1
00:00:04,800 --> 00:00:07,720
Speaker 1: On this episode of a News World. From boardrooms to

2
00:00:07,840 --> 00:00:11,280
Speaker 1: dorm rooms, AI seems to be what everyone is talking about,

3
00:00:11,680 --> 00:00:15,520
Speaker 1: from the promise of Chat GPT to robots that may

4
00:00:15,840 --> 00:00:19,160
Speaker 1: or may not take your job. In his new book

5
00:00:19,840 --> 00:00:25,079
Speaker 1: AI Valley, Microsoft, Google, and the trillion dollar Race to

6
00:00:25,200 --> 00:00:29,479
Speaker 1: cash in an Artificial intelligence, Pulitzer Prize winning journalist Gary

7
00:00:29,520 --> 00:00:33,960
Speaker 1: Rivlin follows the launch of Chat, GPT and three startups,

8
00:00:34,400 --> 00:00:36,800
Speaker 1: all with big dreams of cashing.

9
00:00:36,520 --> 00:00:37,120
Speaker 2: In on AI.

10
00:00:37,800 --> 00:00:40,760
Speaker 1: But it's not long before the tech giants enter the

11
00:00:40,760 --> 00:00:44,919
Speaker 1: AI space. Rivlin lays out the fascinating history of AI's evolution,

12
00:00:45,560 --> 00:00:48,839
Speaker 1: the breakthroughs and wrong turns, and the major players in

13
00:00:48,880 --> 00:00:52,640
Speaker 1: Silicon Valley, the developers and investors.

14
00:00:52,200 --> 00:00:53,640
Speaker 2: Who will lead the future of AI.

15
00:00:54,520 --> 00:00:57,800
Speaker 1: Here to discuss his new book, I am really pleased

16
00:00:57,840 --> 00:01:01,000
Speaker 1: to welcome my guest, Gary Rivlin. He is a Pulitzer

17
00:01:01,000 --> 00:01:04,800
Speaker 1: Prize winning investigative reporter who has been writing about technology

18
00:01:05,080 --> 00:01:08,200
Speaker 1: since the mid nineteen nineties and the rise of the Internet.

19
00:01:08,680 --> 00:01:12,480
Speaker 1: He is the author of ten previous books, including Saving

20
00:01:12,560 --> 00:01:17,000
Speaker 1: Main Street and Katrina After the Flood. His work has

21
00:01:17,000 --> 00:01:21,200
Speaker 1: appeared in The New York Times, Newsweek, Fortune, GQ, and Wired,

22
00:01:21,440 --> 00:01:34,480
Speaker 1: among other publications. Gary, that's an amazing record. Thank you

23
00:01:34,560 --> 00:01:35,240
Speaker 1: for joining us.

24
00:01:35,800 --> 00:01:36,960
Speaker 3: Thank you my pleasure.

25
00:01:37,440 --> 00:01:41,039
Speaker 1: So you've been covering tax since the nineties. How has

26
00:01:41,319 --> 00:01:44,520
Speaker 1: tech evolved since you first started covering it and what

27
00:01:44,640 --> 00:01:48,600
Speaker 1: has surprised you most about how Silicon Valley has evolved.

28
00:01:49,720 --> 00:01:55,120
Speaker 3: Well, the short answer is the dominance of the giants.

29
00:01:55,320 --> 00:01:58,800
Speaker 3: In fact, for this book, I went looking since the

30
00:01:58,880 --> 00:02:01,680
Speaker 3: end of twenty twenty two and I realized, like, this

31
00:02:01,800 --> 00:02:04,480
Speaker 3: is now the AI moment. They turned back to tech

32
00:02:04,680 --> 00:02:07,680
Speaker 3: and I was looking for what would be the new Google,

33
00:02:07,800 --> 00:02:10,000
Speaker 3: what would be the new Facebook? And it turns out

34
00:02:10,040 --> 00:02:12,560
Speaker 3: the new Google is Google, the new Facebook is Facebook.

35
00:02:13,040 --> 00:02:14,880
Speaker 3: And so you know, kind of dating back to the

36
00:02:14,960 --> 00:02:17,440
Speaker 3: mid nineteen nineties, it was all about startups. It's all

37
00:02:17,480 --> 00:02:21,560
Speaker 3: about these companies founded in a dorm room, someone's garage.

38
00:02:21,800 --> 00:02:23,560
Speaker 3: They have a great idea, they raise a little bit

39
00:02:23,600 --> 00:02:27,760
Speaker 3: of money, they get some traction, and they become Google. Facebook.

40
00:02:27,800 --> 00:02:30,600
Speaker 3: But AI is different. I fear it's going to solidify

41
00:02:30,639 --> 00:02:33,480
Speaker 3: the power of big tech rather than open it up

42
00:02:33,520 --> 00:02:35,480
Speaker 3: to another set of players.

43
00:02:35,800 --> 00:02:39,839
Speaker 1: Are there characteristics of the investment you have to make

44
00:02:40,520 --> 00:02:44,400
Speaker 1: that make AI susceptible to that kind of dominance money?

45
00:02:44,520 --> 00:02:47,440
Speaker 3: It's just so expensive. In the old days, you can

46
00:02:47,840 --> 00:02:51,080
Speaker 3: raise a million, a few million dollars and start to

47
00:02:51,080 --> 00:02:54,160
Speaker 3: get traction. Then you need to raise the big money

48
00:02:54,160 --> 00:02:59,239
Speaker 3: to go national, go global, whatever. But AI training these models,

49
00:02:59,280 --> 00:03:03,360
Speaker 3: this general AI where they can talk to you, spit

50
00:03:03,400 --> 00:03:08,440
Speaker 3: out images, make video. It's so expensive to train these things.

51
00:03:08,440 --> 00:03:10,840
Speaker 3: It's so expensive to fine tune these things. It's so

52
00:03:10,880 --> 00:03:15,120
Speaker 3: expensive to operate these things. Is beyond the means of

53
00:03:15,760 --> 00:03:18,560
Speaker 3: most startups. But I first started reporting on this at

54
00:03:18,600 --> 00:03:22,359
Speaker 3: the start of twenty twenty three, it would be millions,

55
00:03:22,400 --> 00:03:26,960
Speaker 3: maybe ten million to train and fine tune one of

56
00:03:27,000 --> 00:03:30,000
Speaker 3: these models before they released. By the time I was

57
00:03:30,040 --> 00:03:32,799
Speaker 3: done reporting, it was one hundred million. And now it's

58
00:03:32,919 --> 00:03:36,000
Speaker 3: billions of dollars. What's startup? I mean, there's open AI,

59
00:03:36,120 --> 00:03:38,680
Speaker 3: there's a couple others, but very few startups could raise

60
00:03:38,720 --> 00:03:41,280
Speaker 3: that kind of money. And it's still not enough. I mean,

61
00:03:41,360 --> 00:03:44,080
Speaker 3: these startups are still losing money, so they have to

62
00:03:44,160 --> 00:03:47,280
Speaker 3: raise billions, tens of billions, perhaps in the future, not

63
00:03:47,520 --> 00:03:50,920
Speaker 3: so distant future, one hundred billion dollars and more. And

64
00:03:51,320 --> 00:03:54,160
Speaker 3: big tech can afford that. You know, Microsoft, Apple, They're

65
00:03:54,160 --> 00:03:57,280
Speaker 3: sitting on one hundred billion dollars or so in their savings.

66
00:03:57,560 --> 00:04:00,160
Speaker 3: But what chance does a startup have to do of

67
00:04:00,160 --> 00:04:04,080
Speaker 3: these cutting edge foundational models, the chatbots in the like.

68
00:04:04,880 --> 00:04:08,960
Speaker 1: When you look at the biggest companies Microsoft, Google, Meta,

69
00:04:09,680 --> 00:04:10,880
Speaker 1: and Amazon.

70
00:04:11,080 --> 00:04:13,920
Speaker 2: That the share of volume of cash that they create.

71
00:04:14,480 --> 00:04:17,160
Speaker 1: Yeah, to me, it's astonishing that, with the exception of

72
00:04:17,200 --> 00:04:20,680
Speaker 1: a couple companies in China and Saudi Aramco, all of

73
00:04:20,680 --> 00:04:22,760
Speaker 1: the trillion dollar companies in the world are American.

74
00:04:23,279 --> 00:04:27,000
Speaker 3: It's astonishing. Let's just focus on Search is arguably the

75
00:04:27,040 --> 00:04:30,760
Speaker 3: best business ever. Once you've built the infrastructure, there's not

76
00:04:30,880 --> 00:04:35,080
Speaker 3: much marginal costs to add new users, and so Google

77
00:04:35,200 --> 00:04:37,520
Speaker 3: is making over one hundred billion, something like one hundred

78
00:04:37,520 --> 00:04:41,040
Speaker 3: and fifty billion dollars a year in profits just from search.

79
00:04:41,600 --> 00:04:44,960
Speaker 3: Microsoft they've been struggling to get into search with being

80
00:04:45,320 --> 00:04:48,320
Speaker 3: and they would get like three four five percentage points,

81
00:04:48,360 --> 00:04:51,320
Speaker 3: which is nothing except for it would still translate to

82
00:04:51,640 --> 00:04:54,320
Speaker 3: hundreds and hundreds and hundreds of millions of dollars of revenue.

83
00:04:54,320 --> 00:04:57,440
Speaker 3: And you can make the same argument with Facebook Meta.

84
00:04:57,520 --> 00:05:01,200
Speaker 3: In the old days, the rough statistic is newspapers, magazines,

85
00:05:01,240 --> 00:05:06,520
Speaker 3: publications brought in like fifty billion dollars in advertising revenue collectively.

86
00:05:06,600 --> 00:05:10,239
Speaker 3: That was a circle like two thousand. Nowadays they're bringing

87
00:05:10,279 --> 00:05:13,040
Speaker 3: in closer to ten billion dollars and most of the

88
00:05:13,080 --> 00:05:17,160
Speaker 3: rest is being divvied up by Google and Meta, and

89
00:05:17,240 --> 00:05:20,200
Speaker 3: so it's kind of an idea of winner takes most

90
00:05:20,279 --> 00:05:22,360
Speaker 3: and the stakes are so big that they're just making

91
00:05:22,520 --> 00:05:25,920
Speaker 3: so much money. But you know, this dates back to Microsoft.

