1
00:00:00,080 --> 00:00:03,920
Speaker 1: Graylock Partner's partner Reid Hoffman says he is thrilled by

2
00:00:03,960 --> 00:00:07,160
Speaker 1: the news that Disney is extending Iiger's time as CEO.

3
00:00:07,600 --> 00:00:11,760
Speaker 1: Hoffman spoke with Bloomberg's Ed Ludlow on the sidelines of

4
00:00:11,920 --> 00:00:17,880
Speaker 1: Alan and Company's annual Sun Valley conference about Disney AI development, metas,

5
00:00:18,120 --> 00:00:19,360
Speaker 1: threads and more.

6
00:00:19,760 --> 00:00:22,799
Speaker 2: Reid Hoffman good off the name Bobyga. Do you know

7
00:00:22,840 --> 00:00:23,319
Speaker 2: who that is?

8
00:00:23,600 --> 00:00:25,840
Speaker 3: Absolutely? Bob and I have been fronds for decades.

9
00:00:25,960 --> 00:00:29,000
Speaker 2: Okay, So Bobyga has extended his stay. It's funny way

10
00:00:29,000 --> 00:00:30,920
Speaker 2: of putting it. At as CEO of Disney through twenty

11
00:00:30,960 --> 00:00:32,440
Speaker 2: twenty six, What do you make of that?

12
00:00:32,840 --> 00:00:35,320
Speaker 3: So Bob is I think one of the great CEOs

13
00:00:35,320 --> 00:00:38,080
Speaker 3: of our time. I think he cares a lot about Disney.

14
00:00:38,120 --> 00:00:41,360
Speaker 3: He's been a great steward. I think he's The news

15
00:00:41,360 --> 00:00:43,920
Speaker 3: doesn't surprise me. I didn't have any I learned about

16
00:00:43,920 --> 00:00:46,879
Speaker 3: it here, didn't. I didn't actually know in advance. But

17
00:00:47,760 --> 00:00:50,760
Speaker 3: I think it's part of what you want CEOs to

18
00:00:50,800 --> 00:00:53,880
Speaker 3: be doing, which is acting like owners and feeling that

19
00:00:53,880 --> 00:00:56,840
Speaker 3: the important thing is the legacy of the company that

20
00:00:56,880 --> 00:01:00,920
Speaker 3: they hand on to their successors, to employees, to society.

21
00:01:00,960 --> 00:01:02,120
Speaker 3: And I think Bob is that guy.

22
00:01:02,520 --> 00:01:05,240
Speaker 2: Excellent? Shall we talk about artificial intelligence?

23
00:01:05,480 --> 00:01:05,960
Speaker 3: Certainly?

24
00:01:06,240 --> 00:01:10,360
Speaker 2: You know, let's start in this single idea AI or

25
00:01:10,360 --> 00:01:13,720
Speaker 2: the development of as a meritocracy. I know that we've

26
00:01:13,760 --> 00:01:19,520
Speaker 2: discussed the principle before, but why is AI and meritocracy?

27
00:01:20,319 --> 00:01:22,840
Speaker 3: So there's a lot of different ways of looking at meritocracy.

28
00:01:22,840 --> 00:01:25,040
Speaker 3: And I actually like the world, even though the word

29
00:01:25,080 --> 00:01:28,280
Speaker 3: has gotten under some disfavor over time. I think of

30
00:01:28,319 --> 00:01:33,959
Speaker 3: AI is amplification intelligence, not artificial intelligence, as an amplifier

31
00:01:34,000 --> 00:01:36,800
Speaker 3: of human abilities. So I wrote this book, the first

32
00:01:36,840 --> 00:01:40,000
Speaker 3: book on AI, co written with AI, earlier this year, impromptu,

33
00:01:40,440 --> 00:01:42,560
Speaker 3: and part of it was to say, the way we

34
00:01:42,600 --> 00:01:46,760
Speaker 3: look at this is not as AI versus people, AI

35
00:01:46,880 --> 00:01:51,800
Speaker 3: versus humankind, but AI with humankind and as such. You say, well, now,

36
00:01:51,920 --> 00:01:54,600
Speaker 3: for example, I'm not a very good poet. I can

37
00:01:54,640 --> 00:01:57,080
Speaker 3: write sonnets. I could write you a sonnet for your birthday.

