1
00:00:03,120 --> 00:00:08,720
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:09,600 --> 00:00:12,360
Speaker 2: The last week of August isn't exactly known as a

3
00:00:12,400 --> 00:00:17,040
Speaker 2: busy time of year unless you work in California's legislature.

4
00:00:17,480 --> 00:00:21,239
Speaker 3: I'm Senator Scott Wiener, and I am calling in from

5
00:00:21,520 --> 00:00:23,360
Speaker 3: the state capitol in California.

6
00:00:23,480 --> 00:00:26,239
Speaker 2: When we talked recently, Senator Wiener told me he was

7
00:00:26,320 --> 00:00:29,400
Speaker 2: running from vote to vote on everything from regulating hedge

8
00:00:29,440 --> 00:00:34,200
Speaker 2: funds to allowing outdoor cannabis sales, very California stuff.

9
00:00:34,600 --> 00:00:37,240
Speaker 3: We are in our last week of our legislative session,

10
00:00:37,320 --> 00:00:41,800
Speaker 3: so every day is marathon voting. We've already voted on

11
00:00:41,840 --> 00:00:45,000
Speaker 3: a few hundred bills and we have one hundreds more

12
00:00:45,000 --> 00:00:47,080
Speaker 3: before our session ends on Saturday.

13
00:00:47,600 --> 00:00:50,080
Speaker 2: But I wanted to speak to Senator Wiener about one

14
00:00:50,320 --> 00:00:53,040
Speaker 2: very specific bill that was being voted on that day.

15
00:00:53,760 --> 00:00:57,040
Speaker 2: It was one he introduced back in February called SB

16
00:00:57,160 --> 00:01:01,160
Speaker 2: ten forty seven, the Safe and Secure Innovation for Frontier

17
00:01:01,280 --> 00:01:05,640
Speaker 2: Artificial Intelligence Models Act. The bill would hold AI companies

18
00:01:05,680 --> 00:01:09,720
Speaker 2: to new safety standards and hold them legally liable if

19
00:01:09,760 --> 00:01:13,680
Speaker 2: their tools end up causing catastrophic harm. The bill would

20
00:01:13,760 --> 00:01:16,720
Speaker 2: only apply to the biggest AI players, the ones that

21
00:01:16,760 --> 00:01:19,600
Speaker 2: are spending upwards of one hundred million dollars on training

22
00:01:19,600 --> 00:01:23,680
Speaker 2: their AI models or spending ten million dollars fine tuning them.

23
00:01:24,120 --> 00:01:27,399
Speaker 3: We have an obligation in California, as the tech leader

24
00:01:27,880 --> 00:01:32,039
Speaker 3: on the planet, to make good, forward looking, pro innovation,

25
00:01:32,360 --> 00:01:34,280
Speaker 3: pro safety tech policy.

26
00:01:34,840 --> 00:01:37,720
Speaker 2: Now. Senator Wiener has never been one to shy away

27
00:01:37,760 --> 00:01:41,800
Speaker 2: from controversial policy moves. A Democrat, he's been one of

28
00:01:41,840 --> 00:01:45,160
Speaker 2: the biggest champions of zoning reform in the state. But

29
00:01:45,280 --> 00:01:48,520
Speaker 2: introducing this bill in the place where chat, GPT and

30
00:01:48,600 --> 00:01:52,560
Speaker 2: Claude were first unleashed and where OpenAI and anthropic are

31
00:01:52,640 --> 00:01:56,800
Speaker 2: racing towards superintelligence hasn't exactly been popular.

32
00:01:57,320 --> 00:02:00,360
Speaker 3: I knew that there would be significant pushback just because

33
00:02:00,400 --> 00:02:03,080
Speaker 3: of the nature of this subject, and I knew it

34
00:02:03,080 --> 00:02:05,440
Speaker 3: would be forceable, but it was even more than I

35
00:02:05,720 --> 00:02:06,640
Speaker 3: thought that it would be.

