1
00:00:00,000 --> 00:00:03,477
And it has to be culturally specific.
If you just take war, the German way of

2
00:00:03,560 --> 00:00:06,837
fighting was different than the French
way, which is different than British way.

3
00:00:06,920 --> 00:00:07,957
Which is different. And you see that

4
00:00:08,040 --> 00:00:11,197
nowadays. But

5
00:00:11,280 --> 00:00:15,237
the the the There there are technologies

6
00:00:15,320 --> 00:00:18,037
that are linear. This is not a linear

7
00:00:18,120 --> 00:00:21,717
technology situation. And if you take a,

8
00:00:21,800 --> 00:00:26,037
you know, and and so it's
just crucial that everybody

9
00:00:26,120 --> 00:00:29,957
has a private space where you're like,
is this working? Why is it working?

10
00:00:30,040 --> 00:00:34,677
What's not working? And then in every
country you have part of the stack you

11
00:00:34,760 --> 00:00:38,437
can realistically power. I mean, China's
a little different. I mean, I think that

12
00:00:38,520 --> 00:00:40,557
part will be mostly powered by Chinese

13
00:00:40,640 --> 00:00:46,277
companies. But every other country
will have to make a decision

14
00:00:46,360 --> 00:00:50,717
of what part of the stack will be powered
by American companies, what part of the

15
00:00:50,800 --> 00:00:54,837
stack will be powered by indigenous
companies, local companies. Who are

16
00:00:54,920 --> 00:00:57,437
those local companies?
Who are the experts?

17
00:00:57,520 --> 00:01:00,117
Um And that expertise

18
00:01:00,200 --> 00:01:03,317
is very very rare.

19
00:01:03,400 --> 00:01:06,317
Um and it it this is also one of the

20
00:01:06,400 --> 00:01:11,037
weird things about this industry is um
there's so many things you It's like

21
00:01:11,120 --> 00:01:14,957
your industry. Like you're a master
in your finance industry. There's so many

22
00:01:15,040 --> 00:01:17,677
things you know about finance

23
00:01:17,760 --> 00:01:19,877
that it would be hard to replicate. Like

24
00:01:19,960 --> 00:01:22,677
say we switch positions. I I am certain

25
00:01:22,760 --> 00:01:25,157
I would It's like And that kind of

26
00:01:25,240 --> 00:01:28,357
knowledge is um just really important to

27
00:01:28,440 --> 00:01:32,477
make sure you have a sovereign ownership
up in your country. And if we can help,

28
00:01:32,560 --> 00:01:35,677
we will. But ultimately, you'll
probably have to do it yourself.

29
00:01:35,760 --> 00:01:38,117
>> So Can you help us understand the

30
00:01:38,200 --> 00:01:41,277
intersection between Palantir

31
00:01:41,360 --> 00:01:43,797
and the AI frontier models? And how how

32
00:01:43,880 --> 00:01:46,077
does that work? Like how do you bring it

33
00:01:46,160 --> 00:01:48,757
together? How should they understand

34
00:01:48,840 --> 00:01:53,837
sort of cuz they're trying to understand
the ecosystems of the AI world. And

35
00:01:53,920 --> 00:01:55,597
that's one of the discussions today.

36
00:01:55,680 --> 00:01:59,717
Like what's that ecosystem?
So maybe you could sort of lay that out.

37
00:01:59,800 --> 00:02:03,317
>> terms, there you have compute,
so I think Jensen was here.

38
00:02:03,400 --> 00:02:07,517
>> Yeah. >> World's expert,
talk to Jensen. Um

39
00:02:07,600 --> 00:02:11,437
and then you have basically then
the caricature of it is you would have

40
00:02:11,520 --> 00:02:14,317
frontier models and it real in real in

41
00:02:14,400 --> 00:02:19,197
reality you're going to have frontier
and non-frontier models and non-frontier

42
00:02:19,280 --> 00:02:24,277
open models that you can make
frontier-like in specific use cases. Um

43
00:02:24,360 --> 00:02:27,757
and and so, what you're
trying to do, let's

44
00:02:27,840 --> 00:02:31,717
say you're trying to solve a problem,
like how do I optimize the supply chain

45
00:02:31,800 --> 00:02:36,157
of my country in a way where I know
every single dependency on a country I

46
00:02:36,240 --> 00:02:40,637
may not trust. Okay, that's a that's
literally an issue every single every

47
00:02:40,720 --> 00:02:42,757
single person in this room has. Um and

48
00:02:42,840 --> 00:02:46,037
and or may not trust tomorrow or if if

49
00:02:46,120 --> 00:02:50,157
this part left, what would be
the replacement? Replacing the parts are

50
00:02:50,240 --> 00:02:53,677
very difficult because a part isn't
a part. Part made by this company is

51
00:02:53,760 --> 00:02:55,837
different than even if the same spec.

