1
00:00:02,720 --> 00:00:10,560
Speaker 1: Bloomberg Audio Studios, podcasts, radio news. You're listening to the

2
00:00:10,560 --> 00:00:14,480
Speaker 1: Bloomberg Intelligence Podcast. Catch us live weekdays at ten am

3
00:00:14,560 --> 00:00:18,479
Speaker 1: Eastern on Applecarplay and Android Auto with the Bloomberg Business App,

4
00:00:18,600 --> 00:00:21,799
Speaker 1: Listen on demand wherever you get your podcasts, or watch

5
00:00:21,920 --> 00:00:23,400
Speaker 1: us live on YouTube.

6
00:00:23,560 --> 00:00:26,880
Speaker 2: Let's take it one step further. Gene Munster, managing partner,

7
00:00:26,920 --> 00:00:29,360
Speaker 2: co founder for Loop Ventures, really one of the leading

8
00:00:29,880 --> 00:00:33,000
Speaker 2: voices on all things technology really for the last twenty

9
00:00:33,040 --> 00:00:37,040
Speaker 2: five thirty years. We appreciate getting a few minutes of

10
00:00:37,120 --> 00:00:39,640
Speaker 2: his time. Gene, I'm just gonna speak for myself, but

11
00:00:39,680 --> 00:00:43,120
Speaker 2: I probably speak for most of my radio and YouTube audience.

12
00:00:43,320 --> 00:00:45,520
Speaker 2: I did not know what deep Sick was, deep Sik

13
00:00:45,680 --> 00:00:50,200
Speaker 2: was before this morning, but now I'm learning. What should

14
00:00:50,200 --> 00:00:52,120
Speaker 2: I be learning? What should I takeaway be today?

15
00:00:54,400 --> 00:00:58,840
Speaker 3: I think that the highest level this whole chapter is about.

16
00:00:59,000 --> 00:01:02,520
Speaker 3: The first takeaway is that investors are nervous that this

17
00:01:03,120 --> 00:01:05,800
Speaker 3: eighty five percent moving the NASDAC over the past couple

18
00:01:05,840 --> 00:01:08,480
Speaker 3: of years is creating an environment where any sort of

19
00:01:09,319 --> 00:01:11,840
Speaker 3: ripple in what that AI trade looks like is going

20
00:01:11,880 --> 00:01:14,399
Speaker 3: to cause a significant impact on some of these valuations.

21
00:01:14,400 --> 00:01:16,640
Speaker 3: So I think that's I think this is a temperature

22
00:01:16,640 --> 00:01:18,520
Speaker 3: in terms of where the market is at as far

23
00:01:18,520 --> 00:01:21,399
Speaker 3: as kind of the company itself is concerned deep seek.

24
00:01:21,760 --> 00:01:26,920
Speaker 3: What's most important here is that they are advancing or

25
00:01:27,120 --> 00:01:30,399
Speaker 3: presumably advancing the cost of training these models you've talked

26
00:01:30,400 --> 00:01:32,800
Speaker 3: a lot about on the show today. And why that's

27
00:01:32,840 --> 00:01:35,920
Speaker 3: important is there's a question about how much hardware is

28
00:01:35,959 --> 00:01:38,920
Speaker 3: needed for that, and that obviously impacts the big hardware companies,

29
00:01:38,920 --> 00:01:42,160
Speaker 3: and then separately, how does that accelerate the adoption? So

30
00:01:42,800 --> 00:01:46,479
Speaker 3: there is the truth is always several layers below the surface,

31
00:01:46,520 --> 00:01:48,840
Speaker 3: and I think when it comes to Dee Sep there

32
00:01:48,880 --> 00:01:52,400
Speaker 3: are still some pretty significant unanswered questions, and I'll start

33
00:01:52,400 --> 00:01:54,360
Speaker 3: with one of probably the biggest one is that this

34
00:01:54,880 --> 00:01:57,520
Speaker 3: five to six million dollar training number that has the

35
00:01:57,560 --> 00:02:02,600
Speaker 3: market upside down today, that was not their most recent model.

36
00:02:02,600 --> 00:02:05,800
Speaker 3: They have an R model. It's unclear, you know, maybe

37
00:02:05,800 --> 00:02:08,440
Speaker 3: that was twenty five fifty one hundred million dollars to

38
00:02:08,480 --> 00:02:12,600
Speaker 3: train it. It probably was below what the most recent

39
00:02:12,680 --> 00:02:16,200
Speaker 3: version of Opening Eyes model, which is anticipated or estimated

40
00:02:16,200 --> 00:02:19,120
Speaker 3: to be about a five hundred million dollars in terms

41
00:02:19,120 --> 00:02:22,280
Speaker 3: of training. So I think big picture here is that

42
00:02:23,280 --> 00:02:26,760
Speaker 3: how we are delivering AI is evolving and there may

43
00:02:26,760 --> 00:02:30,639
Speaker 3: be some pretty significant improvements in terms of the ability

44
00:02:30,680 --> 00:02:32,919
Speaker 3: to do that at a lower cost.

45
00:02:33,400 --> 00:02:36,200
Speaker 4: So to that pointing, how much if we look at

46
00:02:36,280 --> 00:02:38,720
Speaker 4: Envidia in particular, it's down about two hundred and eighty

47
00:02:38,720 --> 00:02:41,920
Speaker 4: three points. What part of that is justified?

