1
00:00:10,200 --> 00:00:14,200
Speaker 1: Hello, and welcome to another episode of the Odd Lots Podcast.

2
00:00:14,240 --> 00:00:16,600
Speaker 2: I'm Joe Wisenthal and I'm Tracy Alloway.

3
00:00:16,840 --> 00:00:19,960
Speaker 1: Tracy, I mean, I think it goes without saying that

4
00:00:20,040 --> 00:00:24,439
Speaker 1: the appetite and interest in anything related to AI and

5
00:00:24,520 --> 00:00:27,280
Speaker 1: making money from AI continues unabated.

6
00:00:27,960 --> 00:00:30,920
Speaker 2: Yes, I think that's accurate. Well, it's interesting because you

7
00:00:31,000 --> 00:00:33,760
Speaker 2: kind of see it from two different sides at the moment.

8
00:00:33,840 --> 00:00:35,760
Speaker 2: So there are a lot of companies that are talking

9
00:00:35,800 --> 00:00:40,440
Speaker 2: about investing internally in AI technology, and then there are

10
00:00:40,440 --> 00:00:45,080
Speaker 2: a lot of investors talking about investing in AI in

11
00:00:45,159 --> 00:00:46,040
Speaker 2: one way or another.

12
00:00:46,560 --> 00:00:48,880
Speaker 1: I mean needless to say. You know, you like throw

13
00:00:48,920 --> 00:00:51,440
Speaker 1: a dart at nasdextoc some they're talking about, you know,

14
00:00:51,720 --> 00:00:54,760
Speaker 1: the way AI incorporated. It's also funny because like you

15
00:00:54,880 --> 00:00:57,400
Speaker 1: like read the uh the earnings transcript of like a

16
00:00:57,440 --> 00:01:01,280
Speaker 1: grocery storagechep. Yes, and the seats we'll be talking about,

17
00:01:01,320 --> 00:01:04,200
Speaker 1: you know, how AI is going to help them.

18
00:01:04,400 --> 00:01:05,280
Speaker 2: Wasn't this Kroger?

19
00:01:05,520 --> 00:01:07,400
Speaker 1: I think so. And to be fair, like I think

20
00:01:07,440 --> 00:01:10,560
Speaker 1: they actually kind of legitimately have been investing. So it's

21
00:01:10,760 --> 00:01:15,039
Speaker 1: totally but like they mentioned AI one company I don't

22
00:01:15,480 --> 00:01:17,240
Speaker 1: like mention it like a dozen times.

23
00:01:17,000 --> 00:01:18,959
Speaker 2: I'm pretty sure it was Kroger. Yeah, But I think

24
00:01:19,000 --> 00:01:21,319
Speaker 2: I've said this on the podcast before. It does feel

25
00:01:21,440 --> 00:01:23,360
Speaker 2: like there are some people out there who at this

26
00:01:23,440 --> 00:01:27,039
Speaker 2: point are basically using AI as a synonym for any

27
00:01:27,080 --> 00:01:30,560
Speaker 2: type of software, just like we use software, we're using AI.

28
00:01:31,040 --> 00:01:36,320
Speaker 2: But it begs the question of how to smartly invest

29
00:01:36,400 --> 00:01:38,720
Speaker 2: in a technology that clearly a lot of people are

30
00:01:38,840 --> 00:01:42,600
Speaker 2: enthused about. But it's also kind of hard to disaggregate

31
00:01:42,680 --> 00:01:45,400
Speaker 2: a lot of the marketing from the reality.

32
00:01:45,560 --> 00:01:47,480
Speaker 1: Well, and the thing that gets me in that I'm

33
00:01:47,520 --> 00:01:50,720
Speaker 1: still trying to wrap my head around too, is like,

34
00:01:51,240 --> 00:01:54,040
Speaker 1: especially for the big tech incumbents, And I'm thinking of

35
00:01:54,120 --> 00:01:57,360
Speaker 1: like Alphabet, Google or whatever, Like they have an amazing

36
00:01:57,400 --> 00:02:00,640
Speaker 1: business model right now, right, like people search for something

37
00:02:00,760 --> 00:02:04,080
Speaker 1: and then you're telling the machine exactly what you are

38
00:02:04,120 --> 00:02:06,360
Speaker 1: looking for, and then the machine knows, okay, well, here

39
00:02:06,360 --> 00:02:08,920
Speaker 1: are some ads that we can put And I know,

40
00:02:09,080 --> 00:02:12,120
Speaker 1: like obviously Google is you know, a head very like

41
00:02:12,240 --> 00:02:15,000
Speaker 1: front of the curve in terms of AI tech, and

42
00:02:15,040 --> 00:02:17,160
Speaker 1: they have their own large language models and all of

43
00:02:17,160 --> 00:02:19,760
Speaker 1: that stuff. But does anyone know that like this is

44
00:02:19,800 --> 00:02:22,120
Speaker 1: going to turn into like a money making thing for them?

45
00:02:22,400 --> 00:02:22,480
Speaker 2: Right?

46
00:02:22,680 --> 00:02:25,720
Speaker 1: Will it be anywhere close to this amazing money printing

47
00:02:25,760 --> 00:02:29,200
Speaker 1: machine that they built when they built the Google search box.

48
00:02:29,360 --> 00:02:31,600
Speaker 2: Well, I think that's a really good question. And also,

49
00:02:31,760 --> 00:02:35,480
Speaker 2: so far we have seen the incumbents come out as

50
00:02:35,560 --> 00:02:39,480
Speaker 2: the big winners of a lot of this new technology,

51
00:02:39,880 --> 00:02:43,560
Speaker 2: and I think a that's been unexpected. If you'd ask someone,

52
00:02:43,720 --> 00:02:45,359
Speaker 2: you know, five years ago, who was going to be

53
00:02:45,400 --> 00:02:47,840
Speaker 2: the big winner in AI, I don't think anyone would

54
00:02:47,880 --> 00:02:52,280
Speaker 2: have said microsom right, right, And so that also raises

55
00:02:52,280 --> 00:02:54,440
Speaker 2: a question of Okay, how did the giants monetize this?

56
00:02:54,639 --> 00:02:57,480
Speaker 2: To your point? And then secondly, are there going to

57
00:02:57,520 --> 00:03:00,600
Speaker 2: be new players who somehow come in and find a

58
00:03:00,600 --> 00:03:01,560
Speaker 2: better way to do it right?

59
00:03:01,639 --> 00:03:03,440
Speaker 1: Like how disruptive it will be?

60
00:03:03,760 --> 00:03:04,200
Speaker 3: Yeah?

61
00:03:04,240 --> 00:03:07,040
Speaker 1: Well I am excited because I believe we do have

62
00:03:08,120 --> 00:03:11,000
Speaker 1: the perfect guest. We're going to be speaking with Josh Wolf,

63
00:03:11,280 --> 00:03:15,000
Speaker 1: founding partner and managing director at Lux Capital, who has

64
00:03:15,080 --> 00:03:19,080
Speaker 1: been investing in AI since long before it was cool,

65
00:03:19,120 --> 00:03:23,040
Speaker 1: long before everyone started asking like chat gpt to like,

66
00:03:23,440 --> 00:03:26,040
Speaker 1: you know, write a song about the FED and the

67
00:03:26,080 --> 00:03:29,000
Speaker 1: style of Johnny Cash or whatever like long before No

68
00:03:29,160 --> 00:03:34,320
Speaker 1: I feel seen, I feel I'm describing myself. I'm describing

69
00:03:34,360 --> 00:03:36,400
Speaker 1: both of us long before we were all doing that,

70
00:03:36,840 --> 00:03:38,560
Speaker 1: and so he's going to talk to us about how

71
00:03:38,600 --> 00:03:42,000
Speaker 1: he thinks about making money in AI and where the

72
00:03:42,080 --> 00:03:45,160
Speaker 1: value is going to crue in identifying investments. So, Josh,

73
00:03:45,200 --> 00:03:47,000
Speaker 1: thank you so much for coming on Odd Laps.

74
00:03:47,320 --> 00:03:48,800
Speaker 3: Joe Tracy, great to be with you.

75
00:03:48,960 --> 00:03:51,000
Speaker 1: Can I ask you a question for real? Like we

76
00:03:51,120 --> 00:03:54,440
Speaker 1: obviously have this boom, etc. But the tech's been around.

77
00:03:54,520 --> 00:03:57,360
Speaker 1: Did people basically just get excited because like someone finally

78
00:03:57,360 --> 00:04:00,120
Speaker 1: put a good ux on top of this technology for

79
00:04:00,120 --> 00:04:02,200
Speaker 1: a while and suddenly like, oh my god, Like was

80
00:04:02,280 --> 00:04:07,280
Speaker 1: that sort of what catalyzed this current stage of enthusiasm.

81
00:04:08,520 --> 00:04:12,200
Speaker 3: I think the lay answer is what catalyzed this was

82
00:04:12,240 --> 00:04:15,560
Speaker 3: a sense of conjuring magic. People felt like they were

83
00:04:15,640 --> 00:04:18,560
Speaker 3: effectively casting spells, like you said, whether it was a

84
00:04:18,640 --> 00:04:20,840
Speaker 3: Johnny Cash song conjured to you know, to talk about

85
00:04:20,839 --> 00:04:24,080
Speaker 3: the FED, but it's the feeling that somebody had a superpower.

86
00:04:24,240 --> 00:04:26,200
Speaker 3: So I think that's what catalyzed that. Where it was

87
00:04:26,240 --> 00:04:29,080
Speaker 3: a feeling of positive surprise where people were like, oh

88
00:04:29,120 --> 00:04:32,080
Speaker 3: my god, like I just created magic and that you know,

89
00:04:32,120 --> 00:04:36,599
Speaker 3: classic cliche that any sufficiently advanced tech is indistinguishable for magic.

90
00:04:36,640 --> 00:04:40,440
Speaker 3: This felt like magic now the roots of it go

91
00:04:40,560 --> 00:04:44,520
Speaker 3: back over a decade, and that's what the public doesn't see.

92
00:04:44,560 --> 00:04:46,560
Speaker 3: And part of that is the bones, and part of

93
00:04:46,600 --> 00:04:49,960
Speaker 3: it is the brains the tech infrastructure. So you start

94
00:04:50,279 --> 00:04:54,039
Speaker 3: with the GPUs. Now, we've known computing for decades and

95
00:04:54,120 --> 00:04:58,400
Speaker 3: Intel was the dominant force. Intel made CPU central processing units,

96
00:04:58,600 --> 00:05:00,120
Speaker 3: and there was this thing off on the side that

97
00:05:00,200 --> 00:05:02,599
Speaker 3: was just doing graphic processing and it was for video games.

98
00:05:03,080 --> 00:05:06,440
Speaker 3: I've got this mental model, this framework where a lot

99
00:05:06,480 --> 00:05:10,120
Speaker 3: of people say that the most dangerous words in investing are,

100
00:05:10,400 --> 00:05:13,560
Speaker 3: this time is different. I actually think that there's a

101
00:05:13,640 --> 00:05:15,960
Speaker 3: secret that people can follow, which is that the most

102
00:05:16,040 --> 00:05:19,719
Speaker 3: valuable words are whenever you hear a parent say it

103
00:05:19,760 --> 00:05:23,520
Speaker 3: will rot your brain, that basically presages it predicts the

104
00:05:23,560 --> 00:05:26,400
Speaker 3: next ten billion dollar industry. So think about this. I

105
00:05:26,400 --> 00:05:30,800
Speaker 3: mean literally nineteen sixties those hip shaken, Johnny Cash, Elvis,

106
00:05:30,839 --> 00:05:32,919
Speaker 3: you know, rock and Roll, It'll rot your brain. Boom

107
00:05:33,000 --> 00:05:37,280
Speaker 3: ten billion dollar industry. You know, seventies, you know personal

108
00:05:37,320 --> 00:05:40,960
Speaker 3: computers and chat rooms, and and you know nineties the

109
00:05:41,000 --> 00:05:45,520
Speaker 3: Internet and these online chat rooms and then gaming. My god,

110
00:05:45,600 --> 00:05:47,880
Speaker 3: you know, these kids are turning into couch potatoes. Get

111
00:05:47,920 --> 00:05:50,839
Speaker 3: them off the video games, the video game players of

112
00:05:50,920 --> 00:05:54,080
Speaker 3: yesterday are today's you know, robotic surgeons and drone pilots.

113
00:05:54,080 --> 00:05:56,600
Speaker 3: But what's really important is the tech that was underlying

114
00:05:56,600 --> 00:06:00,720
Speaker 3: that these massively multiplayer games and people demand and ever

115
00:06:00,839 --> 00:06:07,200
Speaker 3: higher video resolution and PlayStation competing with Xbox, completing with Nintendo.

116
00:06:07,640 --> 00:06:10,640
Speaker 3: It created these chips that took in video from a

117
00:06:10,800 --> 00:06:14,040
Speaker 3: fifteen billion dollar market cap at the time when Intel

118
00:06:14,120 --> 00:06:16,120
Speaker 3: was one hundred and fifty billion a few years ago,

119
00:06:16,480 --> 00:06:20,120
Speaker 3: and today it's a trillion dollar business. We had invested

120
00:06:20,120 --> 00:06:24,039
Speaker 3: in this company going back about eight years. It was

121
00:06:24,120 --> 00:06:27,960
Speaker 3: four people off the Stanford campus, your classic garage, and

122
00:06:28,000 --> 00:06:30,120
Speaker 3: they were literally in a garage at the slack the

123
00:06:30,120 --> 00:06:33,880
Speaker 3: Stanford Linear Accelerator, sort of secret group, and they were

124
00:06:33,920 --> 00:06:36,640
Speaker 3: trying to develop self driving cars. And we had put

125
00:06:36,680 --> 00:06:40,160
Speaker 3: about twenty five million into this small team called Zookes,

126
00:06:40,640 --> 00:06:42,839
Speaker 3: which was a zoo for robotics. It was a silly name.

127
00:06:43,400 --> 00:06:45,960
Speaker 3: And we go in there and I see all these

128
00:06:45,960 --> 00:06:48,400
Speaker 3: people playing video games and I start to get a

129
00:06:48,440 --> 00:06:50,359
Speaker 3: little bit upset. I say, we just put a lot

130
00:06:50,400 --> 00:06:52,960
Speaker 3: of money into this company. What are all these engineers doing.