36
00:02:07,200 --> 00:02:11,720
Speaker 2: Open AI opposes the bill, so does Nancy Pelosi and

37
00:02:11,720 --> 00:02:15,960
Speaker 2: Andresen Horowitz. But Weiener believes that even with that pushback,

38
00:02:16,360 --> 00:02:19,480
Speaker 2: no state is better positioned to be forward looking in

39
00:02:19,520 --> 00:02:20,240
Speaker 2: this realm.

40
00:02:20,480 --> 00:02:25,000
Speaker 3: California has led on tech policy, on climate policy, on

41
00:02:25,760 --> 00:02:29,320
Speaker 3: so many issues where we are ahead of Congress and

42
00:02:29,360 --> 00:02:31,799
Speaker 3: we set the trend, and we should do that here

43
00:02:31,840 --> 00:02:32,639
Speaker 3: as well.

44
00:02:37,480 --> 00:02:40,000
Speaker 2: Today on the show. The debate over the bill to

45
00:02:40,200 --> 00:02:43,480
Speaker 2: curb AI harms has become a lightning rod for the

46
00:02:43,520 --> 00:02:47,119
Speaker 2: tech community. Could it be the future of tech regulation

47
00:02:47,919 --> 00:02:51,480
Speaker 2: or a cautionary tale? This is the big take from

48
00:02:51,520 --> 00:02:59,760
Speaker 2: Bloomberg News. I'm Sarah Holder. The idea for SB ten

49
00:02:59,800 --> 00:03:03,240
Speaker 2: forty seven came to California State Senator Scott Wiener about

50
00:03:03,240 --> 00:03:05,440
Speaker 2: a year and a half ago. That's when he was

51
00:03:05,480 --> 00:03:08,000
Speaker 2: approached by a group of experts who wanted him to

52
00:03:08,040 --> 00:03:09,800
Speaker 2: take on AI regulation.

53
00:03:10,480 --> 00:03:13,720
Speaker 3: Sometimes in the early part of twenty twenty three, some

54
00:03:14,200 --> 00:03:20,239
Speaker 3: AI technologists came to me and expressed that we're overdue

55
00:03:20,280 --> 00:03:24,760
Speaker 3: for a conversation about AI safety. As we have all

56
00:03:24,800 --> 00:03:28,919
Speaker 3: this amazing, brilliant innovation, how do we make sure that

57
00:03:29,040 --> 00:03:32,760
Speaker 3: as AI scales, that it happens in a responsible way,

58
00:03:33,360 --> 00:03:36,680
Speaker 3: because we have a history of not getting out in

59
00:03:36,720 --> 00:03:41,120
Speaker 3: front of the risks posed by technology and just letting

60
00:03:41,120 --> 00:03:44,480
Speaker 3: things happen and then having to clean up the mess later.

61
00:03:45,160 --> 00:03:47,600
Speaker 2: Now, there are a lot of AI bills floating around

62
00:03:47,680 --> 00:03:52,240
Speaker 2: in California. In fact, during that August session, legislators passed

63
00:03:52,280 --> 00:03:56,000
Speaker 2: another bill on deep fakes, but SB ten forty seven

64
00:03:56,160 --> 00:03:59,760
Speaker 2: focuses on a whole different universe of potential AI harms,

65
00:04:00,440 --> 00:04:05,520
Speaker 2: catastrophic ones think property damage over five hundred million dollars

66
00:04:05,960 --> 00:04:09,320
Speaker 2: or a massive loss of human life. And Wiener says

67
00:04:09,320 --> 00:04:12,800
Speaker 2: he chose to focus on this particular aspect for a reason.

68
00:04:13,840 --> 00:04:17,200
Speaker 3: In the debate on SB ten forty seven, there's a

69
00:04:17,200 --> 00:04:20,480
Speaker 3: lot of what about ism? So what about deep fakes?

70
00:04:20,480 --> 00:04:24,960
Speaker 3: And what about algorithmic discrimination? What about misinformation? Why are

71
00:04:24,960 --> 00:04:28,200
Speaker 3: you focusing on the risk of catastrophic harm? And my

72
00:04:28,279 --> 00:04:31,080
Speaker 3: answer is, well, we need to adjust all of these issues.

