1
00:00:03,480 --> 00:00:11,160
Speaker 1: Bloomberg Audio Studios, podcasts, radio news. Is it just me?

2
00:00:11,520 --> 00:00:14,160
Speaker 1: Or is AI suddenly everywhere?

3
00:00:14,720 --> 00:00:17,919
Speaker 2: Right now? Both in the US and China ecosystem, The

4
00:00:18,000 --> 00:00:21,040
Speaker 2: AI space is racing on the breaks, taking.

5
00:00:20,840 --> 00:00:22,360
Speaker 3: Sea is everywhere.

6
00:00:22,520 --> 00:00:24,720
Speaker 4: It's not that big scary thing in the future.

7
00:00:25,040 --> 00:00:28,440
Speaker 1: AI is here with us. You open up Google and

8
00:00:28,600 --> 00:00:31,400
Speaker 1: instead of getting a list of websites, there's now often

9
00:00:31,520 --> 00:00:36,320
Speaker 1: an AI overview. This summary scraped together from a sometimes

10
00:00:36,360 --> 00:00:40,599
Speaker 1: strange variety of sources. One recently told me that cheese

11
00:00:41,159 --> 00:00:44,960
Speaker 1: was great at preventing cavities, citing the Cleveland Clinic, the

12
00:00:44,960 --> 00:00:50,960
Speaker 1: British Heart Foundation, and the noted Vermont cheesemaker Cabot as sources.

13
00:00:51,840 --> 00:00:55,200
Speaker 1: And it's not just Google. Apple wants to power its

14
00:00:55,320 --> 00:01:01,160
Speaker 1: iPhones with Google's Gemini generated AI, Alhibet.

15
00:01:00,680 --> 00:01:04,920
Speaker 2: And Meta, offering millions to Hollywood to partner on artificial intelligence.

16
00:01:05,319 --> 00:01:09,160
Speaker 1: From Instagram to customer service, even to healthcare, there's a

17
00:01:09,280 --> 00:01:13,440
Speaker 1: generative AI bot that seemingly has all the answers, if

18
00:01:13,480 --> 00:01:17,000
Speaker 1: not necessarily the correct ones. All this has led to

19
00:01:17,040 --> 00:01:27,480
Speaker 1: big questions recently about what's going into these models. On

20
00:01:27,560 --> 00:01:31,480
Speaker 1: today's episode, what's behind the recent AI headlines, and what

21
00:01:31,520 --> 00:01:34,040
Speaker 1: does all this mean for our ability to opt in

22
00:01:34,640 --> 00:01:38,679
Speaker 1: or opt out of using AI in the future. I'm

23
00:01:38,720 --> 00:01:41,679
Speaker 1: David Gurra, and this is the big take from Bloomberg News.

24
00:01:47,240 --> 00:01:49,560
Speaker 1: We're going to get back to those AI overview search

25
00:01:49,600 --> 00:01:52,440
Speaker 1: results a little bit later in this episode, but first

26
00:01:52,480 --> 00:01:55,080
Speaker 1: we're going to start unpacking all of this by looking

27
00:01:55,120 --> 00:01:59,120
Speaker 1: at what's been going on with another company, Open Ai. Well,

28
00:01:59,120 --> 00:02:01,760
Speaker 1: six months ago, the company behind chat GPT was in

29
00:02:01,800 --> 00:02:04,880
Speaker 1: the headlines after a leadership struggle broke out between the

30
00:02:04,880 --> 00:02:09,079
Speaker 1: company's founders. Now open Ai is back in the news

31
00:02:09,200 --> 00:02:10,760
Speaker 1: for a whole different reason.

32
00:02:11,120 --> 00:02:12,960
Speaker 4: So well, lot's been going on the last few weeks.

33
00:02:13,520 --> 00:02:15,440
Speaker 4: Sometimes in my head I call it as the Open

34
00:02:15,480 --> 00:02:16,240
Speaker 4: Ai turns.

