1
00:00:00,520 --> 00:00:04,320
Speaker 1: Welcome to Tech Stuff, a production of iHeart Podcasts and Kaleidoscope.

2
00:00:04,720 --> 00:00:07,160
Speaker 1: I'm mos Vloscian and today Cara Price and I will

3
00:00:07,160 --> 00:00:10,680
Speaker 1: bring you the headlines this week, including the fashion models

4
00:00:10,720 --> 00:00:14,320
Speaker 1: getting Ai twins. Then on Tech Support, we'll talk to

5
00:00:14,320 --> 00:00:17,720
Speaker 1: Garrett da vinc a reporter at the Washington Post about

6
00:00:17,760 --> 00:00:20,520
Speaker 1: all the news in AI from Open Ai, it's huge

7
00:00:20,520 --> 00:00:25,920
Speaker 1: new fundraise, to Xai's acquisition of x formerly Twitter, all

8
00:00:25,960 --> 00:00:27,320
Speaker 1: of that on the Weekend.

9
00:00:27,000 --> 00:00:37,400
Speaker 2: Tech It's Friday, April fourth, Cara Price.

10
00:00:37,440 --> 00:00:40,560
Speaker 3: As Volashan, you're looking like me today.

11
00:00:40,680 --> 00:00:43,400
Speaker 1: Yeah, we are matching.

12
00:00:43,720 --> 00:00:46,600
Speaker 3: We are very matchie matchie. I was at.

13
00:00:46,560 --> 00:00:49,600
Speaker 1: McLaren f one's headquarters as you know, earlier this week

14
00:00:49,720 --> 00:00:51,519
Speaker 1: and I sent you a picture that I was very

15
00:00:51,520 --> 00:00:55,760
Speaker 1: proud of me and Zach Brown, the CEO of McLaren Racing,

16
00:00:56,200 --> 00:00:59,280
Speaker 1: And your only comment was, oh, black trainers.

17
00:00:58,920 --> 00:01:02,280
Speaker 3: Always black six sneakers. You can sponsor us, but yes,

18
00:01:02,360 --> 00:01:04,680
Speaker 3: always black A six sneakers, even if he's wearing a tie.

19
00:01:05,800 --> 00:01:08,120
Speaker 3: But it's, you know, for a tech podcast, it's very

20
00:01:08,160 --> 00:01:09,880
Speaker 3: funny how much we talk about fashion. And it's not

21
00:01:09,920 --> 00:01:10,680
Speaker 3: because I'm a woman.

22
00:01:10,920 --> 00:01:13,120
Speaker 1: I mean, it is also ironic that we both have

23
00:01:13,240 --> 00:01:16,960
Speaker 1: our own uniforms. We're wearing blue blue shirts, dark blue pants, shirt,

24
00:01:17,000 --> 00:01:20,520
Speaker 1: blubby pants, and a cool hat.

25
00:01:20,560 --> 00:01:23,280
Speaker 3: Though I am wearing a cool hat, I'm always wearing well,

26
00:01:23,319 --> 00:01:24,760
Speaker 3: I wouldn't call it a cool hat, but if you

27
00:01:24,760 --> 00:01:27,600
Speaker 3: want to call it a cool hat, it's an appropriated

28
00:01:27,840 --> 00:01:31,080
Speaker 3: Yankees city by hat that is completely illegal and was

29
00:01:31,080 --> 00:01:33,280
Speaker 3: taken off the internet the minute it went up, So

30
00:01:33,480 --> 00:01:34,960
Speaker 3: I'm very proud to be wearing it now.

31
00:01:35,000 --> 00:01:36,959
Speaker 1: The reason we're talking about this is not just because

32
00:01:37,000 --> 00:01:40,520
Speaker 1: we're totally self indulgent, but because the topic of tech

33
00:01:40,640 --> 00:01:44,680
Speaker 1: folks and their uniforms is pretty fascinating. Earlier this year,

34
00:01:44,680 --> 00:01:48,160
Speaker 1: we talked about Mark Zuckerberg, who had this famous, weirdly

35
00:01:48,240 --> 00:01:51,880
Speaker 1: shaped T shirt with the Latin phrase out zuk out

36
00:01:52,000 --> 00:01:54,880
Speaker 1: nihil printed on it, which means all zook ll all

37
00:01:54,920 --> 00:01:58,400
Speaker 1: nothing and is a is a reference to Julius Caesar.

38
00:01:58,520 --> 00:01:59,680
Speaker 1: Of course I can't.

39
00:02:00,080 --> 00:02:04,960
Speaker 3: It's just insane to me. The level of It's not narcissism.

40
00:02:05,080 --> 00:02:09,520
Speaker 3: It's this sense of like an obsession with being cool that,

41
00:02:09,680 --> 00:02:14,359
Speaker 3: while chased, is the antithesis of anything cool and fashionable.

42
00:02:14,639 --> 00:02:17,679
Speaker 1: I thought that one of the genuinely coolest tech swipes

43
00:02:17,720 --> 00:02:20,760
Speaker 1: of the year was when Jay Graber, who's the CEO

44
00:02:20,919 --> 00:02:24,200
Speaker 1: of blue Sky, the kind of left leaning social media platform,

45
00:02:24,520 --> 00:02:28,320
Speaker 1: wore a T shirt at south By Southwest that said

46
00:02:28,680 --> 00:02:32,480
Speaker 1: in Latin, no more Caesars, and it was a real

47
00:02:32,639 --> 00:02:35,880
Speaker 1: sort of shots fired at Zuck, I think, and.

48
00:02:35,800 --> 00:02:38,080
Speaker 3: Also just again in Latin. I mean, these are the

49
00:02:38,120 --> 00:02:39,720
Speaker 3: nerdiest people in human history.

50
00:02:39,919 --> 00:02:42,280
Speaker 1: This was a huge best seller for Blue Sky. I

51
00:02:42,320 --> 00:02:43,359
Speaker 1: saw that almost immediately.

52
00:02:43,440 --> 00:02:46,720
Speaker 3: I bet it was, you know, it's tech Titans, as

53
00:02:46,760 --> 00:02:48,840
Speaker 3: I was saying, just seemed to want to have a

54
00:02:48,880 --> 00:02:51,160
Speaker 3: signature look. And I don't know if part of it

55
00:02:51,240 --> 00:02:58,240
Speaker 3: is pathology, psychological pathology, I mean, certainly minus psychological pathology.

56
00:02:58,280 --> 00:03:00,040
Speaker 3: I like to wear the same thing every day. I

57
00:03:00,040 --> 00:03:02,600
Speaker 3: don't have to think about it. Like our friend Elizabeth Holmes,

58
00:03:02,600 --> 00:03:04,720
Speaker 3: who also rocked the Black Turble.

59
00:03:04,560 --> 00:03:05,960
Speaker 1: She flex hard on the Steve Jobs.

60
00:03:06,200 --> 00:03:09,800
Speaker 3: She flexed and copped. She coughed something from Steve Jobs.

61
00:03:09,840 --> 00:03:11,360
Speaker 3: And I guess it's sort of worked for her because

62
00:03:11,360 --> 00:03:12,440
Speaker 3: we're still talking about it.

63
00:03:12,480 --> 00:03:14,600
Speaker 1: But I think the most sort of talked about tech

64
00:03:14,680 --> 00:03:17,160
Speaker 1: CEO of the moment is none other than Jensen Hwang

65
00:03:17,320 --> 00:03:18,840
Speaker 1: of Invidia.

66
00:03:18,520 --> 00:03:20,040
Speaker 3: The man who made a lot of people a lot

67
00:03:20,040 --> 00:03:21,839
Speaker 3: of money and now is making a lot of people

68
00:03:21,880 --> 00:03:22,800
Speaker 3: a little less money.

69
00:03:22,720 --> 00:03:25,600
Speaker 1: And still still a lot, but little less exactly. He

70
00:03:25,760 --> 00:03:29,079
Speaker 1: has a signature. Look, can you describe it.

71
00:03:29,080 --> 00:03:32,120
Speaker 3: It's sort of a Marlon Brando, sort of on the

72
00:03:32,160 --> 00:03:36,480
Speaker 3: waterfront black leather jacket that is I think Tom Ford.

73
00:03:36,680 --> 00:03:38,440
Speaker 1: I think it is Tom forty, which is like a

74
00:03:38,480 --> 00:03:39,320
Speaker 1: ten thousand dollars.

75
00:03:39,320 --> 00:03:41,440
Speaker 3: That's an extremely expensive jacket.

76
00:03:41,680 --> 00:03:43,960
Speaker 1: But there's a lot of Invidia stands. It turns out,

77
00:03:44,080 --> 00:03:46,880
Speaker 1: and if you build it, they will come to buy

78
00:03:46,920 --> 00:03:48,120
Speaker 1: a fake black leather jacket.

79
00:03:48,360 --> 00:03:49,640
Speaker 3: If you build it, they will come and buy it

80
00:03:49,640 --> 00:03:50,960
Speaker 3: a lot cheaper exactly.

81
00:03:51,080 --> 00:03:53,800
Speaker 1: So our friends at four or for Media reported that

82
00:03:53,840 --> 00:03:56,920
Speaker 1: there are all these websites popping up with knockoff Jensen

83
00:03:56,960 --> 00:04:01,240
Speaker 1: Hwang in video jackets for which have creative titles like

84
00:04:01,320 --> 00:04:05,960
Speaker 1: Jensen wang in Video Jacket totally optimized their CEO.

85
00:04:06,080 --> 00:04:07,960
Speaker 3: People, They're going to find my phone and be like

86
00:04:08,160 --> 00:04:09,960
Speaker 3: that was the last thing that she was looking at.

87
00:04:10,960 --> 00:04:12,920
Speaker 1: But I think I don't sure if you saw it.

88
00:04:12,920 --> 00:04:15,600
Speaker 1: But this week Jensen Hwang went to visit a company

89
00:04:15,720 --> 00:04:19,120
Speaker 1: called x one Robots, and they had one of their

90
00:04:19,200 --> 00:04:22,279
Speaker 1: robots come up to Jensen whanging humanoid robot and present

91
00:04:22,360 --> 00:04:25,120
Speaker 1: him with a new black leather jacket, not a tom

92
00:04:25,240 --> 00:04:29,000
Speaker 1: Ford one. This one was bedazzled with the Nvidia stock

93
00:04:29,080 --> 00:04:32,159
Speaker 1: ticker on the front left pocket and the logo in

94
00:04:32,240 --> 00:04:33,920
Speaker 1: kind of Swarovsky style crystals on.

95
00:04:33,839 --> 00:04:36,839
Speaker 3: The back to imagine if the humanoid robot is also sentient,

96
00:04:37,080 --> 00:04:39,120
Speaker 3: if the robot was like, I don't want to do this,

97
00:04:40,640 --> 00:04:43,040
Speaker 3: like the people programming mean, please give me something better

98
00:04:43,080 --> 00:04:43,360
Speaker 3: to do.

99
00:04:44,320 --> 00:04:46,599
Speaker 1: But there's a more serious story about fashion AI here,

100
00:04:46,680 --> 00:04:49,240
Speaker 1: and it comes from The Guardian. So according to their story,

101
00:04:49,320 --> 00:04:51,520
Speaker 1: the clothing retailer H and M announced that they're going

102
00:04:51,560 --> 00:04:55,200
Speaker 1: to create so called AI twins of thirty models, which

103
00:04:55,200 --> 00:04:57,400
Speaker 1: they're going to use for social media marketing posts.

104
00:04:57,440 --> 00:04:59,320
Speaker 3: The way, I wish I had this for a podcast.

105
00:04:59,480 --> 00:05:02,800
Speaker 3: As much as I enjoy doing it with you, I wish.