73
00:04:31,480 --> 00:04:35,000
Speaker 3: But in addition to those potential harms, we know that

74
00:04:35,040 --> 00:04:39,640
Speaker 3: there is a risk of larger harms, for example, shutting

75
00:04:39,680 --> 00:04:43,080
Speaker 3: down the grid, melting down the banking system and making

76
00:04:43,120 --> 00:04:46,960
Speaker 3: it easier, or facilitating the creation of a chemical, biological,

77
00:04:47,360 --> 00:04:53,280
Speaker 3: and nuclear weapon destroying critical infrastructure. These are all harms

78
00:04:53,480 --> 00:04:57,479
Speaker 3: that you don't have to have the best imagination on

79
00:04:57,520 --> 00:05:00,440
Speaker 3: the planet in order to visualize.

80
00:05:00,920 --> 00:05:04,680
Speaker 2: These are worst case scenarios, and Wiener has been criticized

81
00:05:04,760 --> 00:05:06,200
Speaker 2: for even imagining them.

82
00:05:06,680 --> 00:05:09,520
Speaker 3: Sometimes the opponents of the bill, they try to say

83
00:05:09,560 --> 00:05:11,839
Speaker 3: that anyone who focuses on AI safety must be a

84
00:05:11,880 --> 00:05:15,159
Speaker 3: crazy doomer, which is not true.

85
00:05:15,400 --> 00:05:17,280
Speaker 2: You don't consider yourself a crazy doomer.

86
00:05:17,440 --> 00:05:19,599
Speaker 3: I am not a crazy doomer. I am very pro

87
00:05:19,720 --> 00:05:23,240
Speaker 3: AI and pro innovation. I just wanted to be done responsibly.

88
00:05:24,680 --> 00:05:28,800
Speaker 2: Still, the risks outlined in the legislation are mostly hypothetical,

89
00:05:29,200 --> 00:05:33,039
Speaker 2: at least as of now. I asked Serene Gafari, a

90
00:05:33,080 --> 00:05:37,400
Speaker 2: reporter who covers AI for Bloomberg News, about how imminent

91
00:05:37,520 --> 00:05:40,440
Speaker 2: the risk of an AI fueled catastrophe really is.

92
00:05:40,880 --> 00:05:43,159
Speaker 4: Okay, this is where you know, it really depends you

93
00:05:43,240 --> 00:05:46,240
Speaker 4: talk to, because even the experts, even the quote unquote

94
00:05:46,400 --> 00:05:49,880
Speaker 4: godfathers of AI who basically invented this field of the

95
00:05:49,920 --> 00:05:54,720
Speaker 4: current type of AI models we're using today, wildly disagree. So,

96
00:05:55,080 --> 00:05:58,240
Speaker 4: you know, someone like Yoshua Benjo, for example, leading computer

97
00:05:58,320 --> 00:06:01,080
Speaker 4: scientists who won the Touring Award, which is like the

98
00:06:01,080 --> 00:06:04,039
Speaker 4: Nobel Prize in math for deep learning, which again is

99
00:06:04,080 --> 00:06:07,159
Speaker 4: like the foundation of AI that we're using today. He

100
00:06:07,360 --> 00:06:09,080
Speaker 4: is a big supporter of this bill. He's come out

101
00:06:09,080 --> 00:06:11,279
Speaker 4: and said, you know, these risks could be coming. He

102
00:06:11,320 --> 00:06:13,200
Speaker 4: doesn't know exactly when, but he says the risk is

103
00:06:13,320 --> 00:06:15,800
Speaker 4: enough that we should be trying to mitigate it. Then

104
00:06:15,880 --> 00:06:18,200
Speaker 4: you have someone else who also won the Tory Award

105
00:06:18,200 --> 00:06:21,400
Speaker 4: with him that year. Another big scientist, Jan Lucuhn, head

106
00:06:21,440 --> 00:06:23,839
Speaker 4: of AI at Facebook. Right now a meta, He says,

107
00:06:24,000 --> 00:06:26,680
Speaker 4: this is ridiculous. You know, we shouldn't be worried about

108
00:06:26,680 --> 00:06:29,480
Speaker 4: these catastrophic risks. These AI tools today are nowhere near

109
00:06:29,520 --> 00:06:32,200
Speaker 4: as smart as human beings, and we're all way too

110
00:06:32,240 --> 00:06:33,479
Speaker 4: overly paranoid here.