35
00:02:16,880 --> 00:02:19,839
Speaker 1: Rachel Metz is an AI reporter with Bloomberg. She says

36
00:02:19,880 --> 00:02:22,920
Speaker 1: the open Ai soap opera got a new plotline a

37
00:02:22,919 --> 00:02:26,080
Speaker 1: few weeks ago when the company released an update to

38
00:02:26,200 --> 00:02:26,840
Speaker 1: Chat GPT.

39
00:02:27,280 --> 00:02:31,520
Speaker 4: They introduced their latest flagship AI model, This is GPT

40
00:02:31,760 --> 00:02:33,119
Speaker 4: four O.

41
00:02:33,120 --> 00:02:36,360
Speaker 1: Open Ai unveiled human sounding voices for chat GPT in

42
00:02:36,440 --> 00:02:40,280
Speaker 1: September of last year, but the update really only made

43
00:02:40,280 --> 00:02:43,240
Speaker 1: headlines a few weeks ago, after the company hosted a

44
00:02:43,280 --> 00:02:45,480
Speaker 1: live stream to show off what it considers to be

45
00:02:45,680 --> 00:02:46,880
Speaker 1: a big leap forward.

46
00:02:47,400 --> 00:02:50,200
Speaker 5: We'll be showing some live demos today to show the

47
00:02:50,280 --> 00:02:53,600
Speaker 5: full extent of the capabilities of our new model.

48
00:02:53,919 --> 00:02:57,079
Speaker 1: It was during the demonstration when open AI's Mark Chen

49
00:02:57,280 --> 00:03:00,040
Speaker 1: unveiled GPT four Oh's lifelike voices.

50
00:03:00,080 --> 00:03:02,240
Speaker 2: Hey, catching Fatu him Mark? How are you?

51
00:03:03,120 --> 00:03:08,000
Speaker 5: Oh, Mark, I'm doing great, Thanks for asking how about you?

52
00:03:08,040 --> 00:03:10,640
Speaker 1: That a lot of people notice that this voice sounded

53
00:03:10,680 --> 00:03:13,840
Speaker 1: remarkably like the voice of the actress Scarlett Johansson.

54
00:03:14,200 --> 00:03:17,280
Speaker 4: Oh, you're doing a live demb right now. That's awesome.

55
00:03:18,400 --> 00:03:22,680
Speaker 2: Just take a deep breath and remember you're the experts.

56
00:03:22,800 --> 00:03:27,799
Speaker 1: The perception was furthered when Sam Altman, OpenAI's reinstated CEO,

57
00:03:28,200 --> 00:03:32,400
Speaker 1: posted a single word to X after the event, Her,

58
00:03:32,960 --> 00:03:35,200
Speaker 1: which some took as a nod to the twenty thirteen

59
00:03:35,280 --> 00:03:38,560
Speaker 1: Spike Jones film in which johanson voice is the Open

60
00:03:38,680 --> 00:03:41,040
Speaker 1: Operating system Samantha.

61
00:03:40,800 --> 00:03:43,480
Speaker 4: And it turns out Scarlett Johansson thought it also sounded

62
00:03:43,520 --> 00:03:44,200
Speaker 4: a lot like her.

63
00:03:44,320 --> 00:03:48,080
Speaker 1: The thing is Altman had approached Johansson twice about potentially

64
00:03:48,120 --> 00:03:52,280
Speaker 1: working with open Ai, but Johansson turned him down. To

65
00:03:52,360 --> 00:03:55,160
Speaker 1: some watching, it felt like a bunch of Silicon Valley

66
00:03:55,200 --> 00:03:57,760
Speaker 1: types had ignored the wishes of one of Hollywood's most

67
00:03:57,760 --> 00:04:00,760
Speaker 1: famous actresses just so they could make the plot line

68
00:04:00,760 --> 00:04:02,440
Speaker 1: of a movie a reality.

69
00:04:02,760 --> 00:04:06,200
Speaker 4: She seemed quite genuinely upset about it and thought it

70
00:04:06,240 --> 00:04:09,640
Speaker 4: sounded like her, even though she had specifically said I

71
00:04:09,640 --> 00:04:12,600
Speaker 4: don't want to participate in this project on multiple occasions.

