1
00:00:00,520 --> 00:00:03,640
Speaker 1: Already and this is this is the Daily This is

2
00:00:03,680 --> 00:00:06,840
Speaker 1: the Daily OS. Oh, now it makes sense.

3
00:00:14,840 --> 00:00:17,120
Speaker 2: Good morning, and welcome to the Daily OS. It's Friday,

4
00:00:17,239 --> 00:00:19,400
Speaker 2: the twentieth of June. I'm Sam Kazowski.

5
00:00:19,600 --> 00:00:21,480
Speaker 1: I'm Billy Fitzimon's.

6
00:00:21,200 --> 00:00:25,400
Speaker 2: Meta just made its second largest acquisition ever, buying AI

7
00:00:25,440 --> 00:00:29,880
Speaker 2: company Scale Ai for twenty one billion dollars and sparking

8
00:00:29,880 --> 00:00:34,199
Speaker 2: what's being called Silicon Valley's most extreme talent wark. On

9
00:00:34,240 --> 00:00:37,120
Speaker 2: today's podcast, we're going to unpack why tech giants are

10
00:00:37,159 --> 00:00:40,760
Speaker 2: spending hundreds of millions of dollars on AI and offering

11
00:00:40,880 --> 00:00:45,000
Speaker 2: nine figure salaries to law the best minds in artificial intelligence.

12
00:00:48,040 --> 00:00:50,640
Speaker 1: Sam, when you pitched this yesterday, you're basically like, the

13
00:00:50,680 --> 00:00:52,839
Speaker 1: tech bros are fighting and we need to cover it.

14
00:00:52,920 --> 00:00:56,120
Speaker 2: The tech bros are fighting with money. They are really

15
00:00:56,160 --> 00:00:57,120
Speaker 2: fighting with money.

16
00:00:57,280 --> 00:00:59,200
Speaker 1: So I want to understand all of this more. But

17
00:00:59,280 --> 00:01:01,880
Speaker 1: first I want to understand this twenty one billion dollar

18
00:01:02,000 --> 00:01:05,640
Speaker 1: deal that happened last week. So what exactly did Meta buy?

19
00:01:06,040 --> 00:01:08,440
Speaker 2: So Meta bought a forty nine percent stake in this

20
00:01:08,480 --> 00:01:12,200
Speaker 2: company called scale Ai. It was fourteen billion US dollars,

21
00:01:12,240 --> 00:01:15,839
Speaker 2: So twenty one billion. Ossie and scale AI are basically

22
00:01:16,240 --> 00:01:20,679
Speaker 2: a company that specializes in human powered data collection, and

23
00:01:20,959 --> 00:01:23,840
Speaker 2: what that means in reality is they employ thousands of

24
00:01:23,920 --> 00:01:27,280
Speaker 2: humans both in physical workplaces but also lots of remote

25
00:01:27,280 --> 00:01:31,600
Speaker 2: workers to help train machine learning models. And they're the

26
00:01:31,600 --> 00:01:32,960
Speaker 2: systems that power AI.

27
00:01:33,280 --> 00:01:36,679
Speaker 1: So how is that different to other AI companies?

28
00:01:37,280 --> 00:01:41,559
Speaker 2: So they are employed by AI companies to help work

29
00:01:41,600 --> 00:01:46,679
Speaker 2: through information and basically tag the information as a good response,

30
00:01:46,720 --> 00:01:50,880
Speaker 2: a bad response, as particular data being a successful way

31
00:01:50,920 --> 00:01:55,160
Speaker 2: to frame something not successful, basically giving AI feedback. As

32
00:01:55,400 --> 00:01:58,000
Speaker 2: you might say, if you're talking to chat GBT, I

33
00:01:58,040 --> 00:02:01,320
Speaker 2: didn't like that response, Can you go again? These guys

34
00:02:01,360 --> 00:02:04,680
Speaker 2: do it on massive scale, and so they're essentially helping

35
00:02:04,800 --> 00:02:09,000
Speaker 2: AI companies with the quality of the information and their customers.

36
00:02:09,040 --> 00:02:10,520
Speaker 2: To give you a sense of who they're working with,

37
00:02:11,040 --> 00:02:14,079
Speaker 2: their customers include the US Army, Open AI, so they're

38
00:02:14,120 --> 00:02:17,440
Speaker 2: working with what we see at CHATBT pretty much every

39
00:02:17,480 --> 00:02:21,960
Speaker 2: major car brand, so working on internal car technology and meta.

40
00:02:22,040 --> 00:02:25,040
Speaker 2: So this is a case of a customer becoming a

41
00:02:25,080 --> 00:02:28,079
Speaker 2: part owner or nearly a majority owner because they obviously

42
00:02:28,120 --> 00:02:30,560
Speaker 2: see so much value in the product, so.

43
00:02:30,520 --> 00:02:34,280
Speaker 1: They don't necessarily have a product though. It's more their

44
00:02:34,360 --> 00:02:37,560
Speaker 1: employees who are so well trained in AI.