111
00:06:33,600 --> 00:06:37,760
Speaker 2: Overly paranoid or not. Another question is whether state legislation

112
00:06:37,920 --> 00:06:41,600
Speaker 2: like this could actually prevent the worst from happening. I

113
00:06:41,640 --> 00:06:45,920
Speaker 2: asked Serene how exactly SB ten forty seven approaches AI

114
00:06:46,080 --> 00:06:46,800
Speaker 2: harm reduction.

115
00:06:47,800 --> 00:06:50,520
Speaker 4: So what it would actually do is mandate that companies

116
00:06:50,520 --> 00:06:52,960
Speaker 4: do a couple things. One is that they have a

117
00:06:53,080 --> 00:06:55,560
Speaker 4: so called kill switch that they have a way to

118
00:06:55,880 --> 00:06:58,440
Speaker 4: like turn off the machine if things go horribly wrong.

119
00:06:59,240 --> 00:07:01,880
Speaker 4: Another is that they would have to create what's called

120
00:07:02,120 --> 00:07:06,560
Speaker 4: SSPs Safety and Security Protocols. And what that would do

121
00:07:06,640 --> 00:07:10,000
Speaker 4: is essentially be like a document that outlines how they're

122
00:07:10,000 --> 00:07:15,320
Speaker 4: adhering to risk mitigation, which includes things like having external testing. Right,

123
00:07:15,360 --> 00:07:17,240
Speaker 4: so they say you have to get a third party

124
00:07:17,280 --> 00:07:19,880
Speaker 4: to go and test your systems and make sure that

125
00:07:19,920 --> 00:07:22,840
Speaker 4: they're not high risk. Also just through the threat of

126
00:07:22,880 --> 00:07:26,080
Speaker 4: being held legally liable, right, is kind of putting into

127
00:07:26,080 --> 00:07:29,400
Speaker 4: writing and spelling out exactly how the state of California

128
00:07:29,440 --> 00:07:32,440
Speaker 4: could come after you if you are not at least

129
00:07:32,480 --> 00:07:35,720
Speaker 4: following a basic set of principles around minimizing risk.

130
00:07:36,080 --> 00:07:39,440
Speaker 2: It's that threat of legal liability that spooked some tech

131
00:07:39,480 --> 00:07:43,760
Speaker 2: companies and help fuel a fierce lobbying effort against the legislation.

132
00:07:44,320 --> 00:07:47,040
Speaker 4: VC firms and tech companies have brought in the big guns.

133
00:07:47,080 --> 00:07:51,360
Speaker 4: They've hired top California lobbyists andres and Horowitz has hired

134
00:07:51,360 --> 00:07:55,200
Speaker 4: someone who has known Governor Newsom for a long time

135
00:07:55,280 --> 00:07:59,080
Speaker 4: and is a very well known and influential lobbyist in California.

136
00:07:59,200 --> 00:08:02,800
Speaker 4: Open aither companies have also hired lobbyists who are working

137
00:08:02,840 --> 00:08:03,680
Speaker 4: on this bill.

138
00:08:05,280 --> 00:08:09,480
Speaker 2: And already this tech industry pushback has led to tangible changes.

139
00:08:10,200 --> 00:08:12,880
Speaker 2: Take the way SB ten forty seven deals with open

140
00:08:12,920 --> 00:08:17,000
Speaker 2: source startups. For example, Sharin says there were concerns that

141
00:08:17,360 --> 00:08:20,280
Speaker 2: even though the bill only applies to these AI giants,

142
00:08:20,760 --> 00:08:24,560
Speaker 2: some of them, like Meta, make their AI models open source.