131
00:06:53,440 --> 00:06:55,400
Speaker 3: And the founder turned to me and said, no, no,

132
00:06:55,440 --> 00:06:58,640
Speaker 3: you don't understand the cars that you see outside on

133
00:06:58,720 --> 00:07:01,720
Speaker 3: these tracks that are running, they're ingesting information at the

134
00:07:01,800 --> 00:07:05,120
Speaker 3: rate of one second per second, what we call reality,

135
00:07:05,480 --> 00:07:08,560
Speaker 3: and they're taking light ar and radar and visual cues

136
00:07:08,640 --> 00:07:11,720
Speaker 3: and thermal sensors and vibrational sensors and all that and

137
00:07:11,720 --> 00:07:14,920
Speaker 3: they're processing it. But inside these rooms air conditioned, these

138
00:07:14,960 --> 00:07:17,440
Speaker 3: people are not playing video games. This is not grand

139
00:07:17,480 --> 00:07:20,200
Speaker 3: theft or this is not call of duty. The machines

140
00:07:20,240 --> 00:07:23,160
Speaker 3: are actually running simulations and they are training the cars

141
00:07:23,280 --> 00:07:26,080
Speaker 3: thousands of simulations a second, and the machines don't know

142
00:07:26,080 --> 00:07:28,880
Speaker 3: the difference between reality and simulation. And I was like,

143
00:07:28,920 --> 00:07:30,920
Speaker 3: oh my god, okay, this is pretty amazing. What are

144
00:07:30,960 --> 00:07:32,960
Speaker 3: these things running on And they said, those two guys

145
00:07:33,000 --> 00:07:34,880
Speaker 3: over there are from this company, in Vidia, and we

146
00:07:34,920 --> 00:07:37,160
Speaker 3: have chips that aren't on the market yet and they

147
00:07:37,160 --> 00:07:39,960
Speaker 3: are able to do processing like never before. And that

148
00:07:40,160 --> 00:07:43,920
Speaker 3: sent us on this path as investors in AI. And

149
00:07:44,560 --> 00:07:48,080
Speaker 3: from there we found this amazing team that was developing

150
00:07:48,120 --> 00:07:51,920
Speaker 3: like the next gen GPUs and was a guy Navine Row.

151
00:07:52,120 --> 00:07:55,400
Speaker 3: He had a company called Nirvana Systems. Intel buys them

152
00:07:55,440 --> 00:07:57,360
Speaker 3: within a year of US investing for about three hundred

153
00:07:57,360 --> 00:08:01,240
Speaker 3: and fifty million, becomes the core kernel of Intel's AI system.

154
00:08:01,680 --> 00:08:04,320
Speaker 3: We going back that guy again, and just last week

155
00:08:04,560 --> 00:08:07,160
Speaker 3: Data Bricks bought his new company called Mosaic for a

156
00:08:07,200 --> 00:08:10,920
Speaker 3: billion three congratulations, thank you, wild Ride. But you just

157
00:08:10,960 --> 00:08:14,239
Speaker 3: try to find these people that have this irreverent view

158
00:08:14,840 --> 00:08:16,800
Speaker 3: and they sort of see the future and they invented,

159
00:08:16,840 --> 00:08:17,840
Speaker 3: and you get behind them.

160
00:08:18,200 --> 00:08:21,600
Speaker 2: So maybe if I could just step back for a second,

161
00:08:21,960 --> 00:08:24,840
Speaker 2: and this will sort of maybe tell us a little

162
00:08:24,840 --> 00:08:26,920
Speaker 2: bit more about what you're doing in the space. But

163
00:08:27,000 --> 00:08:32,160
Speaker 2: how do you evaluate AI opportunities between the hardware? So

164
00:08:32,320 --> 00:08:35,000
Speaker 2: the chips where they're so far seems to have been

165
00:08:35,200 --> 00:08:38,319
Speaker 2: a lot of excitement and activity versus some of the

166
00:08:38,920 --> 00:08:41,200
Speaker 2: software and the sort of underlying models.

167
00:08:41,520 --> 00:08:43,520
Speaker 3: Today in video really has the lead. It's very hard

168
00:08:43,559 --> 00:08:47,240
Speaker 3: for people to compete. Obviously, there's all kinds of considerations

169
00:08:47,280 --> 00:08:51,680
Speaker 3: of geopolitical dependency and TSMC and ASML and who's helping

170
00:08:51,720 --> 00:08:55,200
Speaker 3: to make these chips. But there's an entire very sophisticated

171
00:08:55,280 --> 00:09:00,960
Speaker 3: stack of semicab equipment manufacturing, IP design so that people

172
00:09:00,960 --> 00:09:03,679
Speaker 3: can make these chips and then the chips themselves. And

173
00:09:03,880 --> 00:09:06,320
Speaker 3: these things are very expensive. I mean, these h one

174
00:09:06,400 --> 00:09:09,199
Speaker 3: hundred chips from Nvidia one hundred thousand dollars. They are

175
00:09:09,200 --> 00:09:12,680
Speaker 3: in scarce supply. One of the other really interesting things

176
00:09:12,720 --> 00:09:15,280
Speaker 3: right now in this chip domain that people should watch for.

177
00:09:15,960 --> 00:09:17,400
Speaker 3: And then I'll tell you where I think in Vidia

178
00:09:17,480 --> 00:09:20,400
Speaker 3: is actually quite vulnerable and they're not just pure monopoly here.

179
00:09:21,400 --> 00:09:23,880
Speaker 3: Anytime that there's hype in a sector, just like you

180
00:09:23,880 --> 00:09:26,480
Speaker 3: were talking about, you know, Kroger's adding AI to their name.

181
00:09:26,640 --> 00:09:28,280
Speaker 3: You saw this in the dot coms, you saw it

182
00:09:28,320 --> 00:09:32,800
Speaker 3: in the Internet, you saw it in mobile everything exactly.

183
00:09:33,200 --> 00:09:35,680
Speaker 3: And you know, look, they lower the cost of capital,

184
00:09:35,960 --> 00:09:40,000
Speaker 3: they take advantage of people's irrationality, they capitalize. And what

185
00:09:40,120 --> 00:09:42,920
Speaker 3: happens in every field, the hype gets high, the cost

186
00:09:42,960 --> 00:09:45,760
Speaker 3: of capital gets low, Hundreds, if not thousands of new

187
00:09:45,760 --> 00:09:48,560
Speaker 3: companies get funded, ninety nine percent of them fail, and

188
00:09:48,600 --> 00:09:51,440
Speaker 3: from that detritus it becomes the comminatorial fodder for the

189
00:09:51,440 --> 00:09:56,000
Speaker 3: next wave. So interestingly, with crypto, crypto people were like

190
00:09:56,200 --> 00:09:58,880
Speaker 3: clamoring for these GPUs. They couldn't get enough of them

191
00:09:58,920 --> 00:10:01,920
Speaker 3: so that they could do bitcoin mining. As that market went,

192
00:10:02,120 --> 00:10:05,439
Speaker 3: you know, hyperbolic, and then crashed. Now you have all

193
00:10:05,440 --> 00:10:08,760
Speaker 3: these xsgpus and you're starting to read headlines about the

194
00:10:08,760 --> 00:10:11,360
Speaker 3: bitcoin miners that are selling their GPUs now to the

195
00:10:11,400 --> 00:10:13,680
Speaker 3: AI researchers, and they're able to do it now, you know,

196
00:10:13,760 --> 00:10:16,080
Speaker 3: cents on the dollar of what they paid before. So

197
00:10:16,240 --> 00:10:20,160
Speaker 3: on the hardware side, that's very interesting. But our bed

198
00:10:20,280 --> 00:10:22,439
Speaker 3: is that there's something different that's happening than betting on

199
00:10:22,480 --> 00:10:26,640
Speaker 3: the next chip. Okay, there's More's law which everybody knows about.

200
00:10:26,720 --> 00:10:29,080
Speaker 3: There's Rocks Law, which basically is that the cost of

201
00:10:29,120 --> 00:10:32,840
Speaker 3: these semiconductor fabs increases exponentially even as our chips get cheaper.

202
00:10:32,840 --> 00:10:33,960
Speaker 1: And sorry, which law?

203
00:10:34,559 --> 00:10:37,840
Speaker 3: Which law Rocks named after Arthur Rock, who is one

204
00:10:37,840 --> 00:10:40,440
Speaker 3: of the you know, first vcs, one of the first

205
00:10:40,679 --> 00:10:45,320
Speaker 3: funders of Intel in East Coast og VC. So they

206
00:10:45,400 --> 00:10:48,040
Speaker 3: called it Rocks law because basically, the cost to build

207
00:10:48,040 --> 00:10:50,960
Speaker 3: these fabs, to make the chips keeps getting more expensive

208
00:10:51,000 --> 00:10:52,839
Speaker 3: every year and a half or two years. It used

209
00:10:52,840 --> 00:10:54,720
Speaker 3: to be you know, one hundred million, then it's a billion,

210
00:10:54,800 --> 00:10:56,840
Speaker 3: and it's ten billion and so on. Okay, why is

211
00:10:56,840 --> 00:11:00,319
Speaker 3: that relevant to make these foundation models that we're all

212
00:11:00,400 --> 00:11:03,920
Speaker 3: using behind the scenes, GPT three and GPT four and

213
00:11:03,960 --> 00:11:06,720
Speaker 3: what comes next. It used to be a few million

214
00:11:06,760 --> 00:11:10,120
Speaker 3: dollars maybe ten million to train GPT three GPT four

215
00:11:10,720 --> 00:11:13,160
Speaker 3: is estimated in the low hundreds of millions of dollars.

216
00:11:13,520 --> 00:11:16,160
Speaker 3: And whatever comes next, people believe is going to cost

217
00:11:16,200 --> 00:11:18,480
Speaker 3: about a billion dollars. Why because they have to buy

218
00:11:18,520 --> 00:11:21,119
Speaker 3: all these in video chips. So in video is telling everybody,

219
00:11:21,280 --> 00:11:23,160
Speaker 3: you've got to get these a one hundred chips. We

220
00:11:23,200 --> 00:11:26,280
Speaker 3: make these h one hundred chips. We make. The reality

221
00:11:26,520 --> 00:11:29,200
Speaker 3: is actually that there's this interesting vulnerability. This is where

222
00:11:29,240 --> 00:11:31,280
Speaker 3: we make our speculative bets. Now we might be wrong,

223
00:11:31,320 --> 00:11:34,000
Speaker 3: but this is what we're betting. We're betting that that's

224
00:11:34,040 --> 00:11:36,120
Speaker 3: not going to be the case, that it's not going

225
00:11:36,160 --> 00:11:39,160
Speaker 3: to be just the domain of open AI, that it's

226
00:11:39,200 --> 00:11:41,320
Speaker 3: not going to be anthropic, that it's not going to

227
00:11:41,320 --> 00:11:43,040
Speaker 3: be just the big giants. And we'll talk about how

228
00:11:43,040 --> 00:11:45,960
Speaker 3: those guys are all intertwined, like you said, with the

229
00:11:46,000 --> 00:11:48,719
Speaker 3: Microsofts and the Googles and the Metas, etc. Because that's

230
00:11:48,720 --> 00:11:52,160
Speaker 3: an interesting dynamic. In Vidia has a language, a computer

231
00:11:52,280 --> 00:11:55,520
Speaker 3: language that people program on and it's called Kuda Cuda,

232
00:11:56,320 --> 00:11:59,520
Speaker 3: and this has been the dominant form, but it's really vulnerable,

233
00:11:59,720 --> 00:12:04,400
Speaker 3: and it's vulnerable interestingly because of Facebook Meta. They came

234
00:12:04,480 --> 00:12:07,240
Speaker 3: up with this language called Pi Torch and a lot

235
00:12:07,280 --> 00:12:10,200
Speaker 3: of the developers are moving to pitorch. It's open source,

236
00:12:11,160 --> 00:12:14,280
Speaker 3: it allows people to do a lot of the AI processing,

237
00:12:14,360 --> 00:12:17,120
Speaker 3: but hardware agnostic, meaning you don't have to use an

238
00:12:17,120 --> 00:12:19,439
Speaker 3: in video chip in video. If you use in video chip,

239
00:12:19,440 --> 00:12:22,480
Speaker 3: you've got a program on Kuda. These guys are saying,

240
00:12:22,679 --> 00:12:25,199
Speaker 3: we can use an AMD chip, we can use an

241
00:12:25,240 --> 00:12:29,880
Speaker 3: ASK an application specific integrated circuit. They are saying, we're

242
00:12:29,920 --> 00:12:32,160
Speaker 3: not going to be beholden to this. So there's two

243
00:12:32,200 --> 00:12:35,360
Speaker 3: competing software languages that are merging sort of quietly. One

244
00:12:35,440 --> 00:12:38,840
Speaker 3: is called PyTorch and one is called Triton. And Triton

245
00:12:38,880 --> 00:12:41,880
Speaker 3: is from open Ai. People probably trust that one a

246
00:12:41,920 --> 00:12:44,480
Speaker 3: little bit less, and Pietorch is totally open source but

247
00:12:45,080 --> 00:12:47,480
Speaker 3: originated from Meta, which is really interesting.

248
00:13:03,800 --> 00:13:06,000
Speaker 1: Man, I'm already learning a lot because I was not

249
00:13:06,120 --> 00:13:10,320
Speaker 1: familiar with PyTorch. You know. So obviously, as you mentioned,

250
00:13:10,360 --> 00:13:14,920
Speaker 1: and we talked about GPT three, GPT four, chat GPT,

251
00:13:15,360 --> 00:13:20,120
Speaker 1: this magical search box that got that got everyone's attention.

252
00:13:20,360 --> 00:13:22,160
Speaker 1: This sort of like what came out of open Ai.

253
00:13:22,280 --> 00:13:24,719
Speaker 1: But there are others that are building chat pots. How

254
00:13:24,720 --> 00:13:27,080
Speaker 1: many winners can there be? A like sort of like

255
00:13:27,360 --> 00:13:30,240
Speaker 1: how do you think about winners? Either at the foundational

256
00:13:30,360 --> 00:13:34,640
Speaker 1: model level, like if we started odd lots GPT, is

257
00:13:34,640 --> 00:13:38,319
Speaker 1: that a worthwhile area or are you thinking like some

258
00:13:38,400 --> 00:13:40,720
Speaker 1: of these problems are kind of solved and it makes

259
00:13:40,760 --> 00:13:44,040
Speaker 1: more sense to focus on building something on top of

260
00:13:44,080 --> 00:13:47,720
Speaker 1: one of these like core winners like GPT four and

261
00:13:47,720 --> 00:13:50,960
Speaker 1: build some sort of a specific application for an industry

262
00:13:51,360 --> 00:13:54,360
Speaker 1: that uses a foundational model that already exists.

263
00:13:54,600 --> 00:13:57,160
Speaker 3: You're thinking exactly right, and you know you should be

264
00:13:57,160 --> 00:13:59,600
Speaker 3: a VC because we're betting on the ladder that you're

265
00:13:59,640 --> 00:14:04,079
Speaker 3: gonna start to get these generalized models which are wowing everybody,

266
00:14:04,200 --> 00:14:05,760
Speaker 3: although you know, you start to look at some of

267
00:14:05,760 --> 00:14:08,680
Speaker 3: the usage pattern classic thing, right, people get really excited

268
00:14:08,679 --> 00:14:10,560
Speaker 3: about the thing, then it starts to die off. Maybe

269
00:14:10,559 --> 00:14:13,080
Speaker 3: it's the summer, but maybe it's reached a little bit

270
00:14:13,120 --> 00:14:16,600
Speaker 3: of a plateau of incremental interest. Okay, so let's break

271
00:14:16,600 --> 00:14:19,160
Speaker 3: this down in terms of the models. You've got open AI,

272
00:14:19,360 --> 00:14:21,280
Speaker 3: which will continue to invest a huge amount of money,

273
00:14:21,320 --> 00:14:24,640
Speaker 3: continue to develop models, continue to wow people. Their next

274
00:14:24,640 --> 00:14:27,400
Speaker 3: thing will be quote unquote multi modal instead of just

275
00:14:27,480 --> 00:14:30,880
Speaker 3: text or voice transcription. You'll start to have all kinds

276
00:14:30,880 --> 00:14:33,280
Speaker 3: of interesting things where like you said, you know, make

277
00:14:33,360 --> 00:14:36,800
Speaker 3: me the Johnny Cash song, give me a full music video,

278
00:14:38,480 --> 00:14:40,560
Speaker 3: print out, all kinds of crazy images. You know, it'll

279
00:14:40,640 --> 00:14:43,440
Speaker 3: it'll do four different things, and that'll be really limited

280
00:14:43,440 --> 00:14:45,680
Speaker 3: by people's creativity. But it's going to be general. Now.

