1
00:00:00,240 --> 00:00:07,200
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:08,119 --> 00:00:11,320
Speaker 2: It's the rarest of product launches, when a company debuts

3
00:00:11,360 --> 00:00:14,239
Speaker 2: something most of the general public couldn't have imagined, and

4
00:00:14,280 --> 00:00:17,000
Speaker 2: then all of a sudden, it seems like no one

5
00:00:17,000 --> 00:00:18,400
Speaker 2: can talk about anything else.

6
00:00:19,239 --> 00:00:22,880
Speaker 3: Techi's everywhere short circuiting with excitement. This is the closest

7
00:00:22,880 --> 00:00:24,560
Speaker 3: thing I've saying to like the star Trek computer.

8
00:00:24,640 --> 00:00:26,400
Speaker 2: Are we going to have this thing filling out like

9
00:00:26,480 --> 00:00:27,240
Speaker 2: dating profiles?

10
00:00:27,280 --> 00:00:27,360
Speaker 3: Now?

11
00:00:27,400 --> 00:00:29,160
Speaker 2: Am I going to fall in love with chat gpt?

12
00:00:29,880 --> 00:00:33,440
Speaker 2: Open Ai released chat GPT in twenty twenty two, and

13
00:00:33,560 --> 00:00:37,640
Speaker 2: as enthusiasm for the chatbot surged, so did the company's ambitions,

14
00:00:37,760 --> 00:00:40,640
Speaker 2: along with interest from investors, which made another round of

15
00:00:40,640 --> 00:00:42,440
Speaker 2: headline grabbing news from open Ai.

16
00:00:42,640 --> 00:00:46,120
Speaker 1: So surprising the company that gave us chat GPT fired

17
00:00:46,240 --> 00:00:49,800
Speaker 1: it's CEO, Sam Altman, who has drawn comparisons to tech

18
00:00:49,840 --> 00:00:53,000
Speaker 1: giants like Steve Jobs, was dismissed by the open Ai

19
00:00:53,040 --> 00:00:56,520
Speaker 1: board Friday. We're reporting their efforts to get Sam Altman back.

20
00:00:56,760 --> 00:00:59,640
Speaker 2: Open Ai rein stinty Sam Altman is CEO and hitting

21
00:00:59,680 --> 00:01:03,080
Speaker 2: go on a bold shake out tapping Altman was fired

22
00:01:03,160 --> 00:01:06,760
Speaker 2: and rehired over the course of a long Weekend. Well,

23
00:01:06,840 --> 00:01:09,800
Speaker 2: it has been and continues to be a wild ride

24
00:01:09,840 --> 00:01:13,360
Speaker 2: for open Ai, a company that's not even ten. A

25
00:01:13,400 --> 00:01:16,520
Speaker 2: famous Altman quotation is the days are long, but the

26
00:01:16,640 --> 00:01:20,280
Speaker 2: decades are short. He recently amended that to the days

27
00:01:20,319 --> 00:01:23,600
Speaker 2: are long and the decades are also being very long.

28
00:01:23,959 --> 00:01:26,640
Speaker 2: At least that's what he told journalist Josh Tieringal recently.

29
00:01:26,760 --> 00:01:28,800
Speaker 2: Josh got an invite from open Ai to spend a

30
00:01:28,800 --> 00:01:32,120
Speaker 2: few days with Altman, someone he's interviewed before, for a

31
00:01:32,120 --> 00:01:34,600
Speaker 2: piece that's the cover story of the latest issue of

32
00:01:34,640 --> 00:01:35,720
Speaker 2: Bloomberg Business Week.

33
00:01:35,920 --> 00:01:39,120
Speaker 3: Two years after the launch of check GPT and a

34
00:01:39,200 --> 00:01:41,480
Speaker 3: year after the board firing, you know, they had some

35
00:01:41,520 --> 00:01:43,640
Speaker 3: stuff that kind of wanted to straighten out for the record.

36
00:01:44,160 --> 00:01:46,800
Speaker 2: Josh went to San Francisco and conducted what is the

37
00:01:46,800 --> 00:01:50,200
Speaker 2: most wide ranging interview Altman has done as open AI's

38
00:01:50,240 --> 00:01:53,440
Speaker 2: chief executive. They talked about the past and looked ahead

39
00:01:53,480 --> 00:01:56,480
Speaker 2: to the future, acknowledging that there is only so much

40
00:01:56,520 --> 00:01:59,280
Speaker 2: faith you can put in predictions about what's next for

41
00:01:59,360 --> 00:02:05,760
Speaker 2: a company like open Ai. I'm David Gera, and this

42
00:02:05,800 --> 00:02:08,480
Speaker 2: is the big take from Bloomberg News Today. On the show,

43
00:02:08,560 --> 00:02:12,000
Speaker 2: Josh Tieringel unpacks his interview with open AI's Sam Altman.