48
00:02:44,080 --> 00:02:46,240
Speaker 3: So I think the big picture here, I think that

49
00:02:46,320 --> 00:02:49,440
Speaker 3: this is an overreaction. I think when you look at

50
00:02:49,440 --> 00:02:52,200
Speaker 3: a company like Nvidia, it's a company we do own,

51
00:02:52,320 --> 00:02:55,080
Speaker 3: and I think that this is an overreaction, and just

52
00:02:55,240 --> 00:02:58,120
Speaker 3: specifically around that, if we can kind of zero in

53
00:02:58,200 --> 00:03:02,120
Speaker 3: on this concept that these models are becoming more efficient,

54
00:03:02,240 --> 00:03:05,040
Speaker 3: let's just take deep seek at face value that they've

55
00:03:05,040 --> 00:03:09,200
Speaker 3: had some sort of improvement. Is that that improvement if

56
00:03:09,240 --> 00:03:13,360
Speaker 3: you believe that the US tech companies are competent, that

57
00:03:13,400 --> 00:03:16,720
Speaker 3: improvement is around some architectures that have been talked about

58
00:03:16,760 --> 00:03:20,720
Speaker 3: for the last couple months. So those companies, the big

59
00:03:21,160 --> 00:03:25,520
Speaker 3: tech companies, the companies that are behind all the announcements

60
00:03:25,600 --> 00:03:31,239
Speaker 3: last week Meta increasing their CAPEX spend, and I think

61
00:03:31,280 --> 00:03:35,160
Speaker 3: that that is all that is a belief that they

62
00:03:35,240 --> 00:03:37,560
Speaker 3: still need this hardware. So when it comes back to

63
00:03:37,840 --> 00:03:41,000
Speaker 3: AI and comes back to the Nvidia trade, is that

64
00:03:41,080 --> 00:03:45,200
Speaker 3: I believe if they are reducing the costs to train

65
00:03:45,280 --> 00:03:49,000
Speaker 3: these models that I actually can build. I think a

66
00:03:49,040 --> 00:03:52,880
Speaker 3: credible case that that could increase the demand for some

67
00:03:52,960 --> 00:03:57,000
Speaker 3: of this hardware if we are getting closer to general intelligence.

68
00:03:57,000 --> 00:03:59,920
Speaker 3: If that potential is even closer, it's just going to

69
00:03:59,960 --> 00:04:02,840
Speaker 3: be an insatiable arms race. It's going to continue. And

70
00:04:02,880 --> 00:04:05,440
Speaker 3: so I understand why and video is down today. I

71
00:04:05,480 --> 00:04:07,800
Speaker 3: think it's actually a healthy thing. But ultimately I think

72
00:04:07,840 --> 00:04:09,480
Speaker 3: that their business is going to be just fine. We

73
00:04:09,520 --> 00:04:11,680
Speaker 3: have to wait twenty one trading days before we hear

74
00:04:11,840 --> 00:04:13,960
Speaker 3: from at least in VIDI on that, but I think

75
00:04:13,960 --> 00:04:14,760
Speaker 3: it's going to be fine.

76
00:04:15,440 --> 00:04:20,240
Speaker 2: So, Gene, was it a coincidence that on Friday, Meta

77
00:04:20,520 --> 00:04:24,640
Speaker 2: upsets capex forecast dramatically, which by a magnitude I've never seen,

78
00:04:24,680 --> 00:04:28,520
Speaker 2: from fifty billion to sixty five billion, and then today

79
00:04:28,560 --> 00:04:31,440
Speaker 2: we get this news on deep sick? Was Deep seek?

80
00:04:31,520 --> 00:04:34,400
Speaker 2: Was that kind of I don't know, coincidence?

81
00:04:35,800 --> 00:04:39,359
Speaker 3: I think, what's I think that's just purely coincidence. But

82
00:04:39,560 --> 00:04:41,840
Speaker 3: I do kind of go back to what we're just

83
00:04:41,880 --> 00:04:43,799
Speaker 3: talking about a minute ago. Is I think it's important

84
00:04:43,800 --> 00:04:47,359
Speaker 3: to note is that that announcement from Meta, if you

85
00:04:47,400 --> 00:04:50,000
Speaker 3: assume a Meta is Meta is competent, and I believe

86
00:04:50,000 --> 00:04:53,640
Speaker 3: that they are. That announcement came with the full knowledge

87
00:04:53,640 --> 00:04:56,159
Speaker 3: of what deep Seek was doing, so it's new to

88
00:04:56,240 --> 00:04:59,080
Speaker 3: most of us today. But for these companies who have

89
00:04:59,200 --> 00:05:02,719
Speaker 3: been making these huge investments, the concept of what deep

90
00:05:02,760 --> 00:05:05,200
Speaker 3: Seak had been trying to build has been They've had

91
00:05:05,240 --> 00:05:07,839
Speaker 3: a model that's been out. This has been something that's

92
00:05:07,880 --> 00:05:10,040
Speaker 3: been aware for the last couple of months, and so

93
00:05:10,080 --> 00:05:13,159
Speaker 3: I think that is important. Is that I think again

94
00:05:13,200 --> 00:05:16,600
Speaker 3: that this commentary from on Friday from Zuckerberg about the

95
00:05:16,640 --> 00:05:23,120
Speaker 3: increased capex factors in some of these potential breakthroughs that

96
00:05:23,160 --> 00:05:25,119
Speaker 3: we've seen with deep Seek. I want to stop short

97
00:05:25,120 --> 00:05:27,960
Speaker 3: of saying it's definitively a breakthrough because a lot of

98
00:05:28,040 --> 00:05:30,240
Speaker 3: unknowns around how good the models are and what the

99
00:05:30,279 --> 00:05:34,800
Speaker 3: true costs are. But I think that the AI investment

100
00:05:34,839 --> 00:05:37,840
Speaker 3: phase is still alive and well, this is still very early.