281
00:14:45,720 --> 00:14:48,560
Speaker 3: One of the problems with the general stuff GPT four

282
00:14:48,680 --> 00:14:51,640
Speaker 3: was Plint was trained on the public Internet, and proud

283
00:14:51,680 --> 00:14:53,320
Speaker 3: of the problem in the public Internet is that you

284
00:14:53,360 --> 00:14:54,880
Speaker 3: got a lot of information, but you also have a

285
00:14:54,920 --> 00:14:57,840
Speaker 3: lot of misinformation. It was trained on Reddit and Twitter

286
00:14:57,880 --> 00:15:00,640
Speaker 3: and all kinds of repositories of public info, and so

287
00:15:00,760 --> 00:15:03,360
Speaker 3: it's going to hallucinate. It's going to give you bs answers.

288
00:15:03,800 --> 00:15:07,320
Speaker 3: So you have people saying, Okay, that's a problem, there's

289
00:15:07,360 --> 00:15:10,320
Speaker 3: white space, let me solve it. And it's probably going

290
00:15:10,400 --> 00:15:13,800
Speaker 3: to be financial and healthcare is our guests, where you

291
00:15:13,880 --> 00:15:18,240
Speaker 3: get very specialized models that need to have high accuracy,

292
00:15:18,680 --> 00:15:20,960
Speaker 3: and they're going to be smaller models. So instead of

293
00:15:21,000 --> 00:15:24,240
Speaker 3: these giant models, they're going to be smaller, more bespoke,

294
00:15:24,720 --> 00:15:28,800
Speaker 3: more industry verticalized. Even Bloomberg Bloomberg GPT people I think

295
00:15:28,840 --> 00:15:31,160
Speaker 3: are really fascinated by what that's going to portend. Because

296
00:15:31,200 --> 00:15:34,400
Speaker 3: you have a proprietary data set, you've got a locked

297
00:15:34,440 --> 00:15:36,320
Speaker 3: and user base and sort of a social network and

298
00:15:36,360 --> 00:15:39,600
Speaker 3: you have reliable, high quality data. So I think that's

299
00:15:39,640 --> 00:15:41,640
Speaker 3: going to be the next wave. It's going to happen

300
00:15:41,680 --> 00:15:44,360
Speaker 3: in financial data. My bet is on Bloomberg. It's going

301
00:15:44,440 --> 00:15:47,240
Speaker 3: to be in healthcare with some of the major healthcare systems,

302
00:15:47,840 --> 00:15:51,360
Speaker 3: and I think that that's sort of the next wave. Now,

303
00:15:51,800 --> 00:15:54,520
Speaker 3: when you look at the big players today with the

304
00:15:54,520 --> 00:15:57,800
Speaker 3: big foundation models, open ai, if you're really honest about it,

305
00:15:58,240 --> 00:16:02,160
Speaker 3: they're captive to Microsoft. Microsoft did an incredibly clever deal.

306
00:16:02,280 --> 00:16:04,920
Speaker 3: They knew that there's no way that the DOJA or

307
00:16:04,960 --> 00:16:07,840
Speaker 3: the FTC would allow them to actually acquire open ai,

308
00:16:08,000 --> 00:16:09,600
Speaker 3: so they structured a deal in a way that they

309
00:16:09,600 --> 00:16:14,040
Speaker 3: effectively control it, but without doing an acquisition. You look

310
00:16:14,080 --> 00:16:17,840
Speaker 3: at Google, they're closely tied up with Anthropic, and a

311
00:16:17,840 --> 00:16:20,120
Speaker 3: lot of these deals are interesting because what happens is

312
00:16:20,160 --> 00:16:23,360
Speaker 3: the company gets a giant equity investment. In this case,

313
00:16:23,360 --> 00:16:25,640
Speaker 3: I think Anthropic got about three hundred million from Google,

314
00:16:25,840 --> 00:16:29,720
Speaker 3: but that money sored around trips. Google gets equity, Anthropic

315
00:16:29,720 --> 00:16:32,640
Speaker 3: gets cash. That cash then goes back to Google and

316
00:16:32,720 --> 00:16:35,040
Speaker 3: is spent on compute, so they get to book it

317
00:16:35,040 --> 00:16:39,480
Speaker 3: as revenue in Google Cloud. Now Meta is really interesting

318
00:16:39,520 --> 00:16:41,960
Speaker 3: because just like you said before, Tracy, nobody would have

319
00:16:42,000 --> 00:16:44,600
Speaker 3: thought Microsoft was the leader or would be a leader

320
00:16:44,720 --> 00:16:49,520
Speaker 3: in AI. Meta, you know, has been under congressional scrutiny

321
00:16:49,600 --> 00:16:52,680
Speaker 3: and has been the sort of evil villain of consumer

322
00:16:52,800 --> 00:16:56,080
Speaker 3: and social media and disrupting and destroying our democracy and

323
00:16:56,120 --> 00:16:58,320
Speaker 3: all this stuff. And then they made this bet on

324
00:16:58,360 --> 00:17:02,160
Speaker 3: the metaverse, which nobody cares about, but this idea of

325
00:17:02,200 --> 00:17:06,080
Speaker 3: this fetiverse that they're starting to talk about with threads.

326
00:17:06,480 --> 00:17:09,840
Speaker 3: It's really interesting because they are embracing this idea of

327
00:17:09,880 --> 00:17:12,760
Speaker 3: open source. Now. They're not doing it benevolently because they

328
00:17:12,760 --> 00:17:14,520
Speaker 3: think it's a good thing. It's in their self interest.

329
00:17:14,880 --> 00:17:16,800
Speaker 3: They want to be the sort of network that is

330
00:17:16,800 --> 00:17:19,440
Speaker 3: connected to everything else. They don't want to be siloed,

331
00:17:19,680 --> 00:17:21,320
Speaker 3: and they get to use it as a little bit

332
00:17:21,320 --> 00:17:24,800
Speaker 3: of achine to stave off the regulatory scrutiny and the

333
00:17:24,840 --> 00:17:27,560
Speaker 3: public criticism. But I think you've got to watch Meta

334
00:17:27,640 --> 00:17:31,119
Speaker 3: really closely. In the coming weeks, new releases of open

335
00:17:31,160 --> 00:17:33,760
Speaker 3: source models that are going to really compete with open AI,

336
00:17:34,359 --> 00:17:38,520
Speaker 3: lots of partnerships with interesting companies knowing that they themselves

337
00:17:38,640 --> 00:17:40,520
Speaker 3: couldn't possibly do an acquisition.

338
00:17:40,840 --> 00:17:43,520
Speaker 2: You mentioned data just then, And this is something I've

339
00:17:43,520 --> 00:17:47,400
Speaker 2: been thinking about. But when it comes to AI technology,

340
00:17:47,520 --> 00:17:53,280
Speaker 2: what's the most important factor? Is it access to reliable

341
00:17:53,320 --> 00:17:56,720
Speaker 2: data as you mentioned, maybe reliable and exclusive sources of

342
00:17:56,720 --> 00:17:59,719
Speaker 2: big data, or is it the sort of like underlying

343
00:18:00,080 --> 00:18:04,240
Speaker 2: modeling technology. And I guess another way of framing it

344
00:18:04,320 --> 00:18:06,560
Speaker 2: is like, are the big winners going to just be

345
00:18:06,880 --> 00:18:11,200
Speaker 2: companies like you know, banks, insurers that have huge data

346
00:18:11,200 --> 00:18:12,439
Speaker 2: sets that they can do things with.

347
00:18:12,840 --> 00:18:16,359
Speaker 3: I think so. I think that today it has been

348
00:18:16,480 --> 00:18:20,439
Speaker 3: a little bit of ignorance arbitrage, meaning the people that

349
00:18:20,600 --> 00:18:22,720
Speaker 3: really were in the know were the model makers, the

350
00:18:22,760 --> 00:18:25,640
Speaker 3: people that could design the algorithms to do the predictive

351
00:18:25,680 --> 00:18:29,760
Speaker 3: analysis and make the models. All those models are either

352
00:18:30,000 --> 00:18:32,960
Speaker 3: held proprietary in the case of like open AI, or

353
00:18:33,000 --> 00:18:34,880
Speaker 3: in the case of one of our companies, which has

354
00:18:35,040 --> 00:18:37,480
Speaker 3: one of the most powerful repositories and one of the

355
00:18:37,520 --> 00:18:41,359
Speaker 3: most ridiculous names, Hugging Face. One of my partner's amazing

356
00:18:41,359 --> 00:18:44,760
Speaker 3: guy Brandon Reeves. He says, you know, there's these French

357
00:18:44,800 --> 00:18:48,400
Speaker 3: PhD computer scientists and mathematicians. They're hanging out in Brooklyn

358
00:18:48,760 --> 00:18:50,880
Speaker 3: and they've got this company called Hugging Face. And here's

359
00:18:50,880 --> 00:18:53,800
Speaker 3: the irony they started out as a chatbot, you know,

360
00:18:53,840 --> 00:18:57,159
Speaker 3: almost like Joaquin Phoenix and her, and then they became

361
00:18:57,560 --> 00:19:00,840
Speaker 3: this open source repository for all all the models, and

362
00:19:00,880 --> 00:19:04,520
Speaker 3: now hundreds of thousands of models, including corporate models that

363
00:19:04,560 --> 00:19:06,840
Speaker 3: are hosted there and constantly improving, and it's all open source.

364
00:19:07,400 --> 00:19:11,119
Speaker 3: And the irony is that OpenAI started as this open

365
00:19:11,200 --> 00:19:13,920
Speaker 3: model company has become the world's greatest chatbot. So it's

366
00:19:13,920 --> 00:19:17,920
Speaker 3: sort of inverse. So hugging Face is making these models,

367
00:19:18,040 --> 00:19:21,280
Speaker 3: and they were the beneficiary very early on of everybody

368
00:19:21,320 --> 00:19:23,760
Speaker 3: trying to deploy the models. They could run them on

369
00:19:23,840 --> 00:19:27,160
Speaker 3: Hugging Face, they could use the cloud compute that they provide.

370
00:19:27,800 --> 00:19:30,640
Speaker 3: And now, Tracy, to your point, you're starting to see

371
00:19:30,680 --> 00:19:33,199
Speaker 3: people saying, Okay, the thing that we want to do

372
00:19:33,320 --> 00:19:35,879
Speaker 3: is build on top of all of these models. What

373
00:19:36,000 --> 00:19:39,280
Speaker 3: was expensive and scarce and rare before was the compute

374
00:19:39,680 --> 00:19:42,919
Speaker 3: and the algorithms, and those are becoming increasingly abundant. So

375
00:19:43,000 --> 00:19:47,159
Speaker 3: what is scarce today reliable data and proprietary data and

376
00:19:47,240 --> 00:19:49,480
Speaker 3: the data sets, like you said, it could be big banks,

377
00:19:49,920 --> 00:19:54,080
Speaker 3: could be consumer data, could be Amazon retail spending information

378
00:19:54,200 --> 00:19:57,960
Speaker 3: could be Spotify with user's behavior, it could be healthcare

379
00:19:58,000 --> 00:20:02,359
Speaker 3: systems appropriately anonymized and protected and compliant with HIPPA, but

380
00:20:02,600 --> 00:20:05,720
Speaker 3: being able to collect all this information and have it

381
00:20:06,680 --> 00:20:10,040
Speaker 3: do high quality inference and training. So you have to

382
00:20:10,080 --> 00:20:12,240
Speaker 3: train the models on the data, and then you have

383
00:20:12,320 --> 00:20:14,600
Speaker 3: to be able to do the predictability, which is the

384
00:20:14,640 --> 00:20:18,000
Speaker 3: inference from somebody putting in a prompt. I will say

385
00:20:18,040 --> 00:20:21,160
Speaker 3: the area that we're probably the most excited about, which

386
00:20:21,200 --> 00:20:23,760
Speaker 3: is not something that the everydayly person is going to

387
00:20:23,760 --> 00:20:25,600
Speaker 3: spend time doing. They're not going to be making those

388
00:20:25,640 --> 00:20:28,600
Speaker 3: Johnny Cash songs or conjuring images on mid Journey and

389
00:20:28,720 --> 00:20:34,200
Speaker 3: Dolly is biology. And the key breakthrough here is there's

390
00:20:34,240 --> 00:20:36,639
Speaker 3: something in all these models called a context window, and

391
00:20:36,640 --> 00:20:38,679
Speaker 3: all it means is basically how much information you can

392
00:20:38,720 --> 00:20:40,600
Speaker 3: put it. And if you ever tried to take like

393
00:20:40,600 --> 00:20:43,160
Speaker 3: a long transcript let's say, of odd lots and throw

394
00:20:43,200 --> 00:20:44,880
Speaker 3: it into a context window, it might say, oh, it's

395
00:20:44,880 --> 00:20:50,080
Speaker 3: too long, right. The context window for open AI has

396
00:20:50,080 --> 00:20:53,639
Speaker 3: been about eight thousand, eight thousand tokens Anthropic now has

397
00:20:53,680 --> 00:20:56,600
Speaker 3: one hundred thousand, and that's growing. What that means is

398
00:20:56,640 --> 00:20:58,400
Speaker 3: the amount of data that you can put into a

399
00:20:58,440 --> 00:21:03,280
Speaker 3: single prompt is growing exponentially. If you think about the

400
00:21:03,320 --> 00:21:05,639
Speaker 3: human genome, if you think about genetic data, where you

401
00:21:05,640 --> 00:21:07,640
Speaker 3: have millions of tokens that you need to be able

402
00:21:07,640 --> 00:21:10,280
Speaker 3: to put into this effectively. That is the next domain

403
00:21:10,320 --> 00:21:12,760
Speaker 3: where you're able to put in huge amounts of information

404
00:21:13,200 --> 00:21:16,000
Speaker 3: and do all kinds of predictive things, from designing new

405
00:21:16,000 --> 00:21:19,320
Speaker 3: proteins to discovering drugs. And that's an area where not

406
00:21:19,359 --> 00:21:23,359
Speaker 3: only are the markets enormous, the information and the expertise

407
00:21:23,440 --> 00:21:25,800
Speaker 3: is very narrow and specialized. And I think it's going

408
00:21:25,840 --> 00:21:28,840
Speaker 3: to completely upturn farm and biotech in a giant way.