72
00:04:12,480 --> 00:04:15,680
Speaker 1: For anyone feeling anxious about the ubiquity of AI and

73
00:04:15,760 --> 00:04:19,479
Speaker 1: about the information these generative models are hoovering up the

74
00:04:19,560 --> 00:04:21,039
Speaker 1: dust up struck a nerve.

75
00:04:21,440 --> 00:04:23,960
Speaker 4: I think there's a few different things that go into

76
00:04:24,000 --> 00:04:28,679
Speaker 4: making people feel really alarmed by this. One is people

77
00:04:28,960 --> 00:04:33,520
Speaker 4: really like Scarlet Johanson. She is an iconic actress. Another

78
00:04:33,520 --> 00:04:35,880
Speaker 4: thing that I think is interesting about her in particular

79
00:04:36,080 --> 00:04:39,080
Speaker 4: is like yourself, David, as you were saying, you talk

80
00:04:39,200 --> 00:04:42,680
Speaker 4: for a living. Your voice is valuable. Her voice is

81
00:04:42,760 --> 00:04:46,120
Speaker 4: valuable as an actor, but it is also iconic. People

82
00:04:46,240 --> 00:04:49,400
Speaker 4: know her voice in part because of the movie her

83
00:04:49,760 --> 00:04:52,440
Speaker 4: but also just because she happens to have a voice

84
00:04:52,480 --> 00:04:56,120
Speaker 4: that is very easily recognizable. I think a less recognizable

85
00:04:56,200 --> 00:05:00,279
Speaker 4: voice people might not be taking as much to this issue.

86
00:05:00,400 --> 00:05:04,120
Speaker 1: Open AI says it never intended to use Johansson's voice

87
00:05:04,240 --> 00:05:05,400
Speaker 1: as one of its assistants.

88
00:05:05,800 --> 00:05:09,560
Speaker 4: Basically, what they said is they started working on the

89
00:05:09,640 --> 00:05:12,599
Speaker 4: voice feature way back in spring of twenty twenty three,

90
00:05:12,960 --> 00:05:16,120
Speaker 4: and they cast a number of voice actors for five

91
00:05:16,120 --> 00:05:18,720
Speaker 4: different voices and said that they were planning on her

92
00:05:18,760 --> 00:05:22,680
Speaker 4: being a sixth voice. So it's not they're saying, Look,

93
00:05:22,680 --> 00:05:24,760
Speaker 4: it's not that she would be that voice, or that

94
00:05:24,839 --> 00:05:27,000
Speaker 4: this voice is meant to sound like her. This is

95
00:05:27,040 --> 00:05:27,760
Speaker 4: another person.

96
00:05:28,240 --> 00:05:31,560
Speaker 1: But after the backlash from Johansson and the public, open

97
00:05:31,600 --> 00:05:35,560
Speaker 1: ai removed the voice, which Rachel says is something of

98
00:05:35,600 --> 00:05:38,640
Speaker 1: a turning point in terms of our ability to opt

99
00:05:38,680 --> 00:05:42,080
Speaker 1: out or at least push back against these generative models

100
00:05:42,480 --> 00:05:45,960
Speaker 1: that seemed to be pushing forward with seemingly few guardrails

101
00:05:45,960 --> 00:05:46,480
Speaker 1: in place.

102
00:05:46,800 --> 00:05:49,080
Speaker 4: A lot of consumers and a lot of artists are saying,

103
00:05:49,320 --> 00:05:51,839
Speaker 4: wait a minute, that's not what I signed up for

104
00:05:51,920 --> 00:05:54,240
Speaker 4: when I put this and that on the internet.

105
00:05:54,680 --> 00:05:57,480
Speaker 1: Just this week, open ai introduced a new safety board

106
00:05:57,680 --> 00:06:01,120
Speaker 1: led by CEO Sam Altman. Move came after the company

107
00:06:01,160 --> 00:06:03,760
Speaker 1: disbanded a team that had been created to focus on

108
00:06:03,800 --> 00:06:07,360
Speaker 1: the long term threat AI could post to humanity. Some

109
00:06:07,400 --> 00:06:11,400
Speaker 1: former insiders are saying that probably won't be enough, including

110
00:06:11,440 --> 00:06:15,080
Speaker 1: former OpenAI board member Helen Toner. Toner told the ted

111
00:06:15,120 --> 00:06:18,160
Speaker 1: Ai Show podcast this week that she had concerns about

112
00:06:18,160 --> 00:06:21,640
Speaker 1: Altman's commitment to safety based on his past behavior.