44
00:02:12,520 --> 00:02:15,400
Speaker 2: New details about the launch of chat gpt and that

45
00:02:15,480 --> 00:02:18,520
Speaker 2: weekend in twenty twenty three when Altman lost his job

46
00:02:18,720 --> 00:02:21,760
Speaker 2: and fought to get it back. Plus what Josh expects

47
00:02:21,840 --> 00:02:24,800
Speaker 2: AI regulation will look like during Donald Trump's second term,

48
00:02:25,240 --> 00:02:28,440
Speaker 2: and what motivated Altman to donate one million dollars to

49
00:02:28,480 --> 00:02:37,959
Speaker 2: Trump's inauguration committee. I'd love to start with some history

50
00:02:37,960 --> 00:02:40,680
Speaker 2: and a bit of biography here. What was Sam Waltman's

51
00:02:40,720 --> 00:02:43,120
Speaker 2: life like before he founded this company.

52
00:02:43,080 --> 00:02:47,280
Speaker 3: Before open ai took off. Sam Altman was a Silicon

53
00:02:47,400 --> 00:02:52,440
Speaker 3: Valley famous guy. He was known as a venture capitalist.

54
00:02:52,520 --> 00:02:55,160
Speaker 3: He was known as a sort of CEO and startup whisperer.

55
00:02:55,800 --> 00:02:59,880
Speaker 3: He had very rigorous standards as far as the kin

56
00:03:00,120 --> 00:03:03,120
Speaker 3: of people he would invest in. But he ran y Combinator,

57
00:03:03,160 --> 00:03:06,160
Speaker 3: which is a really famous fund that took an active

58
00:03:06,240 --> 00:03:09,760
Speaker 3: role in kind of curating leadership. He was also known

59
00:03:09,800 --> 00:03:12,840
Speaker 3: on the kind of circuit of you know, Ted talks

60
00:03:12,880 --> 00:03:16,679
Speaker 3: and conferences like he was that kind of famous. As

61
00:03:16,720 --> 00:03:20,360
Speaker 3: soon as chat gpt launched, I would say within weeks

62
00:03:20,720 --> 00:03:23,000
Speaker 3: he was world famous. And that's what happens when you

63
00:03:23,040 --> 00:03:26,160
Speaker 3: make something that one hundred million people use within six

64
00:03:26,200 --> 00:03:26,680
Speaker 3: eight weeks.

65
00:03:27,040 --> 00:03:29,360
Speaker 2: Can you walk us through what the launch of chat

66
00:03:29,360 --> 00:03:33,440
Speaker 2: gpt was like for him, how he experienced that moment people.

67
00:03:33,360 --> 00:03:36,200
Speaker 3: Knew that there were chatbots. Chatbots were a thing, they

68
00:03:36,200 --> 00:03:40,080
Speaker 3: were generally pretty clunky, and so chat GPT three point

69
00:03:40,080 --> 00:03:45,640
Speaker 3: five shows up, and people inside open Ai were really

70
00:03:45,680 --> 00:03:48,400
Speaker 3: skeptical of why they were launching it. They thought it

71
00:03:48,480 --> 00:03:51,240
Speaker 3: wasn't ready. Basically, they thought it was going to fail.

72
00:03:51,400 --> 00:03:54,040
Speaker 3: And Sam, as he details in the interview, said that,

73
00:03:54,440 --> 00:03:57,040
Speaker 3: you know, he doesn't make a whole lot of I

74
00:03:57,120 --> 00:03:59,520
Speaker 3: say we're doing it, we're doing it kinds of decisions,

75
00:03:59,520 --> 00:04:02,000
Speaker 3: but in this instance, you know, he kind of put

76
00:04:02,000 --> 00:04:04,280
Speaker 3: his finger in the wind. He read the zeitgeist. He

77
00:04:04,320 --> 00:04:07,520
Speaker 3: thought the product was more than good enough, and so

78
00:04:07,560 --> 00:04:09,920
Speaker 3: they launched it. And so for the first handful of

79
00:04:10,040 --> 00:04:13,400
Speaker 3: days it was doing okay, and there was a lot

80
00:04:13,440 --> 00:04:16,600
Speaker 3: of skepticism still within the company, well, well, look, this

81
00:04:16,080 --> 00:04:19,360
Speaker 3: is ridiculous. Why did we do it. It's not taking off.

