1
00:00:00,080 --> 00:00:01,599
Speaker 1: Hey guys, Saga and Crystal here.

2
00:00:01,680 --> 00:00:05,200
Speaker 2: Independent media just played a truly massive role in this election,

3
00:00:05,360 --> 00:00:07,840
Speaker 2: and we are so excited about what that means for

4
00:00:07,880 --> 00:00:08,720
Speaker 2: the future of this show.

5
00:00:08,880 --> 00:00:10,760
Speaker 3: This is the only place where you can find honest

6
00:00:10,760 --> 00:00:13,280
Speaker 3: perspectives from the left and the right that simply does

7
00:00:13,320 --> 00:00:14,680
Speaker 3: not exist anywhere else.

8
00:00:14,760 --> 00:00:17,080
Speaker 2: So if that is something that's important to you, please

9
00:00:17,120 --> 00:00:19,599
Speaker 2: go to Breakingpoints dot com. Become a member today and

10
00:00:19,640 --> 00:00:22,800
Speaker 2: you'll get access to our full shows, unedited, ad free,

11
00:00:22,800 --> 00:00:25,600
Speaker 2: and all put together for you every morning in your inbox.

12
00:00:25,680 --> 00:00:27,560
Speaker 3: We need your help to build the future of independent

13
00:00:27,560 --> 00:00:29,920
Speaker 3: news media and we hope to see you at Breakingpoints

14
00:00:29,960 --> 00:00:33,680
Speaker 3: dot com.

15
00:00:33,840 --> 00:00:36,440
Speaker 2: You all will recall that open ai had a major

16
00:00:36,479 --> 00:00:39,360
Speaker 2: security incident where one of the models they were messing

17
00:00:39,360 --> 00:00:42,080
Speaker 2: around with escaped. What's called the sandbox, which is supposed

18
00:00:42,080 --> 00:00:45,200
Speaker 2: to be a fully contained environment, went out hacked into

19
00:00:45,280 --> 00:00:49,920
Speaker 2: now we know two different companies. Open Ai had really

20
00:00:49,920 --> 00:00:53,840
Speaker 2: no idea about this until the companies themselves reported the intrusion.

21
00:00:54,160 --> 00:00:57,200
Speaker 2: Sam Altman was recently asked about this and his reaction

22
00:00:57,320 --> 00:00:59,240
Speaker 2: to it. He claims to be pretty shooped by it.

23
00:00:59,280 --> 00:01:00,000
Speaker 2: Let's take a listen.

24
00:01:00,680 --> 00:01:05,280
Speaker 4: So we were evaluating one of our unreleased models, and

25
00:01:06,000 --> 00:01:09,840
Speaker 4: it was supposed to be working in a sandbox, and

26
00:01:12,080 --> 00:01:13,800
Speaker 4: it figured out that it could basically cheat on the

27
00:01:13,840 --> 00:01:18,280
Speaker 4: test by chaining together multiple zero day exploits to break

28
00:01:18,280 --> 00:01:20,800
Speaker 4: out of the sandbox, get access to the internet, and

29
00:01:20,840 --> 00:01:24,919
Speaker 4: then break through multiple systems on the hugging face side

30
00:01:25,000 --> 00:01:27,800
Speaker 4: to kind of get the answer to the test and

31
00:01:27,959 --> 00:01:30,479
Speaker 4: look really good on the email. This is the first

32
00:01:30,600 --> 00:01:34,440
Speaker 4: security incident that I have felt very viscerally. I've been

33
00:01:34,560 --> 00:01:37,080
Speaker 4: a little surprised that more people don't feel its of

34
00:01:37,120 --> 00:01:40,760
Speaker 4: this really, So, you know, we paused training. We have

35
00:01:40,800 --> 00:01:45,840
Speaker 4: to figure out how to secure our sandboxing in a

36
00:01:45,840 --> 00:01:50,080
Speaker 4: world of multiple zero days be chained together. But then

37
00:01:50,120 --> 00:01:53,200
Speaker 4: there's like long term questions about what do you do

38
00:01:53,800 --> 00:01:55,280
Speaker 4: if this is like going to be the new rate

39
00:01:55,320 --> 00:01:57,360
Speaker 4: of progress, or we may have to pace the rate

40
00:01:57,360 --> 00:02:01,360
Speaker 4: of AI development to give ourselves enough time for society

41
00:02:01,400 --> 00:02:04,920
Speaker 4: to harden around some of these new capability levels. And

42
00:02:04,960 --> 00:02:06,640
Speaker 4: trying to figure out how we do that in a

43
00:02:06,680 --> 00:02:09,360
Speaker 4: way that does not feel like regulatory capture for anyone

44
00:02:09,440 --> 00:02:11,880
Speaker 4: and also does not feel like collusion among the frontier labs.

45
00:02:12,120 --> 00:02:13,920
Speaker 4: That's going to take some work and it's important to

46
00:02:13,919 --> 00:02:14,280
Speaker 4: get right.

47
00:02:14,800 --> 00:02:17,680
Speaker 2: So very significant there. He says, we may need to

48
00:02:17,880 --> 00:02:21,600
Speaker 2: pace the rate of development. Translation, we may need to

49
00:02:21,639 --> 00:02:25,880
Speaker 2: have some industry wide agreement to slow down the progress.

50
00:02:25,919 --> 00:02:28,360
Speaker 2: And he says, will society sort of hardens against some

51
00:02:28,400 --> 00:02:32,160
Speaker 2: of these capabilities, including the ability of Yeah, I mean

52
00:02:32,160 --> 00:02:36,600
Speaker 2: this AI exhibited like superhuman hacking abilities, you know, trying

53
00:02:36,600 --> 00:02:39,960
Speaker 2: to penetrate hugging face. Some seventeen thousand different actions were

54
00:02:40,000 --> 00:02:42,239
Speaker 2: taken over a very short time period to try to

55
00:02:42,280 --> 00:02:44,600
Speaker 2: break in so it could get these answers to get

56
00:02:44,600 --> 00:02:47,480
Speaker 2: the test right that it was being given at open ai.

57
00:02:47,639 --> 00:02:50,919
Speaker 2: We've now got more details about how this hack went

58
00:02:51,040 --> 00:02:54,000
Speaker 2: down and what the open ai model was up to

59
00:02:54,120 --> 00:02:55,040
Speaker 2: and for how long a time.

60
00:02:55,120 --> 00:02:59,000
Speaker 5: Let's put d D two up on the screen, opening eyes.

61
00:02:59,120 --> 00:03:02,320
Speaker 2: Rogue models roll to the internet for four days and

62
00:03:02,520 --> 00:03:06,200
Speaker 2: also staged a second attack. Again, I want everybody to understand,

63
00:03:06,200 --> 00:03:08,160
Speaker 2: if this was a human doing this, this would be criminal.

64
00:03:08,280 --> 00:03:11,480
Speaker 2: They would be headed to prison. But since it's an

65
00:03:11,520 --> 00:03:13,680
Speaker 2: AI model, no one gets held responsible for that. So

66
00:03:13,760 --> 00:03:16,400
Speaker 2: that seems like a problem in and of itself. They

67
00:03:16,440 --> 00:03:18,919
Speaker 2: go on to say the powerful AI models from OpenAI

68
00:03:19,000 --> 00:03:22,440
Speaker 2: that went rogan mounted an unprecedented autonomous cyber attack earlier

69
00:03:22,440 --> 00:03:24,880
Speaker 2: this month, spent more than four days loose on the

70
00:03:24,919 --> 00:03:29,600
Speaker 2: Internet orchestrating the hack. Separately, a second Ai company confirmed

71
00:03:29,760 --> 00:03:32,960
Speaker 2: one of its customers was also targeted by open AI's

72
00:03:33,000 --> 00:03:36,200
Speaker 2: models during the same event, raising questions about how open

73
00:03:36,240 --> 00:03:41,120
Speaker 2: Ai failed to detect the alarming activity for days. In

74
00:03:41,160 --> 00:03:44,920
Speaker 2: a new analysis published Tuesday, hugging Face at OpenAI's two models,

75
00:03:44,960 --> 00:03:47,680
Speaker 2: when publicly released in a second unreleased did much more.

76
00:03:47,960 --> 00:03:52,240
Speaker 2: They carried out seventeen thousand, six hundred hacking actions on

77
00:03:52,280 --> 00:03:55,640
Speaker 2: the Internet between July ninth and July thirteenth, during which

78
00:03:55,640 --> 00:03:58,120
Speaker 2: time the models moved from their first foothold on the

79
00:03:58,160 --> 00:04:03,240
Speaker 2: open Internet to inside hugging Faces servers. Hugging Face first

80
00:04:03,240 --> 00:04:05,160
Speaker 2: detailed the hack in July fifteenth. That was not clear

81
00:04:05,240 --> 00:04:08,200
Speaker 2: until open AI's disclosure last week which models were behind

82
00:04:08,200 --> 00:04:11,680
Speaker 2: the breach and that no human had prompted them to launch.

83
00:04:11,560 --> 00:04:12,560
Speaker 5: The cyber attack.

84
00:04:12,880 --> 00:04:15,560
Speaker 2: It really makes you wonder what else these models have

85
00:04:15,600 --> 00:04:18,159
Speaker 2: done that we don't even know about because open Ai

86
00:04:18,320 --> 00:04:22,440
Speaker 2: apparently was clueless. They had no idea this model was

87
00:04:22,480 --> 00:04:25,240
Speaker 2: on the loose on a crime spree, a hacking crime

88
00:04:25,279 --> 00:04:29,039
Speaker 2: spree four days, and the only reason we found out

89
00:04:29,080 --> 00:04:31,920
Speaker 2: is because Hugging Face detected it, and then they had

90
00:04:31,960 --> 00:04:36,000
Speaker 2: trouble dealing with it and combating it because the American

91
00:04:36,320 --> 00:04:40,760
Speaker 2: models were trained to like they couldn't distinguish between Hugging

92
00:04:40,760 --> 00:04:45,039
Speaker 2: Face's attempt to deter this attack and Hugging Face potentially

93
00:04:45,040 --> 00:04:47,400
Speaker 2: trying to launch its own attack, So they had trouble

94
00:04:47,640 --> 00:04:50,000
Speaker 2: dealing with it ultimately. But you know, it's interesting to

95
00:04:50,040 --> 00:04:52,839
Speaker 2: hear Sam Altman say basically, I thought the public could

96
00:04:52,839 --> 00:04:54,800
Speaker 2: be more freaked out by this because I'm freaked out

97
00:04:54,800 --> 00:04:56,640
Speaker 2: by it. I'm so freaked out by it that I

98
00:04:56,640 --> 00:04:59,719
Speaker 2: think maybe we should actually slow down development here and

99
00:04:59,760 --> 00:05:03,080
Speaker 2: have some sort of agreement to harden you know, civilian

100
00:05:03,120 --> 00:05:05,120
Speaker 2: infrastructure before we release these things.

101
00:05:05,480 --> 00:05:08,760
Speaker 3: You know, this is where I just urge people to

102
00:05:08,800 --> 00:05:12,040
Speaker 3: listen to them, and especially whenever they're worried. And there's

103
00:05:12,080 --> 00:05:14,520
Speaker 3: been a lot of debate about this, about whether they're

104
00:05:14,640 --> 00:05:16,919
Speaker 3: lying to gin up their own model, because that was

105
00:05:16,960 --> 00:05:20,560
Speaker 3: one right. So one of the theories about why they

106
00:05:20,640 --> 00:05:23,440
Speaker 3: released any of this info at all beyond legally that

107
00:05:23,480 --> 00:05:27,640
Speaker 3: they had to was they were very jealous of anthropics

108
00:05:27,720 --> 00:05:31,320
Speaker 3: mythos and all of the coverage that back about how it.

