versus having it that optimally aligns with the model that

240
00:12:03,840 --> 00:12:07,320
Speaker 2: you're working on, because those are distinct priorities. So what

241
00:12:07,440 --> 00:12:10,319
Speaker 2: is the real rationale for having custom Seligan?

242
00:12:10,559 --> 00:12:13,920
Speaker 4: Yeah, I think thanks for the tech trend in the

243
00:12:14,000 --> 00:12:15,280
Speaker 4: past kind of year and two.

244
00:12:15,600 --> 00:12:18,000
Speaker 5: So if you look at the entire.

245
00:12:18,040 --> 00:12:21,800
Speaker 4: Computing power consumed, for example, last year, most of the

246
00:12:21,840 --> 00:12:26,160
Speaker 4: consumptions actually relating to the training of large foundation model,

247
00:12:26,200 --> 00:12:28,439
Speaker 4: but right now you know many of them going to

248
00:12:28,520 --> 00:12:31,439
Speaker 4: the influence and the complete tasks, and if you look

249
00:12:31,480 --> 00:12:34,080
Speaker 4: at the different stacks, I think right now we are

250
00:12:34,080 --> 00:12:37,600
Speaker 4: actually fitting to an incremental growing piece of the market

251
00:12:37,679 --> 00:12:40,360
Speaker 4: which is well defined with a clear boundary, which is

252
00:12:40,400 --> 00:12:44,600
Speaker 4: not focused on pre training for a very super scale

253
00:12:44,720 --> 00:12:48,760
Speaker 4: foundation model. However, the influence application is important. So to

254
00:12:48,800 --> 00:12:51,960
Speaker 4: a point, I think our cheir products appointing of cloud

255
00:12:52,200 --> 00:12:55,480
Speaker 4: focusing on the inference and application is a unique way

256
00:12:55,520 --> 00:12:58,600
Speaker 4: of we see the positive network effects. I think that's

257
00:12:58,840 --> 00:13:02,080
Speaker 4: where the area will wants and also given the issue

258
00:13:02,080 --> 00:13:05,160
Speaker 4: you mentioned, I think within that defined areas, we I

259
00:13:05,200 --> 00:13:07,880
Speaker 4: feel pretty confident regarding all the issues you mentioned, and

260
00:13:08,040 --> 00:13:10,680
Speaker 4: because on the supply and demand side, we can find

261
00:13:10,800 --> 00:13:14,000
Speaker 4: the good match within the emergent market category, within the

262
00:13:14,000 --> 00:13:28,520
Speaker 4: inference and application markets.

263
00:13:31,360 --> 00:13:35,320
Speaker 3: So hypothetically you could try to do everything right the

264
00:13:35,440 --> 00:13:38,800
Speaker 3: full stack, and I guess the capital investment you would

265
00:13:38,840 --> 00:13:42,800
Speaker 3: need to do that is also hypothetically unlimited at this

266
00:13:42,880 --> 00:13:45,000
Speaker 3: moment in time. And we hear these crazy numbers in

267
00:13:45,040 --> 00:13:48,400
Speaker 3: the US about the hyperscalers spending hundreds of billions of

268
00:13:48,440 --> 00:13:52,240
Speaker 3: dollars this year alone. But when you're looking at the

269
00:13:52,280 --> 00:13:54,240
Speaker 3: different parts of the business. So you have a very

270
00:13:54,280 --> 00:13:57,960
Speaker 3: mature internet search business, and then you have everything that

271
00:13:58,000 --> 00:14:01,760
Speaker 3: you're doing with AI, including in structure, how are you

272
00:14:01,800 --> 00:14:05,360
Speaker 3: actually allocating capital and then how are you actually I guess,

273
00:14:05,400 --> 00:14:09,839
Speaker 3: balancing that with returning capital at the same time to shareholders.

274
00:14:10,600 --> 00:14:13,640
Speaker 4: So I called it impossible triangle. So I kind of

275
00:14:13,880 --> 00:14:16,839
Speaker 4: scratched my head every few months, every few quick weeks,

276
00:14:16,880 --> 00:14:19,960
Speaker 4: depends on I also see the headline numbers brooke news

277
00:14:19,960 --> 00:14:22,400
Speaker 4: with other peers as well, So sometimes I make a

278
00:14:22,440 --> 00:14:25,040
Speaker 4: little bit kind of hesitating to make a statement, but

279
00:14:25,240 --> 00:14:27,400
Speaker 4: I just want to tell the facts and tell the views.

280
00:14:27,440 --> 00:14:29,760
Speaker 4: I want to separate them out. So in the recent

281
00:14:29,800 --> 00:14:32,400
Speaker 4: quarter earnings, we you know, we mentioned we actually solve

282
00:14:32,520 --> 00:14:37,280
Speaker 4: partially on these impossible triangles. One is our operating profit

283
00:14:37,520 --> 00:14:40,640
Speaker 4: increase almost doubled on a Q on Q base and

284
00:14:40,720 --> 00:14:43,640
Speaker 4: number two, our cloud revenue grew about seventy nine percent

285
00:14:43,680 --> 00:14:46,160
Speaker 4: on a WILDWI basis, which is almost double off the

286
00:14:46,200 --> 00:14:49,160
Speaker 4: wildwide growth rate for the cloud market in China. And

287
00:14:49,240 --> 00:14:52,680
Speaker 4: number three is since Q three last year, our operating

288
00:14:52,760 --> 00:14:56,040
Speaker 4: cash flow has turned positive. So positive operating cash flow,

289
00:14:56,280 --> 00:15:00,800
Speaker 4: incremental operating profits and the higher growth and the market. However,

290
00:15:00,960 --> 00:15:03,720
Speaker 4: my capbacks is not seeing kind of double even third

291
00:15:04,200 --> 00:15:07,080
Speaker 4: multiple times. So I think the way of resolving that

292
00:15:07,280 --> 00:15:09,760
Speaker 4: is as a CFO or as a management team of

293
00:15:09,800 --> 00:15:12,760
Speaker 4: a heavy CABACS investor AI tech company right now need

294
00:15:12,800 --> 00:15:15,560
Speaker 4: to find a way on one hand really drive the

295
00:15:15,600 --> 00:15:18,880
Speaker 4: growth but also keep the density of the investment into

296
00:15:18,920 --> 00:15:22,560
Speaker 4: air in a reasonable pacing. However, when you do that,

297
00:15:22,680 --> 00:15:25,360
Speaker 4: you need to keep a conscious regarding OURI and our

298
00:15:25,560 --> 00:15:28,800
Speaker 4: in mind to look at the entire cash cycle. For example,

299
00:15:28,840 --> 00:15:30,880
Speaker 4: every dollar we spend today we probably need to wait

300
00:15:30,920 --> 00:15:33,480
Speaker 4: for another probably twenty thirty forty months depends on the

301
00:15:33,520 --> 00:15:37,200
Speaker 4: category to get full cash back. And during that frame,

302
00:15:38,040 --> 00:15:41,960
Speaker 4: obviously there's a you know, price heights, memories, there's difficulty

303
00:15:41,960 --> 00:15:45,080
Speaker 4: on IDC centers and a huge spending our service. So

304
00:15:45,120 --> 00:15:47,520
Speaker 4: my point is as a SFO on every project, you

305
00:15:47,560 --> 00:15:49,520
Speaker 4: need to look at the entire life cycle, not only

306
00:15:49,520 --> 00:15:52,080
Speaker 4: at one time, but also the pacing important because the

307
00:15:52,120 --> 00:15:54,640
Speaker 4: foundation model R and D always taking a few months. Right,

308
00:15:54,680 --> 00:15:56,440
Speaker 4: So these are the things you need to keep our mind.

309
00:15:56,520 --> 00:15:59,400
Speaker 4: But my statement today is as by do we want

310
00:15:59,480 --> 00:16:04,040
Speaker 4: to invest probably in a more responsible manner to the shareholders,

311
00:16:04,440 --> 00:16:09,360
Speaker 4: but do not diminish our ambitious to investment into AI.

312
00:16:09,600 --> 00:16:12,440
Speaker 4: Keep the right density is important. But given the results

313
00:16:12,440 --> 00:16:14,320
Speaker 4: for this quarter, I think we kind of resolve that

314
00:16:14,400 --> 00:16:16,480
Speaker 4: at least for this quarter. So hopefully you can keep

315
00:16:16,520 --> 00:16:18,400
Speaker 4: on working on that and maybe, you know, a half

316
00:16:18,480 --> 00:16:20,640
Speaker 4: year later when we check on this point, we can

317
00:16:20,680 --> 00:16:22,600
Speaker 4: still keep on the same pattern, you know, high growths,

318
00:16:22,920 --> 00:16:26,080
Speaker 4: less dollars spend, but better II. I think that's probably

319
00:16:26,120 --> 00:16:27,120
Speaker 4: the angle we want to achieve.

320
00:16:27,160 --> 00:16:29,560
Speaker 2: Can I just mention something I'm very curious about? You know,

321
00:16:29,560 --> 00:16:34,480
Speaker 2: I'd say the heads of the American AI labs maybe

322
00:16:34,640 --> 00:16:39,080
Speaker 2: have varying degrees of AI psychosis. They have a lot

323
00:16:39,120 --> 00:16:43,520
Speaker 2: of worries that the what they call alignment research, et cetera.