38
00:01:57,280 --> 00:02:00,440
Speaker 3: That's something I could do. I think what we're going

39
00:02:00,520 --> 00:02:03,280
Speaker 3: to have is every human being is going to have

40
00:02:03,320 --> 00:02:06,640
Speaker 3: a personal intelligence system that will help them on anything

41
00:02:06,720 --> 00:02:09,720
Speaker 3: from work tasks. But also like, oh, I had this

42
00:02:09,760 --> 00:02:12,640
Speaker 3: difficult conversation with a friend. I'm trying to figure it

43
00:02:12,680 --> 00:02:15,400
Speaker 3: out talk to me about it, and so someone who

44
00:02:15,520 --> 00:02:17,240
Speaker 3: talk to you about it with it. And I think

45
00:02:17,600 --> 00:02:24,239
Speaker 3: that's the meritocracy of enabling humanity to essentially anyone who

46
00:02:24,240 --> 00:02:27,760
Speaker 3: has access to a smart device, a smartphone, something else.

47
00:02:27,880 --> 00:02:31,919
Speaker 2: That's the broad principle. Then there's your personal relationship with AI.

48
00:02:32,040 --> 00:02:36,079
Speaker 2: There is Reid Hoffman the investor, Read Hoffman the co founder,

49
00:02:36,639 --> 00:02:39,079
Speaker 2: and Read Hoffman the academic for one of a better expression,

50
00:02:40,480 --> 00:02:42,120
Speaker 2: which one is winning at the moments?

51
00:02:43,800 --> 00:02:45,880
Speaker 3: Well, I hope they win together. I mean this is

52
00:02:45,880 --> 00:02:46,440
Speaker 3: one things.

53
00:02:46,560 --> 00:02:49,240
Speaker 2: Which one are you most focused on being at the moments?

54
00:02:50,000 --> 00:02:54,200
Speaker 3: Well, so these days it's always humanity first, right. So

55
00:02:55,400 --> 00:02:57,640
Speaker 3: but by the way, sometimes people kind of think, oh,

56
00:02:57,720 --> 00:03:01,000
Speaker 3: if you're doing an investment that's a profit company, it

57
00:03:01,040 --> 00:03:03,600
Speaker 3: can't be for humanity. I think, obviously it can be.

58
00:03:04,040 --> 00:03:07,120
Speaker 3: My own company LinkedIn was I think an example of that.

59
00:03:07,440 --> 00:03:09,400
Speaker 3: Airbnb is an example of that. I think there's a

60
00:03:09,440 --> 00:03:13,120
Speaker 3: lot of places where these tech companies amplify humanity in

61
00:03:13,160 --> 00:03:15,880
Speaker 3: really good ways. The same thing with Inflection. So I

62
00:03:15,880 --> 00:03:18,360
Speaker 3: think the same way with these AI companies, which is

63
00:03:18,520 --> 00:03:20,639
Speaker 3: I wouldn't be doing it if I didn't think it

64
00:03:20,680 --> 00:03:23,960
Speaker 3: would make individual people and societies better.

65
00:03:24,840 --> 00:03:27,120
Speaker 2: I had miss suffer soilium and on Bloomberg Technology with

66
00:03:27,200 --> 00:03:30,359
Speaker 2: media other day, your co founder and partner tommit to

67
00:03:30,400 --> 00:03:34,840
Speaker 2: a certain extent, he described having H one hundreds and

68
00:03:35,000 --> 00:03:38,400
Speaker 2: video GPUs in a way that made it seem like

69
00:03:38,480 --> 00:03:43,520
Speaker 2: a badge of honor. Why is having a cluster of

70
00:03:43,680 --> 00:03:46,560
Speaker 2: h one hundreds for our global audience that is in

71
00:03:46,680 --> 00:03:52,000
Speaker 2: Vidia's most powerful AI focused GPU graphics processing unit, a

72
00:03:52,080 --> 00:03:52,800
Speaker 2: badge of honor?

73
00:03:53,440 --> 00:03:56,800
Speaker 3: So what it really is is a badge of a

74
00:03:56,840 --> 00:04:01,960
Speaker 3: declaration of possibility and potential because because with this huge

75
00:04:01,960 --> 00:04:05,960
Speaker 3: cluster H one hundred you can train amazing AIS that

76
00:04:05,960 --> 00:04:09,880
Speaker 3: that Pie Inflections personal intelligence system isn't just going to

77
00:04:09,960 --> 00:04:13,440
Speaker 3: be high IQ but also high EQ, Like how can it?