106
00:05:02,680 --> 00:05:06,760
Speaker 1: You could just have a I'd have a relic. Would

107
00:05:06,760 --> 00:05:08,560
Speaker 1: you rather use your replica or my replica?

108
00:05:09,080 --> 00:05:10,480
Speaker 3: Well, if we wanted me to be on time, I

109
00:05:10,480 --> 00:05:13,919
Speaker 3: would use your replica. But sorry, keep going, because I

110
00:05:13,920 --> 00:05:14,600
Speaker 3: do love the story.

111
00:05:14,720 --> 00:05:17,960
Speaker 1: No, So these models, the real models, have given their

112
00:05:17,960 --> 00:05:20,719
Speaker 1: permission to H and M to use their likeness with AI.

113
00:05:21,600 --> 00:05:23,359
Speaker 1: You know again, as you said, sounds good, you know,

114
00:05:23,400 --> 00:05:25,880
Speaker 1: why why sharp yourself you can send your digital twin.

115
00:05:26,160 --> 00:05:28,880
Speaker 3: Well that and also I think there's something that's very

116
00:05:28,920 --> 00:05:31,520
Speaker 3: interesting here that reminds me of the Atlantic sort of

117
00:05:31,520 --> 00:05:34,400
Speaker 3: selling its data, which is these are people that, like

118
00:05:34,480 --> 00:05:37,839
Speaker 3: in the film industry, are cannibalizing their own job.

119
00:05:39,560 --> 00:05:42,680
Speaker 1: Well, of course, just like with the Atlantic and Open Ai,

120
00:05:43,200 --> 00:05:46,320
Speaker 1: where open Ai actually compensates the Atlantic for using its archive,

121
00:05:46,880 --> 00:05:50,280
Speaker 1: these models are also being compensated for their image. They

122
00:05:50,279 --> 00:05:52,920
Speaker 1: own the rights of their twins, and their twins can

123
00:05:52,920 --> 00:05:55,880
Speaker 1: work for other brands, not just H and M, and

124
00:05:56,160 --> 00:05:59,440
Speaker 1: they'll get paid. One of the thirty models said, quote,

125
00:05:59,600 --> 00:06:01,320
Speaker 1: she's like me without the jet lag.

126
00:06:04,320 --> 00:06:06,400
Speaker 3: So the other thing is, and it's worth saying, the

127
00:06:06,440 --> 00:06:10,320
Speaker 3: images in the business of fashion right up look amazingly similar.

128
00:06:10,360 --> 00:06:11,479
Speaker 1: Like it's not distinguishable.

129
00:06:11,640 --> 00:06:13,520
Speaker 3: Yeah, it's not like who was that person that we

130
00:06:13,600 --> 00:06:16,880
Speaker 3: used to cover on Sleepwalkers. She was a model Lil Michayla.

131
00:06:17,000 --> 00:06:20,240
Speaker 3: Oh yeah, like littl Michayla looked like AI. I think

132
00:06:20,279 --> 00:06:22,760
Speaker 3: you know, after four years, we have now gotten to

133
00:06:22,800 --> 00:06:27,000
Speaker 3: a place where it's just an AI replica, which is incredible.

134
00:06:27,200 --> 00:06:30,480
Speaker 1: Yeah, and there's no sick finger. As you pointed out,

135
00:06:30,640 --> 00:06:32,640
Speaker 1: it does raise concerns about the future of modeling in

136
00:06:32,640 --> 00:06:36,479
Speaker 1: the fashion industry because although these models who are participation

137
00:06:36,520 --> 00:06:38,840
Speaker 1: in the partnership will get paid every time that digital

138
00:06:38,839 --> 00:06:41,279
Speaker 1: twin is used, you know, just like in the film industry,

139
00:06:41,320 --> 00:06:43,920
Speaker 1: also raises questions about what happens to all the people

140
00:06:43,960 --> 00:06:45,840
Speaker 1: who work on sets. I mean the hair siders, the

141
00:06:45,880 --> 00:06:47,400
Speaker 1: makeup artist, the lighting designers.

142
00:06:47,560 --> 00:06:49,840
Speaker 3: Yeah, you know, I was actually talking to our producer Eliza.

143
00:06:49,920 --> 00:06:52,240
Speaker 3: Shout out to Eliza, I like to keep her in

144
00:06:52,279 --> 00:06:55,360
Speaker 3: the mix. And she actually knows a hairstylist who works

145
00:06:55,360 --> 00:06:58,360
Speaker 3: with a company that's testing out replacing many of their

146
00:06:58,520 --> 00:07:02,640
Speaker 3: e commerce models with AI, and you know, work has

147
00:07:02,680 --> 00:07:04,080
Speaker 3: slowed way down for her.

148
00:07:04,480 --> 00:07:05,239
Speaker 1: So this is happening.

149
00:07:05,560 --> 00:07:07,560
Speaker 3: I mean, think about it. You don't need to be

150
00:07:07,640 --> 00:07:09,239
Speaker 3: doing a blood on a digital avatar.

151
00:07:09,720 --> 00:07:11,800
Speaker 1: Yeah, and I mean even for models themselves, if I

152
00:07:11,800 --> 00:07:13,640
Speaker 1: guess if you're you know, for these thirty who have

153
00:07:13,960 --> 00:07:16,360
Speaker 1: gotten first through the door. It's one thing, but this

154
00:07:16,440 --> 00:07:20,000
Speaker 1: may affect the kind of future job landscape for model

155
00:07:20,040 --> 00:07:21,680
Speaker 1: as well as people work around modeling.

156
00:07:22,520 --> 00:07:25,360
Speaker 3: But this is like a This is Zulander three where

157
00:07:25,400 --> 00:07:30,800
Speaker 3: Derek realizes we're in jeopardy. But in all seriousness, you know,

158
00:07:30,880 --> 00:07:34,120
Speaker 3: what what does it look like for models who are

159
00:07:34,120 --> 00:07:36,120
Speaker 3: no longer in demand? And it might not seem important

160
00:07:36,160 --> 00:07:39,720
Speaker 3: to like a lay person, but I do think it's

161
00:07:39,760 --> 00:07:42,440
Speaker 3: a harbinger of things to come, which is like, if

162
00:07:42,480 --> 00:07:45,800
Speaker 3: you can replace something that is so sort of ubiquitous

163
00:07:45,840 --> 00:07:49,160
Speaker 3: for you know, a century, what does that look like?

164
00:07:49,600 --> 00:07:51,840
Speaker 1: Yeah, I mean wile I find most interesting about these

165
00:07:51,880 --> 00:07:54,520
Speaker 1: like AI digital twins or AI actors or whatever, or

166
00:07:55,040 --> 00:07:58,360
Speaker 1: you know, chatbots of famous people from history is their

167
00:07:58,400 --> 00:08:01,440
Speaker 1: effect is kind of to lock in the very few

168
00:08:01,520 --> 00:08:04,480
Speaker 1: most famous people in the world as the only characters

169
00:08:04,520 --> 00:08:07,480
Speaker 1: who are worth interacting with. I mean, you know, if

170
00:08:07,480 --> 00:08:10,280
Speaker 1: you think about an action movie, why would you not

171
00:08:10,280 --> 00:08:12,120
Speaker 1: make it with Tom Cruise? Or talking about doing a

172
00:08:12,120 --> 00:08:14,240
Speaker 1: fashion shoot, why would you not make it with el McPherson.

173
00:08:14,280 --> 00:08:16,600
Speaker 1: So kind of, I think there's a longer term kind

174
00:08:16,600 --> 00:08:19,080
Speaker 1: of chilling effect on the pipeline and new talent in

175
00:08:19,160 --> 00:08:22,480
Speaker 1: creative industries, which you know, which will pretty interesting, you know,

176
00:08:22,560 --> 00:08:24,960
Speaker 1: and somewhat disturbing. To be fair to H and M,

177
00:08:24,960 --> 00:08:27,200
Speaker 1: they're being very upfront about their use of AI. They're

178
00:08:27,200 --> 00:08:29,600
Speaker 1: going to watermark the images of the AI twins in

179
00:08:29,640 --> 00:08:32,600
Speaker 1: their ads so people will know they're the AI versions,

180
00:08:33,120 --> 00:08:35,160
Speaker 1: And by doing this, the company will also be abiding

181
00:08:35,240 --> 00:08:37,959
Speaker 1: with the EU's AI Act, which is coming into effect

182
00:08:38,000 --> 00:08:41,000
Speaker 1: in twenty twenty six, and it will require all AI

183
00:08:41,160 --> 00:08:44,000
Speaker 1: images to be labeled as AI images, to.

184
00:08:44,000 --> 00:08:47,000
Speaker 3: Which I say, who's looking for that? Like, I you

185
00:08:47,040 --> 00:08:48,480
Speaker 3: know what I mean? Like the role of AI and

186
00:08:48,559 --> 00:08:50,800
Speaker 3: creative industries, which you know, is something I'm obsessed with

187
00:08:50,800 --> 00:08:53,040
Speaker 3: and sort of how to regulate it is something that's

188
00:08:53,080 --> 00:08:56,160
Speaker 3: obviously going to keep coming up, and I'm curious to

189
00:08:56,160 --> 00:08:58,560
Speaker 3: see where it goes next. I wonder to what extent

190
00:08:58,720 --> 00:09:02,680
Speaker 3: the lay person cares that, Like where they're buying a shirt,

191
00:09:03,120 --> 00:09:06,240
Speaker 3: is either modeling that shirt with a fake twin or

192
00:09:06,280 --> 00:09:07,440
Speaker 3: the real person.

193
00:09:07,320 --> 00:09:10,880
Speaker 1: Probably not, Probably not. The next headline comes from the

194
00:09:10,880 --> 00:09:14,439
Speaker 1: stuff of nightmares for anyone who's a frequent traveler, and

195
00:09:14,520 --> 00:09:16,520
Speaker 1: it has to do me, and it has to do

196
00:09:16,600 --> 00:09:18,720
Speaker 1: with an airport being shut down for twenty four hours.

197
00:09:18,720 --> 00:09:19,520
Speaker 3: I heard it was Heathrow.

198
00:09:19,640 --> 00:09:21,880
Speaker 1: It was London heath Throw Airport, where I flew out

199
00:09:21,880 --> 00:09:24,559
Speaker 1: of just this week. It shut down, leading to over

200
00:09:24,600 --> 00:09:28,559
Speaker 1: a thousand council flights after a fire caused a power outage,

201
00:09:28,880 --> 00:09:31,400
Speaker 1: and Bloomberg reported that the outage could be traced back

202
00:09:31,440 --> 00:09:35,559
Speaker 1: to a single point of failure, a burned transformer and

203
00:09:35,640 --> 00:09:39,800
Speaker 1: twenty five thousand liters of transformer cooling oil that was

204
00:09:39,840 --> 00:09:42,760
Speaker 1: a blaze for several hours. It's a fascinating story, and

205
00:09:42,800 --> 00:09:45,640
Speaker 1: Bloomberg had such a great headline, which was the device

206
00:09:45,920 --> 00:09:50,000
Speaker 1: throttling the world's electrified future. But Kara, do you know

207
00:09:50,040 --> 00:09:51,120
Speaker 1: what a transformer is?

208
00:09:52,120 --> 00:09:54,439
Speaker 3: I know what the movie Transformers is. I know it's

209
00:09:54,440 --> 00:09:55,880
Speaker 3: a car that turns into a man.