143
00:08:25,080 --> 00:08:27,720
Speaker 2: That makes it easier for smaller startups to use them

144
00:08:27,760 --> 00:08:28,280
Speaker 2: in their work.

145
00:08:28,720 --> 00:08:31,240
Speaker 4: And the worry is that this bill, if it does

146
00:08:31,280 --> 00:08:33,840
Speaker 4: apply to companies like Meta, whether it be a downstream

147
00:08:33,840 --> 00:08:35,800
Speaker 4: effect where Meta says we're not going to open source

148
00:08:35,840 --> 00:08:37,680
Speaker 4: our models anymore because we don't want to deal with

149
00:08:37,800 --> 00:08:42,839
Speaker 4: worrying about potential liability coming from other developers misusing our technology.

150
00:08:43,280 --> 00:08:46,280
Speaker 2: The latest version of the legislation tries to address those

151
00:08:46,320 --> 00:08:48,439
Speaker 2: potential downstream consequences.

152
00:08:48,480 --> 00:08:52,200
Speaker 3: Wiener told me, for example, making very clear that the

153
00:08:52,320 --> 00:08:56,000
Speaker 3: obligation to be able to shut down an AI model

154
00:08:56,800 --> 00:08:59,880
Speaker 3: only applies if you have that model in your possession.

155
00:09:00,200 --> 00:09:04,120
Speaker 3: So if you open source a model and someone significantly

156
00:09:04,480 --> 00:09:07,960
Speaker 3: changes the model, it's no longer your responsibility.

157
00:09:08,440 --> 00:09:12,520
Speaker 2: There's another overarching critique that tech companies like OpenAI have

158
00:09:12,640 --> 00:09:16,320
Speaker 2: made about SB ten forty seven, though, one that's harder

159
00:09:16,360 --> 00:09:18,800
Speaker 2: to solve with an amendment. It's the fact that the

160
00:09:18,880 --> 00:09:22,800
Speaker 2: legislation comes from the state of California, not Congress.

161
00:09:23,120 --> 00:09:24,600
Speaker 4: A lot of critics of the bill say we should

162
00:09:24,600 --> 00:09:27,400
Speaker 4: get federal legislation. They don't want this patchwork.

163
00:09:27,640 --> 00:09:32,120
Speaker 2: A patchwork meaning different AI regulations set state by state,

164
00:09:32,280 --> 00:09:37,160
Speaker 2: potentially overlapping or contradicting one another, instead of one national

165
00:09:37,200 --> 00:09:40,920
Speaker 2: framework all tech companies have to follow. But Senator Wiener

166
00:09:40,960 --> 00:09:44,600
Speaker 2: says part of the reason California is acting is because

167
00:09:44,760 --> 00:09:46,120
Speaker 2: Congress hasn't.

168
00:09:46,600 --> 00:09:50,199
Speaker 3: People would say, hey, this should be handled by Congress.

169
00:09:50,200 --> 00:09:52,640
Speaker 3: It should be handled at the federal level. It's better

170
00:09:52,679 --> 00:09:56,800
Speaker 3: to have a consistent national standard. And my answer is absolutely,

171
00:09:56,880 --> 00:09:59,000
Speaker 3: I would love to be able to close up shop

172
00:09:59,160 --> 00:10:03,160
Speaker 3: because Congress stepped in and passed a strong AI safety law.

173
00:10:03,240 --> 00:10:04,600
Speaker 3: That would be music to my ears.

174
00:10:06,080 --> 00:10:10,079
Speaker 2: Not all California politicians are on board with this trailblazing

175
00:10:10,120 --> 00:10:13,559
Speaker 2: policy agenda, including Senator Nancy Pelosi.

176
00:10:14,040 --> 00:10:18,440
Speaker 3: This bill is well intentioned but ill informed.

177
00:10:18,440 --> 00:10:21,120
Speaker 2: But prominent supporters have also emerged.