78
00:04:13,520 --> 00:04:16,560
Speaker 3: How can it? Like when you say, oh, my friend's

79
00:04:17,680 --> 00:04:20,800
Speaker 3: a treasured dog just died, what should I do? And say, oh,

80
00:04:20,920 --> 00:04:22,760
Speaker 3: that must be really hard for your friend and really

81
00:04:22,839 --> 00:04:25,760
Speaker 3: hard for you. How what you know your friend? What's

82
00:04:25,800 --> 00:04:28,080
Speaker 3: a good way of being there for your friend and

83
00:04:28,120 --> 00:04:30,479
Speaker 3: being that not just like oh, well, normally you would

84
00:04:30,480 --> 00:04:32,479
Speaker 3: send a note and maybe in the note you would

85
00:04:32,480 --> 00:04:35,320
Speaker 3: express your condolences, Like no, that's not a EQ way

86
00:04:35,320 --> 00:04:38,320
Speaker 3: of expressing yes, And so what it is saying, we

87
00:04:38,440 --> 00:04:43,360
Speaker 3: have the computational capabilities to make amazing agents. Okay, one

88
00:04:43,400 --> 00:04:45,359
Speaker 3: of which is already live. You can go to Haypi

89
00:04:45,440 --> 00:04:46,440
Speaker 3: dot com and you can see it.

90
00:04:47,520 --> 00:04:52,040
Speaker 2: You have relationships with many members of the AI ecosystem,

91
00:04:52,040 --> 00:04:55,479
Speaker 2: and I want to understand how that works. You were

92
00:04:56,560 --> 00:04:59,039
Speaker 2: on the board of open AI. You are no longer.

93
00:04:59,400 --> 00:05:01,200
Speaker 2: You took it to decision to step down.

94
00:05:02,480 --> 00:05:06,920
Speaker 3: Explain that first, So open AI great organization, one of

95
00:05:06,960 --> 00:05:09,640
Speaker 3: the leading things that has made I work. I literally

96
00:05:09,720 --> 00:05:12,719
Speaker 3: spent two hours working with Sam Altman yesterday. Still continue

97
00:05:12,720 --> 00:05:15,599
Speaker 3: to help. Part of the challenges I put on a

98
00:05:15,600 --> 00:05:18,279
Speaker 3: blog post is that I was getting more and more

99
00:05:18,279 --> 00:05:20,920
Speaker 3: requests from all of my portfolio companies to get privileged

100
00:05:21,000 --> 00:05:23,120
Speaker 3: access to open AI. And part of when you're on

101
00:05:23,160 --> 00:05:25,479
Speaker 3: a five to one C three, I literally could say

102
00:05:25,520 --> 00:05:27,360
Speaker 3: I cannot help you at all. There's the front door

103
00:05:27,400 --> 00:05:29,159
Speaker 3: to open AI. I can't talk to anybody about it

104
00:05:29,160 --> 00:05:32,640
Speaker 3: because it would be a malfeasans on my five oh

105
00:05:32,640 --> 00:05:36,960
Speaker 3: one C three board of directors, you know, fooduciary responsibilities. Yes, literally,

106
00:05:37,360 --> 00:05:39,880
Speaker 3: the day after I stepped off the board, I said

107
00:05:39,920 --> 00:05:41,680
Speaker 3: that with him, said here are three companies and I

108
00:05:41,680 --> 00:05:44,400
Speaker 3: think would be really good for open AI and for

109
00:05:44,520 --> 00:05:47,560
Speaker 3: the company if you gave them privileged access to the

110
00:05:47,600 --> 00:05:50,440
Speaker 3: APIs and all their ass and so it's important not

111
00:05:50,560 --> 00:05:54,760
Speaker 3: just to have integrity, but to show in all quarters

112
00:05:54,760 --> 00:05:55,960
Speaker 3: that you're acting with integrity.

113
00:05:56,000 --> 00:05:59,039
Speaker 2: Okay, So then we can talk about Inflection AI just

114
00:05:59,120 --> 00:06:06,279
Speaker 2: a monthst first round that they raised they coexist. How

115
00:06:06,320 --> 00:06:08,040
Speaker 2: does that work in your shoes?