409
00:21:29,280 --> 00:21:33,280
Speaker 1: That's really interesting. You mentioned you're talking about some of

410
00:21:33,320 --> 00:21:37,160
Speaker 1: these investments that the hyperscalers, the tech giants have made,

411
00:21:37,560 --> 00:21:40,520
Speaker 1: and I hadn't really appreciated that dynamic before. It's sort

412
00:21:40,560 --> 00:21:44,280
Speaker 1: of like easier to link it easier for Microsoft to

413
00:21:44,359 --> 00:21:46,439
Speaker 1: link up with an open AI than to make a

414
00:21:46,440 --> 00:21:48,520
Speaker 1: big acquisition that's going to get on the you know,

415
00:21:48,560 --> 00:21:52,919
Speaker 1: the headlines for regulators. It's easier for you know, alphabet

416
00:21:53,080 --> 00:21:57,159
Speaker 1: and anthropic et cetera as the VC. And also the

417
00:21:57,200 --> 00:21:59,120
Speaker 1: point about how a lot of that cash just comes

418
00:21:59,160 --> 00:22:01,240
Speaker 1: back in terms of all these companies' compute bill is

419
00:22:01,280 --> 00:22:04,000
Speaker 1: extremely interesting as a VC. Can you talk a little

420
00:22:04,000 --> 00:22:07,119
Speaker 1: bit about this dynamic there seems to be a lot

421
00:22:07,200 --> 00:22:11,880
Speaker 1: because of this of corporate VC in AI specifically, and

422
00:22:11,920 --> 00:22:14,200
Speaker 1: how that sort of changes the game as a non

423
00:22:14,280 --> 00:22:17,720
Speaker 1: corporate VC or as an independent VC firm when you're

424
00:22:17,760 --> 00:22:22,119
Speaker 1: thinking about evaluating companies, the presence of these big the

425
00:22:22,240 --> 00:22:25,600
Speaker 1: VC arms of the large corporations and how that sort

426
00:22:25,600 --> 00:22:26,760
Speaker 1: of changes the game.

427
00:22:27,560 --> 00:22:30,480
Speaker 3: Great, great question, And I'll give you sort of three

428
00:22:30,640 --> 00:22:33,840
Speaker 3: quick angles here. The first is how we think about

429
00:22:33,920 --> 00:22:36,840
Speaker 3: ultimately making money and exiting as a VC, and how

430
00:22:36,840 --> 00:22:39,919
Speaker 3: these people play in the ecosystem. The second is a

431
00:22:39,960 --> 00:22:43,960
Speaker 3: related question about how do we ultimately exit our companies,

432
00:22:44,080 --> 00:22:46,680
Speaker 3: meaning how do we sell them to a large incumbent

433
00:22:47,000 --> 00:22:49,720
Speaker 3: when you have all this congressional scrutiny and it's very

434
00:22:49,760 --> 00:22:54,399
Speaker 3: improbable today that Microsoft or Meta could do a big acquisition.

435
00:22:55,160 --> 00:22:57,639
Speaker 3: It's just you know, it's a regulatory and problem. And

436
00:22:57,640 --> 00:22:59,560
Speaker 3: then the third is a geopolitical angle here that I

437
00:22:59,600 --> 00:23:01,359
Speaker 3: think is stually going to change that. So on the

438
00:23:01,400 --> 00:23:05,040
Speaker 3: first one, I always say that it sounds a little

439
00:23:05,080 --> 00:23:08,000
Speaker 3: bit cheesy, but we do a lot of hard signs investing,

440
00:23:08,040 --> 00:23:09,639
Speaker 3: a lot of deep tech investing, and I like to

441
00:23:09,640 --> 00:23:13,199
Speaker 3: say like the first law of thermodynamics, energy is not

442
00:23:13,240 --> 00:23:17,240
Speaker 3: created or destroyed. Risk and value are not created or destroyed.

443
00:23:17,240 --> 00:23:19,479
Speaker 3: They just change form. And every risk that I can

444
00:23:19,520 --> 00:23:23,080
Speaker 3: identify in an early stage company AI or biotech or aerospace,

445
00:23:23,119 --> 00:23:26,320
Speaker 3: whatever it is, if I can kill that risk, if

446
00:23:26,359 --> 00:23:29,400
Speaker 3: I can actually say, Okay, there's financing risk or tech risk,

447
00:23:29,520 --> 00:23:32,440
Speaker 3: or management risk or product risk or customer risk, whatever

448
00:23:32,440 --> 00:23:36,399
Speaker 3: it is, kill that risk. A later investor coming after

449
00:23:36,520 --> 00:23:39,240
Speaker 3: us should pay a higher price and demand a lower

450
00:23:39,359 --> 00:23:42,040
Speaker 3: quantum of return because they're taking less risk, and I

451
00:23:42,080 --> 00:23:44,600
Speaker 3: should get rewarded for taking the early risk. So I

452
00:23:44,680 --> 00:23:47,359
Speaker 3: sort of think about it as destroying risk to create value.

453
00:23:47,359 --> 00:23:49,399
Speaker 3: Why do I say that, Because if we take an

454
00:23:49,440 --> 00:23:51,760
Speaker 3: early stage risk in a company to prove that the

455
00:23:51,840 --> 00:23:55,159
Speaker 3: tech works, I want those corporate vcs coming in. I

456
00:23:55,200 --> 00:23:57,120
Speaker 3: want them coming in. I want them paying a higher

457
00:23:57,200 --> 00:24:00,360
Speaker 3: price than we did, providing a lower cost of equity

458
00:24:00,400 --> 00:24:03,760
Speaker 3: than we did, and helping to both validate and create

459
00:24:03,800 --> 00:24:07,200
Speaker 3: some competition. So I'll give you an example. Runway mL

460
00:24:07,800 --> 00:24:10,600
Speaker 3: a bunch of interesting scientists. One of them was an

461
00:24:10,600 --> 00:24:13,159
Speaker 3: intern back in the day at Hugging Face became a

462
00:24:13,200 --> 00:24:16,720
Speaker 3: co founder of this company, Runway. Runway basically said, we

463
00:24:16,840 --> 00:24:19,679
Speaker 3: can take the cutting edge models that we're developing. They

464
00:24:19,720 --> 00:24:23,840
Speaker 3: actually were the developer of stable diffusion, and we're gonna

465
00:24:23,840 --> 00:24:26,879
Speaker 3: make videos. We're gonna start with two second videos. You

466
00:24:26,960 --> 00:24:29,760
Speaker 3: talk to the CEO there, Chris, he will say, within

467
00:24:29,800 --> 00:24:32,560
Speaker 3: the next two years, you have a full feature movie

468
00:24:33,000 --> 00:24:36,240
Speaker 3: that is entirely generated by people sitting at a computer

469
00:24:36,359 --> 00:24:41,359
Speaker 3: and just prompting angles, lighting, actors, expressions, the I mean,

470
00:24:41,560 --> 00:24:43,280
Speaker 3: it's like a little bit hard to fathom. It's like

471
00:24:43,320 --> 00:24:46,320
Speaker 3: looking at YouTube when it was two hundred and forty

472
00:24:46,359 --> 00:24:50,000
Speaker 3: pixels versus like eight k today. But it's gonna happen,

473
00:24:50,320 --> 00:24:51,920
Speaker 3: and it's interesting.

474
00:24:52,000 --> 00:24:57,000
Speaker 1: Totally full featured Hollywood film. Everything perfect except the hands.

475
00:24:58,920 --> 00:25:01,040
Speaker 3: Exactly, although I think they'll get the hands right and

476
00:25:01,240 --> 00:25:03,800
Speaker 3: there'll even be some you know, unique special effects. But

477
00:25:03,840 --> 00:25:07,120
Speaker 3: the sound, the lighting, the angles, everything. I think we're

478
00:25:07,160 --> 00:25:10,040
Speaker 3: two years from something that actually you think, oh my god,

479
00:25:10,080 --> 00:25:12,480
Speaker 3: that was made by AI and it'll probably be a

480
00:25:12,960 --> 00:25:15,920
Speaker 3: shorter film, but it's it's coming. Okay. Why do I

481
00:25:15,960 --> 00:25:18,560
Speaker 3: say that. You just had one hundred and forty million

482
00:25:18,600 --> 00:25:22,119
Speaker 3: dollar financing announced a week or two ago, Google in

483
00:25:22,280 --> 00:25:26,920
Speaker 3: Video and Salesforce, and those are three great companies one

484
00:25:27,000 --> 00:25:29,800
Speaker 3: is on the data side, one is on the sort

485
00:25:29,800 --> 00:25:31,879
Speaker 3: of strategic side, one is on the hardware side that

486
00:25:31,920 --> 00:25:34,760
Speaker 3: wants them to use their compute. All of those guys

487
00:25:34,760 --> 00:25:37,440
Speaker 3: are now linked with this company. And so Google's competitors

488
00:25:37,480 --> 00:25:39,680
Speaker 3: are looking at this, and salesforce doesn't want to be

489
00:25:39,760 --> 00:25:42,520
Speaker 3: left behind, and and Vidia is looking, and AMD is looking.

490
00:25:42,520 --> 00:25:45,240
Speaker 3: And so the more corporate strategic folks that you get in,

491
00:25:45,680 --> 00:25:48,600
Speaker 3: the more competitive juices start to flow, and it increases

492
00:25:48,640 --> 00:25:51,040
Speaker 3: the chance for the founders and for us that not

493
00:25:51,080 --> 00:25:53,520
Speaker 3: only do you get good strategic partners, but you set

494
00:25:53,600 --> 00:25:57,280
Speaker 3: up competitive dynamic for future exits. And so that's typically

495
00:25:57,280 --> 00:25:59,400
Speaker 3: you know, great companies get bought, and they get bought

496
00:25:59,400 --> 00:26:03,119
Speaker 3: because there's petitive fervor from a corp dev person at

497
00:26:03,160 --> 00:26:05,080
Speaker 3: one of the big companies that says, we can't let

498
00:26:05,160 --> 00:26:07,960
Speaker 3: our competitors get this. Okay. So that was the second thing,

499
00:26:08,000 --> 00:26:11,679
Speaker 3: which is against this regulatory backdrop. You know, who's going

500
00:26:11,720 --> 00:26:14,800
Speaker 3: to allow these big companies to actually buy these small companies.

501
00:26:15,200 --> 00:26:18,479
Speaker 3: And I'm hopeful, Okay, this is wishful thinking. This is

502
00:26:19,040 --> 00:26:23,040
Speaker 3: more prescriptive than observant. I think that the regime today

503
00:26:23,240 --> 00:26:26,320
Speaker 3: is very focused post twenty sixteen in the election and

504
00:26:26,320 --> 00:26:29,600
Speaker 3: the chaos and social media and all the abuses, particularly

505
00:26:29,680 --> 00:26:32,560
Speaker 3: that you saw at Facebook with users and fake information

506
00:26:32,760 --> 00:26:36,359
Speaker 3: and misinformation. I think that we are turning our turrets

507
00:26:36,400 --> 00:26:40,320
Speaker 3: of attention from Congress on the wrong targets. I think

508
00:26:40,359 --> 00:26:44,040
Speaker 3: that focusing on the domestic industry and trying to slow

509
00:26:44,080 --> 00:26:48,040
Speaker 3: it down and prevent acquisition and prevent failures and prevent

510
00:26:48,080 --> 00:26:51,159
Speaker 3: these companies from buying and competing is exactly what some

511
00:26:51,200 --> 00:26:55,160
Speaker 3: of our peer adversaries overseas would love. China and particularly

512
00:26:55,200 --> 00:26:58,399
Speaker 3: the CCP, would love nothing more than for AI in

513
00:26:58,440 --> 00:27:01,040
Speaker 3: the US to slow down and for all of these

514
00:27:01,119 --> 00:27:04,000
Speaker 3: iterations and experiments to have problems and further to be

515
00:27:04,040 --> 00:27:06,520
Speaker 3: a disincentive for vcs to want to fund these things

516
00:27:06,560 --> 00:27:09,000
Speaker 3: because they'll never get out. And I actually think that

517
00:27:09,040 --> 00:27:11,520
Speaker 3: you'll see some sea change coming in the next few

518
00:27:11,640 --> 00:27:15,280
Speaker 3: quarters year two where people say, Okay, wait a second.

519
00:27:15,720 --> 00:27:17,800
Speaker 3: You know, it isn't that we've met the enemy and

520
00:27:17,840 --> 00:27:21,280
Speaker 3: he is us. We actually have to have domestic competitiveness,

521
00:27:21,280 --> 00:27:23,480
Speaker 3: and one of the great assets that the country has

522
00:27:23,840 --> 00:27:26,560
Speaker 3: is competitive great technology companies, and we need to let

523
00:27:26,600 --> 00:27:30,719
Speaker 3: them thrive so that we can compete, particularly with China's CCP.

524
00:27:31,280 --> 00:27:33,760
Speaker 2: Just going back to what you said about a fully

525
00:27:33,920 --> 00:27:37,840
Speaker 2: AI generated movie. When I hear something like that, it

526
00:27:37,960 --> 00:27:42,560
Speaker 2: sounds incredibly exciting. It also sounds very sci fi and

527
00:27:42,640 --> 00:27:46,080
Speaker 2: difficult to wrap my head around in various ways. But

528
00:27:46,320 --> 00:27:49,840
Speaker 2: it kind of leads into a very basic question, which is,

529
00:27:50,400 --> 00:27:54,359
Speaker 2: what is it like to invest in a an AI

530
00:27:54,560 --> 00:27:58,879
Speaker 2: right now? So how are you actually doing your due diligence?

531
00:27:58,960 --> 00:28:02,040
Speaker 2: You know, if someone comes to you with an opportunity

532
00:28:02,080 --> 00:28:05,520
Speaker 2: for investing in a new technology, is it like all

533
00:28:05,560 --> 00:28:08,359
Speaker 2: of us sat here in the office playing around with

534
00:28:08,680 --> 00:28:11,840
Speaker 2: chat GPT? Is that basically the thrust of due diligence

535
00:28:11,880 --> 00:28:16,560
Speaker 2: on this technology or something else? And then, secondly, how

536
00:28:16,680 --> 00:28:21,160
Speaker 2: competitive is it right now from a venture capital perspective

537
00:28:21,320 --> 00:28:24,879
Speaker 2: to get in on some of these investments, because I imagine,

538
00:28:25,320 --> 00:28:28,080
Speaker 2: given the level of excitement, there is a lot of

539
00:28:28,200 --> 00:28:29,760
Speaker 2: money crowding into the space.

540
00:28:31,800 --> 00:28:35,480
Speaker 3: So the latter all answer first, which is it's very competitive.

541
00:28:35,560 --> 00:28:39,200
Speaker 3: I mean, anybody that can write a check is a competitor. Now,

542
00:28:39,240 --> 00:28:42,239
Speaker 3: if you are a founder, you know, just like if

543
00:28:42,280 --> 00:28:44,640
Speaker 3: you are a star high school athlete or a star

544
00:28:44,720 --> 00:28:47,600
Speaker 3: high school scholar, you want to go to the places

545
00:28:48,000 --> 00:28:52,360
Speaker 3: that reflect the quality of your craft, and so you

546
00:28:52,440 --> 00:28:54,560
Speaker 3: might want to go to Yale or Princeton or Stanford,

547
00:28:54,640 --> 00:28:56,440
Speaker 3: or you might want to go to Vanderbilt or Duke

548
00:28:56,520 --> 00:28:58,959
Speaker 3: and you know in play or Michigan and playball. And

549
00:28:59,000 --> 00:29:01,960
Speaker 3: so I think it's the same thing where great founders

550
00:29:02,000 --> 00:29:05,440
Speaker 3: want to work with great firms. And lux and Sequoia

551
00:29:05,480 --> 00:29:08,040
Speaker 3: and Andresen and a handful of others, you know, have

552
00:29:08,280 --> 00:29:12,320
Speaker 3: brands that confer to a founder that we are highly selective,

553
00:29:12,360 --> 00:29:14,080
Speaker 3: that we have a great network, that we can be

554
00:29:14,160 --> 00:29:16,880
Speaker 3: value add But anybody can fund any of these companies.