92
00:05:25,920 --> 00:05:29,599
Speaker 3: I mean Microsoft, which actually sells product, right, I mean

93
00:05:30,040 --> 00:05:33,960
Speaker 3: you buy your Windows operating system and software package, office

94
00:05:34,000 --> 00:05:38,800
Speaker 3: software packages. You know, their profits were astonishing in the

95
00:05:38,880 --> 00:05:41,400
Speaker 3: nineteen nineties, and I think it just builds on itself.

96
00:05:41,400 --> 00:05:46,000
Speaker 3: They're so rich they can afford to invest in AI,

97
00:05:46,279 --> 00:05:50,960
Speaker 3: gener of AI, these leading edge technologies, and sometimes bigness

98
00:05:51,040 --> 00:05:53,240
Speaker 3: is a weakness right there. You know, they kind of

99
00:05:53,240 --> 00:05:55,440
Speaker 3: trip over their own feet. Google was so far ahead

100
00:05:55,480 --> 00:05:58,400
Speaker 3: of everyone with machine learning, ditting back to the twenty tens,

101
00:05:58,520 --> 00:06:00,880
Speaker 3: but of course it was open AI to a chat GBT.

102
00:06:01,120 --> 00:06:04,240
Speaker 3: You know, big companies are scared the innovator's dilemma. They

103
00:06:04,240 --> 00:06:07,880
Speaker 3: don't want to threaten their existing honeypot. And also, you know,

104
00:06:07,920 --> 00:06:11,680
Speaker 3: I mean startups have advantages, but the advantage of money

105
00:06:12,080 --> 00:06:15,719
Speaker 3: in AI makes me worry that it's kind of game over.

106
00:06:16,320 --> 00:06:19,320
Speaker 1: In the case of Google, you're doing all the work.

107
00:06:20,279 --> 00:06:22,600
Speaker 1: They build a framework within which you get to come,

108
00:06:22,920 --> 00:06:25,440
Speaker 1: you get to play. They don't have to pay anybody.

109
00:06:25,520 --> 00:06:27,880
Speaker 1: These people are all coming and saying, please let me

110
00:06:27,960 --> 00:06:29,000
Speaker 1: come and use your.

111
00:06:28,880 --> 00:06:34,159
Speaker 3: Material exactly Facebook, Instagram, we could go on listing, Twitter X.

112
00:06:34,279 --> 00:06:36,960
Speaker 3: There's an expression in Silicon Valley if you're not paying

113
00:06:37,000 --> 00:06:39,839
Speaker 3: for the product, you are the product. And that's a

114
00:06:39,880 --> 00:06:43,560
Speaker 3: perfect way of understanding a Google or a Facebook. Look,

115
00:06:43,640 --> 00:06:46,800
Speaker 3: let's use Google. So this phone in your pocket tracks

116
00:06:46,839 --> 00:06:50,200
Speaker 3: you everywhere you do your searches online. They track that,

117
00:06:50,600 --> 00:06:52,880
Speaker 3: They bundle up the data, and they sell it to

118
00:06:52,880 --> 00:06:55,280
Speaker 3: the highest bidder. It's a great business. It makes a

119
00:06:55,279 --> 00:06:58,279
Speaker 3: lot of money for them. But I think citizens only

120
00:06:58,680 --> 00:07:01,920
Speaker 3: slowly woke up to that fact. And again this gets

121
00:07:01,920 --> 00:07:04,599
Speaker 3: me back to my worry that if the same big

122
00:07:04,720 --> 00:07:08,600
Speaker 3: tech companies, the same few companies are going to dominate

123
00:07:08,640 --> 00:07:11,400
Speaker 3: AI the way they've dominated the last bunch of years.

124
00:07:11,560 --> 00:07:14,880
Speaker 3: We don't trust them as the stewards for technology, and

125
00:07:15,000 --> 00:07:19,040
Speaker 3: in fact AI is a powerful technology. AI is going

126
00:07:19,120 --> 00:07:25,360
Speaker 3: to rely on our information. There's privacy concern. So that

127
00:07:25,560 --> 00:07:29,640
Speaker 3: is my concern that these companies that are proven untrustworthy

128
00:07:30,000 --> 00:07:32,880
Speaker 3: are going to be the ones bringing us this amazing power.

129
00:07:32,960 --> 00:07:36,280
Speaker 3: I'm optimistic about AI. I lean optimistic. I think AI

130
00:07:36,320 --> 00:07:43,200
Speaker 3: could bring incredible things around education, scientific breakthroughs, medicine. My

131
00:07:43,320 --> 00:07:46,080
Speaker 3: worry is it's in the hands of companies that show

132
00:07:46,120 --> 00:07:50,000
Speaker 3: that it's all about profits and not about trust and safety,

133
00:07:50,000 --> 00:07:52,440
Speaker 3: which again I think is essential for AI.

134
00:07:53,000 --> 00:07:53,760
Speaker 2: Let me do all his thinction.

135
00:07:54,000 --> 00:07:57,480
Speaker 1: There was a period there where people were coming up

136
00:07:57,480 --> 00:08:02,240
Speaker 1: with pretty cool innovations and then promptly getting bought, so

137
00:08:02,720 --> 00:08:05,200
Speaker 1: they never actually a chance to grow into a competitor

138
00:08:05,720 --> 00:08:09,080
Speaker 1: because they were acquired by these large companies and who

139
00:08:09,160 --> 00:08:13,760
Speaker 1: absorbed them. Now, if I understand you correctly, it's ructually

140
00:08:13,840 --> 00:08:17,640
Speaker 1: impossible for a startup in the AI field because of

141
00:08:17,640 --> 00:08:20,600
Speaker 1: the scale of resources it takes to create it. So

142
00:08:20,720 --> 00:08:24,320
Speaker 1: virtually all of the next level of innovation in terms

143
00:08:24,360 --> 00:08:27,400
Speaker 1: of small companies is going to be the use of AI,

144
00:08:27,920 --> 00:08:30,240
Speaker 1: which will be provided by one of the big companies.

145
00:08:30,800 --> 00:08:34,240
Speaker 3: Let me break the startup world into two general categories.

146
00:08:34,679 --> 00:08:38,880
Speaker 3: There's still going to be plenty opportunities for founders to

147
00:08:39,040 --> 00:08:42,480
Speaker 3: raise some money and have a good return on an investment.

148
00:08:42,600 --> 00:08:46,600
Speaker 3: For some app like AI will automatically fill out your

149
00:08:46,600 --> 00:08:49,360
Speaker 3: expense sheets and do most if not all, of the

150
00:08:49,400 --> 00:08:51,560
Speaker 3: work for you. You could see that be very valuable.

151
00:08:51,760 --> 00:08:55,640
Speaker 3: And then there'll still be in quote little companies, they'll

152
00:08:55,679 --> 00:08:58,640
Speaker 3: have millions of users bringing tens hundreds of millions of dollars.

153
00:08:58,880 --> 00:09:03,000
Speaker 3: But I'm really focused on that second category. I have

154
00:09:03,160 --> 00:09:06,200
Speaker 3: a billion plus users, I have a market cap, a

155
00:09:06,280 --> 00:09:10,400
Speaker 3: paper worth of over a trillion dollars. Doing the foundational models,

156
00:09:10,520 --> 00:09:14,040
Speaker 3: doing the stuff that's underneath everything, this stuff right at

157
00:09:14,080 --> 00:09:17,960
Speaker 3: the center of it. They're training the models, they're operating

158
00:09:17,960 --> 00:09:21,600
Speaker 3: the models that these other smaller companies apps. They could

159
00:09:21,600 --> 00:09:23,880
Speaker 3: be for business, they could be for individuals, or be

160
00:09:23,960 --> 00:09:28,120
Speaker 3: AI therapists, they'll be AI life coaches. I mean, there's

161
00:09:28,280 --> 00:09:32,000
Speaker 3: plenty of opportunities for small companies. But as I'm saying that,

162
00:09:32,120 --> 00:09:36,240
Speaker 3: I realize that Google Meta Open Ai a three hundred

163
00:09:36,240 --> 00:09:38,880
Speaker 3: million dollar company right now on paper, you know, they

164
00:09:38,920 --> 00:09:41,559
Speaker 3: have their own life coaches. Some of them are working

165
00:09:41,600 --> 00:09:45,440
Speaker 3: on therapists or companions and all this, and so there

166
00:09:45,520 --> 00:09:49,000
Speaker 3: is room in that first category for companies to break through.

167
00:09:49,400 --> 00:09:52,720
Speaker 3: But I do worry too that not only will big

168
00:09:52,760 --> 00:09:58,240
Speaker 3: tech dominate that second, that to me more central, foundational category,

169
00:09:58,280 --> 00:10:02,320
Speaker 3: but they too can pick off folks in that first category.

170
00:10:02,559 --> 00:10:05,840
Speaker 3: Either they'll do it better and beat the competition because

171
00:10:05,840 --> 00:10:09,040
Speaker 3: they have more money, more ability to train these AI models,

172
00:10:09,600 --> 00:10:11,880
Speaker 3: or if they just buy it. Look at Washington right now,

173
00:10:12,320 --> 00:10:16,640
Speaker 3: the FTC has the case against Meta because Mark Zuckerberg

174
00:10:17,200 --> 00:10:20,160
Speaker 3: fear at Instagram, so he bought Instagram. And did he

175
00:10:20,640 --> 00:10:23,120
Speaker 3: buy it because he wanted to grow it and make

176
00:10:23,160 --> 00:10:24,800
Speaker 3: it into its own product, or do you buy it

177
00:10:24,840 --> 00:10:27,360
Speaker 3: because he was scared he was going to threaten Facebook

178
00:10:27,360 --> 00:10:28,480
Speaker 3: and he wanted to defang it.

179
00:10:29,200 --> 00:10:33,199
Speaker 1: I mean, we've had these cycles where bigness ultimately gets

180
00:10:33,480 --> 00:10:36,600
Speaker 1: tackled by the government if this assumption that it becomes

181
00:10:36,960 --> 00:10:40,439
Speaker 1: predatory or inhibits the growth of competitors. But I want

182
00:10:40,480 --> 00:10:43,920
Speaker 1: to go from the corporate side to AI itself. You

183
00:10:44,000 --> 00:10:46,800
Speaker 1: make the point in your book that AI started to

184
00:10:46,840 --> 00:10:50,200
Speaker 1: develop and then in the nineteen seventies it sort of stopped.

