1
00:00:00,000 --> 00:00:02,137
.

2
00:00:02,220 --> 00:00:09,220
Thank you.

3
00:00:18,240 --> 00:00:25,240
Thank you.

4
00:00:48,240 --> 00:00:49,240
Bye.

5
00:00:50,250 --> 00:00:52,310
you

6
00:00:55,560 --> 00:01:02,560
Thank you.

7
00:01:25,560 --> 00:01:31,627
Thank you.

8
00:01:31,710 --> 00:01:38,710
Thank you.

9
00:02:01,710 --> 00:02:08,710
Thank you.

10
00:02:29,490 --> 00:02:33,500
Thank you. . you

11
00:02:34,830 --> 00:02:35,663
you

12
00:02:35,820 --> 00:02:42,820
Thank you.

13
00:03:05,820 --> 00:03:12,067
Thank you.

14
00:03:12,150 --> 00:03:13,150
Bye.

15
00:03:17,520 --> 00:03:24,520
Thank you.

16
00:03:47,520 --> 00:03:48,760
Thank you.

17
00:03:48,930 --> 00:03:53,840
Every morning, the President of the United
States receives the most exclusive

18
00:03:53,923 --> 00:03:56,530
briefing in the world. the PDB.

19
00:03:56,940 --> 00:04:00,990
the best intelligence on the planet
distilled into what matters most.

20
00:04:01,230 --> 00:04:06,970
But the contest that will define this
century isn't only fought with armies

21
00:04:07,054 --> 00:04:09,457
and spies. It's fought in labs.

22
00:04:09,540 --> 00:04:11,380
and code and chips.

23
00:04:11,670 --> 00:04:16,798
in technology. That's why we created
the President's Tech Brief. Twice a week,

24
00:04:16,881 --> 00:04:20,890
we bring you a briefing
on technology and national security.

25
00:04:21,120 --> 00:04:23,377
the one that we would
put on the Resolute desk.

26
00:04:23,460 --> 00:04:26,208
The world is moving faster, and so are we,

27
00:04:26,292 --> 00:04:31,669
by tracking the current state of the tech
competition with China through our tech

28
00:04:31,753 --> 00:04:36,000
scorecard. Live from the SESP
studio, your briefing starts now.

29
00:04:36,120 --> 00:04:38,420
now. .

30
00:04:42,360 --> 00:04:44,366
Hi, welcome to the President's Tech Brief.

31
00:04:44,449 --> 00:04:47,251
I'm Ily with the SESP and we're
here live from Washington.

32
00:04:47,335 --> 00:04:50,883
I'm here joined by Martin. Martin.
Good to see you, Ily. Good to be back.

33
00:04:50,966 --> 00:04:54,564
Two days ago, you and I, Martin,
talked about the lunch that the President

34
00:04:54,647 --> 00:04:58,570
was having with the AI.
Ily. CEOs after that lunch.

35
00:04:58,860 --> 00:05:03,312
The document was signed, so what
the president calls this morally binding

36
00:05:03,395 --> 00:05:07,474
document. So today we're going
to analyze what that document means.

37
00:05:07,557 --> 00:05:12,692
It has four layers of control that I think
the president expects all these companies

38
00:05:12,776 --> 00:05:17,389
to follow through. And if you have
any questions from the audience or any

39
00:05:17,473 --> 00:05:20,955
comments, please let us know.
Our question for the day is,

40
00:05:21,038 --> 00:05:23,844
do you trust AI companies
to police themselves?

41
00:05:23,927 --> 00:05:28,577
Your response or responses are already
rolling in and we'll read some of them

42
00:05:28,660 --> 00:05:33,270
during the conversation, Martin.
Great, great. So what's your take, Martin?

43
00:05:33,690 --> 00:05:37,817
Well, I think the four layers
are a great starting point, right?

44
00:05:37,900 --> 00:05:43,145
What I really want to then know is if one
of these tests does indicate a failure,

45
00:05:43,229 --> 00:05:47,619
how does that change how the company
is structured, how it operates?

46
00:05:47,702 --> 00:05:51,368
So there's some unknowns there,
implementation deadlines,

47
00:05:51,452 --> 00:05:54,329
who the third-party
auditors are going to be,

48
00:05:54,412 --> 00:05:58,276
and whether or not any of these
reports will be made public.

49
00:05:58,359 --> 00:06:03,457
I think there is a benefit to maximize
the transparency, just so that we can –

50
00:06:03,540 --> 00:06:06,480
increased trust in what
these companies are building.

51
00:06:06,960 --> 00:06:10,028
I think first you need to build trust
between the companies because I think

52
00:06:10,111 --> 00:06:12,367
having an independent auditor
has to be an actor or...

53
00:06:12,450 --> 00:06:15,152
organization or somebody
that really everybody trusts?

54
00:06:15,236 --> 00:06:19,331
And then are the companies really going
to showcase everything they're working on

55
00:06:19,414 --> 00:06:20,910
and what they're building on?

56
00:06:21,510 --> 00:06:26,235
And then thirdly is what happens when
they don't follow this. We saw the FTC,

57
00:06:26,319 --> 00:06:32,140
we talked about this. The FTC,
I think, court ruling, right?

58
00:06:32,580 --> 00:06:36,080
order on that one. Case. Case against
open AI and Anthropic yesterday.

59
00:06:36,330 --> 00:06:39,840
Yeah, exactly. So a lot
of developments still in play.

60
00:06:39,924 --> 00:06:44,765
So as we go through the month, we'll
see how this all starts planning out,

61
00:06:44,848 --> 00:06:49,690
this is shaping out. But overall,
I think this is a good step in the right

62
00:06:49,773 --> 00:06:55,214
direction. And hopefully it'll take some
of the heat out of the debate right now so

63
00:06:55,297 --> 00:06:59,290
we can have more balanced
discussion on how we move forward.

64
00:07:06,330 --> 00:07:10,860
from what we've seen so far. We had two
comments that came on our LinkedIn page.

65
00:07:11,340 --> 00:07:12,720
If our team can pull it up.

66
00:07:16,920 --> 00:07:20,620
I think people had the same questions.
Like you and I have a conversation.

67
00:07:24,420 --> 00:07:27,046
Similar concerns, I believe.
Yeah, I think so.

68
00:07:27,130 --> 00:07:31,759
I think this is largely representative
of how a lot of people are thinking about

69
00:07:31,842 --> 00:07:36,589
these issues. And so, yeah, it's fantastic
to see people engaging with SESPs posts

70
00:07:36,673 --> 00:07:39,476
and on these issues,
how we're thinking about it.

71
00:07:39,559 --> 00:07:44,247
And that's exactly what we want, right?
To just have a mature, robust discussion.

72
00:07:44,330 --> 00:07:47,192
We also have a question
on Signal that I just got.

73
00:07:47,276 --> 00:07:50,280
So what do you think
is going to happen after this?

74
00:07:51,410 --> 00:07:53,910
The President calls it
morally binding document.

75
00:07:54,840 --> 00:07:56,947
Sure. Well,

76
00:07:57,030 --> 00:08:01,614
I want to see the companies then telling
us what the implementation actually looks

77
00:08:01,698 --> 00:08:05,656
like, right? When they start,
what their implementation plan is. Again,

78
00:08:05,739 --> 00:08:06,650
this provides...

79
00:08:07,080 --> 00:08:12,131
the leaders of these tech companies
a great opportunity to show the American

80
00:08:12,215 --> 00:08:15,510
public that they're taking
these issues seriously.

81
00:08:15,593 --> 00:08:20,660
So I would love to see for each of the
companies that are impacted by this.

82
00:08:21,180 --> 00:08:25,010
How are you going to act now
that you've signed this document?

83
00:08:25,094 --> 00:08:29,239
We will cover this topic at length
with our guest, Brandon Daniels,

84
00:08:29,323 --> 00:08:32,100
shortly from Exeger,
who was at the meeting.