324
00:16:43,720 --> 00:16:46,480
Speaker 2: Do you work on similar things or do you have

325
00:16:46,520 --> 00:16:48,520
Speaker 2: the same concerns and do.

326
00:16:48,480 --> 00:16:51,320
Speaker 6: You also have AI side, Like do you like you know,

327
00:16:51,440 --> 00:16:55,920
Speaker 6: for speaking of like trying to make money, Like do

328
00:16:56,000 --> 00:16:58,720
Speaker 6: you invest in or how much do you invest in

329
00:16:58,960 --> 00:17:02,240
Speaker 6: what they would call AI safety or alignment and essentially

330
00:17:02,720 --> 00:17:06,159
Speaker 6: making sure that the models that you're building don't go

331
00:17:06,359 --> 00:17:09,119
Speaker 6: rogue and always work on behalf of human flourishing?

332
00:17:09,400 --> 00:17:11,399
Speaker 2: Is that a thing that you allocate capital to?

333
00:17:11,600 --> 00:17:15,040
Speaker 4: Yeah, that's great question. So there's an emerging area of

334
00:17:15,200 --> 00:17:18,719
Speaker 4: example in this data sanity and all the kind of

335
00:17:18,920 --> 00:17:21,399
Speaker 4: post training efforts need to work on that. You know,

336
00:17:21,440 --> 00:17:24,399
Speaker 4: alignment obviously is one of that. But my point is

337
00:17:25,280 --> 00:17:27,840
Speaker 4: if you look at this new concept of harness, right,

338
00:17:27,880 --> 00:17:31,160
Speaker 4: it's not only about training and getting model on the leaderboard,

339
00:17:31,200 --> 00:17:34,600
Speaker 4: but also more importantly to measure the robustness and all

340
00:17:34,640 --> 00:17:38,119
Speaker 4: the things you mentioned. I think in the context in

341
00:17:38,240 --> 00:17:43,360
Speaker 4: China tech sector, the engineering has to be and has

342
00:17:43,480 --> 00:17:47,600
Speaker 4: been a good competitive advantage. So the harness from the

343
00:17:47,720 --> 00:17:50,879
Speaker 4: data flightwell to the alignment, to the data quality and

344
00:17:50,920 --> 00:17:54,640
Speaker 4: the labeling. I think the entire evil system has been robust.

345
00:17:54,640 --> 00:17:56,560
Speaker 4: For if you think about even in the mobile internet

346
00:17:56,600 --> 00:17:59,679
Speaker 4: work right, so as simple as data labeling to the

347
00:17:59,680 --> 00:18:02,560
Speaker 4: alignment tracks and post training and SFT, I think this

348
00:18:02,840 --> 00:18:05,080
Speaker 4: kind of the full chain of the capability in terms

349
00:18:05,080 --> 00:18:08,520
Speaker 4: of the talents and the pool of resources and the

350
00:18:08,560 --> 00:18:11,840
Speaker 4: cost of data sanity and all the tracks has been

351
00:18:12,040 --> 00:18:15,000
Speaker 4: in my view a little bit kind of more efficient

352
00:18:15,200 --> 00:18:18,120
Speaker 4: in a way that the ecosystem has been in place there,

353
00:18:18,280 --> 00:18:20,960
Speaker 4: so the cost efficiency has been there. So my view

354
00:18:21,000 --> 00:18:26,760
Speaker 4: is this is engineering, not theoretical quantum y right, So

355
00:18:26,800 --> 00:18:30,320
Speaker 4: on that the engineering capability form. The China tech world

356
00:18:30,400 --> 00:18:33,600
Speaker 4: and industry has been there with the key elements I

357
00:18:33,680 --> 00:18:37,320
Speaker 4: mentioned right, talents, lower costs, more efficient I think these

358
00:18:37,320 --> 00:18:38,960
Speaker 4: are the few things I just want to point out

359
00:18:39,040 --> 00:18:41,879
Speaker 4: actually can help solve the issue. But as I mentioned it,

360
00:18:42,000 --> 00:18:44,159
Speaker 4: the things are evolved very quickly, right, so you know,

361
00:18:44,200 --> 00:18:47,359
Speaker 4: we don't worry too much about the issue you mentioned

362
00:18:47,400 --> 00:18:48,080
Speaker 4: in local market.

363
00:18:48,240 --> 00:18:50,359
Speaker 2: Yeah, so this is interesting. I'm curious. I want to

364
00:18:50,640 --> 00:18:53,880
Speaker 2: press further on this because the American air I'm very

365
00:18:53,920 --> 00:18:57,960
Speaker 2: anxious about this, and they publish these model reports and

366
00:18:58,160 --> 00:19:01,080
Speaker 2: it says things that in the chain of thought, we

367
00:19:01,080 --> 00:19:03,720
Speaker 2: were able to see that four percent of the time

368
00:19:04,119 --> 00:19:07,160
Speaker 2: the model was able to identify that it was being tested,

369
00:19:07,640 --> 00:19:11,600
Speaker 2: and therefore it changed its behavior in response to recognizing

370
00:19:11,640 --> 00:19:14,040
Speaker 2: that it was tested. And this is a sign of

371
00:19:14,080 --> 00:19:18,000
Speaker 2: potential misalignment. Are you doing the same sort of research

372
00:19:18,080 --> 00:19:23,000
Speaker 2: and spending to establish that again, the models work for

373
00:19:23,119 --> 00:19:25,640
Speaker 2: people and don't have a rogue goal, So.

374
00:19:26,080 --> 00:19:28,800
Speaker 5: Yeah, sure, I think right now if you look at this, right.

375
00:19:28,920 --> 00:19:31,359
Speaker 4: So, we are also part of the open source community,

376
00:19:31,520 --> 00:19:35,400
Speaker 4: so many of the good model especially publishing recently, also

377
00:19:35,440 --> 00:19:38,320
Speaker 4: will publish their salts. Makes sense, so we kind of

378
00:19:38,440 --> 00:19:41,359
Speaker 4: follow the new sauts, but also doing our own pasts

379
00:19:41,400 --> 00:19:43,840
Speaker 4: as well. So overall, I think people in the open

380
00:19:43,920 --> 00:19:46,439
Speaker 4: source community today, in my view, is very collegial. So

381
00:19:46,480 --> 00:19:48,240
Speaker 4: people still want to do a better model from kil

382
00:19:48,280 --> 00:19:50,639
Speaker 4: model for everyone globally, not really on one country to

383
00:19:51,040 --> 00:19:52,160
Speaker 4: different places yet.

384
00:19:52,760 --> 00:19:55,200
Speaker 3: So actually, related to this, I'm going to ask something.

385
00:19:55,280 --> 00:19:58,120
Speaker 3: Maybe it's slightly sensitive, but I think it's very important.

386
00:19:58,200 --> 00:20:01,359
Speaker 3: So in the US, the AI companies, even as they

387
00:20:01,400 --> 00:20:04,920
Speaker 3: talk about safety, they're basically self regulating, right Like they

388
00:20:05,080 --> 00:20:08,040
Speaker 3: choose to put out these reports and judge their own

389
00:20:08,080 --> 00:20:11,040
Speaker 3: models and things like that. In China, tell me if

390
00:20:11,040 --> 00:20:14,399
Speaker 3: I'm wrong, but it feels very different. It feels like

391
00:20:14,440 --> 00:20:16,920
Speaker 3: the government is more hands on when it comes to AI.

392
00:20:17,040 --> 00:20:19,639
Speaker 3: China has been very explicit about this as an area

393
00:20:20,440 --> 00:20:24,880
Speaker 3: national security, national strategy. So you're operating in an environment

394
00:20:24,920 --> 00:20:30,080
Speaker 3: where you're firmly embedded in China's technological and industrial policy.

395
00:20:30,680 --> 00:20:34,360
Speaker 3: How does that influence the development of your AI models

396
00:20:34,359 --> 00:20:35,520
Speaker 3: and your broader tech.

397
00:20:35,720 --> 00:20:38,600
Speaker 4: So obviously we're not inflation to commut of public policy,

398
00:20:38,640 --> 00:20:40,600
Speaker 4: but I definitely happy to share some of our thoughts

399
00:20:40,640 --> 00:20:43,520
Speaker 4: I have. I think in the world, in the China

400
00:20:43,560 --> 00:20:46,800
Speaker 4: AI today, we believe we have a great group of

401
00:20:47,359 --> 00:20:51,280
Speaker 4: very superior talents, not only the engineers, but also people

402
00:20:51,359 --> 00:20:55,280
Speaker 4: actually design the framework. Right, So that's actually very important

403
00:20:55,400 --> 00:20:58,760
Speaker 4: because it's not only about algorithm selves, it's about whole system,

404
00:20:58,800 --> 00:21:02,840
Speaker 4: regarding infrastructure, regarding the data regulation, regarding the model, and

405
00:21:02,880 --> 00:21:05,160
Speaker 4: the cloud. I think given the past kind of ten

406
00:21:05,240 --> 00:21:08,600
Speaker 4: twenty years in China, giving this entire infrastructure has been

407
00:21:08,640 --> 00:21:11,240
Speaker 4: upgraded to a level that is kind of worth leading.