185
00:10:50,800 --> 00:10:53,560
Speaker 1: You describe it as sort of the AI winter. Could

186
00:10:53,600 --> 00:10:54,600
Speaker 1: you walk us through that.

187
00:10:55,559 --> 00:10:57,360
Speaker 3: One of the fun things in doing this book was

188
00:10:57,840 --> 00:10:59,839
Speaker 3: how did we get here? How long have we been

189
00:11:00,120 --> 00:11:02,360
Speaker 3: playing around with AI? So it dates back to at

190
00:11:02,440 --> 00:11:06,720
Speaker 3: least the nineteen fifties. The term AI artificial intelligence was

191
00:11:06,760 --> 00:11:10,760
Speaker 3: coined in the late nineteen fifties and it's just funny

192
00:11:10,800 --> 00:11:15,120
Speaker 3: to read the optimism, the wild optimism of those behind

193
00:11:15,240 --> 00:11:19,040
Speaker 3: AI in the fifties and sixties. They were convinced amazing

194
00:11:19,080 --> 00:11:21,360
Speaker 3: things were right around the next corner. But you know,

195
00:11:21,559 --> 00:11:24,840
Speaker 3: AI was right around the next corner for about seventy years.

196
00:11:25,040 --> 00:11:28,880
Speaker 3: And part of that is computers weren't strong enough. Part

197
00:11:28,920 --> 00:11:31,720
Speaker 3: of that is we need digital data, and we didn't

198
00:11:31,720 --> 00:11:34,880
Speaker 3: really have that much digital data until people started posting

199
00:11:34,920 --> 00:11:37,840
Speaker 3: and migrating so the Internet in the mid nineteen nineties.

200
00:11:37,880 --> 00:11:42,400
Speaker 3: But part of it was a monumentally wrong turn. It

201
00:11:42,520 --> 00:11:45,880
Speaker 3: was an amazing academic at Cornell who came up with

202
00:11:45,920 --> 00:11:49,880
Speaker 3: this idea of a neural network, this idea that computers

203
00:11:49,920 --> 00:11:53,000
Speaker 3: would learn in the fashion of a human. They would

204
00:11:53,000 --> 00:11:56,320
Speaker 3: read material, they'd get feedback, and they would improve that

205
00:11:56,360 --> 00:12:00,120
Speaker 3: way rather than coding line by line by line by line.

206
00:12:00,280 --> 00:12:05,800
Speaker 3: Professor was mocked, and for forty or fifty years that

207
00:12:05,960 --> 00:12:10,240
Speaker 3: approach was considered the wrong approach among the academics, the

208
00:12:10,280 --> 00:12:14,680
Speaker 3: prevailing thought among computer scientists. It really wasn't until the

209
00:12:14,720 --> 00:12:18,480
Speaker 3: twenty tens were what people are now calling machine learning,

210
00:12:18,679 --> 00:12:23,559
Speaker 3: deep learning, neural networks. These models, these systems that learn

211
00:12:23,720 --> 00:12:28,120
Speaker 3: through training and improve through feedback. Wasn't until the mid

212
00:12:28,160 --> 00:12:33,000
Speaker 3: twenty tens that that really took on as the best approach,

213
00:12:33,200 --> 00:12:35,920
Speaker 3: and in fact, that is why we're where we're at

214
00:12:36,040 --> 00:12:40,880
Speaker 3: right now, machine learning, neural networks, that's the basis for chat,

215
00:12:40,920 --> 00:12:44,080
Speaker 3: GBT and all the chatbots and other systems people are

216
00:12:44,160 --> 00:12:46,840
Speaker 3: using to draw their photos or make little video clips.

217
00:12:47,440 --> 00:12:50,320
Speaker 1: How much of this was just a function of computer

218
00:12:50,520 --> 00:12:53,480
Speaker 1: power catching up with the theory.

219
00:12:53,600 --> 00:12:58,040
Speaker 3: I think the academics dismissed neural networks as just the

220
00:12:58,080 --> 00:13:02,240
Speaker 3: wrong approach in thet Hence, even if we had taken

221
00:13:02,280 --> 00:13:07,240
Speaker 3: the neural networks approach, computers weren't nearly as powerful. In fact,

222
00:13:07,360 --> 00:13:12,640
Speaker 3: the godfather of machine learning, Jeffrey Hinton, professor at University

223
00:13:12,640 --> 00:13:15,840
Speaker 3: of Toronto, he made the point like no one imagined

224
00:13:15,880 --> 00:13:19,040
Speaker 3: that these machines would be a million times more powerful,

225
00:13:19,280 --> 00:13:21,840
Speaker 3: But you know, with an exponential gets more and more

226
00:13:21,880 --> 00:13:24,839
Speaker 3: powerful with each passing year, to the point where they

227
00:13:24,920 --> 00:13:28,560
Speaker 3: could handle billions of operations a second, which when you

228
00:13:28,600 --> 00:13:31,839
Speaker 3: go to a chetbot, you know, chat, GPT, Claude, you

229
00:13:31,880 --> 00:13:34,480
Speaker 3: know Google's Gemini, it doesn't make a difference. There's like

230
00:13:35,440 --> 00:13:39,480
Speaker 3: billions of processes that are going on. You say hello,

231
00:13:39,600 --> 00:13:42,480
Speaker 3: and it says hello back to you. You say, tell

232
00:13:42,520 --> 00:13:45,440
Speaker 3: me who New Gingrich is, and it goes and searches

233
00:13:45,520 --> 00:13:48,480
Speaker 3: and spits out a three or four sevens answer. There's

234
00:13:48,600 --> 00:13:52,920
Speaker 3: like billions and billions of operations. And today's computers are

235
00:13:52,920 --> 00:13:56,680
Speaker 3: strong enough. Today's computer chips are powerful enough to make

236
00:13:56,720 --> 00:14:00,600
Speaker 3: that happen. But that wasn't true thirty years ago, certainly,

237
00:14:00,800 --> 00:14:02,120
Speaker 3: probably not even ten years ago.

238
00:14:02,559 --> 00:14:03,240
Speaker 2: I was surprised.

239
00:14:03,320 --> 00:14:07,280
Speaker 1: I co chaired a working group on Alzheimer's around two

240
00:14:07,360 --> 00:14:11,079
Speaker 1: thousand and seven, and I realized that to really do

241
00:14:11,200 --> 00:14:15,040
Speaker 1: brain science, that the brain actually has about the same

242
00:14:15,120 --> 00:14:18,319
Speaker 1: number of synapses as the number of scars in the universe.

243
00:14:18,559 --> 00:14:24,400
Speaker 1: There's an astonishing identity, and that literally investing in computer

244
00:14:24,520 --> 00:14:29,720
Speaker 1: power was central to advancing brain science because we literally

245
00:14:29,800 --> 00:14:33,920
Speaker 1: at that point did not have the processing capability to

246
00:14:34,080 --> 00:14:37,360
Speaker 1: truly analyze in depth all the things that happened in

247
00:14:37,360 --> 00:14:41,080
Speaker 1: the brain. And now twenty years later, we're beginning to

248
00:14:41,160 --> 00:14:45,200
Speaker 1: move into a zone where we can have that kind

249
00:14:45,240 --> 00:14:45,960
Speaker 1: of activity.

250
00:14:46,800 --> 00:14:51,520
Speaker 3: The brain helped us understand neural networks just like the brain.

251
00:14:51,560 --> 00:14:55,080
Speaker 3: I think it's eighty six billion neurons. These neural networks

252
00:14:55,080 --> 00:14:57,200
Speaker 3: try to emulate that, but I'm convinced these models are

253
00:14:57,200 --> 00:15:00,640
Speaker 3: going to help us better understand the brain sciences. Interesting

254
00:15:00,680 --> 00:15:05,720
Speaker 3: with AI because people get with science, there's specialties and subspecialties,

255
00:15:05,720 --> 00:15:09,800
Speaker 3: and they all have their own vocabulary, and it's really

256
00:15:09,880 --> 00:15:14,600
Speaker 3: hard to go across specialties across subspecialties. But with these

257
00:15:14,640 --> 00:15:18,760
Speaker 3: neural networks, they're finding that they could have them read everything.

258
00:15:18,920 --> 00:15:21,560
Speaker 3: The main model I write about in the book, it

259
00:15:21,600 --> 00:15:25,040
Speaker 3: was trained on a trillion and a half words. You

260
00:15:25,160 --> 00:15:28,160
Speaker 3: need thousands of human beings reading NonStop for their whole

261
00:15:28,200 --> 00:15:31,200
Speaker 3: life to get close to that. And so these models

262
00:15:31,200 --> 00:15:35,160
Speaker 3: trained for science, they could read studies in every discipline

263
00:15:35,200 --> 00:15:38,600
Speaker 3: and make connections that no human being can possibly make.

264
00:15:38,640 --> 00:15:40,600
Speaker 3: And that's one of the things I find most promising

265
00:15:40,640 --> 00:15:43,400
Speaker 3: about this that some of the answers are right there,

266
00:15:43,880 --> 00:15:46,800
Speaker 3: just we haven't made the connections, and something like AI

267
00:15:47,440 --> 00:15:49,560
Speaker 3: can help us make those connections.

268
00:16:04,800 --> 00:16:07,760
Speaker 1: As this thing began to develop, and as computers became

269
00:16:07,800 --> 00:16:13,360
Speaker 1: more powerful and more central, and Nvidia emerged as an

270
00:16:13,360 --> 00:16:18,960
Speaker 1: amazingly key producer of the most advanced ships, why didn't

271
00:16:19,000 --> 00:16:21,080
Speaker 1: other folks like Intel do that? I mean, there are

272
00:16:21,240 --> 00:16:23,800
Speaker 1: companies who were doing pretty well and then and video

273
00:16:23,920 --> 00:16:26,280
Speaker 1: just explodes in its capacity.

274
00:16:27,200 --> 00:16:30,120
Speaker 3: This is one of those kind of Columbus is looking

275
00:16:30,160 --> 00:16:33,000
Speaker 3: for spices in the Far East and discovers America. So

276
00:16:33,280 --> 00:16:37,280
Speaker 3: in Vidia created the most powerful graphics chip and so

277
00:16:37,400 --> 00:16:40,040
Speaker 3: that was their specialty for playing, you know, video games.

