1
00:00:00,000 --> 00:00:04,317
One of the most important [music] things
Palantir's done on the battlefield is be

2
00:00:04,400 --> 00:00:07,917
able to make up for the fact that [music]
half your enterprise doesn't even work.

3
00:00:08,000 --> 00:00:11,677
On the battlefield. It does work
in a lab, on a PowerPoint. [music]

4
00:00:11,760 --> 00:00:13,437
For example, if you just take Ukraine as

5
00:00:13,520 --> 00:00:19,560
an example, how hard could it be
to move a [music] drone from A to B?

6
00:00:22,040 --> 00:00:24,677
>> It's my pleasure
to introduce Alex Karp.

7
00:00:24,760 --> 00:00:27,877
Um I want to just start off on something

8
00:00:27,960 --> 00:00:32,077
that's more intimate between he and I.
I mean, I'm pretty proud of what I created

9
00:00:32,160 --> 00:00:33,677
at BlackRock, but

10
00:00:33,760 --> 00:00:35,997
my total return since I've been a CEO

11
00:00:36,080 --> 00:00:39,157
has only compounded at 21%.

12
00:00:39,240 --> 00:00:43,317
Uh since Alex and Palantir went public,

13
00:00:43,400 --> 00:00:47,117
uh his compounded return is 73%. So,

14
00:00:47,200 --> 00:00:48,877
um

15
00:00:48,960 --> 00:00:51,837
congratulations, Alex.

16
00:00:51,920 --> 00:00:54,757
As co-founder and CEO of Palantir, uh

17
00:00:54,840 --> 00:00:57,357
Alex has closely worked with defense and

18
00:00:57,440 --> 00:00:59,037
government

19
00:00:59,120 --> 00:01:01,997
private private organizations to apply

20
00:01:02,080 --> 00:01:05,117
how AI can be used, and and it's really

21
00:01:05,200 --> 00:01:10,077
important. And I must say, my
conversations that I've had in with Alex

22
00:01:10,160 --> 00:01:13,877
over over the last year
has really enlightened me.

23
00:01:13,960 --> 00:01:16,437
So, I'm looking forward to this, Alex.

24
00:01:16,520 --> 00:01:17,997
So, let me just start off from your

25
00:01:18,080 --> 00:01:20,597
perspective, how is AI supporting

26
00:01:20,680 --> 00:01:24,197
decision-making in defense and security?

27
00:01:24,280 --> 00:01:28,517
>> One of the most important things
Palantir's done on the battlefield is be

28
00:01:28,600 --> 00:01:32,157
able to make up for the fact that half
your enterprise doesn't even work.

29
00:01:32,240 --> 00:01:36,677
On the battlefield. It does work in a lab,
on a PowerPoint. For example, if

30
00:01:36,760 --> 00:01:41,197
you just take Ukraine as an example, how
hard could it be to move a drone from A

31
00:01:41,280 --> 00:01:45,597
to B? Actually, first of all, you're going
to need to know where you want to put the

32
00:01:45,680 --> 00:01:49,477
drone, and that's going to require
synchronizing all your data without

33
00:01:49,560 --> 00:01:51,797
transferring that data to your adversary,
which means you're going to

34
00:01:51,880 --> 00:01:54,357
have to know every single person who
touched it. You're going to have to

35
00:01:54,440 --> 00:01:58,437
obfuscate it to the final thing. Then
you're going to want to do it presumably

36
00:01:58,520 --> 00:02:02,477
either within the accordance of your
strategy or with your ethics. So, where

37
00:02:02,560 --> 00:02:07,277
does it where does it not go? Do you
want to put the drone on your asset? No

38
00:02:07,360 --> 00:02:11,477
one Only two people in Ukraine may know
that one of the generals is your asset.

39
00:02:11,560 --> 00:02:13,117
You can't tell people that your asset.

40
00:02:13,200 --> 00:02:16,957
How do you How do you make it look like
to your soldiers that you actually were

41
00:02:17,040 --> 00:02:20,637
taking people out and you just missed
your asset? And then the war advances.

42
00:02:20,720 --> 00:02:24,317
So, the Russians they began jamming
electronics. So, now you have a

43
00:02:24,400 --> 00:02:27,557
completely different problem. Now it's
not a question of going from A to B.