113
00:06:22,040 --> 00:06:26,160
Speaker 2: On multiple occasions he gave us inaccurate information about the

114
00:06:26,200 --> 00:06:29,280
Speaker 2: small number of formal safety processes that the company did

115
00:06:29,279 --> 00:06:32,320
Speaker 2: have in place, meaning that it was basically impossible for

116
00:06:32,360 --> 00:06:35,000
Speaker 2: the board to know how well those safety processes we're

117
00:06:35,000 --> 00:06:36,960
Speaker 2: working or what might need to change.

118
00:06:37,360 --> 00:06:40,160
Speaker 1: And it's not just open ai that's been getting some pushback.

119
00:06:41,800 --> 00:06:44,400
Speaker 1: After the break, we take a look at another sector

120
00:06:44,560 --> 00:06:47,839
Speaker 1: that's grappling with how and when to fight developments in

121
00:06:47,920 --> 00:06:50,960
Speaker 1: generative AI. Can we weigh up whether or not any

122
00:06:51,000 --> 00:06:53,240
Speaker 1: of this will be enough to shake up what some

123
00:06:53,400 --> 00:06:57,040
Speaker 1: are calling the raw deal at the center of AI.

124
00:07:03,560 --> 00:07:06,599
Speaker 1: Open Ai is certainly not the only company racing to

125
00:07:06,640 --> 00:07:10,880
Speaker 1: release new AI products. There are startups like Anthropic and

126
00:07:11,040 --> 00:07:14,880
Speaker 1: Elon Musk's Xai, which just announced this week it raised

127
00:07:15,000 --> 00:07:19,320
Speaker 1: six billion dollars, giving the company a valuation of twenty

128
00:07:19,320 --> 00:07:23,520
Speaker 1: four billion dollars. There's also Meta, which recently rolled out

129
00:07:23,560 --> 00:07:27,320
Speaker 1: its ask meta feature to Facebook and Instagram and WhatsApp.

130
00:07:27,800 --> 00:07:32,120
Speaker 1: And then there's Alphabet, which just added those AI overviews

131
00:07:32,160 --> 00:07:33,120
Speaker 1: to Google.

132
00:07:33,200 --> 00:07:36,720
Speaker 5: So iioverviews they announced this Developers conference.

133
00:07:37,000 --> 00:07:40,400
Speaker 1: Dave Lee is a Bloomberg opinion columnist. He recently attended

134
00:07:40,400 --> 00:07:43,640
Speaker 1: the Google Developer Conference, where the company unveiled this latest

135
00:07:43,720 --> 00:07:45,120
Speaker 1: change to its search engine.

136
00:07:45,240 --> 00:07:48,240
Speaker 3: The result is a product that does the work for you.

137
00:07:49,080 --> 00:07:53,960
Speaker 3: Google Search is generative AI at the scale of human curiosity,

138
00:07:54,480 --> 00:07:56,960
Speaker 3: and it's our most exciting chapter of search yet.

139
00:07:57,440 --> 00:08:00,440
Speaker 5: And basically what it does is it's all builds on

140
00:08:00,440 --> 00:08:02,920
Speaker 5: something that's already been on Google. If you search for,

141
00:08:03,440 --> 00:08:06,280
Speaker 5: you know, who are the Beatles, you'll get like a

142
00:08:06,280 --> 00:08:08,080
Speaker 5: sort of fact box, and it will have pictures and

143
00:08:08,120 --> 00:08:10,600
Speaker 5: a quote from Wikipedia and all that kind of stuff.

144
00:08:10,880 --> 00:08:13,400
Speaker 5: Aio overviews is sort of that, and then some.