82
00:04:19,920 --> 00:04:22,520
Speaker 3: Because he'd been at y Combinator, he was familiar with

83
00:04:22,560 --> 00:04:24,800
Speaker 3: the sort of pattern of a launch, and so what

84
00:04:24,839 --> 00:04:27,480
Speaker 3: he was seeing was in the first five six days

85
00:04:27,600 --> 00:04:29,479
Speaker 3: that you know, there would be a peak of usage

86
00:04:29,520 --> 00:04:31,320
Speaker 3: during the day and it would go down at night,

87
00:04:31,440 --> 00:04:33,040
Speaker 3: a peak of usage during the day and go down

88
00:04:33,040 --> 00:04:36,320
Speaker 3: at night. But what made it different and where he says,

89
00:04:36,360 --> 00:04:38,680
Speaker 3: I think that people didn't quite realize inside the company

90
00:04:38,760 --> 00:04:40,760
Speaker 3: what they had was that the trough was always higher

91
00:04:41,160 --> 00:04:43,760
Speaker 3: and the peak was always higher. And so after about

92
00:04:43,760 --> 00:04:46,520
Speaker 3: a week he was like, folks, we are failing to

93
00:04:46,600 --> 00:04:48,279
Speaker 3: understand what we have on our hands.

94
00:04:48,640 --> 00:04:52,000
Speaker 2: Central to this story is how this company is structured.

95
00:04:52,000 --> 00:04:54,960
Speaker 2: It's organized at the beginning as a nonprofit. Why was

96
00:04:55,000 --> 00:04:55,599
Speaker 2: that the case?

97
00:04:56,200 --> 00:04:59,320
Speaker 3: You know, the structure of the company is almost as

98
00:04:59,360 --> 00:05:04,359
Speaker 3: complicated explaining artificial general intelligence. It's weird, And I think

99
00:05:04,920 --> 00:05:08,520
Speaker 3: it started from this very sincere place, which was we're

100
00:05:08,520 --> 00:05:11,719
Speaker 3: going to make artificial intelligence that will benefit the world.

101
00:05:12,440 --> 00:05:15,440
Speaker 3: And so what we shouldn't do is have this incredible

102
00:05:15,480 --> 00:05:18,800
Speaker 3: profit motive looming over us at all times. Let's not

103
00:05:18,880 --> 00:05:23,000
Speaker 3: make short term quarterly decisions. Let's make decisions in decade

104
00:05:23,000 --> 00:05:25,080
Speaker 3: long increments. And that was the thing that all of

105
00:05:25,120 --> 00:05:28,479
Speaker 3: the founders agreed to, and among those founders or co

106
00:05:28,560 --> 00:05:31,919
Speaker 3: founders was Elon Musk Right. What they found out along

107
00:05:31,960 --> 00:05:34,120
Speaker 3: the way is not only that that was a sort

108
00:05:34,160 --> 00:05:37,240
Speaker 3: of doomed structure for a lot of just human reasons,

109
00:05:37,680 --> 00:05:41,440
Speaker 3: but that the power of compute, which is the noun

110
00:05:41,600 --> 00:05:43,960
Speaker 3: that we sort of use to talk about what it

111
00:05:44,080 --> 00:05:48,359
Speaker 3: costs to generate an artificial intelligence model. Just the cost

112
00:05:48,600 --> 00:05:53,400
Speaker 3: of GPUs, the cost of energy is so huge that

113
00:05:53,440 --> 00:05:56,640
Speaker 3: a nonprofit couldn't compete. And so at some point they

114
00:05:56,640 --> 00:05:59,560
Speaker 3: were confronted with the decision, which is remain this sort

115
00:05:59,560 --> 00:06:02,760
Speaker 3: of pure nonprofit in which you're kind of a sleepy

116
00:06:02,800 --> 00:06:06,560
Speaker 3: research arm but all the real computing work is happening

117
00:06:06,600 --> 00:06:08,839
Speaker 3: within the big three or four companies in the world,

118
00:06:09,400 --> 00:06:13,600
Speaker 3: or start to compete. And I think the founding origins

119
00:06:13,600 --> 00:06:16,480
Speaker 3: of that were sincere and that they really believed initially

120
00:06:16,480 --> 00:06:18,320
Speaker 3: when we're just going to do this for humanity, and

121
00:06:18,320 --> 00:06:19,520
Speaker 3: then had to confront reality.

122
00:06:20,000 --> 00:06:22,719
Speaker 2: You mentioned the sleepiness of research. It clearly is the

123
00:06:22,760 --> 00:06:25,960
Speaker 2: thing that animates him. He has this obsessiveness with research,

124
00:06:26,000 --> 00:06:30,360
Speaker 2: obsessiveness with artificial general intelligence. What is AGI?

125
00:06:31,160 --> 00:06:35,359
Speaker 3: David, Are you kidding me? You call me and you

126
00:06:35,440 --> 00:06:37,600
Speaker 3: go back and forth, you go back and well, I'll

127
00:06:37,600 --> 00:06:39,800
Speaker 3: tell you. I mean, well, yeah, one of the most

128
00:06:39,880 --> 00:06:44,520
Speaker 3: interesting things and confounding things if you're just a normal person.

129
00:06:44,600 --> 00:06:48,159
Speaker 3: Tuning into this debate is that even the people pursuing

130
00:06:48,320 --> 00:06:54,600
Speaker 3: artificial general intelligence cannot tell you what artificial general intelligence is.