101
00:05:38,920 --> 00:05:41,880
Speaker 4: So if this could be an overreaction, we don't know

102
00:05:42,160 --> 00:05:45,760
Speaker 4: fair enough. In Nvidia, obviously an overreaction that stocked. The

103
00:05:45,800 --> 00:05:48,520
Speaker 4: way you look at it, does it make you think

104
00:05:48,640 --> 00:05:56,560
Speaker 4: though a rating of forty one times estimated PE.

105
00:05:55,839 --> 00:05:58,760
Speaker 3: So from where my head's at and thinking about calendar

106
00:05:58,800 --> 00:06:01,400
Speaker 3: twenty six at this point, and based on the calendar

107
00:06:01,400 --> 00:06:03,359
Speaker 3: twenty six numbers, I think it's closer to like a

108
00:06:03,400 --> 00:06:06,320
Speaker 3: twenty five, and I think that they can grow earnings

109
00:06:06,360 --> 00:06:08,839
Speaker 3: more than that. This is where the basically the rubber

110
00:06:08,880 --> 00:06:12,119
Speaker 3: hits the road, is that ultimately what happens in calendar

111
00:06:12,200 --> 00:06:14,039
Speaker 3: twenty six, all this stuff we're talking about with deep

112
00:06:14,080 --> 00:06:17,320
Speaker 3: seek today really impacts Ynvidia in twenty six. And if

113
00:06:17,360 --> 00:06:20,719
Speaker 3: in fact this dramatically changes how the kind of hardware

114
00:06:20,720 --> 00:06:22,880
Speaker 3: that people need less hardware, then I'm going to be wrong.

115
00:06:23,000 --> 00:06:26,479
Speaker 3: But if this does create this accelerated arms race, then

116
00:06:26,520 --> 00:06:28,120
Speaker 3: I think that they're going to grow faster. And so

117
00:06:28,520 --> 00:06:31,279
Speaker 3: I consider a one peg on that twenty six earnings

118
00:06:31,360 --> 00:06:36,119
Speaker 3: is actually attractive valuation relative to the rest of the group.

119
00:06:36,160 --> 00:06:38,000
Speaker 3: So I'm comfortable with the evaluation.

120
00:06:38,800 --> 00:06:41,280
Speaker 2: Just a red headline crossing the Bloomberg criminals we speak,

121
00:06:41,480 --> 00:06:46,200
Speaker 2: deep Seek says it is subject to large scale malicious attack.

122
00:06:46,360 --> 00:06:48,480
Speaker 2: So that's a headline. We'll have more reporting on that

123
00:06:48,640 --> 00:06:51,520
Speaker 2: going forward. Gene, just real quickly, we're going to hear

124
00:06:51,520 --> 00:06:54,760
Speaker 2: from some of the other big tech companies, Microsoft, Meta, Tesla,

125
00:06:54,760 --> 00:06:56,440
Speaker 2: how do you think they're going to address this issue?

126
00:06:57,360 --> 00:06:59,719
Speaker 3: It's obviously this is going to be front and center

127
00:06:59,760 --> 00:07:02,600
Speaker 3: the topic, and I think they're probably going to address

128
00:07:02,600 --> 00:07:06,200
Speaker 3: it by saying they still have plans to invest meaningfully

129
00:07:06,200 --> 00:07:08,440
Speaker 3: more in counter twenty five over twenty four when it

130
00:07:08,480 --> 00:07:10,920
Speaker 3: comes to Capex, and I think that that will play

131
00:07:11,000 --> 00:07:14,000
Speaker 3: part of a for reassuring we're long ways away from that.

132
00:07:14,040 --> 00:07:15,920
Speaker 3: We got two and a half trading days awave before

133
00:07:15,920 --> 00:07:17,760
Speaker 3: we start to get some of that commentary before the

134
00:07:18,440 --> 00:07:20,160
Speaker 3: mic gets turned over to the other side of the

135
00:07:20,400 --> 00:07:21,080
Speaker 3: equation here.

136
00:07:21,520 --> 00:07:23,360
Speaker 2: All right, Gene, thank you so much for joining us.

137
00:07:23,400 --> 00:07:25,640
Speaker 2: I know you're super busy today. Appreciate getting a few

138
00:07:25,640 --> 00:07:28,560
Speaker 2: minutes of your time. Gen Munster, managing partner co founder

139
00:07:28,880 --> 00:07:31,520
Speaker 2: Lupet Ventures, taking a little bit of a I guess

140
00:07:31,560 --> 00:07:33,720
Speaker 2: a longer term view, in a more broader view of

141
00:07:33,720 --> 00:07:37,840
Speaker 2: what this can mean for the indusuay.