145
00:08:14,800 --> 00:08:19,040
Speaker 1: AI overview synthesizes information from a host of different sources

146
00:08:19,120 --> 00:08:22,280
Speaker 1: into a generated answer, similar to something you'd get if

147
00:08:22,320 --> 00:08:25,480
Speaker 1: you typed a question into say Open AI's chat GPT.

148
00:08:26,280 --> 00:08:28,240
Speaker 1: At the bottom, there are some links to where that

149
00:08:28,280 --> 00:08:31,400
Speaker 1: information came from if you're curious, which is all well

150
00:08:31,400 --> 00:08:34,439
Speaker 1: and good for Google, but critics of this approach point

151
00:08:34,480 --> 00:08:38,560
Speaker 1: out that it deprioritizes going directly to a source, and

152
00:08:38,600 --> 00:08:41,199
Speaker 1: that means less traffic to those sites.

153
00:08:41,600 --> 00:08:44,000
Speaker 5: If you're the website that's crazing that information. Why on

154
00:08:44,040 --> 00:08:47,440
Speaker 5: earth would you continue to do it if you know

155
00:08:47,480 --> 00:08:49,640
Speaker 5: for all these popular searches, you're not going to get

156
00:08:49,640 --> 00:08:52,760
Speaker 5: that kickback of traffic, which means that revenue, which means

157
00:08:53,320 --> 00:08:56,959
Speaker 5: continuing to exist. And that's a really, really big fear

158
00:08:57,040 --> 00:08:58,280
Speaker 5: with AI overviews.

159
00:08:58,760 --> 00:09:01,280
Speaker 1: Google has it rolled out the feature to every search

160
00:09:01,559 --> 00:09:04,160
Speaker 1: just yet, but the mere prospect that it will do

161
00:09:04,280 --> 00:09:08,560
Speaker 1: so has created an existential threat for news publishers. But

162
00:09:08,640 --> 00:09:11,800
Speaker 1: Dave says it's not just news publishers who should be worried.

163
00:09:12,040 --> 00:09:13,920
Speaker 5: I also think there's a big problem for Google as well,

164
00:09:13,960 --> 00:09:18,520
Speaker 5: because because Google needs this web ecosystem to exist. And

165
00:09:18,840 --> 00:09:20,679
Speaker 5: I didn't get the sense, and I tooked to a

166
00:09:20,720 --> 00:09:24,680
Speaker 5: Google executive afterwards their event, I didn't get the sense

167
00:09:24,720 --> 00:09:28,400
Speaker 5: that they really quite comprehended how damaging and how quickly

168
00:09:28,440 --> 00:09:29,200
Speaker 5: damaging it might be.

169
00:09:30,840 --> 00:09:35,120
Speaker 1: This web ecosystem can only exist if publishers can stay

170
00:09:35,120 --> 00:09:38,480
Speaker 1: in business. One way that AI companies have been attempting

171
00:09:38,520 --> 00:09:41,400
Speaker 1: to solve this puzzle is by cutting publishers into some

172
00:09:41,480 --> 00:09:46,400
Speaker 1: of their current and potentially future profits. Take open ai again.

173
00:09:47,240 --> 00:09:49,440
Speaker 5: So what they've been doing is going to going out

174
00:09:49,520 --> 00:09:52,520
Speaker 5: to as many publishers as they can trying to come

175
00:09:52,600 --> 00:09:57,000
Speaker 5: up with these deals to allow open Eye or whoever

176
00:09:58,160 --> 00:10:01,960
Speaker 5: use that information from those publishers on their lash anguish models.

177
00:10:02,160 --> 00:10:05,160
Speaker 1: And some publishers have agreed to take these deals, including

178
00:10:05,160 --> 00:10:08,320
Speaker 1: The Atlantic and Vox Media, both of which ink deals

179
00:10:08,320 --> 00:10:11,240
Speaker 1: on Wednesday. What are the terms of the deals? What's

180
00:10:11,800 --> 00:10:13,000
Speaker 1: what's the incentive for them to do this?

181
00:10:13,120 --> 00:10:15,000
Speaker 5: I wish I'd seen the terms of the deals. I mean,

182
00:10:15,080 --> 00:10:17,000
Speaker 5: this is one of the criticisms of these deals is

183
00:10:17,000 --> 00:10:19,080
Speaker 5: that we don't know a great deal of what, you know,

184
00:10:19,080 --> 00:10:20,720
Speaker 5: the nitty gritty here, what's happening.