45
00:02:37,919 --> 00:02:41,720
Speaker 2: It's basically a service, so if you plug us into

46
00:02:41,720 --> 00:02:44,680
Speaker 2: your system, will make sure that your data is being

47
00:02:44,760 --> 00:02:47,640
Speaker 2: tagged and labeled, and think of it like almost a

48
00:02:47,840 --> 00:02:50,000
Speaker 2: Gmail inbox. Will make sure that the data is going

49
00:02:50,000 --> 00:02:52,960
Speaker 2: in the right folders so that when your customers you

50
00:02:53,040 --> 00:02:56,440
Speaker 2: and I ask CHATBT or whatever you're using for a response,

51
00:02:56,760 --> 00:02:58,640
Speaker 2: it's got a pretty good idea of the types of

52
00:02:58,680 --> 00:03:03,000
Speaker 2: answers it should be giving. And the founder is considered

53
00:03:03,000 --> 00:03:06,359
Speaker 2: one of the world leaders in AI at the tender

54
00:03:06,400 --> 00:03:10,160
Speaker 2: age of twenty eight. His name's Alexander Wang, and he's

55
00:03:10,160 --> 00:03:14,200
Speaker 2: now going to lead Meta's new AI division. And remember

56
00:03:14,200 --> 00:03:18,920
Speaker 2: this is Meta's second biggest acquisition ever. They own Facebook, Instagram,

57
00:03:19,000 --> 00:03:21,600
Speaker 2: and WhatsApp, so this is a massive investment for them.

58
00:03:21,919 --> 00:03:24,240
Speaker 1: Just a quick side note. When we were looking at

59
00:03:24,280 --> 00:03:27,440
Speaker 1: this yesterday, we came across the very fun fact that

60
00:03:27,560 --> 00:03:30,680
Speaker 1: the founder, Alexander Wang, used to be the roommate of

61
00:03:30,840 --> 00:03:34,679
Speaker 1: the CEO of Open Ai, which owns chat GPT Yeh

62
00:03:34,800 --> 00:03:36,080
Speaker 1: Sama Altman Samultman.

63
00:03:36,080 --> 00:03:37,840
Speaker 2: And he's going to play a part in this discussion

64
00:03:37,880 --> 00:03:41,120
Speaker 2: in a sec But can you imagine the discussions in

65
00:03:41,160 --> 00:03:42,200
Speaker 2: that college dorm.

66
00:03:42,320 --> 00:03:45,560
Speaker 1: So two of the biggest names in AI used to

67
00:03:45,560 --> 00:03:46,200
Speaker 1: be roommates.

68
00:03:46,240 --> 00:03:50,440
Speaker 2: It's phenomenal and it's really interesting to remember that as

69
00:03:50,480 --> 00:03:54,920
Speaker 2: we now discuss the talent war that's coming. Because Sam

70
00:03:54,920 --> 00:03:57,160
Speaker 2: Altman made some pretty big comments on a podcast this

71
00:03:57,200 --> 00:04:00,520
Speaker 2: week that's kind of inspiring the way that we're thinking

72
00:04:00,520 --> 00:04:04,200
Speaker 2: about this. He said that Meta are offering one hundred

73
00:04:04,280 --> 00:04:07,440
Speaker 2: million dollar signing bonuses, so one hundred and fifty four

74
00:04:07,480 --> 00:04:12,080
Speaker 2: million Aussie or thereabouts, to try and poach his company's employees.

75
00:04:12,760 --> 00:04:15,640
Speaker 2: And that's not a typo. One hundred million dollars just

76
00:04:15,720 --> 00:04:17,640
Speaker 2: to sign on and then you get paid.

77
00:04:17,839 --> 00:04:21,880
Speaker 1: That's crazy. So Meta is trying to poach the employees

78
00:04:22,000 --> 00:04:25,120
Speaker 1: of open Ai with upwards of one hundred million dollars.

79
00:04:25,320 --> 00:04:27,680
Speaker 2: Yeah, And to understand this from a Meta perspective, there's

80
00:04:27,680 --> 00:04:29,960
Speaker 2: been a lot of reporting that says that Mark Zuckerberg,

81
00:04:30,000 --> 00:04:32,640
Speaker 2: the founder of Meta, has stepped into what people call

82
00:04:32,720 --> 00:04:35,520
Speaker 2: founder mode, which is essentially he is on the tools

83
00:04:35,640 --> 00:04:38,040
Speaker 2: in the room with this new AI team as the

84
00:04:38,160 --> 00:04:40,200
Speaker 2: key priority for the company.

85
00:04:39,960 --> 00:04:42,040
Speaker 1: Kind of making all of the calls exactly.

86
00:04:42,160 --> 00:04:45,440
Speaker 2: One Meta employee told Reuters this week, Zuckerberg is calling

87
00:04:45,480 --> 00:04:49,039
Speaker 2: researchers directly, flying them to his house. Whatever it takes Wow.

88
00:04:49,720 --> 00:04:53,159
Speaker 1: Let's take a step back here. Why are these companies

89
00:04:53,200 --> 00:04:56,200
Speaker 1: spending such astronomical amounts on AI?