85
00:08:32,400 --> 00:08:35,420
And then we also had
a conversation yesterday on our...

86
00:08:35,520 --> 00:08:40,214
weekly segment across the aisle with Mike
Kuyken and Josh Hodges. But first,

87
00:08:40,297 --> 00:08:45,368
you know, this was an interesting week and
probably to a certain degree a sad week

88
00:08:45,451 --> 00:08:47,840
because we left, our forces left Iraq.

89
00:08:48,180 --> 00:08:53,800
where you served, Martin, after how
many years now? 23 years probably since

90
00:08:53,940 --> 00:08:57,058
the operation against Saddam
Hussein from our forces.

91
00:08:57,141 --> 00:09:01,104
We have Ilbert here from our team
joining us who also served there.

92
00:09:01,187 --> 00:09:05,513
So I just want to quickly before we
move on on the next topic just to get

93
00:09:05,596 --> 00:09:06,429
your take.

94
00:09:06,690 --> 00:09:11,075
You were there how many years ago,
Martin? Exactly 16 years ago. So, yeah,

95
00:09:11,158 --> 00:09:14,757
I was... Tell us how you landed there.
What did you do? Yeah,

96
00:09:14,841 --> 00:09:19,044
so I was working with the Central
Intelligence Agency at the time. And,

97
00:09:19,128 --> 00:09:22,365
you know, I was working
on completely different topics,

98
00:09:22,448 --> 00:09:27,195
but I wanted to volunteer and, you know,
contribute myself to the very important

99
00:09:27,278 --> 00:09:29,150
mission that we were there for.

100
00:09:29,880 --> 00:09:35,367
I was originally slated to be in Balad,
but then it turned out that there was

101
00:09:35,451 --> 00:09:38,840
an underlap in staff
coverage in Anbar province.

102
00:09:38,923 --> 00:09:41,600
So I actually found myself in Ramadi.

103
00:09:42,240 --> 00:09:45,558
for the rest of the year,
which was a pretty wild time.

104
00:09:45,642 --> 00:09:49,516
Was that a formative experience
for you, Martin? Oh, absolutely.

105
00:09:49,600 --> 00:09:54,300
I never would have fathomed that I
would be doing the work that I was doing.

106
00:09:54,720 --> 00:09:59,927
how much responsibility I
was entrusted with, got to work with,

107
00:10:00,010 --> 00:10:04,847
Special Operations Forces, with diplomats,

108
00:10:04,930 --> 00:10:10,046
from the United States as well as Iraq and
then a whole bunch of provincial leaders

109
00:10:10,129 --> 00:10:15,839
as well. Lumber, what about you? How did
you end up going there? In Iraq? Well,

110
00:10:15,922 --> 00:10:20,863
I was in Iraq three different times.
One was in the immediate aftermath of

111
00:10:20,946 --> 00:10:24,529
the invasion. This was in 2003.
I was working for U.S.

112
00:10:24,612 --> 00:10:29,757
Institute of Peace at that time, so
we're trying to do some capacity building

113
00:10:29,840 --> 00:10:35,324
for the newly stood-up Iraqi ministries.
And then my second time there was in 2007

114
00:10:35,407 --> 00:10:38,243
during the surge that
General Petraeus led.

115
00:10:38,326 --> 00:10:41,110
I was privileged to be part of his staff.

116
00:10:41,350 --> 00:10:46,510
So I was there for a year. That's really
when we sort of turned the tide against

117
00:10:46,593 --> 00:10:50,680
the a civil war that had erupted in 2006.

118
00:10:51,070 --> 00:10:57,033
And then I traveled quite extensively
to Iraq in 2015-2016 during the height of

119
00:10:57,117 --> 00:11:00,620
our Defeat ISIS campaign.
Secretary Ash Carter.

120
00:11:01,000 --> 00:11:02,500
Yeah, who led that effort.

121
00:11:02,620 --> 00:11:08,163
So this is sort of the third operation
that we're bringing to a conclusion called

122
00:11:08,247 --> 00:11:13,040
inherent resolve. It's bittersweet
in many ways. I think on one hand,

123
00:11:13,390 --> 00:11:18,583
You know, I'm glad that U.S. service
members no longer need to be there,

124
00:11:18,666 --> 00:11:23,210
and the Iraqis can take care
of their own security. We lost...

125
00:11:23,410 --> 00:11:27,290
nearly 4,600 U.S. service members.

126
00:11:28,780 --> 00:11:33,130
300-day coalition members as well. But

127
00:11:33,550 --> 00:11:39,392
I saw twice what
the withdrawal from Iraq led to,

128
00:11:39,475 --> 00:11:44,070
in either instance in 2006 or in 2011.

129
00:11:44,770 --> 00:11:48,549
The outcome was positive.
You know, in 2011,

130
00:11:48,633 --> 00:11:53,110
ISIS came back and engulfed
the country in turmoil.

131
00:11:53,200 --> 00:11:59,204
So I hope that's not the case this time.
I would have preferred that we kept some

132
00:11:59,287 --> 00:12:00,640
residual presence.

133
00:12:01,270 --> 00:12:02,103
sure that

134
00:12:02,620 --> 00:12:08,613
threats in that part of the world don't
impact our interests and the stability

135
00:12:08,697 --> 00:12:09,530
of Iraq.

136
00:12:10,240 --> 00:12:12,440
We are where we are now.

137
00:12:12,550 --> 00:12:15,843
I've never served in Iraq,
but I was there 48 hours, I think,

138
00:12:15,926 --> 00:12:19,690
with the Deputy Secretary of Defense
once on the way to Afghanistan.

139
00:12:19,870 --> 00:12:24,258
It was fascinating. That landing in
Baghdad is something I will never forget.

140
00:12:24,341 --> 00:12:26,490
Did you do the corkscrew? Yeah, yeah.

141
00:12:27,160 --> 00:12:31,650
Yeah. How was it for you, Martin,
this week to see this coming to an end?

142
00:12:32,620 --> 00:12:36,310
It brought back a lot
of memories. Frankly, yeah,

143
00:12:36,393 --> 00:12:42,239
and that story and just seeing those
pictures that I hadn't looked at in many

144
00:12:42,322 --> 00:12:47,937
years, I was like, oh, holy cow,
what a time that was. It was, yeah, just,

145
00:12:48,021 --> 00:12:52,327
again, a very important
experience for me as an American,

146
00:12:52,410 --> 00:12:54,720
for me as a young CIA officer.

147
00:12:55,480 --> 00:12:59,365
So, yeah, very formative.
Fascinating. I mean, you know,

148
00:12:59,449 --> 00:13:05,260
just to thank everybody that served there,
many, many troops, a lot of lost lives.

149
00:13:05,830 --> 00:13:10,448
And, you know, the region is still,
I think, evolving. I mean, we are now,

150
00:13:10,531 --> 00:13:11,365
you know...

151
00:13:11,680 --> 00:13:15,472
in war against Iran. And I think this
doesn't come at the expense of us leaving

152
00:13:15,555 --> 00:13:19,151
Iraq, as you said, like every time we
withdrew or we changed the operation,

153
00:13:19,234 --> 00:13:22,830
you know, other actors filled the vacuum.
So hopefully this time that's not

154
00:13:22,913 --> 00:13:25,713
the case. But we'll be
interested to know your thoughts.

155
00:13:26,260 --> 00:13:30,597
Yeah, I mean, to Martin's point,
this was formative not just for me,

156
00:13:30,680 --> 00:13:34,840
but I think for an entire generation
of policymakers in the U.S.

157
00:13:36,820 --> 00:13:41,623
I mean, Iraq, Afghanistan,
the war on terror. In many ways,

158
00:13:41,706 --> 00:13:48,083
I think the Iraq debate continues to
this day among sort of the foreign policy

159
00:13:48,166 --> 00:13:53,466
elites. You know, I think Iraq
has been through a lot, certainly.