131
00:06:55,240 --> 00:06:57,960
Speaker 3: And so even in the interview, at some point, you know,

132
00:06:58,000 --> 00:07:02,080
Speaker 3: Sam says, if it we were to create a model

133
00:07:02,640 --> 00:07:05,479
Speaker 3: that could do the work of multiple humans, that you

134
00:07:05,520 --> 00:07:07,840
Speaker 3: could assign it a task and it can complete it,

135
00:07:08,120 --> 00:07:12,520
Speaker 3: that would be AGI ish. So the guy, the most

136
00:07:12,600 --> 00:07:17,600
Speaker 3: famous guy pursuing AGI uses ish. Now that is confounding.

137
00:07:17,880 --> 00:07:21,400
Speaker 3: That said, a lot of great scientific discoveries and a

138
00:07:21,440 --> 00:07:23,880
Speaker 3: lot of you know, the things that propel us forward

139
00:07:24,000 --> 00:07:27,760
Speaker 3: in civilization are a little bit ish, but it's very

140
00:07:27,760 --> 00:07:30,080
Speaker 3: hard for me to tell you what AGI is if

141
00:07:30,080 --> 00:07:31,280
Speaker 3: Sam Altman can't do it.

142
00:07:31,600 --> 00:07:33,400
Speaker 2: Let me ask you one more question along these lines,

143
00:07:33,400 --> 00:07:35,600
Speaker 2: which is you mentioned you've spoken with him a number

144
00:07:35,640 --> 00:07:38,600
Speaker 2: of times. How has your sense of him, your understanding

145
00:07:38,600 --> 00:07:40,800
Speaker 2: of him evolved through those interviews.

146
00:07:41,120 --> 00:07:44,080
Speaker 3: I think that the one thing that continues to resonate

147
00:07:44,320 --> 00:07:48,360
Speaker 3: is he is one hundred percent held on getting to

148
00:07:48,400 --> 00:07:52,000
Speaker 3: AGI before anybody else, and he is running the company

149
00:07:52,600 --> 00:07:56,000
Speaker 3: with that singular goal in mind. And so when they're

150
00:07:56,000 --> 00:07:58,640
Speaker 3: doing all these raises, when they're staying focused on the

151
00:07:59,200 --> 00:08:01,640
Speaker 3: latest model, tried to stay out in front of everybody

152
00:08:01,720 --> 00:08:05,360
Speaker 3: I think it's because he's animated by this sense of

153
00:08:05,400 --> 00:08:08,080
Speaker 3: purpose around the science and a belief as a business

154
00:08:08,080 --> 00:08:12,440
Speaker 3: person that getting there first is the only thing that matters.

155
00:08:13,200 --> 00:08:15,320
Speaker 2: There's this crucial moment in the Open Eye story that

156
00:08:15,360 --> 00:08:17,760
Speaker 2: plays out over a weekend. Sam Alton is fired. In

157
00:08:17,840 --> 00:08:20,560
Speaker 2: a few days later, he's rehired. We're now more than

158
00:08:20,600 --> 00:08:23,880
Speaker 2: a year away from that, and I wonder if anyone

159
00:08:23,920 --> 00:08:26,920
Speaker 2: if he has a clear understanding of why things transpired

160
00:08:26,920 --> 00:08:27,480
Speaker 2: the way they did.

161
00:08:27,960 --> 00:08:31,600
Speaker 3: I think he has the clearest understanding of anyone who

162
00:08:31,640 --> 00:08:34,439
Speaker 3: is not on that board. I tried to get at

163
00:08:34,480 --> 00:08:37,040
Speaker 3: it in a number of ways, right. I even offered

164
00:08:37,080 --> 00:08:40,600
Speaker 3: my own theory, which is that basically the board was

165
00:08:40,600 --> 00:08:44,200
Speaker 3: a bunch of purists and they were formed at a

166
00:08:44,240 --> 00:08:49,160
Speaker 3: moment when a nonprofit pursuit of artificial intelligence seemed like

167
00:08:49,200 --> 00:08:52,000
Speaker 3: the right thing, and that they were struggling to adapt

168
00:08:52,000 --> 00:08:55,599
Speaker 3: to the reality that basically being a nonprofit in this

169
00:08:55,679 --> 00:08:58,280
Speaker 3: space was going to doom the company to failure, and

170
00:08:58,360 --> 00:09:01,720
Speaker 3: that Sam was determined to let it fail. I think

171
00:09:01,760 --> 00:09:05,240
Speaker 3: that is actually the crux of the tension. My hunch,

172
00:09:05,440 --> 00:09:08,200
Speaker 3: honestly is that this was something that was doomed to

173
00:09:08,360 --> 00:09:10,600
Speaker 3: happen from the moment they decided to be founded as