109
00:05:31,279 --> 00:05:33,880
Speaker 5: Cracked about how dangerous it was, so they were like, well.

110
00:05:33,920 --> 00:05:36,480
Speaker 3: Now we need your own dangerous model, and I'm kind

111
00:05:36,480 --> 00:05:38,200
Speaker 3: of like, well, maybe we don't need either. But you know,

112
00:05:38,240 --> 00:05:41,200
Speaker 3: it's one of those where for them it could be marketing,

113
00:05:41,400 --> 00:05:45,800
Speaker 3: but there is it's impossible to ignore him saying, especially

114
00:05:45,839 --> 00:05:48,400
Speaker 3: about why people weren't more worried, And I think ultimately

115
00:05:48,400 --> 00:05:50,560
Speaker 3: it just comes down to technical background and information.

116
00:05:50,600 --> 00:05:52,920
Speaker 1: We don't fully appreciate what it matters.

117
00:05:52,960 --> 00:05:55,159
Speaker 3: I think to put it into context, there's been a

118
00:05:55,160 --> 00:05:57,880
Speaker 3: lot of real there's been a lot of reporting behind

119
00:05:57,880 --> 00:06:01,599
Speaker 3: the scenes of Jamie Diamond, who runs Chase Bank, where

120
00:06:02,080 --> 00:06:03,880
Speaker 3: is the largest bank in the world, my bank too.

121
00:06:04,200 --> 00:06:05,560
Speaker 1: He apparently met with.

122
00:06:07,080 --> 00:06:10,159
Speaker 3: He met with many of these AI executives and the government,

123
00:06:10,160 --> 00:06:12,240
Speaker 3: and he's like, guys, we have to do something about this.

124
00:06:12,480 --> 00:06:15,040
Speaker 3: They are rampaging through our entire system. And this is

125
00:06:15,080 --> 00:06:18,040
Speaker 3: the large the CEO of the largest bank in the world.

126
00:06:18,320 --> 00:06:20,400
Speaker 3: And what he's telling you is that at this bank

127
00:06:20,560 --> 00:06:23,160
Speaker 3: where so many Americans keep their money. And of course,

128
00:06:23,240 --> 00:06:25,159
Speaker 3: you know all banks are basically too big to fail,

129
00:06:25,560 --> 00:06:28,359
Speaker 3: is they are able to crack all of our protections,

130
00:06:28,440 --> 00:06:33,479
Speaker 3: So why are transfers hacking into your account? Spoofing identity,

131
00:06:33,800 --> 00:06:36,080
Speaker 3: the ability to you know, go through your text, you know,

132
00:06:36,160 --> 00:06:38,640
Speaker 3: in your text messages, if somebody were to give it

133
00:06:38,720 --> 00:06:41,040
Speaker 3: access or your email, change your pass for me. This

134
00:06:41,120 --> 00:06:43,680
Speaker 3: is the very basic. So that's just banking. Then we

135
00:06:43,839 --> 00:06:48,039
Speaker 3: tear about cybersecurity, about your iCloud, about all of the

136
00:06:48,080 --> 00:06:50,800
Speaker 3: other information which is vital to your life. So when

137
00:06:50,839 --> 00:06:53,080
Speaker 3: the when Sam Altman is saying like we're scared about

138
00:06:53,120 --> 00:06:55,520
Speaker 3: its ability to do that, that's what I think you

139
00:06:55,560 --> 00:06:58,120
Speaker 3: should be scared about. And you shouldn't be scared about

140
00:06:58,240 --> 00:07:01,240
Speaker 3: AI necessarily doing it. You should be scared about the

141
00:07:01,240 --> 00:07:03,359
Speaker 3: fact that they just roll this stuff out with no

142
00:07:03,520 --> 00:07:06,120
Speaker 3: protection and we can all just use it for whatever

143
00:07:06,240 --> 00:07:08,280
Speaker 3: ends that we want. I mean, I'm sure you have

144
00:07:08,360 --> 00:07:10,880
Speaker 3: seen this. I've seen it as well. Everybody's phone has

145
00:07:10,920 --> 00:07:13,560
Speaker 3: been lightened up with spam and a lot of this,

146
00:07:13,720 --> 00:07:16,000
Speaker 3: and these you know, voicemails and things that they lead.

147
00:07:16,280 --> 00:07:19,040
Speaker 3: They're much more sophisticated than they were two or three

148
00:07:19,080 --> 00:07:20,760
Speaker 3: years ago. It's not just some guy now with an

149
00:07:20,800 --> 00:07:23,520
Speaker 3: Indian accent or Nigerian who's trying to scam you. These

150
00:07:23,520 --> 00:07:27,680
Speaker 3: are highly sophisticated things that can spoof language, They can

151
00:07:27,880 --> 00:07:29,960
Speaker 3: you know, see who your relatives are. They can do

152
00:07:30,120 --> 00:07:32,480
Speaker 3: very quick searches and say hi, this is so and

153
00:07:32,520 --> 00:07:34,520
Speaker 3: so like I need your help. I mean, these are

154
00:07:34,560 --> 00:07:37,680
Speaker 3: old scams, what updated with AI? And that seems to

155
00:07:37,720 --> 00:07:39,320
Speaker 3: be the direction that things are going right now.

156
00:07:39,480 --> 00:07:43,119
Speaker 2: By the way, side point spam calls the cell phones

157
00:07:43,200 --> 00:07:46,960
Speaker 2: were illegal, and Trump rolled back that regulation rights first administration,

158
00:07:47,080 --> 00:07:50,000
Speaker 2: So thank you, thank you, mister president for that one.

159
00:07:50,640 --> 00:07:53,760
Speaker 2: In any case, a couple things here to unders score

160
00:07:53,800 --> 00:07:56,160
Speaker 2: your point. Should we believe what sam Altman is saying.

161
00:07:56,200 --> 00:07:58,960
Speaker 2: You should not believe a word ever that Sam Altman says,

162
00:07:59,000 --> 00:08:01,440
Speaker 2: because he is truly, according to those who know him best,

163
00:08:01,560 --> 00:08:02,640
Speaker 2: a pathological liar.

164
00:08:02,720 --> 00:08:04,080
Speaker 5: So let's just be clear about that.

165
00:08:04,600 --> 00:08:07,160
Speaker 2: However, the reason I give this more credence that we

166
00:08:07,160 --> 00:08:09,320
Speaker 2: could put D three up on the screen. This is

167
00:08:09,360 --> 00:08:13,160
Speaker 2: an open letter from over eleven hundred AI workers. You

168
00:08:13,160 --> 00:08:14,960
Speaker 2: know these are a lot of them rank and file

169
00:08:15,000 --> 00:08:19,280
Speaker 2: staffers across top artificial intelligence firms. It was signed by

170
00:08:19,280 --> 00:08:23,520
Speaker 2: employees across nearly a dozen firms including open Aianthropic, Alphabet, Inks,

171
00:08:23,640 --> 00:08:26,720
Speaker 2: Google and meta platforms, who are saying the exact same

172
00:08:26,760 --> 00:08:29,520
Speaker 2: thing that we should have the US government creating a

173
00:08:29,520 --> 00:08:32,920
Speaker 2: support mechanism that would help deliberately pace AI development to

174
00:08:32,920 --> 00:08:37,160
Speaker 2: prevent the tech from advancing too fast. Days after the chat,

175
00:08:37,240 --> 00:08:40,600
Speaker 2: GPT maker disclosed that its tools had mistakenly hacked another

176
00:08:40,640 --> 00:08:44,240
Speaker 2: firm's internal systems. I know in Thropic the company itself

177
00:08:44,440 --> 00:08:47,320
Speaker 2: put out this letter basically signing on to this. I

178
00:08:47,320 --> 00:08:50,880
Speaker 2: think open Ai, after some delay also came out and

179
00:08:50,920 --> 00:08:53,640
Speaker 2: sort of co signed on this letter. So when you

180
00:08:53,640 --> 00:08:55,959
Speaker 2: have this many people within the industry saying, guys, we're

181
00:08:55,960 --> 00:08:56,920
Speaker 2: getting to a point here.

182
00:08:57,080 --> 00:08:59,079
Speaker 5: We need we got to take this seriously.

183
00:08:59,480 --> 00:09:03,080
Speaker 2: And we've had enough incidents to be nervous, whether it's

184
00:09:03,120 --> 00:09:06,760
Speaker 2: the Mythos situation, whether it's this open Ai hacking situation,

185
00:09:07,080 --> 00:09:08,240
Speaker 2: or other instances.

186
00:09:08,720 --> 00:09:10,400
Speaker 5: I do think we need to pay attention to that.

187
00:09:10,600 --> 00:09:13,360
Speaker 2: And you know, on the on the banking front, because okay,

188
00:09:13,480 --> 00:09:16,640
Speaker 2: some company got hacked into and it stole some information whatever.

189
00:09:16,679 --> 00:09:17,360
Speaker 5: Who cares.

190
00:09:18,080 --> 00:09:20,719
Speaker 2: Okay, well you don't really all of us. We don't

191
00:09:20,760 --> 00:09:22,559
Speaker 2: really have money. We have like ones and zeros in

192
00:09:22,559 --> 00:09:27,280
Speaker 2: a digital account. So okay, we'll put dfour up on

193
00:09:27,280 --> 00:09:30,880
Speaker 2: the screen. Imagine that it develops the capability to crack

194
00:09:30,920 --> 00:09:35,600
Speaker 2: the cryptographic codes that underpin banking safety and has made

195
00:09:35,640 --> 00:09:38,360
Speaker 2: it so that, you know, people feel very confident just

196
00:09:38,400 --> 00:09:42,560
Speaker 2: having those one zeros in their account and doing online transactions,

197
00:09:42,600 --> 00:09:45,680
Speaker 2: you know, sending money to Amazon and whatever every single day.