160
00:13:53,549 --> 00:14:00,317
It's certainly in a much better place now
than I think it was in 2003 or 2006, but

161
00:14:00,400 --> 00:14:02,200
or certainly in 2014.

162
00:14:06,670 --> 00:14:12,529
We have different ways of staying engaged
there with the Iraqi government and make

163
00:14:12,613 --> 00:14:18,327
sure that Iraq remains independent. It
certainly doesn't become sort of a stooge

164
00:14:18,410 --> 00:14:23,690
of the Iranian regime and that it
becomes a net contributor of security to

165
00:14:23,773 --> 00:14:24,606
the region.

166
00:14:26,110 --> 00:14:31,561
It looks promising, but again, you know,
I'm cautious because the last two times,

167
00:14:31,644 --> 00:14:35,470
I think, things didn't work
out as the way we had hoped.

168
00:14:35,890 --> 00:14:36,723
it would

169
00:14:36,970 --> 00:14:40,573
Gentlemen, thank you to both of you.
Thank you for your service. You know,

170
00:14:40,657 --> 00:14:43,164
I hope you don't have to go
back anytime soon there,

171
00:14:43,247 --> 00:14:45,197
and I hope that country is on the path.

172
00:14:46,090 --> 00:14:49,850
towards stability and brighter, you know,
perspective for their people first,

173
00:14:49,934 --> 00:14:52,234
and then for the region.
Thank you. Thank you.

174
00:14:54,010 --> 00:15:01,010
*music*

175
00:15:10,080 --> 00:15:12,442
Welcome back.
We're joined by Brandon Daniels,

176
00:15:12,525 --> 00:15:16,163
who is the founder and CEO of Exeger,
one of the leading AI companies.

177
00:15:16,246 --> 00:15:20,416
But the other reason why we wanted to have
you today, Brandon, good to have you.

178
00:15:20,499 --> 00:15:24,167
Good to be here. Yeah, thank you
guys for having me. Yeah, it's great.

179
00:15:24,250 --> 00:15:27,637
is you were one of the CEOs attending the
meeting with the president two days ago.

180
00:15:27,720 --> 00:15:29,838
So we wanted to take...
I came from the White House.

181
00:15:29,921 --> 00:15:33,096
Got to go back to the White House right
after this. Thank you for doing this.

182
00:15:33,180 --> 00:15:33,827
So tell us...

183
00:15:33,910 --> 00:15:38,114
What happened inside that room? What
was the conversation about and then how

184
00:15:38,198 --> 00:15:41,470
that led to the signing of the document?
Yeah, yeah. So...

185
00:15:41,680 --> 00:15:43,600
First, just taking a step back,

186
00:15:43,780 --> 00:15:46,880
you know a lot of people
have talked about that meeting

187
00:15:46,990 --> 00:15:50,990
one of the things that I haven't seen
reported anywhere is that the meeting

188
00:15:51,073 --> 00:15:51,907
was three hours

189
00:15:52,510 --> 00:15:56,650
And that amount of time...

190
00:15:56,740 --> 00:16:01,783
with that group of people,
led to a really robust discussion.

191
00:16:01,866 --> 00:16:07,918
Led to a really robust discussion
around the future of superintelligence,

192
00:16:08,001 --> 00:16:15,001
the name superintelligence. It also led to
a really robust discussion around safety,

193
00:16:15,700 --> 00:16:19,660
around scalability, around...

194
00:16:19,810 --> 00:16:26,775
sort of the infrastructure that needs
to be built around superintelligence,

195
00:16:26,859 --> 00:16:30,430
it led to a very, very vibrant debate.

196
00:16:30,730 --> 00:16:35,080
You know, because a lot of people have
said the meeting was congenial. It was.

197
00:16:35,164 --> 00:16:36,130
It was congenial.

198
00:16:36,460 --> 00:16:41,602
A lot of people have said the meeting was
a lot of people interacting and engaging,

199
00:16:41,685 --> 00:16:45,400
some people that hadn't met before.
Yep, that happened too.

200
00:16:45,940 --> 00:16:50,420
But there was – the president
encouraged robust debate.

201
00:16:50,860 --> 00:16:56,420
So we started the meeting,
and before the meeting –

202
00:16:56,770 --> 00:17:03,770
Mark Zuckerberg and Jensen had put
together a series of principles that they

203
00:17:04,927 --> 00:17:06,430
believed to be

204
00:17:06,550 --> 00:17:11,914
you know inherently superseding principles
that we all could agree to regardless

205
00:17:11,997 --> 00:17:17,020
of where we sat on the super intelligence
existential threat debate. Right.

206
00:17:17,104 --> 00:17:22,331
And so, you know, we had that piece
of paper going in and that had been shared

207
00:17:22,415 --> 00:17:26,825
with Speaker Johnson. It had been
shared with some of the folks in

208
00:17:26,909 --> 00:17:31,097
the administration like we knew
we were going to come out with

209
00:17:31,180 --> 00:17:36,760
an agreement or no agreement, right,
on information that had been shared.

210
00:17:37,240 --> 00:17:41,372
Then the president was that document
coordinated between Mark and Jensen

211
00:17:41,455 --> 00:17:46,114
beforehand or. Yeah. Yeah. But it was also
circulated to the CEOs that were going

212
00:17:46,197 --> 00:17:50,120
to be there. Right. So, you know,
that that that set of principles.

213
00:17:50,230 --> 00:17:54,150
you know, was not new
to the people in the room.

214
00:17:54,970 --> 00:17:59,710
Then the president opened it up
and asked us each for our views.

215
00:17:59,890 --> 00:18:04,364
We all had five to ten minutes to talk,
and then we had an opportunity to exchange

216
00:18:04,447 --> 00:18:06,670
thoughts on the things that people said.

217
00:18:07,150 --> 00:18:10,996
So, you know, we we we all
came to a consensus. I mean,

218
00:18:11,080 --> 00:18:16,212
everybody thinks like, OK, you know,
there's there's so much alignment in

219
00:18:16,296 --> 00:18:21,571
that group because everyone's running
companies in super intelligence in AI

220
00:18:21,655 --> 00:18:25,370
that that all have the same
motive. That's not true.

221
00:18:25,570 --> 00:18:28,750
I mean, the motives of Nikesh

222
00:18:28,930 --> 00:18:35,057
versus the motives of Haktan, versus the
motives of Dario, are all very different.

223
00:18:35,140 --> 00:18:39,146
All of us have different backgrounds. All
of us have different business journeys.

224
00:18:39,229 --> 00:18:41,973
All of us are on different
sides of the spectrum, right?

225
00:18:42,057 --> 00:18:44,480
Any unresolved issues
that you're able to share?

226
00:18:46,480 --> 00:18:51,928
I think there was there was a debate and
there were some issues that didn't have

227
00:18:52,011 --> 00:18:57,529
to be resolved. Right. Like which which
which issue do people believe is the more

228
00:18:57,612 --> 00:19:02,540
important one? the existential
threat of not being the winner in AI?

229
00:19:03,730 --> 00:19:10,730
or the existential threat of a potential
errant or a purposefully set model

230
00:19:11,279 --> 00:19:18,279
doomsday, right? We didn't have to agree
or disagree which one was more likely.

231
00:19:19,990 --> 00:19:24,287
But I will tell you, everyone
in that room said we had to win the race.

232
00:19:24,370 --> 00:19:28,649
There was not, to a man, everyone said
the most important thing is that we win

233
00:19:28,732 --> 00:19:30,130
the race. It for sure is.

234
00:19:30,580 --> 00:19:35,067
But that doesn't mean that some people
didn't have on the same level the risk

235
00:19:35,151 --> 00:19:38,000
of AI doing something
that we didn't anticipate.