555
00:29:16,920 --> 00:29:19,080
Speaker 3: There's always somebody that's got a roommate who's got a

556
00:29:19,120 --> 00:29:21,160
Speaker 3: mother or father that gave them some money and they

557
00:29:21,240 --> 00:29:23,800
Speaker 3: became early investors in this company and they made a ton.

558
00:29:23,600 --> 00:29:26,520
Speaker 3: And so our view is that we are competing on

559
00:29:26,520 --> 00:29:29,960
Speaker 3: one hand with everybody. Now. The second thing is that

560
00:29:30,520 --> 00:29:33,600
Speaker 3: I always say there's this five year psychological bias, which

561
00:29:33,640 --> 00:29:37,040
Speaker 3: is that you want to be invested today where you

562
00:29:37,040 --> 00:29:39,760
Speaker 3: know we were five years ago, and so I'm trying

563
00:29:39,760 --> 00:29:41,840
Speaker 3: to figure out what's the next thing three four, five

564
00:29:41,920 --> 00:29:44,400
Speaker 3: years ahead that people don't yet appreciate. So you know,

565
00:29:44,440 --> 00:29:47,680
Speaker 3: I mentioned biology. Now you know all the listeners can

566
00:29:47,720 --> 00:29:49,920
Speaker 3: go out and try to find the next models in biology.

567
00:29:49,920 --> 00:29:52,920
Speaker 3: But it's a hard or more complex thing, and I'm

568
00:29:52,960 --> 00:29:56,120
Speaker 3: confident that there there's a fewer number of investors that

569
00:29:56,200 --> 00:29:58,720
Speaker 3: actually understand or have the networks of the connection. So

570
00:29:58,760 --> 00:30:02,600
Speaker 3: we have some slight competitive advantage there. When we're evaluating

571
00:30:02,680 --> 00:30:05,440
Speaker 3: these things, you're looking at the credibility of the founders.

572
00:30:05,480 --> 00:30:08,840
Speaker 3: Many of them happened to be academic published papers, so

573
00:30:08,880 --> 00:30:12,760
Speaker 3: you can see, for example, the people behind Runway who

574
00:30:12,880 --> 00:30:16,080
Speaker 3: published the papers that led to stable diffusion, or the

575
00:30:16,200 --> 00:30:18,520
Speaker 3: team that came out of Google that published the paper

576
00:30:18,560 --> 00:30:21,480
Speaker 3: on transformers. Not the robots of course, like optimist prime,

577
00:30:21,520 --> 00:30:25,840
Speaker 3: but the underlying algorithms that led to chat GPT. Every

578
00:30:25,840 --> 00:30:28,280
Speaker 3: one of the people on those papers have basically gone

579
00:30:28,320 --> 00:30:31,320
Speaker 3: on to start companies and raise money. People that were

580
00:30:31,360 --> 00:30:34,040
Speaker 3: at open Ai have the pedigree, they learned what works,

581
00:30:34,040 --> 00:30:36,120
Speaker 3: they went and started anthropic and so there's this sort

582
00:30:36,120 --> 00:30:38,320
Speaker 3: of like just like if you go back in finance

583
00:30:38,360 --> 00:30:41,400
Speaker 3: to like the Drexel days where they spawned you know,

584
00:30:41,440 --> 00:30:44,840
Speaker 3: Apollo and Carlisle and Jeffries and all these it's the

585
00:30:44,880 --> 00:30:47,560
Speaker 3: same sort of thing. There's a diaspora that's coming out

586
00:30:47,560 --> 00:30:50,880
Speaker 3: of a small group of people and you can reference

587
00:30:50,920 --> 00:30:54,000
Speaker 3: the credibility and then yes, you sit with them and

588
00:30:54,040 --> 00:30:56,040
Speaker 3: you look at what their demos are. And we like

589
00:30:56,120 --> 00:31:00,200
Speaker 3: to say that we believe before others understand. And so

590
00:31:00,240 --> 00:31:01,959
Speaker 3: when we had that Runway team in or we had

591
00:31:01,960 --> 00:31:04,200
Speaker 3: the Hugging Face team in, you know, it was very

592
00:31:04,280 --> 00:31:05,920
Speaker 3: raw and very crude, and you have to sort of

593
00:31:06,000 --> 00:31:08,920
Speaker 3: squint and see the future that they're seeing. And then

594
00:31:09,040 --> 00:31:11,480
Speaker 3: we don't fully fund companies. You give a little bit

595
00:31:11,520 --> 00:31:13,920
Speaker 3: of money and you say how much money will accomplish

596
00:31:14,000 --> 00:31:16,680
Speaker 3: what in what period of time? And who will care?

597
00:31:16,800 --> 00:31:18,240
Speaker 3: Are we going to get paid for the risk that

598
00:31:18,280 --> 00:31:21,000
Speaker 3: we're taking and funding you? But I would say right now,

599
00:31:21,200 --> 00:31:25,120
Speaker 3: if you're funding anything that is application focused, anything that

600
00:31:25,200 --> 00:31:29,000
Speaker 3: is to your earlier point in the wrappers around the

601
00:31:29,080 --> 00:31:33,440
Speaker 3: user interface, most of those things are just features. They're

602
00:31:33,480 --> 00:31:35,880
Speaker 3: not companies. Most of those are going to be competed

603
00:31:35,880 --> 00:31:39,360
Speaker 3: away by one hundred other examples. A great example of

604
00:31:39,360 --> 00:31:41,440
Speaker 3: that is and I can't even remember the name of

605
00:31:41,480 --> 00:31:43,600
Speaker 3: it now, but there was something that went out and

606
00:31:43,640 --> 00:31:45,520
Speaker 3: it was an app and you could pay twenty bucks

607
00:31:45,560 --> 00:31:47,680
Speaker 3: and it would give you one hundred versions of yourself,

608
00:31:47,760 --> 00:31:49,880
Speaker 3: you know, as a comic book hero and a cowboy

609
00:31:50,000 --> 00:31:51,600
Speaker 3: and a black and white and long.

610
00:31:52,880 --> 00:31:55,440
Speaker 1: I think we signed up for the one month version

611
00:31:55,480 --> 00:31:57,880
Speaker 1: of that. Yeah, I don't know. I probably forgot to.

612
00:31:57,920 --> 00:32:01,200
Speaker 3: Get and spiked and then it's done. And you're going

613
00:32:01,280 --> 00:32:03,560
Speaker 3: to have tons of those things where it spikes and

614
00:32:03,600 --> 00:32:06,080
Speaker 3: it's done and it ends up being a feature integrated

615
00:32:06,120 --> 00:32:09,080
Speaker 3: into Going back to the earliest point you made, many

616
00:32:09,120 --> 00:32:11,960
Speaker 3: of the big tech companies, the Adobes and the Microsoft's,

617
00:32:11,960 --> 00:32:14,200
Speaker 3: it's going to be in all their suites. I have

618
00:32:14,280 --> 00:32:17,480
Speaker 3: one contrarian take here, which is a bit odd as

619
00:32:17,880 --> 00:32:21,440
Speaker 3: an investors part of a partnership who's funding the deep

620
00:32:21,480 --> 00:32:24,959
Speaker 3: tech roots of the semiconductors and the infrastructure and the

621
00:32:25,000 --> 00:32:28,760
Speaker 3: networks and the models and the algorithms. And despite doing

622
00:32:28,760 --> 00:32:31,040
Speaker 3: all of that, I actually have a view that what

623
00:32:31,320 --> 00:32:34,320
Speaker 3: we are doing right now humans talking, even though it's

624
00:32:34,360 --> 00:32:37,520
Speaker 3: through digital communications in an analog way, that that is

625
00:32:37,560 --> 00:32:40,400
Speaker 3: actually going to become the scarce thing. You are going

626
00:32:40,520 --> 00:32:46,160
Speaker 3: to be flooded, I mean utterly inundated by emails, texts,

627
00:32:46,600 --> 00:32:49,800
Speaker 3: tweets that are not written by humans. And that's not

628
00:32:49,920 --> 00:32:53,000
Speaker 3: like two years away, that's like two weeks. The increasing

629
00:32:53,080 --> 00:32:55,960
Speaker 3: percentage of the communications that you receive even by the way,

630
00:32:55,960 --> 00:32:58,480
Speaker 3: from people that you know and love and trust are

631
00:32:58,520 --> 00:33:00,560
Speaker 3: not going to be written by them. They're not going

632
00:33:00,600 --> 00:33:02,880
Speaker 3: to be spoken by them. You know, voice is going

633
00:33:02,920 --> 00:33:04,680
Speaker 3: to be the next domain where it's going to be

634
00:33:04,800 --> 00:33:07,080
Speaker 3: very hard. You're going to be getting a voicemail and

635
00:33:07,120 --> 00:33:10,520
Speaker 3: the voicemail was not actually spoken by your spouse or

636
00:33:10,560 --> 00:33:14,160
Speaker 3: your cousin or kid. I can already and I've already

637
00:33:14,160 --> 00:33:16,200
Speaker 3: trained a model on my child and I was able

638
00:33:16,240 --> 00:33:18,280
Speaker 3: to trick my wife on It was you know, funny

639
00:33:18,280 --> 00:33:21,560
Speaker 3: and scary. The point of this is if that happens,

640
00:33:21,560 --> 00:33:24,760
Speaker 3: and as that happens, you will start to grow increasingly

641
00:33:24,800 --> 00:33:28,200
Speaker 3: distrustful of many of the communications you get. It'll be

642
00:33:28,320 --> 00:33:30,400
Speaker 3: you know, this form of chat fishing is what I

643
00:33:30,440 --> 00:33:33,479
Speaker 3: call it. And you'll start to just pine for in

644
00:33:33,560 --> 00:33:36,320
Speaker 3: person communications and there will be private clubs that form

645
00:33:36,600 --> 00:33:39,040
Speaker 3: where people come and no devices and you just know

646
00:33:39,120 --> 00:33:42,240
Speaker 3: that you're talking to a human because increasingly that will

647
00:33:42,280 --> 00:33:45,800
Speaker 3: be scarce. So the great irony here is the flood

648
00:33:45,840 --> 00:33:49,360
Speaker 3: of money and talent and productivity in AI and deep

649
00:33:49,400 --> 00:33:53,320
Speaker 3: technology is probably going to bring us closer to our

650
00:33:53,400 --> 00:33:54,280
Speaker 3: innate humanity.

651
00:33:54,480 --> 00:33:56,400
Speaker 1: It sounds like the answer is tracy, and I need

652
00:33:56,480 --> 00:33:59,760
Speaker 1: to do a lot more online pub quizzes and other

653
00:34:00,240 --> 00:34:04,080
Speaker 1: live event I'm that sounds good to me. Can I

654
00:34:04,520 --> 00:34:07,680
Speaker 1: ask a question though about if it's like if all

655
00:34:07,720 --> 00:34:10,239
Speaker 1: you have to do is sort of be a graduate

656
00:34:10,280 --> 00:34:12,799
Speaker 1: from one of the right universities and have your name

657
00:34:12,920 --> 00:34:17,200
Speaker 1: on some paper that's on like the archive website, what

658
00:34:17,239 --> 00:34:21,360
Speaker 1: does that mean for recruitment from the companies that you

659
00:34:21,560 --> 00:34:24,279
Speaker 1: are investing in? And when they want to go out

660
00:34:24,320 --> 00:34:27,279
Speaker 1: and hire someone, how is the challenges? Like, no, that

661
00:34:27,320 --> 00:34:29,640
Speaker 1: person who they probably want to hire can also raise

662
00:34:29,680 --> 00:34:32,600
Speaker 1: one hundred million dollars And how difficult is it to

663
00:34:32,680 --> 00:34:35,000
Speaker 1: recruit if there's just like so much money going to

664
00:34:35,480 --> 00:34:37,240
Speaker 1: you know, a potentially smart founder.

665
00:34:37,560 --> 00:34:40,440
Speaker 3: You know, this has been the plague if you're an

666
00:34:40,480 --> 00:34:44,600
Speaker 3: investor for arguably the past decade in tech broadly, and

667
00:34:45,160 --> 00:34:47,920
Speaker 3: it's a weird phenomenon. And what I mean by that

668
00:34:47,960 --> 00:34:50,479
Speaker 3: plague is everybody thought that they could raise money. Amy

669
00:34:50,520 --> 00:34:52,759
Speaker 3: could They all had a friend or a colleague or

670
00:34:52,760 --> 00:34:56,279
Speaker 3: a former you know, associated or roommate who raised money

671
00:34:56,280 --> 00:34:59,120
Speaker 3: and they literally had this reaction he or she just

672
00:34:59,200 --> 00:35:01,680
Speaker 3: raised that money that valuation? What an idiot? I can

673
00:35:01,719 --> 00:35:04,120
Speaker 3: go and do that? And so that set it comparable

674
00:35:04,120 --> 00:35:05,279
Speaker 3: for them to say I'm going to go do it,

675
00:35:05,320 --> 00:35:08,960
Speaker 3: and you get this sort of collective craze and so

676
00:35:09,480 --> 00:35:13,600
Speaker 3: talent gets disfuse and disperse. It's increasing the cost of

677
00:35:13,920 --> 00:35:18,000
Speaker 3: hiring employees. Valuations are going up, money is being misallocated,

678
00:35:18,120 --> 00:35:20,479
Speaker 3: like the whole thing. Okay, all of that crashed about

679
00:35:20,480 --> 00:35:22,560
Speaker 3: a year and a half ago, once rates started rising

680
00:35:22,600 --> 00:35:25,319
Speaker 3: and the back boom ended and all that except for

681
00:35:25,400 --> 00:35:28,520
Speaker 3: this one domain of AI. And so if you actually

682
00:35:28,520 --> 00:35:30,960
Speaker 3: look at the hiring data, and you know, you have

683
00:35:31,000 --> 00:35:33,400
Speaker 3: to parse this to see whether it's you know, signals

684
00:35:33,400 --> 00:35:36,200
Speaker 3: that they're posting jobs or they're actually hiring. But all

685
00:35:36,239 --> 00:35:38,040
Speaker 3: of the layoffs that you saw, you know, the tens

686
00:35:38,040 --> 00:35:40,879
Speaker 3: of thousands at Meta and Google and Microsoft, et cetera,

687
00:35:41,120 --> 00:35:43,719
Speaker 3: you're starting to see this spike back up in some

688
00:35:43,800 --> 00:35:46,160
Speaker 3: of the data. And what are they doing they're hiring,

689
00:35:46,360 --> 00:35:48,840
Speaker 3: whether it's low level jobs for data entry and processing

690
00:35:48,880 --> 00:35:52,160
Speaker 3: and cleaning, or whether it's cutting edge algorithmic design for AI.

691
00:35:52,800 --> 00:35:55,920
Speaker 3: There was an existential panic at Google and it's been

692
00:35:55,960 --> 00:35:59,000
Speaker 3: reported when open ai came out, you know, the all

693
00:35:59,080 --> 00:36:01,440
Speaker 3: hands deck meeting that people had of my god, we

694
00:36:01,520 --> 00:36:03,360
Speaker 3: got to throw a ton of talent and money at

695
00:36:03,400 --> 00:36:06,120
Speaker 3: this so that we don't get left behind. So right

696
00:36:06,160 --> 00:36:09,080
Speaker 3: now it's a bit of an utter craze. You got

697
00:36:09,080 --> 00:36:12,320
Speaker 3: to be really careful. Most of the incumbents are the winners.