142
00:07:36,680 --> 00:07:40,560
Speaker 1: You're listening to the Bloomberg Intelligence Podcast. Catch the program

143
00:07:40,680 --> 00:07:44,080
Speaker 1: live weekdays at ten am Eastern on Applecarplay and Android

144
00:07:44,080 --> 00:07:47,080
Speaker 1: Auto with the Bloomberg Business app. You can also listen

145
00:07:47,200 --> 00:07:50,440
Speaker 1: live on Amazon Alexa from our flagship New York station,

146
00:07:51,000 --> 00:07:53,680
Speaker 1: Just say Alexa play Bloomberg eleven thirty.

147
00:07:54,240 --> 00:07:56,120
Speaker 4: All right, let's get more on this text sell off

148
00:07:56,120 --> 00:07:59,040
Speaker 4: here Kim Forrest, founder and CIO of Boca Capital Partners.

149
00:07:59,080 --> 00:08:02,560
Speaker 4: She's also a former Chips person, like she did the stuff,

150
00:08:02,600 --> 00:08:03,720
Speaker 4: so she knows the things.

151
00:08:04,120 --> 00:08:04,480
Speaker 5: Clearly.

152
00:08:04,520 --> 00:08:08,240
Speaker 4: The deep Seek news is upending most of the tech industry, right.

153
00:08:08,600 --> 00:08:10,600
Speaker 4: You get the socks and DEXes down hard, you get

154
00:08:10,600 --> 00:08:13,720
Speaker 4: all the hyperscalers are down hard, all the chip makers,

155
00:08:13,760 --> 00:08:17,840
Speaker 4: all the power makers all down hard. Is this the

156
00:08:17,880 --> 00:08:20,200
Speaker 4: reckoning or is this a by the dip moment?

157
00:08:21,880 --> 00:08:26,000
Speaker 5: Well, that is the question, isn't it. I would say

158
00:08:26,000 --> 00:08:28,160
Speaker 5: it's a little bit of both. I think this is

159
00:08:28,200 --> 00:08:33,800
Speaker 5: a good level set, re leveling, resetting our expectations. So

160
00:08:34,240 --> 00:08:36,320
Speaker 5: the first thing is, we have to really see if

161
00:08:36,360 --> 00:08:41,720
Speaker 5: this is true. Now, apparently the company is in deep

162
00:08:41,800 --> 00:08:46,560
Speaker 5: Seek is an open source company. They have released how

163
00:08:46,600 --> 00:08:50,439
Speaker 5: they've done this, but you know, replication is key, so

164
00:08:50,520 --> 00:08:53,360
Speaker 5: we're gonna have to see if this actually works. That's

165
00:08:53,440 --> 00:08:57,760
Speaker 5: thing number one, Like, don't get too upset about major

166
00:08:57,840 --> 00:09:01,680
Speaker 5: developments until you know that they're real major developments, right,

167
00:09:01,760 --> 00:09:05,400
Speaker 5: So that's thing one. Thing two is and I think

168
00:09:05,440 --> 00:09:08,040
Speaker 5: this has been troubling me for a very long time,

169
00:09:09,120 --> 00:09:15,439
Speaker 5: especially with like last week's Stargate announcement. It seems as

170
00:09:15,480 --> 00:09:21,040
Speaker 5: if people have been expecting AI, which we don't know

171
00:09:21,040 --> 00:09:23,719
Speaker 5: what the payoff is yet we have a pretty good

172
00:09:23,760 --> 00:09:26,120
Speaker 5: idea it might help us out. But we're going to

173
00:09:26,160 --> 00:09:29,760
Speaker 5: cover the earth essentially in data centers and use far

174
00:09:29,840 --> 00:09:32,800
Speaker 5: more power than we create right now. So those two

175
00:09:32,840 --> 00:09:35,440
Speaker 5: things are kind of troubling that people have been making

176
00:09:35,480 --> 00:09:39,199
Speaker 5: investments in companies going to produce far more than they

177
00:09:39,240 --> 00:09:42,400
Speaker 5: can produce now. And do you see what I'm saying here?

178
00:09:42,440 --> 00:09:47,080
Speaker 5: We had these really big expectations, and I didn't think

179
00:09:47,080 --> 00:09:49,160
Speaker 5: that they were going to come to happen in short

180
00:09:49,240 --> 00:09:52,520
Speaker 5: or long term because of the well physical limitations of it.

181
00:09:53,160 --> 00:09:55,200
Speaker 2: So Kim, at this very early stage, I think I'm

182
00:09:55,200 --> 00:09:58,120
Speaker 2: probably representative of most of our listeners and viewers had

183
00:09:58,160 --> 00:10:00,679
Speaker 2: not heard of deep Seek before this morning. Sure, and

184
00:10:00,720 --> 00:10:03,120
Speaker 2: so that's calling into question kind of what we think

185
00:10:03,160 --> 00:10:06,560
Speaker 2: we understood about AI and its implication. It's not just

186
00:10:06,600 --> 00:10:09,920
Speaker 2: for technology preference stock market in general. My only question

187
00:10:09,960 --> 00:10:11,720
Speaker 2: that I think I have now that has any relevance

188
00:10:11,800 --> 00:10:16,160
Speaker 2: is do I have to rethink my spending associated with AI.