236
00:19:38,740 --> 00:19:43,353
But the superseding principles
that we came up with were one, look,

237
00:19:43,436 --> 00:19:46,037
everyone in that room is a programmer.

238
00:19:46,120 --> 00:19:50,269
So we've all built these agents. Like I
talked about when the president asked me

239
00:19:50,352 --> 00:19:54,025
for my comments, I said, you know,
we have to go back to the beginning.

240
00:19:54,108 --> 00:19:57,675
The agents that I built, the first
agents that I built in my company,

241
00:19:57,758 --> 00:19:59,937
because I actually sat down and built the.

242
00:20:00,020 --> 00:20:01,720
I know, I remember you telling me.

243
00:20:02,360 --> 00:20:07,725
actually were not for my company. The
first one that I ever built was actually

244
00:20:07,808 --> 00:20:11,580
to diagnose a complex set
of cancers that my wife has.

245
00:20:11,900 --> 00:20:15,727
She's got immunodiseases.
She's got what's called Ehlers-Danlos.

246
00:20:15,810 --> 00:20:18,782
It actually erodes away
at your connective tissue.

247
00:20:18,866 --> 00:20:23,059
And I wanted to understand the link
between this sort of constellation

248
00:20:23,143 --> 00:20:25,167
of cancer that she has and cancer.

249
00:20:25,250 --> 00:20:29,384
these immune issues, right,
or autoimmune diseases, essentially,

250
00:20:29,468 --> 00:20:33,290
and see if there was something
we could do to attack both.

251
00:20:33,710 --> 00:20:37,770
And my wife right now is in clinical
remission, 'cause we did attack both.

252
00:20:38,000 --> 00:20:40,984
And so I said, we can't forget
that in this conversation.

253
00:20:41,067 --> 00:20:44,158
And do we want to be the ones
that have that kind of power?

254
00:20:44,241 --> 00:20:47,900
Do we want someone else to be
the ones that have that kind of power?

255
00:20:48,200 --> 00:20:51,957
And so, you know, we had
a conversation and we said, okay,

256
00:20:52,040 --> 00:20:55,929
we know that we can put governance
and code. So let's do it.

257
00:20:56,012 --> 00:20:59,306
Because AI models, super
intelligence capabilities,

258
00:20:59,389 --> 00:21:01,640
they are impotent without tooling.

259
00:21:01,940 --> 00:21:05,942
So it's actually the tooling and the
software and the access to data that makes

260
00:21:06,025 --> 00:21:08,300
it potent, makes it so
it can be productive.

261
00:21:08,570 --> 00:21:12,868
We can put governance in that tooling. We
can put what I call the surge protector.

262
00:21:12,952 --> 00:21:16,075
In my systems today,
if something is draining too much cost,

263
00:21:16,158 --> 00:21:19,495
draining too much compute,
or is covering too much cybersurface,

264
00:21:19,578 --> 00:21:22,410
there's actually a surge
protector that shuts it off.

265
00:21:22,850 --> 00:21:26,128
That doesn't mean that that AI model
can't keep learning and getting better

266
00:21:26,211 --> 00:21:29,130
and better and better and better.
It doesn't mean I have to give up

267
00:21:29,214 --> 00:21:30,214
my innovative place.

268
00:21:30,410 --> 00:21:35,008
And that's one thing Jensen
and Zuckerberg both said was, hey, like,

269
00:21:35,091 --> 00:21:40,171
I don't think that innovation and safety
are loggerheads. They're actually,

270
00:21:40,254 --> 00:21:45,830
you know, one thing here. So, you know,
I come back to the point of disagreement.

271
00:21:45,950 --> 00:21:51,043
On the point that we all have to know
the model purpose and set up monitoring,

272
00:21:51,126 --> 00:21:54,687
safeguards, and guardrails
to make sure that we are...

273
00:21:54,770 --> 00:21:58,970
We are monitoring whether or not
the model is operating for its purpose.

274
00:21:59,060 --> 00:22:02,623
We all agreed to that. We then all
agreed that we could do it in code.

275
00:22:02,706 --> 00:22:05,001
We know that we can put
that governance into code,

276
00:22:05,084 --> 00:22:07,234
but we have to empower
our people to do it.

277
00:22:07,430 --> 00:22:11,084
And then the thing that I thought
was brilliant and, you know,

278
00:22:11,167 --> 00:22:15,485
even Elon and others were aligned
with this is we should have third party

279
00:22:15,568 --> 00:22:20,066
external auditors. Any discussion on how
those third party auditors would be

280
00:22:20,149 --> 00:22:24,105
selected? Are there specific
criteria that we're going to look for?

281
00:22:24,188 --> 00:22:27,420
I think one of the things
that was interesting is, one,

282
00:22:27,504 --> 00:22:30,374
they have to be experts,
right, to do the audits.

283
00:22:30,457 --> 00:22:33,117
And that's a very small
population of people.

284
00:22:33,200 --> 00:22:37,324
But there's also an opportunity to train
people and have people become fluent

285
00:22:37,407 --> 00:22:41,804
in these tools. And so that's one thing
is we have to be able to attract that kind

286
00:22:41,888 --> 00:22:45,739
of a talent into the auditors.
But one of the things that I also thought

287
00:22:45,822 --> 00:22:49,741
was interesting is, The president
came up with this idea that, you know,

288
00:22:49,824 --> 00:22:53,900
you have a board for those auditors,
and those might not be people that,

289
00:22:54,080 --> 00:22:58,300
that are programmers or developers. Those
might be great lawyers that understand

290
00:22:58,383 --> 00:23:02,066
the technology, or great policy
makers that understand the technology,

291
00:23:02,149 --> 00:23:06,560
which I thought was really brilliant. And
everyone kind of really liked that idea.

292
00:23:06,680 --> 00:23:08,880
I wasn't sure if, you know...

293
00:23:09,140 --> 00:23:12,447
We're all technologists,
and there's a little bit of this, like,

294
00:23:12,531 --> 00:23:15,679
technocrat feel to the meeting,
right? But, I mean, you know,

295
00:23:15,762 --> 00:23:18,592
Vice President Vance said
the same thing. I mean, like,

296
00:23:18,676 --> 00:23:22,884
there was agreement that we should have
people that are thinking about more than,

297
00:23:22,967 --> 00:23:24,980
like, how the technology is effective.

298
00:23:25,160 --> 00:23:29,016
but think about it more from like
a humanitarian impact perspective that I

299
00:23:29,099 --> 00:23:32,240
thought was really interesting
to that third-party auditor.

300
00:23:32,450 --> 00:23:37,610
We have this four-layer framework now.
Which of the layers do you see being most

301
00:23:37,694 --> 00:23:40,250
consequential for how companies will...

302
00:23:40,370 --> 00:23:45,119
behave going forward? I think one,
empowering the people beneath you to make

303
00:23:45,203 --> 00:23:49,888
sure that you have the guardrails
and capabilities set up to make sure it's

304
00:23:49,972 --> 00:23:52,017
safe. We have to do that already.

305
00:23:52,100 --> 00:23:57,040
I like to think of our artificial
intelligence and Palantirs and Nikesh's

306
00:23:57,123 --> 00:24:02,201
from Palo Alto networks. I like to think
of our AI as ruggedized AI, right,

307
00:24:02,284 --> 00:24:07,637
because we deploy to defense networks
and we deploy to, you know, FedRAMP-high,

308
00:24:07,721 --> 00:24:11,904
you know, systems that are
managing cargo and stuff like that.

309
00:24:11,987 --> 00:24:16,460
So we can't have errant decisions.
So all of that stuff is there.

310
00:24:22,100 --> 00:24:26,671
the loop that ends up taking down
a system or crashing a bunch of compute

311
00:24:26,755 --> 00:24:27,588
or whatever.