698
00:36:12,440 --> 00:36:15,480
Speaker 3: And there's going to be some small companies that end

699
00:36:15,560 --> 00:36:18,440
Speaker 3: up with these really interesting novel approaches that are not

700
00:36:18,600 --> 00:36:21,520
Speaker 3: raising a ton of money, almost out of necessity. They're

701
00:36:21,560 --> 00:36:25,399
Speaker 3: doing these cheap, very focused models and arguably, and here's

702
00:36:25,440 --> 00:36:29,759
Speaker 3: another interesting thing, training on distributed computing instead of having

703
00:36:29,800 --> 00:36:33,279
Speaker 3: these big centralized clusters of compute. We for example, back

704
00:36:33,320 --> 00:36:37,040
Speaker 3: to a company called together Compute, and his basic hypothesis was,

705
00:36:37,480 --> 00:36:43,440
Speaker 3: can I train very sophisticated models on all of the

706
00:36:43,520 --> 00:36:47,319
Speaker 3: excess compute from the crypto craze or from idle computers?

707
00:36:47,880 --> 00:36:49,960
Speaker 3: And can I do it for a fraction of the

708
00:36:50,000 --> 00:36:52,080
Speaker 3: cost of what open ai did. And the answer was yes.

709
00:36:52,440 --> 00:36:53,799
Speaker 3: And so you're going to see lots of these small

710
00:36:53,840 --> 00:36:57,120
Speaker 3: companies that come. And I always say, like, whenever the

711
00:36:57,200 --> 00:37:00,839
Speaker 3: DOJ comes in and starts looking at monopoly concerns from

712
00:37:00,880 --> 00:37:04,120
Speaker 3: the big companies, it's never the DOJ that disrupts the

713
00:37:04,120 --> 00:37:06,520
Speaker 3: big company that people are concerned with. It's always some

714
00:37:06,560 --> 00:37:10,040
Speaker 3: small competitor. It happened with Microsoft in the late nineties.

715
00:37:10,080 --> 00:37:12,720
Speaker 3: Google came along. It wasn't the DOJ, it was Google.

716
00:37:12,880 --> 00:37:15,040
Speaker 3: You know, it happened with Facebook and Google. It's happened

717
00:37:15,080 --> 00:37:17,480
Speaker 3: with open Ai and now Facebook, and it'll happen again,

718
00:37:17,840 --> 00:37:20,239
Speaker 3: and it'll be four guys or girls in a room

719
00:37:20,400 --> 00:37:25,480
Speaker 3: in Croatia or Singapore or Mexico City or Silicon Valley

720
00:37:25,840 --> 00:37:27,720
Speaker 3: that come up with the crazy new thing that disrupts

721
00:37:27,719 --> 00:37:28,320
Speaker 3: the big incumbent.

722
00:37:45,719 --> 00:37:49,279
Speaker 2: What impact do you see of AI on how the

723
00:37:49,320 --> 00:37:54,560
Speaker 2: sort of tech industries slash VC is organized or operating

724
00:37:54,760 --> 00:37:57,680
Speaker 2: at the moment, Like, do you see people start to

725
00:37:58,640 --> 00:38:01,120
Speaker 2: respond to the idea that, well, maybe a lot of

726
00:38:01,160 --> 00:38:04,320
Speaker 2: coding is going to be done by AI in the future.

727
00:38:04,360 --> 00:38:08,640
Speaker 2: Are people sort of like reorganizing themselves or reorienting themselves

728
00:38:08,760 --> 00:38:10,600
Speaker 2: ahead of some of this technology.

729
00:38:11,000 --> 00:38:13,680
Speaker 3: Definitely. When you look at the copilot, which is both

730
00:38:13,680 --> 00:38:17,320
Speaker 3: from people like GitHub and open ai and others, it's basically,

731
00:38:17,400 --> 00:38:20,640
Speaker 3: how can we either help you code or how can

732
00:38:20,680 --> 00:38:24,120
Speaker 3: we completely if you look at code Interpreter completely write

733
00:38:24,120 --> 00:38:28,120
Speaker 3: the code for you. You just describe in general lay language

734
00:38:28,280 --> 00:38:30,600
Speaker 3: what you want to do, and it will in Python

735
00:38:31,040 --> 00:38:35,040
Speaker 3: create the code. So pretty much every company will have

736
00:38:36,120 --> 00:38:40,399
Speaker 3: a form of computer programmers in the software that they use,

737
00:38:40,440 --> 00:38:43,760
Speaker 3: whether it's open source or proprietary, that is constantly developing

738
00:38:43,840 --> 00:38:47,359
Speaker 3: and iterating their own applications, and it'll touch everything from

739
00:38:47,800 --> 00:38:52,600
Speaker 3: customer service to radiology and X ray image analysis to

740
00:38:52,920 --> 00:38:57,600
Speaker 3: Bloomberg queries, and people will be able to basically have superpowers.

741
00:38:57,680 --> 00:39:00,640
Speaker 3: What it means is you'll have the elite codevods that

742
00:39:00,680 --> 00:39:02,359
Speaker 3: are sort of always on the edge trying to figure

743
00:39:02,360 --> 00:39:04,560
Speaker 3: out the next thing, and then you'll have the average

744
00:39:04,560 --> 00:39:08,720
Speaker 3: coders that are basically really leveled up and almost indistinguishable

745
00:39:08,719 --> 00:39:12,440
Speaker 3: from the prior elite coders. So I do think that

746
00:39:12,480 --> 00:39:16,080
Speaker 3: what once was really scarce and really valuable, which was

747
00:39:16,760 --> 00:39:20,879
Speaker 3: top notch what people call ten X coders, is now

748
00:39:20,920 --> 00:39:25,080
Speaker 3: going to become increasingly commodity and people will start looking.

749
00:39:25,120 --> 00:39:27,600
Speaker 3: For example, Now the real value is not can you

750
00:39:27,680 --> 00:39:30,360
Speaker 3: code for the current moment. It's like, are you an

751
00:39:30,400 --> 00:39:34,640
Speaker 3: amazing prompt engineer? Like I can't draw, I can't code,

752
00:39:34,680 --> 00:39:38,520
Speaker 3: I can't write, you know, twelve line stanzas, but I

753
00:39:38,600 --> 00:39:42,319
Speaker 3: can prompt pretty well. And so it shifts the capability

754
00:39:42,640 --> 00:39:46,560
Speaker 3: to the creativity of somebody that really wants to describe

755
00:39:46,680 --> 00:39:49,960
Speaker 3: and control the machine again almost more like casting a

756
00:39:50,000 --> 00:39:53,719
Speaker 3: spell than the person that actually has the discrete technical capability.

757
00:39:54,200 --> 00:39:56,600
Speaker 1: Can I ask you know you mentioned that roughly, I

758
00:39:56,600 --> 00:39:58,560
Speaker 1: don't know, a year and a half or so ago,

759
00:39:58,680 --> 00:40:01,600
Speaker 1: or maybe two years ago, this sort of the model

760
00:40:01,640 --> 00:40:05,040
Speaker 1: that had been working like this incredible trade, this incredible

761
00:40:05,120 --> 00:40:08,360
Speaker 1: line go up decade or whatever for tech and VC

762
00:40:08,600 --> 00:40:10,960
Speaker 1: like did sort of crumble to some extent, and we

763
00:40:10,960 --> 00:40:12,799
Speaker 1: saw the big plunge of the NASDAK And I'm sure

764
00:40:12,800 --> 00:40:16,600
Speaker 1: there were like tons of funds raised in twenty twenty

765
00:40:16,640 --> 00:40:19,120
Speaker 1: one and twenty twenty that are like deep underwater and

766
00:40:19,160 --> 00:40:21,560
Speaker 1: all that stuff. And everyone knows that when you're at

767
00:40:21,560 --> 00:40:25,560
Speaker 1: the table now looking at companies, do you still is

768
00:40:25,600 --> 00:40:30,960
Speaker 1: there still pain and paranoia and fear from that? Have

769
00:40:31,040 --> 00:40:33,600
Speaker 1: there been like or has the pain of that been

770
00:40:33,640 --> 00:40:35,680
Speaker 1: internalized in the way or is it you know what?

771
00:40:36,080 --> 00:40:38,840
Speaker 1: We're just back in at games, back on phone mocycle,

772
00:40:38,920 --> 00:40:41,200
Speaker 1: back on let's go like how much are that? How

773
00:40:41,280 --> 00:40:44,239
Speaker 1: much are their scars from the sort of crash of

774
00:40:44,280 --> 00:40:45,080
Speaker 1: twenty twenty one.

775
00:40:46,239 --> 00:40:47,840
Speaker 3: I think people that put a lot of money to

776
00:40:47,880 --> 00:40:51,000
Speaker 3: work in twenty twenty, twenty twenty one feel a lot

777
00:40:51,000 --> 00:40:53,919
Speaker 3: of pain they invested at record high multiples. They made

778
00:40:53,960 --> 00:40:56,680
Speaker 3: the presumption, which was a fair presumption to make that

779
00:40:57,440 --> 00:40:59,440
Speaker 3: if you fund something there's going to be later stage

780
00:40:59,440 --> 00:41:02,440
Speaker 3: capital or a robust public market to follow you on.

781
00:41:02,719 --> 00:41:04,480
Speaker 3: All of that is gone, you know. So so I

782
00:41:04,520 --> 00:41:07,560
Speaker 3: think we went from what I called what everybody called

783
00:41:07,600 --> 00:41:10,240
Speaker 3: fomo if you're missing out to what I call sobs,

784
00:41:10,400 --> 00:41:12,839
Speaker 3: which was the shame of being suckered. And that wasn't

785
00:41:13,160 --> 00:41:15,000
Speaker 3: just that wasn't.

786
00:41:14,680 --> 00:41:17,239
Speaker 1: Just can I just say, you know what, Can I

787
00:41:17,280 --> 00:41:19,759
Speaker 1: just say there's no shame? And you know people buy that.

788
00:41:19,840 --> 00:41:21,319
Speaker 1: It's hard to know at the top is but I

789
00:41:21,360 --> 00:41:23,640
Speaker 1: think about like the person who paid half a million

790
00:41:23,680 --> 00:41:26,239
Speaker 1: dollars or millions of dollars like for the board ape

791
00:41:26,320 --> 00:41:29,839
Speaker 1: nft is that's the ultimate sob. You can never live

792
00:41:29,880 --> 00:41:32,080
Speaker 1: that down. Okay, keep going. Sorry, I just thought about

793
00:41:32,080 --> 00:41:34,439
Speaker 1: that recently and that was in my head. Okay, keep going.

794
00:41:34,719 --> 00:41:37,120
Speaker 3: But you know, I got a friend, Zach bisin Ed

795
00:41:37,120 --> 00:41:38,520
Speaker 3: who wrote the book back in the day on.

796
00:41:38,840 --> 00:41:41,640
Speaker 1: We've had Yeah, that's amazing exactly.

797
00:41:41,560 --> 00:41:43,960
Speaker 3: So, but this is this is one of the things

798
00:41:43,960 --> 00:41:46,280
Speaker 3: that I love to say, which is not some crazy

799
00:41:46,280 --> 00:41:49,080
Speaker 3: personal insight. It's just an observable truth, which is that

800
00:41:49,360 --> 00:41:53,399
Speaker 3: technologies change, and businesses change, and rules change and policies change.

801
00:41:53,480 --> 00:41:56,560
Speaker 3: Human nature is a constant. Greed and fear is a constant.

802
00:41:56,600 --> 00:41:58,800
Speaker 3: That is what made Buffett and Monger brilliant. You know,

803
00:41:58,840 --> 00:42:01,200
Speaker 3: it's what Howard Mark's chronicles all the time. It's what

804
00:42:01,239 --> 00:42:04,279
Speaker 3: you guys cover the excesses of human emotion. And so

805
00:42:04,360 --> 00:42:06,880
Speaker 3: a lot of this is actually capturing. I love finding,

806
00:42:06,920 --> 00:42:11,320
Speaker 3: like where are people not paying attention? Where is attention scarce?

807
00:42:11,360 --> 00:42:14,680
Speaker 3: Because where attention is scarce, valuations are going to be low.

808
00:42:14,880 --> 00:42:16,400
Speaker 3: And we always say, oh, you know, just like everybody

809
00:42:16,400 --> 00:42:19,759
Speaker 3: says they're contrarian investors, we want people to agree with us,

810
00:42:20,120 --> 00:42:23,160
Speaker 3: just later. And that's the key. Okay, going back to

811
00:42:23,520 --> 00:42:26,520
Speaker 3: your question, you know we went from fomo fear of

812
00:42:26,560 --> 00:42:29,399
Speaker 3: missing out to sobs the shame of being suckered. And

813
00:42:29,719 --> 00:42:32,759
Speaker 3: there was an important reason for that, which was the

814
00:42:32,920 --> 00:42:37,840
Speaker 3: disappearance of two major players, at least symbolically, and that

815
00:42:37,960 --> 00:42:41,160
Speaker 3: was SoftBank and Tiger. And why was that important? Because

816
00:42:41,640 --> 00:42:44,200
Speaker 3: you know, you had a venture firm maybe a decade

817
00:42:44,239 --> 00:42:46,360
Speaker 3: ago that said the price you pay for a company

818
00:42:46,400 --> 00:42:48,839
Speaker 3: doesn't really matter because there's only ten companies that matter

819
00:42:48,880 --> 00:42:50,400
Speaker 3: amongst all the ones that are funded. And if you

820
00:42:50,400 --> 00:42:53,360
Speaker 3: would have funded LinkedIn or Facebook at five billion or

821
00:42:53,360 --> 00:42:55,799
Speaker 3: ten billion or twenty, it wouldn't have mattered, right, And

822
00:42:55,880 --> 00:42:58,520
Speaker 3: so that sort of set a precedent that I think

823
00:42:58,600 --> 00:43:01,360
Speaker 3: was a bit insidious and danger to say the price

824
00:43:01,440 --> 00:43:03,680
Speaker 3: doesn't matter, you just have to be in the right companies.

825
00:43:03,840 --> 00:43:06,560
Speaker 3: Of course, that's only obvious in hindsight. So you had

826
00:43:06,600 --> 00:43:12,040
Speaker 3: lots of people that lost valuation discipline. It skewed control

827
00:43:12,080 --> 00:43:15,759
Speaker 3: and leverage to the founders over from the investors, you know,

828
00:43:16,000 --> 00:43:20,080
Speaker 3: and you saw weak governance, and you saw fraud and

829
00:43:20,160 --> 00:43:22,240
Speaker 3: excess and all that kind of stuff, and it's starting

830
00:43:22,280 --> 00:43:25,640
Speaker 3: to wash out, you know, if not fully, but the

831
00:43:25,680 --> 00:43:30,320
Speaker 3: disappearance of those two players. Symbolically they were the top ticking,

832
00:43:30,440 --> 00:43:35,200
Speaker 3: marginal price setting investors. SoftBank was paying insane prices, and

833
00:43:35,200 --> 00:43:36,839
Speaker 3: of course, you know, they were all kind of shenanigans

834
00:43:36,880 --> 00:43:38,920
Speaker 3: of them marking up their own book and pricing up

835
00:43:38,960 --> 00:43:41,200
Speaker 3: again and all that kind of stuff, and Tiger sort

836
00:43:41,239 --> 00:43:44,280
Speaker 3: of took a passive indexation and approach, which was something

837
00:43:44,320 --> 00:43:46,040
Speaker 3: that was widespread in the public markets, but they did

838
00:43:46,120 --> 00:43:48,200
Speaker 3: in the private markets and say they said, we're just

839
00:43:48,200 --> 00:43:50,320
Speaker 3: going to be in all the companies, and the winners

840
00:43:50,320 --> 00:43:52,600
Speaker 3: will make up for the losers, and it'll work when

841
00:43:52,800 --> 00:43:56,000
Speaker 3: those kinds of players disappear. Now, all of a sudden,

842
00:43:56,040 --> 00:43:58,880
Speaker 3: you have a more rational scrutinizing market of people who

843
00:43:58,960 --> 00:44:01,799
Speaker 3: are afraid of paying ex prices, feel like they need

844
00:44:01,840 --> 00:44:04,759
Speaker 3: to get a better deal. You're seeing down rounds in companies.