116
00:06:08,360 --> 00:06:13,600
Speaker 3: So for open AI five one C three beneficial AI

117
00:06:13,720 --> 00:06:18,799
Speaker 3: for humanity, Inflection public Benefit corporation. Right, So it's trying

118
00:06:18,839 --> 00:06:21,080
Speaker 3: to make value out of its equity, trying to make

119
00:06:21,120 --> 00:06:25,880
Speaker 3: up business and is not focused on AGI artificial general intelligence,

120
00:06:26,160 --> 00:06:29,640
Speaker 3: but like a personal intelligence for every human being. So

121
00:06:30,080 --> 00:06:32,400
Speaker 3: even though you get chat GPT where you say, well, okay,

122
00:06:32,440 --> 00:06:35,720
Speaker 3: there's a chatbot, there's lots of chatbots. Microsoft, which has

123
00:06:35,760 --> 00:06:39,400
Speaker 3: a relationship with open AI and Inflection has been chat

124
00:06:39,640 --> 00:06:43,440
Speaker 3: has a chatbot. There's going to be dozens of chatbots

125
00:06:43,480 --> 00:06:46,919
Speaker 3: that will fill different roles in people's lives. And so

126
00:06:47,880 --> 00:06:50,040
Speaker 3: you know, what I think chat GIBT will be was

127
00:06:50,120 --> 00:06:53,200
Speaker 3: kind of the equivalent of instant Wikipedia. What I think

128
00:06:53,240 --> 00:06:56,920
Speaker 3: Pie will be is kind of your personal buddy. What

129
00:06:57,040 --> 00:07:00,360
Speaker 3: I think bing chat will be is an amplification being

130
00:07:01,640 --> 00:07:04,240
Speaker 3: and they're not.

131
00:07:04,320 --> 00:07:07,039
Speaker 2: Koestler, another well known adventure capsist who you may be

132
00:07:07,040 --> 00:07:10,520
Speaker 2: familiar with, said to me a few weeks ago that

133
00:07:10,760 --> 00:07:16,320
Speaker 2: ninety percent of all of these AI focused startups AI

134
00:07:16,480 --> 00:07:20,440
Speaker 2: native startups will not survive. Do you agree with his

135
00:07:22,360 --> 00:07:23,160
Speaker 2: stance on that?

136
00:07:23,640 --> 00:07:27,120
Speaker 3: Well, so that's almost to some degree like an average technology.

137
00:07:28,120 --> 00:07:30,360
Speaker 3: You know, life and death rate with the one out

138
00:07:30,400 --> 00:07:32,600
Speaker 3: of ten succeed is kind of classic ritech company, so

139
00:07:33,280 --> 00:07:35,720
Speaker 3: as an application of a principle from the history might be.

140
00:07:36,240 --> 00:07:41,200
Speaker 3: But since we're in such an AI transformational wave, the

141
00:07:41,240 --> 00:07:44,280
Speaker 3: way I describe this as we're in the steam engine

142
00:07:44,280 --> 00:07:48,320
Speaker 3: of the mind, the cognitive industrial Revolution, my guess is

143
00:07:48,320 --> 00:07:52,080
Speaker 3: that might be many more surviving, just because it's such

144
00:07:52,120 --> 00:07:56,040
Speaker 3: a huge impact on every company, every individual, and every professional,

145
00:07:56,400 --> 00:08:00,160
Speaker 3: and so I would hazard higher percentage. But you know

146
00:08:00,200 --> 00:08:00,560
Speaker 3: who knows.

147
00:08:00,720 --> 00:08:03,960
Speaker 2: I look at gray Locks portfolio, that is what some

148
00:08:04,280 --> 00:08:10,320
Speaker 2: companies you may consider AI native, some AI adjacent perhaps

149
00:08:11,280 --> 00:08:13,560
Speaker 2: do you worry or to what extent do you worry

150
00:08:13,560 --> 00:08:14,880
Speaker 2: that some of them are kind of doing the same

151
00:08:14,880 --> 00:08:18,080
Speaker 2: thing as each other? And how differentiated are those in

152
00:08:18,120 --> 00:08:18,920
Speaker 2: your portfolio?