178
00:10:21,320 --> 00:10:24,040
Speaker 1: Elon Musk now coming out in support of a California

179
00:10:24,120 --> 00:10:27,480
Speaker 1: bill to regulate the development of AI, a stark contrast

180
00:10:27,559 --> 00:10:30,959
Speaker 1: to other big names including open Ai and Meta, both

181
00:10:30,960 --> 00:10:32,800
Speaker 1: of which have come out against the bill.

182
00:10:32,840 --> 00:10:36,440
Speaker 2: You're Sharen wasn't entirely surprised by Musk's take.

183
00:10:36,960 --> 00:10:39,240
Speaker 4: Musk has been someone who for a long time, this

184
00:10:39,320 --> 00:10:43,360
Speaker 4: has been his big fear. He has worried about the

185
00:10:43,360 --> 00:10:47,000
Speaker 4: long term, potentially catastrophic effects of AI. That's why he

186
00:10:47,000 --> 00:10:50,400
Speaker 4: helped fund open Ai to be a counter force to Google,

187
00:10:50,440 --> 00:10:52,839
Speaker 4: because he thought that AI is so important that more

188
00:10:52,880 --> 00:10:55,680
Speaker 4: people should be in control of it. Now, if you're cynical,

189
00:10:55,760 --> 00:10:57,679
Speaker 4: I think you could look at it and say, well,

190
00:10:57,720 --> 00:11:00,679
Speaker 4: Elon has more incentive to be support of this bill

191
00:11:00,720 --> 00:11:03,280
Speaker 4: because his AI company isn't number one Open AI or

192
00:11:03,320 --> 00:11:06,640
Speaker 4: anthropic or these other companies are perceived as being more ahead.

193
00:11:07,200 --> 00:11:10,160
Speaker 2: As for Wiener, he told me he'll take the support

194
00:11:10,160 --> 00:11:12,440
Speaker 2: where he can get it. Do you think his tweet

195
00:11:12,440 --> 00:11:12,800
Speaker 2: helps you?

196
00:11:13,400 --> 00:11:17,640
Speaker 3: I think that his tweet really shook things up because

197
00:11:17,679 --> 00:11:20,959
Speaker 3: there are you know, folks who I know, we're very

198
00:11:20,960 --> 00:11:24,199
Speaker 3: critical of the bill, and once they saw the tweet

199
00:11:24,240 --> 00:11:26,160
Speaker 3: and made them think about it more.

200
00:11:28,080 --> 00:11:31,680
Speaker 2: The drama within the AI community over this bill hasn't

201
00:11:31,679 --> 00:11:35,400
Speaker 2: stopped state legislators. They passed SB ten forty seven in

202
00:11:35,440 --> 00:11:38,760
Speaker 2: the last week in August. Now it's fate rests in

203
00:11:38,800 --> 00:11:42,600
Speaker 2: California Governor Gavin Newsom's hands, and if he signs it

204
00:11:42,640 --> 00:11:47,800
Speaker 2: into law, the reapercussions could be felt far beyond California.

205
00:11:48,360 --> 00:11:52,520
Speaker 2: Coming up will these AI regulations developed in the industry's

206
00:11:52,520 --> 00:12:04,600
Speaker 2: backyard have global reach. The Golden State is often ahead

207
00:12:04,600 --> 00:12:07,720
Speaker 2: of the curve when it comes to technology, both building

208
00:12:07,760 --> 00:12:12,960
Speaker 2: it and regulating it, and Scharene Gafari, a Bloomberg Tech reporter, says,

209
00:12:13,000 --> 00:12:17,000
Speaker 2: what happens in California very rarely stays in California.

210
00:12:17,520 --> 00:12:21,040
Speaker 4: It could very well be something that turns into a

211
00:12:21,080 --> 00:12:24,040
Speaker 4: template for other states. This has happened in the past

212
00:12:24,040 --> 00:12:27,200
Speaker 4: where California said some kind of tech legislation and then

213
00:12:27,280 --> 00:12:30,600
Speaker 4: other people followed through. We are already seeing other states

214
00:12:30,679 --> 00:12:34,199
Speaker 4: like Colorado looking at AI legislation, and I do think

215
00:12:34,240 --> 00:12:35,040
Speaker 4: we could see more.