312
00:24:28,820 --> 00:24:31,406
That empowerment
is really important, right?

313
00:24:31,489 --> 00:24:36,380
And I think it addresses some of what the
FTC has brought up in the investigations

314
00:24:36,463 --> 00:24:38,040
into Anthropic and OpenAI.

315
00:24:38,720 --> 00:24:43,403
And then I thought the idea that we both
have board committees and an external

316
00:24:43,487 --> 00:24:47,742
auditor that reports that board
committee on AI development and safety,

317
00:24:47,826 --> 00:24:50,431
I thought was was also
really consequential.

318
00:24:50,514 --> 00:24:53,792
I know you're heading back
to the White House, Brandon.

319
00:24:53,876 --> 00:24:57,276
Tell us a little bit where
does Exeger stand in this now?

320
00:24:57,359 --> 00:25:01,859
What are you working on? What makes
you excited really about the next steps

321
00:25:01,942 --> 00:25:06,320
in this space? So so the way that the
meeting went when we when we closed

322
00:25:06,403 --> 00:25:08,420
the meeting, folks scattered off.

323
00:25:08,720 --> 00:25:12,795
I don't think Jensen was on that signed
document, was he? I can't remember.

324
00:25:12,878 --> 00:25:17,175
Or six tech CEOs, but I don't remember.
I think Jensen was on it. Was he on it?

325
00:25:17,259 --> 00:25:21,334
I couldn't remember. The way that it
happened is we closed out the meeting,

326
00:25:21,417 --> 00:25:25,104
and I know Alex wasn't on it,
because Alex had to run. I had to run.

327
00:25:25,187 --> 00:25:29,151
Lisa had to run. There were a bunch
of people that had to run right after

328
00:25:29,235 --> 00:25:30,068
the meeting.

329
00:25:38,720 --> 00:25:43,396
and have, and even though, I mean,
Anthropic was not destined to be where

330
00:25:43,479 --> 00:25:48,546
they are. They were just very smart about
creating their models, right? I mean,

331
00:25:48,630 --> 00:25:53,780
they were an underdog in comparison to
OpenAI or NVIDIA or anybody else, right?

332
00:25:54,230 --> 00:25:57,745
So the beauty of where we
are is it's democratized,

333
00:25:57,828 --> 00:26:01,797
and so somebody not in that
room might be the best person

334
00:26:01,880 --> 00:26:06,200
like super intelligence company in three
years, which I think is super cool.

335
00:26:06,380 --> 00:26:10,835
But the idea is now what we're doing is we
all have to agree to those principles.

336
00:26:10,918 --> 00:26:14,196
So in the next few days,
we'll all sign on to that document.

337
00:26:14,279 --> 00:26:18,061
We'll all agree to it, and we will
all implement it in our companies.

338
00:26:18,145 --> 00:26:20,583
Is there an agree-to
implementation timeline,

339
00:26:20,666 --> 00:26:23,048
or is it up to each
individual company to...

340
00:26:23,131 --> 00:26:27,361
I think we will get into those details,
but it's way faster than anyone could

341
00:26:27,445 --> 00:26:31,171
regulate the industry. I mean,
a lot of people talk about, oh, yeah,

342
00:26:31,254 --> 00:26:35,680
let's regulate, let's regulate, let's
regulate. The thing that I thought was...

343
00:26:35,780 --> 00:26:41,112
wild about that meeting is industry
is taking action to provide safety faster

344
00:26:41,195 --> 00:26:45,964
than anyone else could act.
And there was alignment in that room from

345
00:26:46,048 --> 00:26:48,720
the administration to all of the CEOs.

346
00:26:49,400 --> 00:26:51,345
Brandon, thank you.
I know you're a busy man.

347
00:26:51,428 --> 00:26:54,049
Thank you for dropping by and
letting us know what happened.

348
00:26:54,132 --> 00:26:56,843
A huge historic meeting that you
were part of. It was awesome.

349
00:26:56,926 --> 00:27:00,026
Best to you and the company
and we'll see you soon. Thank you.

350
00:27:00,890 --> 00:27:02,950
you

351
00:27:18,260 --> 00:27:22,289
conversation with Mike Kuyken and Josh
Hodges for this particular meeting.

352
00:27:22,372 --> 00:27:25,623
They were actually here in person.
They were here in person.

353
00:27:25,706 --> 00:27:27,540
So let's watch that conversation.

354
00:27:28,010 --> 00:27:32,511
with both of our friends. So Martin,
our next segment is Across the Aisle,

355
00:27:32,594 --> 00:27:34,647
and for the first time actually we have

356
00:27:34,730 --> 00:27:37,437
Both Mike and Josh here in Portland.
Gentlemen, good to have you.

357
00:27:37,520 --> 00:27:41,307
Great to be here. It's great to be here.
We're celebrating five years of SCSP.

358
00:27:41,390 --> 00:27:43,862
Thank you for coming.
Good party today. Yeah.

359
00:27:43,945 --> 00:27:46,670
For those of you that don't
know, Mike Kaiken...

360
00:27:46,760 --> 00:27:49,796
former National Security Advisor
to Senate Leader Chuck Schumer,

361
00:27:49,879 --> 00:27:53,779
Joshua Hodges, former National Security
Advisor to House Speaker Mike Johnson.

362
00:27:54,500 --> 00:27:59,547
Big titles, big people, big personalities.
I'm trying to be less big every day.

363
00:27:59,630 --> 00:28:03,184
All right. Let's jump into
the topic, Martin. Yesterday,

364
00:28:03,267 --> 00:28:05,540
the president hosted major AI CEOs.

365
00:28:06,050 --> 00:28:08,510
The White House, they issued a statement.

366
00:28:09,110 --> 00:28:12,590
Let's get into it. Mike, your take.

367
00:28:13,010 --> 00:28:16,450
Listen, I think it's good that I
always think it's good when industry

368
00:28:16,533 --> 00:28:19,616
and government are in a conversation.
And I think it's simple.

369
00:28:19,699 --> 00:28:22,832
I think government has a good
understanding of a lot of issues.

370
00:28:22,916 --> 00:28:27,122
I think artificial intelligence is one of
these areas where the technology is moving

371
00:28:27,205 --> 00:28:31,104
incredibly quickly. And it's important
that the smart people in the government

372
00:28:31,188 --> 00:28:34,219
actually understand what
the frontier is, what it looks like,

373
00:28:34,302 --> 00:28:37,334
what the what the opportunity
is and what the challenges are.

374
00:28:37,417 --> 00:28:40,500
And so anytime there's a convening
of industry and government,

375
00:28:40,583 --> 00:28:42,370
I generally consider it a positive.

376
00:28:43,010 --> 00:28:47,715
I mean, I'm a Democrat, so I was a little
bummed to make sure to see that there were

377
00:28:47,798 --> 00:28:51,192
no Democrats there.
You'll remember Leader Schumer. Actually,

378
00:28:51,275 --> 00:28:56,037
you were there with with with Eric Schmidt
and a whole bunch of other including Pete,

379
00:28:56,121 --> 00:28:59,116
including Jensen and Elon
Musk and sort of life. Yeah.

380
00:28:59,199 --> 00:29:01,587
And and it was five
or three years ago now.

381
00:29:01,670 --> 00:29:03,110
And so, uh,

382
00:29:03,290 --> 00:29:05,957
You know, this has been done before,
but it's always good to see government

383
00:29:06,040 --> 00:29:07,090
and industry talking.

384
00:29:07,340 --> 00:29:10,407
Yeah, Josh. Yeah, I mean, look,
yes, the meeting obviously happened.

385
00:29:10,490 --> 00:29:12,445
I think there's good
photo evidence of that.

386
00:29:12,528 --> 00:29:16,257
I think what we saw yesterday was really
the industry come together in a new way.