408
00:21:11,600 --> 00:21:15,119
Speaker 4: I think the policy is supporting to getting the moment

409
00:21:15,240 --> 00:21:15,840
Speaker 4: we have today.

410
00:21:15,880 --> 00:21:16,800
Speaker 5: It's already proven.

411
00:21:16,920 --> 00:21:19,960
Speaker 4: We have a proven path to leading not only the

412
00:21:20,000 --> 00:21:23,760
Speaker 4: technology renovation and innovation, but also the way you monitor

413
00:21:23,840 --> 00:21:24,640
Speaker 4: that into the.

414
00:21:24,600 --> 00:21:26,639
Speaker 5: Stage you already have today. So that's my first point.

415
00:21:26,840 --> 00:21:30,840
Speaker 4: My second point is I think today the technology is

416
00:21:30,840 --> 00:21:35,360
Speaker 4: growing very fast, and the tracks on the performance and

417
00:21:35,400 --> 00:21:39,560
Speaker 4: on the data transparency and the rules regarding the data regulation,

418
00:21:40,040 --> 00:21:43,280
Speaker 4: even without the AI model, even on the cloud age

419
00:21:43,320 --> 00:21:46,879
Speaker 4: in the past kind of twenty years, for five years,

420
00:21:47,280 --> 00:21:50,880
Speaker 4: has been getting more robust because if you think about that,

421
00:21:51,000 --> 00:21:54,040
Speaker 4: in a cloud environment, you have almost similar issues, right

422
00:21:54,160 --> 00:21:56,320
Speaker 4: who own the data, who use the data, who can

423
00:21:56,359 --> 00:21:59,840
Speaker 4: access that? But today it's a new tool to actually

424
00:22:00,040 --> 00:22:02,440
Speaker 4: cover all the things we are doing, so I think

425
00:22:02,480 --> 00:22:05,080
Speaker 4: it is not a new concept for the policy makers

426
00:22:05,119 --> 00:22:07,600
Speaker 4: think about. It's a new model, it is a new thing.

427
00:22:07,720 --> 00:22:11,560
Speaker 4: It is really a new and better tools to utilize

428
00:22:11,560 --> 00:22:15,120
Speaker 4: and access resource or already building and existing resources, will

429
00:22:15,240 --> 00:22:17,960
Speaker 4: be building on existing platforms, which has been one hundred

430
00:22:17,960 --> 00:22:19,960
Speaker 4: percent compliant. But also we have a lot of support,

431
00:22:20,040 --> 00:22:24,200
Speaker 4: you know, from the policymakers, industry practitioners, academias, they actually

432
00:22:24,200 --> 00:22:27,600
Speaker 4: all contributing to that. So overall, my feeling is it's

433
00:22:27,720 --> 00:22:31,399
Speaker 4: a very transparent and open environment, not only China but

434
00:22:31,440 --> 00:22:36,280
Speaker 4: also globally, and academias and industry practitioners actually contributing quite

435
00:22:36,320 --> 00:22:38,880
Speaker 4: a lot of the good conversations to this environment. And

436
00:22:39,040 --> 00:22:42,440
Speaker 4: my feeling is the policy makers through different channels has

437
00:22:42,440 --> 00:22:45,640
Speaker 4: been very open also listening to the new frontier issues

438
00:22:45,640 --> 00:22:46,280
Speaker 4: and questions.

439
00:22:46,600 --> 00:22:49,280
Speaker 2: I know this is a business conference and we want

440
00:22:49,320 --> 00:22:52,560
Speaker 2: to keep things very professional here and not engaging gossip,

441
00:22:52,600 --> 00:22:55,000
Speaker 2: et cetera. But I have like one sort of I'm

442
00:22:55,040 --> 00:22:59,920
Speaker 2: just curious about something, which is if the American ai

443
00:23:00,359 --> 00:23:04,560
Speaker 2: CEOs the most hawkish on the sort of like chip

444
00:23:04,600 --> 00:23:08,639
Speaker 2: exports on the China stuff is Dario, who used to

445
00:23:08,680 --> 00:23:09,240
Speaker 2: be a buyer.

446
00:23:09,240 --> 00:23:13,280
Speaker 6: Employee do you ever hear things in the office. Do

447
00:23:13,359 --> 00:23:15,760
Speaker 6: people ever say like, oh, I remember that guy he was?

448
00:23:16,480 --> 00:23:19,920
Speaker 6: You know, is there any Dario gossip that people talk

449
00:23:20,000 --> 00:23:22,320
Speaker 6: about in the office from his stinted bid.

450
00:23:22,840 --> 00:23:24,760
Speaker 4: So that's why I want to put the ball back

451
00:23:24,800 --> 00:23:27,120
Speaker 4: to a court. I want to show another gossip which

452
00:23:27,160 --> 00:23:30,919
Speaker 4: probably want to hear. So probably in the past kind

453
00:23:30,960 --> 00:23:34,000
Speaker 4: of hundred days, right, it's open club become very popular, right,

454
00:23:34,040 --> 00:23:37,080
Speaker 4: and educated market about how AI is really getting to

455
00:23:37,160 --> 00:23:40,720
Speaker 4: the real task and the real word. Obviously in China,

456
00:23:40,840 --> 00:23:43,120
Speaker 4: you know a different way. You have a new way

457
00:23:43,160 --> 00:23:45,600
Speaker 4: calling you know, not only the claw but other nicky names.

458
00:23:45,680 --> 00:23:45,880
Speaker 5: Right.

459
00:23:45,960 --> 00:23:49,960
Speaker 4: So one day I saw Peter, who is the founder

460
00:23:50,040 --> 00:23:52,760
Speaker 4: of the community which drives the open clock to be

461
00:23:52,800 --> 00:23:58,240
Speaker 4: prevailed put in Instagram, yeah, saying you know he actually

462
00:23:58,240 --> 00:24:00,560
Speaker 4: wanted to work with spy do big because you know

463
00:24:00,840 --> 00:24:04,720
Speaker 4: open cloys are too. But the two is it's kind

464
00:24:04,720 --> 00:24:07,679
Speaker 4: of eating up all the capacity. I eat the skills, right,

465
00:24:07,840 --> 00:24:11,399
Speaker 4: so everyone is contributing to the skills, and the cloud

466
00:24:11,480 --> 00:24:13,680
Speaker 4: is actually grab all the skills and do the work.

467
00:24:14,080 --> 00:24:14,280
Speaker 1: Right.

468
00:24:14,760 --> 00:24:17,080
Speaker 4: So I think one day we are pretty happy, you know,

469
00:24:17,119 --> 00:24:20,840
Speaker 4: see Peter drop us a note very in a positive way.

470
00:24:21,000 --> 00:24:24,760
Speaker 4: Because he kind of on one side noticed that the

471
00:24:24,800 --> 00:24:28,240
Speaker 4: search is important capability of the skills I eat the

472
00:24:28,280 --> 00:24:31,359
Speaker 4: skills in the open cloud environment. So actually ask him

473
00:24:31,359 --> 00:24:34,359
Speaker 4: by do to work with him to beef up the

474
00:24:34,400 --> 00:24:37,119
Speaker 4: search skills in order for open cloud to do a

475
00:24:37,119 --> 00:24:40,359
Speaker 4: better work to accessing the real time information. Because today

476
00:24:40,480 --> 00:24:43,040
Speaker 4: the fundational model is one kind of carve out. Is

477
00:24:43,400 --> 00:24:46,399
Speaker 4: every few months you train a new model and the

478
00:24:46,440 --> 00:24:49,440
Speaker 4: model itself in his mindset doesn't have the real time information.

479
00:24:49,520 --> 00:24:52,400
Speaker 4: For example, the fundational model is that it doesn't capture

480
00:24:52,480 --> 00:24:54,679
Speaker 4: you know, Joe and Tracy we're talking about today. You

481
00:24:54,760 --> 00:24:57,920
Speaker 4: need to have a new skills accessing the od slots

482
00:24:58,040 --> 00:25:00,639
Speaker 4: what is happening in real time. So I think you

483
00:25:00,680 --> 00:25:02,760
Speaker 4: know Google globally and a Byeo of China, they are

484
00:25:02,800 --> 00:25:05,680
Speaker 4: the powerhouse for searching real time information. So it has

485
00:25:05,720 --> 00:25:07,480
Speaker 4: to be linking to the open clock. So I think

486
00:25:07,520 --> 00:25:09,480
Speaker 4: that's actually one of the things we're pretty much happy

487
00:25:09,480 --> 00:25:11,520
Speaker 4: to about. So you know, next day we ask our

488
00:25:11,560 --> 00:25:14,000
Speaker 4: engineers to link up with Speter and we're actually part

489
00:25:14,040 --> 00:25:17,320
Speaker 4: of this skill marketplace doing pretty okay. And right now

490
00:25:17,320 --> 00:25:19,840
Speaker 4: I just want to share our foundation model called earning

491
00:25:19,880 --> 00:25:23,520
Speaker 4: five point one right now is ranked as the globally

492
00:25:23,880 --> 00:25:26,760
Speaker 4: number one in the text format of the Global ram

493
00:25:26,800 --> 00:25:29,520
Speaker 4: Arena and the globally number five in the search skill

494
00:25:29,560 --> 00:25:32,760
Speaker 4: capabilities globally in the RAM arena as well. So I

495
00:25:32,800 --> 00:25:35,040
Speaker 4: think that's I want to give you another gossip. So

496
00:25:35,520 --> 00:25:37,280
Speaker 4: but for the previous one, I probably can talk with

497
00:25:37,320 --> 00:25:38,119
Speaker 4: you after this.