52
00:02:55,920 --> 00:02:59,357
Every single every single manufacturing
company in the world has this. Um and

53
00:02:59,440 --> 00:03:02,557
how do I do it while while not exporting

54
00:03:02,640 --> 00:03:05,397
my way of manufacturing um

55
00:03:05,480 --> 00:03:09,757
to another country. Let's just say I
am a French company, I have a way of

56
00:03:09,840 --> 00:03:13,637
building a nuclear reactor. I'm pretty
interested in nuclear reactors are very

57
00:03:13,720 --> 00:03:17,637
complicated. I need to know every single
part of my supply chain that is not

58
00:03:17,720 --> 00:03:21,877
French or not European and I also need
to know if I had how would I have to

59
00:03:21,960 --> 00:03:24,917
rebuild the supply chain if I had to buy
a part from someone else and what would

60
00:03:25,000 --> 00:03:27,997
be the cost. These are So, that's a no
business every business in the world has

61
00:03:28,080 --> 00:03:31,437
this problem. Okay, so um you

62
00:03:31,520 --> 00:03:35,197
if you want to make a frontier model
be able to work on that problem both

63
00:03:35,280 --> 00:03:39,077
passively, meaning you know the answer
when you ask the question, actively,

64
00:03:39,160 --> 00:03:42,397
meaning you know the answer when you
ask question, but also passively. You get

65
00:03:42,480 --> 00:03:46,957
You guys get in a fight and you're
like, I mean, you're tough guy.

66
00:03:47,040 --> 00:03:50,397
So, that's just and you're like, okay,
we got to we got to rebuild the supply

67
00:03:50,480 --> 00:03:52,957
chain. Okay, um

68
00:03:53,040 --> 00:03:55,837
how do you get pre-

69
00:03:55,920 --> 00:03:58,077
these models without an application

70
00:03:58,160 --> 00:04:00,757
layer? Well, A give you an answer, but

71
00:04:00,840 --> 00:04:03,997
that answer even in the the active state

72
00:04:04,080 --> 00:04:07,677
is not 100% accurate. It is a estimation

73
00:04:07,760 --> 00:04:12,037
of the answer and it also degrades over
time. So, first problem is how do you

74
00:04:12,120 --> 00:04:16,597
make sure that answer is correct and you
can decomp it. So, how do you take apart

75
00:04:16,680 --> 00:04:21,037
the question so you can see every single
assumption in the question in the answer

76
00:04:21,120 --> 00:04:24,997
so that if you're a country leader
and you have to decide that that that what

77
00:04:25,080 --> 00:04:27,917
you did in order to do that you will
need an application layer. You will also

78
00:04:28,000 --> 00:04:31,477
need an application layer, which is what
we provide. I'm not just but you'll need

79
00:04:31,560 --> 00:04:33,117
to you you someone will have to provide.

80
00:04:33,200 --> 00:04:35,957
You also need to make sure that your way

81
00:04:36,040 --> 00:04:37,557
of building the supply chain is not

82
00:04:37,640 --> 00:04:41,317
being migrated to a large lab to a lab

83
00:04:41,400 --> 00:04:44,317
model company because that's why you

84
00:04:44,400 --> 00:04:47,397
have a company. So, okay, in reality

85
00:04:47,480 --> 00:04:52,277
what you're going to do is you're going
to bring someone in who understands what

86
00:04:52,360 --> 00:04:56,197
part of the data has to be captured
in this model. And to do that you're going

87
00:04:56,280 --> 00:05:00,037
to have to train either
post train a closed model

88
00:05:00,120 --> 00:05:03,117
or as likely currently you're going to

89
00:05:03,200 --> 00:05:07,797
train a an open model on the precise

90
00:05:07,880 --> 00:05:11,117
questions you need to ask and answer
to and you're going to do it in a closed

91
00:05:11,200 --> 00:05:15,117
environment. And that's how you're going
to solve this problem because otherwise

92
00:05:15,200 --> 00:05:18,877
you are you have to worry
about let's just say

93
00:05:18,960 --> 00:05:21,837
you tell me cross country skiing is evil
and I pull your problem. None of these

94
00:05:21,920 --> 00:05:26,237
things actually happen in practice.
And so, where what does that mean for you?

95
00:05:26,320 --> 00:05:29,037
It means you need someone who
understands the strengths and weaknesses

96
00:05:29,120 --> 00:05:32,557
of every model.
Uh also the legal vulnerabilities that

97
00:05:32,640 --> 00:05:36,077
you're putting yourself into by using
certain models. You need someone who

98
00:05:36,160 --> 00:05:40,117
technically understands the interplay
between your proprietary data and a

99
00:05:40,200 --> 00:05:44,117
model as it exists and a model as you
can train it and can also set up an

100
00:05:44,200 --> 00:05:47,837
architecture that will allow you
to bring in more expensive and cheaper

101
00:05:47,920 --> 00:05:50,837
models, but also more secure
and less secure models because

102
00:05:50,920 --> 00:05:53,437
there are also then you may not care if

103
00:05:53,520 --> 00:05:55,597
a model from America that you don't

104
00:05:55,680 --> 00:05:58,557
trust is running on your ad campaign in

105
00:05:58,640 --> 00:06:00,757
Argentina, which is your biggest market.