845
00:44:04,800 --> 00:44:08,359
Speaker 3: You have a morale spin and decline where employees now

846
00:44:08,400 --> 00:44:11,600
Speaker 3: have underwater stock and need to be refreshed. And here's

847
00:44:11,680 --> 00:44:15,200
Speaker 3: where things get really interesting. We went from this domain

848
00:44:15,239 --> 00:44:18,080
Speaker 3: where I called it the megas and the minnos. The

849
00:44:18,120 --> 00:44:20,480
Speaker 3: megas were the giant funds that were, you know, ten

850
00:44:20,520 --> 00:44:22,440
Speaker 3: billion plus and they were writing these giant checks. And

851
00:44:22,480 --> 00:44:26,200
Speaker 3: the minnos were the thousands of small, sub hundred million

852
00:44:26,200 --> 00:44:28,240
Speaker 3: dollar funds that were just doing all the seed investing.

853
00:44:28,719 --> 00:44:31,160
Speaker 3: Both of those guys have been squeezed out, and so

854
00:44:31,560 --> 00:44:34,319
Speaker 3: now you have a smaller base of capital. You can

855
00:44:34,360 --> 00:44:37,200
Speaker 3: see it in the data. LPs have pulled back the

856
00:44:37,480 --> 00:44:39,720
Speaker 3: you know, the champagne has stopped flowing down the pyramid

857
00:44:39,800 --> 00:44:43,160
Speaker 3: of glasses. Gps are struggling to raise capital. You know,

858
00:44:43,200 --> 00:44:46,839
Speaker 3: we closed a billion two fund in ten weeks, which

859
00:44:46,880 --> 00:44:49,120
Speaker 3: for US was amazing, and it was signal of great

860
00:44:49,120 --> 00:44:52,560
Speaker 3: support of our LPs and great founders. A lot of

861
00:44:52,640 --> 00:44:55,640
Speaker 3: funds out there right now are downsizing. It's taking them

862
00:44:55,640 --> 00:44:57,520
Speaker 3: a lot longer to raise and all of that is

863
00:44:57,560 --> 00:45:01,919
Speaker 3: a rational reaction to a retraction. So I don't think

864
00:45:01,960 --> 00:45:04,960
Speaker 3: you see the same fomo. I think you see a

865
00:45:05,000 --> 00:45:08,080
Speaker 3: lot more fear. People don't want to pay higher prices.

866
00:45:08,760 --> 00:45:12,760
Speaker 3: The only area where there's an exception is inside of AI.

867
00:45:13,719 --> 00:45:17,560
Speaker 2: You know, I just have one more question, And you

868
00:45:17,640 --> 00:45:20,399
Speaker 2: sort of touched on this earlier where we were, well,

869
00:45:20,440 --> 00:45:23,520
Speaker 2: you were talking about parallels between now and the sort

870
00:45:23,520 --> 00:45:26,880
Speaker 2: of dot com era and the idea that well, you know,

871
00:45:27,000 --> 00:45:32,480
Speaker 2: maybe eventually some big winners will emerge from this new technology,

872
00:45:32,520 --> 00:45:35,359
Speaker 2: whether it's AI or the Internet as it was in

873
00:45:35,600 --> 00:45:38,880
Speaker 2: you know, the late nineteen nineties, early two thousands. What's

874
00:45:38,960 --> 00:45:44,359
Speaker 2: the case for investing in AI right now? Rather than

875
00:45:44,520 --> 00:45:48,239
Speaker 2: waiting a little bit to see where the dust settles,

876
00:45:48,400 --> 00:45:52,239
Speaker 2: maybe wait to see who those big winners are, or

877
00:45:52,360 --> 00:45:54,680
Speaker 2: maybe at the very least get a little bit more

878
00:45:54,719 --> 00:45:57,120
Speaker 2: clarity on how this whole thing is going to be

879
00:45:57,160 --> 00:45:58,440
Speaker 2: structured or organized.

880
00:45:59,800 --> 00:46:02,880
Speaker 3: Well, the argument for waiting is, by the time you know,

881
00:46:03,120 --> 00:46:06,040
Speaker 3: it's already fully factored into a price. The contract to

882
00:46:06,080 --> 00:46:08,880
Speaker 3: that is you pay a high price for cheery consensus,

883
00:46:08,920 --> 00:46:12,040
Speaker 3: as Buffett historically said. And so if everybody agrees that

884
00:46:12,080 --> 00:46:14,680
Speaker 3: in Nvidia is the winner, you know, that to me

885
00:46:14,920 --> 00:46:18,480
Speaker 3: gives me pause for concern. You know, Jensen is running high,

886
00:46:18,680 --> 00:46:22,560
Speaker 3: he's got the iconic leather black jacket, he's becoming the

887
00:46:22,600 --> 00:46:24,880
Speaker 3: sort of you know, next profit of tech. Those are

888
00:46:24,880 --> 00:46:27,280
Speaker 3: all signals that are like, okay, just like the classic

889
00:46:27,320 --> 00:46:29,839
Speaker 3: you know, sports illustrated curve, simple reversion to the mean,

890
00:46:29,960 --> 00:46:32,360
Speaker 3: Like what happens? Where's the vulnerability there? To me is

891
00:46:32,360 --> 00:46:36,200
Speaker 3: the question. And I gave you guys and listeners a clue,

892
00:46:36,280 --> 00:46:39,600
Speaker 3: which is that Kuda, their language system, is vulnerable to

893
00:46:39,640 --> 00:46:43,600
Speaker 3: these other ones of py Torch from open source, originated

894
00:46:43,640 --> 00:46:47,200
Speaker 3: from Meta and Triton from open Ai. And that means

895
00:46:47,200 --> 00:46:50,319
Speaker 3: that AMD could actually come from behind and start to

896
00:46:50,320 --> 00:46:52,400
Speaker 3: take share. It's something that people are skeptical about. So

897
00:46:53,239 --> 00:46:57,240
Speaker 3: I would say that if you're thinking about investing now,

898
00:46:57,600 --> 00:47:00,520
Speaker 3: it's too late. It really is, you know, five year

899
00:47:00,560 --> 00:47:03,440
Speaker 3: psychological bias. You want to be invested five years ago,

900
00:47:03,480 --> 00:47:05,719
Speaker 3: where everybody wants to be today and vice versa. So

901
00:47:05,719 --> 00:47:09,040
Speaker 3: i'd be thinking about what are the improbable things that

902
00:47:09,160 --> 00:47:11,480
Speaker 3: are likely to happen in the next wave. I'll give

903
00:47:11,480 --> 00:47:15,640
Speaker 3: you one company that I think is interesting that lux

904
00:47:15,719 --> 00:47:18,759
Speaker 3: is not invested in. It's a public company and we

905
00:47:18,800 --> 00:47:21,840
Speaker 3: do private but cloud Flare. You know, if you go

906
00:47:21,880 --> 00:47:24,800
Speaker 3: back to the Internet early days, one of the winners

907
00:47:24,960 --> 00:47:27,359
Speaker 3: in the infrastructure was Akamai, the people that were sort

908
00:47:27,360 --> 00:47:30,719
Speaker 3: of caching and they were helping to shape the structure

909
00:47:30,719 --> 00:47:33,759
Speaker 3: of the Internet. Cloud Flare is very interesting because they

910
00:47:33,800 --> 00:47:37,279
Speaker 3: have a lot of compute infrastructure at the edge of

911
00:47:37,320 --> 00:47:38,920
Speaker 3: the network. And you hear about this in sort of

912
00:47:38,960 --> 00:47:41,960
Speaker 3: a hype way sometimes the edge edge inference, edge compute.

913
00:47:42,080 --> 00:47:44,840
Speaker 3: It's a real thing. Very simply, you're talking on a

914
00:47:44,840 --> 00:47:47,680
Speaker 3: mobile device or you're on your computer right now, you

915
00:47:47,719 --> 00:47:49,600
Speaker 3: have to go up to the cloud, and the cloud,

916
00:47:49,840 --> 00:47:51,600
Speaker 3: you know, which is basically a bunch of servers somewhere

917
00:47:51,640 --> 00:47:56,200
Speaker 3: with high bandwidth interconnectivity processes. Then you have another domain

918
00:47:56,239 --> 00:47:58,879
Speaker 3: which is on device, so you you know, do something.

919
00:47:58,960 --> 00:48:01,600
Speaker 3: The models get smaller, the chips get better. On your

920
00:48:01,600 --> 00:48:04,200
Speaker 3: Apple device or your Android or your iPad, you're able

921
00:48:04,239 --> 00:48:07,720
Speaker 3: to run the AI model there. Cloud Flare is caching

922
00:48:07,760 --> 00:48:10,160
Speaker 3: a lot of these models and hosting them very close

923
00:48:10,160 --> 00:48:12,920
Speaker 3: to the users, and they're doing it in thousands or

924
00:48:12,920 --> 00:48:14,960
Speaker 3: tens of thousands of places all over the world. So

925
00:48:15,080 --> 00:48:17,520
Speaker 3: I think that they, you know, probably a twenty billion

926
00:48:17,560 --> 00:48:22,760
Speaker 3: dollars market cap company, billion revenue, fifty sixty percent growth.

927
00:48:23,080 --> 00:48:25,319
Speaker 3: I think that they might be poised and aren't one

928
00:48:25,360 --> 00:48:27,120
Speaker 3: of the names that are on the tips of people's

929
00:48:27,120 --> 00:48:29,960
Speaker 3: tongues that are benefiting. But we see them in all

930
00:48:30,000 --> 00:48:31,920
Speaker 3: the infrastructure behind a lot of our companies.

931
00:48:32,160 --> 00:48:37,800
Speaker 1: Interesting and a little investment tip, but for people listening.

932
00:48:37,560 --> 00:48:39,960
Speaker 3: Yeah, it's just it's you know, do your work investigated.

933
00:48:40,000 --> 00:48:42,200
Speaker 3: But it's something that is just not on the front page.

934
00:48:42,200 --> 00:48:43,879
Speaker 3: And I think that they're poised in the same way

935
00:48:43,920 --> 00:48:45,799
Speaker 3: that if I go back ten years when I'm in

936
00:48:45,840 --> 00:48:48,160
Speaker 3: that room in our startup and I got the benefit

937
00:48:48,200 --> 00:48:51,359
Speaker 3: of this legal inside information of seeing these guys from

938
00:48:51,480 --> 00:48:54,680
Speaker 3: Nvidia making these chips that were the soul of the

939
00:48:54,719 --> 00:48:57,799
Speaker 3: new machine, you know, in the proverbial Tracy kittersense, I

940
00:48:57,840 --> 00:49:00,439
Speaker 3: just see that this infrastructure from folks like out Flair

941
00:49:00,560 --> 00:49:01,480
Speaker 3: is probably going to win.

942
00:49:01,680 --> 00:49:04,080
Speaker 1: So I just have one more question as well, and

943
00:49:04,120 --> 00:49:07,160
Speaker 1: it actually also is sort of on the public market side,

944
00:49:07,600 --> 00:49:10,160
Speaker 1: but you know, going back to a company like alphabet

945
00:49:10,280 --> 00:49:13,160
Speaker 1: and obvious and I sort of talked about this in

946
00:49:13,200 --> 00:49:16,359
Speaker 1: my introduction, and you know, obviously they've made a lot

947
00:49:16,400 --> 00:49:19,160
Speaker 1: of AI investments and you know they've been doing research

948
00:49:19,719 --> 00:49:23,960
Speaker 1: for a long time. Nonetheless, though, like the core business

949
00:49:24,000 --> 00:49:27,080
Speaker 1: for now and for probably least the medium term is

950
00:49:27,120 --> 00:49:29,279
Speaker 1: going to be what we call, you know, Google dot

951
00:49:29,320 --> 00:49:31,480
Speaker 1: com or something like that, and enter a search and

952
00:49:31,520 --> 00:49:34,879
Speaker 1: get served a really compelling ad because it's very good

953
00:49:34,880 --> 00:49:38,800
Speaker 1: at that. Like, in your view, how confident should people

954
00:49:38,880 --> 00:49:43,200
Speaker 1: be that some of these big companies can find you know,

955
00:49:43,280 --> 00:49:47,200
Speaker 1: can actually produce revenue and income. I mean, like inference

956
00:49:47,320 --> 00:49:51,360
Speaker 1: is a lot costlier, I presume than a typical search query.

957
00:49:51,400 --> 00:49:53,239
Speaker 1: We don't know what the advertising is going to look

958
00:49:53,280 --> 00:49:55,520
Speaker 1: around it, et cetera. We don't really know. Like, do

959
00:49:55,560 --> 00:49:58,200
Speaker 1: you think it's obvious that these big companies are going

960
00:49:58,239 --> 00:50:02,719
Speaker 1: to find ways to actually sell something profitably from this tech?