174
00:09:10,640 --> 00:09:14,720
Speaker 3: a nonprofit without getting all the parties together in front

175
00:09:14,720 --> 00:09:17,280
Speaker 3: of microphonts. I think we have enough information to know

176
00:09:17,320 --> 00:09:20,600
Speaker 3: that this was a conflict around the purpose of the work,

177
00:09:21,400 --> 00:09:25,240
Speaker 3: and the original board really felt like it was consistent

178
00:09:25,240 --> 00:09:28,040
Speaker 3: with the mission of open Ai to kill the company

179
00:09:28,559 --> 00:09:32,439
Speaker 3: if it couldn't make AI with a sort of rigorous, safe,

180
00:09:33,080 --> 00:09:36,959
Speaker 3: nonprofit standard, and he was not going to let the

181
00:09:37,000 --> 00:09:37,720
Speaker 3: company die.

182
00:09:37,920 --> 00:09:41,040
Speaker 2: You asked Sam what the fallout has been from that moment,

183
00:09:41,080 --> 00:09:45,000
Speaker 2: from that hectic weekend, if he felt like afterward he

184
00:09:45,040 --> 00:09:47,480
Speaker 2: needed to convince his colleagues that he's good, I think

185
00:09:47,559 --> 00:09:49,960
Speaker 2: is how you put it. How did his firing, his

186
00:09:50,040 --> 00:09:52,120
Speaker 2: rehiring effect his ability to sort of work with people

187
00:09:52,120 --> 00:09:54,520
Speaker 2: in open ai and more broadly in the kind of

188
00:09:54,559 --> 00:09:55,680
Speaker 2: nascent AI industry.

189
00:09:56,000 --> 00:09:59,280
Speaker 3: Yeah, look, I'm super fascinated by the human elements behind

190
00:09:59,320 --> 00:10:00,880
Speaker 3: all of this stuff. You know, he said that the

191
00:10:00,880 --> 00:10:03,000
Speaker 3: first couple of days and probably the first couple of

192
00:10:03,040 --> 00:10:06,840
Speaker 3: weeks were super weird. People looked at him funny. People

193
00:10:06,920 --> 00:10:10,000
Speaker 3: didn't know exactly what this was all about. I think

194
00:10:10,000 --> 00:10:13,400
Speaker 3: within the industry itself, you know, it was the classic

195
00:10:14,520 --> 00:10:17,080
Speaker 3: everybody at a competing company, you know, had a bowl

196
00:10:17,120 --> 00:10:19,880
Speaker 3: of popcorn and was like, let's see how this goes, right.

197
00:10:20,720 --> 00:10:23,560
Speaker 3: But I also think that there was an understanding in

198
00:10:23,600 --> 00:10:26,280
Speaker 3: the industry writ large, in the companies that they partner

199
00:10:26,320 --> 00:10:28,920
Speaker 3: with and the companies that they compete with, that Sam

200
00:10:29,240 --> 00:10:33,480
Speaker 3: is a force, and that they didn't question his credibility

201
00:10:33,600 --> 00:10:36,560
Speaker 3: or his credentials to run the company. They were probably

202
00:10:36,600 --> 00:10:39,960
Speaker 3: hoping he would get fired and be forced to start

203
00:10:40,000 --> 00:10:42,000
Speaker 3: all over again somewhere else that they could catch up

204
00:10:42,400 --> 00:10:43,920
Speaker 3: and then as far as it, you know, to return

205
00:10:43,920 --> 00:10:45,800
Speaker 3: to what happens inside the company, I think it's one

206
00:10:45,840 --> 00:10:48,160
Speaker 3: of those things where if you're there day in and

207
00:10:48,240 --> 00:10:52,000
Speaker 3: day out and there is the kind of attrition and

208
00:10:52,040 --> 00:10:55,440
Speaker 3: turnover that you would normally expect in a startup, you know,

209
00:10:55,440 --> 00:10:58,520
Speaker 3: within a couple of months, it was a blip. So

210
00:10:58,600 --> 00:11:00,120
Speaker 3: I think that's how you reckoned with it.

211
00:11:02,679 --> 00:11:05,800
Speaker 2: After the break, we turned from open AI's past to

212
00:11:05,920 --> 00:11:16,880
Speaker 2: its future. When Sam Alton returned as the CEO of

213
00:11:16,920 --> 00:11:20,080
Speaker 2: open Ai, he retook the helm of a rapidly evolving

214
00:11:20,200 --> 00:11:24,240
Speaker 2: and closely watched company. He told journalist Josh Tirerngeal that

215
00:11:24,360 --> 00:11:28,040
Speaker 2: after Altman was reinstated. His attitude was, as he put it,

216
00:11:28,240 --> 00:11:30,480
Speaker 2: we got a complicated job to do. I'm going to

217
00:11:30,600 --> 00:11:33,600
Speaker 2: keep doing this, basically resolved to put his head down

218
00:11:33,720 --> 00:11:36,480
Speaker 2: and get back to work. I asked Josh how that

219
00:11:36,520 --> 00:11:39,200
Speaker 2: played out and what altman's life is like day to day.