106
00:06:00,840 --> 00:06:02,917
You will care if it's running on how you

107
00:06:03,000 --> 00:06:05,637
actually make your nuclear reactor work.

108
00:06:05,720 --> 00:06:10,637
And so that part of the stack, we are
the application layer north, we will put

109
00:06:10,720 --> 00:06:14,117
our an application and south to some
extent the integration layer, we will

110
00:06:14,200 --> 00:06:18,517
put our our ontology in the middle and
allow you to train these models so that

111
00:06:18,600 --> 00:06:22,037
you can control them
and there's no exfiltration.

112
00:06:22,120 --> 00:06:26,037
But whether you used our product or not,
you will need this to make these things

113
00:06:26,120 --> 00:06:28,757
A work and make them secure. And and if

114
00:06:28,840 --> 00:06:32,157
you and one of like just to my biggest

115
00:06:32,240 --> 00:06:35,917
criticism of regulation
outside of America is

116
00:06:36,000 --> 00:06:38,677
please get technical experts to do it.

117
00:06:38,760 --> 00:06:43,237
Like currently the regulation is meaning
it's a meaning the application of it is

118
00:06:43,320 --> 00:06:46,917
you have companies that are essentially
there to deploy large language models.

119
00:06:47,000 --> 00:06:49,357
Those companies are exfiltrating, not

120
00:06:49,440 --> 00:06:53,997
intentionally, your alpha to the
company that provides the model.

121
00:06:54,080 --> 00:06:58,917
It is like the opposite of what you're
trying to It is literally the opposite

122
00:06:59,000 --> 00:07:03,597
of what you're trying to do.
And if this is not sovereignty, this is

123
00:07:03,680 --> 00:07:08,077
anti-sovereignty. And whoever's telling
you this is works needs to be fired

124
00:07:08,160 --> 00:07:09,837
because they are hurting your country.

125
00:07:09,920 --> 00:07:13,717
They're not hurting me, but this is
literally the way to export every one of

126
00:07:13,800 --> 00:07:17,117
your secrets to some other company.
And it's only because the people doing the

127
00:07:17,200 --> 00:07:21,397
regulation don't actually understand
the technical material. And this is very

128
00:07:21,480 --> 00:07:23,157
hard. Every country has this problem.

129
00:07:23,240 --> 00:07:25,197
Like how you know, I think you've been

130
00:07:25,280 --> 00:07:28,037
very effective in kind of tweaking at

131
00:07:28,120 --> 00:07:30,317
the edges when people have gotten out of

132
00:07:30,400 --> 00:07:35,317
control and and in a way
that you know was very effective.

133
00:07:35,400 --> 00:07:39,237
Uh and um but we have
the problem like who's going

134
00:07:39,320 --> 00:07:43,037
to monitor the development of AI? It's
a very hard problem because you need to be

135
00:07:43,120 --> 00:07:46,517
an expert. And how would you why would
you trust the expert? That's a really

136
00:07:46,600 --> 00:07:50,397
hard question. One simple way
of asking and answering the question is

137
00:07:50,480 --> 00:07:52,997
does the expert is the
expert on the payroll?

138
00:07:53,080 --> 00:07:55,797
Cuz a lot of these experts are actually

139
00:07:55,880 --> 00:08:00,797
wrapping a company. I am, too. But I'm
not wrapping buying this product. And

140
00:08:00,880 --> 00:08:04,357
part of the problem is, you know,
we have a very very you know, one of the

141
00:08:04,440 --> 00:08:07,917
things about the development
of nuclear energy and maybe a fortunate

142
00:08:08,000 --> 00:08:11,957
unfortunately war was the people who
did it actually weren't doing it for the

143
00:08:12,040 --> 00:08:16,917
money. And we have a problem now and I
I mean I I I'm kind of more left. I think

144
00:08:17,000 --> 00:08:20,917
it would be fair to say then, you know,
I believe in capitalism as the best

145
00:08:21,000 --> 00:08:22,877
system because everything else is worse.

146
00:08:22,960 --> 00:08:25,957
I think you believe in it as just
the best system. There's a slight

147
00:08:26,040 --> 00:08:27,757
distinction there. Uh

148
00:08:27,840 --> 00:08:30,357
I'm still somewhat of a German social

149
00:08:30,440 --> 00:08:33,037
democrat or a which means things have to

150
00:08:33,120 --> 00:08:35,677
work. I am skeptical of migration.

151
00:08:35,760 --> 00:08:40,317
I think it was a huge mistake. Um I mean
not controlled migration, but what we

152
00:08:40,400 --> 00:08:44,317
got. Um uh but uh which was the opinion of

153
00:08:44,400 --> 00:08:48,277
Helmut Schmidt.
That was Helmut Schmidt's opinion.