90
00:04:56,480 --> 00:04:59,359
Speaker 2: I think this is a really interesting point that we

91
00:04:59,400 --> 00:05:02,240
Speaker 2: don't spend enough time when we cover AI and actually

92
00:05:02,279 --> 00:05:05,240
Speaker 2: considering is we're talking about hundreds of billions here, hundreds

93
00:05:05,240 --> 00:05:07,040
Speaker 2: of billions of their Why like why is there so

94
00:05:07,120 --> 00:05:10,160
Speaker 2: much money flying around? And I think the costs can

95
00:05:10,200 --> 00:05:12,640
Speaker 2: really be broken down into three main areas. So the

96
00:05:12,680 --> 00:05:18,320
Speaker 2: first is data centers, the physical storerooms, warehouses, fields as

97
00:05:18,360 --> 00:05:20,880
Speaker 2: far as the eye can see of servers, and that's

98
00:05:20,920 --> 00:05:23,960
Speaker 2: a lot of computing power. Microsoft and open Ai have

99
00:05:24,040 --> 00:05:27,200
Speaker 2: announced plans for one hundred billion dollar data center project

100
00:05:27,279 --> 00:05:30,400
Speaker 2: called Stargage for twenty twenty eight, and that's going to

101
00:05:30,440 --> 00:05:34,400
Speaker 2: house thousands of chips that process all the data needed

102
00:05:34,440 --> 00:05:35,880
Speaker 2: to train AI systems.

103
00:05:36,120 --> 00:05:38,240
Speaker 1: Can we pause for a second because this is something

104
00:05:38,320 --> 00:05:42,680
Speaker 1: that always confuses me, that the amount of technology, like

105
00:05:42,839 --> 00:05:46,839
Speaker 1: physical technology that is required to power even just something

106
00:05:46,920 --> 00:05:50,720
Speaker 1: like chap GPT, Like how can we kind of conceptualize

107
00:05:50,720 --> 00:05:51,320
Speaker 1: that You.

108
00:05:51,320 --> 00:05:55,800
Speaker 2: Basically need to think about your laptop having one chip

109
00:05:55,839 --> 00:05:58,960
Speaker 2: in it, and when you put your laptop into a bag,

110
00:05:59,080 --> 00:06:01,560
Speaker 2: it's not does way nothing, it's got a bit of

111
00:06:01,560 --> 00:06:04,840
Speaker 2: weight to it. Now think of a million chips and

112
00:06:05,240 --> 00:06:08,480
Speaker 2: laptops with tiny chips, and servers with tiny chips stacked

113
00:06:08,480 --> 00:06:10,880
Speaker 2: on top of each other. And now think of the

114
00:06:10,960 --> 00:06:13,479
Speaker 2: fact that they can't be directly next to each other

115
00:06:13,520 --> 00:06:15,800
Speaker 2: because they might get too hot, so they need to

116
00:06:15,800 --> 00:06:17,800
Speaker 2: be spaced out a little bit. And now think about

117
00:06:17,800 --> 00:06:20,479
Speaker 2: the fact that as we get more and more ambitious

118
00:06:20,520 --> 00:06:22,360
Speaker 2: with what we want AI to do, win in more

119
00:06:22,400 --> 00:06:25,320
Speaker 2: of them, win in more brain power. There are some places,

120
00:06:25,320 --> 00:06:27,720
Speaker 2: particularly in the US, who are really investing a lot

121
00:06:27,760 --> 00:06:30,520
Speaker 2: into these data centers. There are some places which are

122
00:06:31,080 --> 00:06:33,720
Speaker 2: so big that you need cars to drive around inside

123
00:06:33,760 --> 00:06:37,200
Speaker 2: them that are just for the storage of data chips.

124
00:06:37,800 --> 00:06:40,120
Speaker 2: So it's a massive scale. And of course the other

125
00:06:40,160 --> 00:06:42,679
Speaker 2: part of this, though is the energy consumption, and that's

126
00:06:42,839 --> 00:06:46,400
Speaker 2: actually the second real cost to these companies. There is

127
00:06:46,640 --> 00:06:50,280
Speaker 2: enormous amounts of electricity and water needed. Water is needed

128
00:06:50,279 --> 00:06:54,760
Speaker 2: to cool down the systems. Microsoft's water usage jumped thirty

129
00:06:54,760 --> 00:06:57,560
Speaker 2: four percent between twenty twenty one and twenty twenty two,

130
00:06:57,720 --> 00:07:00,240
Speaker 2: and this was at the beginning of the AI revolution.

131
00:07:00,480 --> 00:07:03,680
Speaker 1: And that's because Microsoft partly owns open ai, which as

132
00:07:03,680 --> 00:07:06,160
Speaker 1: we said, is a maker of chat GPT exactly.

133
00:07:06,200 --> 00:07:08,520
Speaker 2: And to put that stat into perspective, that's two and

134
00:07:08,560 --> 00:07:12,320
Speaker 2: a half thousand Olympic swimming pools, just an increase three

135
00:07:12,400 --> 00:07:15,080
Speaker 2: years ago, So can you imagine what it is today.

136
00:07:15,120 --> 00:07:17,160
Speaker 2: There was one study I was reading from a university

137
00:07:17,200 --> 00:07:20,240
Speaker 2: in Amsterdam that said that AI could require the same

138
00:07:20,240 --> 00:07:23,040
Speaker 2: amount of energy as the entire country of the Netherlands

139
00:07:23,360 --> 00:07:26,240
Speaker 2: by twenty twenty seven. So we're talking almost like there's

140
00:07:26,240 --> 00:07:28,800
Speaker 2: another nation on the planet in terms of carbon emissions.