387
00:29:16,340 --> 00:29:19,100
new administration with
a new recognition that...

388
00:29:19,280 --> 00:29:21,632
Really, action has to be
put behind the words here.

389
00:29:21,716 --> 00:29:24,928
I thought it was great that they
got everybody aside the declaration,

390
00:29:25,011 --> 00:29:27,936
the terms of which are still
sort of coming out. But, you know,

391
00:29:28,020 --> 00:29:29,990
this is generally good.
But it's, you know,

392
00:29:30,073 --> 00:29:33,619
we need an actual backstop to it as well.
So we actually have to see sort of

393
00:29:33,703 --> 00:29:36,389
that self-policing, but also
the verification piece of it.

394
00:29:36,473 --> 00:29:39,687
And that's the part that I'm really
kind of looking to see is like –

395
00:29:39,770 --> 00:29:42,761
Are there going to be new authorities
coming forward that help sort of with

396
00:29:42,844 --> 00:29:45,916
that policing of it? Yeah, I think
there's a lot of really interesting issues

397
00:29:46,000 --> 00:29:48,990
that were talked about. I mean, as we
think about these models and we think

398
00:29:49,074 --> 00:29:51,695
about what the frontier is, I mean,
how are we going to do safety?

399
00:29:51,779 --> 00:29:54,564
How are we going to make sure
the government is involved? And I think,

400
00:29:54,648 --> 00:29:57,638
you know, all those topics are sort
of front of mind for people. And again,

401
00:29:57,722 --> 00:29:59,917
just going back to this,
like, very simple point.

402
00:30:00,000 --> 00:30:02,707
Industry and the government
talking is always a good thing.

403
00:30:02,790 --> 00:30:03,623
So,

404
00:30:04,200 --> 00:30:07,263
To push back on Josh a little bit, we
had voluntary commitments in the previous

405
00:30:07,346 --> 00:30:09,246
administration. How is this different?

406
00:30:09,480 --> 00:30:12,025
We're actually going to have
sort of the policing side of it.

407
00:30:12,108 --> 00:30:14,826
So as Trump said yesterday,
we're going to have the DOJ, the FTC.

408
00:30:14,909 --> 00:30:18,187
I think this morning FTC came out with
they're actually doing an investigation

409
00:30:18,270 --> 00:30:21,720
and a review of all of the frontier labs.
So I think we're already starting to see

410
00:30:21,804 --> 00:30:24,607
some of this come to fruition.
And I do think this is going to lead

411
00:30:24,691 --> 00:30:27,451
to congressional action as well.
And to you, Mike, well, you know,

412
00:30:27,535 --> 00:30:30,535
the one last question. Yeah,
sure, sure, sure. To you, Mike.

413
00:30:30,960 --> 00:30:33,989
Last administration did invite all
these CEOs when they signed the voluntary

414
00:30:34,072 --> 00:30:36,572
commitment. They did or didn't?
They didn't. Okay.

415
00:30:37,050 --> 00:30:39,210
So how do you think
this time is different?

416
00:30:40,470 --> 00:30:42,037
I mean, in both...

417
00:30:42,120 --> 00:30:44,583
in both situations, whether it
was the Biden administration or the Trump

418
00:30:44,666 --> 00:30:47,407
administration, they're building this
plane while they're flying it. Right. Uh,

419
00:30:47,490 --> 00:30:50,648
That's hard and it's complicated.
And so you're doing the best you can

420
00:30:50,732 --> 00:30:54,029
in government. And again, I wish it
was bipartisan. But at the same time,

421
00:30:54,112 --> 00:30:57,641
this conversation is important to have.
And the the voluntary commitments will

422
00:30:57,724 --> 00:31:00,929
probably change over time. Maybe
there'll be some congressional action.

423
00:31:01,012 --> 00:31:03,522
I obviously hope there will be,
because all of you know,

424
00:31:03,605 --> 00:31:07,088
in an industry like this, you want
to make sure that you build something that

425
00:31:07,171 --> 00:31:10,607
is enduring. And the only way to have
something enduring is actually through

426
00:31:10,691 --> 00:31:14,358
Congress acting. And I also think in an
industry that could be or that will be so

427
00:31:14,442 --> 00:31:15,275
consequential,

428
00:31:17,490 --> 00:31:19,987
there be a big bipartisan
vote to show that

429
00:31:20,070 --> 00:31:24,380
there is an enduring commitment from
the government to figure out how to be

430
00:31:24,463 --> 00:31:29,125
thoughtful in this ecosystem. What would
your advice be to members of Congress as

431
00:31:29,208 --> 00:31:33,811
a response to this? Do you see them as
being boxed in by the signed declaration?

432
00:31:33,894 --> 00:31:37,350
No, not at all. I think it's
actually the other way around.

433
00:31:37,770 --> 00:31:41,011
I think right now, so the speaker said
sort of they're not planning on bringing

434
00:31:41,094 --> 00:31:44,334
folks back to pass some of what I think
some folks will return to is like these

435
00:31:44,418 --> 00:31:46,606
snap laws where you're seeing
sort of an overreaction.

436
00:31:46,690 --> 00:31:49,507
But what they're trying to do
is take a measured approach.

437
00:31:49,590 --> 00:31:53,462
Yes, the voluntary policing or
self-evaluation combined with government

438
00:31:53,546 --> 00:31:55,858
action. And I think
as the FTC, as the DOJ,

439
00:31:55,941 --> 00:32:00,427
the FBI and others are looking at this,
then you're going to see Congress actually

440
00:32:00,510 --> 00:32:04,438
identifying ways that they can take
rifle shot approaches where it's not

441
00:32:04,521 --> 00:32:08,840
over-regulation and it's not shutting
down the industry and stunting the growth

442
00:32:08,923 --> 00:32:13,352
that we are currently on the cusp of, but
actually enabling that. And by the way,

443
00:32:13,436 --> 00:32:16,890
we do this across all other sectors,
right? Aviation, banking.

444
00:32:19,590 --> 00:32:23,196
We can do it here. Yes,
the technology is moving very, very fast.

445
00:32:23,279 --> 00:32:25,693
We can accomplish this.
Sure. But, you know,

446
00:32:25,776 --> 00:32:29,326
Democrats on the Hill have been
pushing for much more. So, Mike,

447
00:32:29,409 --> 00:32:33,185
what would your recommendation be
to Democratic members of the House

448
00:32:33,268 --> 00:32:37,272
in particular on, you know, how do
you start working with Republicans on

449
00:32:37,355 --> 00:32:38,490
this issue, knowing?

450
00:32:39,420 --> 00:32:43,306
fairly well that, you know, Trump is not
going to sign certain types of bills.

451
00:32:43,389 --> 00:32:47,071
So how should they navigate this?
I mean, the simple first answer, Martin,

452
00:32:47,155 --> 00:32:51,142
is let's get past the election. I think
you're not likely to have a conversation

453
00:32:51,226 --> 00:32:54,806
that doesn't, you know, on purpose
or accidentally result in a political

454
00:32:54,890 --> 00:32:57,808
outcome. But in 30, I don't know,
I can't do math. 34 days.

455
00:32:57,892 --> 00:33:01,167
34 days before the election.
So the first thing I would do is say,

456
00:33:01,250 --> 00:33:05,238
sit on your hands. And then let's maybe
see if we can have a conversation during

457
00:33:05,321 --> 00:33:08,120
the lame duck. I mean,
I would run for 20 lame ducks or

458
00:33:09,420 --> 00:33:12,547
There's not usually a lot of get a lot
that gets done in terms of like long

459
00:33:12,630 --> 00:33:14,987
things that are really truly
enduring. It's usually like,

460
00:33:15,070 --> 00:33:17,298
how can we kick the can into
January? And I, you know,

461
00:33:17,381 --> 00:33:20,894
I don't want to say that Congress won't be
able to get anything done during the lame

462
00:33:20,977 --> 00:33:24,104
duck. But reality is doing something
that is big and consequential is hard.