154
00:08:48,360 --> 00:08:51,637
Maybe we could listen to him.
Uh but any case uh

155
00:08:51,720 --> 00:08:55,557
he was a titan. Um for those
of you who never heard of him. Uh

156
00:08:55,640 --> 00:08:57,677
but any case uh

157
00:08:57,760 --> 00:09:02,157
Um it it's uh it it it's very very hard

158
00:09:02,240 --> 00:09:05,077
to do these things correctly.
And And I'll tell you one of the most

159
00:09:05,160 --> 00:09:08,397
interesting things about regulation
which is on you know, someone who says

160
00:09:08,480 --> 00:09:12,517
one of the things I think is happening
is when we had CDMA and Europe had GSM,

161
00:09:12,600 --> 00:09:14,437
these are the wireless technologies. I

162
00:09:14,520 --> 00:09:17,477
think most most people thought CDMA was

163
00:09:17,560 --> 00:09:21,477
better, but most Europeans thought the
regulation in Europe was better. And so

164
00:09:21,560 --> 00:09:24,957
you ended up with a I think Europeans
would universally say they have a better

165
00:09:25,040 --> 00:09:29,157
product. You can debate it. But I think
people are pretty And this is just a

166
00:09:29,240 --> 00:09:31,157
very different revolution where the

167
00:09:31,240 --> 00:09:33,037
technology is like a breakout

168
00:09:33,120 --> 00:09:36,637
technology. And um and and and and you

169
00:09:36,720 --> 00:09:40,477
just can't intuit what would be good for

170
00:09:40,560 --> 00:09:42,557
your country. You have to really have a

171
00:09:42,640 --> 00:09:46,077
deep understanding of like where where

172
00:09:46,160 --> 00:09:49,397
is the where where is the IP and alpha

173
00:09:49,480 --> 00:09:54,437
actually being monetized? It doesn't
inherently mean because a local person

174
00:09:54,520 --> 00:09:59,157
is doing it, you're safe.
It's actually not true at all.

175
00:09:59,240 --> 00:10:00,197
So,

176
00:10:00,280 --> 00:10:03,397
is sort of a core

177
00:10:03,480 --> 00:10:06,477
teaching moment for the group uh that

178
00:10:06,560 --> 00:10:09,437
they should deeply think about making

179
00:10:09,520 --> 00:10:12,717
sure that their data remains there. You

180
00:10:12,800 --> 00:10:14,597
know, that that what they're training

181
00:10:14,680 --> 00:10:17,997
the model on remains in their control.

182
00:10:18,080 --> 00:10:21,717
That you need an application There's
two basic things. One, I mean, there's a

183
00:10:21,800 --> 00:10:25,397
technical version. You will need
an application layer which prevents which

184
00:10:25,480 --> 00:10:28,197
makes the models valuable, which
helps train them, and which makes them

185
00:10:28,280 --> 00:10:32,597
specific. That will also allow you
to select weaker models that might be

186
00:10:32,680 --> 00:10:36,597
indigenous and control them. So,
like let's say as a matter of theory, you

187
00:10:36,680 --> 00:10:40,157
have I don't know, a pair You have
a model from Chile. It's weaker than the

188
00:10:40,240 --> 00:10:42,517
model in America. You still might need

189
00:10:42,600 --> 00:10:44,197
one to use it. And and and and open

190
00:10:44,280 --> 00:10:47,237
models are we're like we we just we have

191
00:10:47,320 --> 00:10:51,597
a very important client
in America for lots of reasons uh

192
00:10:51,680 --> 00:10:55,837
they wanted to use open models.
The open model is being done by

193
00:10:55,920 --> 00:10:58,917
is actually on certain vectors. Now,
again, that you can't generalize. On

194
00:10:59,000 --> 00:11:01,957
certain vectors, that model is better
than a frontier model. Now, it's only

195
00:11:02,040 --> 00:11:05,197
certain vectors, but it's the vector
they care about. Then, everybody has the

196
00:11:05,280 --> 00:11:10,517
same issue as the US government. Every
single one of you has If you're like, I

197
00:11:10,600 --> 00:11:14,437
don't know, we have a big presence
in there. Do you really want to call large

198
00:11:14,520 --> 00:11:16,637
language model and provider and

199
00:11:16,720 --> 00:11:19,077
negotiate the parameters of what you are

200
00:11:19,160 --> 00:11:21,837
doing on, I don't know, your security

201
00:11:21,920 --> 00:11:26,437
apparatus which in every country
is different. Um, so like, you know, the

202
00:11:26,520 --> 00:11:29,997
security the bounds of what you would do,
for example, just a neutral example.

203
00:11:30,080 --> 00:11:34,757
In Europe, in America, civil
libertarians believe that um

204
00:11:34,840 --> 00:11:39,237
you should not purge data because
by purging the data, it's hard to see what

205
00:11:39,320 --> 00:11:41,357
the government has done with the data.