44
00:02:27,640 --> 00:02:31,637
It's a question of going through
a completely jammed environment where you

45
00:02:31,720 --> 00:02:35,437
have no connectivity while actually
collecting data while you go through

46
00:02:35,520 --> 00:02:37,957
that environment. And every one of those

47
00:02:38,040 --> 00:02:43,317
things is like a dynamic challenge. None
of which were foreseen even before the

48
00:02:43,400 --> 00:02:47,277
Ukraine. And by the way, the other thing
is then people fight differently. The

49
00:02:47,360 --> 00:02:51,757
Ukrainians, it's a small team of people
and very courageous soldiers that are

50
00:02:51,840 --> 00:02:57,037
very technical. They have highly
technical people that have built on top

51
00:02:57,120 --> 00:03:00,877
of our product things that we don't
understand and proprietary ways of using

52
00:03:00,960 --> 00:03:02,797
the product. And then in America you

53
00:03:02,880 --> 00:03:08,037
have just have massive force
but they had to be integrated. So,

54
00:03:08,120 --> 00:03:13,437
every one of those things is different
and so the the twofold role of of

55
00:03:13,520 --> 00:03:16,757
enterprise software on a battlefield
is one to make sure all the underlying

56
00:03:16,840 --> 00:03:20,837
things actually works and then two
to raise the level to a level no one else

57
00:03:20,920 --> 00:03:23,437
has in the world. >> There's
so much technology was created

58
00:03:23,520 --> 00:03:26,877
from defense whether it was the internet

59
00:03:26,960 --> 00:03:29,157
GPS.

60
00:03:29,240 --> 00:03:34,037
How do you see envision this
translating from defense and military

61
00:03:34,120 --> 00:03:38,877
um to corporations
to businesses to society?

62
00:03:38,960 --> 00:03:42,317
>> On the battlefield,
one of the most important issues

63
00:03:42,400 --> 00:03:47,037
is how do you acquire data, process it,
and then put it into a framework. In the

64
00:03:47,120 --> 00:03:48,557
end of the day, what is a business

65
00:03:48,640 --> 00:03:50,677
doing? It's information. It's sorting

66
00:03:50,760 --> 00:03:54,277
the information in a way that you have

67
00:03:54,360 --> 00:03:59,197
a distinct advantage over other
people who are similarly situated

68
00:03:59,280 --> 00:04:03,197
and that that advantage can't be
eviscerated easily. When you approach

69
00:04:03,280 --> 00:04:06,637
the underwriter or the hospital,
we power tons and tons of hospitals.

70
00:04:06,720 --> 00:04:08,077
>> Right. >> They have
an intake problem. They all

71
00:04:08,160 --> 00:04:11,477
have an intake problem. They all have
a shortage of doctors and nurses. They are

72
00:04:11,560 --> 00:04:15,037
working in a low margin environment. But
every single one has a different way of

73
00:04:15,120 --> 00:04:19,237
processing their patients according
to what their specialty is and the kind of

74
00:04:19,320 --> 00:04:23,197
patients they don't do well with and how
do you manage that? And so the intake

75
00:04:23,280 --> 00:04:26,117
flow and into your
enterprise in a way that

76
00:04:26,200 --> 00:04:30,237
you can actually process these things 10,
15 times faster than you could.

77
00:04:30,320 --> 00:04:32,797
>> So what is the basic hindrance in the

78
00:04:32,880 --> 00:04:35,997
adaptation of of the of AI?

79
00:04:36,080 --> 00:04:39,997
>> Our adoption accelerates beyond our
capacity. The The problem with adoption

80
00:04:40,080 --> 00:04:42,637
at this point is people have tried

81
00:04:42,720 --> 00:04:44,957
things that just can never work.

82
00:04:45,040 --> 00:04:47,077
Like by a large language model, you put

83
00:04:47,160 --> 00:04:49,917
it on your stack and you wonder why it's

84
00:04:50,000 --> 00:04:52,277
not working. And so what you're going to

85
00:04:52,360 --> 00:04:55,477
see, especially in America, is people

86
00:04:55,560 --> 00:05:00,197
trying to do something like what we
do with Anthology maybe by hand because

87
00:05:00,280 --> 00:05:03,597
once you build a software layer
to orchestrate and manage the large

88
00:05:03,680 --> 00:05:08,037
language models in in a language
that your enterprise understands, you

89
00:05:08,120 --> 00:05:12,917
actually can create value. And again,
if you go back to the battlefield context,

90
00:05:13,000 --> 00:05:16,277
like everybody in the world
assumed this would not work.

91
00:05:16,360 --> 00:05:20,357
But now it does work.
And so now the question isn't does it

92
00:05:20,440 --> 00:05:24,157
work, but it how can we get it to work
for my country? And this is exactly

93
00:05:24,240 --> 00:05:25,957
what's happening in >> In companies.