210
00:09:56,480 --> 00:10:00,000
Speaker 1: So most simply put, a transformer is something that changed

211
00:10:00,320 --> 00:10:04,000
Speaker 1: is voltage. So when you create electricity in a power plant,

212
00:10:04,320 --> 00:10:06,600
Speaker 1: you actually want to increase the voltage as much as

213
00:10:06,640 --> 00:10:09,040
Speaker 1: you can because that allows it to travel far and

214
00:10:09,200 --> 00:10:12,480
Speaker 1: fast with less loss of electricity along the way. So

215
00:10:12,920 --> 00:10:15,360
Speaker 1: use a transformer on the way out of the power plant.

216
00:10:15,640 --> 00:10:17,240
Speaker 1: But then when it gets to your home or the

217
00:10:17,280 --> 00:10:19,400
Speaker 1: local electric grid or whatever it may be, you actually

218
00:10:19,400 --> 00:10:21,880
Speaker 1: want to use a transformer to turn the voltage back

219
00:10:21,920 --> 00:10:24,200
Speaker 1: down because otherwise it blows up your stuff.

220
00:10:24,559 --> 00:10:27,720
Speaker 3: So ostensibly, it's sort of like when I'm staying at

221
00:10:28,600 --> 00:10:33,280
Speaker 3: not the greatest hotel and I turn a blow dryer

222
00:10:33,360 --> 00:10:36,079
Speaker 3: on and all of a sudden, the entire room short

223
00:10:36,120 --> 00:10:38,640
Speaker 3: circuits and all of the electricity goes off because the

224
00:10:38,720 --> 00:10:39,760
Speaker 3: voltage is too high.

225
00:10:39,960 --> 00:10:42,080
Speaker 1: Yeah, exactly mean that is a short circuit where there's

226
00:10:42,200 --> 00:10:45,360
Speaker 1: likely been a problem with the transformer being bypassed or

227
00:10:45,400 --> 00:10:46,319
Speaker 1: not functioning correctly.

228
00:10:46,480 --> 00:10:48,000
Speaker 3: Right. And I think that in a storm or a

229
00:10:48,080 --> 00:10:51,840
Speaker 3: natural disaster, they can sometimes explode and it's very loud.

230
00:10:51,920 --> 00:10:54,160
Speaker 1: Yeah, it sounds like fireworks or a bomb going off.

231
00:10:54,200 --> 00:10:56,880
Speaker 1: And I mean during hurricanes and other kind of natural disasters,

232
00:10:56,920 --> 00:10:59,160
Speaker 1: these things come under a lot of pressure when they

233
00:10:59,160 --> 00:11:01,800
Speaker 1: do go out. Follow And so what happened to Heathrow

234
00:11:01,840 --> 00:11:04,760
Speaker 1: Airport was this fire broke out in a substation which

235
00:11:04,760 --> 00:11:08,240
Speaker 1: houses transformers, and it took the firefighters seven hours to

236
00:11:08,240 --> 00:11:10,439
Speaker 1: get it under control. The airport, in order to come

237
00:11:10,440 --> 00:11:13,280
Speaker 1: back online, was able to accept power from other substations,

238
00:11:13,760 --> 00:11:15,199
Speaker 1: but even the time it took them to do that,

239
00:11:15,360 --> 00:11:18,800
Speaker 1: many many flights were canceled. It was chaos.

240
00:11:19,840 --> 00:11:22,080
Speaker 3: She made it, but it was not a pretty team.

241
00:11:22,160 --> 00:11:25,200
Speaker 1: Yeah him. Back in twenty thirteen, there was actually a

242
00:11:25,240 --> 00:11:29,400
Speaker 1: sniper attack on a substation in California which caused a

243
00:11:29,440 --> 00:11:32,800
Speaker 1: fire that burned seventeen transformers and almost knocked out all

244
00:11:32,800 --> 00:11:35,480
Speaker 1: of the power of Silicon Valley. Now This actually led

245
00:11:35,520 --> 00:11:38,760
Speaker 1: to a lot more security around substations and even stockpiling

246
00:11:38,800 --> 00:11:43,360
Speaker 1: of transformers, which of course sparked another problem, a shortage

247
00:11:43,360 --> 00:11:46,679
Speaker 1: of transformers, transformer shortage and obviously with the supply chain

248
00:11:46,720 --> 00:11:51,160
Speaker 1: issues recently, the lead time for delivering a new large

249
00:11:51,200 --> 00:11:55,240
Speaker 1: transformer is now about three to five years for a

250
00:11:55,280 --> 00:11:58,520
Speaker 1: single transformer, and bear in mind these can be huge. Nonetheless,

251
00:11:58,559 --> 00:12:00,800
Speaker 1: the scale of transformer that was Heathrow, it takes a

252
00:12:00,840 --> 00:12:03,679
Speaker 1: long hundre replace and they've got a much more expensive

253
00:12:03,760 --> 00:12:05,520
Speaker 1: So you know, now we're living in the error of

254
00:12:05,559 --> 00:12:09,120
Speaker 1: EVS and the AI boom powered by data centers and

255
00:12:09,280 --> 00:12:12,720
Speaker 1: transformers are also required to bring renewable power onto the grid,

256
00:12:13,080 --> 00:12:15,040
Speaker 1: and so you know, one of the interesting implications of

257
00:12:15,080 --> 00:12:18,480
Speaker 1: this story, as Imaza, who's on the show, recently told

258
00:12:18,520 --> 00:12:21,960
Speaker 1: me that actually the US's struggles to onboard new power

259
00:12:21,960 --> 00:12:24,880
Speaker 1: to the grid maybe the reason why the US ultimately

260
00:12:24,880 --> 00:12:26,280
Speaker 1: falls behind China in AI.

261
00:12:26,840 --> 00:12:29,840
Speaker 3: So it's ironic actually that the increase in transformer prices

262
00:12:29,880 --> 00:12:33,160
Speaker 3: could be further increased by President Trump's on again, off

263
00:12:33,200 --> 00:12:36,400
Speaker 3: again relationship with imposing tariffs in Canada and Mexico, which

264
00:12:36,440 --> 00:12:39,520
Speaker 3: is where we import a lot of our large transformers from.

265
00:12:39,520 --> 00:12:41,320
Speaker 1: And that's why I love this Bloomberg story because it's

266
00:12:41,320 --> 00:12:43,680
Speaker 1: fun think about all the sexy stuff like new chips

267
00:12:43,679 --> 00:12:46,640
Speaker 1: and data center construction, you know, but there's still this

268
00:12:46,679 --> 00:12:49,920
Speaker 1: one hundred year old technology that hasn't changed and that

269
00:12:50,040 --> 00:12:53,960
Speaker 1: has to be imported and which is absolutely critical for infrastructure,

270
00:12:54,040 --> 00:12:55,760
Speaker 1: digital and otherwise, you know.

271
00:12:55,720 --> 00:12:58,840
Speaker 3: Speaking of the AI boom and things that use a lot,

272
00:12:58,880 --> 00:13:01,000
Speaker 3: a lot, a lot of energy. This next story that

273
00:13:01,000 --> 00:13:02,920
Speaker 3: I want to tell you is for those people who've

274
00:13:02,920 --> 00:13:06,679
Speaker 3: been sitting here thinking about lms and not really understanding

275
00:13:06,679 --> 00:13:11,040
Speaker 3: how they work. I have news for you. Much like

276
00:13:11,080 --> 00:13:17,040
Speaker 3: Trump's decisions on tariffs, nobody knows how they work. And Anthropic,

277
00:13:17,120 --> 00:13:19,680
Speaker 3: which is the same company that makes the AI model Claud,

278
00:13:20,000 --> 00:13:22,840
Speaker 3: has been trying to figure out how the hell these things.

279
00:13:22,840 --> 00:13:25,040
Speaker 1: This is to solve the so called black books.

280
00:13:24,720 --> 00:13:27,040
Speaker 3: The black box problem. So Anthropic, which is the same

281
00:13:27,080 --> 00:13:29,679
Speaker 3: company that makes the A model Claud, has been trying

282
00:13:29,720 --> 00:13:32,880
Speaker 3: to figure out sort of what's under the hood, and

283
00:13:32,960 --> 00:13:36,679
Speaker 3: they recently released two reports on how llms do things

284
00:13:36,720 --> 00:13:42,800
Speaker 3: like complete sentences, solve math problems, and suppress hallucinations. And

285
00:13:42,840 --> 00:13:45,840
Speaker 3: they use a technique called circuit tracing which let them

286
00:13:45,920 --> 00:13:49,520
Speaker 3: track an LM's decision making process for ten different tasks

287
00:13:49,800 --> 00:13:52,680
Speaker 3: by working back from the solution to the query.

288
00:13:53,040 --> 00:13:56,280
Speaker 1: Huh okay, that makes sense. But I just think more broadly,

289
00:13:56,480 --> 00:13:59,360
Speaker 1: as more and more decisions are taken for us by

290
00:13:59,679 --> 00:14:03,640
Speaker 1: AI or outcomes kind of determined by AI, it's kind

291
00:14:03,640 --> 00:14:06,600
Speaker 1: of remarkable that this huge elephant is still in the

292
00:14:06,640 --> 00:14:09,560
Speaker 1: corner of the room, which is we can't understand how

293
00:14:09,559 --> 00:14:10,520
Speaker 1: they make their decisions.

294
00:14:10,640 --> 00:14:12,560
Speaker 3: But it also makes you think of like, I don't

295
00:14:12,600 --> 00:14:14,840
Speaker 3: really know. I mean, I know how a car works,

296
00:14:15,000 --> 00:14:19,640
Speaker 3: but like the I'm not an engineer, and yet increasingly,

297
00:14:19,880 --> 00:14:22,600
Speaker 3: I mean, you know, since whenever the nineteen twenties, people

298
00:14:22,640 --> 00:14:24,480
Speaker 3: have used cars more and more to get around, and

299
00:14:24,480 --> 00:14:26,760
Speaker 3: we're just kind of like, well, this thing's gonna work

300
00:14:26,800 --> 00:14:29,160
Speaker 3: until it blows up, you know. That's why I feel whatever.

301
00:14:29,400 --> 00:14:33,680
Speaker 3: But it's important to note that lllm's, like Claude, which

302
00:14:33,760 --> 00:14:37,040
Speaker 3: was the focus of this study, are trained, which I

303
00:14:37,080 --> 00:14:40,400
Speaker 3: thought this was really interesting. They're trained, not programmed on

304
00:14:40,440 --> 00:14:43,520
Speaker 3: a bunch of data. They create their own rules based

305
00:14:43,560 --> 00:14:46,600
Speaker 3: on the data they ingest. But up until now we

306
00:14:46,760 --> 00:14:49,560
Speaker 3: haven't been able to see into the models to know

307
00:14:49,600 --> 00:14:52,320
Speaker 3: what those rules actually are. Let alone how the models

308
00:14:52,400 --> 00:14:53,080
Speaker 3: generate them.

309
00:14:53,240 --> 00:14:55,800
Speaker 1: Yeah, and I think what the work the anthropic is

310
00:14:55,840 --> 00:14:59,880
Speaker 1: doing is all about basically understanding decision making. So then

311
00:15:00,120 --> 00:15:02,040
Speaker 1: not yet at the stage of being able to understand

312
00:15:02,040 --> 00:15:04,480
Speaker 1: how the models generate their own rules, But I think

313
00:15:04,520 --> 00:15:07,720
Speaker 1: this story is all about how it's starting to become

314
00:15:08,520 --> 00:15:12,480
Speaker 1: easier or possible to basically backfigure out how a model

315
00:15:12,480 --> 00:15:13,240
Speaker 1: has made a decision.