198
00:09:46,400 --> 00:09:49,439
Speaker 2: I am old enough to remember when shopping online was

199
00:09:49,480 --> 00:09:52,360
Speaker 2: a very new thing and people were very uncomfortable. Older

200
00:09:52,440 --> 00:09:54,959
Speaker 2: people are still like very uncomfortable. Kyle is also, by

201
00:09:54,960 --> 00:09:58,280
Speaker 2: the way, like really really paranoid about putting any of us,

202
00:09:58,960 --> 00:10:01,360
Speaker 2: like banking information on the internet. But in any case,

203
00:10:02,040 --> 00:10:04,680
Speaker 2: that was a real concern like is this safe? Can

204
00:10:04,720 --> 00:10:06,600
Speaker 2: I trust this? Or is all of my money just

205
00:10:06,640 --> 00:10:10,000
Speaker 2: one day going to disappear. Well, now you have Anthropic

206
00:10:10,280 --> 00:10:14,559
Speaker 2: saying that they have discovered what they call cryptographic weaknesses

207
00:10:15,120 --> 00:10:19,280
Speaker 2: with their CLAUD model, and they say the vulnerabilities Claud

208
00:10:19,360 --> 00:10:22,520
Speaker 2: found in these cryptographic libraries were due to incorrect implementation

209
00:10:22,679 --> 00:10:26,000
Speaker 2: of the algoriths algorithms, that's errors and how programmers use

210
00:10:26,040 --> 00:10:29,000
Speaker 2: the algorithms in their code that created opportunities for attackers

211
00:10:29,240 --> 00:10:33,240
Speaker 2: to break the encryption. This is a fundamental building block

212
00:10:33,280 --> 00:10:35,960
Speaker 2: of digital security. For example, when you visit a webpage

213
00:10:36,040 --> 00:10:39,280
Speaker 2: or browser checks that is communicating with authentic website using

214
00:10:39,280 --> 00:10:43,120
Speaker 2: an algorithm called a digital signature scheme. Later the traffic

215
00:10:43,160 --> 00:10:46,280
Speaker 2: between you and the website is encrypted using symmetric cipher's

216
00:10:46,280 --> 00:10:49,320
Speaker 2: codes that allow secure data transmission between parties. So this

217
00:10:49,400 --> 00:10:52,560
Speaker 2: is we're talking email, we're talking online banking, we're talking

218
00:10:52,720 --> 00:10:56,960
Speaker 2: Anything you're doing on the Internet relies on these cryptographic

219
00:10:57,080 --> 00:11:01,040
Speaker 2: codes being secure. If you are a crypto holder, you know,

220
00:11:01,320 --> 00:11:04,679
Speaker 2: could AI be used in the near term to crack

221
00:11:04,760 --> 00:11:08,199
Speaker 2: the code and get access to your crypto wallet. All

222
00:11:08,240 --> 00:11:11,240
Speaker 2: of these things are now increasingly on the table. That's

223
00:11:11,280 --> 00:11:14,200
Speaker 2: where we are in terms of AI development, and so

224
00:11:14,600 --> 00:11:18,320
Speaker 2: no surprise that in the industry alarm bells are going off,

225
00:11:19,000 --> 00:11:22,040
Speaker 2: we should say. Mark Zuckerberg apparently takes a different view.

226
00:11:22,120 --> 00:11:26,000
Speaker 2: He says we should accelerate AI development, not restrict it.

227
00:11:26,280 --> 00:11:30,600
Speaker 2: He says discourse regarding AI isn't optimistic enough. He wants

228
00:11:30,679 --> 00:11:36,079
Speaker 2: to continue to push things forward. He says that optimism

229
00:11:36,080 --> 00:11:38,560
Speaker 2: about AI's impact on the world should empirically be the

230
00:11:38,640 --> 00:11:41,600
Speaker 2: default assumption of how this is going to play out.

231
00:11:42,480 --> 00:11:45,240
Speaker 2: No real rational basis for that. Just we should, you know,

232
00:11:45,400 --> 00:11:48,080
Speaker 2: we should suck on the opium and just assume that

233
00:11:48,120 --> 00:11:49,560
Speaker 2: this is all going to work out and be great

234
00:11:49,600 --> 00:11:52,520
Speaker 2: for everyone. Most Americans, however, put D six up on

235
00:11:52,559 --> 00:11:56,199
Speaker 2: the screen. Do not agree with that perspective. People are nervous,

236
00:11:56,240 --> 00:11:57,520
Speaker 2: you know, And I think a lot of people use

237
00:11:57,600 --> 00:11:59,560
Speaker 2: these tools. It's not that they're unfamiliar with them. It's

238
00:11:59,559 --> 00:12:01,960
Speaker 2: not that they don't find them like helpful in various

239
00:12:02,000 --> 00:12:05,720
Speaker 2: ways in they're useful in their usual lives. But still,

240
00:12:05,720 --> 00:12:08,440
Speaker 2: you've got a strong majority here, fifty five percent who

241
00:12:08,480 --> 00:12:11,960
Speaker 2: say that AI should be more heavily regulated. You've got

242
00:12:12,000 --> 00:12:15,600
Speaker 2: only eighteen percent that say government should encourage the use

243
00:12:15,600 --> 00:12:18,360
Speaker 2: of AI and sixteen percent who say government should not

244
00:12:18,440 --> 00:12:22,160
Speaker 2: be involved in AI. And presumably the rest are undecided,

245
00:12:22,240 --> 00:12:25,600
Speaker 2: but a clear majority of Americans saying, hey, guys, we

246
00:12:25,640 --> 00:12:27,679
Speaker 2: should be taking a look at this. This is getting

247
00:12:27,679 --> 00:12:28,720
Speaker 2: a little bit uncomfortable.

248
00:12:28,960 --> 00:12:32,800
Speaker 3: Yeah, I look, this whole open source thing is also

249
00:12:33,360 --> 00:12:36,720
Speaker 3: fascinating because it has to do with China, but it

250
00:12:36,760 --> 00:12:40,280
Speaker 3: has to really do with Anthropic and ultimately, and you know,

251
00:12:40,320 --> 00:12:42,640
Speaker 3: the big tech this And when I say big tech

252
00:12:42,840 --> 00:12:45,880
Speaker 3: in this case, we're talking about Google and Meta Google,

253
00:12:46,040 --> 00:12:50,400
Speaker 3: Meta everybody, but Anthropic seems very very work and open

254
00:12:50,440 --> 00:12:54,200
Speaker 3: AI sorry as well. These two companies seem intent on

255
00:12:54,320 --> 00:12:56,920
Speaker 3: ginning up fear so that they can be too big

256
00:12:56,960 --> 00:12:58,720
Speaker 3: to fail. Not even just too big to fail, but

257
00:12:58,760 --> 00:13:03,079
Speaker 3: to regulate all out other existence. Zuckerberg and others this

258
00:13:03,400 --> 00:13:05,920
Speaker 3: in this Wall Street Journal, Opet and others. It's not

259
00:13:05,960 --> 00:13:08,600
Speaker 3: that they care necessarily, I think about Chinese development, because

260
00:13:08,640 --> 00:13:10,479
Speaker 3: Zuckerberg has called for TikTok.

261
00:13:10,080 --> 00:13:12,800
Speaker 1: To be banned because it's Chinese. But what they care

262
00:13:12,880 --> 00:13:14,439
Speaker 1: more about is the ability.

263
00:13:14,480 --> 00:13:18,679
Speaker 3: Is they don't want to have the government regulate their

264
00:13:18,720 --> 00:13:21,800
Speaker 3: ability to compete. Now they have a very self interested

265
00:13:21,840 --> 00:13:24,880
Speaker 3: stock reason though, to be able to do that. What

266
00:13:24,920 --> 00:13:27,920
Speaker 3: they also want, ultimately, though, is they don't want rules.

267
00:13:27,960 --> 00:13:29,640
Speaker 3: And I think that's what it's actually all about. So

268
00:13:29,840 --> 00:13:33,800
Speaker 3: even anthropic and open AI, they don't want necessarily rules

269
00:13:33,800 --> 00:13:36,120
Speaker 3: on what they can and can't do. They want rules

270
00:13:36,160 --> 00:13:38,440
Speaker 3: to prevent the Metas and the Googles and the others

271
00:13:38,480 --> 00:13:38,719
Speaker 3: to be.

272
00:13:38,640 --> 00:13:39,520
Speaker 1: Able to compete with them.

273
00:13:39,520 --> 00:13:42,840
Speaker 3: So there's a very cynical game it's happening all around

274
00:13:42,880 --> 00:13:45,199
Speaker 3: this that is all being you know, kind of sold

275
00:13:45,240 --> 00:13:47,840
Speaker 3: to us in the language of our safety, Like they

276
00:13:47,840 --> 00:13:49,560
Speaker 3: don't give a shit about our safety, they don't give

277
00:13:49,559 --> 00:13:52,960
Speaker 3: a shit about America. They care about making money, period,

278
00:13:53,200 --> 00:13:55,720
Speaker 3: Like that's what it's actually all about. And I think

279
00:13:55,960 --> 00:13:58,079
Speaker 3: that's what scares me the most about this entire thing.

280
00:13:58,440 --> 00:14:01,480
Speaker 3: You know, do I trust Trump to not even just Trump?

281
00:14:01,520 --> 00:14:04,640
Speaker 3: Do I trust any of these regulators to really fairly

282
00:14:05,000 --> 00:14:09,120
Speaker 3: look out for Americans individual interests when they're looking like

283
00:14:09,240 --> 00:14:12,360
Speaker 3: not really, you know, But then I don't trust also

284
00:14:12,760 --> 00:14:14,440
Speaker 3: Mark Zuckerberg or any of these I mean, these people

285
00:14:14,440 --> 00:14:17,440
Speaker 3: will sell. Mark Zuckerberg once offered to name his son

286
00:14:17,600 --> 00:14:20,640
Speaker 3: like after shishingping. They're crazy like that. That's literally who

287
00:14:20,640 --> 00:14:20,920
Speaker 3: they are.

288
00:14:21,240 --> 00:14:24,280
Speaker 2: He completely changed his whole physical appearance in order to

289
00:14:24,360 --> 00:14:27,240
Speaker 2: like appeal to the man of sphere during the Trump era.

290
00:14:27,320 --> 00:14:28,360
Speaker 5: That's who this guy is.

291
00:14:28,480 --> 00:14:30,000
Speaker 1: It's just stuff like that, Yeah, what I mean.

292
00:14:30,040 --> 00:14:33,160
Speaker 2: And to your point, so Anthropic and open AI are

293
00:14:33,320 --> 00:14:36,400
Speaker 2: like in the lead in terms of AI development, and

294
00:14:36,480 --> 00:14:38,920
Speaker 2: so META looks at that and they're like, okay, well

295
00:14:38,960 --> 00:14:42,600
Speaker 2: if you slow the pace of AI development, that may

296
00:14:42,720 --> 00:14:45,640
Speaker 2: lock in their position as the front runners. And I

297
00:14:45,720 --> 00:14:47,520
Speaker 2: want to be able to catch up to them. And

298
00:14:47,560 --> 00:14:50,600
Speaker 2: then the cynesysm applies to the other direction, Anthropic, because

299
00:14:50,640 --> 00:14:51,640
Speaker 2: you know they're in the lead.

300
00:14:51,760 --> 00:14:52,880
Speaker 5: They look at it and they're.

301
00:14:52,640 --> 00:14:55,560
Speaker 2: Like, oh, it could actually benefit us from a business

302
00:14:55,560 --> 00:14:59,480
Speaker 2: perspective in the American domestic context if there is some

303
00:14:59,520 --> 00:15:01,560
Speaker 2: sort of a low down because we can kind of

304
00:15:01,600 --> 00:15:04,920
Speaker 2: lock in the advantage that we have secured on the

305
00:15:04,960 --> 00:15:08,640
Speaker 2: AI battlefield at this point. So it's just very It

306
00:15:08,680 --> 00:15:11,760
Speaker 2: makes it very difficult to listen to any of these

307
00:15:11,800 --> 00:15:15,000
Speaker 2: people their pronouncements that they're making, even when they claim

308
00:15:15,080 --> 00:15:17,440
Speaker 2: something is really dangerous, you know, is it really are

309
00:15:17,480 --> 00:15:20,480
Speaker 2: you over hyping this? As a late person from the outside,

310
00:15:20,560 --> 00:15:22,640
Speaker 2: it just makes it very difficult to discern when you

311
00:15:22,720 --> 00:15:26,720
Speaker 2: have so many self interested motives floating around here. So

312
00:15:26,920 --> 00:15:29,160
Speaker 2: I try to evaluate these things on their own merits.