498
00:25:38,720 --> 00:25:39,800
Speaker 5: Open Yeah, I.

499
00:25:39,840 --> 00:25:43,880
Speaker 2: Know you didn't give us any Dario gossip, but implicitly

500
00:25:43,960 --> 00:25:46,480
Speaker 2: because I know that the open clock guy, you know,

501
00:25:46,480 --> 00:25:50,760
Speaker 2: it's originally called open claud and then anthropics suit him.

502
00:25:50,880 --> 00:25:53,760
Speaker 2: And also he's got kind of annoyed because they didn't

503
00:25:53,880 --> 00:25:57,280
Speaker 2: let the API users get full access. So I think

504
00:25:57,720 --> 00:26:00,439
Speaker 2: that fellow who created open claw is not the biggest

505
00:26:00,440 --> 00:26:04,719
Speaker 2: fan of Dario's approach. So I by by giving us answer,

506
00:26:04,880 --> 00:26:08,160
Speaker 2: at least give us a little drama there. Thank you.

507
00:26:08,160 --> 00:26:12,280
Speaker 3: You mentioned search a number of times already, and data

508
00:26:12,359 --> 00:26:15,160
Speaker 3: is obviously very important to AI. We spoke with Gray

509
00:26:15,320 --> 00:26:18,280
Speaker 3: Shao earlier in the week. She writes about AI on

510
00:26:18,320 --> 00:26:20,560
Speaker 3: her sub stack, and I asked her if China has

511
00:26:20,600 --> 00:26:24,879
Speaker 3: an edge when it comes to data collection, and she

512
00:26:25,040 --> 00:26:27,320
Speaker 3: said she thought not really because a lot of the

513
00:26:27,400 --> 00:26:30,399
Speaker 3: data that's been collected was unstructured, and so it was

514
00:26:30,520 --> 00:26:34,639
Speaker 3: hard to harness for AI model training and inference purposes.

515
00:26:35,000 --> 00:26:36,800
Speaker 3: He talked a little bit more about how you did

516
00:26:36,800 --> 00:26:40,080
Speaker 3: that at BYDO because you've got a lot of data

517
00:26:40,119 --> 00:26:44,560
Speaker 3: you're using it for Ernie. How is that transition process

518
00:26:44,640 --> 00:26:45,960
Speaker 3: like actually carried out.

519
00:26:46,240 --> 00:26:49,040
Speaker 4: Yeah, sure, I probably would talk about something that market

520
00:26:49,040 --> 00:26:52,280
Speaker 4: has not noticed enough, and I will talk about what's

521
00:26:52,280 --> 00:26:54,879
Speaker 4: a real challenge, right, So it's always two sides of

522
00:26:54,880 --> 00:26:57,360
Speaker 4: a story. I seem I'm very happy to talk about

523
00:26:57,400 --> 00:27:00,360
Speaker 4: Google versus what we think about by DO in some

524
00:27:00,680 --> 00:27:01,639
Speaker 4: certain formats.

525
00:27:02,200 --> 00:27:02,879
Speaker 5: A few things.

526
00:27:03,000 --> 00:27:05,919
Speaker 4: I think the markets, not only the capital markets, but

527
00:27:06,000 --> 00:27:09,119
Speaker 4: also the industry has kind of on the value a

528
00:27:09,160 --> 00:27:12,240
Speaker 4: little bit regarding the same components we actually imagine with

529
00:27:12,280 --> 00:27:16,400
Speaker 4: the same structure. Google is monetizing and have this integrated capacity.

530
00:27:16,880 --> 00:27:21,480
Speaker 4: So first of all, Google has its own TPU, right,

531
00:27:21,480 --> 00:27:25,040
Speaker 4: which empowered that cloud. So the GCP growth, the Google

532
00:27:25,040 --> 00:27:28,440
Speaker 4: cloud is growing faster, which is part of the reason

533
00:27:28,520 --> 00:27:31,240
Speaker 4: is the TPU. And so by Do we have our

534
00:27:31,280 --> 00:27:34,640
Speaker 4: own trip department, and you know, based on the public information,

535
00:27:34,800 --> 00:27:37,119
Speaker 4: we recently did the public filing on the spring of

536
00:27:37,200 --> 00:27:39,320
Speaker 4: these assets, right, So the trip we have the theme

537
00:27:39,359 --> 00:27:41,280
Speaker 4: with Google for the foundation model.

538
00:27:41,280 --> 00:27:43,600
Speaker 5: We have our earnings sus mentioned.

539
00:27:43,560 --> 00:27:47,199
Speaker 4: However in the physical AI cord applications or called the

540
00:27:47,240 --> 00:27:50,560
Speaker 4: work model. So we have our robotaxi called Apollo Goal.

541
00:27:51,000 --> 00:27:52,800
Speaker 4: Just want to share one number. I think the market

542
00:27:52,960 --> 00:27:55,360
Speaker 4: sometimes I tell even my friends, it's kind of surprised

543
00:27:55,359 --> 00:27:59,159
Speaker 4: that each week, including San Francisco and including all the

544
00:27:59,200 --> 00:28:03,720
Speaker 4: cities asting taxes in US, we MO from Google delivered

545
00:28:03,720 --> 00:28:07,199
Speaker 4: about five hundred thousand trips per week, and in the

546
00:28:07,240 --> 00:28:11,240
Speaker 4: last quorder by do appollable in globally twenty seven cities

547
00:28:11,320 --> 00:28:14,119
Speaker 4: delivered about three hundred and fifty thousand trips, which is

548
00:28:14,119 --> 00:28:17,760
Speaker 4: only about twenty five percent field than Google. The part

549
00:28:17,880 --> 00:28:20,479
Speaker 4: of that is not only about robotaxi. It's about how

550
00:28:20,520 --> 00:28:24,639
Speaker 4: we're using the data empower our own foundations, models, trainings

551
00:28:24,680 --> 00:28:26,600
Speaker 4: and also do influence but also have a lot of

552
00:28:26,640 --> 00:28:29,280
Speaker 4: know how regarding the multi modile contents and all the

553
00:28:29,280 --> 00:28:31,760
Speaker 4: different things and the more important if you look at

554
00:28:31,880 --> 00:28:35,600
Speaker 4: the traditional search right now on this border also on

555
00:28:35,640 --> 00:28:37,960
Speaker 4: the market, didn't notice that, you know, still even my

556
00:28:38,040 --> 00:28:41,520
Speaker 4: friends telling me, oh, Henry, congratulations for that for fiel earnings.

557
00:28:41,520 --> 00:28:44,160
Speaker 4: But you search it probably still eighty ninety percent of

558
00:28:44,160 --> 00:28:46,240
Speaker 4: the revenue. But the chooses for this border is declined

559
00:28:46,280 --> 00:28:49,040
Speaker 4: about forty eight percent, so it's already blow fifty percent.

560
00:28:49,320 --> 00:28:50,960
Speaker 5: So the new growing area.

561
00:28:50,680 --> 00:28:54,840
Speaker 4: For example, the digital humans and also our application software

562
00:28:54,920 --> 00:28:57,640
Speaker 4: is becoming a powerhouse and growing very fast. So my

563
00:28:57,720 --> 00:29:01,240
Speaker 4: point on that is if you look had key components

564
00:29:01,280 --> 00:29:05,040
Speaker 4: or the blocks from the trip cloud, ROBOTAXI for AI,

565
00:29:05,040 --> 00:29:07,840
Speaker 4: physical air applications and the software. And also one more

566
00:29:07,840 --> 00:29:10,080
Speaker 4: thing I want to mention is Google linking with the

567
00:29:10,120 --> 00:29:13,280
Speaker 4: YouTube for the multi model contents, and we actually have

568
00:29:13,400 --> 00:29:16,760
Speaker 4: our control the subsidy called it iq in China, which

569
00:29:16,800 --> 00:29:18,680
Speaker 4: is actually over fifty per cent of market share in

570
00:29:18,760 --> 00:29:21,440
Speaker 4: China for certain long form contents in China as well,

571
00:29:21,760 --> 00:29:24,360
Speaker 4: So we also have our closed loop of the data

572
00:29:24,360 --> 00:29:27,120
Speaker 4: flight well as well, probably at a different scale, but

573
00:29:27,200 --> 00:29:29,080
Speaker 4: I think it's still in the same format and the

574
00:29:29,080 --> 00:29:32,680
Speaker 4: same model. So my view is yes, I think on