316
00:15:13,400 --> 00:15:18,520
Speaker 3: Yes, and it's not that simple. I think the researchers

317
00:15:18,560 --> 00:15:21,880
Speaker 3: in this particular case were inspired by brain scan techniques

318
00:15:22,240 --> 00:15:25,360
Speaker 3: that are used in neuroscience. They found that llms, which

319
00:15:25,400 --> 00:15:28,520
Speaker 3: again I'm so interested in how we anthropomorphize lms, But

320
00:15:28,840 --> 00:15:33,080
Speaker 3: they found that llms store different constellations of knowledge in

321
00:15:33,120 --> 00:15:35,400
Speaker 3: different parts of their model. So, for example, the concept

322
00:15:35,440 --> 00:15:39,800
Speaker 3: of smallness or the idea of a rabbit. Anthropic was

323
00:15:39,840 --> 00:15:43,560
Speaker 3: actually able to identify and then turn off certain parts

324
00:15:43,600 --> 00:15:47,120
Speaker 3: of the model. So like the idea rabbit and tune

325
00:15:47,120 --> 00:15:49,160
Speaker 3: it down so that it couldn't be part of a

326
00:15:49,240 --> 00:15:50,160
Speaker 3: queer result.

327
00:15:50,360 --> 00:15:52,760
Speaker 1: And for exac like what eats carrots? It would be like.

328
00:15:54,080 --> 00:15:58,080
Speaker 3: People right, a dog right exactly, And so the same

329
00:15:58,200 --> 00:16:01,480
Speaker 3: query would have a different answer if the rabbit part

330
00:16:01,520 --> 00:16:03,280
Speaker 3: of the model was dealt up or down, which you.

331
00:16:03,280 --> 00:16:06,720
Speaker 1: Say said by contrast, you might you might say, you know,

332
00:16:07,000 --> 00:16:09,000
Speaker 1: what's a mammal? It always asks rabbit if you dial

333
00:16:09,000 --> 00:16:10,240
Speaker 1: it up, and if you if you dial it down,

334
00:16:10,240 --> 00:16:11,120
Speaker 1: it would never own a rabbit.

335
00:16:11,280 --> 00:16:11,640
Speaker 4: Correct.

336
00:16:11,720 --> 00:16:13,680
Speaker 3: So I mean, and it's similar. I mean it's similar

337
00:16:13,680 --> 00:16:15,960
Speaker 3: in human beings, which is like I don't drink anymore,

338
00:16:15,960 --> 00:16:17,560
Speaker 3: so I go to a bar and I'm like water.

339
00:16:19,520 --> 00:16:20,680
Speaker 3: It's a similar kind of thing.

340
00:16:21,120 --> 00:16:23,080
Speaker 1: I mean that that actually clarifies for me, and I

341
00:16:23,240 --> 00:16:25,520
Speaker 1: kind of We've done a bunch of coverage of AI

342
00:16:25,600 --> 00:16:28,000
Speaker 1: and spoken to Jeffrey Hinton and others, but like that

343
00:16:28,120 --> 00:16:31,640
Speaker 1: idea of a neural network is to me really clarified

344
00:16:31,640 --> 00:16:32,280
Speaker 1: by this study.

345
00:16:32,400 --> 00:16:34,080
Speaker 3: Well, and how could it not be. It's something that

346
00:16:34,080 --> 00:16:36,000
Speaker 3: we've created, you know what I mean, it's going to

347
00:16:36,040 --> 00:16:40,400
Speaker 3: reflect a sort of human centric way of working. I

348
00:16:40,440 --> 00:16:44,480
Speaker 3: think this story actually came from the MIT Technology Review

349
00:16:44,480 --> 00:16:47,800
Speaker 3: with the headline anthropic can now track the bizarre inner

350
00:16:47,840 --> 00:16:51,280
Speaker 3: workings of a large language model, to which I said, Now,

351
00:16:52,480 --> 00:16:54,280
Speaker 3: what really blew my mind, though, is the way that

352
00:16:54,360 --> 00:16:57,280
Speaker 3: this thing solves math problems, because it's not the way

353
00:16:57,280 --> 00:16:58,200
Speaker 3: that humans do math.

354
00:16:58,320 --> 00:16:59,400
Speaker 1: Okay, tell me more.

355
00:16:59,360 --> 00:17:04,400
Speaker 3: So, research asked it to solve the equation thirty six

356
00:17:04,480 --> 00:17:07,640
Speaker 3: plus fifty nine, and while Claude came up with there,

357
00:17:07,760 --> 00:17:13,880
Speaker 3: what's the correct answer? Well, well, Claude took a very

358
00:17:13,920 --> 00:17:17,440
Speaker 3: circuitous route to get that answer. Anthropic found that Claude

359
00:17:17,520 --> 00:17:22,280
Speaker 3: used multiple computation paths in parallel to get its final answer,

360
00:17:22,359 --> 00:17:25,280
Speaker 3: unlike you, who just used your brain. So one path

361
00:17:25,359 --> 00:17:28,720
Speaker 3: added a bunch of numbers. I love this. This is

362
00:17:28,760 --> 00:17:31,600
Speaker 3: like nerd alert close to thirty six and fifty nine

363
00:17:31,720 --> 00:17:36,000
Speaker 3: to approximate the total like thirty five and sixty, while

364
00:17:36,160 --> 00:17:40,920
Speaker 3: another path focused on determining the last digit of the sum,

365
00:17:41,040 --> 00:17:43,879
Speaker 3: so it actually added the last two digits of thirty

366
00:17:43,880 --> 00:17:47,040
Speaker 3: six and fifty nine, six and nine to know that

367
00:17:47,119 --> 00:17:50,200
Speaker 3: the answer had to be a multiple of five. So

368
00:17:50,240 --> 00:17:52,480
Speaker 3: Claude used these two paths to come up with the

369
00:17:52,520 --> 00:17:54,800
Speaker 3: correct answer, which is as you said.

370
00:17:54,800 --> 00:17:58,320
Speaker 1: Ninety five. You know, I find particularly fascinating about this,

371
00:17:58,760 --> 00:18:01,320
Speaker 1: so I always struggled to be with math. My dad,

372
00:18:01,359 --> 00:18:04,679
Speaker 1: on the other hand, is a crazy math murder. He

373
00:18:04,800 --> 00:18:07,880
Speaker 1: was like the under thirteen chess champion of Britain, etc.

374
00:18:08,840 --> 00:18:10,560
Speaker 1: You know, he said to me about the most important

375
00:18:10,640 --> 00:18:14,200
Speaker 1: thing about how to do math well, is how approximate

376
00:18:14,240 --> 00:18:16,480
Speaker 1: the answer before you work it out. He basically told

377
00:18:16,560 --> 00:18:20,000
Speaker 1: me to do exactly what this model does. Basically break

378
00:18:20,000 --> 00:18:22,320
Speaker 1: it down to much simpler calculation, and then when you

379
00:18:22,359 --> 00:18:24,560
Speaker 1: do the actual work, you will know whether or not

380
00:18:24,600 --> 00:18:25,359
Speaker 1: you're in the range.

381
00:18:25,440 --> 00:18:25,640
Speaker 4: Right.

382
00:18:25,760 --> 00:18:28,919
Speaker 3: Well, Claude does math like your dad. When asked by

383
00:18:28,960 --> 00:18:31,639
Speaker 3: a user how it got the answer ninety five, it

384
00:18:31,680 --> 00:18:34,840
Speaker 3: claims to do it by the book, for example, simple addition,

385
00:18:35,000 --> 00:18:38,760
Speaker 3: carrying the one and one of the explanations anthropic positive.

386
00:18:38,440 --> 00:18:41,200
Speaker 1: Positive for the fact that, when asked how it got

387
00:18:41,200 --> 00:18:44,280
Speaker 1: to the answer, did it shared a response that was not,

388
00:18:44,320 --> 00:18:45,840
Speaker 1: in fact, how it got to the answer.

389
00:18:45,760 --> 00:18:49,560
Speaker 3: Right because Claude's written answer and I quote may reflect

390
00:18:49,560 --> 00:18:52,920
Speaker 3: the fact that the model learns to explain math by

391
00:18:53,000 --> 00:18:56,800
Speaker 3: simulating explanations written by people. So when asked to do

392
00:18:56,880 --> 00:19:00,120
Speaker 3: math without being taught how to do it, it may

393
00:19:00,119 --> 00:19:04,600
Speaker 3: develop its own internal strategies to do so. Remember it's

394
00:19:04,640 --> 00:19:05,720
Speaker 3: trade not programmed.

395
00:19:12,280 --> 00:19:14,359
Speaker 1: We're going to take a quick break, but stick around

396
00:19:14,520 --> 00:19:17,439
Speaker 1: and we'll be back with this week's tech support, all

397
00:19:17,480 --> 00:19:34,240
Speaker 1: about AI competition among tech giants. For our next segment,

398
00:19:34,240 --> 00:19:37,719
Speaker 1: we're going to be talking about all things AI. Surprise, Surprise,

399
00:19:37,760 --> 00:19:41,280
Speaker 1: on tech stuff Rise tell surprise, But in all seriousness,

400
00:19:41,400 --> 00:19:43,960
Speaker 1: there's just so much happening in AI all the time.

401
00:19:44,040 --> 00:19:45,760
Speaker 1: I find it even though it's our job, but it's

402
00:19:45,800 --> 00:19:46,439
Speaker 1: hard to keep up with.

403
00:19:46,880 --> 00:19:50,120
Speaker 3: We know. I literally work on a podcast about technology

404
00:19:50,160 --> 00:19:52,320
Speaker 3: and half the time other people are telling me what's

405
00:19:52,359 --> 00:19:54,240
Speaker 3: going on in the world of technology. Because I've come

406
00:19:54,640 --> 00:19:57,640
Speaker 3: I've become kind of like a sieve for technology news

407
00:19:57,680 --> 00:19:58,440
Speaker 3: from my friends.

408
00:19:58,720 --> 00:20:01,280
Speaker 1: Absolutely, So this week we're going to talk to somebody

409
00:20:01,359 --> 00:20:04,719
Speaker 1: who can help navigate a whirlwind of headlines. And you've

410
00:20:04,720 --> 00:20:08,760
Speaker 1: got the big open Ai fundraise, You've got elon integrating

411
00:20:09,119 --> 00:20:12,640
Speaker 1: x into Xai, and then you've got the struggles at

412
00:20:12,680 --> 00:20:13,360
Speaker 1: Apple with.

413
00:20:13,400 --> 00:20:16,560
Speaker 2: Their AI products or lack thereof.

414
00:20:17,000 --> 00:20:20,040
Speaker 1: Here to help us decode the current AI landscape is

415
00:20:20,119 --> 00:20:23,800
Speaker 1: Garrett Devinc who's a technology reporter for The Washington Post. Garrett,

416
00:20:23,840 --> 00:20:24,679
Speaker 1: Welcome to Tech Stuff.

417
00:20:24,800 --> 00:20:25,520
Speaker 4: Happy to be here.

418
00:20:25,680 --> 00:20:28,560
Speaker 1: Let's start with a big one. The volume of news

419
00:20:28,680 --> 00:20:32,480
Speaker 1: around AI today feels quite overwhelming. I mean, also, it

420
00:20:32,520 --> 00:20:37,800
Speaker 1: comes from so many different companies Open Ai, Google, Xai, Microsoft, Amazon, Apple,

421
00:20:38,119 --> 00:20:40,480
Speaker 1: The list goes on and on. How do you keep

422
00:20:40,600 --> 00:20:42,840
Speaker 1: up with all of this and also figure out how

423
00:20:42,840 --> 00:20:44,280
Speaker 1: to sort the signal from the noise.