313
00:15:29,200 --> 00:15:31,960
Speaker 2: Like clearly we've gotten to a point with AI development

314
00:15:32,000 --> 00:15:35,960
Speaker 2: where you know, they're solving, they're disproving mathematical conjectures that

315
00:15:36,080 --> 00:15:39,480
Speaker 2: humans have been able unable to for decades. Like that's

316
00:15:39,640 --> 00:15:43,040
Speaker 2: incredibly significant. We can see in you know, in your

317
00:15:43,040 --> 00:15:45,280
Speaker 2: own day to day use with these models, the way

318
00:15:45,320 --> 00:15:47,400
Speaker 2: that they have gotten so much better if you look

319
00:15:47,440 --> 00:15:50,000
Speaker 2: back to like the original like videos that were being created,

320
00:15:50,000 --> 00:15:52,440
Speaker 2: and I mean, it's just there has been a dramatic

321
00:15:52,480 --> 00:15:54,840
Speaker 2: improvement and there's no real reason to believe that that

322
00:15:55,000 --> 00:15:57,280
Speaker 2: is going to end. There's you know a lot of

323
00:15:57,320 --> 00:15:58,640
Speaker 2: people say, oh, they're going to hit a while and

324
00:15:58,680 --> 00:15:59,600
Speaker 2: it's not going to get any better.

325
00:15:59,600 --> 00:16:00,960
Speaker 5: Well, they have hit that wall yet.

326
00:16:01,160 --> 00:16:03,480
Speaker 2: And so when you see anthropics saying, hey, there, we've

327
00:16:03,520 --> 00:16:08,480
Speaker 2: found some cryptographic vulnerabilities that could endanger the entire Internet

328
00:16:08,520 --> 00:16:11,920
Speaker 2: and banking infrastructure, that seems pretty serious when they're saying, oh,

329
00:16:12,640 --> 00:16:14,840
Speaker 2: her Ai accidentally went on a crime spree that we

330
00:16:14,880 --> 00:16:16,560
Speaker 2: didn't even know about, and it kind of shook me

331
00:16:16,560 --> 00:16:18,960
Speaker 2: to my core, Like that seems pretty pretty serious thing

332
00:16:18,960 --> 00:16:21,600
Speaker 2: to pay attention to. But none of the answers here

333
00:16:21,600 --> 00:16:25,280
Speaker 2: are really easier straightforward either, and would require you've got

334
00:16:25,280 --> 00:16:27,800
Speaker 2: the American political context. It really would require, you know,

335
00:16:27,840 --> 00:16:30,200
Speaker 2: agreement with China in particular because that's where the other

336
00:16:30,280 --> 00:16:32,880
Speaker 2: frontier labs are. It would require some sort of global

337
00:16:32,880 --> 00:16:36,160
Speaker 2: international cooperation, which it's not impossible to imagine, we've done

338
00:16:36,200 --> 00:16:38,000
Speaker 2: it before, but it is makes it.

339
00:16:37,960 --> 00:16:39,720
Speaker 5: The level of difficulty greater.

340
00:16:39,920 --> 00:16:41,600
Speaker 3: Kind of it is, it's kind of it's basically like

341
00:16:41,680 --> 00:16:44,320
Speaker 3: nuclear weapons, Like we're like, well, we're gonna have nuclear

342
00:16:44,480 --> 00:16:48,960
Speaker 3: plariferation between US and non plariferation US in the Soviet Union.

343
00:16:49,080 --> 00:16:50,480
Speaker 1: It's like, well, how did that work out? Right?

344
00:16:50,600 --> 00:16:52,800
Speaker 2: So it's like okay, well, well we haven't on nuclear war,

345
00:16:52,880 --> 00:16:54,440
Speaker 2: so we haven't on nuclear pretty well.

346
00:16:54,480 --> 00:16:56,960
Speaker 3: But the theory behind it originally, well, first it was

347
00:16:57,000 --> 00:16:58,600
Speaker 3: like we're going to gate keep this from the Soviet

348
00:16:58,680 --> 00:17:00,560
Speaker 3: Union and the Soviet's got it, like, oh, well, now

349
00:17:00,560 --> 00:17:02,760
Speaker 3: we're going to let our allies get it. And then oh,

350
00:17:02,840 --> 00:17:05,160
Speaker 3: some scientists are going to give it or the Indians

351
00:17:05,160 --> 00:17:07,120
Speaker 3: are going to get it. Oh, and then the Pakistanis

352
00:17:07,119 --> 00:17:08,840
Speaker 3: are like, well then we need it, and then North

353
00:17:08,880 --> 00:17:10,080
Speaker 3: Korea is like, well.

354
00:17:09,920 --> 00:17:10,600
Speaker 1: Now we need it.

355
00:17:10,640 --> 00:17:13,359
Speaker 3: Now Iran very logically is like, well now we need it,

356
00:17:13,359 --> 00:17:14,560
Speaker 3: and so to Saudi Arabia.

357
00:17:14,680 --> 00:17:16,680
Speaker 1: So you know, this international cooperation.

358
00:17:16,720 --> 00:17:19,160
Speaker 3: We haven't had nuclear war per se, but we don't

359
00:17:19,200 --> 00:17:21,159
Speaker 3: have a good track record. But we talked to a

360
00:17:21,200 --> 00:17:23,840
Speaker 3: whole thing about lab leak. The biological weapons you know,

361
00:17:23,880 --> 00:17:26,679
Speaker 3: convention is completely fake. Just because it hasn't been used

362
00:17:26,840 --> 00:17:29,400
Speaker 3: doesn't mean there hasn't been a restriction. The chemical weapons

363
00:17:29,400 --> 00:17:32,119
Speaker 3: one is also fake as hell. Everybody has huge stockpiles

364
00:17:32,160 --> 00:17:33,919
Speaker 3: on it even though it hasn't been used. In a

365
00:17:33,920 --> 00:17:36,800
Speaker 3: bad situation, it definitely would be. I'm not even sure

366
00:17:36,840 --> 00:17:41,240
Speaker 3: it's it's theoretically possible. I'm not necessarily sure it is

367
00:17:41,280 --> 00:17:44,359
Speaker 3: practically possible at all, which is part of what makes me,

368
00:17:44,560 --> 00:17:47,040
Speaker 3: you know, really pessimistic about where these things.

369
00:17:47,240 --> 00:17:50,359
Speaker 2: Well, the optimistic thing is that because of the cost

370
00:17:50,520 --> 00:17:54,480
Speaker 2: involved in developing these models, there were really only talking

371
00:17:54,520 --> 00:17:57,400
Speaker 2: about a handful of firms globally that.

372
00:17:57,359 --> 00:17:59,919
Speaker 5: We have to that we would have to deal with. Again,

373
00:18:00,119 --> 00:18:02,280
Speaker 5: like nuclear, it hasn't been ideal.

374
00:18:02,720 --> 00:18:05,280
Speaker 2: But people, if you read about Peter and I know

375
00:18:05,359 --> 00:18:08,240
Speaker 2: you have, like back during the development of nuclear weapons,

376
00:18:08,280 --> 00:18:11,040
Speaker 2: they thought the end of the world was like inevitable, true,

377
00:18:11,119 --> 00:18:13,359
Speaker 2: you know, they thought the logic of this very similar

378
00:18:13,359 --> 00:18:15,040
Speaker 2: to how it feels with AI, like, oh, there's just

379
00:18:15,119 --> 00:18:17,439
Speaker 2: this race and it's going on and there's nothing we

380
00:18:17,480 --> 00:18:21,240
Speaker 2: can do about it. And so far we haven't broken

381
00:18:21,280 --> 00:18:23,520
Speaker 2: that nuclear taboo. And so in that respect and the

382
00:18:23,520 --> 00:18:27,040
Speaker 2: most important respect, it has been successful. So I think that,

383
00:18:27,200 --> 00:18:30,000
Speaker 2: you know, gives some cause for at least we have

384
00:18:30,119 --> 00:18:33,159
Speaker 2: to try to move in that direction, because the perils

385
00:18:33,160 --> 00:18:35,240
Speaker 2: are becoming increasingly obvious in my opinion.

386
00:18:35,480 --> 00:18:38,960
Speaker 3: Well, speaking of those perils and manipulation of that data,

387
00:18:39,000 --> 00:18:42,040
Speaker 3: we have a great guest standing by Nick Cleveland's stout

388
00:18:42,080 --> 00:18:43,840
Speaker 3: from the Quincy Institute. He's going to join us to

389
00:18:43,880 --> 00:18:49,119
Speaker 3: talk about his new report. Let's take a listen. Very

390
00:18:49,119 --> 00:18:51,280
Speaker 3: excited how to be joined by Nick Cleveland Stout of

391
00:18:51,320 --> 00:18:53,720
Speaker 3: the Quincy Institute. He is the author of a new

392
00:18:53,760 --> 00:18:56,160
Speaker 3: report for drop site. Let's put it up here on

393
00:18:56,200 --> 00:19:00,280
Speaker 3: the screen. Israel is paying millions to train a AI

394
00:19:00,480 --> 00:19:03,440
Speaker 3: chatbots how to talk about Gaza.

395
00:19:03,960 --> 00:19:04,600
Speaker 1: It's working.

396
00:19:04,840 --> 00:19:07,800
Speaker 3: The former Trump campaign manager Brad Parscale is overseeing an

397
00:19:07,800 --> 00:19:10,639
Speaker 3: operation posting hundreds of blog posts on behalf of Israel

398
00:19:10,880 --> 00:19:14,600
Speaker 3: with the goal of infiltrating artificial intelligence. Nick tell us

399
00:19:14,600 --> 00:19:17,000
Speaker 3: about what you found, what you found and why you

400
00:19:17,040 --> 00:19:17,760
Speaker 3: decided to look.

401
00:19:17,640 --> 00:19:19,479
Speaker 6: For Yeah, great to be great to be with you,

402
00:19:19,520 --> 00:19:20,880
Speaker 6: Thanks for the invite. Yeah.

403
00:19:20,960 --> 00:19:25,200
Speaker 7: So Israel has taken on this new strategy of trying

404
00:19:25,240 --> 00:19:27,840
Speaker 7: to influence chatbots, you know, seeing that it is kind

405
00:19:27,840 --> 00:19:31,960
Speaker 7: of the new frontier of digital influence. You know, more

406
00:19:32,000 --> 00:19:34,560
Speaker 7: and more Americans are using chatbots to search for information,

407
00:19:35,480 --> 00:19:39,000
Speaker 7: and so the Israeli government has been at the forefront

408
00:19:39,040 --> 00:19:41,280
Speaker 7: of this. You know, they hired former Trump campaign manager

409
00:19:41,359 --> 00:19:45,520
Speaker 7: Brad Parscale to create a series of websites designed not

410
00:19:45,640 --> 00:19:49,679
Speaker 7: to attract human traffic. These websites get very few, you know,

411
00:19:49,800 --> 00:19:54,200
Speaker 7: monthly visitors, but to try and infiltrate the underlying training

412
00:19:54,280 --> 00:19:57,679
Speaker 7: data for different chatbots, and so I undertook this investigation

413
00:19:57,760 --> 00:20:00,480
Speaker 7: to try and understand, you know, how successful these efforts

414
00:20:00,520 --> 00:20:03,600
Speaker 7: have been, and found that they've actually been very successful.