206
00:11:41,440 --> 00:11:45,557
Core to European concept of your of of
civil liberties is that you do purge the

207
00:11:45,640 --> 00:11:49,157
data. I'm kind of a little bit
on the European side because

208
00:11:49,240 --> 00:11:52,957
maybe you have an affair. Maybe
something like that happens. Do you need

209
00:11:53,040 --> 00:11:56,757
everyone to know that? So, I kind
of like the purge the data thing, but I

210
00:11:56,840 --> 00:12:00,397
think it's hard to say one or the other
is actually better. If you If you purge

211
00:12:00,480 --> 00:12:04,877
the data, you can't really see what the
police did. If you don't purge uh if you

212
00:12:04,960 --> 00:12:07,157
don't purge the data, you can, but it's

213
00:12:07,240 --> 00:12:12,597
harder to have what might be viewed as a
normal personal life. Uh and most people

214
00:12:12,680 --> 00:12:15,957
do have a normal personal life
at some time in their life. Um

215
00:12:16,040 --> 00:12:20,757
of course no one in this room,
but um I may have had one. But uh um

216
00:12:20,840 --> 00:12:22,357
and uh uh

217
00:12:22,440 --> 00:12:25,117
so, those are but if you have to

218
00:12:25,200 --> 00:12:29,637
negotiate the parameters
of the model or take

219
00:12:29,720 --> 00:12:32,837
I mean, I I like I you know, we have
it's commonly known we have a lot of big

220
00:12:32,920 --> 00:12:38,037
presence in French internal security.
Um and um but we don't have one in other

221
00:12:38,120 --> 00:12:42,837
places. Um Do French
the the French internal

222
00:12:42,920 --> 00:12:45,997
security are very
effective at what they do.

223
00:12:46,080 --> 00:12:49,637
Do they really need to negotiate
when they use a model with an American

224
00:12:49,720 --> 00:12:53,317
frontier lab of the parameters that you
they're using the model? By the way,

225
00:12:53,400 --> 00:12:57,117
they might have a way of doing something
that allows them to catch people no

226
00:12:57,200 --> 00:12:59,197
one's ever thought of. If you exportate

227
00:12:59,280 --> 00:13:01,277
that, you no longer have a way of doing

228
00:13:01,360 --> 00:13:04,677
it. And so, the the the those things are

229
00:13:04,760 --> 00:13:06,437
I would also say then the second point

230
00:13:06,520 --> 00:13:09,117
is not that related. Every single one of

231
00:13:09,200 --> 00:13:13,477
you has a a national in this country who

232
00:13:13,560 --> 00:13:18,357
is a titan in industry.
And I I told Merkel this. I've got I've

233
00:13:18,440 --> 00:13:21,917
got someone I I I talked
I told a lot of you have to

234
00:13:22,000 --> 00:13:26,317
get them to move home.
Whatever it I don't know how you do it,

235
00:13:26,400 --> 00:13:30,357
but you got to get them to come and say
you are now going to do this for my

236
00:13:30,440 --> 00:13:34,477
country and you're going to set policy
in my country. And those technical

237
00:13:34,560 --> 00:13:39,837
experts are with the exception really
of China, they are sitting here.

238
00:13:39,920 --> 00:13:43,237
And they're sitting here for lots
of reasons. We, you know, like again, the

239
00:13:43,320 --> 00:13:46,677
reason why I think it's important
to remember how crucial the time is, we

240
00:13:46,760 --> 00:13:50,117
don't have time to debate why people
are coming here. You can debate it. I can

241
00:13:50,200 --> 00:13:53,917
tell you why they're coming here,
but it doesn't matter. They are here.

242
00:13:54,000 --> 00:13:57,437
Almost everybody who
knows this is here and

243
00:13:57,520 --> 00:14:02,437
they you need to get them to come home
to your country and sit next to you and

244
00:14:02,520 --> 00:14:06,117
explain what regulation should look
in your terms. What should regulation look

245
00:14:06,200 --> 00:14:08,157
like? What should the stack look like?

246
00:14:08,240 --> 00:14:12,997
What model should we use? What software
we should use? What chips we should buy?

247
00:14:13,080 --> 00:14:16,917
And what are and what is the cost of not,
you know, it may not be possible to

248
00:14:17,000 --> 00:14:21,037
buy the best stack. Like I'm I'm
negotiating this with a lot of times of

249
00:14:21,120 --> 00:14:26,077
with of countries abroad that quite
frankly are you know, like they're like

250
00:14:26,160 --> 00:14:29,717
we don't want to be dependent
on the Palantir stack completely.

251
00:14:29,800 --> 00:14:33,157
Which is, you know,
it's a heavy lift because

252
00:14:33,240 --> 00:14:36,877
there are certain things that just we
have especially on the battlefield. But

253
00:14:36,960 --> 00:14:39,197
I'm helping them to build a plan B.

254
00:14:39,280 --> 00:14:42,877
For because of course they don't want to
be dependent on me. I mean, in this case

255
00:14:42,960 --> 00:14:45,957
I'm wildly pro-European despite

256
00:14:46,040 --> 00:14:50,437
I feel like being slightly maligned. But
I'm I'm supplying things to Europe that

257
00:14:50,520 --> 00:14:52,357
are not even in my economic interest.

258
00:14:52,440 --> 00:14:55,277
What if I get hit by a car?
How are you going to like dealing with

259
00:14:55,360 --> 00:14:57,957
the next person running the company
who's like, "Are you not paying me as

260
00:14:58,040 --> 00:15:01,717
much as the gas station?"
That's what's going to happen. So, like

261
00:15:01,800 --> 00:15:05,077
developing these things
and a technical expert

262
00:15:05,160 --> 00:15:08,357
which you have, can really really drill

263
00:15:08,440 --> 00:15:10,637
into these issues in hours. A

264
00:15:10,720 --> 00:15:13,397
non-technical expert is very likely

265
00:15:13,480 --> 00:15:17,677
working for a company
that wants you to spend money.