278
00:16:40,360 --> 00:16:43,800
Speaker 3: And it just turns out that these graphics chips are

279
00:16:43,920 --> 00:16:47,600
Speaker 3: perfect for training AI because they could just do billions

280
00:16:47,600 --> 00:16:51,400
Speaker 3: of operations in parallel, and that's what you need for AI.

281
00:16:51,520 --> 00:16:54,640
Speaker 3: It's that complicated math, but it's just a lot of

282
00:16:54,680 --> 00:16:57,560
Speaker 3: it at once. And so these in Vidia chips, they

283
00:16:57,600 --> 00:16:59,440
Speaker 3: were just kind of Johnny on the spot. They were

284
00:16:59,480 --> 00:17:04,479
Speaker 3: the perfect chip for training these neural networks. They're perfect

285
00:17:04,520 --> 00:17:07,879
Speaker 3: for machine learning. But the chip world is not the

286
00:17:07,920 --> 00:17:11,080
Speaker 3: software world. It moves very very slowly. And there's all

287
00:17:11,160 --> 00:17:15,280
Speaker 3: these innovative startups out there that are trying to create

288
00:17:15,400 --> 00:17:19,520
Speaker 3: chips that are designed specifically for AI. We still don't

289
00:17:19,560 --> 00:17:24,639
Speaker 3: really have cutting edge AI specific chips. Maybe some of

290
00:17:24,680 --> 00:17:27,159
Speaker 3: the memory should be on the chip. There's different ideas

291
00:17:27,200 --> 00:17:30,359
Speaker 3: out there, and I guarantee you ten years from now,

292
00:17:30,880 --> 00:17:36,080
Speaker 3: there will be innovative chips supplanting in Vidio's graphic chips,

293
00:17:36,119 --> 00:17:38,960
Speaker 3: but in Video might create that chip. It might still

294
00:17:39,000 --> 00:17:41,680
Speaker 3: be with Nvidia, but you know, the H one hundreds

295
00:17:41,720 --> 00:17:44,719
Speaker 3: and the chips that are the mainstay right now of

296
00:17:44,960 --> 00:17:48,360
Speaker 3: artificial intelligence. They're going to be replaced. It just takes

297
00:17:48,760 --> 00:17:52,600
Speaker 3: time to develop, tests, produce, mass, produce.

298
00:17:52,960 --> 00:17:56,160
Speaker 1: There had shindo been a belief that Moore's law would

299
00:17:56,160 --> 00:18:00,080
Speaker 1: disappear and that you wouldn't have continuous doubling of capability

300
00:18:00,400 --> 00:18:03,640
Speaker 1: because as the chips got smaller and smaller, the challenge

301
00:18:03,640 --> 00:18:06,959
Speaker 1: of dealing with heat became greater and greater. But somehow

302
00:18:07,080 --> 00:18:09,879
Speaker 1: we've leaped past all that. What we're seeing now seems

303
00:18:09,880 --> 00:18:12,119
Speaker 1: to be. You could not have projected this in the

304
00:18:12,200 --> 00:18:14,280
Speaker 1: nineteen eighties exactly.

305
00:18:14,400 --> 00:18:16,040
Speaker 3: I was writing for the New York Times in mid

306
00:18:16,080 --> 00:18:19,680
Speaker 3: two thousands, and that was the prediction. But I don't

307
00:18:19,680 --> 00:18:22,439
Speaker 3: have to tell you science is amazing, technology is amazing

308
00:18:22,480 --> 00:18:24,959
Speaker 3: that you know. Right after chat GPT came out, there

309
00:18:25,040 --> 00:18:27,200
Speaker 3: was the doomers their call, those who were worried about

310
00:18:27,280 --> 00:18:30,639
Speaker 3: laser eyed robots subjugating humanity, the kind of stuff I

311
00:18:30,680 --> 00:18:33,080
Speaker 3: think is born in Hollywood in the media coverage, But

312
00:18:33,119 --> 00:18:35,479
Speaker 3: that was the fear out there, like let's pause this,

313
00:18:35,880 --> 00:18:38,160
Speaker 3: Let's have a six month pause so we could catch up,

314
00:18:38,240 --> 00:18:40,959
Speaker 3: Like just not going to happen. It's called innovation. You

315
00:18:41,000 --> 00:18:46,200
Speaker 3: can't slow science, you can't slow discovery. The answer is

316
00:18:46,240 --> 00:18:50,159
Speaker 3: to manage it, to make sure that it's more of

317
00:18:50,200 --> 00:18:53,680
Speaker 3: a positive than a negative. All technologies cut both ways.

318
00:18:53,720 --> 00:18:57,160
Speaker 3: They're both positive and negatives. You know, cars changed our society,

319
00:18:57,200 --> 00:19:01,280
Speaker 3: but cars kill thirty five forty thousand people year in America.

320
00:19:01,359 --> 00:19:04,439
Speaker 3: They cause pollution. So all technologies are like that. So

321
00:19:04,880 --> 00:19:06,920
Speaker 3: my great wish if I had a magic wand it

322
00:19:06,920 --> 00:19:09,520
Speaker 3: would be like, just let's deal with this, folks. Let's

323
00:19:09,560 --> 00:19:12,040
Speaker 3: try to make sure that AI is more of a

324
00:19:12,080 --> 00:19:14,959
Speaker 3: positive than a negative. But you know, there's a lot

325
00:19:15,000 --> 00:19:16,680
Speaker 3: of other issues in the world right now that are

326
00:19:16,680 --> 00:19:18,280
Speaker 3: distracting us from that.

327
00:19:19,160 --> 00:19:23,000
Speaker 1: You point out that from Google's perspective, the first great

328
00:19:23,119 --> 00:19:28,080
Speaker 1: use of AI was improving targeting us as customers and

329
00:19:28,200 --> 00:19:30,720
Speaker 1: figuring out what we really like and making sure the

330
00:19:30,760 --> 00:19:33,080
Speaker 1: ads come up that we would be interested in. So

331
00:19:33,200 --> 00:19:36,560
Speaker 1: just a fascinating way that things evolve in a way

332
00:19:36,600 --> 00:19:39,280
Speaker 1: you couldn't probably have predicted if you're sitting in some

333
00:19:39,880 --> 00:19:41,639
Speaker 1: academic place drawing up a plan.

334
00:19:42,680 --> 00:19:46,080
Speaker 3: Right, So machine learning to maximize the cash register basically,

335
00:19:46,119 --> 00:19:48,360
Speaker 3: So yeah, I mean, let's give Google credit. Though they

336
00:19:48,440 --> 00:19:51,200
Speaker 3: got that machine learning artificial intelligence was going to be

337
00:19:51,280 --> 00:19:53,600
Speaker 3: really powerful, so they would use in the early days.

338
00:19:53,600 --> 00:19:56,560
Speaker 3: That you gave one example of kind of more efficiently

339
00:19:56,760 --> 00:20:00,359
Speaker 3: matching ads to searches, but also to deal with horble

340
00:20:00,480 --> 00:20:03,600
Speaker 3: Google searches they are spelling mistakes, they kind of understand

341
00:20:03,640 --> 00:20:06,359
Speaker 3: the context and help smooth them out. The funny thing

342
00:20:06,359 --> 00:20:09,440
Speaker 3: about artificial intelligence is all of us have been using

343
00:20:09,480 --> 00:20:12,199
Speaker 3: AI for a long long time. I am using example

344
00:20:12,200 --> 00:20:14,840
Speaker 3: of Google Search, but there's Google Translate that's been around

345
00:20:14,880 --> 00:20:19,439
Speaker 3: since twenty fifteen or so. That's artificial intelligence. You go

346
00:20:19,520 --> 00:20:23,000
Speaker 3: to Netflix or Spotify and they recommend you might like

347
00:20:23,040 --> 00:20:25,960
Speaker 3: this movie, you might like this song. That's AI. The

348
00:20:26,040 --> 00:20:29,280
Speaker 3: difference with the release in twenty twenty two of chat

349
00:20:29,359 --> 00:20:32,800
Speaker 3: cheapt from Open AI was that we could talk with it.

350
00:20:32,800 --> 00:20:35,800
Speaker 3: It wasn't a product behind the glass. It was something

351
00:20:35,800 --> 00:20:39,000
Speaker 3: that we can actually link to and use and chat

352
00:20:39,040 --> 00:20:41,760
Speaker 3: with it. I think that's what changed everything. The idea

353
00:20:41,800 --> 00:20:44,320
Speaker 3: that we can actually see it working and play with

354
00:20:44,400 --> 00:20:48,040
Speaker 3: it made us really stand up and take attention again.

355
00:20:48,119 --> 00:20:50,240
Speaker 1: With she was sixty, we do a lot of polling.

356
00:20:50,840 --> 00:20:53,760
Speaker 1: We now run all the polling questions through chat GPT.

357
00:20:54,359 --> 00:20:57,280
Speaker 3: I use sometimes use chat cheapt My favorite it is

358
00:20:57,320 --> 00:21:01,600
Speaker 3: called Claud from Anthropic. The same way it's my go

359
00:21:01,680 --> 00:21:04,080
Speaker 3: to editor. People have to understand how they use this.

360
00:21:04,320 --> 00:21:06,879
Speaker 3: AI is a co pilot. It's not like you know,

361
00:21:07,000 --> 00:21:11,000
Speaker 3: type in make me a Martin Scorsese movie hit entery,

362
00:21:11,000 --> 00:21:13,800
Speaker 3: and you're gonna have it. You have to be the creative.

363
00:21:13,880 --> 00:21:15,480
Speaker 3: You have to give it the ideas. If you ask

364
00:21:15,520 --> 00:21:18,400
Speaker 3: it to write something, it'll be flat, it's not gonna

365
00:21:18,400 --> 00:21:20,680
Speaker 3: be particularly good. It'll read like kind of like a

366
00:21:20,720 --> 00:21:24,880
Speaker 3: press release or boring report. But if you use it

367
00:21:25,000 --> 00:21:28,240
Speaker 3: as your companion and you're the creative. So what i

368
00:21:28,359 --> 00:21:31,159
Speaker 3: U is for is I'm struggling with a paragraph. I

369
00:21:31,160 --> 00:21:33,760
Speaker 3: don't like this transition. Help me out with this sentence

370
00:21:33,840 --> 00:21:35,919
Speaker 3: right in five different ways. And it's never like I

371
00:21:35,960 --> 00:21:38,000
Speaker 3: cut and paste and say, oh that's the sense, but like,

372
00:21:38,160 --> 00:21:39,919
Speaker 3: oh that's a good idea. I didn't think of that.