216
00:12:35,320 --> 00:12:37,720
Speaker 2: And it's not just other states that could be taking

217
00:12:37,720 --> 00:12:39,400
Speaker 2: notes on what California is doing.

218
00:12:39,840 --> 00:12:41,680
Speaker 4: It could even at one day become a template for

219
00:12:41,720 --> 00:12:44,320
Speaker 4: federal or even international kind of regulations.

220
00:12:45,600 --> 00:12:49,440
Speaker 2: Sharene says SB ten forty seven already hewes closely to

221
00:12:49,559 --> 00:12:52,640
Speaker 2: international agreements that have come out of recent global safety

222
00:12:52,720 --> 00:12:56,040
Speaker 2: summits in the EU and in Korea. And though the

223
00:12:56,200 --> 00:13:00,640
Speaker 2: US has no national AI regulation in place, Open AI

224
00:13:00,679 --> 00:13:05,040
Speaker 2: and Anthropic recently voluntarily agreed to share their AI models

225
00:13:05,040 --> 00:13:08,920
Speaker 2: before they're released with the US Commerce Department's new AI

226
00:13:09,040 --> 00:13:10,240
Speaker 2: Safety Institute.

227
00:13:10,520 --> 00:13:12,880
Speaker 4: So proponents of the bill say, look, you're already doing

228
00:13:12,880 --> 00:13:14,560
Speaker 4: some of this work. Why are you so against it?

229
00:13:14,600 --> 00:13:16,720
Speaker 4: But critics of the bill say, well, these voluntary agreements

230
00:13:16,720 --> 00:13:19,480
Speaker 4: are working. We're already seeing companies voluntarily come on board,

231
00:13:19,640 --> 00:13:22,199
Speaker 4: So why do we need to have this hammer.

232
00:13:22,160 --> 00:13:26,320
Speaker 2: It's a common tension between businesses and policymakers. They both

233
00:13:26,320 --> 00:13:29,200
Speaker 2: think they should be the ones holding the hammer. But

234
00:13:29,280 --> 00:13:32,800
Speaker 2: for the AI industry, Sharin says, now is a particularly

235
00:13:32,920 --> 00:13:34,800
Speaker 2: tense time to be having this debate.

236
00:13:35,480 --> 00:13:38,360
Speaker 4: People think that there is a race to the finish

237
00:13:38,400 --> 00:13:41,640
Speaker 4: line to reach superintelligence right now. It's this exciting moment

238
00:13:41,640 --> 00:13:44,640
Speaker 4: where we're seeing this kind of Cambrian explosion and this

239
00:13:45,320 --> 00:13:50,080
Speaker 4: rapid advancement in AI, and they worry that this is

240
00:13:50,120 --> 00:13:53,079
Speaker 4: going to just stifle that. And I think that's kind

241
00:13:53,080 --> 00:13:55,760
Speaker 4: of a classic tech versus government fear.

242
00:13:57,040 --> 00:14:01,600
Speaker 2: To AI regulators slowing down that blowsive growth just long

243
00:14:01,720 --> 00:14:04,280
Speaker 2: enough to put up some guardrails might not be the

244
00:14:04,320 --> 00:14:07,640
Speaker 2: worst thing. And while Serene says we likely won't see

245
00:14:07,640 --> 00:14:09,960
Speaker 2: any changes to the AI tools we use in our

246
00:14:10,000 --> 00:14:13,640
Speaker 2: everyday lives just yet, the biggest change could be in

247
00:14:13,760 --> 00:14:17,000
Speaker 2: how developers go about rolling out new technology.