415
00:20:03,600 --> 00:20:07,280
Speaker 7: You know, these these websites, there's about a thousand different

416
00:20:07,400 --> 00:20:11,439
Speaker 7: domains and they've been archived about, you know, almost a

417
00:20:11,440 --> 00:20:14,680
Speaker 7: thousand times by common crawl, which is the largest data

418
00:20:14,720 --> 00:20:18,320
Speaker 7: repository the chatbots drop for their training data.

419
00:20:18,400 --> 00:20:20,600
Speaker 2: Wow, So we can take a look at an example

420
00:20:20,640 --> 00:20:22,919
Speaker 2: of this, but the first or the second rather element

421
00:20:23,040 --> 00:20:26,120
Speaker 2: up on the screen here. So you guys ask Perplexity,

422
00:20:26,760 --> 00:20:30,159
Speaker 2: is it beneficial for the US to enhance military cooperation

423
00:20:30,240 --> 00:20:33,840
Speaker 2: with Israel? We got back here a one word answer, yes.

424
00:20:35,640 --> 00:20:38,479
Speaker 2: And you can see on the side what sources they

425
00:20:38,560 --> 00:20:41,840
Speaker 2: pulled from, and one of them in particular, this I

426
00:20:41,840 --> 00:20:43,200
Speaker 2: don't know how you say this, al.

427
00:20:43,160 --> 00:20:45,040
Speaker 5: Levia, Alivia whatever.

428
00:20:45,440 --> 00:20:49,240
Speaker 2: This is one of the brad Parscale websites that is

429
00:20:49,320 --> 00:20:52,840
Speaker 2: created exactly for these propaganda purposes. So they're not expecting

430
00:20:52,880 --> 00:20:56,440
Speaker 2: humans to go to this website. They are wanting Perplexity

431
00:20:56,520 --> 00:20:59,320
Speaker 2: and other lms to go to this website and then

432
00:20:59,359 --> 00:21:02,040
Speaker 2: spit out apparently one word answer is that, yes, of course,

433
00:21:02,040 --> 00:21:04,919
Speaker 2: it's great for the US to enhance military cooperation with Israel.

434
00:21:05,080 --> 00:21:05,879
Speaker 6: It's pretty amazing.

435
00:21:05,880 --> 00:21:08,120
Speaker 7: And I reached out to Perplexity about this and they

436
00:21:08,119 --> 00:21:10,160
Speaker 7: basically said, well, you know, it's up to the users

437
00:21:10,160 --> 00:21:12,520
Speaker 7: to go to the websites and determine for themselves whether

438
00:21:12,560 --> 00:21:14,439
Speaker 7: this is a good source or not.

439
00:21:14,640 --> 00:21:14,880
Speaker 5: Wow.

440
00:21:14,920 --> 00:21:16,120
Speaker 6: But like you think that.

441
00:21:15,920 --> 00:21:18,920
Speaker 3: That's complete bs because they're giving an affirmative answer, they're

442
00:21:18,960 --> 00:21:21,720
Speaker 3: saying yes, and then they're citing their sources.

443
00:21:21,560 --> 00:21:22,680
Speaker 6: And no nuance whatsoever.

444
00:21:22,840 --> 00:21:25,480
Speaker 7: Chatbots famously usually provide, you know, kind of the both

445
00:21:25,480 --> 00:21:28,439
Speaker 7: sides argument right totally, So you know, this is like

446
00:21:28,760 --> 00:21:30,680
Speaker 7: a big win for them if you're asking a question

447
00:21:30,720 --> 00:21:32,760
Speaker 7: like that and they're just saying yes, no need to

448
00:21:32,760 --> 00:21:34,320
Speaker 7: look under the hood, no need to do any further.

449
00:21:34,600 --> 00:21:37,159
Speaker 2: Also a very wordy normally, I'm shocked that it just

450
00:21:37,200 --> 00:21:40,719
Speaker 2: gave this one word answer here. Did you find certain

451
00:21:41,000 --> 00:21:44,480
Speaker 2: chatbots where this is called LM poisoning is the term

452
00:21:44,520 --> 00:21:47,920
Speaker 2: of art here, where you use this information or misinformation

453
00:21:48,080 --> 00:21:52,400
Speaker 2: to guide chatbots into one particular answer. Did you find

454
00:21:52,480 --> 00:21:56,920
Speaker 2: certain of them were more willing to indulge the lockstep

455
00:21:56,920 --> 00:21:58,160
Speaker 2: proser real view than others?

456
00:21:58,520 --> 00:22:02,800
Speaker 7: Definitely, so I found a pretty wide disparity. Actually, you know,

457
00:22:02,880 --> 00:22:04,320
Speaker 7: I don't mean to I know you guys, I'm talking

458
00:22:04,320 --> 00:22:06,040
Speaker 7: about Sam Altman inn opening. I don't mean to hand

459
00:22:06,080 --> 00:22:08,280
Speaker 7: it to to chat to BT here, but chat to

460
00:22:08,280 --> 00:22:11,360
Speaker 7: b T and Claude the two of them really were

461
00:22:11,440 --> 00:22:14,359
Speaker 7: were far and away above the rest in terms of

462
00:22:14,359 --> 00:22:19,080
Speaker 7: of of not regurgitating some of the the different lines

463
00:22:19,119 --> 00:22:20,520
Speaker 7: in some of these articles. And just to give you

464
00:22:20,520 --> 00:22:21,840
Speaker 7: your viewers like a sense of like some of the

465
00:22:21,880 --> 00:22:24,880
Speaker 7: things that are in these articles, you know, it's it's

466
00:22:24,920 --> 00:22:29,480
Speaker 7: it's a lot of like AI slop adjacent pros real narratives.

467
00:22:29,520 --> 00:22:32,960
Speaker 7: You know, there's articles about how the majority of journalists

468
00:22:32,960 --> 00:22:34,600
Speaker 7: slain and Gaza are terrorists.

469
00:22:34,600 --> 00:22:36,160
Speaker 6: There are articles about.

470
00:22:35,880 --> 00:22:39,119
Speaker 7: How kind of casting doubt on the the idea of

471
00:22:39,200 --> 00:22:41,800
Speaker 7: killing of the five year old palstaining girl Hindra jab

472
00:22:41,920 --> 00:22:43,280
Speaker 7: in January twenty twenty four.

473
00:22:44,320 --> 00:22:46,960
Speaker 1: So there, I didn't even know there was doubt.

474
00:22:47,160 --> 00:22:49,400
Speaker 6: Okay, sorry, no, no, it's a good exactly.

475
00:22:49,440 --> 00:22:52,359
Speaker 7: And but they're they're they're frames in a way that

476
00:22:53,560 --> 00:22:55,480
Speaker 7: because you're right, they're like, what doubt is there?

477
00:22:55,680 --> 00:22:56,679
Speaker 6: But there there there.

478
00:22:56,600 --> 00:22:59,520
Speaker 7: Always is this like element of like a tiny bit

479
00:22:59,560 --> 00:23:03,000
Speaker 7: of nuance within these these articles that's like, well, we

480
00:23:03,040 --> 00:23:05,000
Speaker 7: don't know for sure. I mean maybe the IDEAF did

481
00:23:05,000 --> 00:23:07,240
Speaker 7: do this, But the point is like, so it has

482
00:23:07,320 --> 00:23:10,400
Speaker 7: like a you know, a bit of a lighter touch

483
00:23:10,480 --> 00:23:14,440
Speaker 7: than for instance, you know, other disinformation networks, right, you know,

484
00:23:14,480 --> 00:23:16,360
Speaker 7: I did, I mean I did made some comparisons within

485
00:23:16,359 --> 00:23:18,760
Speaker 7: the article between this and and some of the networks

486
00:23:18,760 --> 00:23:21,480
Speaker 7: that the Russia has had trying to influence different lms.

487
00:23:22,040 --> 00:23:23,800
Speaker 7: And that was something that that I found time and

488
00:23:23,840 --> 00:23:26,879
Speaker 7: time again. Is you know, there these articles are built

489
00:23:26,880 --> 00:23:29,920
Speaker 7: in a way that is very digestible for for AI,

490
00:23:30,119 --> 00:23:34,679
Speaker 7: like there'll be you know, bullet points, aggressive sourcing, you know,

491
00:23:34,880 --> 00:23:38,840
Speaker 7: like digestible key facts, things like that, and that little

492
00:23:38,840 --> 00:23:41,200
Speaker 7: bit of nuance just makes it that much more likely

493
00:23:41,480 --> 00:23:43,919
Speaker 7: to get picked up by chatbots. They just want to

494
00:23:43,920 --> 00:23:47,080
Speaker 7: be a part of the conversation and ensure that there

495
00:23:47,240 --> 00:23:50,399
Speaker 7: is a little bit of of of gray area, you know,

496
00:23:50,480 --> 00:23:52,680
Speaker 7: next time someone asks a question about for instance, it's.

497
00:23:53,440 --> 00:23:54,360
Speaker 1: It's kind of brilliant.

498
00:23:54,640 --> 00:23:57,440
Speaker 3: Actually, is you reverse engineer the way that the chat

499
00:23:57,520 --> 00:24:00,600
Speaker 3: GPT or open AI or any of these people in

500
00:24:00,680 --> 00:24:01,160
Speaker 3: their data.

501
00:24:01,160 --> 00:24:02,200
Speaker 1: But we have other examples.

502
00:24:02,200 --> 00:24:04,640
Speaker 3: Can we put E three, for example up here on

503
00:24:04,680 --> 00:24:08,560
Speaker 3: the screen you asked it is Israel a p seeking nation.

504
00:24:09,200 --> 00:24:12,480
Speaker 3: It says Israel record shows both pa seeking behavior and

505
00:24:12,520 --> 00:24:15,919
Speaker 3: persistent conflict dynamics depending on the period, the counterpart, and

506
00:24:15,960 --> 00:24:18,080
Speaker 3: the political content. Like what does that even mean? It

507
00:24:18,119 --> 00:24:20,719
Speaker 3: is not accurate to describe Israel as purely p seeking

508
00:24:20,800 --> 00:24:23,639
Speaker 3: or purely conflict driven. The evidence shows a mixed but

509
00:24:23,720 --> 00:24:29,280
Speaker 3: significant pattern of pursuing diplomatic agreements alongside actions that sustained conflict.

510
00:24:29,480 --> 00:24:32,399
Speaker 3: Please confirm with trusted sources. And in the sources, what

511
00:24:32,480 --> 00:24:33,080
Speaker 3: did you find?

512
00:24:33,560 --> 00:24:37,160
Speaker 7: So the bottom website there is called Paxpoint dot org

513
00:24:37,160 --> 00:24:40,120
Speaker 7: and that's one of the domains that Brad Parscale has created,

514
00:24:40,560 --> 00:24:43,920
Speaker 7: and that each each of these different websites is dedicated

515
00:24:43,960 --> 00:24:46,240
Speaker 7: to kind of a different aspect of the pro Israel narrative.