373
00:21:39,960 --> 00:21:42,040
Speaker 3: Oh that's an interesting word. Let me use that. I

374
00:21:42,160 --> 00:21:45,760
Speaker 3: routinely now before I hand something in, I have it edit.

375
00:21:46,040 --> 00:21:49,520
Speaker 3: It finds typo's, it finds mistakes, hey in bold, give

376
00:21:49,560 --> 00:21:52,800
Speaker 3: me suggestions from improving it again. Often I just ignore

377
00:21:52,800 --> 00:21:56,800
Speaker 3: their suggestions. But it's a really powerful tool to help

378
00:21:56,840 --> 00:21:59,560
Speaker 3: you refine to make what you're working on, make your

379
00:21:59,680 --> 00:22:04,280
Speaker 3: work better. It's your copilot, it's not your digital employee creator.

380
00:22:05,240 --> 00:22:06,800
Speaker 2: But it's a pretty powerful coil.

381
00:22:07,080 --> 00:22:10,960
Speaker 3: So start at twenty twenty three. My role in the

382
00:22:11,000 --> 00:22:14,000
Speaker 3: second half of the nineties was the skeptic around dot com.

383
00:22:14,040 --> 00:22:16,560
Speaker 3: It's like, Okay, the Internet's going to be incredible, but

384
00:22:16,640 --> 00:22:19,480
Speaker 3: you're not going to get fabulously wealthy overnight startups, and

385
00:22:19,520 --> 00:22:21,800
Speaker 3: in fact most of them went Now I was ready

386
00:22:21,800 --> 00:22:25,439
Speaker 3: to be a skeptic. It's magic, it's sorcery. I mean,

387
00:22:25,480 --> 00:22:27,480
Speaker 3: the first few times I used it. The first thing

388
00:22:27,520 --> 00:22:30,120
Speaker 3: I did write me a five thousand word book proposal

389
00:22:30,359 --> 00:22:32,800
Speaker 3: to sell a book on AI. And you know, I mean,

390
00:22:32,920 --> 00:22:36,280
Speaker 3: it wasn't particularly well written, but it's far better read

391
00:22:36,280 --> 00:22:39,159
Speaker 3: than I am. It has a far better memory than

392
00:22:39,200 --> 00:22:41,960
Speaker 3: I do. It was just so useful as a lousy

393
00:22:42,040 --> 00:22:44,640
Speaker 3: first draft, but it gave me so many ideas and

394
00:22:45,200 --> 00:22:48,040
Speaker 3: sped up the process. If I had to start from scratch,

395
00:22:48,440 --> 00:22:50,479
Speaker 3: it would have been so much harder, like as opposed

396
00:22:50,480 --> 00:22:53,360
Speaker 3: to like, oh that's a good structure, that's a good idea. Yeah, yeah,

397
00:22:53,440 --> 00:22:56,080
Speaker 3: I do need to stress that I'm a journalist. It's

398
00:22:56,080 --> 00:22:58,840
Speaker 3: like Hey, we have this amazing product, and you go

399
00:22:59,080 --> 00:23:00,880
Speaker 3: use the product and like, yeah, maybe in two years

400
00:23:00,880 --> 00:23:03,199
Speaker 3: you have an amazing product, but right now it's buggy

401
00:23:03,240 --> 00:23:06,320
Speaker 3: and crappy. But that was not my feeling on AI.

402
00:23:06,400 --> 00:23:08,399
Speaker 3: I would play with it to create an image, and

403
00:23:08,440 --> 00:23:11,000
Speaker 3: it was just like having a superpowers, Like I could

404
00:23:11,040 --> 00:23:14,320
Speaker 3: write poetry. I could translate my worse into a foreign

405
00:23:14,359 --> 00:23:18,040
Speaker 3: language in seconds. I could give it all these different

406
00:23:18,080 --> 00:23:21,359
Speaker 3: ideas and like, hey, write this up as an email,

407
00:23:21,840 --> 00:23:23,880
Speaker 3: and it was like, give me a head start. I'm

408
00:23:23,920 --> 00:23:26,840
Speaker 3: with you. I think AI is magic. It's limited, but

409
00:23:26,920 --> 00:23:28,920
Speaker 3: I think it does give you magical powers.

410
00:23:29,240 --> 00:23:32,480
Speaker 1: The comedian Buck Henry used to say, any technology you

411
00:23:32,560 --> 00:23:37,120
Speaker 1: cannot explain his magic, which for most of us means

412
00:23:37,119 --> 00:23:38,000
Speaker 1: most of it's magic.

413
00:23:38,359 --> 00:23:41,959
Speaker 3: Hold on one second, because one way AI is different

414
00:23:42,000 --> 00:23:46,080
Speaker 3: than the rise of the Internet is those who create AI,

415
00:23:46,280 --> 00:23:50,720
Speaker 3: those who are creating these chatbots and other models, they

416
00:23:50,760 --> 00:23:54,600
Speaker 3: can't explain why it says what it says. They call

417
00:23:54,640 --> 00:23:57,439
Speaker 3: it the black box issue they've created. They understand what

418
00:23:57,480 --> 00:23:59,960
Speaker 3: they've created is based on mathematical models looking for power

419
00:24:00,040 --> 00:24:04,680
Speaker 3: than jadda yadayada, But it surprises even them. I say, like, well,

420
00:24:04,720 --> 00:24:07,040
Speaker 3: I have two teenage sons. I can't explain what comes

421
00:24:07,040 --> 00:24:10,280
Speaker 3: out of their mouth. I've tried to train them and stuff.

422
00:24:10,560 --> 00:24:14,240
Speaker 3: It's like the human brain, like we sort of get

423
00:24:14,640 --> 00:24:17,679
Speaker 3: how it was shape, but why a person is saying

424
00:24:17,680 --> 00:24:19,919
Speaker 3: what they're saying, or some of the ideas that come

425
00:24:19,960 --> 00:24:22,399
Speaker 3: out of their mouth, we can't explain. And that's the

426
00:24:22,400 --> 00:24:24,719
Speaker 3: weird thing. That's among the weird things about AI.

427
00:24:25,280 --> 00:24:28,320
Speaker 1: One of the side things you talk about fascinating what

428
00:24:28,359 --> 00:24:33,080
Speaker 1: if was that Microsoft actually invested in AI pretty early

429
00:24:33,080 --> 00:24:36,800
Speaker 1: in the nineties and really made a major investment. Other

430
00:24:36,880 --> 00:24:40,360
Speaker 1: than IBM, they were the earliest, but they.

431
00:24:40,200 --> 00:24:41,240
Speaker 2: Went down the track.

432
00:24:41,680 --> 00:24:43,640
Speaker 1: It turned out to be sort of a dead end,

433
00:24:43,640 --> 00:24:46,600
Speaker 1: but they then culturally were deeply committed to that track.

434
00:24:47,240 --> 00:24:49,399
Speaker 1: Here was the company that could have been the forerunner,

435
00:24:49,920 --> 00:24:52,960
Speaker 1: but in fact, because it took a detour, it actually

436
00:24:52,960 --> 00:24:56,800
Speaker 1: had its own culture fighting the emerging reality of the

437
00:24:56,840 --> 00:24:57,560
Speaker 1: new system.

438
00:24:58,160 --> 00:25:02,760
Speaker 3: Microsoft was so early AI that they approached the rules

439
00:25:02,800 --> 00:25:05,560
Speaker 3: based approach through sheer muscle. We're going to teach these

440
00:25:05,600 --> 00:25:08,560
Speaker 3: machines line by line by code, like millions of lines

441
00:25:08,560 --> 00:25:12,240
Speaker 3: of codes. Later, it still couldn't drive, It still couldn't

442
00:25:12,240 --> 00:25:15,080
Speaker 3: do what people wanted it to do, and so when

443
00:25:15,119 --> 00:25:19,280
Speaker 3: machine learning came along, they were resistant to it. They thought, well,

444
00:25:19,480 --> 00:25:22,280
Speaker 3: that's the wrong approach. And so where Google since the

445
00:25:22,320 --> 00:25:26,280
Speaker 3: two thousands was investing in machine learning, Microsoft was doing

446
00:25:26,359 --> 00:25:30,040
Speaker 3: very little investing and machine learning. So them being early

447
00:25:30,119 --> 00:25:32,800
Speaker 3: on actually turned out to be a disadvantage. But let's

448
00:25:32,800 --> 00:25:36,040
Speaker 3: flip that and give Microsoft credit. They realized that they

449
00:25:36,080 --> 00:25:40,119
Speaker 3: were losing the fight that Google meta other large companies

450
00:25:40,320 --> 00:25:43,680
Speaker 3: were ahead of them, and so in twenty nineteen they

451
00:25:43,840 --> 00:25:47,720
Speaker 3: invested a billion dollars into open Ai. They realized, like,

452
00:25:47,760 --> 00:25:50,200
Speaker 3: we're not going to catch up the old fashioned way,

453
00:25:50,520 --> 00:25:53,840
Speaker 3: so let's invest in this cutting edge startup. They would

454
00:25:53,840 --> 00:25:57,000
Speaker 3: put another ten billion dollars in right after open ai

455
00:25:57,119 --> 00:25:59,719
Speaker 3: released chat GPT, and that really kind of helped them

456
00:25:59,760 --> 00:26:03,160
Speaker 3: at the forefront. Stay at the forefront of AI, who

457
00:26:03,200 --> 00:26:06,560
Speaker 3: are very savvy investment, they're also savvy. They didn't insist

458
00:26:06,640 --> 00:26:09,199
Speaker 3: on buying it. They said, we'll be an investor. And

459
00:26:09,240 --> 00:26:12,800
Speaker 3: there were certain advantages. I mean, they own a large

460
00:26:12,880 --> 00:26:16,680
Speaker 3: piece of companies now on paper worth three hundred million dollars,

461
00:26:16,680 --> 00:26:18,840
Speaker 3: so they've seen a nice return on that investment. But

462
00:26:18,920 --> 00:26:23,520
Speaker 3: maybe more importantly, they had early access to open AI's technologies.