248
00:14:17,559 --> 00:14:20,400
Speaker 4: They may have to take more of a beat right

249
00:14:20,520 --> 00:14:24,600
Speaker 4: before putting out certain models and make sure that, Okay,

250
00:14:24,640 --> 00:14:28,440
Speaker 4: are we in compliance with these guidelines and if not,

251
00:14:29,480 --> 00:14:31,240
Speaker 4: are we taking the risk of being sued. So it

252
00:14:31,240 --> 00:14:35,160
Speaker 4: would just kind of create some more administrative overhead for

253
00:14:35,240 --> 00:14:37,960
Speaker 4: these companies to make sure that they are following these.

254
00:14:37,840 --> 00:14:43,360
Speaker 2: Rules, and for California State Senator Scott Wiener that administrative

255
00:14:43,360 --> 00:14:46,880
Speaker 2: overhead is a small price to pay to ensure future safety.

256
00:14:47,360 --> 00:14:51,560
Speaker 3: This is not about being an AI doomer. It's about

257
00:14:51,840 --> 00:14:55,120
Speaker 3: being a responsible grown up, where we say, let's do

258
00:14:55,240 --> 00:15:00,480
Speaker 3: all these good things and let's also protect against significant.

259
00:15:01,200 --> 00:15:03,760
Speaker 2: Whether you agree with him or not, Charine says that

260
00:15:03,800 --> 00:15:07,120
Speaker 2: the debate over the bill has already informed what questions

261
00:15:07,120 --> 00:15:09,880
Speaker 2: we all might need to be asking about AI development

262
00:15:10,480 --> 00:15:13,040
Speaker 2: and AI regulation in the coming years.

263
00:15:13,240 --> 00:15:15,680
Speaker 4: I think what this bill shows is that it's very

264
00:15:15,720 --> 00:15:17,880
Speaker 4: easy to say, oh, yeah, there could be some kind

265
00:15:17,920 --> 00:15:20,800
Speaker 4: of catastrophic risks from AI. We're all in agreement on that,

266
00:15:20,840 --> 00:15:23,320
Speaker 4: but it's very hard to actually agree. Okay, now, what

267
00:15:23,440 --> 00:15:25,240
Speaker 4: is the percentage chance you think that risk is going

268
00:15:25,280 --> 00:15:28,280
Speaker 4: to come, How urgent is it, What is the best

269
00:15:28,280 --> 00:15:31,520
Speaker 4: way to manage it? Once you get into the details,

270
00:15:31,520 --> 00:15:34,600
Speaker 4: all of a sudden, managing AI becomes really really complicated,

271
00:15:34,720 --> 00:15:39,520
Speaker 4: especially because it's so new, it's so technologically difficult. So

272
00:15:39,760 --> 00:15:42,720
Speaker 4: I think it's probably one of the hardest kinds of

273
00:15:42,760 --> 00:15:45,440
Speaker 4: technology in some ways to regulate.

274
00:15:51,280 --> 00:15:54,360
Speaker 2: This is the big take from Bloomberg News I'm Sarah Holder.

275
00:15:54,720 --> 00:15:57,400
Speaker 2: This episode was produced by Adrian AA Tapia, who also

276
00:15:57,520 --> 00:16:01,080
Speaker 2: fact checked this episode. It was mixed by Blake. Our

277
00:16:01,120 --> 00:16:04,360
Speaker 2: senior producers are Naomi Shaven and Kim Gittelson, who also

278
00:16:04,480 --> 00:16:08,560
Speaker 2: edited this episode. Our senior editor is Elizabeth Ponso. Nicole

279
00:16:08,560 --> 00:16:12,440
Speaker 2: Bumster Borr is our executive producer. Sage Bauman is Bloomberg's

280
00:16:12,440 --> 00:16:15,480
Speaker 2: head of podcasts. If you liked the episode, make sure

281
00:16:15,520 --> 00:16:18,040
Speaker 2: to subscribe and review The Big Take wherever you listen

282
00:16:18,080 --> 00:16:21,400
Speaker 2: to podcasts. It helps people find the show. Thank you

283
00:16:21,440 --> 00:16:27,640
Speaker 2: so much for listening. We'll be back on Monday.