463
00:33:24,187 --> 00:33:27,057
So that's the first thing I would say.
Sorry, go ahead. No, go ahead.

464
00:33:27,141 --> 00:33:29,857
I was just I didn't expect to agree
with Mike on anything today.

465
00:33:29,940 --> 00:33:33,383
But this scenario, I'm pleased to hear
because I do think getting through

466
00:33:33,466 --> 00:33:35,116
the election is really important.

467
00:33:35,670 --> 00:33:38,842
In addition to that, like immediately
we already have things in the works

468
00:33:38,925 --> 00:33:42,453
that Congress is working on, and I think
there's opportunity to find a bipartisan

469
00:33:42,536 --> 00:33:45,663
approach on the NDA. Again, the whole
purpose of this is we want to win.

470
00:33:45,747 --> 00:33:48,963
We want to compete against every nation
that's trying to develop AI and be

471
00:33:49,046 --> 00:33:52,307
the lead to own that marketplace.
Right now there are provisions in the NDA

472
00:33:52,390 --> 00:33:55,696
that would help us maintain that lead,
and I think that is something that on

473
00:33:55,779 --> 00:33:57,787
a bipartisan basis we
can rally around and do.

474
00:33:57,870 --> 00:33:59,901
Yeah, I mean, the things
that Josh mentioned in the NDA,

475
00:33:59,984 --> 00:34:01,901
I think there's a lot
of really good things in there.

476
00:34:01,985 --> 00:34:03,713
And interestingly,
they're different capacities,

477
00:34:03,797 --> 00:34:06,318
a lot of things that Josh and I
have been working on for a long time.

478
00:34:06,401 --> 00:34:09,527
So those are good. I think in terms of how
do you have this conversation either after

479
00:34:09,610 --> 00:34:12,132
the election or in the new year,
when there's a new Congress, I mean,

480
00:34:12,215 --> 00:34:14,321
the first thing you have to do
is sort of figure out like,

481
00:34:14,404 --> 00:34:15,997
where are we actually aligned? And I think

482
00:34:16,080 --> 00:34:19,315
that one of the conversations you just
can't have 34 days or 35 days before

483
00:34:19,398 --> 00:34:21,704
an election is like, where
is their alignment? I mean,

484
00:34:21,787 --> 00:34:24,624
every political actor, regardless
of their Democrat or Republican,

485
00:34:24,707 --> 00:34:27,942
is going to be like, where can I find
the opportunity to drive a wedge into

486
00:34:28,026 --> 00:34:31,216
this issue? And how can I how can I
get political gain from it? So, again,

487
00:34:31,300 --> 00:34:34,579
just this this is not the time to do it,
I think that but that doesn't mean,

488
00:34:34,662 --> 00:34:38,207
you know, 39 days from now or in January
that we shouldn't have that conversation.

489
00:34:38,290 --> 00:34:40,286
But I think there's a lot
of alignment. I mean,

490
00:34:40,370 --> 00:34:43,320
some of the things that Josh
talked about with DOJ and FTC.

491
00:34:46,080 --> 00:34:49,036
We'll have to have transparency into
that process. What does it look like?

492
00:34:49,120 --> 00:34:52,117
What are the outcomes? What are the risks?
And how do we think about those?

493
00:34:52,200 --> 00:34:55,537
The one thing I'm still puzzled is what
role would DOJ would play in this?

494
00:34:55,620 --> 00:34:58,980
There's so many agencies
that can put on top of this.

495
00:34:59,160 --> 00:35:01,988
to do some kind of a vetting,
control testing and stuff like that.

496
00:35:02,071 --> 00:35:04,458
What is DOJ going to do?
I think you're partially right.

497
00:35:04,541 --> 00:35:05,857
I think DOJ is sort of the...

498
00:35:05,940 --> 00:35:08,886
the muscle, so to speak, to make sure
that the companies are compliant. No,

499
00:35:08,970 --> 00:35:11,027
I know how the sound of that sounds.
It's the muscle.

500
00:35:11,110 --> 00:35:13,047
You're going to have
a major IC component on this.

501
00:35:13,130 --> 00:35:16,417
You're going to have the FBI involved.
They're going to have to be involved.

502
00:35:16,500 --> 00:35:18,944
specifically with some
of the WMD stuff, the bioweapons.

503
00:35:19,027 --> 00:35:22,329
And so they're going to have to be
involved across the interagency on this.

504
00:35:22,412 --> 00:35:24,811
It's not just going to be
those two. And that's, again,

505
00:35:24,894 --> 00:35:26,857
why I think there's
a lot of tools right now

506
00:35:26,940 --> 00:35:30,184
I think the FDC is relying on sort of some
statutes that go a little bit beyond

507
00:35:30,267 --> 00:35:33,457
their remit. And so they're going
to have to work with Congress on that.

508
00:35:33,540 --> 00:35:36,066
I mean, we have the departments
and agencies that we have.

509
00:35:36,149 --> 00:35:39,530
So somebody someone who's going to be
picked as the as the I don't know if we

510
00:35:39,613 --> 00:35:42,948
say lucky person or unlucky person
in this case. And so the agent, you know,

511
00:35:43,032 --> 00:35:45,153
how it works readily.
Someone's got it. You know,

512
00:35:45,236 --> 00:35:48,436
the president has to assign somebody
in the piece of paper that he forms.

513
00:35:48,520 --> 00:35:51,720
I mean, hopefully it is an interagency
effort. You know, the FTC. I mean,

514
00:35:51,804 --> 00:35:55,319
it's great. Consumers and the American
people are going to have have have should

515
00:35:55,402 --> 00:35:58,738
have an incredible role in this. And
understanding the impact on on citizens

516
00:35:58,821 --> 00:36:02,021
is important. I think the point
Josh raised is really important.

517
00:36:03,540 --> 00:36:06,104
international intelligence play,
the cyber community play,

518
00:36:06,187 --> 00:36:08,477
and what's sort of the red
teaming capacity of this?

519
00:36:08,560 --> 00:36:11,900
How are we thinking about auditing?
There's a lot of questions to be asked.

520
00:36:11,983 --> 00:36:12,067
And

521
00:36:12,150 --> 00:36:17,312
Someone's got to lead it, so here we are.
So one other thing I wanted to ask you

522
00:36:17,395 --> 00:36:21,639
guys was – so yesterday we released
America.gov, we being the U.S.

523
00:36:21,722 --> 00:36:25,537
government. So it's a chatbot
that should allow people to –

524
00:36:25,620 --> 00:36:28,121
engage with government
services more easily.

525
00:36:28,204 --> 00:36:31,703
So love to get your guys take.
Well, what do you think of it?

526
00:36:31,787 --> 00:36:36,343
Is this the right way forward? Are there
any risks involved with something like

527
00:36:36,426 --> 00:36:40,627
this? I'll let my esteemed
colleague go first on the attack.

528
00:36:40,710 --> 00:36:45,975
No, listen, I think... When I
think about the American government,

529
00:36:46,058 --> 00:36:49,171
one thing I never think is AI ready.
And the other thing I never think about

530
00:36:49,254 --> 00:36:52,409
is data ready. So I think in both places,
I think the administration, I mean,

531
00:36:52,492 --> 00:36:55,605
is trying to do some work here. But
there's so much work to be done and none

532
00:36:55,688 --> 00:36:58,212
of it is being done urgently.
And so, like, there's a website,

533
00:36:58,295 --> 00:37:01,408
did you know this part, called data.gov,
where there's all kinds of the U.S.

534
00:37:01,491 --> 00:37:04,391
government and state and local
data that's made available.