141
00:07:29,320 --> 00:07:32,080
Speaker 2: And then the third cost is cloud infrastructure. So I've

142
00:07:32,120 --> 00:07:34,520
Speaker 2: taken you through the hardware, the stuff on the ground,

143
00:07:34,680 --> 00:07:37,840
Speaker 2: then the power that's needed to power that stuff, but

144
00:07:37,880 --> 00:07:40,520
Speaker 2: we haven't even talked about the cloud yet. And building

145
00:07:40,800 --> 00:07:44,400
Speaker 2: the systems to deliver AI services to users takes a

146
00:07:44,440 --> 00:07:46,560
Speaker 2: lot of power in creating connections. I mean, you and

147
00:07:46,560 --> 00:07:48,280
Speaker 2: I don't have a data ten to run through our desk.

148
00:07:48,560 --> 00:07:51,520
Speaker 2: We tap into it wirelessly. That takes a lot of

149
00:07:51,560 --> 00:07:53,360
Speaker 2: power and energy and cost as well.

150
00:07:53,440 --> 00:07:56,880
Speaker 1: And so that's why buying something like scale AI, which

151
00:07:56,960 --> 00:07:58,560
Speaker 1: Meta has done, is so expensive.

152
00:07:58,800 --> 00:08:02,360
Speaker 2: Partly, I think there's a huge amount of the what

153
00:08:02,520 --> 00:08:06,320
Speaker 2: the company is required to spend to keep it operational

154
00:08:06,400 --> 00:08:09,520
Speaker 2: that contributes to that price tag. The other part of it, though,

155
00:08:09,640 --> 00:08:13,120
Speaker 2: is the talent and the intellectual property, and that at

156
00:08:13,120 --> 00:08:14,800
Speaker 2: the moment you almost can't put a price on.

157
00:08:15,240 --> 00:08:18,280
Speaker 1: I think something that's interesting to talk about here is

158
00:08:18,360 --> 00:08:21,120
Speaker 1: kind of the historical context of all of this, because

159
00:08:21,320 --> 00:08:24,960
Speaker 1: you know, we're seeing technology kind of revolutions quite a bit,

160
00:08:25,080 --> 00:08:29,120
Speaker 1: especially over the past century. Sure is this different to that?

161
00:08:29,440 --> 00:08:29,520
Speaker 2: Like?

162
00:08:29,600 --> 00:08:32,680
Speaker 1: How important is this in the context of what we

163
00:08:32,760 --> 00:08:35,080
Speaker 1: have seen in terms of advancements of the last one

164
00:08:35,160 --> 00:08:35,800
Speaker 1: hundred years.

165
00:08:35,960 --> 00:08:37,600
Speaker 2: I feel like we're at the stage where it's not

166
00:08:37,679 --> 00:08:41,160
Speaker 2: controversial to say this is the real deal. This is

167
00:08:41,240 --> 00:08:44,200
Speaker 2: really big, And if we just look at the behavior

168
00:08:44,240 --> 00:08:47,640
Speaker 2: of the world's biggest companies as an indicator, then we

169
00:08:47,800 --> 00:08:51,320
Speaker 2: can see that they think AI is the next big shift,

170
00:08:51,400 --> 00:08:53,800
Speaker 2: the next seismic shift in the way that we live

171
00:08:53,840 --> 00:08:57,360
Speaker 2: our lives. And I think what's really interesting about where

172
00:08:57,400 --> 00:09:00,920
Speaker 2: we are in the AI tech world is that there's

173
00:09:00,920 --> 00:09:03,600
Speaker 2: so much investment because we still don't know what AI

174
00:09:03,720 --> 00:09:06,800
Speaker 2: could be capable of. So it's not like there's a

175
00:09:06,840 --> 00:09:09,480
Speaker 2: benchmark that's been set that now everybody's trying to reach

176
00:09:09,520 --> 00:09:11,800
Speaker 2: and spend heaps of money trying to get to it's

177
00:09:12,080 --> 00:09:14,800
Speaker 2: that people are spending money because they don't know what

178
00:09:14,840 --> 00:09:17,280
Speaker 2: they don't know, and that's what's really interesting. There's this

179
00:09:17,360 --> 00:09:22,160
Speaker 2: idea called artificial general intelligence or AGI, and that's basically

180
00:09:22,480 --> 00:09:24,600
Speaker 2: AI that can do the same thing that you and

181
00:09:24,640 --> 00:09:26,520
Speaker 2: I can do. So it can write stuff, and it

182
00:09:26,559 --> 00:09:28,600
Speaker 2: can make decisions and all of that kind of stuff.

183
00:09:28,960 --> 00:09:32,720
Speaker 2: But there's also a thing called artificial superintelligence, which is

184
00:09:33,040 --> 00:09:35,839
Speaker 2: AI that is smarter than you and I. Not smarter

185
00:09:35,880 --> 00:09:38,480
Speaker 2: than you. You're the smartest person ever. AI that's smarter

186
00:09:38,559 --> 00:09:43,160
Speaker 2: than me. And in February, talking about artificial superintelligence, Zuckerberg said,

187
00:09:43,200 --> 00:09:45,680
Speaker 2: this year is going to set the course for the future.