961
00:50:03,080 --> 00:50:06,160
Speaker 3: I do, if there's good, strong leadership. And that sounds

962
00:50:06,200 --> 00:50:09,600
Speaker 3: like a you know, a weasel answer, but you know, historically,

963
00:50:09,719 --> 00:50:13,480
Speaker 3: if you look at Satia and Microsoft and you look

964
00:50:13,560 --> 00:50:17,279
Speaker 3: at Google, I just feel like Google was run by

965
00:50:17,280 --> 00:50:19,880
Speaker 3: the inmates for a very long time, and this competitive

966
00:50:19,960 --> 00:50:22,919
Speaker 3: near existential threat from open Ai has given a sense

967
00:50:22,960 --> 00:50:25,640
Speaker 3: of urgency for them to refocus and say, Okay, we

968
00:50:25,719 --> 00:50:28,120
Speaker 3: got to stop with all of the social stuff that

969
00:50:28,280 --> 00:50:31,080
Speaker 3: is happening internally and we got to really focus on

970
00:50:31,320 --> 00:50:34,239
Speaker 3: what we are roots were. Google hasn't really had a

971
00:50:34,360 --> 00:50:37,399
Speaker 3: killer product, I mean, a true new product in over

972
00:50:37,440 --> 00:50:41,560
Speaker 3: ten years. But what's interesting is YouTube is a big

973
00:50:41,600 --> 00:50:44,080
Speaker 3: winning I mean, that was a great acquisition. It's a

974
00:50:44,120 --> 00:50:47,239
Speaker 3: thriving product. It's generating a lot of money. Hopefully they

975
00:50:47,280 --> 00:50:49,440
Speaker 3: don't go crazy and spend, you know, like everybody else

976
00:50:49,480 --> 00:50:52,520
Speaker 3: in the streaming wars. But just like Facebook, right, Facebook

977
00:50:52,520 --> 00:50:55,719
Speaker 3: dot com is dead right, what makes money for Facebook

978
00:50:55,760 --> 00:50:58,680
Speaker 3: is everything else the Instagram and WhatsApp and if Threads

979
00:50:58,680 --> 00:51:00,799
Speaker 3: takes off, you know who knows, you know, and they're

980
00:51:00,800 --> 00:51:02,960
Speaker 3: able to capture some modicum of the enterprise value that

981
00:51:02,960 --> 00:51:05,440
Speaker 3: has been destroyed by Elon with Twitter. So it's all

982
00:51:05,480 --> 00:51:09,520
Speaker 3: of these ancillary product categories inside of the mothership that

983
00:51:09,600 --> 00:51:11,680
Speaker 3: I think that people are cranking and figuring out how

984
00:51:11,719 --> 00:51:14,759
Speaker 3: do we make this work. Google's prominence and search I

985
00:51:14,800 --> 00:51:17,080
Speaker 3: think is going to persist. I think it'll extend into

986
00:51:17,080 --> 00:51:19,840
Speaker 3: other domains. I think it's less likely to be threatened

987
00:51:19,840 --> 00:51:22,200
Speaker 3: by a lot of the AI stuff. They'll integrate it

988
00:51:22,800 --> 00:51:26,120
Speaker 3: barred when it first launched, socked. Now it's not bad,

989
00:51:26,560 --> 00:51:30,239
Speaker 3: you know, Incremental search results are pretty good, but the

990
00:51:30,239 --> 00:51:34,160
Speaker 3: corpus of information that they have from my photos, to

991
00:51:34,280 --> 00:51:37,359
Speaker 3: my emails to my calendar, I'm pretty locked in. And

992
00:51:37,680 --> 00:51:42,359
Speaker 3: I'm relatively trusting of Google, also relatively, if not high,

993
00:51:42,360 --> 00:51:46,120
Speaker 3: trusting of Apple, and I've historically been very low trusting

994
00:51:46,320 --> 00:51:48,320
Speaker 3: of Meta. I always say that whenever Meta launched is

995
00:51:48,360 --> 00:51:51,560
Speaker 3: a product, the one feature that it lacks is trust.

996
00:51:52,120 --> 00:51:55,719
Speaker 3: And I think they're realizing, even if it's a little

997
00:51:55,760 --> 00:51:58,880
Speaker 3: bit of a showcase facade for both the regulator regulators

998
00:51:58,880 --> 00:52:01,319
Speaker 3: and the critics, that they really have to double down

999
00:52:01,320 --> 00:52:02,839
Speaker 3: on trust. And one of the ways to do that

1000
00:52:03,000 --> 00:52:06,000
Speaker 3: is a lot of open source stuff. So really watch

1001
00:52:06,040 --> 00:52:08,600
Speaker 3: for Meta to embrace open source in a giant way.

1002
00:52:09,040 --> 00:52:12,600
Speaker 1: Josh Wolf Lux Capital. That was a great conversation, great

1003
00:52:12,719 --> 00:52:15,719
Speaker 1: overview of sort of the market right now. Thank you

1004
00:52:15,760 --> 00:52:17,520
Speaker 1: so much for coming on out Law. It's got to

1005
00:52:17,520 --> 00:52:18,240
Speaker 1: have you back again.

1006
00:52:18,400 --> 00:52:19,680
Speaker 3: Joe Tracy, great to be with you.

1007
00:52:32,800 --> 00:52:35,640
Speaker 1: Tracy, can I just say, you know, I don't know,

1008
00:52:36,080 --> 00:52:39,560
Speaker 1: listeners might not. I'm an amateur, you know, songwriter, and

1009
00:52:39,719 --> 00:52:43,400
Speaker 1: my only goal is to get something published before like

1010
00:52:43,480 --> 00:52:45,640
Speaker 1: the computers are just like so good at it. That's

1011
00:52:45,640 --> 00:52:47,239
Speaker 1: like my you know, I've like maybe I have like

1012
00:52:47,280 --> 00:52:48,680
Speaker 1: a window of like a year or two. I just

1013
00:52:48,680 --> 00:52:51,759
Speaker 1: want like one public you know. I just want to

1014
00:52:51,760 --> 00:52:54,560
Speaker 1: have like one something someone singing one of my songs,

1015
00:52:54,600 --> 00:52:57,879
Speaker 1: and then the computers can do their things.

1016
00:52:57,960 --> 00:53:00,400
Speaker 2: I mean, I do think this is kind of most

1017
00:53:00,440 --> 00:53:04,600
Speaker 2: disturbing aspect of this whole AI discussion, which is that

1018
00:53:04,840 --> 00:53:07,399
Speaker 2: so far it seems to mostly apply to the fun

1019
00:53:07,440 --> 00:53:12,880
Speaker 2: stuff songwriting, poetry, making movies, and we're still sort of

1020
00:53:12,960 --> 00:53:17,600
Speaker 2: doing all the dredge work ourselves. But that was a

1021
00:53:17,640 --> 00:53:21,360
Speaker 2: really interesting conversation, I do think. So, I don't know,

1022
00:53:21,440 --> 00:53:26,040
Speaker 2: I take Josh's point about getting in early on some

1023
00:53:26,120 --> 00:53:29,279
Speaker 2: of this, but I'm looking at a chart of what

1024
00:53:29,320 --> 00:53:32,719
Speaker 2: am I looking at Google, you know, since the IPO,

1025
00:53:33,080 --> 00:53:35,439
Speaker 2: and if you got in in two thousand and seven,

1026
00:53:35,480 --> 00:53:38,280
Speaker 2: two thousand and eight, like I think you'd still be okay.

1027
00:53:38,880 --> 00:53:43,000
Speaker 2: You would have missed like maybe the life changing money,

1028
00:53:43,200 --> 00:53:47,400
Speaker 2: but you'd still be up significantly on your investment. So

1029
00:53:47,880 --> 00:53:51,640
Speaker 2: I do wonder. Obviously there's a lot of excitement around

1030
00:53:51,719 --> 00:53:54,440
Speaker 2: the prospects of AI and what it means for various companies.

1031
00:53:54,440 --> 00:53:56,560
Speaker 2: But I also feel like if you waited for the

1032
00:53:56,680 --> 00:54:00,880
Speaker 2: dust to settle a little bit, you wouldn't necessarily be

1033
00:54:01,520 --> 00:54:02,600
Speaker 2: automatically losing it.

1034
00:54:02,719 --> 00:54:05,640
Speaker 1: I like how your question and your point here is

1035
00:54:05,719 --> 00:54:10,480
Speaker 1: not really is basically like questioning the entire premise of

1036
00:54:10,560 --> 00:54:15,480
Speaker 1: venture venture capital, Like that is actually the entire subtext

1037
00:54:15,480 --> 00:54:16,080
Speaker 1: of question.

1038
00:54:16,160 --> 00:54:19,399
Speaker 2: After twenty twenty two, I think that's a valid question.

1039
00:54:19,400 --> 00:54:21,120
Speaker 1: I would just wait for it all to go public

1040
00:54:21,120 --> 00:54:23,879
Speaker 1: and yeah, yeah, it's fine by the index. The other

1041
00:54:23,960 --> 00:54:27,040
Speaker 1: thing which I hadn't appreciated, which I thought was really interesting.

1042
00:54:27,239 --> 00:54:30,439
Speaker 1: You know, obviously I know that you know, Microsoft, Open AI,

1043
00:54:30,800 --> 00:54:34,200
Speaker 1: Google or alphabet anthropic, but the sort of like the

1044
00:54:34,239 --> 00:54:36,120
Speaker 1: way in which some of this may be a function

1045
00:54:36,160 --> 00:54:39,160
Speaker 1: of the regulatory department. I had not really like appreciated

1046
00:54:39,239 --> 00:54:42,800
Speaker 1: that and why, Like, Okay, it's it's gonna be hard

1047
00:54:42,880 --> 00:54:46,080
Speaker 1: to like make big acquisitions, so you just invest in

1048
00:54:46,080 --> 00:54:50,239
Speaker 1: companies who spend most of their money with you, Like yeah, yeah,

1049
00:54:50,239 --> 00:54:52,240
Speaker 1: it's like sort of I hadn't appreciated that element.

1050
00:54:52,520 --> 00:54:55,239
Speaker 2: No, I think that was a really interesting angle and

1051
00:54:55,320 --> 00:54:58,520
Speaker 2: actually explains a lot of the choices and decisions that

1052
00:54:58,560 --> 00:55:00,960
Speaker 2: are being made at the moment, because sometimes you look

1053
00:55:00,960 --> 00:55:04,239
Speaker 2: at them and you're like, this is this is interesting.

1054
00:55:04,360 --> 00:55:06,759
Speaker 2: But I'm not sure I completely see what's happening here.

1055
00:55:06,760 --> 00:55:08,840
Speaker 2: But if you look at it from a regulatory slash

1056
00:55:08,920 --> 00:55:12,400
Speaker 2: reputational angle, it makes a lot of sense.

1057
00:55:12,800 --> 00:55:15,680
Speaker 1: Uh, you know what, I'm really excited about what Bloomberg

1058
00:55:15,760 --> 00:55:19,440
Speaker 1: Bloomberg GPT did. Josh said it's going to be one

1059
00:55:19,440 --> 00:55:20,000
Speaker 1: of the winners.

1060
00:55:20,120 --> 00:55:22,879
Speaker 2: I feel like we should be doing a disclaimer here.

1061
00:55:23,200 --> 00:55:26,760
Speaker 1: We work for Bloomberg. It's fairly obvious we worked for Bloomberg,

1062
00:55:26,920 --> 00:55:30,120
Speaker 1: but we did not tell Josh to say that Bloomberg GPT.

1063
00:55:30,160 --> 00:55:32,920
Speaker 1: But there's point about who has actually had well quality

1064
00:55:33,040 --> 00:55:33,840
Speaker 1: data is interesting?

1065
00:55:34,000 --> 00:55:36,160
Speaker 2: Yeah, And I would say so far, a lot of

1066
00:55:36,200 --> 00:55:39,080
Speaker 2: the excitement is around the chip makers, some of the

1067
00:55:39,120 --> 00:55:43,880
Speaker 2: incumbents like Microsoft. I haven't seen people get really excited

1068
00:55:43,880 --> 00:55:47,840
Speaker 2: about like insurance companies as an AI play yet, but

1069
00:55:48,120 --> 00:55:50,400
Speaker 2: I think there's something there. The other thing I wanted

1070
00:55:50,440 --> 00:55:54,919
Speaker 2: to say, and you asked this question about the user interphase,

1071
00:55:55,760 --> 00:55:59,640
Speaker 2: and I actually think it's really important in the story here.

1072
00:55:59,760 --> 00:56:02,160
Speaker 2: And this is where I would draw a parallel with

1073
00:56:02,280 --> 00:56:06,080
Speaker 2: blockchain and crypto, which is the interesting thing about crypto

1074
00:56:06,360 --> 00:56:09,799
Speaker 2: was that you could participate in this as a sort

1075
00:56:09,840 --> 00:56:12,800
Speaker 2: of normal person. You know, you could open a wallet

1076
00:56:12,920 --> 00:56:16,360
Speaker 2: of some sort and buy whatever your preferred cryptocurrency is,

1077
00:56:16,400 --> 00:56:20,000
Speaker 2: so you could participate in it. And I think having

1078
00:56:20,040 --> 00:56:23,120
Speaker 2: something like open ai and various other models that you

1079
00:56:23,160 --> 00:56:27,560
Speaker 2: can play around with like clearly has drawn in that

1080
00:56:27,719 --> 00:56:30,960
Speaker 2: additional Oh yeah, like that is a big part of it.

1081
00:56:31,080 --> 00:56:33,719
Speaker 1: Absolutely. I do think like it's just like we've all

1082
00:56:33,760 --> 00:56:36,439
Speaker 1: had the jaw dropping moment which is like you didn't

1083
00:56:36,440 --> 00:56:38,239
Speaker 1: really get with crypto. It's like, yeah, you could do it,

1084
00:56:38,280 --> 00:56:39,839
Speaker 1: but then it's like, Okay, now I have this coin

1085
00:56:39,920 --> 00:56:43,080
Speaker 1: in my wallet, right, that is true, And then but

1086
00:56:43,120 --> 00:56:45,040
Speaker 1: then you just like you know, literally it takes you

1087
00:56:45,120 --> 00:56:47,320
Speaker 1: ten seconds to like be blown away.

1088
00:56:47,480 --> 00:56:49,960
Speaker 3: Is just so powerful. Yeah, shall we leave it there?

1089
00:56:50,040 --> 00:56:50,640
Speaker 1: Let's leave it there?

1090
00:56:50,640 --> 00:56:51,040
Speaker 3: All right.

1091
00:56:51,120 --> 00:56:54,080
Speaker 2: This has been another episode of the All Thoughts Podcast.

1092
00:56:54,200 --> 00:56:56,640
Speaker 2: I'm Tracy Alloway. You can follow me on Twitter at

1093
00:56:56,719 --> 00:56:57,960
Speaker 2: Tracy Alloway.

1094
00:56:57,880 --> 00:56:59,840
Speaker 1: And I'm Jill Why Isn't Though? You can follow me

1095
00:57:00,080 --> 00:57:03,240
Speaker 1: on Twitter at the Stalwart. Follow our guest Josh Wolf

1096
00:57:03,320 --> 00:57:07,160
Speaker 1: on Twitter. He's at wolf Josh. Follow our producers Carmen

1097
00:57:07,280 --> 00:57:11,080
Speaker 1: Rodriguez at Carmen Arman and dash Ol Bennett at dashbot.

1098
00:57:11,280 --> 00:57:14,040
Speaker 1: And check out all of the Bloomberg podcasts under the

1099
00:57:14,040 --> 00:57:17,760
Speaker 1: handle ad Podcasts. And for more odd Lots content, go

1100
00:57:17,800 --> 00:57:20,160
Speaker 1: to Bloomberg dot com slash odd Lots, where we have

1101
00:57:20,240 --> 00:57:23,760
Speaker 1: transcripts and a blog and a weekly newsletter. And check

1102
00:57:23,800 --> 00:57:26,080
Speaker 1: out the discord where we chat about all these things

1103
00:57:26,120 --> 00:57:28,360
Speaker 1: twenty four to seven. We even have an AI room

1104
00:57:28,360 --> 00:57:31,400
Speaker 1: in there. Some of the questions from the conversation I

1105
00:57:31,560 --> 00:57:34,440
Speaker 1: sourced from there, so go check it out Discord dot

1106
00:57:34,560 --> 00:57:35,920
Speaker 1: gg slash odd Lots.

1107
00:57:36,080 --> 00:57:39,760
Speaker 2: Also, if you enjoy odd Lots, if you appreciate conversations

1108
00:57:39,840 --> 00:57:42,240
Speaker 2: like the one we just had with Josh, please leave

1109
00:57:42,320 --> 00:57:46,480
Speaker 2: us a positive review on your favorite podcast platform. Thanks

1110
00:57:46,480 --> 00:58:04,080
Speaker 2: for listening in