185
00:10:21,320 --> 00:10:22,360
Speaker 1: Some publishers, yes.

186
00:10:22,880 --> 00:10:25,600
Speaker 5: Have made deals. The most notable one is News Corporation

187
00:10:25,920 --> 00:10:27,440
Speaker 5: for a lot of money, for a lot of money,

188
00:10:27,600 --> 00:10:30,520
Speaker 5: two hundred and fifty million over five years. Now that

189
00:10:30,559 --> 00:10:32,760
Speaker 5: doesn't necessarily mean they're getting a nice big check for

190
00:10:32,760 --> 00:10:34,839
Speaker 5: two hundred and fifty million, because part of that deal

191
00:10:35,000 --> 00:10:38,880
Speaker 5: is using open ai technology within news Corps and what

192
00:10:38,920 --> 00:10:40,720
Speaker 5: that looks like we don't know. We don't know whether

193
00:10:40,760 --> 00:10:42,959
Speaker 5: open Ai has said, guess what, using our technology is

194
00:10:43,000 --> 00:10:45,400
Speaker 5: worth one hundred million dollars a year. We don't have

195
00:10:45,440 --> 00:10:49,280
Speaker 5: those details, which is very frustrating, But you know, there

196
00:10:49,360 --> 00:10:52,520
Speaker 5: is there is a willingness to make friends of open

197
00:10:52,559 --> 00:10:56,200
Speaker 5: AI because there's a general feeling. And the chairman of Lamont,

198
00:10:56,360 --> 00:10:59,760
Speaker 5: the French newspaper. I'm paraphrasing here, but he basically said, look,

199
00:10:59,760 --> 00:11:02,840
Speaker 5: even we do this with them and we get some money,

200
00:11:03,320 --> 00:11:05,600
Speaker 5: or they do it anyway with our content and we

201
00:11:05,640 --> 00:11:08,520
Speaker 5: have no control over it and we don't benefit, and

202
00:11:08,520 --> 00:11:10,160
Speaker 5: it's going to harm us as a publisher.

203
00:11:10,600 --> 00:11:13,559
Speaker 1: Dave says he believes publishers are making a big mistake

204
00:11:13,840 --> 00:11:17,120
Speaker 1: by agreeing to sell their content to open AI, even

205
00:11:17,160 --> 00:11:21,000
Speaker 1: if he understands the economic incentives pushing them to strike

206
00:11:21,040 --> 00:11:24,600
Speaker 1: deals that could help insulate them against potential revenue short

207
00:11:24,640 --> 00:11:27,800
Speaker 1: falls in the future. And there are some holdouts.

208
00:11:28,360 --> 00:11:31,040
Speaker 5: The big one at the moment is the New York Times.

209
00:11:31,320 --> 00:11:35,160
Speaker 5: They suit open AI. They said them for copyright infringement,

210
00:11:35,200 --> 00:11:37,120
Speaker 5: which is going to be an interesting one to make

211
00:11:37,160 --> 00:11:40,199
Speaker 5: in court because their argument is that when you said

212
00:11:40,200 --> 00:11:44,120
Speaker 5: certain things the chat GPT, it would recreate bits of

213
00:11:44,200 --> 00:11:47,600
Speaker 5: timed journalism and that was copyright infringement. There won't be

214
00:11:47,640 --> 00:11:49,360
Speaker 5: a preliminary hearing in that case for it for a

215
00:11:49,400 --> 00:11:51,480
Speaker 5: number of months, which is you know, just kind of

216
00:11:51,480 --> 00:11:53,640
Speaker 5: goes to show. I mean, think how much has changed

217
00:11:53,640 --> 00:11:55,640
Speaker 5: an AI in just the last two years or so.

218
00:11:56,120 --> 00:11:57,960
Speaker 5: You know, with every passing day, it feels like this

219
00:11:58,040 --> 00:12:00,240
Speaker 5: challenge is getting bigger and bigger and bigger.