266
00:15:17,760 --> 00:15:22,877
And then last lesson, you don't
need money. You need knowledge. I

267
00:15:22,960 --> 00:15:26,237
One of the things I won't pick
on countries here, but a country I care a

268
00:15:26,320 --> 00:15:30,397
lot about is always telling me how
much money they're going to spend.

269
00:15:30,480 --> 00:15:32,237
I don't want to hear how much money
you're going to spend. Who's going to

270
00:15:32,320 --> 00:15:35,357
get the money? Like the money
doesn't matter as who as

271
00:15:35,440 --> 00:15:37,117
much as who gets it.

272
00:15:37,200 --> 00:15:39,477
And that is crucial. The person who is

273
00:15:39,560 --> 00:15:44,877
in charge of spending, investing
in tech needs to be someone who really is

274
00:15:44,960 --> 00:15:48,077
thinking, "Why am I doing this? I could
be making even more money in America."

275
00:15:48,160 --> 00:15:52,037
And you remind them,
"Because your country is great."

276
00:15:52,120 --> 00:15:54,837
And that's how you do it.

277
00:15:54,920 --> 00:15:59,197
>> One of the earlier conversations today

278
00:15:59,280 --> 00:16:03,237
said that having access to data centers

279
00:16:03,320 --> 00:16:06,277
>> Mhm. >> is non-negotiable.
Whether you built it

280
00:16:06,360 --> 00:16:11,077
or you rented it from someone else,
that the ability to have compute for your

281
00:16:11,160 --> 00:16:16,037
country is non-negotiable.
Do you agree? Do you

282
00:16:16,120 --> 00:16:17,637
buy that? What What's your take?

283
00:16:17,720 --> 00:16:19,957
>> Um I I I just want
to say for the record

284
00:16:20,040 --> 00:16:24,117
your palantir I get on my so crazy I
get on my line. There's this insane guy

285
00:16:24,200 --> 00:16:27,637
saying we're building data centers like
in Florida. We don't even build data

286
00:16:27,720 --> 00:16:32,517
centers. So just in case anyone's
watching this, um I It's like, you know,

287
00:16:32,600 --> 00:16:36,717
yeah, look, you you are going to have
to have You need access to compute, you

288
00:16:36,800 --> 00:16:40,437
need access to your own data.
So the stack is compute,

289
00:16:40,520 --> 00:16:44,757
data, ability to manage data, ability to
control the models, ability to train the

290
00:16:44,840 --> 00:16:47,077
models, and ability to compete against

291
00:16:47,160 --> 00:16:52,197
your competitors and adversaries. And one
crucial part of that is compute. And

292
00:16:52,280 --> 00:16:55,437
countries are going to figure
out different ways to do it. I

293
00:16:55,520 --> 00:16:59,917
I I would My go to We're
partnered with Jensen and and

294
00:17:00,000 --> 00:17:04,397
Nvidia. I would I think he's the world's
best at this. I would take I I also So

295
00:17:04,480 --> 00:17:07,157
you know, like there's different ways
we're doing some pretty interesting

296
00:17:07,240 --> 00:17:10,517
things that are are not yet public um

297
00:17:10,600 --> 00:17:14,877
around how you can build compute in in
kind of very sovereign architectures,

298
00:17:14,960 --> 00:17:19,397
both commercially and in government.
Um so I that I think that is a starting

299
00:17:19,480 --> 00:17:21,877
point. You also need electricity.

300
00:17:21,960 --> 00:17:25,077
Um, and uh, if you look
at, you know, just like

301
00:17:25,160 --> 00:17:27,957
again, I I would say, um, you know, the

302
00:17:28,040 --> 00:17:32,597
move in France to build nuclear
and stick with nuclear is probably what's

303
00:17:32,680 --> 00:17:34,877
powering like it's just that like we

304
00:17:34,960 --> 00:17:37,197
have to have a non-politically

305
00:17:37,280 --> 00:17:41,997
uh, it like Luddite conversation about
these things. Like you are going to need

306
00:17:42,080 --> 00:17:47,197
a lot of electricity. How are you going
to get it? And again, I where like where

307
00:17:47,280 --> 00:17:50,597
I I tend to think all these things come
with negative externalities. So I don't

308
00:17:50,680 --> 00:17:52,357
think there's like I don't think the AI

309
00:17:52,440 --> 00:17:56,157
revolution is 100% like good. But but

310
00:17:56,240 --> 00:17:58,757
there's a huge dangers. But and with all

311
00:17:58,840 --> 00:18:02,117
these technologies, but you but but then

312
00:18:02,200 --> 00:18:06,597
being left behind is probably the biggest
danger now. The and and then we

313
00:18:06,680 --> 00:18:08,717
all have and you know, some countries

314
00:18:08,800 --> 00:18:10,717
here I know pretty well. We all have

315
00:18:10,800 --> 00:18:13,717
political insanity in our countries on

316
00:18:13,800 --> 00:18:16,997
both sides. By which I do not mean you.