188
00:09:45,960 --> 00:09:48,959
Speaker 2: AI is potentially one of the most important innovations in history.

189
00:09:49,200 --> 00:09:51,240
Speaker 1: So what I'm hearing is they're kind of buying into

190
00:09:51,320 --> 00:09:52,800
Speaker 1: the potential.

191
00:09:52,559 --> 00:09:56,640
Speaker 2: Of it exactly. They're buying into the potential of understanding

192
00:09:56,760 --> 00:10:00,320
Speaker 2: that the moves they make now could dictate their ponomic

193
00:10:00,360 --> 00:10:02,400
Speaker 2: future and how much proper they make for the next

194
00:10:03,120 --> 00:10:03,720
Speaker 2: fifty years.

195
00:10:03,880 --> 00:10:06,520
Speaker 1: Yeah, and they don't know exactly what that potential is,

196
00:10:06,559 --> 00:10:08,440
Speaker 1: but they're pretty sure it's very big.

197
00:10:08,559 --> 00:10:11,040
Speaker 2: Yeah. I've been thinking about the space race, for example,

198
00:10:11,120 --> 00:10:13,040
Speaker 2: Like we know that there was a race to get

199
00:10:13,080 --> 00:10:15,360
Speaker 2: to the moon you could see the moon. Whoever got

200
00:10:15,360 --> 00:10:17,480
Speaker 2: their first would win the race. This is a different

201
00:10:17,520 --> 00:10:20,079
Speaker 2: type of race because nobody actually knows where the finish

202
00:10:20,120 --> 00:10:22,480
Speaker 2: line is here, and so the amount of money being

203
00:10:22,520 --> 00:10:24,160
Speaker 2: spent is not spent on just trying to get to

204
00:10:24,160 --> 00:10:25,760
Speaker 2: that endpoint. It's what's beyond that.

205
00:10:26,480 --> 00:10:28,760
Speaker 1: Sam, I feel like we've only just touched the surface.

206
00:10:28,800 --> 00:10:30,920
Speaker 1: But before we go on, just a quick message from

207
00:10:30,960 --> 00:10:36,280
Speaker 1: our sponsor. Okay, and so I want to go back

208
00:10:36,320 --> 00:10:39,200
Speaker 1: to this talent war that we spoke about, the race

209
00:10:39,320 --> 00:10:41,319
Speaker 1: to get people in the room who are the best

210
00:10:41,320 --> 00:10:44,840
Speaker 1: and brightest are working out the capabilities of AI in

211
00:10:44,880 --> 00:10:47,280
Speaker 1: real time. Sam, I'm most surprised that no one has

212
00:10:47,320 --> 00:10:48,600
Speaker 1: tapped you on the shoulder yet.

213
00:10:48,400 --> 00:10:48,960
Speaker 2: But I'm sure.

214
00:10:49,840 --> 00:10:52,720
Speaker 1: I'm sure it's coming. Mark Zuckerberg just a call away.

215
00:10:52,760 --> 00:10:54,959
Speaker 2: I'm sure when I talk you through these salaries, you're

216
00:10:54,960 --> 00:10:56,480
Speaker 2: going to understand why I would do that.

217
00:10:57,000 --> 00:11:00,720
Speaker 1: I'm sure you would. How extreme are we talking.

218
00:11:00,880 --> 00:11:03,440
Speaker 2: We're talking nine figures, so we are talking salaries in

219
00:11:03,480 --> 00:11:07,679
Speaker 2: the hundreds of millions in US dollars. So I mean,

220
00:11:07,960 --> 00:11:09,400
Speaker 2: if you're at the top of the game in the

221
00:11:09,400 --> 00:11:11,680
Speaker 2: AI space, you could easily be looking at the salary

222
00:11:11,679 --> 00:11:14,040
Speaker 2: of four to five hundred million dollars a year.

223
00:11:14,360 --> 00:11:16,280
Speaker 1: Just so hard to comprehend.

224
00:11:16,320 --> 00:11:18,600
Speaker 2: That so hard to comprehend, especially with a sign on

225
00:11:18,640 --> 00:11:20,320
Speaker 2: bonus of one hundred million that I talked to you

226
00:11:20,320 --> 00:11:23,040
Speaker 2: about earlier. There was a line, a great line in

227
00:11:23,080 --> 00:11:25,880
Speaker 2: an article I was reading from a tech recruiter. He said,

228
00:11:25,920 --> 00:11:29,960
Speaker 2: we're seeing signing bonuses that exceed what CEOs made a

229
00:11:30,000 --> 00:11:32,240
Speaker 2: decade ago, and this is just the beginning of all

230
00:11:32,280 --> 00:11:35,480
Speaker 2: of this. There was another researcher from Stanford who said,

231
00:11:35,480 --> 00:11:38,280
Speaker 2: the key to winning isn't just computing power, it's the

232
00:11:38,320 --> 00:11:40,880
Speaker 2: people who know how to use it. And I think