153
00:08:19,080 --> 00:08:24,320
Speaker 3: So you always have a worry that two companies will

154
00:08:24,320 --> 00:08:27,120
Speaker 3: track and competing on the exact same thing, and then

155
00:08:27,160 --> 00:08:30,120
Speaker 3: you have to navigate that if it happens, because technology

156
00:08:30,160 --> 00:08:33,080
Speaker 3: is always looking for what's the right way to create

157
00:08:33,120 --> 00:08:36,720
Speaker 3: an amazing you know, a product, service, an amazing business,

158
00:08:36,720 --> 00:08:39,400
Speaker 3: a transformation of an industry. But on the other hand,

159
00:08:39,640 --> 00:08:42,600
Speaker 3: more often than not, at the beginning stages, you think

160
00:08:42,600 --> 00:08:46,439
Speaker 3: they're gonna be competitive, like chatbot and chatbot, chatbots competing,

161
00:08:46,840 --> 00:08:49,600
Speaker 3: But this chatbot's gonna be end up being something that's

162
00:08:49,640 --> 00:08:52,360
Speaker 3: really great for like students, and this one's gonna end

163
00:08:52,440 --> 00:08:55,840
Speaker 3: up being really great for you know, large company workers,

164
00:08:55,920 --> 00:08:58,240
Speaker 3: and they're gonna have different functions, and this one's gonna

165
00:08:58,240 --> 00:09:01,680
Speaker 3: be you know, everyone's personal assistant, you know. So I

166
00:09:01,840 --> 00:09:06,079
Speaker 3: tend to think that they you we overworry about them competing,

167
00:09:06,160 --> 00:09:09,960
Speaker 3: and when the surface area is ten percent competing ninety

168
00:09:09,960 --> 00:09:11,480
Speaker 3: percent different, that's different.

169
00:09:11,960 --> 00:09:15,559
Speaker 2: I want you to think, as read Hoffmann the bench catalysts.

170
00:09:15,920 --> 00:09:18,760
Speaker 2: If you'll allow me to go down that route, LinkedIn

171
00:09:19,520 --> 00:09:25,480
Speaker 2: was acquired. Every Silicon Valley bounder's dream is the mega IPO.

172
00:09:26,160 --> 00:09:30,439
Speaker 2: But I do wonder if with so many AI startups

173
00:09:30,440 --> 00:09:34,520
Speaker 2: being founded, how much consolidation we might see an M and.

174
00:09:34,559 --> 00:09:37,040
Speaker 3: A oh, well, you're certainly going to see a bunch

175
00:09:37,080 --> 00:09:41,360
Speaker 3: of M and A because every company, certainly every tech

176
00:09:41,400 --> 00:09:43,760
Speaker 3: company and probably many other companies as well, are going

177
00:09:43,800 --> 00:09:46,360
Speaker 3: to need to have an AI strategy, right, and the

178
00:09:46,480 --> 00:09:48,560
Speaker 3: question is what's your AI strategy? How are you doing

179
00:09:48,559 --> 00:09:50,560
Speaker 3: it now? Some of them may work with a company

180
00:09:50,640 --> 00:09:54,280
Speaker 3: like Microsoft or or open ai or Inflection or other

181
00:09:54,360 --> 00:09:58,480
Speaker 3: and make that work. Others may you know, homegrown their own,

182
00:09:58,520 --> 00:10:00,400
Speaker 3: Others may buy. You're going to see a lot of

183
00:10:00,440 --> 00:10:02,840
Speaker 3: him and A. There's no question, you know, over the

184
00:10:02,880 --> 00:10:04,920
Speaker 3: next two to three years there'll be a whole bunch

185
00:10:04,960 --> 00:10:05,800
Speaker 3: of many activity.

186
00:10:06,000 --> 00:10:09,200
Speaker 2: There are many names in this space. One of those

187
00:10:09,280 --> 00:10:12,320
Speaker 2: names is Elon Musk and I don't know if you saw,

188
00:10:12,400 --> 00:10:16,880
Speaker 2: but he on x do AI published in information about

189
00:10:16,880 --> 00:10:19,240
Speaker 2: what they're up to and who they are as a team.

190
00:10:19,960 --> 00:10:21,080
Speaker 2: What do you make of all of that?

191
00:10:21,280 --> 00:10:24,520
Speaker 3: So Elon is one of the great entrepreneurs of our times.

192
00:10:25,600 --> 00:10:29,680
Speaker 3: I think, you know, open AI owes its foundational existence

193
00:10:29,720 --> 00:10:32,000
Speaker 3: to his work, so I think it's you know, he

194
00:10:32,120 --> 00:10:34,920
Speaker 3: was a massive contributor. I think Sam was the was

195
00:10:34,920 --> 00:10:37,560
Speaker 3: one of the drivers, but Elon was there shaping it,

196
00:10:37,679 --> 00:10:40,360
Speaker 3: helping it. So I think it's great. I think that

197
00:10:41,160 --> 00:10:44,480
Speaker 3: he cares deeply about humanity, which I think is also

198
00:10:44,600 --> 00:10:48,400
Speaker 3: great news. You know, I kind of look a little

199
00:10:48,440 --> 00:10:51,160
Speaker 3: bit askuy and sit signing a six month pause letter

200
00:10:51,280 --> 00:10:53,800
Speaker 3: while you're trying to accelerate your own effort and things

201
00:10:53,880 --> 00:10:56,760
Speaker 3: like that. But I think he's I think he is

202
00:10:56,800 --> 00:11:00,680
Speaker 3: a great entrepreneur to have amongst the entrepreneurs doing these things.