575
00:29:33,000 --> 00:29:37,080
Speaker 4: website it is right that certain data and elements are

576
00:29:37,120 --> 00:29:40,600
Speaker 4: in their own kind of pockets. But however, for by

577
00:29:40,680 --> 00:29:44,960
Speaker 4: do we still have access to those pockets, probably better

578
00:29:44,960 --> 00:29:48,320
Speaker 4: than other peers. But I want also very honest admit, right,

579
00:29:48,400 --> 00:29:51,080
Speaker 4: given Joey's my girlfriend, I still own him a little

580
00:29:51,200 --> 00:29:53,760
Speaker 4: kind of gossip after the session. I want to share

581
00:29:53,880 --> 00:29:56,880
Speaker 4: my own challenge or it's in China. You have different camps, right,

582
00:29:56,920 --> 00:30:00,200
Speaker 4: different camps, they kind of don't open up enough how

583
00:30:00,240 --> 00:30:02,440
Speaker 4: to share the data which is reality known to the

584
00:30:02,480 --> 00:30:05,280
Speaker 4: market for everyone. But my point is right now this

585
00:30:06,480 --> 00:30:10,000
Speaker 4: you know, the agents and the foundation model become super

586
00:30:10,040 --> 00:30:13,920
Speaker 4: smart and it has a great push to move everything

587
00:30:14,000 --> 00:30:17,160
Speaker 4: to a public cloud. It's actually helped resolving that issue

588
00:30:17,200 --> 00:30:20,080
Speaker 4: to be accessing more information. Last note I want to

589
00:30:20,080 --> 00:30:23,920
Speaker 4: share is before AI can the public car penetration in

590
00:30:24,000 --> 00:30:26,760
Speaker 4: China is about twenty thirty percent versus in US is

591
00:30:26,880 --> 00:30:29,160
Speaker 4: kind of ninety percent. So that's why your comments I

592
00:30:29,200 --> 00:30:32,600
Speaker 4: can totally understand because without AI, the gap is like this,

593
00:30:33,320 --> 00:30:36,360
Speaker 4: but right now it's actually getting closer, but still there's

594
00:30:36,400 --> 00:30:39,640
Speaker 4: a gap. But my confident coming from this gap will

595
00:30:39,720 --> 00:30:43,960
Speaker 4: further narrowed because everything will be on cloud environment, everyone's

596
00:30:43,960 --> 00:30:46,720
Speaker 4: access real time information. But also for by do we

597
00:30:46,800 --> 00:30:49,880
Speaker 4: have a four stack and each components. Given the Google

598
00:30:49,960 --> 00:30:52,440
Speaker 4: Pass has been proven to be right and more efficient,

599
00:30:52,600 --> 00:30:55,400
Speaker 4: we just want to follow the same pattern and access

600
00:30:55,440 --> 00:31:03,640
Speaker 4: and benefits from the different layer of the data itself.

601
00:31:14,200 --> 00:31:17,920
Speaker 2: So I am very glad that you brought up the robotaxis,

602
00:31:17,960 --> 00:31:20,480
Speaker 2: because first of all, I just love robo taxis period.

603
00:31:20,520 --> 00:31:21,840
Speaker 2: They're very fun to ride in.

604
00:31:22,120 --> 00:31:24,600
Speaker 3: You took me on my first I remember, and I

605
00:31:24,680 --> 00:31:25,160
Speaker 3: was a big.

606
00:31:25,040 --> 00:31:26,760
Speaker 2: I was like, Tracy, you gotta ride in Weimo, you

607
00:31:26,760 --> 00:31:29,080
Speaker 2: gotta ride in a Weimo. And I think you're convinced.

608
00:31:28,760 --> 00:31:29,960
Speaker 3: They're now we have to write it.

609
00:31:31,920 --> 00:31:35,040
Speaker 2: But here's another besides. I'm also excited about them as

610
00:31:35,040 --> 00:31:38,280
Speaker 2: a business story for a very specific reason, which is

611
00:31:38,320 --> 00:31:40,880
Speaker 2: that when I think of like by Do, people call

612
00:31:40,920 --> 00:31:43,680
Speaker 2: it the Google of China's but you know, separate markets, right,

613
00:31:43,720 --> 00:31:46,880
Speaker 2: Google is something China. People in the US don't buy

614
00:31:46,920 --> 00:31:48,640
Speaker 2: in large use by Do as far as I know.

615
00:31:48,960 --> 00:31:51,480
Speaker 2: But there's gonna be cities now where there's gonna be

616
00:31:51,560 --> 00:31:57,120
Speaker 2: direct competition between Weimo and Apollo, and including London, I

617
00:31:57,160 --> 00:31:59,360
Speaker 2: think is going to be the first city where there's

618
00:31:59,400 --> 00:32:02,280
Speaker 2: going to be head to head beut. And so this

619
00:32:02,400 --> 00:32:05,360
Speaker 2: is exciting because I feel like, Okay, the tech Internet,

620
00:32:05,600 --> 00:32:08,920
Speaker 2: the consumer facing Internet giants of the US, the consumer

621
00:32:08,960 --> 00:32:11,800
Speaker 2: facing Internet giants of China are for the first time

622
00:32:12,200 --> 00:32:16,600
Speaker 2: going to really be competing in certain identical consumer markets.

623
00:32:16,680 --> 00:32:19,560
Speaker 2: And so what I'm curious about is like, not who

624
00:32:20,000 --> 00:32:21,840
Speaker 2: you think is gonna win. I presume you think you're

625
00:32:21,840 --> 00:32:25,920
Speaker 2: gonna win, But like, what is the dimension upon which

626
00:32:26,680 --> 00:32:29,720
Speaker 2: the winner will emerge? Will it be the quality of

627
00:32:29,760 --> 00:32:34,000
Speaker 2: the application, will it be who can produce and secure

628
00:32:34,120 --> 00:32:38,760
Speaker 2: automobiles in volume? What is the most important dimension that

629
00:32:38,880 --> 00:32:42,080
Speaker 2: will determine the winner, either globally or in.

630
00:32:42,040 --> 00:32:44,880
Speaker 4: A specific city, So be fugetting that, just you know. Also,

631
00:32:45,040 --> 00:32:47,280
Speaker 4: the car is one of my hobby too, just you know,

632
00:32:47,360 --> 00:32:50,960
Speaker 4: Joe probably know I'm actually a risk car driver, so

633
00:32:51,000 --> 00:32:53,040
Speaker 4: I got my risk my license as well, so every

634
00:32:53,120 --> 00:32:56,640
Speaker 4: time I actually drive a little bit, so it's quite nice,

635
00:32:56,760 --> 00:32:59,920
Speaker 4: enjoyable because I think that probably the remaining KAI can

636
00:33:00,200 --> 00:33:03,560
Speaker 4: use gasoling and drive yourself probably twenty years from now,

637
00:33:03,960 --> 00:33:07,720
Speaker 4: so before getting there, I think the key world is change.

638
00:33:07,400 --> 00:33:11,960
Speaker 5: The car ownership. Okay, my view is in my own calculation.

639
00:33:12,240 --> 00:33:17,480
Speaker 4: In US as an example, right now, each mile including insurance,

640
00:33:17,680 --> 00:33:21,720
Speaker 4: gas price, parkings average is about sixty to eighty cents

641
00:33:21,760 --> 00:33:25,080
Speaker 4: per mile. Is the tipping points between rent a car.

642
00:33:25,080 --> 00:33:25,920
Speaker 5: Versus owner car.

643
00:33:26,120 --> 00:33:28,880
Speaker 4: Okay, so if cheaper than that, people will buy a car,

644
00:33:29,000 --> 00:33:31,480
Speaker 4: but more expensive people will rent a car. Right so,

645
00:33:31,640 --> 00:33:33,760
Speaker 4: right now, for global player, I don't want the name

646
00:33:33,840 --> 00:33:36,960
Speaker 4: name but the average ROBOTAXI costs today because the scale

647
00:33:37,000 --> 00:33:41,000
Speaker 4: server still very small. It's about you know, one two

648
00:33:41,320 --> 00:33:43,440
Speaker 4: or two point five dollars per mile, so that's a

649
00:33:43,520 --> 00:33:45,600
Speaker 4: range about one to two dollars. So my point is

650
00:33:46,080 --> 00:33:49,400
Speaker 4: this curve, just like agents getting more prevailing, is coming

651
00:33:49,440 --> 00:33:53,000
Speaker 4: down very fast. So assuming when some point, you know,

652
00:33:53,120 --> 00:33:57,000
Speaker 4: five years, six years, whatever, ten years, if globally the

653
00:33:57,080 --> 00:34:01,240
Speaker 4: robotax deliver average price per mile coming down to let's

654
00:34:01,240 --> 00:34:04,640
Speaker 4: say sixty eighty cents US per mile, then many people

655
00:34:04,640 --> 00:34:06,960
Speaker 4: were thinking buying a car because today's very thankful, you know,

656
00:34:07,040 --> 00:34:08,520
Speaker 4: Joe and Tricity. You're probably in Hong Kong. You know,

657
00:34:08,560 --> 00:34:11,800
Speaker 4: the parking is so expensive, even more expensive than the gas,

658
00:34:12,000 --> 00:34:13,200
Speaker 4: and the gas in Hong Kong.