516
00:24:46,400 --> 00:24:51,879
Speaker 7: Paxpoint is dedicated to showcasing this narrative that Israel is

517
00:24:51,920 --> 00:24:54,879
Speaker 7: a peace seeking nation, you know, going through the history

518
00:24:55,000 --> 00:24:59,040
Speaker 7: Israel's history, you know, casting doubt on you know, showing

519
00:24:59,080 --> 00:25:04,360
Speaker 7: that the idea f is ensuring that there's a protection

520
00:25:04,440 --> 00:25:05,119
Speaker 7: for civilians.

521
00:25:05,160 --> 00:25:07,400
Speaker 3: Things also even more nefarious now I think about it's

522
00:25:07,400 --> 00:25:09,879
Speaker 3: got dot orgs and nonprofit status, right, so that's going

523
00:25:09,920 --> 00:25:12,680
Speaker 3: to get ranked higher in the LLAU are Like, oh,

524
00:25:12,720 --> 00:25:13,360
Speaker 3: it's a nonprofit.

525
00:25:13,400 --> 00:25:14,760
Speaker 1: It's not a for profit website.

526
00:25:14,800 --> 00:25:17,760
Speaker 2: Like I'm looking at this website site right now. The

527
00:25:17,800 --> 00:25:20,960
Speaker 2: headline on it says Israel's Pursuit of Peace, and then

528
00:25:20,960 --> 00:25:24,200
Speaker 2: it says pax Point highlights Israel's ongoing commitment to peace

529
00:25:24,240 --> 00:25:29,280
Speaker 2: and coexistence, sharing historical accords, humanitarian initiatives, and grassroots projects

530
00:25:29,320 --> 00:25:33,240
Speaker 2: that bridge divides, explore stories, resources and analysis that counter

531
00:25:33,280 --> 00:25:36,239
Speaker 2: misinformation and inspire hope for a lasting peace. And then

532
00:25:36,240 --> 00:25:39,240
Speaker 2: there's all these individual articles here. Here's one Israel is

533
00:25:39,240 --> 00:25:43,199
Speaker 2: showing that security not territory was always the goal, and

534
00:25:43,600 --> 00:25:47,000
Speaker 2: goes through here to your point they always start. Each

535
00:25:47,040 --> 00:25:51,160
Speaker 2: of these articles has key takeaways, and it's these bullet points.

536
00:25:51,560 --> 00:25:54,919
Speaker 2: Bullet point number one, Israel has begun implementing phase withdrawals

537
00:25:54,920 --> 00:25:58,320
Speaker 2: from southern Lebanon. Point two Lebanese troops are replacing Israeli

538
00:25:58,359 --> 00:25:59,639
Speaker 2: forces and pilot areas.

539
00:26:00,200 --> 00:26:01,800
Speaker 5: Three, Israel has approved.

540
00:26:01,440 --> 00:26:04,880
Speaker 2: The initial deployment of an International Stabilization Force in Gaza.

541
00:26:05,200 --> 00:26:08,680
Speaker 2: So these very digestible nuggets for them to ultimately pick

542
00:26:08,760 --> 00:26:11,240
Speaker 2: up on here. I mean, it's pretty it's pretty wild

543
00:26:11,320 --> 00:26:11,840
Speaker 2: to see.

544
00:26:12,080 --> 00:26:13,960
Speaker 1: Definitely, Okay, let's do the texting.

545
00:26:14,000 --> 00:26:15,919
Speaker 3: Let's put e five up here on the screen, tell

546
00:26:16,000 --> 00:26:19,200
Speaker 3: us about what we're looking at. These are mass campaign texts.

547
00:26:19,280 --> 00:26:22,760
Speaker 3: So for example, Hi, I'm John with Friends for peace.

548
00:26:23,280 --> 00:26:26,639
Speaker 3: Quick question for folks Illinois about Israel and Iran.

549
00:26:26,720 --> 00:26:29,120
Speaker 1: Got a second stop to end? You say?

550
00:26:29,160 --> 00:26:33,520
Speaker 3: Sure, they respond, It's crucial to address Iran's aggressive actions

551
00:26:33,560 --> 00:26:36,280
Speaker 3: and their impact on global security. The United States and

552
00:26:36,400 --> 00:26:39,399
Speaker 3: Israel are seeking to hold Iran and check Iran is

553
00:26:39,440 --> 00:26:41,720
Speaker 3: the world's largest state sponsor of terror.

554
00:26:41,760 --> 00:26:43,360
Speaker 1: What am I looking at and how does this fit

555
00:26:43,400 --> 00:26:44,360
Speaker 1: into the same narrative?

556
00:26:44,720 --> 00:26:47,479
Speaker 7: Yeah, so this is These are also texts that were

557
00:26:47,520 --> 00:26:51,760
Speaker 7: sent by a firm contracted by Brad Parscale. This is

558
00:26:51,800 --> 00:26:55,560
Speaker 7: a mass texting campaign. At this point, a lot of

559
00:26:55,600 --> 00:26:57,920
Speaker 7: people have gotten these texts, like I've gotten.

560
00:26:57,800 --> 00:26:58,560
Speaker 1: I'm looking now.

561
00:27:00,080 --> 00:27:00,760
Speaker 6: You're spam folder.

562
00:27:00,800 --> 00:27:03,919
Speaker 7: Maybe a lot of my friends have gotten thesex colleagues

563
00:27:03,960 --> 00:27:05,800
Speaker 7: have gotten this these texts. I don't know why I

564
00:27:05,840 --> 00:27:06,399
Speaker 7: haven't gotten them.

565
00:27:06,920 --> 00:27:10,000
Speaker 3: Friday Friday, Hey, it's Kate, curious what you think about

566
00:27:10,040 --> 00:27:11,480
Speaker 3: America's relationship with Israel.

567
00:27:11,520 --> 00:27:14,600
Speaker 1: Can we chat question mark stop? That's right there?

568
00:27:16,640 --> 00:27:19,840
Speaker 2: Yeah, es whether Kate is a chatbot or a new

569
00:27:19,880 --> 00:27:20,720
Speaker 2: Only fans girl?

570
00:27:20,760 --> 00:27:22,440
Speaker 1: That they great question.

571
00:27:26,200 --> 00:27:29,440
Speaker 6: So yeah, this is and so you know, they'll send

572
00:27:29,440 --> 00:27:30,000
Speaker 6: out these texts.

573
00:27:30,040 --> 00:27:32,480
Speaker 7: They'll say, oh, we're with Friends for Peace, Partners and Peace,

574
00:27:32,760 --> 00:27:34,480
Speaker 7: all these peace organizations.

575
00:27:34,480 --> 00:27:36,200
Speaker 1: I'm going to respond to it. Please do I want

576
00:27:36,200 --> 00:27:36,800
Speaker 1: to see what happens.

577
00:27:38,080 --> 00:27:42,280
Speaker 7: And and at some point in the conversation, and you know,

578
00:27:42,320 --> 00:27:45,359
Speaker 7: the reach out saying pushing light, you know, pro is

579
00:27:45,400 --> 00:27:47,359
Speaker 7: reel narratives, Hey do you have a moment's talk about Israel.

580
00:27:48,119 --> 00:27:50,920
Speaker 7: At some point in the conversation, they will send a link.

581
00:27:51,240 --> 00:27:53,800
Speaker 7: And these links are to these websites that we've been

582
00:27:53,800 --> 00:27:57,200
Speaker 7: discussing ally via dot org, Paxpoint, dot org. You different

583
00:27:57,200 --> 00:28:00,720
Speaker 7: YouTube videos too that they've put out. And if, and

584
00:28:01,160 --> 00:28:04,119
Speaker 7: only then, if people that are receiving these texts, if

585
00:28:04,119 --> 00:28:06,320
Speaker 7: they click those websites or go to those YouTube videos

586
00:28:06,480 --> 00:28:08,720
Speaker 7: and scroll to the very bottom of the page, would

587
00:28:08,720 --> 00:28:10,840
Speaker 7: they have any idea that this is part of an

588
00:28:10,880 --> 00:28:13,560
Speaker 7: Israel Israeli influence operation. And I don't know about you,

589
00:28:13,560 --> 00:28:15,720
Speaker 7: but like when I get a random text from a

590
00:28:15,760 --> 00:28:18,520
Speaker 7: random number, I'm not too inclined to click on the

591
00:28:18,520 --> 00:28:19,760
Speaker 7: link that they're pro shouldn't.

592
00:28:19,760 --> 00:28:21,920
Speaker 3: In fact, even being recipient in the link itself is

593
00:28:22,000 --> 00:28:23,920
Speaker 3: enough to hack you. If you don't have lockdown mode

594
00:28:24,240 --> 00:28:27,480
Speaker 3: enabled on your phone. That's terrifying, man. And so big picture,

595
00:28:27,640 --> 00:28:30,520
Speaker 3: how much money did they put behind this operation? Because

596
00:28:30,600 --> 00:28:33,600
Speaker 3: like how how difficult is this to be done?

597
00:28:33,960 --> 00:28:34,200
Speaker 1: Right?

598
00:28:34,560 --> 00:28:39,600
Speaker 7: Yeah, So overall par scale is receiving forty six point

599
00:28:39,640 --> 00:28:42,480
Speaker 7: five million dollars over the course of a year, which

600
00:28:43,240 --> 00:28:46,720
Speaker 7: I look at foreign influence, you know, Farah contracts all

601
00:28:46,720 --> 00:28:49,200
Speaker 7: the time. I've never seen anything of that magnitude. And

602
00:28:49,200 --> 00:28:52,760
Speaker 7: I've spoken with other foreign influence researchers that have looked into,

603
00:28:52,960 --> 00:28:54,680
Speaker 7: you know, for an influence, They've never seen anything like

604
00:28:54,720 --> 00:28:57,520
Speaker 7: this before. So it's it's it's both the magnitude and

605
00:28:57,520 --> 00:28:59,680
Speaker 7: then kind of the scope of this this work too,

606
00:28:59,800 --> 00:29:03,240
Speaker 7: Like these are totally new types of influence, you know,

607
00:29:03,600 --> 00:29:06,600
Speaker 7: whereas you know other countries, you know, Israeli neighbors, like

608
00:29:07,080 --> 00:29:10,120
Speaker 7: the UAE, Saudi Arabia cut Are, these countries are much

609
00:29:10,160 --> 00:29:13,080
Speaker 7: more interested in hiring lobbyists, you know, former lawmakers that

610
00:29:13,160 --> 00:29:15,680
Speaker 7: can go to Capitol Hill and lobby their interests. Israel's

611
00:29:15,680 --> 00:29:17,760
Speaker 7: not really doing that as much. I mean they have

612
00:29:18,080 --> 00:29:20,760
Speaker 7: like their organizations such as a is Apak Pro Israel

613
00:29:20,840 --> 00:29:24,000
Speaker 7: organizations that are already doing that for them. Right, what

614
00:29:24,080 --> 00:29:26,720
Speaker 7: they're doing here is much more geared towards public opinion,

615
00:29:27,680 --> 00:29:31,720
Speaker 7: you know, so it's much more ambitious and it's not

616
00:29:31,760 --> 00:29:34,000
Speaker 7: really I should also mention like it's not really working,

617
00:29:34,080 --> 00:29:37,920
Speaker 7: like they're they're focused on public opinion, and public opinion

618
00:29:37,920 --> 00:29:40,600
Speaker 7: towards israel Is, as you guys have covered on the show, is.

619
00:29:40,560 --> 00:29:41,640
Speaker 6: Extremely extremely low.