220
00:11:39,840 --> 00:11:42,120
Speaker 2: There's a moment where he shows you his calendar, and

221
00:11:42,160 --> 00:11:43,600
Speaker 2: I wonder if you could just describe sort of what

222
00:11:43,679 --> 00:11:44,080
Speaker 2: that's like.

223
00:11:44,480 --> 00:11:47,560
Speaker 3: Yeah, so, Sam, when I asked about how he runs

224
00:11:47,600 --> 00:11:49,600
Speaker 3: the company, just sort of day to day where he's

225
00:11:49,600 --> 00:11:52,040
Speaker 3: spending his time, you know, he just flipped his laptop

226
00:11:52,080 --> 00:11:56,280
Speaker 3: around and pulled up his Google calendar and it's a mess.

227
00:11:56,559 --> 00:12:01,040
Speaker 3: I mean, it's just an absolute mess of colors, conflicts

228
00:12:01,240 --> 00:12:04,240
Speaker 3: starting from about seven am going to about nine fifteen,

229
00:12:04,360 --> 00:12:09,760
Speaker 3: with some dinners after even that, and lots of overlapping meetings,

230
00:12:09,800 --> 00:12:12,959
Speaker 3: lots of small meetings, lots of one on ones with engineers.

231
00:12:13,320 --> 00:12:16,040
Speaker 3: I'm sure many of your listeners have calendars that look similar.

232
00:12:16,520 --> 00:12:18,160
Speaker 3: What I would say is that it was just day

233
00:12:18,520 --> 00:12:22,440
Speaker 3: after day after day prescheduled. There's not a lot of

234
00:12:22,480 --> 00:12:25,760
Speaker 3: walking around time. It's indicative of a company that is

235
00:12:25,880 --> 00:12:30,680
Speaker 3: in a full competitive sprint to get someplace. So yeah,

236
00:12:30,720 --> 00:12:32,120
Speaker 3: it was pretty daunting.

237
00:12:32,520 --> 00:12:34,040
Speaker 2: Let me ask you a bit about the future of

238
00:12:34,080 --> 00:12:36,520
Speaker 2: the company and his plans for it. Opening Eye has

239
00:12:36,520 --> 00:12:38,960
Speaker 2: already changed the history the trajectory of the AI industry.

240
00:12:39,800 --> 00:12:42,400
Speaker 2: What did Sam tell you about his plans for the

241
00:12:42,400 --> 00:12:44,600
Speaker 2: future of the company, sort of where he sees things going.

242
00:12:45,240 --> 00:12:48,720
Speaker 3: He used the word protect, that the company is structured

243
00:12:48,760 --> 00:12:51,679
Speaker 3: to protect research, and so I think those are the

244
00:12:51,720 --> 00:12:54,880
Speaker 3: words I would look to as I monitor the company

245
00:12:54,920 --> 00:12:57,040
Speaker 3: over the next few years. He is hell bent on

246
00:12:57,200 --> 00:13:00,920
Speaker 3: protecting the research and getting to AGI. I think he

247
00:13:01,000 --> 00:13:04,400
Speaker 3: believes everything else will take care of itself if they

248
00:13:04,440 --> 00:13:07,439
Speaker 3: can do that. And in that way, you know, even

249
00:13:07,480 --> 00:13:11,560
Speaker 3: though the tech is wildly new and unconventional, the business

250
00:13:11,559 --> 00:13:15,040
Speaker 3: approach is fairly conventional, right. It's not that different to

251
00:13:15,080 --> 00:13:18,240
Speaker 3: a late nineties, you know, web startup, which is we

252
00:13:18,320 --> 00:13:20,440
Speaker 3: got to get audience, right, We got to get as

253
00:13:20,480 --> 00:13:22,400
Speaker 3: big as possible, as quickly as possible, and then the

254
00:13:22,400 --> 00:13:27,520
Speaker 3: finances will sort themselves out. It's Amazon's strategy, It's Facebook strategy.

255
00:13:27,600 --> 00:13:30,240
Speaker 3: So that's how I would project with the next few

256
00:13:30,320 --> 00:13:34,920
Speaker 3: years old protect the research, productize effectively. See where we

257
00:13:34,960 --> 00:13:36,360
Speaker 3: are in eighteen months to two years.

258
00:13:36,720 --> 00:13:39,079
Speaker 2: Josh, what are the biggest challenges that he and OpenAI

259
00:13:39,200 --> 00:13:42,200
Speaker 2: face sort of seeking that objective. Is it bandwidth? Is

260
00:13:42,200 --> 00:13:43,200
Speaker 2: it getting compute?