317
00:18:17,080 --> 00:18:18,077
>> Thank you.

318
00:18:18,160 --> 00:18:21,317
>> I I I I but like how do you negotiate

319
00:18:21,400 --> 00:18:24,597
with wide swaths of your populace that

320
00:18:24,680 --> 00:18:28,677
you know, don't Again, I think they're
legitimate. I mean, I personally and I'm

321
00:18:28,760 --> 00:18:33,477
I'm speaking for on my own behalf and not
in I think a big problem we have in

322
00:18:33,560 --> 00:18:36,717
this country is I don't
think people are down for me

323
00:18:36,800 --> 00:18:39,237
being 20 times wealthier.

324
00:18:39,320 --> 00:18:44,237
It it it's just it it leaves a bad taste
in people's mouth. And I come from an

325
00:18:44,320 --> 00:18:47,837
academic family. I'm not even sure how
I ended up with all this money. And so

326
00:18:47,920 --> 00:18:51,437
like I do think that's a that's
like that's like a real issue for our

327
00:18:51,520 --> 00:18:54,757
societies. And I do think like some

328
00:18:54,840 --> 00:18:59,357
societies in this room have less of that
issue because it's like very much like

329
00:18:59,440 --> 00:19:02,437
we're looking out for you. I think
the Arab Middle East here has a huge

330
00:19:02,520 --> 00:19:07,157
advantage. Um, because
it's like uh, um, we are

331
00:19:07,240 --> 00:19:09,077
like on the political front, we are

332
00:19:09,160 --> 00:19:11,397
under enormous pressure because the big

333
00:19:11,480 --> 00:19:14,277
technology is not egalitarian. And then

334
00:19:14,360 --> 00:19:16,357
it's I think easier to organize if you

335
00:19:16,440 --> 00:19:21,957
are a a country with a one with
a very understanding of a order

336
00:19:22,040 --> 00:19:24,797
that everyone in the country that belongs
is being looked after. You just

337
00:19:24,880 --> 00:19:28,237
have a lot less of these issues. And
there's no getting around these issues.

338
00:19:28,320 --> 00:19:32,357
And so we have to figure out a ways
to bring people along and explain they'll

339
00:19:32,440 --> 00:19:34,477
participate, but

340
00:19:34,560 --> 00:19:36,877
but but I actually believe this, but I'm

341
00:19:36,960 --> 00:19:40,077
not the right person to explain it.

342
00:19:40,160 --> 00:19:43,477
And so it's like, "Guys, it's going
to be great for everybody." And I'm not

343
00:19:43,560 --> 00:19:46,997
That's actually not my message. I think
it There are going to be some disruptive

344
00:19:47,080 --> 00:19:50,157
things. I don't believe in the doom
message. Now, part of the problem we

345
00:19:50,240 --> 00:19:52,517
have in this country is a certain

346
00:19:52,600 --> 00:19:57,197
company was doing a doom message every
day. And that doom message is really

347
00:19:57,280 --> 00:20:01,197
really scared people and and then
they're and it scares people because

348
00:20:01,280 --> 00:20:03,437
there are elements of truth in it.

349
00:20:03,520 --> 00:20:07,477
But the elements of truth have been
exaggerated to the point where it causes

350
00:20:07,560 --> 00:20:11,197
huge political problems
and I I I think you will

351
00:20:11,280 --> 00:20:13,037
you you see these across the Western

352
00:20:13,120 --> 00:20:16,237
world. Uh maybe less actually uh

353
00:20:16,320 --> 00:20:19,557
definitely less in Japan, Korea, China

354
00:20:19,640 --> 00:20:22,037
than than in other places, but there I

355
00:20:22,120 --> 00:20:27,117
think also um you just have you have
a very different setup. I mean, all those

356
00:20:27,200 --> 00:20:31,837
countries are very different from one
another, but in the end I think um we'll

357
00:20:31,920 --> 00:20:33,957
have political advantages on the

358
00:20:34,040 --> 00:20:37,037
adoption of AI. Uh whereas the American

359
00:20:37,120 --> 00:20:41,517
advantage is the production of AI. So,
that's our We have an enormous advantage

360
00:20:41,600 --> 00:20:43,397
in the production. The dissemination of

361
00:20:43,480 --> 00:20:45,877
AI and getting people to adopt without

362
00:20:45,960 --> 00:20:49,117
political insanity, that's where we and

363
00:20:49,200 --> 00:20:52,037
I would I would say without overreach

364
00:20:52,120 --> 00:20:54,437
European countries have a pretty extreme

365
00:20:54,520 --> 00:20:59,717
disadvantage. And so, how do
you how do you get that to work?

366
00:20:59,800 --> 00:21:02,997
Is a real Maybe we should see if they
have any questionings. I don't know, but

367
00:21:03,080 --> 00:21:06,037
I love your questions.
We can keep going with yours.