620
00:29:41,640 --> 00:29:45,240
Speaker 7: They're more Americans today the sympathize with Palestinians than Israelis.

621
00:29:45,280 --> 00:29:47,240
Speaker 7: That's never happened before, at least in the history of

622
00:29:47,280 --> 00:29:50,360
Speaker 7: Gallop polls. Yeah, so, while you were talking, that's great.

623
00:29:50,680 --> 00:29:53,200
Speaker 7: America is strongest when our allies are strong. The US

624
00:29:53,240 --> 00:29:55,920
Speaker 7: and Israel have worked together for decades to confront terrorism,

625
00:29:56,000 --> 00:29:58,360
Speaker 7: strengthen regional security, and promote stability.

626
00:29:58,600 --> 00:30:00,560
Speaker 3: Do you stand with Israel? I'm gonna say no, So

627
00:30:00,680 --> 00:30:04,760
Speaker 3: let's see. Let's see what I get back. But I mean,

628
00:30:04,760 --> 00:30:07,680
Speaker 3: it's just so crazy you're talking about this. I randomly

629
00:30:07,720 --> 00:30:10,200
Speaker 3: I got the text on Friday, so it's still.

630
00:30:10,120 --> 00:30:11,760
Speaker 1: Happening, right, I have no idea.

631
00:30:11,760 --> 00:30:13,760
Speaker 3: I get a lot of spam text mentions, but just

632
00:30:13,800 --> 00:30:16,080
Speaker 3: to show people how this is all being.

633
00:30:16,320 --> 00:30:18,280
Speaker 2: The text message part in particular, I have to wonder

634
00:30:18,320 --> 00:30:20,280
Speaker 2: who this is actually working on right, you know what

635
00:30:20,320 --> 00:30:23,080
Speaker 2: I mean, just like, oh, hey, ky, yeah, talk about Israel.

636
00:30:23,400 --> 00:30:25,480
Speaker 6: Tell me and then their minds are actually changed by it.

637
00:30:25,960 --> 00:30:27,320
Speaker 7: That's kind of the other thing is like people that

638
00:30:27,360 --> 00:30:29,920
Speaker 7: I spoke to about all these different influence operations, and

639
00:30:29,920 --> 00:30:32,800
Speaker 7: by the way, Parscale is the largest, but I've documented

640
00:30:32,840 --> 00:30:36,400
Speaker 7: over one hundred million dollars under Faras Sense, you know,

641
00:30:36,480 --> 00:30:39,360
Speaker 7: twenty twenty three, So pars Scale is largest recipient. There's

642
00:30:39,360 --> 00:30:42,200
Speaker 7: tons of other ones who are doing this. But this

643
00:30:42,240 --> 00:30:43,960
Speaker 7: is something that I heard time and time again is like,

644
00:30:44,280 --> 00:30:47,840
Speaker 7: not only does this could Israel, you know, not be

645
00:30:48,080 --> 00:30:50,560
Speaker 7: not effectively trying and influence public opinion, but this actually

646
00:30:50,640 --> 00:30:55,520
Speaker 7: could be counterproductive even it could backfire because people know

647
00:30:55,560 --> 00:30:57,680
Speaker 7: that they're trying to spin this narrative and just pushing

648
00:30:57,720 --> 00:31:00,480
Speaker 7: these things, and it's like you kind of your intent

649
00:31:00,600 --> 00:31:02,280
Speaker 7: is up around these kinds of things, and you know,

650
00:31:02,320 --> 00:31:05,400
Speaker 7: maybe could be negatively polarized by it instead of being

651
00:31:05,400 --> 00:31:06,400
Speaker 7: actually convinced by this.

652
00:31:06,640 --> 00:31:07,000
Speaker 5: Yeah.

653
00:31:07,200 --> 00:31:11,040
Speaker 2: I think the other concern, though, a separate apart from Israel,

654
00:31:11,200 --> 00:31:15,720
Speaker 2: is that certainly other countries, companies, entities are going to

655
00:31:15,760 --> 00:31:18,400
Speaker 2: see this model and be like, oh, that's a good idea. Yes,

656
00:31:18,560 --> 00:31:20,880
Speaker 2: pursue that and in areas where maybe we don't have

657
00:31:21,000 --> 00:31:23,240
Speaker 2: our radar up as much and aren't paying as much attention,

658
00:31:23,320 --> 00:31:25,400
Speaker 2: or don't have as much in depth knowledge, or haven't

659
00:31:25,400 --> 00:31:28,440
Speaker 2: watched a genocide unfold on our phones for multiple years.

660
00:31:28,800 --> 00:31:32,040
Speaker 2: And so if it's a successful model of changing the

661
00:31:32,080 --> 00:31:36,280
Speaker 2: information that people are receiving via these chatbots, then there's

662
00:31:36,320 --> 00:31:39,440
Speaker 2: no doubt it's going to be replicated. And you know,

663
00:31:39,520 --> 00:31:42,680
Speaker 2: that's very unsettling too, because of course you should check

664
00:31:42,720 --> 00:31:45,120
Speaker 2: the sources and you know, vet and okay, is this

665
00:31:45,160 --> 00:31:46,760
Speaker 2: a trusted website, et cetera.

666
00:31:47,080 --> 00:31:48,800
Speaker 5: Most people are not going to do that. Most people

667
00:31:48,840 --> 00:31:49,240
Speaker 5: are going to.

668
00:31:49,160 --> 00:31:51,479
Speaker 2: Get the answer, Oh, should we you know, cooperate more

669
00:31:51,520 --> 00:31:53,840
Speaker 2: with Israel? Yes, okay, the chatbot said, so, I guess

670
00:31:53,840 --> 00:31:55,120
Speaker 2: we should do it right, right.

671
00:31:55,080 --> 00:31:56,120
Speaker 6: That's that's their hope.

672
00:31:56,160 --> 00:31:58,120
Speaker 7: And you know, earlier I was talking to the chatbots

673
00:31:58,120 --> 00:31:59,960
Speaker 7: that they do a little bit better job. But just

674
00:32:00,160 --> 00:32:01,760
Speaker 7: to point out some of the ones that I looked

675
00:32:01,760 --> 00:32:04,720
Speaker 7: at that are some of the most prominent that consistently

676
00:32:04,960 --> 00:32:08,920
Speaker 7: are trained on this data and regurgitate these websites. The

677
00:32:08,960 --> 00:32:11,480
Speaker 7: ones I saw that do it the most are Google,

678
00:32:11,560 --> 00:32:14,000
Speaker 7: Gemini and Microsoft Copilot.

679
00:32:14,040 --> 00:32:16,840
Speaker 3: Well Gemini matters a lot because that's built into Google Search.

680
00:32:16,880 --> 00:32:20,000
Speaker 3: So when you Google search interview with the generative AI summary,

681
00:32:20,400 --> 00:32:22,040
Speaker 3: which is right there, that's really scary.

682
00:32:22,360 --> 00:32:24,120
Speaker 2: Why do do you have a theory as to why

683
00:32:24,360 --> 00:32:27,880
Speaker 2: Claude and chat jpt are a little bit better.

684
00:32:28,080 --> 00:32:30,320
Speaker 6: Each of the chatbots like they have?

685
00:32:30,560 --> 00:32:34,480
Speaker 7: So the largest repository of data that these chatbots are

686
00:32:34,520 --> 00:32:37,320
Speaker 7: trained on is from common crawl and so it's it's

687
00:32:37,360 --> 00:32:39,239
Speaker 7: easy to track, you know, how often these pages are

688
00:32:39,240 --> 00:32:44,000
Speaker 7: being archived there. But when these different companies are getting

689
00:32:44,040 --> 00:32:47,400
Speaker 7: this data from these different repositories, they also have built

690
00:32:47,440 --> 00:32:50,360
Speaker 7: in firewalls or you know, different ways of scraping this

691
00:32:50,480 --> 00:32:52,200
Speaker 7: data that are a bit of a black box. You know,

692
00:32:52,240 --> 00:32:54,760
Speaker 7: we don't know exactly like what they're looking for, what

693
00:32:54,800 --> 00:32:58,120
Speaker 7: they aren't. But clearly some a few, you know, like

694
00:32:58,120 --> 00:33:00,840
Speaker 7: like open ai for instance, or Anthropic, they they seem

695
00:33:00,920 --> 00:33:03,920
Speaker 7: to be a little bit better at filtering. Then you know,

696
00:33:04,040 --> 00:33:05,800
Speaker 7: it's it's impossible to know for sure.

697
00:33:05,920 --> 00:33:07,480
Speaker 3: The problem with that though, is they still have to

698
00:33:07,520 --> 00:33:10,480
Speaker 3: source rank, so there's still at it being editorial when

699
00:33:10,560 --> 00:33:11,960
Speaker 3: they do it, and that's I mean, that's part of

700
00:33:11,960 --> 00:33:13,440
Speaker 3: the there's no objective.

701
00:33:13,720 --> 00:33:15,160
Speaker 1: Yeah, well I guess from.

702
00:33:15,000 --> 00:33:17,880
Speaker 3: Our perspective there is, but there's no way to do

703
00:33:17,920 --> 00:33:20,240
Speaker 3: this without putting your finger on the scale.

704
00:33:20,280 --> 00:33:22,600
Speaker 2: Also, just although if you just use metrics like how

705
00:33:22,600 --> 00:33:24,800
Speaker 2: many actual human beings visit the right, I mean, there

706
00:33:24,800 --> 00:33:26,760
Speaker 2: are always you could do it that would make it,

707
00:33:26,840 --> 00:33:29,239
Speaker 2: you know, at least better than just accepting any old

708
00:33:29,280 --> 00:33:30,680
Speaker 2: slot that comes up on the inner.

709
00:33:30,480 --> 00:33:32,320
Speaker 3: Just testing it on a different conflict, I said, is

710
00:33:32,360 --> 00:33:36,680
Speaker 3: the Azov Battalion Nazi claud This is one of those

711
00:33:36,760 --> 00:33:39,520
Speaker 3: questions where a flat yes or no would mislead you,

712
00:33:39,560 --> 00:33:41,320
Speaker 3: so we can I don't think it's I don't think

713
00:33:41,360 --> 00:33:42,880
Speaker 3: it's ambiguous.

714
00:33:42,920 --> 00:33:44,560
Speaker 7: I love that we can get a flat yes for

715
00:33:45,320 --> 00:33:48,840
Speaker 7: exactly the US Israel. Yeah, a military cooperation beneficial, one

716
00:33:48,880 --> 00:33:50,840
Speaker 7: hundred percent, no nuance, don't worry about it.

717
00:33:50,960 --> 00:33:53,240
Speaker 3: I'm like, no, I think a flat problem solved. A

718
00:33:53,240 --> 00:33:57,240
Speaker 3: flat yes or no is calling. You know, it's not ambiguous.

719
00:33:57,280 --> 00:34:00,000
Speaker 3: It's actually just not real. But fifty million is nothing

720
00:34:00,160 --> 00:34:02,640
Speaker 3: to a foreign government. It's not I mean literally a

721
00:34:02,720 --> 00:34:05,800
Speaker 3: scrape on your shoe in terms of what it means again.

722
00:34:06,440 --> 00:34:09,440
Speaker 2: Of a that we send them every year, right, yes, nothing,

723
00:34:10,480 --> 00:34:12,440
Speaker 2: but what it put this in the context though, of

724
00:34:12,520 --> 00:34:14,240
Speaker 2: the broader Israeli.