189
00:10:18,280 --> 00:10:21,280
Speaker 5: It depends on how long your timeframe is. If it's

190
00:10:21,320 --> 00:10:24,760
Speaker 5: the very short term, probably not you you know you'll

191
00:10:24,760 --> 00:10:27,920
Speaker 5: get rewarded because well orders are in and all that

192
00:10:28,000 --> 00:10:30,320
Speaker 5: kind of good stuff. And by short term I'm talking

193
00:10:30,320 --> 00:10:33,080
Speaker 5: a year, but let's say three to five years maybe,

194
00:10:33,920 --> 00:10:36,720
Speaker 5: And why is that? Well, if we can really train

195
00:10:36,880 --> 00:10:42,600
Speaker 5: models more rapidly with less input, which is a good thing, right,

196
00:10:42,679 --> 00:10:45,040
Speaker 5: where were we going to get all this electricity? That

197
00:10:45,240 --> 00:10:49,680
Speaker 5: was like my biggest question. But anyhow, back to your question,

198
00:10:50,240 --> 00:10:56,760
Speaker 5: I think you know this is part of being an

199
00:10:56,800 --> 00:11:00,960
Speaker 5: investors understanding how long was I play? I'm keeping this,

200
00:11:01,559 --> 00:11:05,080
Speaker 5: So that's part of the problem. We have two more

201
00:11:05,120 --> 00:11:09,200
Speaker 5: problems that are not being clarified by deep Seek. One is,

202
00:11:09,720 --> 00:11:12,080
Speaker 5: and remember I used to do this, so I'm way

203
00:11:12,080 --> 00:11:16,400
Speaker 5: down the road from most investors in thinking about things.

204
00:11:16,679 --> 00:11:20,200
Speaker 5: But as I use it, I'm not getting the results

205
00:11:20,240 --> 00:11:22,960
Speaker 5: that I need, And by that I mean it's not

206
00:11:23,200 --> 00:11:26,600
Speaker 5: right enough. The error rate that I get back from

207
00:11:26,640 --> 00:11:30,080
Speaker 5: the questions that I'm asking are anywhere between five and

208
00:11:30,160 --> 00:11:34,040
Speaker 5: maybe thirty percent, which that's huge. Like I can't just

209
00:11:34,280 --> 00:11:37,920
Speaker 5: use AI. I have to babysit AI. So these are

210
00:11:38,000 --> 00:11:41,520
Speaker 5: problems that aren't solved by anybody at this point. And

211
00:11:41,640 --> 00:11:44,760
Speaker 5: then there is how are we actually going to use it?

212
00:11:44,800 --> 00:11:48,040
Speaker 5: I think there was I think a Bloomberg report that

213
00:11:48,200 --> 00:11:51,319
Speaker 5: was saying that probably two hundred thousand people could be

214
00:11:51,440 --> 00:11:56,480
Speaker 5: laid off those lower level entry level people into finance

215
00:11:56,800 --> 00:12:00,480
Speaker 5: and then that mid tier that are now you know,

216
00:12:00,600 --> 00:12:05,720
Speaker 5: kind of like the AI for investment banking and equity research.

217
00:12:06,640 --> 00:12:09,839
Speaker 5: But if we were that wrong, we wouldn't last right, Like,

218
00:12:10,200 --> 00:12:13,439
Speaker 5: if you're five percent wrong in investment banking, your career

219
00:12:13,520 --> 00:12:17,440
Speaker 5: is really short. So you know, these are problems that

220
00:12:17,559 --> 00:12:22,000
Speaker 5: aren't being solved. So that's another issue for investors. How

221
00:12:22,360 --> 00:12:24,440
Speaker 5: right How long is it going to be till these

222
00:12:24,480 --> 00:12:26,160
Speaker 5: get right enough to replace a person?

223
00:12:26,640 --> 00:12:28,400
Speaker 4: That's such a great point because I was also reading

224
00:12:28,400 --> 00:12:30,640
Speaker 4: an article that talked about the limitations so far that

225
00:12:30,679 --> 00:12:35,520
Speaker 4: have been seen about Gee, what is it deep Seek?

226
00:12:35,520 --> 00:12:39,640
Speaker 4: I want to say, geek see deep Seek was its

227
00:12:40,080 --> 00:12:44,120
Speaker 4: lack of inability to discuss jijenping or to give real

228
00:12:44,200 --> 00:12:47,959
Speaker 4: handswers on Tanam and square, just those topics that are

229
00:12:47,960 --> 00:12:51,200
Speaker 4: near and dear to China's heart. And that's in part

230
00:12:51,240 --> 00:12:53,920
Speaker 4: some of the issue that you're talking about. So who

231
00:12:53,960 --> 00:12:55,480
Speaker 4: wins in this right now?

232
00:12:56,559 --> 00:13:00,760
Speaker 5: Who wins? Well, I think investors are losing in the

233
00:13:00,840 --> 00:13:04,800
Speaker 5: short term. But if you like me think that AI

234
00:13:05,080 --> 00:13:09,800
Speaker 5: is a path to productivity, and productivity always wins. I

235
00:13:09,840 --> 00:13:14,359
Speaker 5: think you could be a winner. I think maybe companies

236
00:13:14,520 --> 00:13:18,880
Speaker 5: like open Ai and Microsoft, you know, by dint of

237
00:13:19,679 --> 00:13:23,200
Speaker 5: you know, investing in them, and then everybody else who's

238
00:13:23,240 --> 00:13:28,000
Speaker 5: developing AI, they might be winners. Maybe the what is

239
00:13:28,040 --> 00:13:30,200
Speaker 5: it Stargate? I don't know why. I can never remember.