220
00:12:00,320 --> 00:12:03,600
Speaker 1: There's also another group called old in Global Capital, which

221
00:12:03,679 --> 00:12:05,840
Speaker 1: owns a bunch of titles like The New York Daily

222
00:12:05,920 --> 00:12:08,840
Speaker 1: News and The Chicago Tribune. They're using the same sort

223
00:12:08,880 --> 00:12:11,880
Speaker 1: of logic to sue open AI and adding another element.

224
00:12:12,520 --> 00:12:14,440
Speaker 1: It has to do with a well known flaw in

225
00:12:14,520 --> 00:12:17,160
Speaker 1: a lot of generative AI tools we've been using, which

226
00:12:17,280 --> 00:12:20,720
Speaker 1: is their propensity to hallucinate or make up answers.

227
00:12:21,120 --> 00:12:24,720
Speaker 5: That's another interesting thing because these hallucinations couldn't be damaging

228
00:12:24,720 --> 00:12:27,560
Speaker 5: to these news brands as well. If these machines are saying,

229
00:12:27,760 --> 00:12:29,920
Speaker 5: you know, crunching two bits of information and saying, well,

230
00:12:29,920 --> 00:12:31,560
Speaker 5: that came from such and such and it might not.

231
00:12:31,960 --> 00:12:35,200
Speaker 1: What does that mean for the news industry. If publishers

232
00:12:35,200 --> 00:12:38,560
Speaker 1: are reluctant to or unable to opt out from what

233
00:12:38,800 --> 00:12:41,000
Speaker 1: these AI companies are trying to do.

234
00:12:42,120 --> 00:12:46,319
Speaker 5: That's a profound question. Right, publishers have had perilous business

235
00:12:46,400 --> 00:12:49,400
Speaker 5: models for so long now, and sort of one by

236
00:12:49,440 --> 00:12:51,839
Speaker 5: one they've been they've been sort of removed by bits

237
00:12:51,880 --> 00:12:54,480
Speaker 5: of tech, you know, whether it was Craigslist for classifieds

238
00:12:54,559 --> 00:12:56,439
Speaker 5: or you know, Google just getting a lot of the

239
00:12:56,480 --> 00:13:00,000
Speaker 5: advertising revenue from from one over the place. You've got

240
00:13:00,120 --> 00:13:03,080
Speaker 5: this situation where I think, already, and we've seen this

241
00:13:03,160 --> 00:13:05,920
Speaker 5: from the reaction to AI overviews from just regular users,

242
00:13:06,320 --> 00:13:09,439
Speaker 5: there is a suspicion of these sort of computer generated

243
00:13:09,559 --> 00:13:12,000
Speaker 5: fits of information. I saw a great quote and it's

244
00:13:12,000 --> 00:13:13,600
Speaker 5: gone around the Internet so much now that I've no

245
00:13:13,640 --> 00:13:15,600
Speaker 5: idea who first said it, which is kind of good

246
00:13:15,600 --> 00:13:17,200
Speaker 5: because it kind of just belongs to us all now.

247
00:13:17,760 --> 00:13:19,120
Speaker 5: And one of them was, you know, why would I

248
00:13:19,160 --> 00:13:21,640
Speaker 5: be bothered to read something you couldn't be bothered to write?

249
00:13:22,160 --> 00:13:25,080
Speaker 5: And I kind of hope. I hope that sort of

250
00:13:25,720 --> 00:13:28,200
Speaker 5: attitude is going to persist. I hold I hope that

251
00:13:28,960 --> 00:13:32,000
Speaker 5: people will care about human made things, whether it's newspapers

252
00:13:32,120 --> 00:13:33,440
Speaker 5: or movies or books or whatever.

253
00:13:33,880 --> 00:13:36,360
Speaker 1: But for now, at least, there isn't an easy way

254
00:13:36,400 --> 00:13:39,240
Speaker 1: to opt down. Here's Bloomberg's Rachel Mets again.