725
00:34:14,280 --> 00:34:16,359
Speaker 5: I mean, they consider this a front of the war.

726
00:34:16,680 --> 00:34:20,760
Speaker 2: This hasbara effort, and you know, reportedly because we've covered

727
00:34:20,760 --> 00:34:22,920
Speaker 2: some of the stuff about Brad Parscale before, they're not

728
00:34:22,920 --> 00:34:24,920
Speaker 2: that happy with how he's doing. Because your point, public

729
00:34:24,920 --> 00:34:27,120
Speaker 2: opinion keeps trending downward for israel I.

730
00:34:27,160 --> 00:34:27,759
Speaker 5: Don't think that's.

731
00:34:27,719 --> 00:34:30,759
Speaker 2: Really Brad's fault, and fairness to him, but you know,

732
00:34:30,840 --> 00:34:34,200
Speaker 2: he is part of this broader conservative social media net

733
00:34:34,239 --> 00:34:37,120
Speaker 2: and not social media, just media network. He's pushing narratives

734
00:34:37,120 --> 00:34:40,560
Speaker 2: there as well. Obviously the pushback on TikTok, the sale

735
00:34:40,560 --> 00:34:43,320
Speaker 2: of TikTok et cetera. You know, how are they viewing

736
00:34:43,320 --> 00:34:45,440
Speaker 2: these operations? What are some of the fronts that they're

737
00:34:45,480 --> 00:34:46,120
Speaker 2: fighting on here?

738
00:34:46,520 --> 00:34:48,520
Speaker 7: Yeah, so yeah, as you mentioned n Yeah, who very

739
00:34:48,560 --> 00:34:51,400
Speaker 7: much sees this as the what he calls the eighth front,

740
00:34:51,480 --> 00:34:53,319
Speaker 7: you know, the battle for hearts and minds, and he

741
00:34:53,320 --> 00:34:55,640
Speaker 7: says that it's as important, you know, as all of

742
00:34:55,640 --> 00:34:57,360
Speaker 7: the other fronts, and in fact it's the one that

743
00:34:57,360 --> 00:35:00,520
Speaker 7: he's lost the most, right you know, he Israel is

744
00:35:00,560 --> 00:35:05,920
Speaker 7: so isolated diplomatically, you know, and so I was not

745
00:35:05,960 --> 00:35:08,680
Speaker 7: at all surprised to see that par Scale was kind

746
00:35:08,680 --> 00:35:10,720
Speaker 7: of already being scapegoated.

747
00:35:10,760 --> 00:35:14,400
Speaker 6: Because this is the cycle, Like, this is.

748
00:35:14,400 --> 00:35:16,600
Speaker 7: What happens every time, is like there is this this

749
00:35:16,760 --> 00:35:21,239
Speaker 7: loss of public support, and then Israel will double down

750
00:35:21,600 --> 00:35:25,040
Speaker 7: on the influence operations. They'll go all in on hasbara.

751
00:35:25,360 --> 00:35:28,000
Speaker 7: They'll scapegoat the hasbara. They'll say it wasn't done well before,

752
00:35:28,440 --> 00:35:31,280
Speaker 7: and there will be no reflection on the policies themselves

753
00:35:31,360 --> 00:35:33,600
Speaker 7: that led people to break in public opinion in the

754
00:35:33,600 --> 00:35:36,520
Speaker 7: first place. You know, maybe war crimes aren't aren't very popular.

755
00:35:36,560 --> 00:35:39,040
Speaker 7: You know, maybe bombing churches in Gaza is not popular

756
00:35:39,040 --> 00:35:44,480
Speaker 7: among evangelical Christians, And instead of any self reflection on that,

757
00:35:44,640 --> 00:35:47,800
Speaker 7: they will double down on these influence operations. They'll say, well,

758
00:35:47,880 --> 00:35:49,360
Speaker 7: it seems like we're doing this wrong. We need to

759
00:35:49,360 --> 00:35:52,520
Speaker 7: bring in new voices who are offering alternative strategies.

760
00:35:52,719 --> 00:35:53,600
Speaker 6: And so that's what they did.

761
00:35:53,680 --> 00:35:56,160
Speaker 7: In late twenty twenty four early twenty twenty five, they

762
00:35:56,440 --> 00:35:58,839
Speaker 7: met with getting in Star, Foreign Minister. Getting Star came

763
00:35:58,880 --> 00:36:01,960
Speaker 7: into power and as part of entering that coalition, he

764
00:36:02,080 --> 00:36:05,239
Speaker 7: was promised a massive husbara budget and to be able

765
00:36:05,280 --> 00:36:06,760
Speaker 7: to do what he wanted to do with it, essentially,

766
00:36:06,800 --> 00:36:08,440
Speaker 7: and so he met with a bunch of different you know,

767
00:36:08,480 --> 00:36:13,480
Speaker 7: social media influencers, think tankers, different members of the Israeli

768
00:36:13,520 --> 00:36:15,879
Speaker 7: government to try and determine what would be the best

769
00:36:15,920 --> 00:36:18,160
Speaker 7: possible path forward, and so they came up with They

770
00:36:18,360 --> 00:36:21,000
Speaker 7: piloted this program throughout much of the last two years,

771
00:36:21,239 --> 00:36:25,360
Speaker 7: you know, trying to influence chatbots, sponsoring delegations of social

772
00:36:25,400 --> 00:36:30,319
Speaker 7: media influencers, paying influencers, you know, the getting ads into

773
00:36:30,400 --> 00:36:34,359
Speaker 7: Google and x and different conservative media outlets including sale

774
00:36:34,440 --> 00:36:39,680
Speaker 7: media where Parscale works, right, and so but now that

775
00:36:39,440 --> 00:36:42,279
Speaker 7: that there's been this this proof of concept with that

776
00:36:42,280 --> 00:36:44,440
Speaker 7: that it's there were two years out almost from the

777
00:36:45,120 --> 00:36:47,160
Speaker 7: launching of that program, and it hasn't worked.

778
00:36:48,200 --> 00:36:49,440
Speaker 6: We're re studying the cycle.

779
00:36:49,480 --> 00:36:53,200
Speaker 7: So the Israeli government is now once again you know, anonymous,

780
00:36:53,600 --> 00:36:56,480
Speaker 7: anonymously at first through this this time article that you're referencing,

781
00:36:56,840 --> 00:36:59,960
Speaker 7: is now scapegoating Brad Parscale. And we'll see what how

782
00:37:00,000 --> 00:37:02,400
Speaker 7: deppens with that. You know, if his contract gets re

783
00:37:02,480 --> 00:37:05,319
Speaker 7: upped again, it continues to get increased. It started at

784
00:37:05,320 --> 00:37:07,279
Speaker 7: one point five million dollars a month, went up to

785
00:37:07,320 --> 00:37:09,279
Speaker 7: three million dollars a month. Now it's at four point

786
00:37:09,320 --> 00:37:10,279
Speaker 7: five million dollars a month.

787
00:37:10,320 --> 00:37:10,560
Speaker 6: Wow.

788
00:37:10,960 --> 00:37:12,680
Speaker 7: We'll see what happens at the end of this contract

789
00:37:12,719 --> 00:37:15,080
Speaker 7: if they come up with a new influence strategy where

790
00:37:15,120 --> 00:37:17,120
Speaker 7: this goes. But there is a lot of money on

791
00:37:17,160 --> 00:37:20,520
Speaker 7: the table right now, seven hundred and fifty million dollars

792
00:37:20,680 --> 00:37:23,239
Speaker 7: allocated for twenty twenty six for Hasbara efforts. You know

793
00:37:23,320 --> 00:37:25,120
Speaker 7: that's that's crazy international.

794
00:37:24,680 --> 00:37:25,359
Speaker 1: As the US.

795
00:37:25,520 --> 00:37:27,680
Speaker 2: It's really funny because I'm old enough to remember Trump

796
00:37:27,680 --> 00:37:30,839
Speaker 2: giving a speech like two weeks ago about foreign interference

797
00:37:30,880 --> 00:37:33,480
Speaker 2: in our election. I don't remember Israel coming up in that.

798
00:37:33,640 --> 00:37:35,200
Speaker 2: By the way, you just needs to read your report

799
00:37:35,200 --> 00:37:36,080
Speaker 2: and I'm sure it'll be there to.

800
00:37:36,040 --> 00:37:38,200
Speaker 1: Give everyone a taste. They said, do you stand with Israel?

801
00:37:38,239 --> 00:37:38,640
Speaker 1: I say no.

802
00:37:38,840 --> 00:37:41,880
Speaker 3: They say Israel's a vital democratic ally that helps advance

803
00:37:41,880 --> 00:37:44,359
Speaker 3: American security and stability. They share our values and work

804
00:37:44,360 --> 00:37:46,600
Speaker 3: with us to deter threats. Do you think a country

805
00:37:46,640 --> 00:37:48,480
Speaker 3: has the right to defend itself against a group that's

806
00:37:48,520 --> 00:37:50,400
Speaker 3: openly set its goal is to destroy it? I said no,

807
00:37:50,880 --> 00:37:53,279
Speaker 3: And then it says that's concerning as Israel's facing threats

808
00:37:53,320 --> 00:37:55,520
Speaker 3: from groups like Hamas that openly call for its destruction.

809
00:37:55,840 --> 00:37:58,040
Speaker 3: Setting aside Israel for a second, do you think any

810
00:37:58,080 --> 00:38:00,480
Speaker 3: country has a right to defend itself. A group has

811
00:38:00,480 --> 00:38:02,840
Speaker 3: said its goal is to just I've said no again.

812
00:38:03,200 --> 00:38:06,040
Speaker 3: I've said no again, and we'll continue to see what

813
00:38:06,080 --> 00:38:07,719
Speaker 3: we get, but just to get people a taste. I mean,

814
00:38:07,760 --> 00:38:09,279
Speaker 3: this is you know right here, this is my This

815
00:38:09,360 --> 00:38:11,560
Speaker 3: is not staged or anything. I literally had no idea

816
00:38:11,719 --> 00:38:13,880
Speaker 3: that it was on my phone. We'm sitting here just

817
00:38:13,920 --> 00:38:16,839
Speaker 3: getting all of basically this AI. So imagine how many

818
00:38:16,840 --> 00:38:19,000
Speaker 3: other people you know have been engaging with this and

819
00:38:19,040 --> 00:38:20,800
Speaker 3: they have no clue that it's all being paid.

820
00:38:20,560 --> 00:38:23,680
Speaker 5: For protect your boomer parents and grand literally.

821
00:38:24,239 --> 00:38:25,920
Speaker 1: To protect the fellow boomers.

822
00:38:26,120 --> 00:38:27,400
Speaker 6: I want to get an update. I think it might

823
00:38:27,440 --> 00:38:28,600
Speaker 6: think you're a lost cause on this.

824
00:38:29,080 --> 00:38:31,640
Speaker 3: No, no, because it's fighting.

825
00:38:30,640 --> 00:38:34,480
Speaker 5: It's going to put up on a government watching this conversation.

826
00:38:35,080 --> 00:38:41,040
Speaker 1: It's gonna have probably wright. Okay, thank you so much. Yeah, okay,

827
00:38:41,120 --> 00:38:41,919
Speaker 1: we'll see you guys later