424
00:20:44,560 --> 00:20:46,879
Speaker 5: Yeah, I mean, I think this is kind of a

425
00:20:46,920 --> 00:20:50,080
Speaker 5: key question. I'm sure a lot of people are asking themselves,

426
00:20:50,080 --> 00:20:52,520
Speaker 5: and I think first thing is, just take a deep breath,

427
00:20:52,600 --> 00:20:55,639
Speaker 5: calm down. AI is not coming for your job. AI

428
00:20:55,760 --> 00:20:58,560
Speaker 5: is not going to take over the world tomorrow, even

429
00:20:58,640 --> 00:21:02,040
Speaker 5: if really smart people or powerful people or rich companies

430
00:21:02,040 --> 00:21:05,919
Speaker 5: are saying that. Essentially, what we're seeing here is the

431
00:21:05,960 --> 00:21:09,320
Speaker 5: tech industry is this huge conglomeration of a bunch of

432
00:21:09,359 --> 00:21:12,440
Speaker 5: powerful people and billions and billions of dollars that are

433
00:21:12,480 --> 00:21:14,679
Speaker 5: always looking for the next thing, always looking for the

434
00:21:14,680 --> 00:21:17,640
Speaker 5: next way to make money. And they look back at

435
00:21:17,720 --> 00:21:22,440
Speaker 5: tech trends over the years, the Internet, cloud computing, moving

436
00:21:22,520 --> 00:21:25,399
Speaker 5: to mobile phones, and the tech industry has sort of

437
00:21:25,440 --> 00:21:30,360
Speaker 5: collectively decided that AI is the next one of those stages, right,

438
00:21:30,400 --> 00:21:32,760
Speaker 5: And so when the mobile phone came out, a bunch

439
00:21:32,760 --> 00:21:34,120
Speaker 5: of new people were able to make money.

440
00:21:34,200 --> 00:21:34,360
Speaker 1: Right.

441
00:21:34,400 --> 00:21:36,320
Speaker 4: We didn't have Uber, we didn't have.

442
00:21:36,200 --> 00:21:39,040
Speaker 5: Those kind of mobile first door Dash, those kinds of

443
00:21:39,080 --> 00:21:42,120
Speaker 5: companies before the mobile phone came out, and people made

444
00:21:42,280 --> 00:21:45,439
Speaker 5: huge amounts of money during that tech transition. And so

445
00:21:45,840 --> 00:21:49,320
Speaker 5: what's happening now is the tech industry believes and is

446
00:21:49,400 --> 00:21:52,439
Speaker 5: convincing themselves and trying to convince all of us that

447
00:21:52,560 --> 00:21:55,240
Speaker 5: AI is that next step, and so they're pouring money

448
00:21:55,240 --> 00:21:58,200
Speaker 5: into it, they're pouring marketing dollars into it, but at

449
00:21:58,200 --> 00:22:01,440
Speaker 5: the same time, there's still trying to build the plane

450
00:22:01,560 --> 00:22:04,480
Speaker 5: as it's taking off. And so that's why you might

451
00:22:04,520 --> 00:22:06,680
Speaker 5: see a lot of products that maybe.

452
00:22:06,440 --> 00:22:07,440
Speaker 4: Don't work very well.

453
00:22:07,560 --> 00:22:09,879
Speaker 5: You don't know really how they fit into your life,

454
00:22:10,200 --> 00:22:11,840
Speaker 5: and you're not sure whether you should be paying for

455
00:22:11,880 --> 00:22:13,600
Speaker 5: them yet. And so I think the first thing is

456
00:22:13,640 --> 00:22:15,639
Speaker 5: to say, you know, don't worry. You're not missing the

457
00:22:15,640 --> 00:22:18,040
Speaker 5: boat if you're not using a million AI apps right now.

458
00:22:18,119 --> 00:22:20,520
Speaker 5: And so, yes, this is happening. There's a lot of

459
00:22:20,720 --> 00:22:23,440
Speaker 5: hype and interest here, but it's not as if AI

460
00:22:23,640 --> 00:22:26,600
Speaker 5: is just sort of gonna change everything immediately.

461
00:22:26,880 --> 00:22:28,800
Speaker 1: And one of the things that Karen and I sometimes

462
00:22:28,800 --> 00:22:31,159
Speaker 1: talk about is like, is it speeding up or is

463
00:22:31,200 --> 00:22:33,880
Speaker 1: it slowing down? Because like November December, all the headlines

464
00:22:33,920 --> 00:22:36,880
Speaker 1: were slowing down CHACKGBT five is not coming, and now

465
00:22:36,880 --> 00:22:39,160
Speaker 1: it feels like we're in a big speeding up moment. Again,

466
00:22:39,760 --> 00:22:41,679
Speaker 1: is it even a relevant question? And where do you

467
00:22:41,680 --> 00:22:42,199
Speaker 1: fall on it?

468
00:22:42,359 --> 00:22:42,560
Speaker 4: Yeah?

469
00:22:42,560 --> 00:22:43,879
Speaker 5: I mean I think it's a great question, and I

470
00:22:43,960 --> 00:22:46,720
Speaker 5: think that analysis is correct, right. I mean, everyone is

471
00:22:46,760 --> 00:22:49,240
Speaker 5: trying to either build something up or tear it down.

472
00:22:49,280 --> 00:22:50,640
Speaker 4: That's how these things work.

473
00:22:50,720 --> 00:22:53,920
Speaker 5: And so the reason we had this huge boost in

474
00:22:54,000 --> 00:22:57,720
Speaker 5: interest was because chat gipt came out. It was definitely

475
00:22:57,840 --> 00:23:00,639
Speaker 5: better than what most people had you used before and

476
00:23:00,640 --> 00:23:02,560
Speaker 5: being able to experience directly, and they were able to

477
00:23:02,560 --> 00:23:05,240
Speaker 5: put it in a format that regular people could actually

478
00:23:05,359 --> 00:23:08,840
Speaker 5: understand and have a conversation with it. And so that's

479
00:23:08,840 --> 00:23:11,960
Speaker 5: sort of what fired the starting gun. And then they said, okay,

480
00:23:12,000 --> 00:23:14,320
Speaker 5: how do we make that better? And then the techniques

481
00:23:14,320 --> 00:23:17,520
Speaker 5: they were using, which was essentially to use way more data,

482
00:23:17,840 --> 00:23:20,560
Speaker 5: to just shove more data into these AI models and

483
00:23:20,720 --> 00:23:23,399
Speaker 5: hope that they get smarter. That had been working up

484
00:23:23,400 --> 00:23:25,840
Speaker 5: into a certain point, and then they kind of ran

485
00:23:25,920 --> 00:23:28,520
Speaker 5: out of data and that method slowed down, and so

486
00:23:28,640 --> 00:23:32,720
Speaker 5: they've now pivoted to using different techniques where they're actually

487
00:23:32,720 --> 00:23:35,640
Speaker 5: spending a lot more time training the model to.

488
00:23:35,600 --> 00:23:36,399
Speaker 4: Do different things.

489
00:23:36,400 --> 00:23:38,800
Speaker 5: They're sort of doing more coding so that it can

490
00:23:39,080 --> 00:23:42,120
Speaker 5: be a bit more efficient and strategic, and now they're

491
00:23:42,119 --> 00:23:44,959
Speaker 5: seeing a boost in capability.

492
00:23:44,280 --> 00:23:45,320
Speaker 4: From that technique.

493
00:23:45,320 --> 00:23:49,080
Speaker 5: And if people remember the Chinese AI model called deep Seek.

494
00:23:49,440 --> 00:23:51,440
Speaker 5: They really had a huge breakthrough where they were able

495
00:23:51,480 --> 00:23:54,000
Speaker 5: to use a little bit less data and less computing

496
00:23:54,080 --> 00:23:56,440
Speaker 5: power to come up with a model that was really

497
00:23:56,520 --> 00:23:59,439
Speaker 5: quite capable. And so now everyone is saying, oh, we're

498
00:23:59,480 --> 00:24:02,520
Speaker 5: speeding up again because we found new ways of increasing

499
00:24:02,560 --> 00:24:04,200
Speaker 5: the capabilities.

500
00:24:03,960 --> 00:24:07,680
Speaker 3: Which is I mean, ultimately a more long term effective

501
00:24:07,720 --> 00:24:10,040
Speaker 3: way to get these things to work than having to

502
00:24:10,119 --> 00:24:14,520
Speaker 3: rely on humongous data sets that might not be replenishable.

503
00:24:14,600 --> 00:24:16,080
Speaker 4: Yeah, yeah, absolutely.

504
00:24:16,080 --> 00:24:18,320
Speaker 5: And I mean the other aspect here is that AI

505
00:24:18,440 --> 00:24:21,800
Speaker 5: is very compute intensive, which is essentially just a way

506
00:24:21,840 --> 00:24:23,840
Speaker 5: of saying they need a lot of computers and a

507
00:24:23,840 --> 00:24:26,919
Speaker 5: lot of computer chips in order to do AI in

508
00:24:26,960 --> 00:24:29,919
Speaker 5: the first place. Takes a lot of energy, and so

509
00:24:30,000 --> 00:24:33,159
Speaker 5: there's a lot of potential environmental concerns. Even here in

510
00:24:33,200 --> 00:24:35,840
Speaker 5: the United States, coal power plants that were slated to

511
00:24:35,880 --> 00:24:38,879
Speaker 5: be shut down have actually been ramped back up in

512
00:24:39,080 --> 00:24:41,880
Speaker 5: order to serve all those AI data centers. So there

513
00:24:41,960 --> 00:24:44,919
Speaker 5: is a lot of interest and pressure in making AI

514
00:24:45,040 --> 00:24:48,240
Speaker 5: more efficient so that it's cheaper and more environmentally friendly.

515
00:24:48,960 --> 00:24:51,280
Speaker 3: One of the biggest names in the industry is open AI,

516
00:24:51,720 --> 00:24:55,919
Speaker 3: and for anyone who is living under a rock and

517
00:24:56,040 --> 00:24:58,679
Speaker 3: wasn't on social media this week. Why is everybody talking

518
00:24:58,720 --> 00:25:00,960
Speaker 3: so much about Studio Ghibli.

519
00:25:01,720 --> 00:25:06,400
Speaker 5: Yeah, so, open Ai they released a new image generator, right.

520
00:25:06,480 --> 00:25:08,400
Speaker 4: So one of the things that people have been able.

521
00:25:08,200 --> 00:25:10,240
Speaker 5: To use AI for over the last couple of years

522
00:25:10,320 --> 00:25:12,520
Speaker 5: is you make a short description and it spits out

523
00:25:12,560 --> 00:25:14,760
Speaker 5: an image, and you know, it's been getting better and

524
00:25:14,800 --> 00:25:18,440
Speaker 5: better over the months, and essentially they released a big

525
00:25:18,520 --> 00:25:23,440
Speaker 5: update to THEIRS and people realize that they could use photos,

526
00:25:23,560 --> 00:25:27,119
Speaker 5: upload photos and get the model to recreate that image

527
00:25:27,119 --> 00:25:31,119
Speaker 5: and sort of the same design styles from iconic animators

528
00:25:31,119 --> 00:25:33,840
Speaker 5: like you mentioned Studio Ghibli or even like the movie

529
00:25:33,920 --> 00:25:36,520
Speaker 5: Wallace and Grommet or the Muppets, and.

530
00:25:37,359 --> 00:25:40,200
Speaker 3: I like the Lego Family, the Lego Family or exactly.

531
00:25:41,080 --> 00:25:43,600
Speaker 5: It was this moment where the technology allowed people to

532
00:25:43,640 --> 00:25:45,560
Speaker 5: apply their creativity.

533
00:25:45,320 --> 00:25:47,639
Speaker 4: In a new way, and that's why it went viral.