240
00:13:30,200 --> 00:13:32,560
Speaker 5: That's my break up here. I can ever remember that

241
00:13:33,000 --> 00:13:35,160
Speaker 5: because it has nothing to do with stars or gates,

242
00:13:35,520 --> 00:13:39,280
Speaker 5: that Stargate might be the loser in those people because

243
00:13:39,280 --> 00:13:42,880
Speaker 5: we're not necessarily going to need to cover the earth

244
00:13:42,880 --> 00:13:43,760
Speaker 5: in data centers.

245
00:13:43,920 --> 00:13:45,240
Speaker 4: Right, yep, all.

246
00:13:45,240 --> 00:13:47,559
Speaker 2: Right, Kim, thanks so much for joining us. Always appreciate

247
00:13:47,559 --> 00:13:51,000
Speaker 2: getting your perspective of your technology background. Kim Farst, founder

248
00:13:51,320 --> 00:13:56,080
Speaker 2: and chief investment officer of Book Capital Partners, joining us

249
00:13:56,160 --> 00:13:57,920
Speaker 2: from Pittsburgh via the zoom Thing.

250
00:14:00,040 --> 00:14:04,000
Speaker 1: Listening to the Bloomberg Intelligence Podcast. Catch us live weekdays

251
00:14:04,000 --> 00:14:07,040
Speaker 1: at ten am Eastern on Applecarplay and Android Auto with

252
00:14:07,120 --> 00:14:10,240
Speaker 1: the Bloomberg Business App. Listen on demand wherever you get

253
00:14:10,280 --> 00:14:13,520
Speaker 1: your podcasts, or watch us live on YouTube.

254
00:14:14,160 --> 00:14:16,720
Speaker 4: Happy mindy, everybody, Alex Deal here alongside Paul Sweeny. This

255
00:14:16,720 --> 00:14:19,400
Speaker 4: is Bloomberg Intelligence Radio. We are broadcasting to live for

256
00:14:19,560 --> 00:14:23,040
Speaker 4: Interactive Brookers Studio right here in Midtown Manhattan. Also check

257
00:14:23,120 --> 00:14:25,920
Speaker 4: us out on YouTube dot com. Or clearly have the

258
00:14:25,920 --> 00:14:29,880
Speaker 4: tech sel off underway really concentrated in certain names the

259
00:14:29,920 --> 00:14:33,040
Speaker 4: mag seven obviously, in Vidia, the chip stocks, although Apple's up,

260
00:14:33,080 --> 00:14:35,200
Speaker 4: I should say, but in Vidia obviously getting pounded, the

261
00:14:35,240 --> 00:14:37,400
Speaker 4: chip stocks getting pounded as well. So we wanted to

262
00:14:37,800 --> 00:14:40,960
Speaker 4: continue the conversation about AI and you really need to

263
00:14:41,000 --> 00:14:45,040
Speaker 4: rethink this investment thesis. Joining us now is Josh Pantoni.

264
00:14:45,120 --> 00:14:48,840
Speaker 4: He's CEO boosted dot Ai Now. The company has helped

265
00:14:48,920 --> 00:14:53,080
Speaker 4: dozens of investment managers who's AUM totals over one trillion

266
00:14:53,120 --> 00:14:56,440
Speaker 4: dollars and how to implement machine learning in their portfolios.

267
00:14:56,880 --> 00:15:01,720
Speaker 4: First off, Josh, have you seen this new geek seek? Geeks?

268
00:15:01,720 --> 00:15:04,000
Speaker 4: Why do I keep saying geek seek? It's not geeks

269
00:15:04,000 --> 00:15:10,680
Speaker 4: seek deep see see if you try deep seak? Have

270
00:15:10,720 --> 00:15:12,720
Speaker 4: you talked to your clients about it? What do you think?

271
00:15:13,720 --> 00:15:16,240
Speaker 6: Yeah? So, I mean it came out I think last Monday.

272
00:15:16,680 --> 00:15:19,200
Speaker 6: We've already ran a whole suite of benchmarks against it.

273
00:15:19,200 --> 00:15:20,480
Speaker 6: Try that out over a bunch of things, and I'd

274
00:15:20,520 --> 00:15:24,120
Speaker 6: say we're pretty familiar with it. I think it's impressive

275
00:15:24,440 --> 00:15:27,560
Speaker 6: in a number of ways. It's impressive how how much

276
00:15:27,600 --> 00:15:29,800
Speaker 6: power you get for the cost. I think the flip

277
00:15:29,840 --> 00:15:33,520
Speaker 6: side is it's actually not quite something that we could

278
00:15:33,560 --> 00:15:36,000
Speaker 6: even consider implementing it to our process. And I think

279
00:15:36,000 --> 00:15:37,360
Speaker 6: there's a lot of clients that will reach the same

280
00:15:37,400 --> 00:15:40,240
Speaker 6: conclusion after they start working with it. Why is that.

281
00:15:42,080 --> 00:15:42,160
Speaker 5: So?