203
00:11:00,840 --> 00:11:06,520
Speaker 2: I tweeted that Xai had published this list of people.

204
00:11:06,840 --> 00:11:09,640
Speaker 2: What surprised me. So many people responded straight away saying,

205
00:11:10,320 --> 00:11:12,200
Speaker 2: every single one of the people in that list is

206
00:11:12,200 --> 00:11:14,720
Speaker 2: a man. There are no women on that list, and

207
00:11:14,760 --> 00:11:18,319
Speaker 2: they were really concerned by that in the context of

208
00:11:18,360 --> 00:11:21,839
Speaker 2: developing large language models. Was your thoughts on that?

209
00:11:22,120 --> 00:11:25,040
Speaker 3: So, I mean, these things we're going to have as

210
00:11:25,080 --> 00:11:28,480
Speaker 3: kind of transformative impact, it's very important that they are

211
00:11:29,280 --> 00:11:33,760
Speaker 3: kind of inclusive of different voices and considerations. And so

212
00:11:33,960 --> 00:11:36,240
Speaker 3: every effort that I'm part of, we make a real

213
00:11:36,280 --> 00:11:38,160
Speaker 3: effort not just on gender, which is of course an

214
00:11:38,160 --> 00:11:41,559
Speaker 3: important thing, yes, but also on race, on geographic regions,

215
00:11:42,240 --> 00:11:45,079
Speaker 3: all of that kind of diversity inclusion actually matters because

216
00:11:45,559 --> 00:11:48,240
Speaker 3: these things are going to have such an impact on humanity.

217
00:11:48,360 --> 00:11:51,720
Speaker 3: If you don't do that, you may do something bad.

218
00:11:52,800 --> 00:11:55,960
Speaker 2: The United States has a lot going on in the

219
00:11:55,960 --> 00:12:00,120
Speaker 2: theater of AI that the question increasingly asked is what

220
00:12:00,120 --> 00:12:02,839
Speaker 2: do we know about China's progress in the field of AI?

221
00:12:03,160 --> 00:12:06,480
Speaker 2: What is your assessment of where they are technologically? Can

222
00:12:06,520 --> 00:12:07,720
Speaker 2: you is there a way of knowing?

223
00:12:08,400 --> 00:12:10,960
Speaker 3: There's no way of fully knowing, because one of the

224
00:12:10,960 --> 00:12:14,000
Speaker 3: benefits and challenges we have of being part of this

225
00:12:14,360 --> 00:12:17,240
Speaker 3: great Western system as we have a free press, lots

226
00:12:17,240 --> 00:12:18,840
Speaker 3: of reporting all the rest. I think one of the

227
00:12:18,840 --> 00:12:21,760
Speaker 3: thing is it's great about our society, and that's more

228
00:12:21,840 --> 00:12:24,319
Speaker 3: challenged within China, so you don't have as much visibility.

229
00:12:24,640 --> 00:12:27,959
Speaker 3: Now that being said, they have amazing technologists, they have

230
00:12:28,040 --> 00:12:33,119
Speaker 3: amazing technical companies. They are very dedicated to building technology

231
00:12:33,160 --> 00:12:35,920
Speaker 3: for the future, and so I think they will have

232
00:12:35,960 --> 00:12:39,240
Speaker 3: a number of great AI efforts. Their AI efforts are

233
00:12:39,240 --> 00:12:42,480
Speaker 3: obviously somewhat challenged by the fact that large language models

234
00:12:42,520 --> 00:12:46,040
Speaker 3: are difficult to control, and so can you get can

235
00:12:46,080 --> 00:12:48,679
Speaker 3: you make sure a very scale large language model can

236
00:12:48,720 --> 00:12:53,680
Speaker 3: both generate all of this interesting material and also not say, well,

237
00:12:53,679 --> 00:12:57,199
Speaker 3: here is ten funny jokes about choosing ping, or here's

238
00:12:57,240 --> 00:13:00,520
Speaker 3: the argument about why the Chinese Communist Party is actually

239
00:13:00,520 --> 00:13:02,480
Speaker 3: not a democracy that you had to vote for once,

240
00:13:03,000 --> 00:13:07,240
Speaker 3: and the Chinese government system doesn't want that in any

241
00:13:07,280 --> 00:13:09,240
Speaker 3: of its Internet or digital things. So that's more challenge.