659
00:34:13,040 --> 00:34:14,000
Speaker 5: Also very expensive.

660
00:34:14,440 --> 00:34:16,480
Speaker 4: So first of all, the car right now is all

661
00:34:16,480 --> 00:34:19,560
Speaker 4: Evy drive car Number two, you don't have a buyer

662
00:34:19,960 --> 00:34:21,640
Speaker 4: parking because you and we can drive in a car

663
00:34:21,680 --> 00:34:24,200
Speaker 4: can go out and number three, while we're having this

664
00:34:24,320 --> 00:34:26,480
Speaker 4: forty minutes, my car actually can go up pick up

665
00:34:26,480 --> 00:34:28,080
Speaker 4: passengers and I can make some money for me.

666
00:34:28,200 --> 00:34:29,200
Speaker 5: Right it's a new agent.

667
00:34:29,280 --> 00:34:31,319
Speaker 4: So that's what I'm saying. It's a physical agents on

668
00:34:31,360 --> 00:34:34,160
Speaker 4: the road to making money for myself. So my view

669
00:34:34,239 --> 00:34:38,040
Speaker 4: is ROBOTAXI will change the human behavior getting out in

670
00:34:38,120 --> 00:34:41,640
Speaker 4: terms of behavior pattern of transportations. That's one thing, right,

671
00:34:41,800 --> 00:34:44,800
Speaker 4: So we and way More and all other players globally

672
00:34:44,800 --> 00:34:47,680
Speaker 4: are going to that direction. So that's my vision for

673
00:34:48,040 --> 00:34:51,480
Speaker 4: the market going forward. However, as you mentioned about the

674
00:34:51,520 --> 00:34:55,000
Speaker 4: market as a player in the near chime competition, my

675
00:34:55,239 --> 00:34:57,839
Speaker 4: view is right now the markets do.

676
00:34:58,239 --> 00:35:01,320
Speaker 5: Very early and the ten is very high.

677
00:35:01,840 --> 00:35:04,640
Speaker 4: So in the last quarter we shap our car in

678
00:35:04,680 --> 00:35:07,200
Speaker 4: London and as you know, you know both way More

679
00:35:07,239 --> 00:35:09,880
Speaker 4: and Us are starting open the market London. Hopefully next

680
00:35:09,960 --> 00:35:12,200
Speaker 4: year you will see a car. And you know we

681
00:35:12,280 --> 00:35:14,560
Speaker 4: have go partner with both Uber and the Lift and

682
00:35:14,600 --> 00:35:17,319
Speaker 4: also with a Grab in South East countries, so next

683
00:35:17,360 --> 00:35:19,960
Speaker 4: time you probably call a car from Uber or lift apps,

684
00:35:20,000 --> 00:35:22,359
Speaker 4: you'll get a buy those car. So I think it's

685
00:35:22,440 --> 00:35:27,400
Speaker 4: actually helping increasing the services because one interesting notice the

686
00:35:27,480 --> 00:35:29,839
Speaker 4: human driver probably don't work in the midnight right in

687
00:35:29,840 --> 00:35:32,120
Speaker 4: certain cities, but right now the car actually can work

688
00:35:32,160 --> 00:35:35,480
Speaker 4: twenty four hours. So expanding a new market and right

689
00:35:35,520 --> 00:35:38,160
Speaker 4: now is still a very low percentage of penetration, so

690
00:35:38,320 --> 00:35:40,560
Speaker 4: still we have a lot of room to go on

691
00:35:40,560 --> 00:35:42,959
Speaker 4: the other end to your question about the success factor,

692
00:35:43,040 --> 00:35:45,399
Speaker 4: I seek the two things. When is a technology need

693
00:35:45,480 --> 00:35:47,719
Speaker 4: to cutting edge and improving, The number.

694
00:35:47,440 --> 00:35:48,600
Speaker 5: Two is operation efficiency.

695
00:35:48,680 --> 00:35:51,440
Speaker 4: Right, so it's actually have a lot of work need

696
00:35:51,520 --> 00:35:55,440
Speaker 4: to be operational driven. For example, how many locations you

697
00:35:55,520 --> 00:35:57,240
Speaker 4: pick up passengers to meet more efficient?

698
00:35:57,440 --> 00:35:57,600
Speaker 5: Right?

699
00:35:57,680 --> 00:36:00,680
Speaker 4: The charging stations and all the different networks is actually

700
00:36:00,760 --> 00:36:03,560
Speaker 4: very important. But given we are working on this business

701
00:36:03,600 --> 00:36:06,759
Speaker 4: for kind of thirteen years so far, and I can

702
00:36:06,920 --> 00:36:09,600
Speaker 4: tell you one interesting fact. Globally, there are only two

703
00:36:09,640 --> 00:36:14,880
Speaker 4: cities right now have over you know, thousand cars in

704
00:36:14,960 --> 00:36:17,320
Speaker 4: that scale, which is San Francisco and a win city

705
00:36:17,400 --> 00:36:20,239
Speaker 4: in China, which is Google operating in San Francisco and

706
00:36:20,600 --> 00:36:24,480
Speaker 4: Apollo Go from Baidu operating one city in China. But

707
00:36:24,600 --> 00:36:27,560
Speaker 4: I think our kind of partnership with both you know,

708
00:36:27,719 --> 00:36:30,920
Speaker 4: a Lift and the Uber globally with different cities, I

709
00:36:30,960 --> 00:36:33,560
Speaker 4: think has been very collegial because the demand is much

710
00:36:33,600 --> 00:36:34,840
Speaker 4: higher than the supply.

711
00:36:36,120 --> 00:36:39,840
Speaker 3: Joe, I'm going to admit something slightly embarrassing. Actually you

712
00:36:39,880 --> 00:36:43,000
Speaker 3: already know this, but I never learned to drive, partly

713
00:36:43,040 --> 00:36:45,080
Speaker 3: because I grew up in Tokyo and then I moved

714
00:36:45,080 --> 00:36:46,960
Speaker 3: to a bunch of other big cities and so I

715
00:36:47,000 --> 00:36:49,680
Speaker 3: never needed to, and now I always joke that I'm

716
00:36:49,719 --> 00:36:50,520
Speaker 3: basically I'm never.

717
00:36:50,520 --> 00:36:50,960
Speaker 5: Going to learn.

718
00:36:51,000 --> 00:36:53,360
Speaker 3: I'm just going to hold out for the self driving cars.

719
00:36:53,400 --> 00:36:57,840
Speaker 3: So you know, fingers crossed, I hope. So I wanted

720
00:36:57,880 --> 00:37:00,880
Speaker 3: to ask something about you know, you've mentioned agents a

721
00:37:00,960 --> 00:37:03,200
Speaker 3: number of times and this seems to be becoming the

722
00:37:03,239 --> 00:37:05,840
Speaker 3: hot new thing in AI, and I know your CEO

723
00:37:05,960 --> 00:37:08,279
Speaker 3: has talked about how one of the key metrics for

724
00:37:08,320 --> 00:37:12,879
Speaker 3: bydo is daily active agents. And my question is how

725
00:37:12,920 --> 00:37:17,279
Speaker 3: does that actually turn into revenue or return from a

726
00:37:17,320 --> 00:37:20,719
Speaker 3: cfo's perspective, because I understand with search, you know, you

727
00:37:20,760 --> 00:37:24,440
Speaker 3: type something in you see the ads advertisers are paying

728
00:37:24,480 --> 00:37:27,440
Speaker 3: you for that, but I'm very unclear how it works

729
00:37:27,640 --> 00:37:29,960
Speaker 3: if the agent is actually going out and doing something.

730
00:37:30,160 --> 00:37:30,680
Speaker 5: Yeah.

731
00:37:30,800 --> 00:37:34,200
Speaker 4: So in the mobile internet, where you know everyone look have,

732
00:37:34,320 --> 00:37:38,600
Speaker 4: for example, the DAO the daily active users because that

733
00:37:38,760 --> 00:37:44,560
Speaker 4: either fulfilled information, quority demand and individual are primary users

734
00:37:44,600 --> 00:37:47,600
Speaker 4: for many of the mobile applications in app stores, so

735
00:37:47,640 --> 00:37:51,760
Speaker 4: that DAO was the primary matrix to measure that. However,

736
00:37:51,880 --> 00:37:55,279
Speaker 4: in the recent conference, our chairman and founder of by Do,

737
00:37:55,960 --> 00:38:00,480
Speaker 4: Robin mentioned, as in based on his leadership, that DAA,

738
00:38:00,640 --> 00:38:03,759
Speaker 4: which is a daily active agents are the new kind

739
00:38:03,800 --> 00:38:07,759
Speaker 4: of matrix to defining the success of agents. So I

740
00:38:07,880 --> 00:38:10,640
Speaker 4: kind of very much agree on that. The reason is

741
00:38:11,120 --> 00:38:13,680
Speaker 4: if you look at the tasks, it's actually spread out

742
00:38:13,760 --> 00:38:17,360
Speaker 4: into different verticals right so right now it's very difficult

743
00:38:17,360 --> 00:38:20,600
Speaker 4: to find a new way to identify how much people

744
00:38:20,719 --> 00:38:24,000
Speaker 4: using especially how much value coming out from using AI.