94
00:05:26,040 --> 00:05:29,317
>> In companies. It's like, oh, this
company it worked, mine didn't. What are

95
00:05:29,400 --> 00:05:33,317
you doing? >> How rapidly can AI

96
00:05:33,400 --> 00:05:36,557
change the growth trajectory? >>
In a lot of these things there's a speed

97
00:05:36,640 --> 00:05:38,757
question. I think like with a lot of our

98
00:05:38,840 --> 00:05:41,677
companies we can take out

99
00:05:41,760 --> 00:05:44,557
in the area we go in up to

100
00:05:44,640 --> 00:05:49,197
80% of your cost and improve your top
line dramatically. But it really depends

101
00:05:49,280 --> 00:05:51,117
on the use case and what we're doing.

102
00:05:51,200 --> 00:05:54,757
In the past, 5 years ago, that would
take us a year. Now, it could take a week.

103
00:05:54,840 --> 00:05:58,957
>> Week, yeah. Is AI going to create
jobs or destroy jobs overall?

104
00:05:59,040 --> 00:06:02,877
>> There are will be more than enough
jobs for the citizens of your nation,

105
00:06:02,960 --> 00:06:07,037
especially those with vocational training.
I do think these these trends

106
00:06:07,120 --> 00:06:10,917
really do make it hard to imagine why
we should have large-scale immigration

107
00:06:11,000 --> 00:06:13,197
unless you have a very specialized skill.

108
00:06:13,280 --> 00:06:16,637
>> Are you also insinuating we're
probably going to need less white-collar?

109
00:06:16,720 --> 00:06:18,157
>> You know, there are a lot of people

110
00:06:18,240 --> 00:06:21,117
doing X that should be doing Y. It's

111
00:06:21,200 --> 00:06:24,237
like just the person managing our Maven

112
00:06:24,320 --> 00:06:26,717
system in the US Army is a former police

113
00:06:26,800 --> 00:06:29,277
officer who I think went to a junior

114
00:06:29,360 --> 00:06:34,037
college. And they're doing very
high-end, very complicated targeting

115
00:06:34,120 --> 00:06:38,277
globally. And that person actually is

116
00:06:38,360 --> 00:06:42,277
irreplaceable.
And I think in the past, the way we

117
00:06:42,360 --> 00:06:44,157
tested for aptitude

118
00:06:44,240 --> 00:06:47,317
uh would not have fully exposed how

119
00:06:47,400 --> 00:06:49,437
irreplaceable that person's talents are.

120
00:06:49,520 --> 00:06:52,197
And would they been as talented if
[clears throat] they had not gone to

121
00:06:52,280 --> 00:06:55,797
their college? Yes.
If you look at inside Palantir, what am

122
00:06:55,880 --> 00:07:00,837
I really doing all day? I'm one walking
around figuring out what is someone's

123
00:07:00,920 --> 00:07:04,757
outlier aptitude. And then
I'm putting them on that thing

124
00:07:04,840 --> 00:07:08,317
and trying to get them to stay on that
thing and not on the five other things

125
00:07:08,400 --> 00:07:10,637
they think they're great at. >> So,

126
00:07:10,720 --> 00:07:12,677
is AI going to create greater imbalance

127
00:07:12,760 --> 00:07:15,597
in our in our world in terms of growth?

128
00:07:15,680 --> 00:07:18,277
>> One way to look at the unfairness of AI

129
00:07:18,360 --> 00:07:21,517
is it pentests, meaning it load bears on

130
00:07:21,600 --> 00:07:25,797
things. So, societies that can
and organizations and companies that can

131
00:07:25,880 --> 00:07:30,117
bear that load have a huge
advantage. The problem is if

132
00:07:30,200 --> 00:07:34,357
you've been pretending you're bearing
a load you're not, it collapses and that's

133
00:07:34,440 --> 00:07:36,557
where you have to start. And there's a

134
00:07:36,640 --> 00:07:39,557
certain honesty that is painful for all

135
00:07:39,640 --> 00:07:42,917
of us in in this technology. It you just

136
00:07:43,000 --> 00:07:45,317
cannot obfuscate what can bear the load

137
00:07:45,400 --> 00:07:50,037
and what can't. And then political
structures are built to do just that.

138
00:07:50,120 --> 00:07:54,117
Like I can't fix anything,
but I can give you some line that you

139
00:07:54,200 --> 00:07:57,797
want to hear that's going to make you
not care about how bad your life is and

140
00:07:57,880 --> 00:07:59,757
how much worse it's going to be tomorrow.

141
00:07:59,840 --> 00:08:02,117
And the best thing you can do and you

142
00:08:02,200 --> 00:08:03,477
really care for the people you're

143
00:08:03,560 --> 00:08:07,197
representing is to say let's

144
00:08:07,280 --> 00:08:12,237
You have to kind of look closely
at what what load we can bear.

145
00:08:12,320 --> 00:08:13,757
>> Thank you Alex. >> Thank you.

146
00:08:13,840 --> 00:08:17,200
>> everyone. >> [applause]