534
00:25:48,359 --> 00:25:52,040
Speaker 5: And open Ai I think they did hope that this

535
00:25:52,080 --> 00:25:55,720
Speaker 5: would happen. They themselves were using some of these Studio

536
00:25:55,760 --> 00:26:00,639
Speaker 5: Ghibili examples early on. But also they can't really predict

537
00:26:00,720 --> 00:26:02,520
Speaker 5: or control how this is going to go, and some

538
00:26:02,600 --> 00:26:04,760
Speaker 5: of their releases have just kind of fallen flat. Everyone's

539
00:26:04,800 --> 00:26:06,600
Speaker 5: been like boring we don't care, but.

540
00:26:06,640 --> 00:26:07,720
Speaker 4: This one a lot of people.

541
00:26:07,840 --> 00:26:09,800
Speaker 5: They found it really fun, they found it really interesting,

542
00:26:09,800 --> 00:26:12,320
Speaker 5: and it definitely went viral, and it actually brought a

543
00:26:12,359 --> 00:26:14,000
Speaker 5: lot more people who.

544
00:26:13,920 --> 00:26:16,240
Speaker 4: Hadn't been using chat GPT before.

545
00:26:16,680 --> 00:26:19,000
Speaker 5: But now, of course this raises all sorts of questions

546
00:26:19,000 --> 00:26:22,800
Speaker 5: about art and copyright and big problems like that.

547
00:26:23,600 --> 00:26:29,640
Speaker 3: What's interesting, though, is while they're trying to grow very quickly,

548
00:26:30,080 --> 00:26:32,800
Speaker 3: we have someone like Sam Altman come out and say,

549
00:26:33,080 --> 00:26:35,320
Speaker 3: please slow down with this image generation.

550
00:26:35,640 --> 00:26:38,679
Speaker 1: He said that the GPUs are being melted by the demand, right.

551
00:26:38,840 --> 00:26:41,840
Speaker 5: Yeah, I mean it's possible some GPUs actually did melt

552
00:26:41,840 --> 00:26:43,360
Speaker 5: a little bit. I mean, you know, when you're using

553
00:26:43,400 --> 00:26:45,920
Speaker 5: your laptop, you've got two hundred Chrome tabs open and

554
00:26:45,960 --> 00:26:48,480
Speaker 5: you're trying to listen to YouTube, it gets hot, right,

555
00:26:48,520 --> 00:26:50,360
Speaker 5: and so that's exactly what happens.

556
00:26:50,560 --> 00:26:52,080
Speaker 3: My phone does get quite hot.

557
00:26:52,280 --> 00:26:56,280
Speaker 5: Yeah, you know that same thing happens with AI, right.

558
00:26:56,280 --> 00:26:57,920
Speaker 5: I mean, so this is still a very physical thing.

559
00:26:57,960 --> 00:26:59,879
Speaker 5: Every single time everyone in the world says, hey, make

560
00:27:00,119 --> 00:27:02,440
Speaker 5: an image of this. Hey write me a resume for this,

561
00:27:02,520 --> 00:27:06,080
Speaker 5: Hey answer my test questions for me. That needs to

562
00:27:06,119 --> 00:27:08,960
Speaker 5: go to a data center, It needs to be computed

563
00:27:09,160 --> 00:27:12,720
Speaker 5: on these computer chips on GPUs, which is the technical

564
00:27:12,800 --> 00:27:15,840
Speaker 5: term for the computer chips, and that heats them up

565
00:27:15,920 --> 00:27:18,040
Speaker 5: and so I don't know if they were actually melting,

566
00:27:18,040 --> 00:27:20,560
Speaker 5: But essentially what sam Altman was referring to there is

567
00:27:20,600 --> 00:27:22,840
Speaker 5: just that so many people wanted to use this that

568
00:27:22,920 --> 00:27:27,199
Speaker 5: it was becoming very expensive for open ai to run it.

569
00:27:27,200 --> 00:27:29,040
Speaker 5: And this kind of gets to a central problem for

570
00:27:29,080 --> 00:27:31,639
Speaker 5: them because the more people use it, the more it

571
00:27:31,760 --> 00:27:34,920
Speaker 5: costs them in computer chip costs, and so they want

572
00:27:34,960 --> 00:27:37,200
Speaker 5: people to use it, but at the same time they

573
00:27:37,240 --> 00:27:39,199
Speaker 5: need to figure out, you know, how can we convince

574
00:27:39,240 --> 00:27:42,280
Speaker 5: people force people to pay for these things so that

575
00:27:42,320 --> 00:27:44,320
Speaker 5: we can actually grow as a business. And this is

576
00:27:44,400 --> 00:27:47,960
Speaker 5: a huge question mark around open ai and other AI companies.

577
00:27:48,359 --> 00:27:51,480
Speaker 1: As fun as it is making a studio ghibli portrait yourself,

578
00:27:51,520 --> 00:27:54,280
Speaker 1: it's hard to measure it being a huge, huge business

579
00:27:54,320 --> 00:27:56,520
Speaker 1: driver of people buying tokens to do that.

580
00:27:56,520 --> 00:27:58,479
Speaker 5: That's sort of a question for open Ai, right, I mean,

581
00:27:58,520 --> 00:28:01,520
Speaker 5: they've been able to have these viral moments. Chat GPT

582
00:28:01,920 --> 00:28:04,800
Speaker 5: itself was a viral moment, and I do think people

583
00:28:05,000 --> 00:28:07,880
Speaker 5: are using these technologies. I mean, in some ways open

584
00:28:07,880 --> 00:28:10,800
Speaker 5: ai chatch GPT is one of the fastest growing, if

585
00:28:10,800 --> 00:28:16,040
Speaker 5: not the fastest growing consumer Internet products ever. But we're

586
00:28:16,080 --> 00:28:18,120
Speaker 5: not quite at that point yet where I think regular

587
00:28:18,200 --> 00:28:20,600
Speaker 5: people are saying, oh, like, this is so important to

588
00:28:20,680 --> 00:28:24,320
Speaker 5: my life, I need it so badly to write emails

589
00:28:24,560 --> 00:28:28,240
Speaker 5: or to have fun like generating these images that I'm

590
00:28:28,240 --> 00:28:31,159
Speaker 5: willing to pay two hundred dollars a month for it.

591
00:28:31,200 --> 00:28:33,640
Speaker 5: I mean, I do think that the company is still

592
00:28:33,640 --> 00:28:36,200
Speaker 5: in this world where they're trying to figure out how

593
00:28:36,200 --> 00:28:38,760
Speaker 5: do we convince people that they need this so badly

594
00:28:39,160 --> 00:28:41,400
Speaker 5: that they're willing to spend hundreds of dollars a year

595
00:28:41,440 --> 00:28:41,760
Speaker 5: on it.

596
00:28:42,320 --> 00:28:45,600
Speaker 1: So you may be somewhat skeptical, Garrett, but the market

597
00:28:45,640 --> 00:28:47,800
Speaker 1: is not, or at least soft Bank is not. You

598
00:28:47,840 --> 00:28:48,600
Speaker 1: talk a bit about that.

599
00:28:49,000 --> 00:28:53,640
Speaker 5: Yeah, I mean open Ai raised forty billion dollars and

600
00:28:54,000 --> 00:28:57,520
Speaker 5: at a three hundred billion dollar valuation, and I mean it's.

601
00:28:57,360 --> 00:29:00,640
Speaker 1: The largest ever private financing of a US company that.

602
00:29:01,200 --> 00:29:03,760
Speaker 3: A company that's not a private company a little while ago.

603
00:29:03,920 --> 00:29:07,440
Speaker 5: Yeah, And the only private company that is actually worth

604
00:29:07,520 --> 00:29:10,880
Speaker 5: more than three hundred billion dollars is SpaceX. So that's

605
00:29:10,960 --> 00:29:14,720
Speaker 5: Elon Musk's Space company. But they build big physical rockets

606
00:29:14,760 --> 00:29:17,800
Speaker 5: that cost hundreds of millions of dollars, So that company's

607
00:29:17,800 --> 00:29:20,240
Speaker 5: worth three hundred fifty billion. Opening Eye is now, according

608
00:29:20,240 --> 00:29:23,320
Speaker 5: to its investors, worth three hundred billion dollars, right, And

609
00:29:23,400 --> 00:29:26,200
Speaker 5: so I think this mostly goes to the to that

610
00:29:26,360 --> 00:29:29,720
Speaker 5: question about how expensive it is to do AI, right,

611
00:29:29,760 --> 00:29:33,480
Speaker 5: and so they need that money in order to buy

612
00:29:33,560 --> 00:29:37,240
Speaker 5: data centers, buy computer chips to keep helping people make

613
00:29:37,320 --> 00:29:39,600
Speaker 5: these these studio ghibli images and all the other things

614
00:29:39,640 --> 00:29:41,200
Speaker 5: that Opening Eye is working for.

615
00:29:41,240 --> 00:29:42,120
Speaker 4: And so so it's.

616
00:29:41,960 --> 00:29:44,920
Speaker 1: Just the uber model where you basically subsidized users and

617
00:29:45,000 --> 00:29:47,040
Speaker 1: then once you get them hooked, you start to charge them,

618
00:29:47,080 --> 00:29:48,000
Speaker 1: but at huge scale.

619
00:29:48,160 --> 00:29:48,360
Speaker 4: Yeah.

620
00:29:48,360 --> 00:29:50,680
Speaker 5: I mean that's a playbook that tech companies have used

621
00:29:50,680 --> 00:29:53,480
Speaker 5: for years now, right, and get people hooked on something

622
00:29:53,520 --> 00:29:56,720
Speaker 5: that is fun, cheap, easy, free work into their lives

623
00:29:56,720 --> 00:29:59,240
Speaker 5: and so that they feel like they needed every single day,

624
00:29:59,560 --> 00:30:02,040
Speaker 5: and then start to increase the costs. I mean Google

625
00:30:02,120 --> 00:30:04,840
Speaker 5: has done this. I'm paying for my Gmail storage. I

626
00:30:04,880 --> 00:30:05,800
Speaker 5: don't know if you guys are.

627
00:30:05,840 --> 00:30:11,080
Speaker 3: I mean, yes, it's a question of when does something

628
00:30:11,120 --> 00:30:14,200
Speaker 3: go from the gimmick phase to the business phase. I think,

629
00:30:14,200 --> 00:30:16,160
Speaker 3: at least in terms of chat GBT.

630
00:30:16,440 --> 00:30:18,800
Speaker 5: Yeah, And I think the other thing to point out

631
00:30:18,840 --> 00:30:21,960
Speaker 5: here is that open ai does have a business where

632
00:30:22,000 --> 00:30:25,320
Speaker 5: they sell access to their AI to other businesses, right,

633
00:30:25,360 --> 00:30:28,239
Speaker 5: and so there's the consumer question, which is exactly what

634
00:30:28,240 --> 00:30:30,960
Speaker 5: we're talking about. Then they are actually selling to businesses

635
00:30:30,960 --> 00:30:34,120
Speaker 5: who want to put AI technology into their own apps

636
00:30:34,200 --> 00:30:38,000
Speaker 5: and into their own technology, and that's also a big

637
00:30:38,040 --> 00:30:39,960
Speaker 5: part of what open ai is trying to do here.

638
00:30:40,440 --> 00:30:43,320
Speaker 5: But that's also, you know, a big question mark because

639
00:30:43,320 --> 00:30:47,440
Speaker 5: we have these open source AI models as well. Deep Seek,

640
00:30:47,520 --> 00:30:50,280
Speaker 5: the Chinese one we mentioned earlier is an example of that.