255
00:13:40,440 --> 00:13:43,440
Speaker 4: These companies that are ingesting tons and tons and tons

256
00:13:43,679 --> 00:13:47,160
Speaker 4: of data using that to train their AI systems, And

257
00:13:47,240 --> 00:13:49,520
Speaker 4: if we want these AI systems to get better, the

258
00:13:49,559 --> 00:13:53,720
Speaker 4: current prevailing thought is we need more data and more

259
00:13:53,720 --> 00:13:56,760
Speaker 4: compute to make them more better. That is how it's

260
00:13:56,760 --> 00:13:58,800
Speaker 4: working right now. I have my own thoughts about whether

261
00:13:58,840 --> 00:14:00,640
Speaker 4: that will be true in the future, but right now,

262
00:14:00,679 --> 00:14:03,120
Speaker 4: that's what we're seeing, is we're getting better results from

263
00:14:03,160 --> 00:14:06,280
Speaker 4: more data and more compute. And in that sense, we're

264
00:14:06,400 --> 00:14:09,640
Speaker 4: all sort of opted in in various ways, depending on

265
00:14:09,920 --> 00:14:11,760
Speaker 4: how much of our lives have been lived on the

266
00:14:11,760 --> 00:14:14,160
Speaker 4: Internet and how much of that data these companies are using.

267
00:14:14,600 --> 00:14:17,280
Speaker 1: But just because we're all effectively opted in, whether we

268
00:14:17,440 --> 00:14:19,640
Speaker 1: like it or not, doesn't mean that we don't have

269
00:14:19,680 --> 00:14:22,360
Speaker 1: a say in how our data are written, words, and

270
00:14:22,400 --> 00:14:25,560
Speaker 1: our voices are being used, even if we're not all

271
00:14:25,560 --> 00:14:28,720
Speaker 1: as famous as Scarlett Johansson. Rachel says, the fact that

272
00:14:28,760 --> 00:14:32,040
Speaker 1: OpenAI listened, that it pulled the Sky Voice in the

273
00:14:32,040 --> 00:14:36,240
Speaker 1: face of a public outcry, means there's room to well

274
00:14:36,800 --> 00:14:40,560
Speaker 1: have a voice in how generative AI gets developed and used.

275
00:14:40,960 --> 00:14:44,800
Speaker 4: I don't like to think that anything with technology is inevitable.

276
00:14:45,760 --> 00:14:47,920
Speaker 4: I like to think that we have a lot of

277
00:14:47,920 --> 00:14:50,920
Speaker 4: control over what happens, and I think that what we're

278
00:14:50,920 --> 00:14:54,720
Speaker 4: seeing with some of this pushback recently against this voice

279
00:14:54,880 --> 00:14:58,360
Speaker 4: that people feel and Scarlet Johansson feels sounds like Scarlett Johansson.

280
00:14:58,600 --> 00:15:00,480
Speaker 4: I feel like that's a really good examp love it.

281
00:15:00,720 --> 00:15:03,080
Speaker 4: So I feel like that's sort of a hopeful sign,

282
00:15:03,160 --> 00:15:05,640
Speaker 4: right that we still have agency and we still have control.

283
00:15:10,160 --> 00:15:13,000
Speaker 1: This is the Big Take from Bloomberg News. I'm David Gura.

284
00:15:13,600 --> 00:15:16,680
Speaker 1: This episode was produced by Thomas lu It was edited

285
00:15:16,680 --> 00:15:19,640
Speaker 1: by Aaron Edwards. It was mixed by Blake Maples. It

286
00:15:19,720 --> 00:15:23,240
Speaker 1: was fact checked by Adriana Tapia. Our senior producers are

287
00:15:23,280 --> 00:15:27,440
Speaker 1: Kim Gittleson and Naomi Shaven. Our senior editor is Elizabeth Ponso.

288
00:15:28,120 --> 00:15:31,400
Speaker 1: Nicole Beemster bor is our executive producer. Sage Bauman is

289
00:15:31,400 --> 00:15:35,080
Speaker 1: our head of podcasts. Thanks for listening. Please follow and

290
00:15:35,120 --> 00:15:37,840
Speaker 1: review The Big Take wherever you get your podcasts. It

291
00:15:37,920 --> 00:15:40,920
Speaker 1: helps new listeners find the show. We'll be back tomorrow.