242
00:13:09,240 --> 00:13:11,520
Speaker 2: And the data that they're training the model zone, why

243
00:13:11,559 --> 00:13:14,959
Speaker 2: is that an important point of differentiation, not just the

244
00:13:15,040 --> 00:13:17,520
Speaker 2: volume of data, but what is in the data.

245
00:13:17,679 --> 00:13:21,079
Speaker 3: Well, data is a complicated story when it comes to

246
00:13:21,120 --> 00:13:23,199
Speaker 3: a high So when you're doing these large language models,

247
00:13:23,520 --> 00:13:26,120
Speaker 3: what you really need is volume of data versus the

248
00:13:27,360 --> 00:13:30,600
Speaker 3: specific like, oh do you have this really high quality data? Now,

249
00:13:30,640 --> 00:13:34,040
Speaker 3: high quality data become useful, useful for medical applications, useful

250
00:13:34,080 --> 00:13:37,719
Speaker 3: for coding, et cetera. And by the way, the Chinese

251
00:13:38,480 --> 00:13:42,080
Speaker 3: general lack of privacy controls on data could make some

252
00:13:42,120 --> 00:13:44,800
Speaker 3: of their efforts much stronger. The challenge, of course, so

253
00:13:44,800 --> 00:13:47,560
Speaker 3: when you do that with large language model approaches is

254
00:13:47,600 --> 00:13:52,080
Speaker 3: those large language model approaches don't necessarily go here's how

255
00:13:52,120 --> 00:13:55,640
Speaker 3: I can obey CCP regulations. Yes, right, and that's going

256
00:13:55,679 --> 00:13:56,880
Speaker 3: to be one of the challenges that are going to have.

257
00:13:57,640 --> 00:14:00,360
Speaker 2: We could talk for the rest of the week AI

258
00:14:00,960 --> 00:14:03,880
Speaker 2: for sure, but we won't threads.

259
00:14:04,200 --> 00:14:07,800
Speaker 3: Yes, So look, I think Facebook has done a meta

260
00:14:07,920 --> 00:14:11,040
Speaker 3: has done a great product launch. It's a high quality product.

261
00:14:11,480 --> 00:14:15,439
Speaker 3: I think it's because of turbulence around social media, they

262
00:14:15,440 --> 00:14:19,280
Speaker 3: have a natural launch for that. I think by the way,

263
00:14:19,280 --> 00:14:22,440
Speaker 3: Instagram has done things like this before, like they launched Stories,

264
00:14:22,520 --> 00:14:25,320
Speaker 3: and their Stories launch was very successful and was part

265
00:14:25,320 --> 00:14:29,400
Speaker 3: of became part of the Instagram ecosystem. So we'll see

266
00:14:29,440 --> 00:14:32,400
Speaker 3: how it stays, but its launch was amazing.

267
00:14:32,720 --> 00:14:36,080
Speaker 2: Mark Zuckerberg himself has been really active and engaged on

268
00:14:36,240 --> 00:14:39,080
Speaker 2: the in the early days of the platform with its users.

269
00:14:40,040 --> 00:14:41,840
Speaker 2: I just wondered if you could reflect on what you

270
00:14:41,880 --> 00:14:45,800
Speaker 2: see in Mark now versus when all those years ago

271
00:14:45,880 --> 00:14:47,120
Speaker 2: when Facebook started.

272
00:14:47,720 --> 00:14:51,920
Speaker 3: Well, one of Mark's strengths, which is the strength of

273
00:14:52,320 --> 00:14:55,640
Speaker 3: great entrepreneurs, is he's an infinite learner, so he's been

274
00:14:55,720 --> 00:14:58,320
Speaker 3: learning since then, Like what's the way that you what

275
00:14:58,400 --> 00:15:00,200
Speaker 3: are the things that you pay attention to? How do

276
00:15:00,200 --> 00:15:03,600
Speaker 3: you launch good products? Like this is an enormously successful

277
00:15:03,640 --> 00:15:07,000
Speaker 3: product launch from an established companies, and established companies don't

278
00:15:07,040 --> 00:15:10,400
Speaker 3: that often have new successful product launches, and this is

279
00:15:10,440 --> 00:15:13,160
Speaker 3: one of them. And so that's I think a testimony

280
00:15:13,200 --> 00:15:14,400
Speaker 3: to what he's learned over the years.