641
00:30:50,680 --> 00:30:53,360
Speaker 5: Facebook also provides these tools where they just essentially put

642
00:30:53,400 --> 00:30:56,080
Speaker 5: the AI out there for free for other businesses to

643
00:30:56,120 --> 00:30:58,560
Speaker 5: take and use in their own ways. And so open

644
00:30:58,600 --> 00:31:02,960
Speaker 5: ai is a very strong business. They have incredible technology,

645
00:31:02,960 --> 00:31:04,680
Speaker 5: they have some of the smartest people in the world

646
00:31:04,760 --> 00:31:07,840
Speaker 5: on AI, and they have huge funding and backing.

647
00:31:07,520 --> 00:31:08,360
Speaker 4: From their investors.

648
00:31:08,360 --> 00:31:11,280
Speaker 5: But at the same time, that doesn't guarantee that they're

649
00:31:11,320 --> 00:31:13,720
Speaker 5: going to continue to grow or even be around in

650
00:31:13,800 --> 00:31:18,840
Speaker 5: five years.

651
00:31:21,080 --> 00:31:23,280
Speaker 1: When we come back, we'll hear about how other big

652
00:31:23,280 --> 00:31:26,600
Speaker 1: tech companies like Amazon and Apple are fairing in the

653
00:31:26,680 --> 00:31:44,680
Speaker 1: scramble to develop AI products. Yeah, we want to ask

654
00:31:44,680 --> 00:31:46,520
Speaker 1: you a little bit about what Amazon and Apple are

655
00:31:46,520 --> 00:31:48,640
Speaker 1: doing in the realm of AI. But just before we

656
00:31:48,720 --> 00:31:51,720
Speaker 1: get there, there's another huge deal this week, also with

657
00:31:51,840 --> 00:31:53,840
Speaker 1: a potentially dubious price tag.

658
00:31:54,280 --> 00:31:57,920
Speaker 5: Yes, so I think you're referring to Elon Musk's merger

659
00:31:57,960 --> 00:31:59,960
Speaker 5: of two of his companies little confusing.

660
00:32:00,080 --> 00:32:02,000
Speaker 4: They're both kind of called XX.

661
00:32:02,360 --> 00:32:04,040
Speaker 1: I couldn't even get through the read at the beginning

662
00:32:04,080 --> 00:32:06,560
Speaker 1: of the shows. That's a tongue twist. Yeah, exactly.

663
00:32:06,640 --> 00:32:10,880
Speaker 4: I mean I think that sort of speaks to Elon Musk.

664
00:32:11,040 --> 00:32:13,720
Speaker 5: He has many companies at this point, and he's been

665
00:32:13,760 --> 00:32:16,600
Speaker 5: known to sort of move assets around a little bit.

666
00:32:16,680 --> 00:32:20,200
Speaker 5: And so when it comes to x which is formally Twitter,

667
00:32:20,280 --> 00:32:22,320
Speaker 5: so that's the social media platform that he bought a

668
00:32:22,320 --> 00:32:25,040
Speaker 5: couple of years ago for forty four billion dollars, he

669
00:32:25,360 --> 00:32:30,000
Speaker 5: has now sold that company to his AI company, which

670
00:32:30,040 --> 00:32:34,200
Speaker 5: is called Xai. And sort of formally, the social media

671
00:32:34,200 --> 00:32:37,240
Speaker 5: company is now owned by the AI company. And so

672
00:32:37,440 --> 00:32:39,600
Speaker 5: I say formally, because these companies had kind of been

673
00:32:39,600 --> 00:32:42,720
Speaker 5: working together in a lot of ways. User data from

674
00:32:42,840 --> 00:32:45,920
Speaker 5: the social media company was already being used to train

675
00:32:46,400 --> 00:32:50,120
Speaker 5: the AI on the AI company. The AI company's main product,

676
00:32:50,120 --> 00:32:53,160
Speaker 5: which is called Grock, which is a chat GPT competitor,

677
00:32:53,720 --> 00:32:57,520
Speaker 5: was available through the social media company, and so in

678
00:32:57,560 --> 00:33:00,520
Speaker 5: a lot of ways, these companies were already the same thing.

679
00:33:00,640 --> 00:33:02,920
Speaker 5: And what he did here is he said, look, the

680
00:33:03,000 --> 00:33:04,920
Speaker 5: AI company is able to raise a lot of money

681
00:33:04,920 --> 00:33:08,120
Speaker 5: because everyone loves AI, everyone wants to boost AI. So

682
00:33:08,200 --> 00:33:10,240
Speaker 5: I'm going to use that money that the AI company

683
00:33:10,280 --> 00:33:12,840
Speaker 5: is able to raise to bail out and kind of

684
00:33:13,120 --> 00:33:15,480
Speaker 5: give me more time when it comes to the social

685
00:33:15,520 --> 00:33:19,840
Speaker 5: media company, which is very influential and maybe Elon Musk's

686
00:33:19,880 --> 00:33:22,960
Speaker 5: most important company right now because of the political influence

687
00:33:23,000 --> 00:33:26,200
Speaker 5: it gives him. But from a business perspective, the social

688
00:33:26,200 --> 00:33:29,080
Speaker 5: media company has struggled and sort of been, you know,

689
00:33:29,240 --> 00:33:31,840
Speaker 5: going through the wilderness a little bit since Elon Musk

690
00:33:31,960 --> 00:33:34,360
Speaker 5: bought it because most of its users left, a whole

691
00:33:34,400 --> 00:33:36,720
Speaker 5: bunch of new users came in, advertisers left.

692
00:33:36,760 --> 00:33:37,920
Speaker 4: Maybe the advertisers are going.

693
00:33:37,880 --> 00:33:39,880
Speaker 5: To come back, and so it's a way for Elon

694
00:33:39,960 --> 00:33:42,840
Speaker 5: Musk to sort of use the AI hype to kind

695
00:33:42,840 --> 00:33:46,360
Speaker 5: of help shore up the finances of his social media company.

696
00:33:47,280 --> 00:33:50,840
Speaker 3: You know another major tech company that isn't mentioned as

697
00:33:50,920 --> 00:33:53,240
Speaker 3: much in the AI raise, and that is Amazon. They

698
00:33:53,280 --> 00:33:58,880
Speaker 3: recently unveiled Amazon Nova Act. They're AI agent. So what

699
00:33:58,920 --> 00:34:01,280
Speaker 3: does the lay person need to know about this.

700
00:34:01,600 --> 00:34:04,200
Speaker 4: So okay, a couple of things. I think AI agent

701
00:34:04,480 --> 00:34:05,240
Speaker 4: is a term.

702
00:34:05,080 --> 00:34:07,920
Speaker 5: That people are probably already hearing, and I guarantee you

703
00:34:07,960 --> 00:34:09,680
Speaker 5: they're going to be hearing more about it in the

704
00:34:09,719 --> 00:34:12,600
Speaker 5: coming months and years. An AI agent all that is

705
00:34:12,600 --> 00:34:15,080
Speaker 5: is just you know, you can say, okay, well, if

706
00:34:15,120 --> 00:34:17,120
Speaker 5: I can have a conversation with chat GPT, I can

707
00:34:17,160 --> 00:34:19,600
Speaker 5: ask it things. Can I ask it to then go

708
00:34:20,080 --> 00:34:22,360
Speaker 5: and read the internet for me? Can I ask it

709
00:34:22,440 --> 00:34:24,319
Speaker 5: to go do things on the internet for me?

710
00:34:24,440 --> 00:34:24,600
Speaker 1: Right?

711
00:34:24,680 --> 00:34:28,400
Speaker 5: If chat GPT is able to read an e commerce website,

712
00:34:28,719 --> 00:34:31,759
Speaker 5: can't I just tell chat GPT, hey, go buy me

713
00:34:31,920 --> 00:34:34,600
Speaker 5: the cheapest sofa for my new furniture that you can

714
00:34:34,680 --> 00:34:38,240
Speaker 5: find from my new apartment that's green and seats three people.

715
00:34:38,320 --> 00:34:38,480
Speaker 4: Right.

716
00:34:38,560 --> 00:34:41,800
Speaker 5: And so that's what an AI agent is, is essentially using

717
00:34:41,840 --> 00:34:45,360
Speaker 5: AI to go and help people to do things on

718
00:34:45,480 --> 00:34:48,080
Speaker 5: the internet for them. And so this is something a

719
00:34:48,120 --> 00:34:50,279
Speaker 5: lot of AI companies are talking about. Obviously, there's a

720
00:34:50,280 --> 00:34:52,719
Speaker 5: lot of problems. The technology is not quite there. The

721
00:34:52,800 --> 00:34:55,560
Speaker 5: last thing you want is to say to your AI agent,

722
00:34:55,640 --> 00:34:57,960
Speaker 5: go buy me a sofa for under one thousand dollars,

723
00:34:58,400 --> 00:35:01,040
Speaker 5: and then suddenly eight sofas that cost ten thousand dollars

724
00:35:01,040 --> 00:35:03,000
Speaker 5: each show up at your door a week later, right,

725
00:35:03,040 --> 00:35:04,960
Speaker 5: I mean we need to be really careous. Is that?

726
00:35:04,960 --> 00:35:06,439
Speaker 3: What is that the phase we're in right now?

727
00:35:06,760 --> 00:35:09,239
Speaker 5: I mean a colleague of mine ran an experiment where

728
00:35:09,239 --> 00:35:11,319
Speaker 5: he asked some of these AI agents to go and

729
00:35:11,400 --> 00:35:13,440
Speaker 5: find him the cheapest eggs.

730
00:35:13,440 --> 00:35:15,720
Speaker 4: This was sort of at the height of the egg panic,

731
00:35:15,760 --> 00:35:16,400
Speaker 4: and eggs.

732
00:35:16,200 --> 00:35:20,160
Speaker 5: Were selling for twenty dollars a dozen and the agent

733
00:35:20,239 --> 00:35:23,480
Speaker 5: actually went and bought I think it was like thirty

734
00:35:23,520 --> 00:35:26,880
Speaker 5: dollars eggs and had them delivered to his home before

735
00:35:26,960 --> 00:35:29,640
Speaker 5: he even had the chance to say, like, yes, make

736
00:35:29,680 --> 00:35:30,240
Speaker 5: that purchase.

737
00:35:30,280 --> 00:35:32,560
Speaker 4: And so it's definitely in the experimental phase.

738
00:35:32,880 --> 00:35:36,720
Speaker 5: But Amazon they sort of see that this is maybe

739
00:35:36,760 --> 00:35:38,400
Speaker 5: the next frontier of the technology.

740
00:35:38,520 --> 00:35:41,200
Speaker 1: Is that what the key interest is to basically do

741
00:35:41,280 --> 00:35:42,239
Speaker 1: your shopping for you?

742
00:35:42,560 --> 00:35:44,520
Speaker 4: I mean, I think that would make sense.

743
00:35:44,560 --> 00:35:47,719
Speaker 5: I mean, they are the shopping company, and we all

744
00:35:47,760 --> 00:35:49,480
Speaker 5: know Amazon Alexa.

745
00:35:49,239 --> 00:35:50,520
Speaker 4: Was an early version of this.