261
00:13:43,679 --> 00:13:47,400
Speaker 3: The three challenges that the industry as a whole phases

262
00:13:47,640 --> 00:13:51,280
Speaker 3: are just getting the compute right, getting access to the

263
00:13:51,320 --> 00:13:55,439
Speaker 3: GPUs that you need the energy to power those GPUs.

264
00:13:55,960 --> 00:13:58,760
Speaker 3: And then the biggest question and most unknowable is are

265
00:13:58,800 --> 00:14:03,360
Speaker 3: the models plateau Are you continuing to see artificial intelligence

266
00:14:04,160 --> 00:14:09,600
Speaker 3: gain steam in training the models right? He's extremely confident

267
00:14:09,640 --> 00:14:12,680
Speaker 3: that their models are not plateauing. There's some debate within

268
00:14:12,720 --> 00:14:16,440
Speaker 3: the industry whether that's bluster or not. On energy, you know,

269
00:14:16,880 --> 00:14:22,360
Speaker 3: I'd like to be a fairly prepared, smooth interviewer. I

270
00:14:22,400 --> 00:14:26,680
Speaker 3: will admit I was rendered a momentarily mute when I

271
00:14:26,680 --> 00:14:31,200
Speaker 3: asked him about energy. He's like, Fusion's gonna work. I'm sorry,

272
00:14:31,200 --> 00:14:35,880
Speaker 3: what now? Which fusion? When? Huh? But you know, he

273
00:14:35,960 --> 00:14:38,360
Speaker 3: is a co founder of a company called Helion with

274
00:14:38,560 --> 00:14:42,040
Speaker 3: reied Hoffman and some others and believes in fusion is

275
00:14:42,040 --> 00:14:44,400
Speaker 3: coming and fusion will be a sort of silver bullet

276
00:14:44,440 --> 00:14:47,320
Speaker 3: for our energy issues. And then on chips, you know, look,

277
00:14:47,360 --> 00:14:51,320
Speaker 3: they like everybody else, they are finding and buying as

278
00:14:51,360 --> 00:14:53,400
Speaker 3: many chips as they possibly can they have their own

279
00:14:53,480 --> 00:14:56,760
Speaker 3: fab effort going to make sure that they're never beholding

280
00:14:56,760 --> 00:14:59,520
Speaker 3: to one supplier. So those are the three things. He

281
00:14:59,640 --> 00:15:02,440
Speaker 3: feels very confident that their position to address all three.

282
00:15:02,760 --> 00:15:04,800
Speaker 3: And then the fourth, which is unique to open AI,

283
00:15:04,920 --> 00:15:07,560
Speaker 3: which I asked about, is like, right, let's talk about

284
00:15:07,920 --> 00:15:11,640
Speaker 3: governance because we now are in a place where the

285
00:15:11,680 --> 00:15:14,400
Speaker 3: man who really is calling himself a kind of co

286
00:15:14,520 --> 00:15:19,000
Speaker 3: president has a competing AI company was once Sam Altman's

287
00:15:19,040 --> 00:15:22,000
Speaker 3: co founder and is currently suing the company. That is

288
00:15:22,240 --> 00:15:25,680
Speaker 3: a level of volatility. You know, I can't predict, he

289
00:15:25,720 --> 00:15:30,200
Speaker 3: can't predict, but Elon Musk's existence is a factor in

290
00:15:30,480 --> 00:15:32,760
Speaker 3: how companies will do in developing AI.

291
00:15:33,400 --> 00:15:36,000
Speaker 2: You mentioned governance, Let me ask you about his relationship

292
00:15:36,040 --> 00:15:39,760
Speaker 2: with Washington and policymakers and politicians there. He donated a

293
00:15:39,760 --> 00:15:45,160
Speaker 2: million dollars to Donald Trump's inauguration fund. He told you

294
00:15:45,200 --> 00:15:47,760
Speaker 2: he supports any president. How authentic did that sound to

295
00:15:47,800 --> 00:15:50,320
Speaker 2: you when he said that that he, you know, would

296
00:15:50,360 --> 00:15:52,360
Speaker 2: support a Democrat or Republican, whoever's in.

297
00:15:52,320 --> 00:15:54,400
Speaker 3: The White House. You know, it's hard for me to

298
00:15:54,440 --> 00:15:57,160
Speaker 3: speculate about his authenticity. What I can say is that

299
00:15:57,720 --> 00:16:02,680
Speaker 3: taking a step back strategy whether I'm Sam Waltman or

300
00:16:02,720 --> 00:16:06,160
Speaker 3: anyone else who is competing with Elon Musk, I think

301
00:16:06,200 --> 00:16:10,080
Speaker 3: the smartest approach is to lavish praise on the President

302
00:16:10,840 --> 00:16:13,280
Speaker 3: and to try and create any sort of friction I

303
00:16:13,320 --> 00:16:17,200
Speaker 3: possibly could between Elon Musk and the President. And so

304
00:16:17,360 --> 00:16:19,960
Speaker 3: by saying oh, I support of course, I support the

305
00:16:19,960 --> 00:16:23,280
Speaker 3: American President. I may not agree with everything, but he's

306
00:16:23,320 --> 00:16:26,200
Speaker 3: the president. I wish for his great success, and then

307
00:16:26,280 --> 00:16:29,360
Speaker 3: saying oh, you know Elon, you know he's going to

308
00:16:29,400 --> 00:16:31,680
Speaker 3: do what he's going to do. He's the co president.