233
00:11:40,920 --> 00:11:42,840
Speaker 2: this is the key to this talent war. It's that

234
00:11:43,120 --> 00:11:47,640
Speaker 2: anyone could buy warehouses of massive machinery and ships, and

235
00:11:47,679 --> 00:11:50,320
Speaker 2: by anyone, I mean anyone with that sort of money

236
00:11:50,320 --> 00:11:53,000
Speaker 2: at their disposal. But the people who know how to

237
00:11:53,000 --> 00:11:56,160
Speaker 2: harness that and quick they're going to be the really

238
00:11:56,240 --> 00:11:59,640
Speaker 2: valuable assets in Silicon Valley. And most of those people

239
00:11:59,640 --> 00:12:02,200
Speaker 2: already have jobs in AI. And that's why you're seeing

240
00:12:02,200 --> 00:12:05,160
Speaker 2: this bidding war with talent trying to drag people from

241
00:12:05,200 --> 00:12:06,000
Speaker 2: one place to another.

242
00:12:06,200 --> 00:12:08,360
Speaker 1: And so when I mentioned at the start that you

243
00:12:08,440 --> 00:12:11,000
Speaker 1: kind of pitch this as the tech bros are fighting.

244
00:12:11,040 --> 00:12:13,680
Speaker 1: You also mentioned that Sam Oltman had some things to

245
00:12:13,679 --> 00:12:17,160
Speaker 1: say about Meta kind of stealing or poaching his employees.

246
00:12:17,559 --> 00:12:20,439
Speaker 1: What exactly did Sam Oltman say about this and about

247
00:12:20,440 --> 00:12:21,520
Speaker 1: the sign on bonuses?

248
00:12:21,920 --> 00:12:23,959
Speaker 2: Well, that's the interesting bit we learned from Sam mommon

249
00:12:24,000 --> 00:12:25,880
Speaker 2: this week was we didn't know about these one hundred

250
00:12:25,920 --> 00:12:28,720
Speaker 2: million dollars sign on bonuses before an interview that Sam

251
00:12:28,760 --> 00:12:32,040
Speaker 2: Oltman gave on a podcast, and he didn't hold back.

252
00:12:32,120 --> 00:12:34,800
Speaker 2: He called these offers crazy. He said that Meta was

253
00:12:34,840 --> 00:12:39,280
Speaker 2: offering not just these signing bonuses, but these hundreds of

254
00:12:39,280 --> 00:12:42,679
Speaker 2: millions of dollars worth of salary. And the really interesting

255
00:12:42,720 --> 00:12:46,079
Speaker 2: part is that Oldman also said that despite these massive offers,

256
00:12:46,559 --> 00:12:49,360
Speaker 2: none of open AI's best people have taken Meta up

257
00:12:49,400 --> 00:12:51,120
Speaker 2: on them, at least not yet.

258
00:12:51,240 --> 00:12:53,600
Speaker 1: Why would they not accept those offers?

259
00:12:53,800 --> 00:12:56,280
Speaker 2: Well, according to Oltman, he said that it's about culture

260
00:12:56,320 --> 00:12:57,840
Speaker 2: and mission and you can start to see a bit

261
00:12:57,880 --> 00:13:01,079
Speaker 2: of competition really emerging between these tech As he said,

262
00:13:01,200 --> 00:13:03,960
Speaker 2: and I'm paraphrasing here, that when a company focuses so

263
00:13:04,040 --> 00:13:08,320
Speaker 2: heavily on upfront guaranteed compensation rather than the work and

264
00:13:08,360 --> 00:13:11,120
Speaker 2: the mission and the growth of the company. It doesn't

265
00:13:11,120 --> 00:13:14,240
Speaker 2: set up a great culture. Everyone's kind of there, according

266
00:13:14,320 --> 00:13:17,240
Speaker 2: to Altman, because they got a very sweet package to

267
00:13:17,400 --> 00:13:19,640
Speaker 2: land there, and when they're there, they don't really want

268
00:13:19,640 --> 00:13:22,480
Speaker 2: to be there. And he took a direct shot at

269
00:13:22,520 --> 00:13:25,800
Speaker 2: Meta's ability to innovate. He said that metas push into

270
00:13:25,840 --> 00:13:28,400
Speaker 2: AI is the same as when Google tried to push

271
00:13:28,400 --> 00:13:31,320
Speaker 2: into social media a decade or so ago, and it

272
00:13:31,360 --> 00:13:33,959
Speaker 2: was clear to people that Google weren't really going to

273
00:13:34,120 --> 00:13:36,880
Speaker 2: compete with an Instagram or a Facebook. And he said

274
00:13:36,880 --> 00:13:39,560
Speaker 2: that it feels a bit similar here with metas ai efforts,

275
00:13:39,600 --> 00:13:41,520
Speaker 2: that they're kind of in the realm that they don't

276
00:13:41,559 --> 00:13:43,920
Speaker 2: serve to be in, which is spicy.

277
00:13:44,040 --> 00:13:46,840
Speaker 1: That is quite a big claiming and quite a big criticism.