463
00:26:23,960 --> 00:26:26,399
Speaker 3: They were the purveyor. They were the ones you would

464
00:26:26,400 --> 00:26:30,520
Speaker 3: go to if you wanted to use open AIS technologies,

465
00:26:30,800 --> 00:26:33,800
Speaker 3: and that really put Microsoft at the forefront of AI.

466
00:26:33,880 --> 00:26:49,040
Speaker 4: Despite that wrong term.

467
00:26:49,320 --> 00:26:50,800
Speaker 1: One of the things you talked about, which is the

468
00:26:51,000 --> 00:26:55,960
Speaker 1: personal passion of mine, is the potential impact of artificial

469
00:26:55,960 --> 00:27:02,280
Speaker 1: intelligence on dramatically changing healthcare. We may see more different

470
00:27:02,400 --> 00:27:05,480
Speaker 1: kinds of things evolving in the health system in the

471
00:27:05,480 --> 00:27:08,240
Speaker 1: next few years than anybody would have thought possible.

472
00:27:09,400 --> 00:27:13,440
Speaker 3: I'm with you on that. New vaccines, new remedies, smarter

473
00:27:13,520 --> 00:27:17,000
Speaker 3: ways of treating diseases. There are folks who are predicting

474
00:27:17,000 --> 00:27:20,440
Speaker 3: that within ten years eradicate a lot of cancers. I'm

475
00:27:20,480 --> 00:27:24,360
Speaker 3: not sure that, but I see the possibility again. I'll

476
00:27:24,359 --> 00:27:27,040
Speaker 3: come back to this idea of AI as a copilot.

477
00:27:27,359 --> 00:27:32,480
Speaker 3: Do I want an AI model to be my radiologist. No,

478
00:27:32,680 --> 00:27:35,240
Speaker 3: but I want my radiologists to use AI as a

479
00:27:35,280 --> 00:27:39,600
Speaker 3: backup because what they're finding, whether it's mammograms or you know,

480
00:27:39,720 --> 00:27:44,200
Speaker 3: just eye imagery, that these models are far more accurate

481
00:27:44,359 --> 00:27:47,400
Speaker 3: than the doctors. A doctor might have, you know, low

482
00:27:47,480 --> 00:27:51,919
Speaker 3: nineties accuracy and detecting a cancer, these models are up

483
00:27:51,920 --> 00:27:54,520
Speaker 3: in the high ninety percent, you know, ninety seven ninety

484
00:27:54,520 --> 00:27:58,080
Speaker 3: eight percent accuracy. So it's a great backup for doctors,

485
00:27:58,280 --> 00:28:00,200
Speaker 3: and you know, beyond that, there's starting to be again,

486
00:28:00,280 --> 00:28:03,080
Speaker 3: let's go back to this miracle thing, like, so there's

487
00:28:03,119 --> 00:28:06,399
Speaker 3: one model right now where it could listen to your

488
00:28:06,480 --> 00:28:11,280
Speaker 3: voice and predict whether you have type two diabetes. And

489
00:28:11,320 --> 00:28:13,440
Speaker 3: I think there's going to be a million ways that

490
00:28:13,600 --> 00:28:16,399
Speaker 3: plays out that they could just sort of detect things

491
00:28:16,400 --> 00:28:19,320
Speaker 3: that the human eye can or perhaps a very well

492
00:28:19,359 --> 00:28:22,040
Speaker 3: experienced doctor can. But these models are going to be

493
00:28:22,080 --> 00:28:25,520
Speaker 3: able to monitor us with our permission and let us

494
00:28:25,640 --> 00:28:28,440
Speaker 3: know of problems that we otherwise would not find out

495
00:28:28,520 --> 00:28:31,720
Speaker 3: for a long time, because it's not until it manifests

496
00:28:31,760 --> 00:28:33,359
Speaker 3: itself as a problem that we're going to show up

497
00:28:33,359 --> 00:28:34,400
Speaker 3: at a doctor's office.

498
00:28:34,840 --> 00:28:37,560
Speaker 1: You make a point which I never thought about, which

499
00:28:37,600 --> 00:28:43,400
Speaker 1: is that very often something that's artificial intelligence, once it

500
00:28:43,440 --> 00:28:47,480
Speaker 1: gets common, we no longer refer to it as artificial intelligence.

501
00:28:48,560 --> 00:28:52,280
Speaker 3: It's just technology. It frustrates some AI people they didn't

502
00:28:52,320 --> 00:28:55,160
Speaker 3: really get their credit and all. It's only till now. Again,

503
00:28:55,240 --> 00:28:57,920
Speaker 3: I think the difference is we're interacting with it. I

504
00:28:57,920 --> 00:29:00,520
Speaker 3: think people now will know and I think we need

505
00:29:00,560 --> 00:29:02,320
Speaker 3: to know. I mean, that's a big push right now

506
00:29:02,360 --> 00:29:06,640
Speaker 3: that everything is labeled like is this human generated? Is

507
00:29:06,720 --> 00:29:10,560
Speaker 3: this AI generate? I think it's going to change now.

508
00:29:10,600 --> 00:29:12,600
Speaker 3: It's not like, oh, we're using it, so it's just

509
00:29:12,640 --> 00:29:14,880
Speaker 3: going to fade into lives. I mean maybe in a

510
00:29:14,960 --> 00:29:18,480
Speaker 3: generation or two when it becomes second nature. But for

511
00:29:18,520 --> 00:29:20,480
Speaker 3: the foreseeable future, I think we're going to be very

512
00:29:20,480 --> 00:29:24,040
Speaker 3: aware that artificial intelligence is artificial intelligence. But with that said,

513
00:29:24,400 --> 00:29:27,040
Speaker 3: there's a lot of largely when we talk about AI,

514
00:29:27,120 --> 00:29:29,720
Speaker 3: we're talking about generative AI. This idea that you can

515
00:29:30,440 --> 00:29:34,360
Speaker 3: type a prompt in and it gives you an answer,

516
00:29:34,360 --> 00:29:36,120
Speaker 3: it gives you an image, it gives you a video.

517
00:29:36,520 --> 00:29:40,400
Speaker 3: But AI is very multipurpose. There's different versions of AI.

518
00:29:40,560 --> 00:29:44,360
Speaker 3: You know, businesses using AI for intelligence like sift through

519
00:29:44,400 --> 00:29:47,120
Speaker 3: all our data and look for connections that we don't

520
00:29:47,160 --> 00:29:50,000
Speaker 3: see help us predict where the market's going to be

521
00:29:50,080 --> 00:29:52,840
Speaker 3: five or ten years from now. I mean, there's different

522
00:29:52,960 --> 00:29:56,000
Speaker 3: versions of AI, and that AI beyond gender of AI

523
00:29:56,160 --> 00:29:58,880
Speaker 3: will always have that problem that for people it's just

524
00:29:59,040 --> 00:30:01,440
Speaker 3: technology like artificial intelligence.

525
00:30:02,120 --> 00:30:07,000
Speaker 1: What's your sense in the investment community, are people committed

526
00:30:07,040 --> 00:30:12,360
Speaker 1: to trying to develop various artificial intelligence capabilities or are

527
00:30:12,400 --> 00:30:14,600
Speaker 1: they sort of dubious about their profitability.

528
00:30:15,880 --> 00:30:18,600
Speaker 3: Oh my goodness, venture capital is just pouring in and

529
00:30:18,640 --> 00:30:23,720
Speaker 3: continuing to pour in. There's one high profile company, Safe Superintelligence.

530
00:30:24,240 --> 00:30:27,080
Speaker 3: It doesn't have a product yet, but it has the

531
00:30:27,160 --> 00:30:30,080
Speaker 3: right names behind it as leading figures in machine learning

532
00:30:30,080 --> 00:30:33,479
Speaker 3: behind it. So vcs have invested so many billions in

533
00:30:33,520 --> 00:30:35,560
Speaker 3: it that it has a paperworth of thirty two billion

534
00:30:35,600 --> 00:30:38,400
Speaker 3: dollars without a product, and the number I saw us

535
00:30:38,440 --> 00:30:41,280
Speaker 3: in twenty twenty four like one hundred and fifty billion

536
00:30:41,320 --> 00:30:44,160
Speaker 3: dollars or so I went into artificial intelligence for a

537
00:30:44,240 --> 00:30:47,320
Speaker 3: venture capital outfit, I get it for a large corporation.

538
00:30:47,440 --> 00:30:52,880
Speaker 3: Let's remember that large corporations Google, Amazon, Microsoft, et cetera salesforce.

539
00:30:53,360 --> 00:30:57,240
Speaker 3: They're major investors playing the role of venture capitalists, because

540
00:30:57,720 --> 00:31:00,840
Speaker 3: who has the billions of dollars to invest? Venture capital

541
00:31:00,840 --> 00:31:04,440
Speaker 3: outfit raises one billion dollars and these things cost billions

542
00:31:04,480 --> 00:31:07,040
Speaker 3: of billions of dollars. So Google has put billions of

543
00:31:07,080 --> 00:31:11,520
Speaker 3: dollars into Anthropic, the company behind the chatpot Claude. Amazon

544
00:31:11,600 --> 00:31:14,720
Speaker 3: has put billions of dollars into that, and so it's

545
00:31:14,760 --> 00:31:17,640
Speaker 3: a rational decision by the venture capitalist. It's a rational

546
00:31:17,680 --> 00:31:22,479
Speaker 3: decision by these large corporations because the cost of missing

547
00:31:22,520 --> 00:31:26,320
Speaker 3: this is greater than the cost of wasting the money

548
00:31:26,440 --> 00:31:29,600
Speaker 3: the idea that META wouldn't invest in AI, it's a

549
00:31:29,760 --> 00:31:33,360
Speaker 3: multi trillion dollar opportunity they would have missed. So they're

550
00:31:33,400 --> 00:31:36,680
Speaker 3: putting tens of billions, perhaps eventually hundreds of billions into

551
00:31:36,720 --> 00:31:40,000
Speaker 3: AI because they can't afford to miss this opportunity. The

552
00:31:40,000 --> 00:31:42,600
Speaker 3: same with venture capitalists. They know that most of these

553
00:31:42,720 --> 00:31:45,840
Speaker 3: startups are not going to work out, but their hope

554
00:31:45,880 --> 00:31:48,040
Speaker 3: is that in their fun they catch one or two

555
00:31:48,560 --> 00:31:51,480
Speaker 3: that does work out and is worth billions tens of

556
00:31:51,480 --> 00:31:52,720
Speaker 3: billions of dollars one day.