746
00:35:50,600 --> 00:35:53,160
Speaker 5: I mean you could ask Amazon Alexa to buy things

747
00:35:53,160 --> 00:35:55,680
Speaker 5: for you on Amazon. You could ask it obviously to

748
00:35:55,840 --> 00:35:58,359
Speaker 5: remind you about the weather, and people never really use

749
00:35:58,400 --> 00:36:00,799
Speaker 5: it for more than that. And so people have been

750
00:36:00,840 --> 00:36:03,920
Speaker 5: saying why was Amazon not ahead of this trend? Why

751
00:36:04,120 --> 00:36:07,440
Speaker 5: is Amazon Alexa not smarter? Why can chat GPT do

752
00:36:07,640 --> 00:36:10,959
Speaker 5: things that Amazon Alexa can't. So Amazon has been under

753
00:36:11,000 --> 00:36:13,799
Speaker 5: a huge amount of pressure from its investors, from its

754
00:36:13,800 --> 00:36:16,400
Speaker 5: own employees, from other people in the tech industry to

755
00:36:16,560 --> 00:36:18,960
Speaker 5: show that they are riding this AI wave just like

756
00:36:19,000 --> 00:36:21,880
Speaker 5: these other companies. And I think this Nova Act product,

757
00:36:21,880 --> 00:36:24,280
Speaker 5: which is very new, it's still in the experimental phase,

758
00:36:24,760 --> 00:36:26,640
Speaker 5: is a sign that they're trying to do that.

759
00:36:26,880 --> 00:36:29,680
Speaker 1: All of this, of course brings us to Siri. I mean,

760
00:36:29,719 --> 00:36:32,799
Speaker 1: the irony being that Amazon Alexa and Apple Siri were

761
00:36:32,840 --> 00:36:35,800
Speaker 1: kind of ahead of the curve in terms of voice

762
00:36:35,880 --> 00:36:41,200
Speaker 1: driven assistance, essentially agentic type features, and now Amazon feels

763
00:36:41,200 --> 00:36:43,920
Speaker 1: like it's behind the curve a little bit, and certainly

764
00:36:44,000 --> 00:36:46,919
Speaker 1: Apple does too when it comes to AI. What's going

765
00:36:46,960 --> 00:36:47,400
Speaker 1: on there?

766
00:36:47,680 --> 00:36:50,400
Speaker 5: Yeah, I mean, I think there's this dynamic that you're

767
00:36:50,840 --> 00:36:53,879
Speaker 5: putting your finger on where these companies are really really

768
00:36:53,880 --> 00:36:56,280
Speaker 5: invested in their own way of doing things and then

769
00:36:56,600 --> 00:36:58,920
Speaker 5: there's a new way that kind of comes out of

770
00:36:58,960 --> 00:37:01,880
Speaker 5: left field and to adjust right. This is the classic

771
00:37:02,360 --> 00:37:05,520
Speaker 5: innovator's dilemma. And so I would say for both companies.

772
00:37:05,560 --> 00:37:10,080
Speaker 5: Don't count them out. Amazon, Apple. They are both massive companies.

773
00:37:10,120 --> 00:37:13,319
Speaker 5: They are bigger than pretty much anything we've seen in

774
00:37:13,440 --> 00:37:18,360
Speaker 5: the history of business. They are in people's lives every minute,

775
00:37:18,400 --> 00:37:20,279
Speaker 5: every day, and so I think, first of all, we

776
00:37:20,320 --> 00:37:22,520
Speaker 5: should be careful not to just write them off and say, oh,

777
00:37:22,520 --> 00:37:25,919
Speaker 5: they're behind, therefore they will fail. But I do think

778
00:37:25,960 --> 00:37:28,919
Speaker 5: that this is really five alarm fire moment for them.

779
00:37:28,960 --> 00:37:31,319
Speaker 1: So you, I mean, they've been having internal sort of

780
00:37:31,360 --> 00:37:33,759
Speaker 1: ol hands meetings. What's the kind of internal drummer at

781
00:37:33,760 --> 00:37:34,279
Speaker 1: apput on this?

782
00:37:34,719 --> 00:37:38,280
Speaker 5: They are saying, Wow, we need to get on board

783
00:37:38,320 --> 00:37:40,439
Speaker 5: with this. And Google had the same thing when chat

784
00:37:40,480 --> 00:37:41,640
Speaker 5: GPT eventually came out.

785
00:37:41,680 --> 00:37:43,720
Speaker 4: Right, Google is the AI company.

786
00:37:43,760 --> 00:37:46,360
Speaker 5: A lot of this technology we're talking about was actually

787
00:37:46,360 --> 00:37:48,719
Speaker 5: developed by Google and then just shared with the world

788
00:37:48,760 --> 00:37:50,719
Speaker 5: because they didn't quite know what to do with it.

789
00:37:51,239 --> 00:37:54,400
Speaker 5: And so for Apple, people have said, okay, well Series

790
00:37:54,440 --> 00:37:56,560
Speaker 5: been around, are you going to make series smarter? This

791
00:37:56,600 --> 00:38:00,279
Speaker 5: seems like an obvious application. They stayed quiet for about

792
00:38:00,280 --> 00:38:03,040
Speaker 5: a year after chat GPT came out, and then they said.

793
00:38:03,040 --> 00:38:05,000
Speaker 4: Yes, we're doing it. We're going all in.

794
00:38:05,160 --> 00:38:08,360
Speaker 5: We're an AI company like everyone else. And now they've

795
00:38:08,360 --> 00:38:11,200
Speaker 5: delayed some of those releases, right, they've said, oh, actually,

796
00:38:11,320 --> 00:38:13,600
Speaker 5: we might need a bit more time to get it

797
00:38:13,640 --> 00:38:15,680
Speaker 5: to the level where we want it. And we can

798
00:38:15,719 --> 00:38:19,680
Speaker 5: see some of Apple's AI experiments have already failed pretty spectacularly.

799
00:38:19,719 --> 00:38:22,680
Speaker 5: They had a bot that summarized some of the messages

800
00:38:22,719 --> 00:38:24,879
Speaker 5: that were coming into your phone. Well I remember that, yeah,

801
00:38:25,239 --> 00:38:28,040
Speaker 5: last year Apple Intelligence. And what they would do is

802
00:38:28,080 --> 00:38:30,160
Speaker 5: they would see news alerts and they and said, oh,

803
00:38:30,239 --> 00:38:32,200
Speaker 5: let me be helpful. Let me summarize the three or

804
00:38:32,200 --> 00:38:35,520
Speaker 5: four news alerts you have. And the summaries were incorrect, right,

805
00:38:35,560 --> 00:38:38,759
Speaker 5: And so that is upsetting as journalists. It's upsetting as

806
00:38:38,760 --> 00:38:42,120
Speaker 5: a user because you want accurate information and Apple is

807
00:38:42,280 --> 00:38:43,240
Speaker 5: muddling the waters.

808
00:38:43,280 --> 00:38:44,879
Speaker 4: And so they had to pull that back.

809
00:38:44,920 --> 00:38:47,239
Speaker 5: And people are starting to seriously ask the question, is

810
00:38:47,280 --> 00:38:50,360
Speaker 5: this the moment where a company like Apple kind of

811
00:38:50,400 --> 00:38:53,840
Speaker 5: falls from grace and loses its steam, loses its power.

812
00:38:54,560 --> 00:38:57,200
Speaker 1: But here's the thing. Fifty percent of Apple's revenues come

813
00:38:57,239 --> 00:39:01,680
Speaker 1: from setting the iPhone. And I can't imagine why do

814
00:39:01,719 --> 00:39:03,120
Speaker 1: they need to be a monket eater in AI? Why

815
00:39:03,120 --> 00:39:04,840
Speaker 1: can't they license other people they eye products?

816
00:39:04,920 --> 00:39:07,239
Speaker 5: Yeah, I mean Apple doesn't really like doing that. They

817
00:39:07,560 --> 00:39:10,360
Speaker 5: like to do everything on their own. They now build

818
00:39:10,360 --> 00:39:13,160
Speaker 5: their own computer chips, they build their own software.

819
00:39:13,160 --> 00:39:15,560
Speaker 4: And the whole point of Apple was.

820
00:39:15,520 --> 00:39:18,040
Speaker 5: That it wasn't a PC, right, they had their own

821
00:39:18,160 --> 00:39:21,520
Speaker 5: operating system. And so that is a huge part of

822
00:39:21,640 --> 00:39:25,520
Speaker 5: Apple's value where they say, come into our walld garden.

823
00:39:25,600 --> 00:39:27,719
Speaker 5: You're going to pay a lot of money, but you're

824
00:39:27,800 --> 00:39:30,200
Speaker 5: going to be more secure, it's going to be cool,

825
00:39:30,400 --> 00:39:33,040
Speaker 5: it's going to be more intuitive. We're doing things our way,

826
00:39:33,120 --> 00:39:36,280
Speaker 5: and so that is their whole pitch to the consumer.

827
00:39:36,600 --> 00:39:38,879
Speaker 5: And if they suddenly have to start saying, come into

828
00:39:38,880 --> 00:39:42,200
Speaker 5: our world garden and use Google's AI, come into our

829
00:39:42,200 --> 00:39:43,520
Speaker 5: world garden and use opening.

830
00:39:43,760 --> 00:39:45,640
Speaker 3: Which I do now by the way, Yeah, I mean

831
00:39:45,680 --> 00:39:48,319
Speaker 3: I use my Apple products to use chatchibt. But it's

832
00:39:48,360 --> 00:39:50,960
Speaker 3: also like can you get everything from one guy? I

833
00:39:50,960 --> 00:39:51,319
Speaker 3: don't know?

834
00:39:51,600 --> 00:39:53,479
Speaker 5: Yeah, And I mean this is another big question because

835
00:39:53,480 --> 00:39:55,680
Speaker 5: that's been the struggle over the last ten years between

836
00:39:55,680 --> 00:39:58,040
Speaker 5: these giant tech companies to sort of box each other

837
00:39:58,080 --> 00:40:02,160
Speaker 5: out and try to corral people within their own ecosystems.

838
00:40:02,600 --> 00:40:05,880
Speaker 5: And maybe AI is the technology that kind of blows

839
00:40:05,920 --> 00:40:06,680
Speaker 5: that up in a way.

840
00:40:07,800 --> 00:40:09,120
Speaker 1: Gary, what a great place to end.

841
00:40:09,280 --> 00:40:11,160
Speaker 4: Thank you so much for thank you so much, of course,

842
00:40:11,160 --> 00:40:11,880
Speaker 4: it is my pleasure.

843
00:40:17,320 --> 00:40:20,400
Speaker 2: That's it for this week for tech Stuff, I'm mos Vloshin.

844
00:40:20,000 --> 00:40:23,360
Speaker 3: And I'm Kara Price. This episode was produced by Eliza Dennis,

845
00:40:23,440 --> 00:40:27,440
Speaker 3: Victoria Dominguez, and Adriana Tapia. It was executive produced by

846
00:40:27,480 --> 00:40:31,440
Speaker 3: me Oz Valashan and Kate Osbourne for Kaleidoscope and Katrina

847
00:40:31,480 --> 00:40:35,160
Speaker 3: Norvel for iHeart Podcasts. Jess Crinchich is the engineer and

848
00:40:35,280 --> 00:40:38,759
Speaker 3: Jack Insley mix the episode. Kyle Murdoch wrote our theme song.

849
00:40:39,239 --> 00:40:42,600
Speaker 1: Join us next Wednesday for tech Stuff The Story when

850
00:40:42,640 --> 00:40:46,360
Speaker 1: we'll share an in depth conversation with Reid Hoffmann, legendary

851
00:40:46,360 --> 00:40:50,040
Speaker 1: founder of LinkedIn, venture capitalist at gray Lock, and author

852
00:40:50,080 --> 00:40:53,200
Speaker 1: of the new book Super Agency. What could possibly go

853
00:40:53,320 --> 00:40:54,800
Speaker 1: right with our AI future?

854
00:40:55,120 --> 00:40:57,600
Speaker 3: Please rate, review, and reach out to us at tech

855
00:40:57,600 --> 00:41:00,719
Speaker 3: Stuff podcast at gmail dot com. We want to hear

856
00:41:00,719 --> 00:41:01,640
Speaker 3: from you.