281
00:15:14,400 --> 00:15:16,560
Speaker 2: Have you had a chance to discuss threads with him?

282
00:15:16,600 --> 00:15:17,600
Speaker 3: I have not. I hope too.

283
00:15:18,560 --> 00:15:22,320
Speaker 2: What is it that you think threads has an advantage

284
00:15:22,400 --> 00:15:26,840
Speaker 2: over other similar social media platforms and what is its disadvantage?

285
00:15:28,000 --> 00:15:30,520
Speaker 3: So I think its advantage is that it can do

286
00:15:30,680 --> 00:15:33,400
Speaker 3: a it can do. It can try to do a

287
00:15:33,440 --> 00:15:37,760
Speaker 3: new launch without some of the baggage that that other

288
00:15:37,880 --> 00:15:40,560
Speaker 3: social media like we all learn as we go, Like

289
00:15:40,600 --> 00:15:43,200
Speaker 3: there's things like I would have launched LinkedIn with a

290
00:15:43,200 --> 00:15:47,200
Speaker 3: whole influencers network if I'd had that idea at the beginning, right,

291
00:15:47,240 --> 00:15:50,600
Speaker 3: it's a vital part of keeping this kind of it's

292
00:15:50,640 --> 00:15:54,600
Speaker 3: about business and vision and leadership on LinkedIn. Undoubtedly they

293
00:15:54,600 --> 00:15:57,080
Speaker 3: have similar kind of whatever their plans are for threads,

294
00:15:57,320 --> 00:16:01,200
Speaker 3: I think it's applying those from a from an from

295
00:16:01,440 --> 00:16:05,720
Speaker 3: a launch like part of it is immediately benefiting from

296
00:16:05,760 --> 00:16:08,520
Speaker 3: your Instagram account as you knew that when you launched it,

297
00:16:08,600 --> 00:16:10,840
Speaker 3: you didn't have a cold start problem. That was one

298
00:16:10,880 --> 00:16:13,680
Speaker 3: of the things that is a classic network building that

299
00:16:13,720 --> 00:16:18,320
Speaker 3: they've navigated around I think fairly elegantly. But you know,

300
00:16:18,520 --> 00:16:22,280
Speaker 3: I I would anticipate that they have a whole bunch

301
00:16:22,280 --> 00:16:22,880
Speaker 3: of plans that.

302
00:16:22,840 --> 00:16:25,080
Speaker 2: We have yet to see and on the disadvantage that

303
00:16:25,120 --> 00:16:28,600
Speaker 2: it may have over a guess for itself against other platforms.

304
00:16:28,880 --> 00:16:32,040
Speaker 3: So disadvantage, well, typically it's when you try to start

305
00:16:32,080 --> 00:16:34,920
Speaker 3: from fresh, you're starting cold start network problem. They've solved

306
00:16:34,920 --> 00:16:39,120
Speaker 3: that one. Look, I personally hope that they are doing

307
00:16:39,160 --> 00:16:42,560
Speaker 3: more on the try to keep it to a more

308
00:16:43,360 --> 00:16:47,600
Speaker 3: positive communications culture versus which I think Instagram is generally

309
00:16:47,960 --> 00:16:51,000
Speaker 3: more than you know, kind of Facebook itself. So I

310
00:16:51,040 --> 00:16:53,120
Speaker 3: hope for that, but that would be a that's a question.

311
00:16:54,480 --> 00:16:58,680
Speaker 2: Final thing is can threads and Twitter both exist together?

312
00:16:59,760 --> 00:17:03,440
Speaker 3: For sure, they can, but they may end up competing too,

313
00:17:03,480 --> 00:17:05,639
Speaker 3: but that still means they could exist together.

314
00:17:06,240 --> 00:17:11,320
Speaker 1: That's Reid Hoffman, partner at Greylock Partners, speaking with Bloomberg's

315
00:17:11,480 --> 00:17:15,199
Speaker 1: Ed Ludlow in Sun Valley. For more conversations like this,

316
00:17:15,359 --> 00:17:20,040
Speaker 1: subscribe to the Bloomberg Talks podcast, available on your favorite

317
00:17:20,040 --> 00:17:21,359
Speaker 1: podcast platform