535
00:37:05,040 --> 00:37:08,295
Where are we in terms of making
all this other data available,

536
00:37:08,379 --> 00:37:12,173
and how can we think about really
mobilizing the entire executive branch

537
00:37:12,256 --> 00:37:16,373
to think about data in a real way where
we can make sure that these models are

538
00:37:16,456 --> 00:37:19,442
actually training on America.gov,
on other American data?

539
00:37:19,526 --> 00:37:23,643
Because at least before the frontier
models sort of got out and started really

540
00:37:23,726 --> 00:37:27,980
improving, you know who the biggest holder
of data was in the world? Uncle Sam.

541
00:37:35,040 --> 00:37:37,240
some really interesting
work with that data.

542
00:37:37,980 --> 00:37:38,377
Thank you.

543
00:37:38,460 --> 00:37:42,656
So, yes, this is one of those areas where
I think it probably sounds very boring

544
00:37:42,739 --> 00:37:46,240
to the average person, but it's
absolutely essential and necessary.

545
00:37:46,323 --> 00:37:49,771
The United States has been
drastically behind on this for decades.

546
00:37:49,854 --> 00:37:53,194
I mean, literally, like the fact
that we have a separate system,

547
00:37:53,277 --> 00:37:57,420
a separate website for every little thing
you have to do across the government.

548
00:37:57,503 --> 00:38:00,630
A country like Estonia has done
this decades ago. I was in –

549
00:38:00,713 --> 00:38:03,120
we were speaking to a
foreign minister from –

550
00:38:03,210 --> 00:38:06,133
Kazakhstan, and we were talking
about AI and tech and everything,

551
00:38:06,216 --> 00:38:08,815
and he pulled out his phone
and showed me that on one app,

552
00:38:08,898 --> 00:38:12,145
he could go from working on a visa
application to paying his water bill,

553
00:38:12,228 --> 00:38:15,197
to paying his electrical utilities.
It was all there in one place.

554
00:38:15,281 --> 00:38:17,556
He didn't have to go
to several different websites.

555
00:38:17,639 --> 00:38:19,545
It's all sort of there
together. Obviously,

556
00:38:19,628 --> 00:38:22,782
we had a more complicated system
here in the United States, but again,

557
00:38:22,865 --> 00:38:25,742
I think it's a great initiative.
It probably sounds very boring,

558
00:38:25,825 --> 00:38:29,442
probably sounds like it's nothing, but I
think it's going to make people's lives

559
00:38:29,525 --> 00:38:30,358
much

560
00:38:33,210 --> 00:38:37,321
they're working with performs yeah I mean
Martin one of the things I recently did

561
00:38:37,405 --> 00:38:41,620
was I was driving so in a city that's not
Washington DC and the traffic lights were

562
00:38:41,703 --> 00:38:45,866
a disaster can you imagine if you could
synchronize all the data in just a city oh

563
00:38:45,949 --> 00:38:50,164
yeah and I can take two and a half minutes
off your commute every morning and every

564
00:38:50,248 --> 00:38:53,510
afternoon morning I was sitting
at an intersection for a minute

565
00:38:54,000 --> 00:38:57,280
red light, no cars, no pedestrians,
like there's a better way. I mean,

566
00:38:57,364 --> 00:39:01,028
this is a solvable problem, right? And so
as you start thinking about the data

567
00:39:01,112 --> 00:39:03,142
that government has
available to it, I mean,

568
00:39:03,226 --> 00:39:05,426
there's all kinds
of really incredible ways.

569
00:39:06,000 --> 00:39:09,319
kind of small ways that have a real
impact on citizens, right? I mean,

570
00:39:09,402 --> 00:39:13,012
in Washington, D.C., I'm told, and I
learned this a long time ago when I was

571
00:39:13,096 --> 00:39:16,852
in the Senate, that they actually have
the ability to control all of the lights

572
00:39:16,935 --> 00:39:20,011
in the city, which is astounding
given how dysfunctional the D.C.

573
00:39:20,094 --> 00:39:23,753
government can sometimes be. But can
you imagine taking my traffic light idea

574
00:39:23,836 --> 00:39:26,377
and applying that to the
District of Columbia? I mean,

575
00:39:26,460 --> 00:39:29,876
it would be bananas, Martin. And if
you could then say, like, oh, honey,

576
00:39:29,960 --> 00:39:32,112
I'm home five minutes
early because, you know,

577
00:39:32,195 --> 00:39:35,595
Claude or Chachy PT somehow got
the lights in the city to get going.

578
00:39:36,000 --> 00:39:38,504
Hopefully you can scale
that idea with Uncle Sam's data,

579
00:39:38,587 --> 00:39:41,414
and so hopefully you can figure
out how to get services faster.

580
00:39:41,498 --> 00:39:44,972
It would be pretty astounding. I think
it's in general a positive initiative,

581
00:39:45,055 --> 00:39:47,355
but like everything's
in the execution, right?

582
00:39:47,580 --> 00:39:49,761
Yeah, and it actually has to work,
right? Because ultimately,

583
00:39:49,845 --> 00:39:52,843
if it gives people time back in their day
to spend with their families or do things

584
00:39:52,926 --> 00:39:55,293
that they need to be doing to create
a business or something else,

585
00:39:55,376 --> 00:39:58,115
that's amazing. At the same time,
there is an inherent sort of security risk

586
00:39:58,198 --> 00:39:59,927
and vulnerability to it.
Obviously, we all know.

587
00:40:00,010 --> 00:40:02,747
America's like the largest
cyber target in the world.

588
00:40:02,830 --> 00:40:06,769
And so really important that that piece of
it goes along and moves forward with it.

589
00:40:06,852 --> 00:40:08,998
But again, like an amazing
initiative. I mean,

590
00:40:09,081 --> 00:40:12,244
the cyber piece is totally right.
I'm actually glad Josh raised it.

591
00:40:12,328 --> 00:40:14,667
Something we worked
on in another capacity, again,

592
00:40:14,751 --> 00:40:18,108
is just like how do you deploy
cybersecurity in a really meaningful way

593
00:40:18,191 --> 00:40:20,628
at scale and quickly.
So it's a really hard problem.

594
00:40:20,711 --> 00:40:23,535
But that doesn't mean we shouldn't do it.
That's right, yes.

595
00:40:23,618 --> 00:40:27,121
You can't let some of these things
stop you from implementing these things

596
00:40:27,204 --> 00:40:30,755
that people have had in their mind
with respect to government services for,

597
00:40:30,839 --> 00:40:32,089
you know, decades. Right.

598
00:40:32,380 --> 00:40:35,997
Thank you, gentlemen. That's all the time
we have for this week's segment across

599
00:40:36,080 --> 00:40:39,142
the aisle. Thank you for being here
in the studio. Getting the hook.

600
00:40:39,225 --> 00:40:41,454
Thanks for having us.
Great to be here personally.

601
00:40:41,538 --> 00:40:43,638
You should come see Josh and I more often.

602
00:40:45,100 --> 00:40:46,200
It's great to see you.

603
00:40:59,890 --> 00:41:04,783
Martin, this concludes our Thursday show.
We'll be back on Tuesday at noon across

604
00:41:04,867 --> 00:41:08,592
all our platforms. Make sure
you check out our next big event,

605
00:41:08,676 --> 00:41:13,385
which is the AI Plus Robotics. This is
our second big AI Plus Robotics Summit.

606
00:41:13,468 --> 00:41:17,378
It's the time when we have to really
pay attention on this issue,

607
00:41:17,462 --> 00:41:20,327
October 14th. Please go
and register at scsp.ai,

608
00:41:20,411 --> 00:41:24,647
and Martin and I will be back here
on Tuesday to cover the next week.

609
00:41:24,730 --> 00:41:27,868
Everything that the president needs
to know about tech and national security.

610
00:41:27,951 --> 00:41:29,001
Have a great weekend.

611
00:41:29,410 --> 00:41:36,410
*music*