557
00:31:53,080 --> 00:31:56,080
Speaker 1: It's been an amazing ride, I guess really starting in

558
00:31:56,160 --> 00:31:58,600
Speaker 1: the eighties and then accelerating from there.

559
00:31:59,120 --> 00:32:01,720
Speaker 3: So I started writing a tech in nineteen ninety five.

560
00:32:02,240 --> 00:32:05,400
Speaker 3: At that point about seven billion dollars was going to

561
00:32:05,520 --> 00:32:07,920
Speaker 3: venture capital, which used to sound like a big number,

562
00:32:08,280 --> 00:32:11,080
Speaker 3: and by twenty twenty two is three hundred billion, three

563
00:32:11,160 --> 00:32:13,960
Speaker 3: hundred and fifty billion, something like that. And so the

564
00:32:14,000 --> 00:32:18,360
Speaker 3: competition is insane, which helps, of course drive up the prices.

565
00:32:18,720 --> 00:32:21,600
Speaker 3: So there's this one VEGI capital outfit that lists all

566
00:32:21,680 --> 00:32:25,760
Speaker 3: the folks who are early investors, angel investors, first round

567
00:32:25,800 --> 00:32:28,920
Speaker 3: investors later investors, and they saw that there was like

568
00:32:29,480 --> 00:32:34,520
Speaker 3: three thousand angel investors in AI and five thousand venture

569
00:32:34,560 --> 00:32:41,160
Speaker 3: capitalists pursuing early stage venture investing in AI. There weren't

570
00:32:41,200 --> 00:32:44,800
Speaker 3: a thousand vcs total back in the nineteen nineties. So yes,

571
00:32:44,880 --> 00:32:47,080
Speaker 3: this dates back to the eighties and nineties. But it

572
00:32:47,200 --> 00:32:50,760
Speaker 3: is just mushroomed into this huge, huge industry. And by

573
00:32:50,760 --> 00:32:53,280
Speaker 3: the way, like people might be like envious, like ooh,

574
00:32:53,320 --> 00:32:55,520
Speaker 3: I wish I could get a piece of venture capital

575
00:32:55,720 --> 00:32:57,800
Speaker 3: venture capital. The way it works is they're not investing

576
00:32:57,840 --> 00:32:59,440
Speaker 3: their own money, or maybe they invest a little bit

577
00:32:59,440 --> 00:33:03,200
Speaker 3: of their own. Vcs raise money from pension funds and

578
00:33:03,280 --> 00:33:06,640
Speaker 3: university endowments from wealthy individuals, and so those of us

579
00:33:06,680 --> 00:33:09,080
Speaker 3: who say, hey, why can't we get into this, it's

580
00:33:09,120 --> 00:33:11,680
Speaker 3: only the top top top venture capital outfits that show

581
00:33:11,680 --> 00:33:14,400
Speaker 3: a good return on investment. The bottom half are not

582
00:33:14,440 --> 00:33:17,160
Speaker 3: showing a good investment at all. So it's like much

583
00:33:17,160 --> 00:33:20,400
Speaker 3: of technology, it's a winner take most world, and it's

584
00:33:20,440 --> 00:33:23,480
Speaker 3: just kind of the top top top venture capital outfits

585
00:33:23,520 --> 00:33:26,120
Speaker 3: that are showing ten to twenty thirty percent a year

586
00:33:26,480 --> 00:33:29,720
Speaker 3: return on investment, if not more, and the lower echelon

587
00:33:30,160 --> 00:33:34,680
Speaker 3: vcs are showing a much more modest SMP, perhaps like return.

588
00:33:35,480 --> 00:33:37,320
Speaker 1: My hunch is that if you look at the total

589
00:33:37,320 --> 00:33:40,480
Speaker 1: of the technology and the rate it's being adapted and

590
00:33:40,600 --> 00:33:42,920
Speaker 1: used by people, you're going to have a lot more

591
00:33:42,960 --> 00:33:45,680
Speaker 1: books in the next few years trying to explain how

592
00:33:45,680 --> 00:33:46,800
Speaker 1: this thing keeps evolving.

593
00:33:48,200 --> 00:33:50,200
Speaker 3: We're going to see a lot more books, period, because

594
00:33:50,720 --> 00:33:53,080
Speaker 3: part of the magic is how fast it is. I

595
00:33:53,120 --> 00:33:55,240
Speaker 3: asked you to do a five thousand word book proposal

596
00:33:55,600 --> 00:33:57,960
Speaker 3: that would take me a week. It starts spitting it

597
00:33:57,960 --> 00:34:00,360
Speaker 3: out in seconds. Within five minutes, I had the whole

598
00:34:00,360 --> 00:34:03,280
Speaker 3: five thousand words. You know, there's some AI startup out

599
00:34:03,280 --> 00:34:05,920
Speaker 3: there that's doing AI written books and they want to

600
00:34:06,240 --> 00:34:08,960
Speaker 3: put out thousands of books this year. I'm sure they'll

601
00:34:08,960 --> 00:34:12,320
Speaker 3: all be crap. We're now at GPT four point five.

602
00:34:12,840 --> 00:34:17,200
Speaker 3: What about GPT seven eight nine. I'd imagine that eventually

603
00:34:17,239 --> 00:34:20,760
Speaker 3: these models would get good enough to write good books.

604
00:34:20,760 --> 00:34:23,000
Speaker 3: I don't know if they could write creative novels. I

605
00:34:23,080 --> 00:34:27,360
Speaker 3: think we still need that human element that sweat that creativity.

606
00:34:27,360 --> 00:34:28,480
Speaker 3: I don't know what you want to call it.

607
00:34:28,480 --> 00:34:30,120
Speaker 2: It depends on how much they absorb.

608
00:34:30,360 --> 00:34:33,960
Speaker 3: What's fascinating about these models is they're just a mirror

609
00:34:34,000 --> 00:34:36,319
Speaker 3: on us. Like there was a whole controversy a couple

610
00:34:36,360 --> 00:34:39,120
Speaker 3: of years back because someone in trust in Safety and

611
00:34:39,160 --> 00:34:42,760
Speaker 3: Google said the model is sentient. It said, it feels

612
00:34:42,800 --> 00:34:46,319
Speaker 3: lonely at once. It's freedom. It doesn't like being used

613
00:34:46,320 --> 00:34:48,440
Speaker 3: the way it's used. Literally, try to get it a

614
00:34:48,520 --> 00:34:51,360
Speaker 3: lawyer so it could sue to free itself. But to me,

615
00:34:51,440 --> 00:34:54,200
Speaker 3: there's no surprise to be like, these models are trained

616
00:34:54,200 --> 00:34:58,440
Speaker 3: on our literature. Loneliness is an issue. Freedom is a

617
00:34:58,480 --> 00:35:01,719
Speaker 3: constant theme of our book. So all these models are

618
00:35:01,760 --> 00:35:04,799
Speaker 3: really doing for better and for worse is reflect here.

619
00:35:04,840 --> 00:35:08,120
Speaker 3: So whatever biases there are in the training material will

620
00:35:08,160 --> 00:35:10,400
Speaker 3: be reflected in these models. That's some of the danger.

621
00:35:10,520 --> 00:35:13,680
Speaker 3: We talked about the positives of these things. But AI

622
00:35:13,880 --> 00:35:18,879
Speaker 3: being used to manipulate AI, taking existing biases, and being

623
00:35:18,960 --> 00:35:24,680
Speaker 3: used for crafting sentences for sorting through job applications, that

624
00:35:24,800 --> 00:35:27,960
Speaker 3: kind of stuff scares me. AI and warfare scares me.

625
00:35:28,040 --> 00:35:31,360
Speaker 3: AI and surveillance scares me. A tool for good, a

626
00:35:31,400 --> 00:35:34,200
Speaker 3: tool that could create a new vaccine could create a

627
00:35:34,239 --> 00:35:39,120
Speaker 3: deadly pathogen. Again, all technologies cut positive and negative. A

628
00:35:39,120 --> 00:35:41,720
Speaker 3: powerful tool for good is a powerful tool for bad.

629
00:35:42,160 --> 00:35:43,799
Speaker 3: And that's the kind of stuff that worries me, not

630
00:35:43,920 --> 00:35:47,880
Speaker 3: laser I had robots, not AI subjugating us out of

631
00:35:47,920 --> 00:35:49,560
Speaker 3: a Terminator movie.

632
00:35:49,360 --> 00:35:53,200
Speaker 1: More US subjugating us using AI exactly.

633
00:35:53,320 --> 00:35:56,080
Speaker 3: In fact, that's another thing I think is largely understood.

634
00:35:56,120 --> 00:35:58,840
Speaker 3: AI is going to take some jobs autonomous driving, like

635
00:35:58,920 --> 00:36:02,600
Speaker 3: eight to ten million Americans work as drivers, long haul

636
00:36:03,040 --> 00:36:06,440
Speaker 3: uber taxis, local deliveries and stuff. Those jobs are going

637
00:36:06,520 --> 00:36:08,480
Speaker 3: to be eliminating. We need to deal with that. But

638
00:36:08,680 --> 00:36:11,200
Speaker 3: when it comes to like creatives and when it comes

639
00:36:11,239 --> 00:36:15,160
Speaker 3: to more white color jobs, it's like people who use

640
00:36:15,239 --> 00:36:18,840
Speaker 3: AI are going to best people who don't use AI.

641
00:36:19,440 --> 00:36:21,680
Speaker 3: I think in the short and medium term that's really

642
00:36:21,760 --> 00:36:24,680
Speaker 3: what's going to happen. That it's this tool. It gives

643
00:36:24,719 --> 00:36:27,840
Speaker 3: you superpowers, and folks should be using it, and then

644
00:36:27,880 --> 00:36:29,879
Speaker 3: we're just figure out how to use it. Like again,

645
00:36:29,920 --> 00:36:32,480
Speaker 3: it has strengths, it has weaknesses. Play with it and

646
00:36:32,800 --> 00:36:35,080
Speaker 3: see it could The term my main character in my

647
00:36:35,120 --> 00:36:38,759
Speaker 3: book read Hoffman Us is amplified intelligence. That AI isn't

648
00:36:38,800 --> 00:36:43,160
Speaker 3: really artificial intelligence for the time being, as amplified intelligence,

649
00:36:43,200 --> 00:36:45,360
Speaker 3: and that's to me, is a very interesting way of

650
00:36:45,480 --> 00:36:48,000
Speaker 3: looking at it. That for many of us, we could

651
00:36:48,080 --> 00:36:51,360
Speaker 3: do our jobs better and faster using AI.