745
00:38:24,520 --> 00:38:28,480
Speaker 4: So the agents today is basically can deliver a final task,

746
00:38:28,760 --> 00:38:32,120
Speaker 4: not only using as a tool for human beings. The

747
00:38:32,200 --> 00:38:35,600
Speaker 4: agent is smart enoughs can think about that, planning the

748
00:38:35,680 --> 00:38:38,680
Speaker 4: task and completing the task, and obviously in the way

749
00:38:38,760 --> 00:38:41,520
Speaker 4: of interacting with human beings, it actually become more smarter

750
00:38:41,640 --> 00:38:43,960
Speaker 4: and in the way that working with more efficient planning

751
00:38:43,960 --> 00:38:48,400
Speaker 4: of that. So the truer question overall my thinking is

752
00:38:48,640 --> 00:38:54,000
Speaker 4: the DAA will measure not only how many people using that,

753
00:38:54,120 --> 00:38:56,880
Speaker 4: but also how difficult it is. True a question the

754
00:38:56,920 --> 00:38:59,960
Speaker 4: result driven payment is coming up in the near trail.

755
00:39:00,560 --> 00:39:02,880
Speaker 4: So I just want to share a few things. For example,

756
00:39:03,200 --> 00:39:06,000
Speaker 4: right now we have three or four different key products,

757
00:39:06,040 --> 00:39:10,760
Speaker 4: one of them in China qualify or it's really solving

758
00:39:11,160 --> 00:39:15,839
Speaker 4: complicated issues for enterprises. It's very similar to our fur goal,

759
00:39:15,880 --> 00:39:18,800
Speaker 4: which actually in the previous years to do the planning.

760
00:39:18,800 --> 00:39:20,520
Speaker 4: But right now it's actually coming to the real world.

761
00:39:20,880 --> 00:39:23,759
Speaker 4: So we install this agent to one of the biggest

762
00:39:24,040 --> 00:39:27,560
Speaker 4: pots in China and help them deploy and planning for

763
00:39:27,640 --> 00:39:31,360
Speaker 4: the shipments, the logistics. It's saving the cost of the

764
00:39:31,400 --> 00:39:35,279
Speaker 4: idle time and improving the revenue of their parts. So

765
00:39:35,400 --> 00:39:39,160
Speaker 4: the pots actually willing to share a certain profit generation

766
00:39:39,320 --> 00:39:42,680
Speaker 4: with us. So the key things I'm observing is in

767
00:39:42,760 --> 00:39:45,120
Speaker 4: the previous meetings. Even I'm a sapphold, but actually I'm

768
00:39:45,120 --> 00:39:47,439
Speaker 4: tending a lot of you know, the meetings to meet

769
00:39:47,440 --> 00:39:50,480
Speaker 4: with clients. In the previous meeting without AI, most of

770
00:39:50,520 --> 00:39:53,120
Speaker 4: the meetings we are talking with is the CTO and

771
00:39:53,160 --> 00:39:56,160
Speaker 4: the CIO of that company. Because it was a tool,

772
00:39:56,280 --> 00:39:59,000
Speaker 4: it was a cost center, so they need to find

773
00:39:59,040 --> 00:40:01,759
Speaker 4: a budget internally, Joe, Unit's not easy, right So they

774
00:40:01,760 --> 00:40:04,640
Speaker 4: have their CFO and their CEO. But right now, most

775
00:40:04,680 --> 00:40:06,400
Speaker 4: of the meeting we are having today is with the

776
00:40:06,440 --> 00:40:09,680
Speaker 4: CEO himself. Because AI right now is not only about

777
00:40:09,880 --> 00:40:12,040
Speaker 4: by Doo, It's really about helping our clients. So the

778
00:40:12,080 --> 00:40:14,400
Speaker 4: client has to be a top down level of the

779
00:40:14,400 --> 00:40:18,560
Speaker 4: initiatives to really drive AI internally. So I think our

780
00:40:18,800 --> 00:40:21,759
Speaker 4: sales process become relatively more efficient in a way that

781
00:40:21,800 --> 00:40:24,520
Speaker 4: we get into the number one decision makers. He has

782
00:40:24,560 --> 00:40:27,440
Speaker 4: a budget and he knows that driving the part efficiency

783
00:40:27,480 --> 00:40:30,240
Speaker 4: is important for his task, so he's willing to share

784
00:40:30,280 --> 00:40:34,080
Speaker 4: certain economics with us. So I think the customization is

785
00:40:34,080 --> 00:40:36,560
Speaker 4: also diminished because right now the agents can be used

786
00:40:36,600 --> 00:40:39,319
Speaker 4: different pasts. You can repeating that success lower the unit

787
00:40:39,320 --> 00:40:42,359
Speaker 4: basis of the cost, and the agents become more real

788
00:40:42,600 --> 00:40:45,560
Speaker 4: and the clients see the value and the profits, so

789
00:40:45,600 --> 00:40:48,920
Speaker 4: they have a higher willingness to pay and high ability

790
00:40:49,000 --> 00:40:52,360
Speaker 4: to achieve that payment. So I think these four cycles

791
00:40:52,360 --> 00:40:54,239
Speaker 4: actually in the AI world is very different with the

792
00:40:54,280 --> 00:40:54,840
Speaker 4: traditional it.

793
00:40:55,400 --> 00:40:58,440
Speaker 2: This is actually an interesting question because I've seen debate

794
00:40:58,560 --> 00:41:02,520
Speaker 2: on this within AI about what is revenue look like

795
00:41:02,800 --> 00:41:05,840
Speaker 2: or what is you know, a sales price look like?

796
00:41:06,120 --> 00:41:10,040
Speaker 2: Because another thing people talk about is, for example, using

797
00:41:10,080 --> 00:41:13,520
Speaker 2: an AI agent to say, resolve an insurance claim or

798
00:41:13,560 --> 00:41:16,960
Speaker 2: something like that, and then the AI provider gets paid

799
00:41:17,080 --> 00:41:21,480
Speaker 2: on say like you know, the number of successful claims resolved,

800
00:41:21,480 --> 00:41:26,040
Speaker 2: et cetera. Are you bullish on that basic model where

801
00:41:26,120 --> 00:41:29,560
Speaker 2: the payment is as you said, Okay, maybe they'll share

802
00:41:29,640 --> 00:41:33,839
Speaker 2: revenue with you because they can measure that savings. Is

803
00:41:33,920 --> 00:41:36,360
Speaker 2: that the model that you see across a range of

804
00:41:36,400 --> 00:41:40,320
Speaker 2: AI applications where it's like a sort of per task

805
00:41:40,480 --> 00:41:43,120
Speaker 2: or sort of very clearly linked to the efficiency game.

806
00:41:43,320 --> 00:41:45,960
Speaker 4: Yeah, we have another kind of line up business we

807
00:41:46,120 --> 00:41:50,600
Speaker 4: call the digital employees, which you know, Joe, probably you

808
00:41:50,640 --> 00:41:53,320
Speaker 4: have the similar experience that if you have one season

809
00:41:53,400 --> 00:41:55,880
Speaker 4: of the podcast, you're probably very energetic.

810
00:41:55,960 --> 00:41:56,120
Speaker 3: Right.

811
00:41:56,120 --> 00:41:58,200
Speaker 4: If you do that like ten twenty times in two

812
00:41:58,239 --> 00:42:02,040
Speaker 4: weeks is very exhaustive, right, because you think about you're.

813
00:42:01,880 --> 00:42:04,440
Speaker 2: Seeing you share my views, right, so way is that

814
00:42:04,480 --> 00:42:06,880
Speaker 2: what you're insinuated we're going to get place because we

815
00:42:06,920 --> 00:42:09,040
Speaker 2: get exhausted. But the AI didn't want.

816
00:42:08,920 --> 00:42:11,479
Speaker 4: Get it, so so you know. So, So my point

817
00:42:11,560 --> 00:42:15,880
Speaker 4: is the humans there, the motivation and the knowledge base

818
00:42:16,680 --> 00:42:20,080
Speaker 4: have their own kind of territory to be frank, but

819
00:42:20,640 --> 00:42:23,840
Speaker 4: if you look at the conversation, look at the quality

820
00:42:23,880 --> 00:42:26,000
Speaker 4: of them, know how, if you really tap in a

821
00:42:26,040 --> 00:42:29,160
Speaker 4: good manner, of course, the digital human actually can deliver

822
00:42:29,440 --> 00:42:33,960
Speaker 4: efficiency and better execution quality. So one example I just

823
00:42:33,960 --> 00:42:36,480
Speaker 4: want to share is e commerce is a big industry

824
00:42:36,480 --> 00:42:39,560
Speaker 4: in China. Yeah, and a lot of live performance is

825
00:42:39,640 --> 00:42:41,680
Speaker 4: really selling the products. It looks fun, but you know,

826
00:42:41,760 --> 00:42:46,080
Speaker 4: the come out is the kol cannot work like twenty

827
00:42:46,120 --> 00:42:49,440
Speaker 4: four hours, right, so and people cannot buying stuff like

828
00:42:49,440 --> 00:42:50,360
Speaker 4: twenty four hours.