282
00:15:42,480 --> 00:15:44,880
Speaker 6: A lot of how we help our clients is with

283
00:15:44,960 --> 00:15:48,640
Speaker 6: things like trying to identify risk and trying to identify

284
00:15:49,400 --> 00:15:51,440
Speaker 6: to build out these different workflows for different parts of

285
00:15:51,440 --> 00:15:53,840
Speaker 6: the financial process. And a really big part of that

286
00:15:54,200 --> 00:15:56,520
Speaker 6: is related to trust. You need to be able to

287
00:15:56,560 --> 00:15:58,480
Speaker 6: give answers that you're very confident in. You need to

288
00:15:58,480 --> 00:16:01,480
Speaker 6: give answers that you can sort of deeply into it

289
00:16:01,520 --> 00:16:03,200
Speaker 6: where it's coming from. And I think one of the

290
00:16:03,320 --> 00:16:05,200
Speaker 6: challenge of this model is that it has some very

291
00:16:05,280 --> 00:16:08,160
Speaker 6: clear political biases right off the bat. So if I'm

292
00:16:08,200 --> 00:16:10,880
Speaker 6: trying to use it to say understand what my portfolio

293
00:16:10,880 --> 00:16:14,400
Speaker 6: exposure is to China's going to be a challenge to

294
00:16:14,520 --> 00:16:17,400
Speaker 6: use it for things like that. Also, a lot of

295
00:16:17,440 --> 00:16:20,080
Speaker 6: the things we're most excited about are things like being

296
00:16:20,120 --> 00:16:21,960
Speaker 6: able to do computer use, being able to sort of

297
00:16:21,960 --> 00:16:24,480
Speaker 6: teach the machine how to like interact with different apps

298
00:16:24,480 --> 00:16:27,960
Speaker 6: and things on the computer, image recognition, trying to extract

299
00:16:27,960 --> 00:16:30,520
Speaker 6: out charts and things like that, and a lot of

300
00:16:30,560 --> 00:16:32,760
Speaker 6: those kind of capabilities that actually doesn't really have it

301
00:16:32,760 --> 00:16:34,680
Speaker 6: and seem to be as sophisticated in so it can

302
00:16:34,720 --> 00:16:37,920
Speaker 6: do very well in certain benchmarks, but when it actually

303
00:16:37,960 --> 00:16:39,560
Speaker 6: comes to trying to apply it for at least most

304
00:16:39,600 --> 00:16:41,800
Speaker 6: of the use cases that we're doing, it's not quite

305
00:16:41,880 --> 00:16:42,440
Speaker 6: there yet.

306
00:16:42,640 --> 00:16:44,920
Speaker 4: What is there for you right now? Like what does work?

307
00:16:46,440 --> 00:16:48,400
Speaker 6: I think the most interesting thing for me is actually

308
00:16:48,440 --> 00:16:50,440
Speaker 6: on the reasoning side. So of course, like one of

309
00:16:50,440 --> 00:16:52,480
Speaker 6: the really big pushes Opening Eyes had right now is

310
00:16:52,520 --> 00:16:55,600
Speaker 6: like with one and three, you've got these models coming

311
00:16:55,640 --> 00:16:58,080
Speaker 6: in where in theory they can take a task, break

312
00:16:58,120 --> 00:17:01,520
Speaker 6: it into subcomponents, and then start executing those components. And

313
00:17:01,560 --> 00:17:05,239
Speaker 6: I think in that area it's actually quite good. I

314
00:17:05,280 --> 00:17:08,160
Speaker 6: also think, you know, just from a cost basis, it's

315
00:17:08,200 --> 00:17:10,680
Speaker 6: extremely impressive they've been able to do. There's been some

316
00:17:11,160 --> 00:17:14,000
Speaker 6: controversy about exactly how much it costs to actually train

317
00:17:14,040 --> 00:17:16,440
Speaker 6: the model. What I can confirm though, is the actual

318
00:17:16,480 --> 00:17:19,440
Speaker 6: inference cost. The cost of running the model is extremely

319
00:17:19,520 --> 00:17:21,199
Speaker 6: cheap compared to what you're getting with like oh one

320
00:17:21,240 --> 00:17:23,240
Speaker 6: and oh three, and so just the fact is even

321
00:17:23,320 --> 00:17:25,480
Speaker 6: possible to do that as a major technology breakthrough.

322
00:17:26,320 --> 00:17:31,600
Speaker 2: If nothing else does Deep Seek just highlight the cost

323
00:17:31,840 --> 00:17:36,840
Speaker 2: issue and potentially the I don't know, the movement down

324
00:17:37,000 --> 00:17:38,960
Speaker 2: in costs of implementing AI.

325
00:17:40,480 --> 00:17:42,920
Speaker 6: Yeah, the way I tend to think about it is

326
00:17:43,320 --> 00:17:46,760
Speaker 6: the cheaper it is to train AI systems, the more

327
00:17:47,040 --> 00:17:49,320
Speaker 6: advanced capabilities you can train.

328
00:17:49,359 --> 00:17:52,240
Speaker 4: In the stress you lost your audio Josh, you there, Yes,

329
00:17:52,320 --> 00:17:52,680
Speaker 4: I am.

330
00:17:52,560 --> 00:17:54,120
Speaker 5: Here, still hear me. We're good?