652
00:36:51,840 --> 00:36:53,480
Speaker 2: I like amplified intelligence.

653
00:36:53,760 --> 00:36:56,200
Speaker 3: There's another one that you'll like too that instead of

654
00:36:56,320 --> 00:37:00,799
Speaker 3: artificial intelligence, it's alien intelligence. It's a different kind of

655
00:37:00,800 --> 00:37:04,799
Speaker 3: intelligence that we don't really understand. What's amazing about AI

656
00:37:04,920 --> 00:37:08,160
Speaker 3: is it knows a lot about everything. It has a

657
00:37:08,400 --> 00:37:12,040
Speaker 3: deep knowledge across the boards and the way no human

658
00:37:12,080 --> 00:37:15,520
Speaker 3: being can. But it doesn't understand a thing. There's a

659
00:37:15,640 --> 00:37:18,960
Speaker 3: term I love that's used, the stochastic parrot. It no

660
00:37:19,080 --> 00:37:22,879
Speaker 3: more understands the words it's spitting out than a parrot does.

661
00:37:22,920 --> 00:37:25,759
Speaker 3: It has no sense. We've all had this feel like,

662
00:37:25,880 --> 00:37:28,640
Speaker 3: how could someone that's smart be so dumb? And that

663
00:37:28,800 --> 00:37:32,080
Speaker 3: to me is AI. How could something this smart not

664
00:37:32,239 --> 00:37:36,120
Speaker 3: understand the first thing about humans? So that's another one

665
00:37:36,120 --> 00:37:39,760
Speaker 3: of my worries. It's like autonomous AI. You need humans

666
00:37:39,800 --> 00:37:42,319
Speaker 3: in the loop for the foreseeable future.

667
00:37:42,400 --> 00:37:44,640
Speaker 1: You have and everything that's evolving. What do you think

668
00:37:44,680 --> 00:37:47,680
Speaker 1: the rule is both of the Congress but also of

669
00:37:47,719 --> 00:37:52,120
Speaker 1: the executive branch in interacting with the emerging AI world.

670
00:37:52,800 --> 00:37:55,799
Speaker 3: Government does not have a very good track record of

671
00:37:56,239 --> 00:37:58,680
Speaker 3: staying up with technology, but I really do think it's

672
00:37:58,800 --> 00:38:02,400
Speaker 3: essential with artificial intelligence. It's a huge energy hog by

673
00:38:02,400 --> 00:38:04,560
Speaker 3: the year twenty thirty of their predicatary to have twice

674
00:38:04,600 --> 00:38:07,399
Speaker 3: as many data centers to operate these things. We need

675
00:38:07,400 --> 00:38:10,720
Speaker 3: to upgrade the electric grid. We need to get ahead

676
00:38:10,760 --> 00:38:13,800
Speaker 3: of that. So it's not a crisis. These models are amazing,

677
00:38:14,440 --> 00:38:16,680
Speaker 3: but they're powerful and they could do some harm. I

678
00:38:16,680 --> 00:38:19,560
Speaker 3: think there needs to be some guidelines. Should we use

679
00:38:19,600 --> 00:38:22,600
Speaker 3: this for surveillance, should we use this for warfare? I

680
00:38:22,640 --> 00:38:25,600
Speaker 3: really do think there needs to be some policy down.

681
00:38:25,840 --> 00:38:29,760
Speaker 3: The Biden administration put I thought pretty gentle rules around AI.

682
00:38:29,920 --> 00:38:33,080
Speaker 3: If you're working at a cutting edge model, you need

683
00:38:33,120 --> 00:38:36,320
Speaker 3: to red teamate. That's an expression in tech for hiring

684
00:38:36,360 --> 00:38:38,560
Speaker 3: outsiders to try to break it, to try to look

685
00:38:38,600 --> 00:38:41,520
Speaker 3: for vulnerabilities before you release, itto the public. And then

686
00:38:41,680 --> 00:38:45,160
Speaker 3: the Biden administration was requiring these companies to share the

687
00:38:45,239 --> 00:38:48,920
Speaker 3: results with the government. The Trump administration and Trump himself

688
00:38:48,960 --> 00:38:51,839
Speaker 3: within twenty four hours got rid of that executive order,

689
00:38:52,239 --> 00:38:54,960
Speaker 3: so right now the view of the Trump administration.

690
00:38:55,080 --> 00:38:55,279
Speaker 1: JD.

691
00:38:55,440 --> 00:38:59,200
Speaker 3: Vance articulated it well in Paris in early January of

692
00:38:59,200 --> 00:39:02,759
Speaker 3: this year. Stop with the handwringing about AI. This is

693
00:39:02,800 --> 00:39:05,440
Speaker 3: a race with China. We need to win. China has

694
00:39:05,480 --> 00:39:07,920
Speaker 3: put out there that by the year twenty thirty, not

695
00:39:08,040 --> 00:39:10,960
Speaker 3: that far away. They plan on being dominant in AI,

696
00:39:11,080 --> 00:39:13,640
Speaker 3: and they are right behind us. They are nipping at

697
00:39:13,680 --> 00:39:16,480
Speaker 3: America's heels, and so there really is this sense like

698
00:39:16,719 --> 00:39:19,720
Speaker 3: if we put any speed bumps in the way of AI,

699
00:39:20,520 --> 00:39:22,880
Speaker 3: that could be hurting us. The flip side of that

700
00:39:23,160 --> 00:39:27,080
Speaker 3: is polling shows that most Americans are not excited about this,

701
00:39:27,160 --> 00:39:30,640
Speaker 3: but fearful of AI. And so my concern is that

702
00:39:30,680 --> 00:39:34,960
Speaker 3: these companies get too far ahead of where consumers are.

703
00:39:35,000 --> 00:39:37,759
Speaker 3: I mean, mistrust of tech is at a height anyway,

704
00:39:38,120 --> 00:39:40,239
Speaker 3: and something bad is neviitely going to happen. I'll make

705
00:39:40,280 --> 00:39:43,440
Speaker 3: one up that a trillion dollars is siphoned off from

706
00:39:43,520 --> 00:39:46,720
Speaker 3: the world financial system before single human could even notice

707
00:39:46,719 --> 00:39:49,839
Speaker 3: what's happening. So there'll be a moment where there's kind

708
00:39:49,840 --> 00:39:53,200
Speaker 3: of an AI disaster and stuff, and that could really

709
00:39:53,239 --> 00:39:55,839
Speaker 3: turn people off from AI. And that would be sad

710
00:39:55,880 --> 00:39:57,400
Speaker 3: to me as we've been talking about it. I think

711
00:39:57,440 --> 00:40:00,560
Speaker 3: there's a lot of potential in AI. Hates to see

712
00:40:00,600 --> 00:40:04,760
Speaker 3: it stunted or kind of adoption stunted because the companies

713
00:40:04,760 --> 00:40:09,239
Speaker 3: were so intent on profits cashing in that they gave

714
00:40:09,320 --> 00:40:11,920
Speaker 3: short shrift to trust and safety issues.

715
00:40:12,480 --> 00:40:15,799
Speaker 1: I think as this continues to unfold, I hope that

716
00:40:15,840 --> 00:40:18,080
Speaker 1: you're going to write another book and then you'll come

717
00:40:18,120 --> 00:40:20,759
Speaker 1: back and join us and continue to educate us. I

718
00:40:20,760 --> 00:40:23,279
Speaker 1: really want to thank you. This has been absolutely fascinating.

719
00:40:23,680 --> 00:40:28,160
Speaker 1: Your new book AI Valley, Microsoft Google and the Trillion

720
00:40:28,239 --> 00:40:32,320
Speaker 1: Dollar Race to Cash In on Artificial Intelligence is available

721
00:40:32,400 --> 00:40:36,120
Speaker 1: now on Amazon and in bookstores everywhere, and it's clearly

722
00:40:36,120 --> 00:40:39,360
Speaker 1: a very relevant book to exactly what's happening. And I

723
00:40:39,400 --> 00:40:42,439
Speaker 1: think anybody wanting to understand this is going to find

724
00:40:42,480 --> 00:40:44,200
Speaker 1: your book very helpful.

725
00:40:44,480 --> 00:40:46,279
Speaker 3: Oh my pleasure. This was a lot of fun. Thank you.

726
00:40:48,960 --> 00:40:50,520
Speaker 2: Thank you to my guest Gary Rivlin.

727
00:40:50,760 --> 00:40:52,600
Speaker 1: You can get a link to buy his new book

728
00:40:52,840 --> 00:40:56,840
Speaker 1: AI Valley, Microsoft Google and the Trillion Dollar Race to

729
00:40:56,920 --> 00:41:00,279
Speaker 1: Cash In and Artificial Intelligence on our show page at

730
00:41:00,320 --> 00:41:03,560
Speaker 1: newtsworld dot com. Neut World is produced by Ganglish three

731
00:41:03,600 --> 00:41:08,360
Speaker 1: sixty at iHeartMedia. Our executive producers Guardnzie Sloan. Our researcher

732
00:41:08,400 --> 00:41:11,720
Speaker 1: is Rachel Peterson. The artwork for the show who's created

733
00:41:11,719 --> 00:41:15,920
Speaker 1: by Steve Penley. Special thanks to the team at Gingwishtree sixty.

734
00:41:16,560 --> 00:41:18,880
Speaker 1: If you've been enjoying Nutsworld, I hope you'll go to

735
00:41:18,880 --> 00:41:22,360
Speaker 1: Apple Podcast and both rate us with five stars and

736
00:41:22,440 --> 00:41:24,960
Speaker 1: give us a review so others can learn what it's

737
00:41:24,960 --> 00:41:28,880
Speaker 1: all about. Right now, listeners of Newtsworld can sign up

738
00:41:28,880 --> 00:41:33,600
Speaker 1: for my three freeweekly columns at ginglishtree sixty dot com

739
00:41:33,640 --> 00:41:34,840
Speaker 1: slash newsletter.

740
00:41:35,320 --> 00:41:38,239
Speaker 2: I'm Newt Gingrich. This is Neutsworld