368
00:21:06,120 --> 00:21:10,517
Well, I I think uh I'd like
to sort of end it with are

369
00:21:10,600 --> 00:21:14,677
you tell me about optimism. Is this uh an

370
00:21:14,760 --> 00:21:16,397
optimistic time? You know, there are

371
00:21:16,480 --> 00:21:19,677
always risks with change, but isn't this

372
00:21:19,760 --> 00:21:23,997
an incredibly optimistic time? The
opportunity that they all have for their

373
00:21:24,080 --> 00:21:26,317
citizens to be able to

374
00:21:26,400 --> 00:21:29,997
uh extend education, knowledge,

375
00:21:30,080 --> 00:21:32,997
uh democratize knowledge, the capacity

376
00:21:33,080 --> 00:21:37,437
to change things. I'm interested in your
take. Is it Is it worrisome to you or is

377
00:21:37,520 --> 00:21:41,157
it optimistic? >> I ain't ready to put it.

378
00:21:41,240 --> 00:21:44,317
I mean, like in our like my I I am

379
00:21:44,400 --> 00:21:48,477
optimistic under the condition that
everyone goes home and says, "Okay,

380
00:21:48,560 --> 00:21:53,237
guys, let's get real. It we're not
really taking this seriously yet."

381
00:21:53,320 --> 00:21:57,317
Uh I think it I would
say it's very similar to

382
00:21:57,400 --> 00:22:00,197
battlefield context. You Under the

383
00:22:00,280 --> 00:22:04,877
condition that everyone goes home and
says, "Okay, we didn't really take this

384
00:22:04,960 --> 00:22:07,997
seriously, and there were lots
of reasons not to take it seriously, and

385
00:22:08,080 --> 00:22:12,077
some of it is a joke." And I mean,
as I'm speaking as someone who's lived

386
00:22:12,160 --> 00:22:16,757
abroad a lot of my life, many times people
think something is American marketing.

387
00:22:16,840 --> 00:22:19,837
Like, you know, it's a thing. People
like think we're the best at this, and

388
00:22:19,920 --> 00:22:23,557
maybe it's, you know,
that. This isn't that.

389
00:22:23,640 --> 00:22:26,197
And uh and so that's where it under the

390
00:22:26,280 --> 00:22:29,037
condition that people like you're very

391
00:22:29,120 --> 00:22:32,077
important people in the world, and it

392
00:22:32,160 --> 00:22:33,597
going to your

393
00:22:33,680 --> 00:22:37,317
very justifiably important, historically

394
00:22:37,400 --> 00:22:40,157
important, in the present important uh

395
00:22:40,240 --> 00:22:44,837
culture and history-shaping countries,
of which you all represent in one form

396
00:22:44,920 --> 00:22:49,277
or another here, uh and saying,
"This is game time.

397
00:22:49,360 --> 00:22:52,437
How serious are we?" And and I do
think you should get access to all the

398
00:22:52,520 --> 00:22:56,437
classified data. What has worked on
the battlefield? Is a very important You

399
00:22:56,520 --> 00:22:59,877
know, historically, the test
of technology has been what has worked on

400
00:22:59,960 --> 00:23:02,597
the battlefield. Get access to the class

401
00:23:02,680 --> 00:23:05,317
your most your actually classified data.

402
00:23:05,400 --> 00:23:07,237
What has worked, what hasn't worked. And

403
00:23:07,320 --> 00:23:10,357
then build something around we are going

404
00:23:10,440 --> 00:23:13,837
to make this work and find actual

405
00:23:13,920 --> 00:23:18,157
patriots in your country.
Then I'm very optimistic.

406
00:23:18,240 --> 00:23:22,397
Um and then I'm also in then
in the American context,

407
00:23:22,480 --> 00:23:26,477
you know, finding ways to educate people
about where things will work and where

408
00:23:26,560 --> 00:23:29,397
they won't have a more
honest conversation.

409
00:23:29,480 --> 00:23:32,637
I would be even more optimistic. But the

410
00:23:32,720 --> 00:23:35,037
net net, I'm of course optimistic.

411
00:23:35,120 --> 00:23:37,117
Um and you know, I guess I'm still an

412
00:23:37,200 --> 00:23:40,037
academic, so I can't be 100% academic as

413
00:23:40,120 --> 00:23:44,357
as cheery and lively as you are, but I I

414
00:23:44,440 --> 00:23:47,877
I'm aspiring to it. I'm aspiring to it.

415
00:23:47,960 --> 00:23:52,120
>> Thank you. Okay.
>> Thank you very much.

416
00:24:27,440 --> 00:24:31,157
>> Good afternoon, everyone. Um

417
00:24:31,240 --> 00:24:34,397
part of our objective uh

418
00:24:34,480 --> 00:24:38,117
was to bring about a variety of views

419
00:24:38,200 --> 00:24:39,677
and ideas

420
00:24:39,760 --> 00:24:42,637
um without filter.

421
00:24:42,720 --> 00:24:46,157
Right? AI is without filter. And so I

422
00:24:46,240 --> 00:24:51,960
didn't sit aside and plan what my