278
00:13:47,160 --> 00:13:50,440
Speaker 2: Yeah, and it's particularly a big criticism to give to

279
00:13:50,480 --> 00:13:54,160
Speaker 2: somebody who's offering your employees hundreds of millions of dollars

280
00:13:54,600 --> 00:13:57,560
Speaker 2: to go over there. But Altman raised the possibility that

281
00:13:57,679 --> 00:14:00,640
Speaker 2: even if they do join Meta, and even if they

282
00:14:00,679 --> 00:14:05,960
Speaker 2: do try and achieve this superintelligence that Mark Zuckerberg's talking about,

283
00:14:06,600 --> 00:14:09,120
Speaker 2: we aren't going to see AI change the world as

284
00:14:09,200 --> 00:14:11,400
Speaker 2: much as we think it could. I thought that was

285
00:14:11,440 --> 00:14:14,840
Speaker 2: so interesting from the head of open AI. He said

286
00:14:14,880 --> 00:14:16,840
Speaker 2: that people will live their lives pretty much in the

287
00:14:16,840 --> 00:14:19,480
Speaker 2: same way in two years time. But when we start

288
00:14:19,520 --> 00:14:23,600
Speaker 2: to have AI discovering new science, which is about five

289
00:14:23,640 --> 00:14:25,920
Speaker 2: to ten years away, that's when things could really change.

290
00:14:26,000 --> 00:14:28,440
Speaker 1: What does that mean discovering new science?

291
00:14:28,640 --> 00:14:31,880
Speaker 2: So right now, scientific discoveries are made by human researchers

292
00:14:32,000 --> 00:14:35,360
Speaker 2: who run experiments in a lab and form hypotheses, they

293
00:14:35,480 --> 00:14:38,240
Speaker 2: analyze results. They're using AI at the moment to analyze

294
00:14:38,320 --> 00:14:41,200
Speaker 2: results and to speed up their processes. But what Oltman

295
00:14:41,280 --> 00:14:45,560
Speaker 2: is suggesting is that AI could actually make scientific discoveries itself.

296
00:14:45,880 --> 00:14:47,640
Speaker 2: So if you think about it this way, AI could

297
00:14:47,640 --> 00:14:51,760
Speaker 2: potentially analyze all the data, spot patterns that humans can

298
00:14:51,840 --> 00:14:56,280
Speaker 2: miss and propose entirely new theories which could lead to

299
00:14:56,640 --> 00:15:00,480
Speaker 2: new materials for batteries, or new compounds that go into drugs,

300
00:15:00,880 --> 00:15:04,920
Speaker 2: or new physics theories that humans haven't thought of yet.

301
00:15:05,240 --> 00:15:08,600
Speaker 2: And there are examples of this. We've seen AI discover

302
00:15:08,800 --> 00:15:13,760
Speaker 2: new antibiotics. We've seen deep minds AI predict protein structures

303
00:15:13,800 --> 00:15:17,760
Speaker 2: that had stumped scientists for decades. So if AI is

304
00:15:17,840 --> 00:15:20,320
Speaker 2: starting to actually do the research itself and form its

305
00:15:20,320 --> 00:15:22,520
Speaker 2: own hypotheses, that's an entirely new wave.

306
00:15:22,920 --> 00:15:24,880
Speaker 1: I feel like I've said so many times this podcast

307
00:15:24,960 --> 00:15:27,560
Speaker 1: already that it is kind of hard to wrap your

308
00:15:27,680 --> 00:15:30,480
Speaker 1: head around and to kind of understand what it all means.

309
00:15:31,120 --> 00:15:35,680
Speaker 1: Are companies doing anything else besides throwing money at individual researchers.

310
00:15:36,040 --> 00:15:38,720
Speaker 2: Well, the interesting thing about the scale ai acquisition this

311
00:15:38,760 --> 00:15:41,320
Speaker 2: week by Meta is that clearly if they can't get

312
00:15:41,640 --> 00:15:45,480
Speaker 2: the person to come over and lead their AI project

313
00:15:45,520 --> 00:15:47,680
Speaker 2: and AI innovation, they're just going to buy the company

314
00:15:48,120 --> 00:15:51,760
Speaker 2: and bring everybody over. And so we've seen that with Microsoft,

315
00:15:51,800 --> 00:15:54,200
Speaker 2: as you mentioned earlier, they invested in open Ai to

316
00:15:54,240 --> 00:15:57,080
Speaker 2: get a seat at the table. Google acquired DeepMind for

317
00:15:57,120 --> 00:15:59,840
Speaker 2: five hundred million dollars, which seems kind of cheap in

318
00:16:00,080 --> 00:16:02,400
Speaker 2: the scheme of what we're talking about today. But we're

319
00:16:02,440 --> 00:16:03,840
Speaker 2: going to see more of this. We're going to see

320
00:16:03,840 --> 00:16:07,240
Speaker 2: more companies getting acquired when there's an individual at the

321
00:16:07,240 --> 00:16:08,760
Speaker 2: heart of the company they really want.

322
00:16:09,160 --> 00:16:12,840
Speaker 1: You mentioned Google just then, and I also saw that

323
00:16:12,960 --> 00:16:15,960
Speaker 1: Google was one of the clients of Scale Ai, And

324
00:16:16,040 --> 00:16:19,720
Speaker 1: after this acquisition by Meta was announced, Google then dropped

325
00:16:19,960 --> 00:16:22,560
Speaker 1: Scale ai also they will no longer work with them.