331
00:17:54,160 --> 00:17:57,199
Speaker 6: Okay, perfect? Yeah. What the way I like to think

332
00:17:57,200 --> 00:18:00,160
Speaker 6: about it is the cheaper it is to train these

333
00:18:00,400 --> 00:18:04,600
Speaker 6: AI systems, Then the more advanced capabilities you can train,

334
00:18:04,760 --> 00:18:07,360
Speaker 6: the more advanced capabilities you can train, the more use

335
00:18:07,400 --> 00:18:10,280
Speaker 6: cases you unlock. The more use cases unlock, the more

336
00:18:10,320 --> 00:18:12,520
Speaker 6: you actually see it accelerate. So I actually think this

337
00:18:12,560 --> 00:18:16,040
Speaker 6: model will probably cause an acceleration in the capabilities of

338
00:18:16,040 --> 00:18:18,720
Speaker 6: other models that are getting built as more folks start

339
00:18:18,760 --> 00:18:20,199
Speaker 6: to adopt it. So for me, I actually see it

340
00:18:20,200 --> 00:18:21,120
Speaker 6: as completely positive.

341
00:18:22,480 --> 00:18:26,159
Speaker 4: What outside of deep Seek and other models, like, have

342
00:18:26,280 --> 00:18:27,800
Speaker 4: you worked with that you did?

343
00:18:27,920 --> 00:18:31,280
Speaker 6: Like yeah, so, I mean behind the scenes, we work

344
00:18:31,320 --> 00:18:33,640
Speaker 6: with in thropic models we work with open AI models.

345
00:18:34,280 --> 00:18:36,760
Speaker 6: We work with a bunch of fine tune models that

346
00:18:36,800 --> 00:18:40,040
Speaker 6: we build in house ourselves. You know. I like to

347
00:18:40,040 --> 00:18:42,840
Speaker 6: sort of describe it as an orchestra of different types

348
00:18:42,840 --> 00:18:44,840
Speaker 6: of models. And one of the things we've sort of

349
00:18:44,880 --> 00:18:48,520
Speaker 6: noticed is there's a lot of differences in capabilities. So

350
00:18:48,520 --> 00:18:50,840
Speaker 6: something like the inthroduct models and thropic models tend to

351
00:18:50,880 --> 00:18:53,120
Speaker 6: be better at very long context window where you try

352
00:18:53,119 --> 00:18:56,600
Speaker 6: and handle like huge amounts of text, whereas something like

353
00:18:56,640 --> 00:18:58,479
Speaker 6: the GPD form models tend to be a little bit

354
00:18:58,480 --> 00:19:01,800
Speaker 6: better at like foreign language and just sort of the

355
00:19:03,240 --> 00:19:06,240
Speaker 6: general verbiage it uses as it's giving outputs. So we

356
00:19:06,520 --> 00:19:07,280
Speaker 6: use a whole bunch.

357
00:19:08,760 --> 00:19:11,280
Speaker 2: Where how are your clients and the asset management business,

358
00:19:12,240 --> 00:19:14,320
Speaker 2: Josh using AI these days, in these early days.

359
00:19:15,680 --> 00:19:17,320
Speaker 6: Yeah, So the way I like to think about it

360
00:19:17,359 --> 00:19:18,920
Speaker 6: is we give them the ability to create this sort

361
00:19:18,960 --> 00:19:22,399
Speaker 6: of team of little AI workers where you teach them

362
00:19:22,440 --> 00:19:24,200
Speaker 6: how to do some kind of task, and then they're

363
00:19:24,200 --> 00:19:26,439
Speaker 6: going to do that task on a continuous basis. So

364
00:19:26,880 --> 00:19:29,040
Speaker 6: let's say you to do something like write an investment

365
00:19:29,040 --> 00:19:31,080
Speaker 6: thesis on a company, or let's say you had to

366
00:19:31,119 --> 00:19:34,080
Speaker 6: do something like write an ESG report or Let's say

367
00:19:34,119 --> 00:19:38,800
Speaker 6: you wanted to continuously monitor the world for any kinds

368
00:19:38,840 --> 00:19:41,520
Speaker 6: of updates to something that might happen in the air space.

369
00:19:42,040 --> 00:19:45,119
Speaker 6: These are all examples of sort of workflows you can

370
00:19:45,119 --> 00:19:47,000
Speaker 6: teach the system and have the system start to automate.

371
00:19:48,200 --> 00:19:50,160
Speaker 4: Well, we really appreciate your time. It was so good

372
00:19:50,160 --> 00:19:52,080
Speaker 4: to get that perspective. It's good to kind of get

373
00:19:52,080 --> 00:19:55,600
Speaker 4: the user mindset in for this. Josh Josh Pantoni, CEO

374
00:19:55,640 --> 00:19:59,119
Speaker 4: a boosted dot AI on deep Seek and sort of

375
00:19:59,160 --> 00:20:00,560
Speaker 4: the pros and the hans there.

376
00:20:01,480 --> 00:20:06,199
Speaker 1: This is the Bloomberg Intelligence podcast, available on Apple, Spotify,

377
00:20:06,400 --> 00:20:09,840
Speaker 1: and anywhere else you get your podcasts. Listen live each

378
00:20:09,880 --> 00:20:13,640
Speaker 1: weekday ten am to noon Eastern on Bloomberg dot com,

379
00:20:13,760 --> 00:20:17,280
Speaker 1: the iHeartRadio app, tune In, and the Bloomberg Business app.

380
00:20:17,720 --> 00:20:20,800
Speaker 1: You can also watch us live every weekday on YouTube

381
00:20:21,040 --> 00:20:23,280
Speaker 1: and always on the Bloomberg terminal.