829
00:42:50,640 --> 00:42:51,440
Speaker 5: But if you think.

830
00:42:51,320 --> 00:42:54,480
Speaker 4: About you have a great quality of the human employee

831
00:42:54,480 --> 00:42:57,680
Speaker 4: can help the merchant owners to sell into different time

832
00:42:57,760 --> 00:43:01,560
Speaker 4: zones and also can speak Chinese English and for the

833
00:43:01,600 --> 00:43:04,000
Speaker 4: different parts of audience to have the little jokes from

834
00:43:04,040 --> 00:43:06,799
Speaker 4: their own countries. They actually can help you know, the

835
00:43:06,800 --> 00:43:09,640
Speaker 4: ecommas revenue. So that's why we have their own kind

836
00:43:09,640 --> 00:43:13,960
Speaker 4: of product called digital Employees. We actually help our merchant

837
00:43:14,520 --> 00:43:17,560
Speaker 4: and the e commace store owners to really pushing on

838
00:43:17,600 --> 00:43:21,719
Speaker 4: that and selling all the goods. And it actually can

839
00:43:21,760 --> 00:43:24,080
Speaker 4: perform pretty well because on the you know, the Q

840
00:43:24,200 --> 00:43:27,239
Speaker 4: and A sessions on the questions is actually reacting to

841
00:43:27,320 --> 00:43:30,799
Speaker 4: the random users asking a wide range of questions. That

842
00:43:30,960 --> 00:43:33,319
Speaker 4: knowledge actually is very fluent in a way that for

843
00:43:33,400 --> 00:43:35,719
Speaker 4: the foundation model, it is the way it works, right,

844
00:43:36,000 --> 00:43:37,880
Speaker 4: So I think that actually has different user case and

845
00:43:37,920 --> 00:43:40,800
Speaker 4: we actually monitized by charging for example, the result improvement

846
00:43:40,840 --> 00:43:41,800
Speaker 4: and all the different things.

847
00:43:42,480 --> 00:43:45,280
Speaker 3: I've tried to buy things for twenty four hours straight before.

848
00:43:45,440 --> 00:43:48,520
Speaker 3: I think when I first used Cowbow, I think I

849
00:43:48,560 --> 00:43:51,440
Speaker 3: had Tawboo psychosis or something, and that's how my husband

850
00:43:51,440 --> 00:43:54,280
Speaker 3: and I ended up with three couches in our apartment.

851
00:43:54,719 --> 00:43:56,480
Speaker 3: There was about five hundred square feet large.

852
00:43:56,480 --> 00:43:58,799
Speaker 4: So my little suggestion you need to have another agent

853
00:43:58,880 --> 00:44:03,279
Speaker 4: help you to sell it right product. That's another way are.

854
00:44:03,200 --> 00:44:05,719
Speaker 3: You going to spin out the chips business. We're at

855
00:44:05,719 --> 00:44:08,640
Speaker 3: Bloomberg invest it's our first live recording of all lots.

856
00:44:08,760 --> 00:44:11,239
Speaker 4: Let's break some news, all right, So it's coming to

857
00:44:11,360 --> 00:44:15,560
Speaker 4: the to the money part. So I was not to

858
00:44:15,600 --> 00:44:18,000
Speaker 4: be Frank. I was not even trained by by Finance.

859
00:44:18,080 --> 00:44:22,080
Speaker 4: I become self by accident. So actually my BOSTH Bachelor

860
00:44:22,080 --> 00:44:26,160
Speaker 4: and the master training was a trip designer myself. So

861
00:44:26,239 --> 00:44:29,279
Speaker 4: I think after joining by DO, I definitely realized the

862
00:44:29,320 --> 00:44:32,919
Speaker 4: Bayous chip product has been really really high quality and

863
00:44:33,160 --> 00:44:35,319
Speaker 4: really good for the influence, for the for the all

864
00:44:35,320 --> 00:44:38,600
Speaker 4: the things we talked about helping our DAA to grow

865
00:44:38,680 --> 00:44:41,920
Speaker 4: as well. So based on the power of information, we

866
00:44:42,040 --> 00:44:45,359
Speaker 4: already filed the confidential filing for spin off of our

867
00:44:45,400 --> 00:44:49,080
Speaker 4: trip assets in Hong Kong and we are doing that

868
00:44:49,160 --> 00:44:53,719
Speaker 4: and processing that process on track, and that is one

869
00:44:53,800 --> 00:44:56,480
Speaker 4: part of the assets we try to unlock at this moment.

870
00:44:56,640 --> 00:44:59,759
Speaker 4: But however, as I mentioned the cloud of foundation model,

871
00:44:59,760 --> 00:45:02,319
Speaker 4: they are all very important. So after this being off,

872
00:45:02,360 --> 00:45:05,799
Speaker 4: we hopefully can enhancing that ego system. And as you'll know,

873
00:45:05,920 --> 00:45:08,880
Speaker 4: Chip is not only the hardware. I deeply understand, it

874
00:45:08,960 --> 00:45:12,239
Speaker 4: is about ecosystems. We need to work pretty well with

875
00:45:12,280 --> 00:45:16,160
Speaker 4: our customers and the suppliers and the software developers all

876
00:45:16,239 --> 00:45:19,440
Speaker 4: at one goal, and I think to be a separalistic

877
00:45:19,520 --> 00:45:22,480
Speaker 4: public company. It will help to achieve that goal not

878
00:45:22,560 --> 00:45:25,239
Speaker 4: only as a hardware, but also the entire ecosystem as well.

879
00:45:25,440 --> 00:45:28,720
Speaker 4: And our customer will view our trip products more neutral

880
00:45:28,719 --> 00:45:31,840
Speaker 4: and independent products that can actually do more testing and

881
00:45:31,960 --> 00:45:33,640
Speaker 4: more usage on their own cases.

882
00:45:33,760 --> 00:45:36,319
Speaker 2: Yeah, it seems to call reading that you figured out

883
00:45:36,360 --> 00:45:41,000
Speaker 2: a way that developers can easily port over their Kuda

884
00:45:41,200 --> 00:45:45,040
Speaker 2: stack over to your stack without much trouble. Henry, thank

885
00:45:45,080 --> 00:45:47,720
Speaker 2: you so much for coming on Our Loves our first

886
00:45:47,800 --> 00:45:51,440
Speaker 2: live recording anywhere in Asia. Really appreciate it. Again, it's

887
00:45:51,440 --> 00:45:52,120
Speaker 2: the perfect guest.

888
00:45:52,200 --> 00:45:53,360
Speaker 5: Yeah again, thanks for having me.

889
00:45:53,440 --> 00:45:59,280
Speaker 4: Thank you, thank you, Tracy, thank you.

890
00:46:08,040 --> 00:46:11,720
Speaker 3: That was our conversation with Henry Hoot, the CFO of Baidu,

891
00:46:11,920 --> 00:46:16,319
Speaker 3: recorded live at Bloomberg Asia invest I'm Tracy Alloway. You

892
00:46:16,320 --> 00:46:18,400
Speaker 3: can follow me at Tracy Alloway.

893
00:46:18,080 --> 00:46:20,719
Speaker 2: And I'm Joe Wisenthal. You can follow me at the Stalwart,

894
00:46:20,960 --> 00:46:24,080
Speaker 2: follow our producers Carmen Rodriguez at Carmen Erman, dash Ol

895
00:46:24,080 --> 00:46:27,759
Speaker 2: Bennett at Dashbot, Kilbrooks at Kilbrooks and Kevin Lozano at

896
00:46:27,840 --> 00:46:30,560
Speaker 2: Kevin Lloyd Lozano. And from our Odd Laws content. Go

897
00:46:30,600 --> 00:46:32,919
Speaker 2: to Bloomberg dot com slash odd Lots for the daily

898
00:46:33,000 --> 00:46:35,279
Speaker 2: newsletter and all of our episodes, and you can shut

899
00:46:35,320 --> 00:46:37,239
Speaker 2: about all of these topics twenty four to seven in

900
00:46:37,400 --> 00:46:40,800
Speaker 2: our discord discord dot gg slash outlines.

901
00:46:40,480 --> 00:46:42,359
Speaker 3: And if you enjoy odd Lots, if you like it

902
00:46:42,400 --> 00:46:45,399
Speaker 3: when we talk to Chinese tech executives, and please leave

903
00:46:45,440 --> 00:46:49,000
Speaker 3: us a positive review on your favorite podcast platform. And remember,

904
00:46:49,040 --> 00:46:51,960
Speaker 3: if you are a Bloomberg subscriber, you can listen to

905
00:46:52,040 --> 00:46:54,920
Speaker 3: all of our episodes absolutely ad free. All you need

906
00:46:54,960 --> 00:46:57,560
Speaker 3: to do is find the Bloomberg channel on Apple Podcasts

907
00:46:57,600 --> 00:47:21,919
Speaker 3: and follow the instructions there. Thanks for listening, Yeah,