309
00:16:31,960 --> 00:16:35,000
Speaker 3: I think anything that creates that friction is probably beneficial

310
00:16:35,040 --> 00:16:37,920
Speaker 3: to people competing with Elon. It will not surprise me

311
00:16:38,000 --> 00:16:41,320
Speaker 3: if that's a tactic taken by others as well. But

312
00:16:41,360 --> 00:16:42,760
Speaker 3: I you know, as I said, I can't speak to

313
00:16:42,760 --> 00:16:45,400
Speaker 3: the authenticity of it. I just think there's not the

314
00:16:45,400 --> 00:16:46,200
Speaker 3: worst strategy.

315
00:16:46,400 --> 00:16:49,800
Speaker 2: How does he see the role of Washington in regulating

316
00:16:50,000 --> 00:16:52,800
Speaker 2: this new technology? Does he see a role for Washington here?

317
00:16:53,360 --> 00:16:55,680
Speaker 2: And what does that role look like under this new administration?

318
00:16:56,000 --> 00:16:58,920
Speaker 3: Yeah? Famously, Look, Sam is a he thought this should

319
00:16:58,920 --> 00:17:03,040
Speaker 3: be a nationalized technology similar to nuclear power. He thought

320
00:17:03,080 --> 00:17:06,440
Speaker 3: it was that powerful, that dangerous, and that important to

321
00:17:06,480 --> 00:17:10,040
Speaker 3: the national interest and he got no buyers, no bites,

322
00:17:10,160 --> 00:17:12,800
Speaker 3: no interest, and he needed the money, I mean, the

323
00:17:12,880 --> 00:17:15,720
Speaker 3: open AI needed the money in the investment. Under the

324
00:17:15,720 --> 00:17:19,600
Speaker 3: Biden administration, he was very close with Secretary of Commerce

325
00:17:19,640 --> 00:17:23,040
Speaker 3: Jeania Romando. He was very close to the commission that

326
00:17:23,160 --> 00:17:26,920
Speaker 3: worked on the executive order around AI. He wants regulation.

327
00:17:27,400 --> 00:17:30,280
Speaker 3: I think some of that is ideological. I think some

328
00:17:30,359 --> 00:17:34,200
Speaker 3: of it's competitive when you're out in front, Hey, let's

329
00:17:34,200 --> 00:17:36,879
Speaker 3: tap the brakes on everybody else, right, But in the

330
00:17:36,960 --> 00:17:40,000
Speaker 3: last four years he has been an active participant and

331
00:17:40,080 --> 00:17:42,960
Speaker 3: collaborator with the federal government in figuring out what to

332
00:17:42,960 --> 00:17:46,240
Speaker 3: do with artificial intelligence. But he's no dummy. I think

333
00:17:46,240 --> 00:17:48,240
Speaker 3: he's going to take the temperature of the Trump administration

334
00:17:48,440 --> 00:17:53,440
Speaker 3: see where it is, react accordingly. But yeah, historically he's

335
00:17:53,440 --> 00:18:00,480
Speaker 3: been very much in favor of regulation of these models.

336
00:18:04,760 --> 00:18:07,240
Speaker 2: This is the Big Take from Bloomberg News. I'm David Gura.

337
00:18:07,520 --> 00:18:09,919
Speaker 2: This episode was produced by Alex Tie and it was

338
00:18:10,040 --> 00:18:13,080
Speaker 2: edited by our senior producer, Naomi Shavin. It was mixed

339
00:18:13,080 --> 00:18:15,520
Speaker 2: and sound designed by Alex Segura and fact checked by

340
00:18:15,560 --> 00:18:19,760
Speaker 2: Adriana Tapia. Our senior editor is Elizabeth Ponso. Our executive

341
00:18:19,760 --> 00:18:23,000
Speaker 2: producer is Nicole Beamster Boor Sage Bauman is Bloomberg's head

342
00:18:23,040 --> 00:18:25,879
Speaker 2: of podcasts. If you liked this episode, make sure to

343
00:18:25,920 --> 00:18:28,840
Speaker 2: subscribe and review The Big Take wherever you listen to podcasts.

344
00:18:28,880 --> 00:18:31,760
Speaker 2: It helps people find the show. Thanks for listening, we'll

345
00:18:31,760 --> 00:18:32,480
Speaker 2: be back tomorrow