326
00:16:22,600 --> 00:16:24,520
Speaker 2: You've got to remember there's so much data and there's

327
00:16:24,520 --> 00:16:27,000
Speaker 2: so much under the hood information that's plugged into all

328
00:16:27,040 --> 00:16:30,000
Speaker 2: of these systems, and if a competitor buys something that

329
00:16:30,040 --> 00:16:31,600
Speaker 2: you're using, I don't think you're going to use it

330
00:16:31,640 --> 00:16:32,480
Speaker 2: for very much longer.

331
00:16:32,720 --> 00:16:34,280
Speaker 1: Where does this go from.

332
00:16:34,120 --> 00:16:37,760
Speaker 2: Here, Well, industry experts say that this talent war, these

333
00:16:37,800 --> 00:16:41,240
Speaker 2: sign on bonuses, these salaries are just going to get bigger.

334
00:16:41,480 --> 00:16:43,120
Speaker 2: I mean, a lot of the commentary I was reading

335
00:16:43,200 --> 00:16:45,040
Speaker 2: was saying that we are not too far away from

336
00:16:45,080 --> 00:16:49,840
Speaker 2: billion dollar salaries in the AI space, and the pool

337
00:16:49,880 --> 00:16:53,000
Speaker 2: of experts will grow as well. But so all the

338
00:16:53,040 --> 00:16:56,000
Speaker 2: demands and what's at stake for the companies to make

339
00:16:56,040 --> 00:16:59,160
Speaker 2: money from all of this. And the other thing to

340
00:16:59,200 --> 00:17:02,480
Speaker 2: remember is that the companies have placed their bets. They've

341
00:17:02,480 --> 00:17:05,040
Speaker 2: gone all in on AI. They've gone all in on

342
00:17:05,160 --> 00:17:08,520
Speaker 2: these innovations. They've told shareholders, they've told customers that this

343
00:17:08,600 --> 00:17:11,120
Speaker 2: is the next big thing, and so now they're almost

344
00:17:11,400 --> 00:17:14,520
Speaker 2: too far away from the start line to go backwards,

345
00:17:14,880 --> 00:17:16,879
Speaker 2: but they can't quite see the finish line yet, and

346
00:17:16,920 --> 00:17:19,919
Speaker 2: they just have no choice but to spend billions and

347
00:17:19,960 --> 00:17:21,320
Speaker 2: billions trying to push forward.

348
00:17:21,119 --> 00:17:24,119
Speaker 1: As fast as possible just before we end. I do

349
00:17:24,240 --> 00:17:27,480
Speaker 1: think it is worth mentioning the impact that this has

350
00:17:27,640 --> 00:17:32,040
Speaker 1: on employees at these companies that aren't in the AI space.

351
00:17:32,400 --> 00:17:35,800
Speaker 1: We saw this week that Amazon has announced redundancies for

352
00:17:35,960 --> 00:17:39,080
Speaker 1: people who aren't in the AI space, and I think

353
00:17:39,119 --> 00:17:42,199
Speaker 1: that's a really important thing to acknowledge as well in

354
00:17:42,320 --> 00:17:43,400
Speaker 1: all of these conversations.

355
00:17:43,400 --> 00:17:47,480
Speaker 2: Definitely, and there's a law floating around in amongst US

356
00:17:47,640 --> 00:17:53,040
Speaker 2: legislators proposing that employers have to declare when AI is

357
00:17:53,080 --> 00:17:57,679
Speaker 2: replacing the jobs of humans in any redundancy announcements that

358
00:17:57,720 --> 00:18:00,680
Speaker 2: they make, So you're right, this is a major as well.

359
00:18:00,880 --> 00:18:03,760
Speaker 1: Sam. Thanks for coming on and explaining to us your

360
00:18:03,800 --> 00:18:04,480
Speaker 1: favorite topic.

361
00:18:04,600 --> 00:18:07,080
Speaker 2: Thanks so much for having me what an indulgence.

362
00:18:07,320 --> 00:18:09,800
Speaker 1: And thank you so much for listening to this episode

363
00:18:09,840 --> 00:18:12,280
Speaker 1: of The Daily Os. We'll be back this afternoon with

364
00:18:12,359 --> 00:18:15,360
Speaker 1: your evening headlines, but until then, have a great day,

365
00:18:15,440 --> 00:18:20,880
Speaker 1: See you later. My name is Lily Madden and I'm

366
00:18:20,880 --> 00:18:25,359
Speaker 1: a proud Arunda Bungelung Calcuttin woman from Gadighl Country. The

367
00:18:25,440 --> 00:18:28,560
Speaker 1: Daily oz acknowledges that this podcast is recorded on the

368
00:18:28,600 --> 00:18:31,280
Speaker 1: lands of the Gadighl people and pays respect to all

369
00:18:31,440 --> 00:18:34,520
Speaker 1: Aboriginal and Torres Strait Island and nations. We pay our

370
00:18:34,520 --> 00:18:37,680
Speaker 1: respects to the first peoples of these countries, both past

371
00:18:37,760 --> 00:18:38,280
Speaker 1: and present,